{ "storage": "unilm_joint.espresso.weights", "analyses": { "N8Espresso23analysis_debug_metadataE": { "path": "", "bundle": "", "product": "", "use_case": "", "name": "LOCALHOST-2023-05-01-ANE-quant", "version": "" } }, "properties": { "mldb_token" : "mldb-5qbg63zgxe" }, "format_version": 200, "metadata_in_weights": [], "layers": [ { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": -1, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 3, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "input", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "16", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "16", "begin_48": 0, "top": "16_0,16_1,16_2", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "nB": 15000, "top": "52", "has_biases": 0, "weights": { "Q": 1, "Qscale_t": 3, "W_t_int8": 5 }, "nC": 512, "is_lookup": 1, "quantization_mode": 2, "type": "inner_product", "has_relu": 0, "bottom": "16_0", "debug_info": "", "has_tanh": 0, "nd_mode": true, "name": "52", "has_prelu": 0 }, { "nB": 128, "top": "55", "has_biases": 0, "weights": { "Q": 1, "Qscale_t": 327, "W_t_int8": 329 }, "nC": 512, "is_lookup": 1, "quantization_mode": 2, "type": "inner_product", "has_relu": 0, "bottom": "16_1", "debug_info": "", "has_tanh": 0, "nd_mode": true, "name": "55", "has_prelu": 0 }, { "nB": 2, "top": "58", "has_biases": 0, "weights": { "Q": 1, "Qscale_t": 11, "W_t_int8": 13 }, "nC": 512, "is_lookup": 1, "quantization_mode": 2, "type": "inner_product", "has_relu": 0, "bottom": "16_2", "debug_info": "", "has_tanh": 0, "nd_mode": true, "name": "58", "has_prelu": 0 }, { "bottom": "52,55", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "60", "top": "60", "type": "elementwise", "name": "60", "beta": 0 }, { "bottom": "60,58", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "61", "top": "61", "type": "elementwise", "name": "61", "beta": 0 }, { "axis_seq": 4, "top": "transpose_7", "name": "transpose_7", "axis_n": 3, "axis_h": 1, "type": "transpose", "axis_w": 2, "bottom": "61", "hint_fallback_from_cpu": 1, "axis_k": 0, "debug_info": "transpose_7", "weights": {} }, { "axis_mode": 4, "bottom": "transpose_7", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.3", "use_version": 1, "top": "channels_mean.3", "type": "reduce", "name": "channels_mean.3" }, { "alpha": -1, "bottom": "channels_mean.3", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.3", "top": "_neg_y_zero_mean.3", "type": "activation", "name": "_neg_y_zero_mean.3", "beta": 0 }, { "bottom": "transpose_7,_neg_y_zero_mean.3", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.3", "top": "zero_mean.3", "type": "elementwise", "name": "zero_mean.3", "beta": 0 }, { "bottom": "zero_mean.3,zero_mean.3", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.3", "top": "zero_mean_sq.3", "type": "elementwise", "name": "zero_mean_sq.3", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.3", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "89", "use_version": 1, "top": "89", "type": "reduce", "name": "89" }, { "bottom": "89", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "91", "top": "91", "type": "elementwise", "name": "91", "beta": 0 }, { "bottom": "91", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.3", "top": "denom.3", "type": "elementwise", "name": "denom.3", "beta": 0 }, { "bottom": "zero_mean.3,denom.3", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.3", "top": "out.3", "type": "elementwise", "name": "out.3", "beta": 0 }, { "top": "95", "w": 1, "h": 1, "name": "95", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "95", "n": 1, "weights": {}, "constant_blob": 575 }, { "bottom": "out.3,95", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "96", "top": "96", "type": "elementwise", "name": "96", "beta": 0 }, { "top": "98", "w": 1, "h": 1, "name": "98", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "98", "n": 1, "weights": {}, "constant_blob": 333 }, { "bottom": "96,98", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "k.3", "top": "k.3", "type": "elementwise", "name": "k.3", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "113", "K": 512, "blob_biases": 21, "name": "113", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.3", "weights": { "per_ch_qscale": 23, "W_S8": 19, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "k.5", "K": 512, "blob_biases": 27, "name": "k.5", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.3", "weights": { "per_ch_qscale": 29, "W_S8": 25, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "127", "K": 512, "blob_biases": 33, "name": "127", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.3", "weights": { "per_ch_qscale": 35, "W_S8": 31, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 1, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "113", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "128", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "128", "begin_48": 0, "top": "128_0,128_1,128_2,128_3,128_4,128_5,128_6,128_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 1, "axis_w": 2, "bottom": "k.5", "axis_k": 0, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "transpose_6", "top": "transpose_6", "type": "transpose", "name": "transpose_6" }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 3, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "transpose_6", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "138", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "138", "begin_48": 0, "top": "138_0,138_1,138_2,138_3,138_4,138_5,138_6,138_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 1, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "127", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "147", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "147", "begin_48": 0, "top": "147_0,147_1,147_2,147_3,147_4,147_5,147_6,147_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "138_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "157_transpose_x", "top": "138_0_transposed", "type": "transpose", "name": "157_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "128_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "157_transpose_y", "top": "128_0_transposed", "type": "transpose", "name": "157_transpose_y" }, { "bottom": "138_0_transposed,128_0_transposed", "weights": {}, "debug_info": "157_batch_matmul", "top": "157_pre_transpose", "type": "batch_matmul", "name": "157_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "157_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "157", "top": "157", "type": "transpose", "name": "157" }, { "bottom": "157", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "159", "top": "159", "type": "elementwise", "name": "159", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "138_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "161_transpose_x", "top": "138_1_transposed", "type": "transpose", "name": "161_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "128_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "161_transpose_y", "top": "128_1_transposed", "type": "transpose", "name": "161_transpose_y" }, { "bottom": "138_1_transposed,128_1_transposed", "weights": {}, "debug_info": "161_batch_matmul", "top": "161_pre_transpose", "type": "batch_matmul", "name": "161_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "161_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "161", "top": "161", "type": "transpose", "name": "161" }, { "bottom": "161", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "163", "top": "163", "type": "elementwise", "name": "163", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "138_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "165_transpose_x", "top": "138_2_transposed", "type": "transpose", "name": "165_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "128_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "165_transpose_y", "top": "128_2_transposed", "type": "transpose", "name": "165_transpose_y" }, { "bottom": "138_2_transposed,128_2_transposed", "weights": {}, "debug_info": "165_batch_matmul", "top": "165_pre_transpose", "type": "batch_matmul", "name": "165_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "165_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "165", "top": "165", "type": "transpose", "name": "165" }, { "bottom": "165", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "167", "top": "167", "type": "elementwise", "name": "167", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "138_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "169_transpose_x", "top": "138_3_transposed", "type": "transpose", "name": "169_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "128_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "169_transpose_y", "top": "128_3_transposed", "type": "transpose", "name": "169_transpose_y" }, { "bottom": "138_3_transposed,128_3_transposed", "weights": {}, "debug_info": "169_batch_matmul", "top": "169_pre_transpose", "type": "batch_matmul", "name": "169_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "169_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "169", "top": "169", "type": "transpose", "name": "169" }, { "bottom": "169", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "171", "top": "171", "type": "elementwise", "name": "171", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "138_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "173_transpose_x", "top": "138_4_transposed", "type": "transpose", "name": "173_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "128_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "173_transpose_y", "top": "128_4_transposed", "type": "transpose", "name": "173_transpose_y" }, { "bottom": "138_4_transposed,128_4_transposed", "weights": {}, "debug_info": "173_batch_matmul", "top": "173_pre_transpose", "type": "batch_matmul", "name": "173_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "173_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "173", "top": "173", "type": "transpose", "name": "173" }, { "bottom": "173", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "175", "top": "175", "type": "elementwise", "name": "175", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "138_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "177_transpose_x", "top": "138_5_transposed", "type": "transpose", "name": "177_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "128_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "177_transpose_y", "top": "128_5_transposed", "type": "transpose", "name": "177_transpose_y" }, { "bottom": "138_5_transposed,128_5_transposed", "weights": {}, "debug_info": "177_batch_matmul", "top": "177_pre_transpose", "type": "batch_matmul", "name": "177_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "177_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "177", "top": "177", "type": "transpose", "name": "177" }, { "bottom": "177", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "179", "top": "179", "type": "elementwise", "name": "179", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "138_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "181_transpose_x", "top": "138_6_transposed", "type": "transpose", "name": "181_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "128_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "181_transpose_y", "top": "128_6_transposed", "type": "transpose", "name": "181_transpose_y" }, { "bottom": "138_6_transposed,128_6_transposed", "weights": {}, "debug_info": "181_batch_matmul", "top": "181_pre_transpose", "type": "batch_matmul", "name": "181_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "181_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "181", "top": "181", "type": "transpose", "name": "181" }, { "bottom": "181", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "183", "top": "183", "type": "elementwise", "name": "183", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "138_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "185_transpose_x", "top": "138_7_transposed", "type": "transpose", "name": "185_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "128_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "185_transpose_y", "top": "128_7_transposed", "type": "transpose", "name": "185_transpose_y" }, { "bottom": "138_7_transposed,128_7_transposed", "weights": {}, "debug_info": "185_batch_matmul", "top": "185_pre_transpose", "type": "batch_matmul", "name": "185_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "185_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "185", "top": "185", "type": "transpose", "name": "185" }, { "bottom": "185", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "187", "top": "187", "type": "elementwise", "name": "187", "beta": 0 }, { "bottom": "159,163,167,171,175,179,183,187", "weights": {}, "nd_mode": true, "axis": 2, "debug_info": "attn_weights.2", "top": "attn_weights.2", "type": "general_concat", "name": "attn_weights.2" }, { "bottom": "qk_mask", "alpha": -10000, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "", "top": "qk_mask.1", "type": "elementwise", "name": "qk_mask/scaling", "beta": 0 }, { "bottom": "attn_weights.2,qk_mask.1", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "attn_weights0.2", "nd_mode": true, "top": "attn_weights0.2", "type": "elementwise", "name": "attn_weights0.2", "beta": 0 }, { "bottom": "attn_weights0.2", "weights": {}, "debug_info": "input.5", "top": "input.5", "C": 2, "type": "softmax", "name": "input.5" }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 2, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "input.5", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "193", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "193", "begin_48": 0, "top": "193_0,193_1,193_2,193_3,193_4,193_5,193_6,193_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "147_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "203_transpose_x", "top": "147_0_transposed", "type": "transpose", "name": "203_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "193_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "203_transpose_y", "top": "193_0_transposed", "type": "transpose", "name": "203_transpose_y" }, { "bottom": "147_0_transposed,193_0_transposed", "weights": {}, "debug_info": "203_batch_matmul", "top": "203_pre_transpose", "type": "batch_matmul", "name": "203_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "203_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "203", "top": "203", "type": "transpose", "name": "203" }, { "axis_h": 2, "axis_w": 0, "bottom": "147_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "205_transpose_x", "top": "147_1_transposed", "type": "transpose", "name": "205_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "193_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "205_transpose_y", "top": "193_1_transposed", "type": "transpose", "name": "205_transpose_y" }, { "bottom": "147_1_transposed,193_1_transposed", "weights": {}, "debug_info": "205_batch_matmul", "top": "205_pre_transpose", "type": "batch_matmul", "name": "205_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "205_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "205", "top": "205", "type": "transpose", "name": "205" }, { "axis_h": 2, "axis_w": 0, "bottom": "147_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "207_transpose_x", "top": "147_2_transposed", "type": "transpose", "name": "207_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "193_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "207_transpose_y", "top": "193_2_transposed", "type": "transpose", "name": "207_transpose_y" }, { "bottom": "147_2_transposed,193_2_transposed", "weights": {}, "debug_info": "207_batch_matmul", "top": "207_pre_transpose", "type": "batch_matmul", "name": "207_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "207_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "207", "top": "207", "type": "transpose", "name": "207" }, { "axis_h": 2, "axis_w": 0, "bottom": "147_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "209_transpose_x", "top": "147_3_transposed", "type": "transpose", "name": "209_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "193_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "209_transpose_y", "top": "193_3_transposed", "type": "transpose", "name": "209_transpose_y" }, { "bottom": "147_3_transposed,193_3_transposed", "weights": {}, "debug_info": "209_batch_matmul", "top": "209_pre_transpose", "type": "batch_matmul", "name": "209_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "209_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "209", "top": "209", "type": "transpose", "name": "209" }, { "axis_h": 2, "axis_w": 0, "bottom": "147_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "211_transpose_x", "top": "147_4_transposed", "type": "transpose", "name": "211_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "193_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "211_transpose_y", "top": "193_4_transposed", "type": "transpose", "name": "211_transpose_y" }, { "bottom": "147_4_transposed,193_4_transposed", "weights": {}, "debug_info": "211_batch_matmul", "top": "211_pre_transpose", "type": "batch_matmul", "name": "211_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "211_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "211", "top": "211", "type": "transpose", "name": "211" }, { "axis_h": 2, "axis_w": 0, "bottom": "147_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "213_transpose_x", "top": "147_5_transposed", "type": "transpose", "name": "213_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "193_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "213_transpose_y", "top": "193_5_transposed", "type": "transpose", "name": "213_transpose_y" }, { "bottom": "147_5_transposed,193_5_transposed", "weights": {}, "debug_info": "213_batch_matmul", "top": "213_pre_transpose", "type": "batch_matmul", "name": "213_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "213_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "213", "top": "213", "type": "transpose", "name": "213" }, { "axis_h": 2, "axis_w": 0, "bottom": "147_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "215_transpose_x", "top": "147_6_transposed", "type": "transpose", "name": "215_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "193_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "215_transpose_y", "top": "193_6_transposed", "type": "transpose", "name": "215_transpose_y" }, { "bottom": "147_6_transposed,193_6_transposed", "weights": {}, "debug_info": "215_batch_matmul", "top": "215_pre_transpose", "type": "batch_matmul", "name": "215_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "215_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "215", "top": "215", "type": "transpose", "name": "215" }, { "axis_h": 2, "axis_w": 0, "bottom": "147_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "217_transpose_x", "top": "147_7_transposed", "type": "transpose", "name": "217_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "193_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "217_transpose_y", "top": "193_7_transposed", "type": "transpose", "name": "217_transpose_y" }, { "bottom": "147_7_transposed,193_7_transposed", "weights": {}, "debug_info": "217_batch_matmul", "top": "217_pre_transpose", "type": "batch_matmul", "name": "217_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "217_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "217", "top": "217", "type": "transpose", "name": "217" }, { "weights": {}, "debug_info": "attn.5", "top": "attn.5", "type": "concat", "name": "attn.5", "bottom": "203,205,207,209,211,213,215,217" }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "inputs.5", "K": 512, "blob_biases": 39, "name": "inputs.5", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "attn.5", "weights": { "per_ch_qscale": 41, "W_S8": 37, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "axis_mode": 4, "bottom": "inputs.5", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.5", "use_version": 1, "top": "channels_mean.5", "type": "reduce", "name": "channels_mean.5" }, { "alpha": -1, "bottom": "channels_mean.5", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.5", "top": "_neg_y_zero_mean.5", "type": "activation", "name": "_neg_y_zero_mean.5", "beta": 0 }, { "bottom": "inputs.5,_neg_y_zero_mean.5", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.5", "top": "zero_mean.5", "type": "elementwise", "name": "zero_mean.5", "beta": 0 }, { "bottom": "zero_mean.5,zero_mean.5", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.5", "top": "zero_mean_sq.5", "type": "elementwise", "name": "zero_mean_sq.5", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.5", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "240", "use_version": 1, "top": "240", "type": "reduce", "name": "240" }, { "bottom": "240", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "242", "top": "242", "type": "elementwise", "name": "242", "beta": 0 }, { "bottom": "242", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.5", "top": "denom.5", "type": "elementwise", "name": "denom.5", "beta": 0 }, { "bottom": "zero_mean.5,denom.5", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.5", "top": "out.5", "type": "elementwise", "name": "out.5", "beta": 0 }, { "top": "246", "w": 1, "h": 1, "name": "246", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "246", "n": 1, "weights": {}, "constant_blob": 577 }, { "bottom": "out.5,246", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "247", "top": "247", "type": "elementwise", "name": "247", "beta": 0 }, { "top": "249", "w": 1, "h": 1, "name": "249", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "249", "n": 1, "weights": {}, "constant_blob": 337 }, { "bottom": "247,249", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.7", "top": "input.7", "type": "elementwise", "name": "input.7", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.6", "K": 512, "blob_biases": 341, "name": "x.6", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 64, "bottom": "input.7", "weights": { "per_ch_qscale": 343, "W_S8": 339, "per_ch_qbias": 579 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.6", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "259", "top": "259", "type": "elementwise", "name": "259", "beta": 0 }, { "bottom": "259", "weights": {}, "mode": 3, "debug_info": "260", "top": "260", "type": "activation", "name": "260" }, { "bottom": "x.6,260", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.13", "top": "input.13", "type": "elementwise", "name": "input.13", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.8", "K": 64, "blob_biases": 347, "name": "x.8", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.13", "weights": { "per_ch_qscale": 349, "W_S8": 345, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.8,inputs.5", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "attn.7", "top": "attn.7", "type": "elementwise", "name": "attn.7", "beta": 0 }, { "bottom": "transpose_7,attn.7", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "inputs.2", "top": "inputs.2", "type": "elementwise", "name": "inputs.2", "beta": 0 }, { "axis_mode": 4, "bottom": "inputs.2", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.7", "use_version": 1, "top": "channels_mean.7", "type": "reduce", "name": "channels_mean.7" }, { "alpha": -1, "bottom": "channels_mean.7", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.7", "top": "_neg_y_zero_mean.7", "type": "activation", "name": "_neg_y_zero_mean.7", "beta": 0 }, { "bottom": "inputs.2,_neg_y_zero_mean.7", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.7", "top": "zero_mean.7", "type": "elementwise", "name": "zero_mean.7", "beta": 0 }, { "bottom": "zero_mean.7,zero_mean.7", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.7", "top": "zero_mean_sq.7", "type": "elementwise", "name": "zero_mean_sq.7", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.7", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "278", "use_version": 1, "top": "278", "type": "reduce", "name": "278" }, { "bottom": "278", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "280", "top": "280", "type": "elementwise", "name": "280", "beta": 0 }, { "bottom": "280", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.7", "top": "denom.7", "type": "elementwise", "name": "denom.7", "beta": 0 }, { "bottom": "zero_mean.7,denom.7", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.7", "top": "out.7", "type": "elementwise", "name": "out.7", "beta": 0 }, { "top": "284", "w": 1, "h": 1, "name": "284", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "284", "n": 1, "weights": {}, "constant_blob": 581 }, { "bottom": "out.7,284", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "285", "top": "285", "type": "elementwise", "name": "285", "beta": 0 }, { "top": "287", "w": 1, "h": 1, "name": "287", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "287", "n": 1, "weights": {}, "constant_blob": 353 }, { "bottom": "285,287", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.15", "top": "input.15", "type": "elementwise", "name": "input.15", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.10", "K": 512, "blob_biases": 49, "name": "x.10", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 2048, "bottom": "input.15", "weights": { "per_ch_qscale": 51, "W_S8": 47, "per_ch_qbias": 299 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.10", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "299", "top": "299", "type": "elementwise", "name": "299", "beta": 0 }, { "bottom": "299", "weights": {}, "mode": 3, "debug_info": "300", "top": "300", "type": "activation", "name": "300" }, { "bottom": "x.10,300", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.17", "top": "input.17", "type": "elementwise", "name": "input.17", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "input0.3", "K": 2048, "blob_biases": 55, "name": "input0.3", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.17", "weights": { "per_ch_qscale": 57, "W_S8": 53, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "axis_mode": 4, "bottom": "input0.3", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.9", "use_version": 1, "top": "channels_mean.9", "type": "reduce", "name": "channels_mean.9" }, { "alpha": -1, "bottom": "channels_mean.9", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.9", "top": "_neg_y_zero_mean.9", "type": "activation", "name": "_neg_y_zero_mean.9", "beta": 0 }, { "bottom": "input0.3,_neg_y_zero_mean.9", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.9", "top": "zero_mean.9", "type": "elementwise", "name": "zero_mean.9", "beta": 0 }, { "bottom": "zero_mean.9,zero_mean.9", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.9", "top": "zero_mean_sq.9", "type": "elementwise", "name": "zero_mean_sq.9", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.9", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "320", "use_version": 1, "top": "320", "type": "reduce", "name": "320" }, { "bottom": "320", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "322", "top": "322", "type": "elementwise", "name": "322", "beta": 0 }, { "bottom": "322", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.9", "top": "denom.9", "type": "elementwise", "name": "denom.9", "beta": 0 }, { "bottom": "zero_mean.9,denom.9", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.9", "top": "out.9", "type": "elementwise", "name": "out.9", "beta": 0 }, { "top": "326", "w": 1, "h": 1, "name": "326", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "326", "n": 1, "weights": {}, "constant_blob": 583 }, { "bottom": "out.9,326", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "327", "top": "327", "type": "elementwise", "name": "327", "beta": 0 }, { "top": "329", "w": 1, "h": 1, "name": "329", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "329", "n": 1, "weights": {}, "constant_blob": 357 }, { "bottom": "327,329", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.19", "top": "input.19", "type": "elementwise", "name": "input.19", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.12", "K": 512, "blob_biases": 361, "name": "x.12", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 64, "bottom": "input.19", "weights": { "per_ch_qscale": 363, "W_S8": 359, "per_ch_qbias": 579 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.12", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "339", "top": "339", "type": "elementwise", "name": "339", "beta": 0 }, { "bottom": "339", "weights": {}, "mode": 3, "debug_info": "340", "top": "340", "type": "activation", "name": "340" }, { "bottom": "x.12,340", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.21", "top": "input.21", "type": "elementwise", "name": "input.21", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.14", "K": 64, "blob_biases": 367, "name": "x.14", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.21", "weights": { "per_ch_qscale": 369, "W_S8": 365, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.14,input0.3", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "f.2", "top": "f.2", "type": "elementwise", "name": "f.2", "beta": 0 }, { "bottom": "f.2,inputs.2", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "x1.2", "top": "x1.2", "type": "elementwise", "name": "x1.2", "beta": 0 }, { "bottom": "transpose_7", "alpha": 0, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "352", "top": "352", "type": "elementwise", "name": "352", "beta": 0 }, { "bottom": "352,x1.2", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "inputs.3", "top": "inputs.3", "type": "elementwise", "name": "inputs.3", "beta": 0 }, { "axis_mode": 4, "bottom": "inputs.3", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.11", "use_version": 1, "top": "channels_mean.11", "type": "reduce", "name": "channels_mean.11" }, { "alpha": -1, "bottom": "channels_mean.11", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.11", "top": "_neg_y_zero_mean.11", "type": "activation", "name": "_neg_y_zero_mean.11", "beta": 0 }, { "bottom": "inputs.3,_neg_y_zero_mean.11", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.11", "top": "zero_mean.11", "type": "elementwise", "name": "zero_mean.11", "beta": 0 }, { "bottom": "zero_mean.11,zero_mean.11", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.11", "top": "zero_mean_sq.11", "type": "elementwise", "name": "zero_mean_sq.11", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.11", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "369", "use_version": 1, "top": "369", "type": "reduce", "name": "369" }, { "bottom": "369", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "371", "top": "371", "type": "elementwise", "name": "371", "beta": 0 }, { "bottom": "371", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.11", "top": "denom.11", "type": "elementwise", "name": "denom.11", "beta": 0 }, { "bottom": "zero_mean.11,denom.11", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.11", "top": "out.11", "type": "elementwise", "name": "out.11", "beta": 0 }, { "top": "375", "w": 1, "h": 1, "name": "375", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "375", "n": 1, "weights": {}, "constant_blob": 585 }, { "bottom": "out.11,375", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "376", "top": "376", "type": "elementwise", "name": "376", "beta": 0 }, { "top": "378", "w": 1, "h": 1, "name": "378", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "378", "n": 1, "weights": {}, "constant_blob": 373 }, { "bottom": "376,378", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "k.7", "top": "k.7", "type": "elementwise", "name": "k.7", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "393", "K": 512, "blob_biases": 65, "name": "393", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.7", "weights": { "per_ch_qscale": 67, "W_S8": 63, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "k.9", "K": 512, "blob_biases": 71, "name": "k.9", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.7", "weights": { "per_ch_qscale": 73, "W_S8": 69, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "407", "K": 512, "blob_biases": 77, "name": "407", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.7", "weights": { "per_ch_qscale": 79, "W_S8": 75, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 1, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "393", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "408", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "408", "begin_48": 0, "top": "408_0,408_1,408_2,408_3,408_4,408_5,408_6,408_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 1, "axis_w": 2, "bottom": "k.9", "axis_k": 0, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "transpose_5", "top": "transpose_5", "type": "transpose", "name": "transpose_5" }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 3, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "transpose_5", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "418", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "418", "begin_48": 0, "top": "418_0,418_1,418_2,418_3,418_4,418_5,418_6,418_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 1, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "407", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "427", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "427", "begin_48": 0, "top": "427_0,427_1,427_2,427_3,427_4,427_5,427_6,427_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "418_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "437_transpose_x", "top": "418_0_transposed", "type": "transpose", "name": "437_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "408_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "437_transpose_y", "top": "408_0_transposed", "type": "transpose", "name": "437_transpose_y" }, { "bottom": "418_0_transposed,408_0_transposed", "weights": {}, "debug_info": "437_batch_matmul", "top": "437_pre_transpose", "type": "batch_matmul", "name": "437_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "437_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "437", "top": "437", "type": "transpose", "name": "437" }, { "bottom": "437", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "439", "top": "439", "type": "elementwise", "name": "439", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "418_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "441_transpose_x", "top": "418_1_transposed", "type": "transpose", "name": "441_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "408_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "441_transpose_y", "top": "408_1_transposed", "type": "transpose", "name": "441_transpose_y" }, { "bottom": "418_1_transposed,408_1_transposed", "weights": {}, "debug_info": "441_batch_matmul", "top": "441_pre_transpose", "type": "batch_matmul", "name": "441_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "441_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "441", "top": "441", "type": "transpose", "name": "441" }, { "bottom": "441", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "443", "top": "443", "type": "elementwise", "name": "443", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "418_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "445_transpose_x", "top": "418_2_transposed", "type": "transpose", "name": "445_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "408_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "445_transpose_y", "top": "408_2_transposed", "type": "transpose", "name": "445_transpose_y" }, { "bottom": "418_2_transposed,408_2_transposed", "weights": {}, "debug_info": "445_batch_matmul", "top": "445_pre_transpose", "type": "batch_matmul", "name": "445_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "445_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "445", "top": "445", "type": "transpose", "name": "445" }, { "bottom": "445", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "447", "top": "447", "type": "elementwise", "name": "447", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "418_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "449_transpose_x", "top": "418_3_transposed", "type": "transpose", "name": "449_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "408_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "449_transpose_y", "top": "408_3_transposed", "type": "transpose", "name": "449_transpose_y" }, { "bottom": "418_3_transposed,408_3_transposed", "weights": {}, "debug_info": "449_batch_matmul", "top": "449_pre_transpose", "type": "batch_matmul", "name": "449_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "449_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "449", "top": "449", "type": "transpose", "name": "449" }, { "bottom": "449", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "451", "top": "451", "type": "elementwise", "name": "451", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "418_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "453_transpose_x", "top": "418_4_transposed", "type": "transpose", "name": "453_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "408_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "453_transpose_y", "top": "408_4_transposed", "type": "transpose", "name": "453_transpose_y" }, { "bottom": "418_4_transposed,408_4_transposed", "weights": {}, "debug_info": "453_batch_matmul", "top": "453_pre_transpose", "type": "batch_matmul", "name": "453_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "453_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "453", "top": "453", "type": "transpose", "name": "453" }, { "bottom": "453", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "455", "top": "455", "type": "elementwise", "name": "455", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "418_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "457_transpose_x", "top": "418_5_transposed", "type": "transpose", "name": "457_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "408_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "457_transpose_y", "top": "408_5_transposed", "type": "transpose", "name": "457_transpose_y" }, { "bottom": "418_5_transposed,408_5_transposed", "weights": {}, "debug_info": "457_batch_matmul", "top": "457_pre_transpose", "type": "batch_matmul", "name": "457_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "457_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "457", "top": "457", "type": "transpose", "name": "457" }, { "bottom": "457", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "459", "top": "459", "type": "elementwise", "name": "459", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "418_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "461_transpose_x", "top": "418_6_transposed", "type": "transpose", "name": "461_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "408_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "461_transpose_y", "top": "408_6_transposed", "type": "transpose", "name": "461_transpose_y" }, { "bottom": "418_6_transposed,408_6_transposed", "weights": {}, "debug_info": "461_batch_matmul", "top": "461_pre_transpose", "type": "batch_matmul", "name": "461_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "461_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "461", "top": "461", "type": "transpose", "name": "461" }, { "bottom": "461", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "463", "top": "463", "type": "elementwise", "name": "463", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "418_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "465_transpose_x", "top": "418_7_transposed", "type": "transpose", "name": "465_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "408_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "465_transpose_y", "top": "408_7_transposed", "type": "transpose", "name": "465_transpose_y" }, { "bottom": "418_7_transposed,408_7_transposed", "weights": {}, "debug_info": "465_batch_matmul", "top": "465_pre_transpose", "type": "batch_matmul", "name": "465_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "465_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "465", "top": "465", "type": "transpose", "name": "465" }, { "bottom": "465", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "467", "top": "467", "type": "elementwise", "name": "467", "beta": 0 }, { "bottom": "439,443,447,451,455,459,463,467", "weights": {}, "nd_mode": true, "axis": 2, "debug_info": "attn_weights.4", "top": "attn_weights.4", "type": "general_concat", "name": "attn_weights.4" }, { "bottom": "attn_weights.4,qk_mask.1", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "attn_weights0.4", "nd_mode": true, "top": "attn_weights0.4", "type": "elementwise", "name": "attn_weights0.4", "beta": 0 }, { "bottom": "attn_weights0.4", "weights": {}, "debug_info": "input.23", "top": "input.23", "C": 2, "type": "softmax", "name": "input.23" }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 2, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "input.23", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "473", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "473", "begin_48": 0, "top": "473_0,473_1,473_2,473_3,473_4,473_5,473_6,473_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "427_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "483_transpose_x", "top": "427_0_transposed", "type": "transpose", "name": "483_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "473_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "483_transpose_y", "top": "473_0_transposed", "type": "transpose", "name": "483_transpose_y" }, { "bottom": "427_0_transposed,473_0_transposed", "weights": {}, "debug_info": "483_batch_matmul", "top": "483_pre_transpose", "type": "batch_matmul", "name": "483_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "483_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "483", "top": "483", "type": "transpose", "name": "483" }, { "axis_h": 2, "axis_w": 0, "bottom": "427_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "485_transpose_x", "top": "427_1_transposed", "type": "transpose", "name": "485_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "473_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "485_transpose_y", "top": "473_1_transposed", "type": "transpose", "name": "485_transpose_y" }, { "bottom": "427_1_transposed,473_1_transposed", "weights": {}, "debug_info": "485_batch_matmul", "top": "485_pre_transpose", "type": "batch_matmul", "name": "485_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "485_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "485", "top": "485", "type": "transpose", "name": "485" }, { "axis_h": 2, "axis_w": 0, "bottom": "427_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "487_transpose_x", "top": "427_2_transposed", "type": "transpose", "name": "487_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "473_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "487_transpose_y", "top": "473_2_transposed", "type": "transpose", "name": "487_transpose_y" }, { "bottom": "427_2_transposed,473_2_transposed", "weights": {}, "debug_info": "487_batch_matmul", "top": "487_pre_transpose", "type": "batch_matmul", "name": "487_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "487_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "487", "top": "487", "type": "transpose", "name": "487" }, { "axis_h": 2, "axis_w": 0, "bottom": "427_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "489_transpose_x", "top": "427_3_transposed", "type": "transpose", "name": "489_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "473_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "489_transpose_y", "top": "473_3_transposed", "type": "transpose", "name": "489_transpose_y" }, { "bottom": "427_3_transposed,473_3_transposed", "weights": {}, "debug_info": "489_batch_matmul", "top": "489_pre_transpose", "type": "batch_matmul", "name": "489_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "489_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "489", "top": "489", "type": "transpose", "name": "489" }, { "axis_h": 2, "axis_w": 0, "bottom": "427_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "491_transpose_x", "top": "427_4_transposed", "type": "transpose", "name": "491_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "473_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "491_transpose_y", "top": "473_4_transposed", "type": "transpose", "name": "491_transpose_y" }, { "bottom": "427_4_transposed,473_4_transposed", "weights": {}, "debug_info": "491_batch_matmul", "top": "491_pre_transpose", "type": "batch_matmul", "name": "491_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "491_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "491", "top": "491", "type": "transpose", "name": "491" }, { "axis_h": 2, "axis_w": 0, "bottom": "427_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "493_transpose_x", "top": "427_5_transposed", "type": "transpose", "name": "493_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "473_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "493_transpose_y", "top": "473_5_transposed", "type": "transpose", "name": "493_transpose_y" }, { "bottom": "427_5_transposed,473_5_transposed", "weights": {}, "debug_info": "493_batch_matmul", "top": "493_pre_transpose", "type": "batch_matmul", "name": "493_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "493_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "493", "top": "493", "type": "transpose", "name": "493" }, { "axis_h": 2, "axis_w": 0, "bottom": "427_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "495_transpose_x", "top": "427_6_transposed", "type": "transpose", "name": "495_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "473_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "495_transpose_y", "top": "473_6_transposed", "type": "transpose", "name": "495_transpose_y" }, { "bottom": "427_6_transposed,473_6_transposed", "weights": {}, "debug_info": "495_batch_matmul", "top": "495_pre_transpose", "type": "batch_matmul", "name": "495_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "495_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "495", "top": "495", "type": "transpose", "name": "495" }, { "axis_h": 2, "axis_w": 0, "bottom": "427_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "497_transpose_x", "top": "427_7_transposed", "type": "transpose", "name": "497_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "473_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "497_transpose_y", "top": "473_7_transposed", "type": "transpose", "name": "497_transpose_y" }, { "bottom": "427_7_transposed,473_7_transposed", "weights": {}, "debug_info": "497_batch_matmul", "top": "497_pre_transpose", "type": "batch_matmul", "name": "497_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "497_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "497", "top": "497", "type": "transpose", "name": "497" }, { "weights": {}, "debug_info": "attn.11", "top": "attn.11", "type": "concat", "name": "attn.11", "bottom": "483,485,487,489,491,493,495,497" }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "inputs.9", "K": 512, "blob_biases": 83, "name": "inputs.9", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "attn.11", "weights": { "per_ch_qscale": 85, "W_S8": 81, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "axis_mode": 4, "bottom": "inputs.9", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.13", "use_version": 1, "top": "channels_mean.13", "type": "reduce", "name": "channels_mean.13" }, { "alpha": -1, "bottom": "channels_mean.13", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.13", "top": "_neg_y_zero_mean.13", "type": "activation", "name": "_neg_y_zero_mean.13", "beta": 0 }, { "bottom": "inputs.9,_neg_y_zero_mean.13", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.13", "top": "zero_mean.13", "type": "elementwise", "name": "zero_mean.13", "beta": 0 }, { "bottom": "zero_mean.13,zero_mean.13", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.13", "top": "zero_mean_sq.13", "type": "elementwise", "name": "zero_mean_sq.13", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.13", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "520", "use_version": 1, "top": "520", "type": "reduce", "name": "520" }, { "bottom": "520", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "522", "top": "522", "type": "elementwise", "name": "522", "beta": 0 }, { "bottom": "522", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.13", "top": "denom.13", "type": "elementwise", "name": "denom.13", "beta": 0 }, { "bottom": "zero_mean.13,denom.13", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.13", "top": "out.13", "type": "elementwise", "name": "out.13", "beta": 0 }, { "top": "526", "w": 1, "h": 1, "name": "526", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "526", "n": 1, "weights": {}, "constant_blob": 587 }, { "bottom": "out.13,526", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "527", "top": "527", "type": "elementwise", "name": "527", "beta": 0 }, { "top": "529", "w": 1, "h": 1, "name": "529", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "529", "n": 1, "weights": {}, "constant_blob": 377 }, { "bottom": "527,529", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.25", "top": "input.25", "type": "elementwise", "name": "input.25", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.16", "K": 512, "blob_biases": 381, "name": "x.16", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 64, "bottom": "input.25", "weights": { "per_ch_qscale": 383, "W_S8": 379, "per_ch_qbias": 579 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.16", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "539", "top": "539", "type": "elementwise", "name": "539", "beta": 0 }, { "bottom": "539", "weights": {}, "mode": 3, "debug_info": "540", "top": "540", "type": "activation", "name": "540" }, { "bottom": "x.16,540", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.27", "top": "input.27", "type": "elementwise", "name": "input.27", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.18", "K": 64, "blob_biases": 387, "name": "x.18", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.27", "weights": { "per_ch_qscale": 389, "W_S8": 385, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.18,inputs.9", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "attn.13", "top": "attn.13", "type": "elementwise", "name": "attn.13", "beta": 0 }, { "bottom": "inputs.3,attn.13", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "inputs0.4", "top": "inputs0.4", "type": "elementwise", "name": "inputs0.4", "beta": 0 }, { "axis_mode": 4, "bottom": "inputs0.4", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.15", "use_version": 1, "top": "channels_mean.15", "type": "reduce", "name": "channels_mean.15" }, { "alpha": -1, "bottom": "channels_mean.15", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.15", "top": "_neg_y_zero_mean.15", "type": "activation", "name": "_neg_y_zero_mean.15", "beta": 0 }, { "bottom": "inputs0.4,_neg_y_zero_mean.15", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.15", "top": "zero_mean.15", "type": "elementwise", "name": "zero_mean.15", "beta": 0 }, { "bottom": "zero_mean.15,zero_mean.15", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.15", "top": "zero_mean_sq.15", "type": "elementwise", "name": "zero_mean_sq.15", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.15", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "558", "use_version": 1, "top": "558", "type": "reduce", "name": "558" }, { "bottom": "558", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "560", "top": "560", "type": "elementwise", "name": "560", "beta": 0 }, { "bottom": "560", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.15", "top": "denom.15", "type": "elementwise", "name": "denom.15", "beta": 0 }, { "bottom": "zero_mean.15,denom.15", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.15", "top": "out.15", "type": "elementwise", "name": "out.15", "beta": 0 }, { "top": "564", "w": 1, "h": 1, "name": "564", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "564", "n": 1, "weights": {}, "constant_blob": 589 }, { "bottom": "out.15,564", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "565", "top": "565", "type": "elementwise", "name": "565", "beta": 0 }, { "top": "567", "w": 1, "h": 1, "name": "567", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "567", "n": 1, "weights": {}, "constant_blob": 393 }, { "bottom": "565,567", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.29", "top": "input.29", "type": "elementwise", "name": "input.29", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.20", "K": 512, "blob_biases": 93, "name": "x.20", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 2048, "bottom": "input.29", "weights": { "per_ch_qscale": 95, "W_S8": 91, "per_ch_qbias": 299 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.20", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "579", "top": "579", "type": "elementwise", "name": "579", "beta": 0 }, { "bottom": "579", "weights": {}, "mode": 3, "debug_info": "580", "top": "580", "type": "activation", "name": "580" }, { "bottom": "x.20,580", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.31", "top": "input.31", "type": "elementwise", "name": "input.31", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "input0.7", "K": 2048, "blob_biases": 99, "name": "input0.7", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.31", "weights": { "per_ch_qscale": 101, "W_S8": 97, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "axis_mode": 4, "bottom": "input0.7", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.17", "use_version": 1, "top": "channels_mean.17", "type": "reduce", "name": "channels_mean.17" }, { "alpha": -1, "bottom": "channels_mean.17", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.17", "top": "_neg_y_zero_mean.17", "type": "activation", "name": "_neg_y_zero_mean.17", "beta": 0 }, { "bottom": "input0.7,_neg_y_zero_mean.17", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.17", "top": "zero_mean.17", "type": "elementwise", "name": "zero_mean.17", "beta": 0 }, { "bottom": "zero_mean.17,zero_mean.17", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.17", "top": "zero_mean_sq.17", "type": "elementwise", "name": "zero_mean_sq.17", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.17", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "600", "use_version": 1, "top": "600", "type": "reduce", "name": "600" }, { "bottom": "600", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "602", "top": "602", "type": "elementwise", "name": "602", "beta": 0 }, { "bottom": "602", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.17", "top": "denom.17", "type": "elementwise", "name": "denom.17", "beta": 0 }, { "bottom": "zero_mean.17,denom.17", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.17", "top": "out.17", "type": "elementwise", "name": "out.17", "beta": 0 }, { "top": "606", "w": 1, "h": 1, "name": "606", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "606", "n": 1, "weights": {}, "constant_blob": 591 }, { "bottom": "out.17,606", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "607", "top": "607", "type": "elementwise", "name": "607", "beta": 0 }, { "top": "609", "w": 1, "h": 1, "name": "609", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "609", "n": 1, "weights": {}, "constant_blob": 397 }, { "bottom": "607,609", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.33", "top": "input.33", "type": "elementwise", "name": "input.33", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.22", "K": 512, "blob_biases": 401, "name": "x.22", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 64, "bottom": "input.33", "weights": { "per_ch_qscale": 403, "W_S8": 399, "per_ch_qbias": 579 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.22", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "619", "top": "619", "type": "elementwise", "name": "619", "beta": 0 }, { "bottom": "619", "weights": {}, "mode": 3, "debug_info": "620", "top": "620", "type": "activation", "name": "620" }, { "bottom": "x.22,620", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.35", "top": "input.35", "type": "elementwise", "name": "input.35", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.24", "K": 64, "blob_biases": 407, "name": "x.24", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.35", "weights": { "per_ch_qscale": 409, "W_S8": 405, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.24,input0.7", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "f.4", "top": "f.4", "type": "elementwise", "name": "f.4", "beta": 0 }, { "bottom": "f.4,inputs0.4", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "x1.4", "top": "x1.4", "type": "elementwise", "name": "x1.4", "beta": 0 }, { "bottom": "inputs.3", "alpha": 0, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "632", "top": "632", "type": "elementwise", "name": "632", "beta": 0 }, { "bottom": "632,x1.4", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "inputs0.2", "top": "inputs0.2", "type": "elementwise", "name": "inputs0.2", "beta": 0 }, { "axis_mode": 4, "bottom": "inputs0.2", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.19", "use_version": 1, "top": "channels_mean.19", "type": "reduce", "name": "channels_mean.19" }, { "alpha": -1, "bottom": "channels_mean.19", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.19", "top": "_neg_y_zero_mean.19", "type": "activation", "name": "_neg_y_zero_mean.19", "beta": 0 }, { "bottom": "inputs0.2,_neg_y_zero_mean.19", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.19", "top": "zero_mean.19", "type": "elementwise", "name": "zero_mean.19", "beta": 0 }, { "bottom": "zero_mean.19,zero_mean.19", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.19", "top": "zero_mean_sq.19", "type": "elementwise", "name": "zero_mean_sq.19", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.19", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "649", "use_version": 1, "top": "649", "type": "reduce", "name": "649" }, { "bottom": "649", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "651", "top": "651", "type": "elementwise", "name": "651", "beta": 0 }, { "bottom": "651", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.19", "top": "denom.19", "type": "elementwise", "name": "denom.19", "beta": 0 }, { "bottom": "zero_mean.19,denom.19", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.19", "top": "out.19", "type": "elementwise", "name": "out.19", "beta": 0 }, { "top": "655", "w": 1, "h": 1, "name": "655", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "655", "n": 1, "weights": {}, "constant_blob": 593 }, { "bottom": "out.19,655", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "656", "top": "656", "type": "elementwise", "name": "656", "beta": 0 }, { "top": "658", "w": 1, "h": 1, "name": "658", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "658", "n": 1, "weights": {}, "constant_blob": 413 }, { "bottom": "656,658", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "k.11", "top": "k.11", "type": "elementwise", "name": "k.11", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "673", "K": 512, "blob_biases": 109, "name": "673", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.11", "weights": { "per_ch_qscale": 111, "W_S8": 107, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "k.13", "K": 512, "blob_biases": 115, "name": "k.13", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.11", "weights": { "per_ch_qscale": 117, "W_S8": 113, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "687", "K": 512, "blob_biases": 121, "name": "687", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.11", "weights": { "per_ch_qscale": 123, "W_S8": 119, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 1, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "673", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "688", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "688", "begin_48": 0, "top": "688_0,688_1,688_2,688_3,688_4,688_5,688_6,688_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 1, "axis_w": 2, "bottom": "k.13", "axis_k": 0, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "transpose_4", "top": "transpose_4", "type": "transpose", "name": "transpose_4" }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 3, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "transpose_4", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "698", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "698", "begin_48": 0, "top": "698_0,698_1,698_2,698_3,698_4,698_5,698_6,698_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 1, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "687", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "707", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "707", "begin_48": 0, "top": "707_0,707_1,707_2,707_3,707_4,707_5,707_6,707_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "698_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "717_transpose_x", "top": "698_0_transposed", "type": "transpose", "name": "717_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "688_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "717_transpose_y", "top": "688_0_transposed", "type": "transpose", "name": "717_transpose_y" }, { "bottom": "698_0_transposed,688_0_transposed", "weights": {}, "debug_info": "717_batch_matmul", "top": "717_pre_transpose", "type": "batch_matmul", "name": "717_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "717_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "717", "top": "717", "type": "transpose", "name": "717" }, { "bottom": "717", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "719", "top": "719", "type": "elementwise", "name": "719", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "698_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "721_transpose_x", "top": "698_1_transposed", "type": "transpose", "name": "721_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "688_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "721_transpose_y", "top": "688_1_transposed", "type": "transpose", "name": "721_transpose_y" }, { "bottom": "698_1_transposed,688_1_transposed", "weights": {}, "debug_info": "721_batch_matmul", "top": "721_pre_transpose", "type": "batch_matmul", "name": "721_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "721_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "721", "top": "721", "type": "transpose", "name": "721" }, { "bottom": "721", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "723", "top": "723", "type": "elementwise", "name": "723", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "698_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "725_transpose_x", "top": "698_2_transposed", "type": "transpose", "name": "725_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "688_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "725_transpose_y", "top": "688_2_transposed", "type": "transpose", "name": "725_transpose_y" }, { "bottom": "698_2_transposed,688_2_transposed", "weights": {}, "debug_info": "725_batch_matmul", "top": "725_pre_transpose", "type": "batch_matmul", "name": "725_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "725_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "725", "top": "725", "type": "transpose", "name": "725" }, { "bottom": "725", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "727", "top": "727", "type": "elementwise", "name": "727", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "698_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "729_transpose_x", "top": "698_3_transposed", "type": "transpose", "name": "729_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "688_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "729_transpose_y", "top": "688_3_transposed", "type": "transpose", "name": "729_transpose_y" }, { "bottom": "698_3_transposed,688_3_transposed", "weights": {}, "debug_info": "729_batch_matmul", "top": "729_pre_transpose", "type": "batch_matmul", "name": "729_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "729_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "729", "top": "729", "type": "transpose", "name": "729" }, { "bottom": "729", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "731", "top": "731", "type": "elementwise", "name": "731", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "698_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "733_transpose_x", "top": "698_4_transposed", "type": "transpose", "name": "733_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "688_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "733_transpose_y", "top": "688_4_transposed", "type": "transpose", "name": "733_transpose_y" }, { "bottom": "698_4_transposed,688_4_transposed", "weights": {}, "debug_info": "733_batch_matmul", "top": "733_pre_transpose", "type": "batch_matmul", "name": "733_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "733_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "733", "top": "733", "type": "transpose", "name": "733" }, { "bottom": "733", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "735", "top": "735", "type": "elementwise", "name": "735", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "698_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "737_transpose_x", "top": "698_5_transposed", "type": "transpose", "name": "737_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "688_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "737_transpose_y", "top": "688_5_transposed", "type": "transpose", "name": "737_transpose_y" }, { "bottom": "698_5_transposed,688_5_transposed", "weights": {}, "debug_info": "737_batch_matmul", "top": "737_pre_transpose", "type": "batch_matmul", "name": "737_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "737_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "737", "top": "737", "type": "transpose", "name": "737" }, { "bottom": "737", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "739", "top": "739", "type": "elementwise", "name": "739", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "698_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "741_transpose_x", "top": "698_6_transposed", "type": "transpose", "name": "741_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "688_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "741_transpose_y", "top": "688_6_transposed", "type": "transpose", "name": "741_transpose_y" }, { "bottom": "698_6_transposed,688_6_transposed", "weights": {}, "debug_info": "741_batch_matmul", "top": "741_pre_transpose", "type": "batch_matmul", "name": "741_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "741_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "741", "top": "741", "type": "transpose", "name": "741" }, { "bottom": "741", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "743", "top": "743", "type": "elementwise", "name": "743", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "698_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "745_transpose_x", "top": "698_7_transposed", "type": "transpose", "name": "745_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "688_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "745_transpose_y", "top": "688_7_transposed", "type": "transpose", "name": "745_transpose_y" }, { "bottom": "698_7_transposed,688_7_transposed", "weights": {}, "debug_info": "745_batch_matmul", "top": "745_pre_transpose", "type": "batch_matmul", "name": "745_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "745_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "745", "top": "745", "type": "transpose", "name": "745" }, { "bottom": "745", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "747", "top": "747", "type": "elementwise", "name": "747", "beta": 0 }, { "bottom": "719,723,727,731,735,739,743,747", "weights": {}, "nd_mode": true, "axis": 2, "debug_info": "attn_weights.6", "top": "attn_weights.6", "type": "general_concat", "name": "attn_weights.6" }, { "bottom": "attn_weights.6,qk_mask.1", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "attn_weights0.6", "nd_mode": true, "top": "attn_weights0.6", "type": "elementwise", "name": "attn_weights0.6", "beta": 0 }, { "bottom": "attn_weights0.6", "weights": {}, "debug_info": "input.37", "top": "input.37", "C": 2, "type": "softmax", "name": "input.37" }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 2, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "input.37", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "753", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "753", "begin_48": 0, "top": "753_0,753_1,753_2,753_3,753_4,753_5,753_6,753_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "707_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "763_transpose_x", "top": "707_0_transposed", "type": "transpose", "name": "763_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "753_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "763_transpose_y", "top": "753_0_transposed", "type": "transpose", "name": "763_transpose_y" }, { "bottom": "707_0_transposed,753_0_transposed", "weights": {}, "debug_info": "763_batch_matmul", "top": "763_pre_transpose", "type": "batch_matmul", "name": "763_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "763_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "763", "top": "763", "type": "transpose", "name": "763" }, { "axis_h": 2, "axis_w": 0, "bottom": "707_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "765_transpose_x", "top": "707_1_transposed", "type": "transpose", "name": "765_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "753_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "765_transpose_y", "top": "753_1_transposed", "type": "transpose", "name": "765_transpose_y" }, { "bottom": "707_1_transposed,753_1_transposed", "weights": {}, "debug_info": "765_batch_matmul", "top": "765_pre_transpose", "type": "batch_matmul", "name": "765_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "765_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "765", "top": "765", "type": "transpose", "name": "765" }, { "axis_h": 2, "axis_w": 0, "bottom": "707_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "767_transpose_x", "top": "707_2_transposed", "type": "transpose", "name": "767_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "753_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "767_transpose_y", "top": "753_2_transposed", "type": "transpose", "name": "767_transpose_y" }, { "bottom": "707_2_transposed,753_2_transposed", "weights": {}, "debug_info": "767_batch_matmul", "top": "767_pre_transpose", "type": "batch_matmul", "name": "767_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "767_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "767", "top": "767", "type": "transpose", "name": "767" }, { "axis_h": 2, "axis_w": 0, "bottom": "707_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "769_transpose_x", "top": "707_3_transposed", "type": "transpose", "name": "769_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "753_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "769_transpose_y", "top": "753_3_transposed", "type": "transpose", "name": "769_transpose_y" }, { "bottom": "707_3_transposed,753_3_transposed", "weights": {}, "debug_info": "769_batch_matmul", "top": "769_pre_transpose", "type": "batch_matmul", "name": "769_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "769_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "769", "top": "769", "type": "transpose", "name": "769" }, { "axis_h": 2, "axis_w": 0, "bottom": "707_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "771_transpose_x", "top": "707_4_transposed", "type": "transpose", "name": "771_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "753_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "771_transpose_y", "top": "753_4_transposed", "type": "transpose", "name": "771_transpose_y" }, { "bottom": "707_4_transposed,753_4_transposed", "weights": {}, "debug_info": "771_batch_matmul", "top": "771_pre_transpose", "type": "batch_matmul", "name": "771_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "771_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "771", "top": "771", "type": "transpose", "name": "771" }, { "axis_h": 2, "axis_w": 0, "bottom": "707_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "773_transpose_x", "top": "707_5_transposed", "type": "transpose", "name": "773_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "753_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "773_transpose_y", "top": "753_5_transposed", "type": "transpose", "name": "773_transpose_y" }, { "bottom": "707_5_transposed,753_5_transposed", "weights": {}, "debug_info": "773_batch_matmul", "top": "773_pre_transpose", "type": "batch_matmul", "name": "773_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "773_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "773", "top": "773", "type": "transpose", "name": "773" }, { "axis_h": 2, "axis_w": 0, "bottom": "707_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "775_transpose_x", "top": "707_6_transposed", "type": "transpose", "name": "775_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "753_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "775_transpose_y", "top": "753_6_transposed", "type": "transpose", "name": "775_transpose_y" }, { "bottom": "707_6_transposed,753_6_transposed", "weights": {}, "debug_info": "775_batch_matmul", "top": "775_pre_transpose", "type": "batch_matmul", "name": "775_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "775_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "775", "top": "775", "type": "transpose", "name": "775" }, { "axis_h": 2, "axis_w": 0, "bottom": "707_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "777_transpose_x", "top": "707_7_transposed", "type": "transpose", "name": "777_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "753_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "777_transpose_y", "top": "753_7_transposed", "type": "transpose", "name": "777_transpose_y" }, { "bottom": "707_7_transposed,753_7_transposed", "weights": {}, "debug_info": "777_batch_matmul", "top": "777_pre_transpose", "type": "batch_matmul", "name": "777_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "777_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "777", "top": "777", "type": "transpose", "name": "777" }, { "weights": {}, "debug_info": "attn.17", "top": "attn.17", "type": "concat", "name": "attn.17", "bottom": "763,765,767,769,771,773,775,777" }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "inputs.13", "K": 512, "blob_biases": 127, "name": "inputs.13", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "attn.17", "weights": { "per_ch_qscale": 129, "W_S8": 125, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "axis_mode": 4, "bottom": "inputs.13", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.21", "use_version": 1, "top": "channels_mean.21", "type": "reduce", "name": "channels_mean.21" }, { "alpha": -1, "bottom": "channels_mean.21", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.21", "top": "_neg_y_zero_mean.21", "type": "activation", "name": "_neg_y_zero_mean.21", "beta": 0 }, { "bottom": "inputs.13,_neg_y_zero_mean.21", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.21", "top": "zero_mean.21", "type": "elementwise", "name": "zero_mean.21", "beta": 0 }, { "bottom": "zero_mean.21,zero_mean.21", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.21", "top": "zero_mean_sq.21", "type": "elementwise", "name": "zero_mean_sq.21", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.21", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "800", "use_version": 1, "top": "800", "type": "reduce", "name": "800" }, { "bottom": "800", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "802", "top": "802", "type": "elementwise", "name": "802", "beta": 0 }, { "bottom": "802", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.21", "top": "denom.21", "type": "elementwise", "name": "denom.21", "beta": 0 }, { "bottom": "zero_mean.21,denom.21", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.21", "top": "out.21", "type": "elementwise", "name": "out.21", "beta": 0 }, { "top": "806", "w": 1, "h": 1, "name": "806", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "806", "n": 1, "weights": {}, "constant_blob": 595 }, { "bottom": "out.21,806", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "807", "top": "807", "type": "elementwise", "name": "807", "beta": 0 }, { "top": "809", "w": 1, "h": 1, "name": "809", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "809", "n": 1, "weights": {}, "constant_blob": 417 }, { "bottom": "807,809", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.39", "top": "input.39", "type": "elementwise", "name": "input.39", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.26", "K": 512, "blob_biases": 421, "name": "x.26", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 64, "bottom": "input.39", "weights": { "per_ch_qscale": 423, "W_S8": 419, "per_ch_qbias": 579 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.26", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "819", "top": "819", "type": "elementwise", "name": "819", "beta": 0 }, { "bottom": "819", "weights": {}, "mode": 3, "debug_info": "820", "top": "820", "type": "activation", "name": "820" }, { "bottom": "x.26,820", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.41", "top": "input.41", "type": "elementwise", "name": "input.41", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.28", "K": 64, "blob_biases": 427, "name": "x.28", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.41", "weights": { "per_ch_qscale": 429, "W_S8": 425, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.28,inputs.13", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "attn.19", "top": "attn.19", "type": "elementwise", "name": "attn.19", "beta": 0 }, { "bottom": "inputs0.2,attn.19", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "inputs0.6", "top": "inputs0.6", "type": "elementwise", "name": "inputs0.6", "beta": 0 }, { "axis_mode": 4, "bottom": "inputs0.6", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.23", "use_version": 1, "top": "channels_mean.23", "type": "reduce", "name": "channels_mean.23" }, { "alpha": -1, "bottom": "channels_mean.23", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.23", "top": "_neg_y_zero_mean.23", "type": "activation", "name": "_neg_y_zero_mean.23", "beta": 0 }, { "bottom": "inputs0.6,_neg_y_zero_mean.23", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.23", "top": "zero_mean.23", "type": "elementwise", "name": "zero_mean.23", "beta": 0 }, { "bottom": "zero_mean.23,zero_mean.23", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.23", "top": "zero_mean_sq.23", "type": "elementwise", "name": "zero_mean_sq.23", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.23", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "838", "use_version": 1, "top": "838", "type": "reduce", "name": "838" }, { "bottom": "838", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "840", "top": "840", "type": "elementwise", "name": "840", "beta": 0 }, { "bottom": "840", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.23", "top": "denom.23", "type": "elementwise", "name": "denom.23", "beta": 0 }, { "bottom": "zero_mean.23,denom.23", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.23", "top": "out.23", "type": "elementwise", "name": "out.23", "beta": 0 }, { "top": "844", "w": 1, "h": 1, "name": "844", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "844", "n": 1, "weights": {}, "constant_blob": 597 }, { "bottom": "out.23,844", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "845", "top": "845", "type": "elementwise", "name": "845", "beta": 0 }, { "top": "847", "w": 1, "h": 1, "name": "847", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "847", "n": 1, "weights": {}, "constant_blob": 433 }, { "bottom": "845,847", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.43", "top": "input.43", "type": "elementwise", "name": "input.43", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.30", "K": 512, "blob_biases": 137, "name": "x.30", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 2048, "bottom": "input.43", "weights": { "per_ch_qscale": 139, "W_S8": 135, "per_ch_qbias": 299 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.30", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "859", "top": "859", "type": "elementwise", "name": "859", "beta": 0 }, { "bottom": "859", "weights": {}, "mode": 3, "debug_info": "860", "top": "860", "type": "activation", "name": "860" }, { "bottom": "x.30,860", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.45", "top": "input.45", "type": "elementwise", "name": "input.45", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "input0.11", "K": 2048, "blob_biases": 143, "name": "input0.11", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.45", "weights": { "per_ch_qscale": 145, "W_S8": 141, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "axis_mode": 4, "bottom": "input0.11", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.25", "use_version": 1, "top": "channels_mean.25", "type": "reduce", "name": "channels_mean.25" }, { "alpha": -1, "bottom": "channels_mean.25", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.25", "top": "_neg_y_zero_mean.25", "type": "activation", "name": "_neg_y_zero_mean.25", "beta": 0 }, { "bottom": "input0.11,_neg_y_zero_mean.25", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.25", "top": "zero_mean.25", "type": "elementwise", "name": "zero_mean.25", "beta": 0 }, { "bottom": "zero_mean.25,zero_mean.25", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.25", "top": "zero_mean_sq.25", "type": "elementwise", "name": "zero_mean_sq.25", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.25", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "880", "use_version": 1, "top": "880", "type": "reduce", "name": "880" }, { "bottom": "880", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "882", "top": "882", "type": "elementwise", "name": "882", "beta": 0 }, { "bottom": "882", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.25", "top": "denom.25", "type": "elementwise", "name": "denom.25", "beta": 0 }, { "bottom": "zero_mean.25,denom.25", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.25", "top": "out.25", "type": "elementwise", "name": "out.25", "beta": 0 }, { "top": "886", "w": 1, "h": 1, "name": "886", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "886", "n": 1, "weights": {}, "constant_blob": 599 }, { "bottom": "out.25,886", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "887", "top": "887", "type": "elementwise", "name": "887", "beta": 0 }, { "top": "889", "w": 1, "h": 1, "name": "889", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "889", "n": 1, "weights": {}, "constant_blob": 437 }, { "bottom": "887,889", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.47", "top": "input.47", "type": "elementwise", "name": "input.47", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.32", "K": 512, "blob_biases": 441, "name": "x.32", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 64, "bottom": "input.47", "weights": { "per_ch_qscale": 443, "W_S8": 439, "per_ch_qbias": 579 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.32", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "899", "top": "899", "type": "elementwise", "name": "899", "beta": 0 }, { "bottom": "899", "weights": {}, "mode": 3, "debug_info": "900", "top": "900", "type": "activation", "name": "900" }, { "bottom": "x.32,900", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.49", "top": "input.49", "type": "elementwise", "name": "input.49", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.34", "K": 64, "blob_biases": 447, "name": "x.34", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.49", "weights": { "per_ch_qscale": 449, "W_S8": 445, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.34,input0.11", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "f.6", "top": "f.6", "type": "elementwise", "name": "f.6", "beta": 0 }, { "bottom": "f.6,inputs0.6", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "x1.6", "top": "x1.6", "type": "elementwise", "name": "x1.6", "beta": 0 }, { "bottom": "inputs0.2", "alpha": 0, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "912", "top": "912", "type": "elementwise", "name": "912", "beta": 0 }, { "bottom": "912,x1.6", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "inputs1.1", "top": "inputs1.1", "type": "elementwise", "name": "inputs1.1", "beta": 0 }, { "axis_mode": 4, "bottom": "inputs1.1", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.27", "use_version": 1, "top": "channels_mean.27", "type": "reduce", "name": "channels_mean.27" }, { "alpha": -1, "bottom": "channels_mean.27", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.27", "top": "_neg_y_zero_mean.27", "type": "activation", "name": "_neg_y_zero_mean.27", "beta": 0 }, { "bottom": "inputs1.1,_neg_y_zero_mean.27", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.27", "top": "zero_mean.27", "type": "elementwise", "name": "zero_mean.27", "beta": 0 }, { "bottom": "zero_mean.27,zero_mean.27", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.27", "top": "zero_mean_sq.27", "type": "elementwise", "name": "zero_mean_sq.27", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.27", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "929", "use_version": 1, "top": "929", "type": "reduce", "name": "929" }, { "bottom": "929", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "931", "top": "931", "type": "elementwise", "name": "931", "beta": 0 }, { "bottom": "931", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.27", "top": "denom.27", "type": "elementwise", "name": "denom.27", "beta": 0 }, { "bottom": "zero_mean.27,denom.27", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.27", "top": "out.27", "type": "elementwise", "name": "out.27", "beta": 0 }, { "top": "935", "w": 1, "h": 1, "name": "935", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "935", "n": 1, "weights": {}, "constant_blob": 601 }, { "bottom": "out.27,935", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "936", "top": "936", "type": "elementwise", "name": "936", "beta": 0 }, { "top": "938", "w": 1, "h": 1, "name": "938", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "938", "n": 1, "weights": {}, "constant_blob": 453 }, { "bottom": "936,938", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "k.15", "top": "k.15", "type": "elementwise", "name": "k.15", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "953", "K": 512, "blob_biases": 153, "name": "953", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.15", "weights": { "per_ch_qscale": 155, "W_S8": 151, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "k.17", "K": 512, "blob_biases": 159, "name": "k.17", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.15", "weights": { "per_ch_qscale": 161, "W_S8": 157, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "967", "K": 512, "blob_biases": 165, "name": "967", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.15", "weights": { "per_ch_qscale": 167, "W_S8": 163, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 1, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "953", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "968", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "968", "begin_48": 0, "top": "968_0,968_1,968_2,968_3,968_4,968_5,968_6,968_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 1, "axis_w": 2, "bottom": "k.17", "axis_k": 0, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "transpose_3", "top": "transpose_3", "type": "transpose", "name": "transpose_3" }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 3, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "transpose_3", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "978", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "978", "begin_48": 0, "top": "978_0,978_1,978_2,978_3,978_4,978_5,978_6,978_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 1, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "967", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "987", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "987", "begin_48": 0, "top": "987_0,987_1,987_2,987_3,987_4,987_5,987_6,987_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "978_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "997_transpose_x", "top": "978_0_transposed", "type": "transpose", "name": "997_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "968_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "997_transpose_y", "top": "968_0_transposed", "type": "transpose", "name": "997_transpose_y" }, { "bottom": "978_0_transposed,968_0_transposed", "weights": {}, "debug_info": "997_batch_matmul", "top": "997_pre_transpose", "type": "batch_matmul", "name": "997_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "997_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "997", "top": "997", "type": "transpose", "name": "997" }, { "bottom": "997", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "999", "top": "999", "type": "elementwise", "name": "999", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "978_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1001_transpose_x", "top": "978_1_transposed", "type": "transpose", "name": "1001_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "968_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1001_transpose_y", "top": "968_1_transposed", "type": "transpose", "name": "1001_transpose_y" }, { "bottom": "978_1_transposed,968_1_transposed", "weights": {}, "debug_info": "1001_batch_matmul", "top": "1001_pre_transpose", "type": "batch_matmul", "name": "1001_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1001_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1001", "top": "1001", "type": "transpose", "name": "1001" }, { "bottom": "1001", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1003", "top": "1003", "type": "elementwise", "name": "1003", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "978_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1005_transpose_x", "top": "978_2_transposed", "type": "transpose", "name": "1005_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "968_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1005_transpose_y", "top": "968_2_transposed", "type": "transpose", "name": "1005_transpose_y" }, { "bottom": "978_2_transposed,968_2_transposed", "weights": {}, "debug_info": "1005_batch_matmul", "top": "1005_pre_transpose", "type": "batch_matmul", "name": "1005_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1005_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1005", "top": "1005", "type": "transpose", "name": "1005" }, { "bottom": "1005", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1007", "top": "1007", "type": "elementwise", "name": "1007", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "978_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1009_transpose_x", "top": "978_3_transposed", "type": "transpose", "name": "1009_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "968_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1009_transpose_y", "top": "968_3_transposed", "type": "transpose", "name": "1009_transpose_y" }, { "bottom": "978_3_transposed,968_3_transposed", "weights": {}, "debug_info": "1009_batch_matmul", "top": "1009_pre_transpose", "type": "batch_matmul", "name": "1009_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1009_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1009", "top": "1009", "type": "transpose", "name": "1009" }, { "bottom": "1009", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1011", "top": "1011", "type": "elementwise", "name": "1011", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "978_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1013_transpose_x", "top": "978_4_transposed", "type": "transpose", "name": "1013_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "968_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1013_transpose_y", "top": "968_4_transposed", "type": "transpose", "name": "1013_transpose_y" }, { "bottom": "978_4_transposed,968_4_transposed", "weights": {}, "debug_info": "1013_batch_matmul", "top": "1013_pre_transpose", "type": "batch_matmul", "name": "1013_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1013_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1013", "top": "1013", "type": "transpose", "name": "1013" }, { "bottom": "1013", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1015", "top": "1015", "type": "elementwise", "name": "1015", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "978_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1017_transpose_x", "top": "978_5_transposed", "type": "transpose", "name": "1017_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "968_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1017_transpose_y", "top": "968_5_transposed", "type": "transpose", "name": "1017_transpose_y" }, { "bottom": "978_5_transposed,968_5_transposed", "weights": {}, "debug_info": "1017_batch_matmul", "top": "1017_pre_transpose", "type": "batch_matmul", "name": "1017_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1017_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1017", "top": "1017", "type": "transpose", "name": "1017" }, { "bottom": "1017", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1019", "top": "1019", "type": "elementwise", "name": "1019", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "978_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1021_transpose_x", "top": "978_6_transposed", "type": "transpose", "name": "1021_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "968_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1021_transpose_y", "top": "968_6_transposed", "type": "transpose", "name": "1021_transpose_y" }, { "bottom": "978_6_transposed,968_6_transposed", "weights": {}, "debug_info": "1021_batch_matmul", "top": "1021_pre_transpose", "type": "batch_matmul", "name": "1021_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1021_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1021", "top": "1021", "type": "transpose", "name": "1021" }, { "bottom": "1021", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1023", "top": "1023", "type": "elementwise", "name": "1023", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "978_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1025_transpose_x", "top": "978_7_transposed", "type": "transpose", "name": "1025_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "968_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1025_transpose_y", "top": "968_7_transposed", "type": "transpose", "name": "1025_transpose_y" }, { "bottom": "978_7_transposed,968_7_transposed", "weights": {}, "debug_info": "1025_batch_matmul", "top": "1025_pre_transpose", "type": "batch_matmul", "name": "1025_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1025_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1025", "top": "1025", "type": "transpose", "name": "1025" }, { "bottom": "1025", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1027", "top": "1027", "type": "elementwise", "name": "1027", "beta": 0 }, { "bottom": "999,1003,1007,1011,1015,1019,1023,1027", "weights": {}, "nd_mode": true, "axis": 2, "debug_info": "attn_weights.8", "top": "attn_weights.8", "type": "general_concat", "name": "attn_weights.8" }, { "bottom": "attn_weights.8,qk_mask.1", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "attn_weights0.8", "nd_mode": true, "top": "attn_weights0.8", "type": "elementwise", "name": "attn_weights0.8", "beta": 0 }, { "bottom": "attn_weights0.8", "weights": {}, "debug_info": "input.51", "top": "input.51", "C": 2, "type": "softmax", "name": "input.51" }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 2, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "input.51", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "1033", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "1033", "begin_48": 0, "top": "1033_0,1033_1,1033_2,1033_3,1033_4,1033_5,1033_6,1033_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "987_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1043_transpose_x", "top": "987_0_transposed", "type": "transpose", "name": "1043_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1033_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1043_transpose_y", "top": "1033_0_transposed", "type": "transpose", "name": "1043_transpose_y" }, { "bottom": "987_0_transposed,1033_0_transposed", "weights": {}, "debug_info": "1043_batch_matmul", "top": "1043_pre_transpose", "type": "batch_matmul", "name": "1043_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1043_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1043", "top": "1043", "type": "transpose", "name": "1043" }, { "axis_h": 2, "axis_w": 0, "bottom": "987_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1045_transpose_x", "top": "987_1_transposed", "type": "transpose", "name": "1045_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1033_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1045_transpose_y", "top": "1033_1_transposed", "type": "transpose", "name": "1045_transpose_y" }, { "bottom": "987_1_transposed,1033_1_transposed", "weights": {}, "debug_info": "1045_batch_matmul", "top": "1045_pre_transpose", "type": "batch_matmul", "name": "1045_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1045_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1045", "top": "1045", "type": "transpose", "name": "1045" }, { "axis_h": 2, "axis_w": 0, "bottom": "987_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1047_transpose_x", "top": "987_2_transposed", "type": "transpose", "name": "1047_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1033_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1047_transpose_y", "top": "1033_2_transposed", "type": "transpose", "name": "1047_transpose_y" }, { "bottom": "987_2_transposed,1033_2_transposed", "weights": {}, "debug_info": "1047_batch_matmul", "top": "1047_pre_transpose", "type": "batch_matmul", "name": "1047_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1047_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1047", "top": "1047", "type": "transpose", "name": "1047" }, { "axis_h": 2, "axis_w": 0, "bottom": "987_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1049_transpose_x", "top": "987_3_transposed", "type": "transpose", "name": "1049_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1033_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1049_transpose_y", "top": "1033_3_transposed", "type": "transpose", "name": "1049_transpose_y" }, { "bottom": "987_3_transposed,1033_3_transposed", "weights": {}, "debug_info": "1049_batch_matmul", "top": "1049_pre_transpose", "type": "batch_matmul", "name": "1049_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1049_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1049", "top": "1049", "type": "transpose", "name": "1049" }, { "axis_h": 2, "axis_w": 0, "bottom": "987_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1051_transpose_x", "top": "987_4_transposed", "type": "transpose", "name": "1051_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1033_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1051_transpose_y", "top": "1033_4_transposed", "type": "transpose", "name": "1051_transpose_y" }, { "bottom": "987_4_transposed,1033_4_transposed", "weights": {}, "debug_info": "1051_batch_matmul", "top": "1051_pre_transpose", "type": "batch_matmul", "name": "1051_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1051_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1051", "top": "1051", "type": "transpose", "name": "1051" }, { "axis_h": 2, "axis_w": 0, "bottom": "987_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1053_transpose_x", "top": "987_5_transposed", "type": "transpose", "name": "1053_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1033_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1053_transpose_y", "top": "1033_5_transposed", "type": "transpose", "name": "1053_transpose_y" }, { "bottom": "987_5_transposed,1033_5_transposed", "weights": {}, "debug_info": "1053_batch_matmul", "top": "1053_pre_transpose", "type": "batch_matmul", "name": "1053_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1053_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1053", "top": "1053", "type": "transpose", "name": "1053" }, { "axis_h": 2, "axis_w": 0, "bottom": "987_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1055_transpose_x", "top": "987_6_transposed", "type": "transpose", "name": "1055_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1033_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1055_transpose_y", "top": "1033_6_transposed", "type": "transpose", "name": "1055_transpose_y" }, { "bottom": "987_6_transposed,1033_6_transposed", "weights": {}, "debug_info": "1055_batch_matmul", "top": "1055_pre_transpose", "type": "batch_matmul", "name": "1055_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1055_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1055", "top": "1055", "type": "transpose", "name": "1055" }, { "axis_h": 2, "axis_w": 0, "bottom": "987_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1057_transpose_x", "top": "987_7_transposed", "type": "transpose", "name": "1057_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1033_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1057_transpose_y", "top": "1033_7_transposed", "type": "transpose", "name": "1057_transpose_y" }, { "bottom": "987_7_transposed,1033_7_transposed", "weights": {}, "debug_info": "1057_batch_matmul", "top": "1057_pre_transpose", "type": "batch_matmul", "name": "1057_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1057_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1057", "top": "1057", "type": "transpose", "name": "1057" }, { "weights": {}, "debug_info": "attn.23", "top": "attn.23", "type": "concat", "name": "attn.23", "bottom": "1043,1045,1047,1049,1051,1053,1055,1057" }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "inputs.17", "K": 512, "blob_biases": 171, "name": "inputs.17", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "attn.23", "weights": { "per_ch_qscale": 173, "W_S8": 169, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "axis_mode": 4, "bottom": "inputs.17", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.29", "use_version": 1, "top": "channels_mean.29", "type": "reduce", "name": "channels_mean.29" }, { "alpha": -1, "bottom": "channels_mean.29", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.29", "top": "_neg_y_zero_mean.29", "type": "activation", "name": "_neg_y_zero_mean.29", "beta": 0 }, { "bottom": "inputs.17,_neg_y_zero_mean.29", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.29", "top": "zero_mean.29", "type": "elementwise", "name": "zero_mean.29", "beta": 0 }, { "bottom": "zero_mean.29,zero_mean.29", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.29", "top": "zero_mean_sq.29", "type": "elementwise", "name": "zero_mean_sq.29", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.29", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "1080", "use_version": 1, "top": "1080", "type": "reduce", "name": "1080" }, { "bottom": "1080", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1082", "top": "1082", "type": "elementwise", "name": "1082", "beta": 0 }, { "bottom": "1082", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.29", "top": "denom.29", "type": "elementwise", "name": "denom.29", "beta": 0 }, { "bottom": "zero_mean.29,denom.29", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.29", "top": "out.29", "type": "elementwise", "name": "out.29", "beta": 0 }, { "top": "1086", "w": 1, "h": 1, "name": "1086", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1086", "n": 1, "weights": {}, "constant_blob": 603 }, { "bottom": "out.29,1086", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1087", "top": "1087", "type": "elementwise", "name": "1087", "beta": 0 }, { "top": "1089", "w": 1, "h": 1, "name": "1089", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1089", "n": 1, "weights": {}, "constant_blob": 457 }, { "bottom": "1087,1089", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.53", "top": "input.53", "type": "elementwise", "name": "input.53", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.36", "K": 512, "blob_biases": 461, "name": "x.36", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 64, "bottom": "input.53", "weights": { "per_ch_qscale": 463, "W_S8": 459, "per_ch_qbias": 579 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.36", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1099", "top": "1099", "type": "elementwise", "name": "1099", "beta": 0 }, { "bottom": "1099", "weights": {}, "mode": 3, "debug_info": "1100", "top": "1100", "type": "activation", "name": "1100" }, { "bottom": "x.36,1100", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.55", "top": "input.55", "type": "elementwise", "name": "input.55", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.38", "K": 64, "blob_biases": 467, "name": "x.38", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.55", "weights": { "per_ch_qscale": 469, "W_S8": 465, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.38,inputs.17", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "attn.25", "top": "attn.25", "type": "elementwise", "name": "attn.25", "beta": 0 }, { "bottom": "inputs1.1,attn.25", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "inputs0.8", "top": "inputs0.8", "type": "elementwise", "name": "inputs0.8", "beta": 0 }, { "axis_mode": 4, "bottom": "inputs0.8", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.31", "use_version": 1, "top": "channels_mean.31", "type": "reduce", "name": "channels_mean.31" }, { "alpha": -1, "bottom": "channels_mean.31", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.31", "top": "_neg_y_zero_mean.31", "type": "activation", "name": "_neg_y_zero_mean.31", "beta": 0 }, { "bottom": "inputs0.8,_neg_y_zero_mean.31", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.31", "top": "zero_mean.31", "type": "elementwise", "name": "zero_mean.31", "beta": 0 }, { "bottom": "zero_mean.31,zero_mean.31", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.31", "top": "zero_mean_sq.31", "type": "elementwise", "name": "zero_mean_sq.31", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.31", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "1118", "use_version": 1, "top": "1118", "type": "reduce", "name": "1118" }, { "bottom": "1118", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1120", "top": "1120", "type": "elementwise", "name": "1120", "beta": 0 }, { "bottom": "1120", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.31", "top": "denom.31", "type": "elementwise", "name": "denom.31", "beta": 0 }, { "bottom": "zero_mean.31,denom.31", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.31", "top": "out.31", "type": "elementwise", "name": "out.31", "beta": 0 }, { "top": "1124", "w": 1, "h": 1, "name": "1124", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1124", "n": 1, "weights": {}, "constant_blob": 605 }, { "bottom": "out.31,1124", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1125", "top": "1125", "type": "elementwise", "name": "1125", "beta": 0 }, { "top": "1127", "w": 1, "h": 1, "name": "1127", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1127", "n": 1, "weights": {}, "constant_blob": 473 }, { "bottom": "1125,1127", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.57", "top": "input.57", "type": "elementwise", "name": "input.57", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.40", "K": 512, "blob_biases": 181, "name": "x.40", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 2048, "bottom": "input.57", "weights": { "per_ch_qscale": 183, "W_S8": 179, "per_ch_qbias": 299 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.40", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1139", "top": "1139", "type": "elementwise", "name": "1139", "beta": 0 }, { "bottom": "1139", "weights": {}, "mode": 3, "debug_info": "1140", "top": "1140", "type": "activation", "name": "1140" }, { "bottom": "x.40,1140", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.59", "top": "input.59", "type": "elementwise", "name": "input.59", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "input0.15", "K": 2048, "blob_biases": 187, "name": "input0.15", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.59", "weights": { "per_ch_qscale": 189, "W_S8": 185, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "axis_mode": 4, "bottom": "input0.15", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.33", "use_version": 1, "top": "channels_mean.33", "type": "reduce", "name": "channels_mean.33" }, { "alpha": -1, "bottom": "channels_mean.33", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.33", "top": "_neg_y_zero_mean.33", "type": "activation", "name": "_neg_y_zero_mean.33", "beta": 0 }, { "bottom": "input0.15,_neg_y_zero_mean.33", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.33", "top": "zero_mean.33", "type": "elementwise", "name": "zero_mean.33", "beta": 0 }, { "bottom": "zero_mean.33,zero_mean.33", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.33", "top": "zero_mean_sq.33", "type": "elementwise", "name": "zero_mean_sq.33", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.33", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "1160", "use_version": 1, "top": "1160", "type": "reduce", "name": "1160" }, { "bottom": "1160", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1162", "top": "1162", "type": "elementwise", "name": "1162", "beta": 0 }, { "bottom": "1162", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.33", "top": "denom.33", "type": "elementwise", "name": "denom.33", "beta": 0 }, { "bottom": "zero_mean.33,denom.33", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.33", "top": "out.33", "type": "elementwise", "name": "out.33", "beta": 0 }, { "top": "1166", "w": 1, "h": 1, "name": "1166", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1166", "n": 1, "weights": {}, "constant_blob": 607 }, { "bottom": "out.33,1166", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1167", "top": "1167", "type": "elementwise", "name": "1167", "beta": 0 }, { "top": "1169", "w": 1, "h": 1, "name": "1169", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1169", "n": 1, "weights": {}, "constant_blob": 477 }, { "bottom": "1167,1169", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.61", "top": "input.61", "type": "elementwise", "name": "input.61", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.42", "K": 512, "blob_biases": 481, "name": "x.42", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 64, "bottom": "input.61", "weights": { "per_ch_qscale": 483, "W_S8": 479, "per_ch_qbias": 579 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.42", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1179", "top": "1179", "type": "elementwise", "name": "1179", "beta": 0 }, { "bottom": "1179", "weights": {}, "mode": 3, "debug_info": "1180", "top": "1180", "type": "activation", "name": "1180" }, { "bottom": "x.42,1180", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.63", "top": "input.63", "type": "elementwise", "name": "input.63", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.44", "K": 64, "blob_biases": 487, "name": "x.44", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.63", "weights": { "per_ch_qscale": 489, "W_S8": 485, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.44,input0.15", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "f.8", "top": "f.8", "type": "elementwise", "name": "f.8", "beta": 0 }, { "bottom": "f.8,inputs0.8", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "x1.8", "top": "x1.8", "type": "elementwise", "name": "x1.8", "beta": 0 }, { "bottom": "inputs1.1", "alpha": 0, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1192", "top": "1192", "type": "elementwise", "name": "1192", "beta": 0 }, { "bottom": "1192,x1.8", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "inputs2.1", "top": "inputs2.1", "type": "elementwise", "name": "inputs2.1", "beta": 0 }, { "axis_mode": 4, "bottom": "inputs2.1", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.35", "use_version": 1, "top": "channels_mean.35", "type": "reduce", "name": "channels_mean.35" }, { "alpha": -1, "bottom": "channels_mean.35", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.35", "top": "_neg_y_zero_mean.35", "type": "activation", "name": "_neg_y_zero_mean.35", "beta": 0 }, { "bottom": "inputs2.1,_neg_y_zero_mean.35", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.35", "top": "zero_mean.35", "type": "elementwise", "name": "zero_mean.35", "beta": 0 }, { "bottom": "zero_mean.35,zero_mean.35", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.35", "top": "zero_mean_sq.35", "type": "elementwise", "name": "zero_mean_sq.35", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.35", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "1209", "use_version": 1, "top": "1209", "type": "reduce", "name": "1209" }, { "bottom": "1209", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1211", "top": "1211", "type": "elementwise", "name": "1211", "beta": 0 }, { "bottom": "1211", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.35", "top": "denom.35", "type": "elementwise", "name": "denom.35", "beta": 0 }, { "bottom": "zero_mean.35,denom.35", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.35", "top": "out.35", "type": "elementwise", "name": "out.35", "beta": 0 }, { "top": "1215", "w": 1, "h": 1, "name": "1215", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1215", "n": 1, "weights": {}, "constant_blob": 609 }, { "bottom": "out.35,1215", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1216", "top": "1216", "type": "elementwise", "name": "1216", "beta": 0 }, { "top": "1218", "w": 1, "h": 1, "name": "1218", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1218", "n": 1, "weights": {}, "constant_blob": 493 }, { "bottom": "1216,1218", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "k.19", "top": "k.19", "type": "elementwise", "name": "k.19", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "1233", "K": 512, "blob_biases": 197, "name": "1233", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.19", "weights": { "per_ch_qscale": 199, "W_S8": 195, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "k.21", "K": 512, "blob_biases": 203, "name": "k.21", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.19", "weights": { "per_ch_qscale": 205, "W_S8": 201, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "1247", "K": 512, "blob_biases": 209, "name": "1247", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.19", "weights": { "per_ch_qscale": 211, "W_S8": 207, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 1, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "1233", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "1248", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "1248", "begin_48": 0, "top": "1248_0,1248_1,1248_2,1248_3,1248_4,1248_5,1248_6,1248_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 1, "axis_w": 2, "bottom": "k.21", "axis_k": 0, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "transpose_2", "top": "transpose_2", "type": "transpose", "name": "transpose_2" }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 3, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "transpose_2", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "1258", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "1258", "begin_48": 0, "top": "1258_0,1258_1,1258_2,1258_3,1258_4,1258_5,1258_6,1258_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 1, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "1247", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "1267", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "1267", "begin_48": 0, "top": "1267_0,1267_1,1267_2,1267_3,1267_4,1267_5,1267_6,1267_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1258_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1277_transpose_x", "top": "1258_0_transposed", "type": "transpose", "name": "1277_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1248_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1277_transpose_y", "top": "1248_0_transposed", "type": "transpose", "name": "1277_transpose_y" }, { "bottom": "1258_0_transposed,1248_0_transposed", "weights": {}, "debug_info": "1277_batch_matmul", "top": "1277_pre_transpose", "type": "batch_matmul", "name": "1277_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1277_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1277", "top": "1277", "type": "transpose", "name": "1277" }, { "bottom": "1277", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1279", "top": "1279", "type": "elementwise", "name": "1279", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1258_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1281_transpose_x", "top": "1258_1_transposed", "type": "transpose", "name": "1281_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1248_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1281_transpose_y", "top": "1248_1_transposed", "type": "transpose", "name": "1281_transpose_y" }, { "bottom": "1258_1_transposed,1248_1_transposed", "weights": {}, "debug_info": "1281_batch_matmul", "top": "1281_pre_transpose", "type": "batch_matmul", "name": "1281_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1281_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1281", "top": "1281", "type": "transpose", "name": "1281" }, { "bottom": "1281", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1283", "top": "1283", "type": "elementwise", "name": "1283", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1258_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1285_transpose_x", "top": "1258_2_transposed", "type": "transpose", "name": "1285_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1248_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1285_transpose_y", "top": "1248_2_transposed", "type": "transpose", "name": "1285_transpose_y" }, { "bottom": "1258_2_transposed,1248_2_transposed", "weights": {}, "debug_info": "1285_batch_matmul", "top": "1285_pre_transpose", "type": "batch_matmul", "name": "1285_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1285_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1285", "top": "1285", "type": "transpose", "name": "1285" }, { "bottom": "1285", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1287", "top": "1287", "type": "elementwise", "name": "1287", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1258_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1289_transpose_x", "top": "1258_3_transposed", "type": "transpose", "name": "1289_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1248_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1289_transpose_y", "top": "1248_3_transposed", "type": "transpose", "name": "1289_transpose_y" }, { "bottom": "1258_3_transposed,1248_3_transposed", "weights": {}, "debug_info": "1289_batch_matmul", "top": "1289_pre_transpose", "type": "batch_matmul", "name": "1289_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1289_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1289", "top": "1289", "type": "transpose", "name": "1289" }, { "bottom": "1289", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1291", "top": "1291", "type": "elementwise", "name": "1291", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1258_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1293_transpose_x", "top": "1258_4_transposed", "type": "transpose", "name": "1293_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1248_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1293_transpose_y", "top": "1248_4_transposed", "type": "transpose", "name": "1293_transpose_y" }, { "bottom": "1258_4_transposed,1248_4_transposed", "weights": {}, "debug_info": "1293_batch_matmul", "top": "1293_pre_transpose", "type": "batch_matmul", "name": "1293_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1293_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1293", "top": "1293", "type": "transpose", "name": "1293" }, { "bottom": "1293", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1295", "top": "1295", "type": "elementwise", "name": "1295", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1258_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1297_transpose_x", "top": "1258_5_transposed", "type": "transpose", "name": "1297_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1248_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1297_transpose_y", "top": "1248_5_transposed", "type": "transpose", "name": "1297_transpose_y" }, { "bottom": "1258_5_transposed,1248_5_transposed", "weights": {}, "debug_info": "1297_batch_matmul", "top": "1297_pre_transpose", "type": "batch_matmul", "name": "1297_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1297_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1297", "top": "1297", "type": "transpose", "name": "1297" }, { "bottom": "1297", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1299", "top": "1299", "type": "elementwise", "name": "1299", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1258_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1301_transpose_x", "top": "1258_6_transposed", "type": "transpose", "name": "1301_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1248_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1301_transpose_y", "top": "1248_6_transposed", "type": "transpose", "name": "1301_transpose_y" }, { "bottom": "1258_6_transposed,1248_6_transposed", "weights": {}, "debug_info": "1301_batch_matmul", "top": "1301_pre_transpose", "type": "batch_matmul", "name": "1301_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1301_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1301", "top": "1301", "type": "transpose", "name": "1301" }, { "bottom": "1301", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1303", "top": "1303", "type": "elementwise", "name": "1303", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1258_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1305_transpose_x", "top": "1258_7_transposed", "type": "transpose", "name": "1305_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1248_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1305_transpose_y", "top": "1248_7_transposed", "type": "transpose", "name": "1305_transpose_y" }, { "bottom": "1258_7_transposed,1248_7_transposed", "weights": {}, "debug_info": "1305_batch_matmul", "top": "1305_pre_transpose", "type": "batch_matmul", "name": "1305_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1305_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1305", "top": "1305", "type": "transpose", "name": "1305" }, { "bottom": "1305", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1307", "top": "1307", "type": "elementwise", "name": "1307", "beta": 0 }, { "bottom": "1279,1283,1287,1291,1295,1299,1303,1307", "weights": {}, "nd_mode": true, "axis": 2, "debug_info": "attn_weights.10", "top": "attn_weights.10", "type": "general_concat", "name": "attn_weights.10" }, { "bottom": "attn_weights.10,qk_mask.1", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "attn_weights0.10", "nd_mode": true, "top": "attn_weights0.10", "type": "elementwise", "name": "attn_weights0.10", "beta": 0 }, { "bottom": "attn_weights0.10", "weights": {}, "debug_info": "input.65", "top": "input.65", "C": 2, "type": "softmax", "name": "input.65" }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 2, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "input.65", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "1313", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "1313", "begin_48": 0, "top": "1313_0,1313_1,1313_2,1313_3,1313_4,1313_5,1313_6,1313_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1267_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1323_transpose_x", "top": "1267_0_transposed", "type": "transpose", "name": "1323_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1313_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1323_transpose_y", "top": "1313_0_transposed", "type": "transpose", "name": "1323_transpose_y" }, { "bottom": "1267_0_transposed,1313_0_transposed", "weights": {}, "debug_info": "1323_batch_matmul", "top": "1323_pre_transpose", "type": "batch_matmul", "name": "1323_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1323_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1323", "top": "1323", "type": "transpose", "name": "1323" }, { "axis_h": 2, "axis_w": 0, "bottom": "1267_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1325_transpose_x", "top": "1267_1_transposed", "type": "transpose", "name": "1325_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1313_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1325_transpose_y", "top": "1313_1_transposed", "type": "transpose", "name": "1325_transpose_y" }, { "bottom": "1267_1_transposed,1313_1_transposed", "weights": {}, "debug_info": "1325_batch_matmul", "top": "1325_pre_transpose", "type": "batch_matmul", "name": "1325_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1325_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1325", "top": "1325", "type": "transpose", "name": "1325" }, { "axis_h": 2, "axis_w": 0, "bottom": "1267_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1327_transpose_x", "top": "1267_2_transposed", "type": "transpose", "name": "1327_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1313_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1327_transpose_y", "top": "1313_2_transposed", "type": "transpose", "name": "1327_transpose_y" }, { "bottom": "1267_2_transposed,1313_2_transposed", "weights": {}, "debug_info": "1327_batch_matmul", "top": "1327_pre_transpose", "type": "batch_matmul", "name": "1327_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1327_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1327", "top": "1327", "type": "transpose", "name": "1327" }, { "axis_h": 2, "axis_w": 0, "bottom": "1267_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1329_transpose_x", "top": "1267_3_transposed", "type": "transpose", "name": "1329_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1313_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1329_transpose_y", "top": "1313_3_transposed", "type": "transpose", "name": "1329_transpose_y" }, { "bottom": "1267_3_transposed,1313_3_transposed", "weights": {}, "debug_info": "1329_batch_matmul", "top": "1329_pre_transpose", "type": "batch_matmul", "name": "1329_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1329_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1329", "top": "1329", "type": "transpose", "name": "1329" }, { "axis_h": 2, "axis_w": 0, "bottom": "1267_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1331_transpose_x", "top": "1267_4_transposed", "type": "transpose", "name": "1331_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1313_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1331_transpose_y", "top": "1313_4_transposed", "type": "transpose", "name": "1331_transpose_y" }, { "bottom": "1267_4_transposed,1313_4_transposed", "weights": {}, "debug_info": "1331_batch_matmul", "top": "1331_pre_transpose", "type": "batch_matmul", "name": "1331_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1331_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1331", "top": "1331", "type": "transpose", "name": "1331" }, { "axis_h": 2, "axis_w": 0, "bottom": "1267_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1333_transpose_x", "top": "1267_5_transposed", "type": "transpose", "name": "1333_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1313_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1333_transpose_y", "top": "1313_5_transposed", "type": "transpose", "name": "1333_transpose_y" }, { "bottom": "1267_5_transposed,1313_5_transposed", "weights": {}, "debug_info": "1333_batch_matmul", "top": "1333_pre_transpose", "type": "batch_matmul", "name": "1333_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1333_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1333", "top": "1333", "type": "transpose", "name": "1333" }, { "axis_h": 2, "axis_w": 0, "bottom": "1267_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1335_transpose_x", "top": "1267_6_transposed", "type": "transpose", "name": "1335_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1313_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1335_transpose_y", "top": "1313_6_transposed", "type": "transpose", "name": "1335_transpose_y" }, { "bottom": "1267_6_transposed,1313_6_transposed", "weights": {}, "debug_info": "1335_batch_matmul", "top": "1335_pre_transpose", "type": "batch_matmul", "name": "1335_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1335_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1335", "top": "1335", "type": "transpose", "name": "1335" }, { "axis_h": 2, "axis_w": 0, "bottom": "1267_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1337_transpose_x", "top": "1267_7_transposed", "type": "transpose", "name": "1337_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1313_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1337_transpose_y", "top": "1313_7_transposed", "type": "transpose", "name": "1337_transpose_y" }, { "bottom": "1267_7_transposed,1313_7_transposed", "weights": {}, "debug_info": "1337_batch_matmul", "top": "1337_pre_transpose", "type": "batch_matmul", "name": "1337_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1337_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1337", "top": "1337", "type": "transpose", "name": "1337" }, { "weights": {}, "debug_info": "attn.29", "top": "attn.29", "type": "concat", "name": "attn.29", "bottom": "1323,1325,1327,1329,1331,1333,1335,1337" }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "inputs.21", "K": 512, "blob_biases": 215, "name": "inputs.21", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "attn.29", "weights": { "per_ch_qscale": 217, "W_S8": 213, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "axis_mode": 4, "bottom": "inputs.21", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.37", "use_version": 1, "top": "channels_mean.37", "type": "reduce", "name": "channels_mean.37" }, { "alpha": -1, "bottom": "channels_mean.37", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.37", "top": "_neg_y_zero_mean.37", "type": "activation", "name": "_neg_y_zero_mean.37", "beta": 0 }, { "bottom": "inputs.21,_neg_y_zero_mean.37", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.37", "top": "zero_mean.37", "type": "elementwise", "name": "zero_mean.37", "beta": 0 }, { "bottom": "zero_mean.37,zero_mean.37", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.37", "top": "zero_mean_sq.37", "type": "elementwise", "name": "zero_mean_sq.37", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.37", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "1360", "use_version": 1, "top": "1360", "type": "reduce", "name": "1360" }, { "bottom": "1360", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1362", "top": "1362", "type": "elementwise", "name": "1362", "beta": 0 }, { "bottom": "1362", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.37", "top": "denom.37", "type": "elementwise", "name": "denom.37", "beta": 0 }, { "bottom": "zero_mean.37,denom.37", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.37", "top": "out.37", "type": "elementwise", "name": "out.37", "beta": 0 }, { "top": "1366", "w": 1, "h": 1, "name": "1366", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1366", "n": 1, "weights": {}, "constant_blob": 611 }, { "bottom": "out.37,1366", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1367", "top": "1367", "type": "elementwise", "name": "1367", "beta": 0 }, { "top": "1369", "w": 1, "h": 1, "name": "1369", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1369", "n": 1, "weights": {}, "constant_blob": 497 }, { "bottom": "1367,1369", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.67", "top": "input.67", "type": "elementwise", "name": "input.67", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.46", "K": 512, "blob_biases": 501, "name": "x.46", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 64, "bottom": "input.67", "weights": { "per_ch_qscale": 503, "W_S8": 499, "per_ch_qbias": 579 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.46", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1379", "top": "1379", "type": "elementwise", "name": "1379", "beta": 0 }, { "bottom": "1379", "weights": {}, "mode": 3, "debug_info": "1380", "top": "1380", "type": "activation", "name": "1380" }, { "bottom": "x.46,1380", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.69", "top": "input.69", "type": "elementwise", "name": "input.69", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.48", "K": 64, "blob_biases": 507, "name": "x.48", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.69", "weights": { "per_ch_qscale": 509, "W_S8": 505, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.48,inputs.21", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "attn.31", "top": "attn.31", "type": "elementwise", "name": "attn.31", "beta": 0 }, { "bottom": "inputs2.1,attn.31", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "inputs0.10", "top": "inputs0.10", "type": "elementwise", "name": "inputs0.10", "beta": 0 }, { "axis_mode": 4, "bottom": "inputs0.10", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.39", "use_version": 1, "top": "channels_mean.39", "type": "reduce", "name": "channels_mean.39" }, { "alpha": -1, "bottom": "channels_mean.39", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.39", "top": "_neg_y_zero_mean.39", "type": "activation", "name": "_neg_y_zero_mean.39", "beta": 0 }, { "bottom": "inputs0.10,_neg_y_zero_mean.39", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.39", "top": "zero_mean.39", "type": "elementwise", "name": "zero_mean.39", "beta": 0 }, { "bottom": "zero_mean.39,zero_mean.39", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.39", "top": "zero_mean_sq.39", "type": "elementwise", "name": "zero_mean_sq.39", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.39", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "1398", "use_version": 1, "top": "1398", "type": "reduce", "name": "1398" }, { "bottom": "1398", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1400", "top": "1400", "type": "elementwise", "name": "1400", "beta": 0 }, { "bottom": "1400", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.39", "top": "denom.39", "type": "elementwise", "name": "denom.39", "beta": 0 }, { "bottom": "zero_mean.39,denom.39", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.39", "top": "out.39", "type": "elementwise", "name": "out.39", "beta": 0 }, { "top": "1404", "w": 1, "h": 1, "name": "1404", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1404", "n": 1, "weights": {}, "constant_blob": 613 }, { "bottom": "out.39,1404", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1405", "top": "1405", "type": "elementwise", "name": "1405", "beta": 0 }, { "top": "1407", "w": 1, "h": 1, "name": "1407", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1407", "n": 1, "weights": {}, "constant_blob": 513 }, { "bottom": "1405,1407", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.71", "top": "input.71", "type": "elementwise", "name": "input.71", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.50", "K": 512, "blob_biases": 225, "name": "x.50", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 2048, "bottom": "input.71", "weights": { "per_ch_qscale": 227, "W_S8": 223, "per_ch_qbias": 299 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.50", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1419", "top": "1419", "type": "elementwise", "name": "1419", "beta": 0 }, { "bottom": "1419", "weights": {}, "mode": 3, "debug_info": "1420", "top": "1420", "type": "activation", "name": "1420" }, { "bottom": "x.50,1420", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.73", "top": "input.73", "type": "elementwise", "name": "input.73", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "input0.19", "K": 2048, "blob_biases": 231, "name": "input0.19", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.73", "weights": { "per_ch_qscale": 233, "W_S8": 229, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "axis_mode": 4, "bottom": "input0.19", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.41", "use_version": 1, "top": "channels_mean.41", "type": "reduce", "name": "channels_mean.41" }, { "alpha": -1, "bottom": "channels_mean.41", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.41", "top": "_neg_y_zero_mean.41", "type": "activation", "name": "_neg_y_zero_mean.41", "beta": 0 }, { "bottom": "input0.19,_neg_y_zero_mean.41", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.41", "top": "zero_mean.41", "type": "elementwise", "name": "zero_mean.41", "beta": 0 }, { "bottom": "zero_mean.41,zero_mean.41", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.41", "top": "zero_mean_sq.41", "type": "elementwise", "name": "zero_mean_sq.41", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.41", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "1440", "use_version": 1, "top": "1440", "type": "reduce", "name": "1440" }, { "bottom": "1440", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1442", "top": "1442", "type": "elementwise", "name": "1442", "beta": 0 }, { "bottom": "1442", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.41", "top": "denom.41", "type": "elementwise", "name": "denom.41", "beta": 0 }, { "bottom": "zero_mean.41,denom.41", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.41", "top": "out.41", "type": "elementwise", "name": "out.41", "beta": 0 }, { "top": "1446", "w": 1, "h": 1, "name": "1446", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1446", "n": 1, "weights": {}, "constant_blob": 615 }, { "bottom": "out.41,1446", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1447", "top": "1447", "type": "elementwise", "name": "1447", "beta": 0 }, { "top": "1449", "w": 1, "h": 1, "name": "1449", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1449", "n": 1, "weights": {}, "constant_blob": 517 }, { "bottom": "1447,1449", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.75", "top": "input.75", "type": "elementwise", "name": "input.75", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.52", "K": 512, "blob_biases": 521, "name": "x.52", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 64, "bottom": "input.75", "weights": { "per_ch_qscale": 523, "W_S8": 519, "per_ch_qbias": 579 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.52", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1459", "top": "1459", "type": "elementwise", "name": "1459", "beta": 0 }, { "bottom": "1459", "weights": {}, "mode": 3, "debug_info": "1460", "top": "1460", "type": "activation", "name": "1460" }, { "bottom": "x.52,1460", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.77", "top": "input.77", "type": "elementwise", "name": "input.77", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.54", "K": 64, "blob_biases": 527, "name": "x.54", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.77", "weights": { "per_ch_qscale": 529, "W_S8": 525, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.54,input0.19", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "f.10", "top": "f.10", "type": "elementwise", "name": "f.10", "beta": 0 }, { "bottom": "f.10,inputs0.10", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "x1.10", "top": "x1.10", "type": "elementwise", "name": "x1.10", "beta": 0 }, { "bottom": "inputs2.1", "alpha": 0, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1472", "top": "1472", "type": "elementwise", "name": "1472", "beta": 0 }, { "bottom": "1472,x1.10", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "inputs3.1", "top": "inputs3.1", "type": "elementwise", "name": "inputs3.1", "beta": 0 }, { "axis_mode": 4, "bottom": "inputs3.1", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.2", "use_version": 1, "top": "channels_mean.2", "type": "reduce", "name": "channels_mean.2" }, { "alpha": -1, "bottom": "channels_mean.2", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.2", "top": "_neg_y_zero_mean.2", "type": "activation", "name": "_neg_y_zero_mean.2", "beta": 0 }, { "bottom": "inputs3.1,_neg_y_zero_mean.2", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.2", "top": "zero_mean.2", "type": "elementwise", "name": "zero_mean.2", "beta": 0 }, { "bottom": "zero_mean.2,zero_mean.2", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.2", "top": "zero_mean_sq.2", "type": "elementwise", "name": "zero_mean_sq.2", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.2", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "1489", "use_version": 1, "top": "1489", "type": "reduce", "name": "1489" }, { "bottom": "1489", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1491", "top": "1491", "type": "elementwise", "name": "1491", "beta": 0 }, { "bottom": "1491", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.2", "top": "denom.2", "type": "elementwise", "name": "denom.2", "beta": 0 }, { "bottom": "zero_mean.2,denom.2", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.2", "top": "out.2", "type": "elementwise", "name": "out.2", "beta": 0 }, { "top": "1495", "w": 1, "h": 1, "name": "1495", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1495", "n": 1, "weights": {}, "constant_blob": 617 }, { "bottom": "out.2,1495", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1496", "top": "1496", "type": "elementwise", "name": "1496", "beta": 0 }, { "top": "1498", "w": 1, "h": 1, "name": "1498", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1498", "n": 1, "weights": {}, "constant_blob": 533 }, { "bottom": "1496,1498", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "k.2", "top": "k.2", "type": "elementwise", "name": "k.2", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "1513", "K": 512, "blob_biases": 241, "name": "1513", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.2", "weights": { "per_ch_qscale": 243, "W_S8": 239, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "k.1", "K": 512, "blob_biases": 247, "name": "k.1", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.2", "weights": { "per_ch_qscale": 249, "W_S8": 245, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "1527", "K": 512, "blob_biases": 253, "name": "1527", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "k.2", "weights": { "per_ch_qscale": 255, "W_S8": 251, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 1, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "1513", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "1528", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "1528", "begin_48": 0, "top": "1528_0,1528_1,1528_2,1528_3,1528_4,1528_5,1528_6,1528_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 1, "axis_w": 2, "bottom": "k.1", "axis_k": 0, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "transpose_1", "top": "transpose_1", "type": "transpose", "name": "transpose_1" }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 3, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "transpose_1", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "1538", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "1538", "begin_48": 0, "top": "1538_0,1538_1,1538_2,1538_3,1538_4,1538_5,1538_6,1538_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 1, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "1527", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "1547", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "1547", "begin_48": 0, "top": "1547_0,1547_1,1547_2,1547_3,1547_4,1547_5,1547_6,1547_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1538_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1557_transpose_x", "top": "1538_0_transposed", "type": "transpose", "name": "1557_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1528_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1557_transpose_y", "top": "1528_0_transposed", "type": "transpose", "name": "1557_transpose_y" }, { "bottom": "1538_0_transposed,1528_0_transposed", "weights": {}, "debug_info": "1557_batch_matmul", "top": "1557_pre_transpose", "type": "batch_matmul", "name": "1557_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1557_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1557", "top": "1557", "type": "transpose", "name": "1557" }, { "bottom": "1557", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1559", "top": "1559", "type": "elementwise", "name": "1559", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1538_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1561_transpose_x", "top": "1538_1_transposed", "type": "transpose", "name": "1561_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1528_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1561_transpose_y", "top": "1528_1_transposed", "type": "transpose", "name": "1561_transpose_y" }, { "bottom": "1538_1_transposed,1528_1_transposed", "weights": {}, "debug_info": "1561_batch_matmul", "top": "1561_pre_transpose", "type": "batch_matmul", "name": "1561_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1561_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1561", "top": "1561", "type": "transpose", "name": "1561" }, { "bottom": "1561", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1563", "top": "1563", "type": "elementwise", "name": "1563", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1538_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1565_transpose_x", "top": "1538_2_transposed", "type": "transpose", "name": "1565_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1528_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1565_transpose_y", "top": "1528_2_transposed", "type": "transpose", "name": "1565_transpose_y" }, { "bottom": "1538_2_transposed,1528_2_transposed", "weights": {}, "debug_info": "1565_batch_matmul", "top": "1565_pre_transpose", "type": "batch_matmul", "name": "1565_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1565_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1565", "top": "1565", "type": "transpose", "name": "1565" }, { "bottom": "1565", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1567", "top": "1567", "type": "elementwise", "name": "1567", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1538_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1569_transpose_x", "top": "1538_3_transposed", "type": "transpose", "name": "1569_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1528_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1569_transpose_y", "top": "1528_3_transposed", "type": "transpose", "name": "1569_transpose_y" }, { "bottom": "1538_3_transposed,1528_3_transposed", "weights": {}, "debug_info": "1569_batch_matmul", "top": "1569_pre_transpose", "type": "batch_matmul", "name": "1569_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1569_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1569", "top": "1569", "type": "transpose", "name": "1569" }, { "bottom": "1569", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1571", "top": "1571", "type": "elementwise", "name": "1571", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1538_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1573_transpose_x", "top": "1538_4_transposed", "type": "transpose", "name": "1573_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1528_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1573_transpose_y", "top": "1528_4_transposed", "type": "transpose", "name": "1573_transpose_y" }, { "bottom": "1538_4_transposed,1528_4_transposed", "weights": {}, "debug_info": "1573_batch_matmul", "top": "1573_pre_transpose", "type": "batch_matmul", "name": "1573_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1573_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1573", "top": "1573", "type": "transpose", "name": "1573" }, { "bottom": "1573", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1575", "top": "1575", "type": "elementwise", "name": "1575", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1538_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1577_transpose_x", "top": "1538_5_transposed", "type": "transpose", "name": "1577_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1528_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1577_transpose_y", "top": "1528_5_transposed", "type": "transpose", "name": "1577_transpose_y" }, { "bottom": "1538_5_transposed,1528_5_transposed", "weights": {}, "debug_info": "1577_batch_matmul", "top": "1577_pre_transpose", "type": "batch_matmul", "name": "1577_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1577_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1577", "top": "1577", "type": "transpose", "name": "1577" }, { "bottom": "1577", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1579", "top": "1579", "type": "elementwise", "name": "1579", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1538_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1581_transpose_x", "top": "1538_6_transposed", "type": "transpose", "name": "1581_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1528_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1581_transpose_y", "top": "1528_6_transposed", "type": "transpose", "name": "1581_transpose_y" }, { "bottom": "1538_6_transposed,1528_6_transposed", "weights": {}, "debug_info": "1581_batch_matmul", "top": "1581_pre_transpose", "type": "batch_matmul", "name": "1581_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1581_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1581", "top": "1581", "type": "transpose", "name": "1581" }, { "bottom": "1581", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1583", "top": "1583", "type": "elementwise", "name": "1583", "beta": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1538_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1585_transpose_x", "top": "1538_7_transposed", "type": "transpose", "name": "1585_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1528_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1585_transpose_y", "top": "1528_7_transposed", "type": "transpose", "name": "1585_transpose_y" }, { "bottom": "1538_7_transposed,1528_7_transposed", "weights": {}, "debug_info": "1585_batch_matmul", "top": "1585_pre_transpose", "type": "batch_matmul", "name": "1585_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1585_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1585", "top": "1585", "type": "transpose", "name": "1585" }, { "bottom": "1585", "alpha": 0.125, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1587", "top": "1587", "type": "elementwise", "name": "1587", "beta": 0 }, { "bottom": "1559,1563,1567,1571,1575,1579,1583,1587", "weights": {}, "nd_mode": true, "axis": 2, "debug_info": "attn_weights.1", "top": "attn_weights.1", "type": "general_concat", "name": "attn_weights.1" }, { "bottom": "attn_weights.1,qk_mask.1", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "attn_weights0.1", "nd_mode": true, "top": "attn_weights0.1", "type": "elementwise", "name": "attn_weights0.1", "beta": 0 }, { "bottom": "attn_weights0.1", "weights": {}, "debug_info": "input.4", "top": "input.4", "C": 2, "type": "softmax", "name": "input.4" }, { "begin_104": 0, "begin_63": 0, "begin_59": 0, "begin_64": 0, "begin_105": 0, "begin_65": 0, "begin_0": 0, "begin_70": 0, "begin_66": 0, "begin_106": 0, "begin_110": 0, "begin_2": 0, "begin_71": 0, "begin_67": 0, "begin_68": 0, "begin_107": 0, "begin_111": 0, "begin_4": 0, "begin_72": 0, "begin_73": 0, "begin_69": 0, "begin_6": 0, "begin_74": 0, "begin_108": 0, "type": "split_nd", "begin_8": 0, "begin_112": 0, "begin_75": 0, "begin_10": 0, "begin_80": 0, "begin_76": 0, "begin_109": 0, "begin_113": 0, "begin_11": 0, "nd_axis": 2, "begin_81": 0, "begin_77": 0, "begin_114": 0, "begin_12": 0, "begin_78": 0, "begin_82": 0, "begin_13": 0, "begin_83": 0, "begin_79": 0, "begin_115": 0, "begin_14": 0, "begin_84": 0, "begin_15": 0, "begin_85": 0, "num_splits": 8, "begin_116": 0, "begin_120": 0, "begin_20": 0, "begin_86": 0, "begin_16": 0, "begin_90": 0, "begin_21": 0, "begin_91": 0, "begin_87": 0, "begin_17": 0, "begin_117": 0, "begin_121": 0, "begin_18": 0, "begin_92": 0, "begin_88": 0, "begin_22": 0, "begin_23": 0, "begin_89": 0, "begin_19": 0, "begin_93": 0, "begin_118": 0, "begin_122": 0, "begin_24": 0, "begin_94": 0, "begin_25": 0, "begin_95": 0, "begin_119": 0, "begin_123": 0, "begin_30": 0, "begin_26": 0, "begin_96": 0, "begin_124": 0, "begin_31": 0, "begin_97": 0, "begin_27": 0, "begin_28": 0, "begin_98": 0, "begin_32": 0, "begin_125": 0, "begin_33": 0, "begin_29": 0, "begin_99": 0, "begin_1": 0, "weights": {}, "begin_34": 0, "bottom": "input.4", "begin_3": 0, "begin_126": 0, "begin_35": 0, "begin_5": 0, "begin_36": 0, "begin_40": 0, "begin_127": 0, "name": "1593", "begin_41": 0, "begin_37": 0, "begin_7": 0, "begin_42": 0, "begin_38": 0, "begin_9": 0, "begin_39": 0, "begin_43": 0, "begin_44": 0, "begin_45": 0, "begin_50": 0, "begin_46": 0, "begin_51": 0, "begin_47": 0, "begin_100": 0, "begin_52": 0, "debug_info": "1593", "begin_48": 0, "top": "1593_0,1593_1,1593_2,1593_3,1593_4,1593_5,1593_6,1593_7", "begin_101": 0, "begin_53": 0, "begin_49": 0, "begin_54": 0, "begin_55": 0, "begin_102": 0, "begin_60": 0, "begin_56": 0, "begin_103": 0, "begin_57": 0, "begin_61": 0, "begin_62": 0, "begin_58": 0 }, { "axis_h": 2, "axis_w": 0, "bottom": "1547_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1603_transpose_x", "top": "1547_0_transposed", "type": "transpose", "name": "1603_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1593_0", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1603_transpose_y", "top": "1593_0_transposed", "type": "transpose", "name": "1603_transpose_y" }, { "bottom": "1547_0_transposed,1593_0_transposed", "weights": {}, "debug_info": "1603_batch_matmul", "top": "1603_pre_transpose", "type": "batch_matmul", "name": "1603_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1603_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1603", "top": "1603", "type": "transpose", "name": "1603" }, { "axis_h": 2, "axis_w": 0, "bottom": "1547_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1605_transpose_x", "top": "1547_1_transposed", "type": "transpose", "name": "1605_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1593_1", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1605_transpose_y", "top": "1593_1_transposed", "type": "transpose", "name": "1605_transpose_y" }, { "bottom": "1547_1_transposed,1593_1_transposed", "weights": {}, "debug_info": "1605_batch_matmul", "top": "1605_pre_transpose", "type": "batch_matmul", "name": "1605_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1605_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1605", "top": "1605", "type": "transpose", "name": "1605" }, { "axis_h": 2, "axis_w": 0, "bottom": "1547_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1607_transpose_x", "top": "1547_2_transposed", "type": "transpose", "name": "1607_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1593_2", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1607_transpose_y", "top": "1593_2_transposed", "type": "transpose", "name": "1607_transpose_y" }, { "bottom": "1547_2_transposed,1593_2_transposed", "weights": {}, "debug_info": "1607_batch_matmul", "top": "1607_pre_transpose", "type": "batch_matmul", "name": "1607_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1607_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1607", "top": "1607", "type": "transpose", "name": "1607" }, { "axis_h": 2, "axis_w": 0, "bottom": "1547_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1609_transpose_x", "top": "1547_3_transposed", "type": "transpose", "name": "1609_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1593_3", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1609_transpose_y", "top": "1593_3_transposed", "type": "transpose", "name": "1609_transpose_y" }, { "bottom": "1547_3_transposed,1593_3_transposed", "weights": {}, "debug_info": "1609_batch_matmul", "top": "1609_pre_transpose", "type": "batch_matmul", "name": "1609_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1609_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1609", "top": "1609", "type": "transpose", "name": "1609" }, { "axis_h": 2, "axis_w": 0, "bottom": "1547_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1611_transpose_x", "top": "1547_4_transposed", "type": "transpose", "name": "1611_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1593_4", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1611_transpose_y", "top": "1593_4_transposed", "type": "transpose", "name": "1611_transpose_y" }, { "bottom": "1547_4_transposed,1593_4_transposed", "weights": {}, "debug_info": "1611_batch_matmul", "top": "1611_pre_transpose", "type": "batch_matmul", "name": "1611_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1611_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1611", "top": "1611", "type": "transpose", "name": "1611" }, { "axis_h": 2, "axis_w": 0, "bottom": "1547_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1613_transpose_x", "top": "1547_5_transposed", "type": "transpose", "name": "1613_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1593_5", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1613_transpose_y", "top": "1593_5_transposed", "type": "transpose", "name": "1613_transpose_y" }, { "bottom": "1547_5_transposed,1593_5_transposed", "weights": {}, "debug_info": "1613_batch_matmul", "top": "1613_pre_transpose", "type": "batch_matmul", "name": "1613_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1613_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1613", "top": "1613", "type": "transpose", "name": "1613" }, { "axis_h": 2, "axis_w": 0, "bottom": "1547_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1615_transpose_x", "top": "1547_6_transposed", "type": "transpose", "name": "1615_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1593_6", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1615_transpose_y", "top": "1593_6_transposed", "type": "transpose", "name": "1615_transpose_y" }, { "bottom": "1547_6_transposed,1593_6_transposed", "weights": {}, "debug_info": "1615_batch_matmul", "top": "1615_pre_transpose", "type": "batch_matmul", "name": "1615_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1615_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1615", "top": "1615", "type": "transpose", "name": "1615" }, { "axis_h": 2, "axis_w": 0, "bottom": "1547_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1617_transpose_x", "top": "1547_7_transposed", "type": "transpose", "name": "1617_transpose_x" }, { "axis_h": 2, "axis_w": 0, "bottom": "1593_7", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1617_transpose_y", "top": "1593_7_transposed", "type": "transpose", "name": "1617_transpose_y" }, { "bottom": "1547_7_transposed,1593_7_transposed", "weights": {}, "debug_info": "1617_batch_matmul", "top": "1617_pre_transpose", "type": "batch_matmul", "name": "1617_batch_matmul", "channel_mode": false }, { "axis_h": 2, "axis_w": 0, "bottom": "1617_pre_transpose", "axis_k": 1, "axis_n": 3, "axis_seq": 4, "weights": {}, "debug_info": "1617", "top": "1617", "type": "transpose", "name": "1617" }, { "weights": {}, "debug_info": "attn.4", "top": "attn.4", "type": "concat", "name": "attn.4", "bottom": "1603,1605,1607,1609,1611,1613,1615,1617" }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "inputs.4", "K": 512, "blob_biases": 259, "name": "inputs.4", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "attn.4", "weights": { "per_ch_qscale": 261, "W_S8": 257, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "axis_mode": 4, "bottom": "inputs.4", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.4", "use_version": 1, "top": "channels_mean.4", "type": "reduce", "name": "channels_mean.4" }, { "alpha": -1, "bottom": "channels_mean.4", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.4", "top": "_neg_y_zero_mean.4", "type": "activation", "name": "_neg_y_zero_mean.4", "beta": 0 }, { "bottom": "inputs.4,_neg_y_zero_mean.4", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.4", "top": "zero_mean.4", "type": "elementwise", "name": "zero_mean.4", "beta": 0 }, { "bottom": "zero_mean.4,zero_mean.4", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.4", "top": "zero_mean_sq.4", "type": "elementwise", "name": "zero_mean_sq.4", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.4", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "1640", "use_version": 1, "top": "1640", "type": "reduce", "name": "1640" }, { "bottom": "1640", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1642", "top": "1642", "type": "elementwise", "name": "1642", "beta": 0 }, { "bottom": "1642", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.4", "top": "denom.4", "type": "elementwise", "name": "denom.4", "beta": 0 }, { "bottom": "zero_mean.4,denom.4", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.4", "top": "out.4", "type": "elementwise", "name": "out.4", "beta": 0 }, { "top": "1646", "w": 1, "h": 1, "name": "1646", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1646", "n": 1, "weights": {}, "constant_blob": 619 }, { "bottom": "out.4,1646", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1647", "top": "1647", "type": "elementwise", "name": "1647", "beta": 0 }, { "top": "1649", "w": 1, "h": 1, "name": "1649", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1649", "n": 1, "weights": {}, "constant_blob": 537 }, { "bottom": "1647,1649", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.6", "top": "input.6", "type": "elementwise", "name": "input.6", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.5", "K": 512, "blob_biases": 541, "name": "x.5", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 64, "bottom": "input.6", "weights": { "per_ch_qscale": 543, "W_S8": 539, "per_ch_qbias": 579 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.5", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1659", "top": "1659", "type": "elementwise", "name": "1659", "beta": 0 }, { "bottom": "1659", "weights": {}, "mode": 3, "debug_info": "1660", "top": "1660", "type": "activation", "name": "1660" }, { "bottom": "x.5,1660", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.12", "top": "input.12", "type": "elementwise", "name": "input.12", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.3", "K": 64, "blob_biases": 547, "name": "x.3", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.12", "weights": { "per_ch_qscale": 549, "W_S8": 545, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.3,inputs.4", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "attn.1", "top": "attn.1", "type": "elementwise", "name": "attn.1", "beta": 0 }, { "bottom": "inputs3.1,attn.1", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "inputs0.1", "top": "inputs0.1", "type": "elementwise", "name": "inputs0.1", "beta": 0 }, { "axis_mode": 4, "bottom": "inputs0.1", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.6", "use_version": 1, "top": "channels_mean.6", "type": "reduce", "name": "channels_mean.6" }, { "alpha": -1, "bottom": "channels_mean.6", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.6", "top": "_neg_y_zero_mean.6", "type": "activation", "name": "_neg_y_zero_mean.6", "beta": 0 }, { "bottom": "inputs0.1,_neg_y_zero_mean.6", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.6", "top": "zero_mean.6", "type": "elementwise", "name": "zero_mean.6", "beta": 0 }, { "bottom": "zero_mean.6,zero_mean.6", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.6", "top": "zero_mean_sq.6", "type": "elementwise", "name": "zero_mean_sq.6", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.6", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "1678", "use_version": 1, "top": "1678", "type": "reduce", "name": "1678" }, { "bottom": "1678", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1680", "top": "1680", "type": "elementwise", "name": "1680", "beta": 0 }, { "bottom": "1680", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.6", "top": "denom.6", "type": "elementwise", "name": "denom.6", "beta": 0 }, { "bottom": "zero_mean.6,denom.6", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.6", "top": "out.6", "type": "elementwise", "name": "out.6", "beta": 0 }, { "top": "1684", "w": 1, "h": 1, "name": "1684", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1684", "n": 1, "weights": {}, "constant_blob": 621 }, { "bottom": "out.6,1684", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1685", "top": "1685", "type": "elementwise", "name": "1685", "beta": 0 }, { "top": "1687", "w": 1, "h": 1, "name": "1687", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1687", "n": 1, "weights": {}, "constant_blob": 553 }, { "bottom": "1685,1687", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.8", "top": "input.8", "type": "elementwise", "name": "input.8", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.7", "K": 512, "blob_biases": 269, "name": "x.7", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 2048, "bottom": "input.8", "weights": { "per_ch_qscale": 271, "W_S8": 267, "per_ch_qbias": 299 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.7", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1699", "top": "1699", "type": "elementwise", "name": "1699", "beta": 0 }, { "bottom": "1699", "weights": {}, "mode": 3, "debug_info": "1700", "top": "1700", "type": "activation", "name": "1700" }, { "bottom": "x.7,1700", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.10", "top": "input.10", "type": "elementwise", "name": "input.10", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "input0.1", "K": 2048, "blob_biases": 275, "name": "input0.1", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.10", "weights": { "per_ch_qscale": 277, "W_S8": 273, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "axis_mode": 4, "bottom": "input0.1", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.43", "use_version": 1, "top": "channels_mean.43", "type": "reduce", "name": "channels_mean.43" }, { "alpha": -1, "bottom": "channels_mean.43", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.43", "top": "_neg_y_zero_mean.43", "type": "activation", "name": "_neg_y_zero_mean.43", "beta": 0 }, { "bottom": "input0.1,_neg_y_zero_mean.43", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.43", "top": "zero_mean.43", "type": "elementwise", "name": "zero_mean.43", "beta": 0 }, { "bottom": "zero_mean.43,zero_mean.43", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.43", "top": "zero_mean_sq.43", "type": "elementwise", "name": "zero_mean_sq.43", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.43", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "1720", "use_version": 1, "top": "1720", "type": "reduce", "name": "1720" }, { "bottom": "1720", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1722", "top": "1722", "type": "elementwise", "name": "1722", "beta": 0 }, { "bottom": "1722", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.43", "top": "denom.43", "type": "elementwise", "name": "denom.43", "beta": 0 }, { "bottom": "zero_mean.43,denom.43", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.43", "top": "out.43", "type": "elementwise", "name": "out.43", "beta": 0 }, { "top": "1726", "w": 1, "h": 1, "name": "1726", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1726", "n": 1, "weights": {}, "constant_blob": 623 }, { "bottom": "out.43,1726", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1727", "top": "1727", "type": "elementwise", "name": "1727", "beta": 0 }, { "top": "1729", "w": 1, "h": 1, "name": "1729", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1729", "n": 1, "weights": {}, "constant_blob": 557 }, { "bottom": "1727,1729", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.1", "top": "input.1", "type": "elementwise", "name": "input.1", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.2", "K": 512, "blob_biases": 561, "name": "x.2", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 64, "bottom": "input.1", "weights": { "per_ch_qscale": 563, "W_S8": 559, "per_ch_qbias": 579 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.2", "alpha": 1.7020000219345093, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1739", "top": "1739", "type": "elementwise", "name": "1739", "beta": 0 }, { "bottom": "1739", "weights": {}, "mode": 3, "debug_info": "1740", "top": "1740", "type": "activation", "name": "1740" }, { "bottom": "x.2,1740", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "input.2", "top": "input.2", "type": "elementwise", "name": "input.2", "beta": 0 }, { "pad_r": 0, "fused_relu": 0, "fused_tanh": 0, "debug_info": "", "pad_fill_mode": 0, "pad_b": 0, "pad_l": 0, "top": "x.1", "K": 64, "blob_biases": 567, "name": "x.1", "has_batch_norm": 0, "type": "convolution", "n_groups": 1, "pad_t": 0, "has_biases": 1, "C": 512, "bottom": "input.2", "weights": { "per_ch_qscale": 569, "W_S8": 565, "per_ch_qbias": 295 }, "Nx": 1, "pad_mode": 0, "pad_value": 0, "Ny": 1, "n_parallel": 1 }, { "bottom": "x.1,input0.1", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "f.1", "top": "f.1", "type": "elementwise", "name": "f.1", "beta": 0 }, { "bottom": "f.1,inputs0.1", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "x1.1", "top": "x1.1", "type": "elementwise", "name": "x1.1", "beta": 0 }, { "bottom": "inputs3.1", "alpha": 0, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "1752", "top": "1752", "type": "elementwise", "name": "1752", "beta": 0 }, { "bottom": "1752,x1.1", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "inputs4.1", "top": "inputs4.1", "type": "elementwise", "name": "inputs4.1", "beta": 0 }, { "axis_mode": 4, "bottom": "inputs4.1", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "channels_mean.1", "use_version": 1, "top": "channels_mean.1", "type": "reduce", "name": "channels_mean.1" }, { "alpha": -1, "bottom": "channels_mean.1", "weights": {}, "mode": 6, "debug_info": "_neg_y_zero_mean.1", "top": "_neg_y_zero_mean.1", "type": "activation", "name": "_neg_y_zero_mean.1", "beta": 0 }, { "bottom": "inputs4.1,_neg_y_zero_mean.1", "alpha": 0.010416666977107525, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean.1", "top": "zero_mean.1", "type": "elementwise", "name": "zero_mean.1", "beta": 0 }, { "bottom": "zero_mean.1,zero_mean.1", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "zero_mean_sq.1", "top": "zero_mean_sq.1", "type": "elementwise", "name": "zero_mean_sq.1", "beta": 0 }, { "axis_mode": 4, "bottom": "zero_mean_sq.1", "weights": {}, "mode": 1, "nd_axis": 1, "nd_mode": true, "debug_info": "1763", "use_version": 1, "top": "1763", "type": "reduce", "name": "1763" }, { "bottom": "1763", "alpha": 1.0850693676900391e-09, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1765", "top": "1765", "type": "elementwise", "name": "1765", "beta": 0 }, { "bottom": "1765", "alpha": 1, "operation": 12, "eps": 9.999999960041972e-13, "weights": {}, "fused_relu": 0, "debug_info": "denom.1", "top": "denom.1", "type": "elementwise", "name": "denom.1", "beta": 0 }, { "bottom": "zero_mean.1,denom.1", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "out.1", "top": "out.1", "type": "elementwise", "name": "out.1", "beta": 0 }, { "top": "1769", "w": 1, "h": 1, "name": "1769", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1769", "n": 1, "weights": {}, "constant_blob": 625 }, { "bottom": "out.1,1769", "alpha": 1, "operation": 0, "weights": {}, "fused_relu": 0, "debug_info": "1770", "top": "1770", "type": "elementwise", "name": "1770", "beta": 0 }, { "top": "1772", "w": 1, "h": 1, "name": "1772", "nd_rank": 4, "type": "load_constant", "k": 512, "bottom": "", "debug_info": "1772", "n": 1, "weights": {}, "constant_blob": 573 }, { "bottom": "1770,1772", "alpha": 1, "operation": 1, "weights": {}, "fused_relu": 0, "debug_info": "embeddings.1", "top": "embeddings.1", "type": "elementwise", "name": "embeddings.1", "beta": 0 }, { "axis_seq": 4, "name": "transpose_0", "axis_n": 3, "axis_h": 1, "type": "transpose", "attributes": { "is_output": 1 }, "bottom": "embeddings.1", "axis_w": 2, "axis_k": 0, "debug_info": "transpose_0", "weights": {}, "top": "embedding_out" } ] }