{ "input_info": [ { "sample_size": [ 1, 128 ], "type": "long", "keyword": "input_ids" }, { "sample_size": [ 1, 128 ], "type": "long", "keyword": "token_type_ids" }, { "sample_size": [ 1, 128 ], "type": "long", "keyword": "attention_mask" } ], "bootstrapNAS": { "training": { "algorithm": "progressive_shrinking", "progressivity_of_elasticity": [ "width" ], "batchnorm_adaptation": { "num_bn_adaptation_samples": 0 }, "schedule": { "list_stage_descriptions": [ { "train_dims": [ "width" ], "epochs": 4, "depth_indicator": 1, "width_indicator": 5, "init_lr": 3e-05, "epochs_lr": 4, "sample_rate": 1 } ] }, "elasticity": { "available_elasticity_dims": [ "width" ], "width": { "overwrite_groups": [ [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[0]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[query]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[0]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[key]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[0]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[value]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[1]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[query]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[1]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[key]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[1]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[value]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[2]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[query]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[2]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[key]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[2]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[value]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[3]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[query]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[3]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[key]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[3]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[value]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[4]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[query]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[4]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[key]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[4]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[value]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[5]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[query]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[5]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[key]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[5]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[value]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[6]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[query]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[6]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[key]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[6]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[value]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[7]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[query]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[7]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[key]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[7]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[value]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[8]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[query]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[8]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[key]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[8]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[value]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[9]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[query]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[9]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[key]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[9]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[value]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[10]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[query]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[10]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[key]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[10]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[value]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[11]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[query]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[11]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[key]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[11]/BertAttention[attention]/BertSelfAttention[self]/NNCFLinear[value]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[0]/BertIntermediate[intermediate]/NNCFLinear[dense]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[1]/BertIntermediate[intermediate]/NNCFLinear[dense]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[2]/BertIntermediate[intermediate]/NNCFLinear[dense]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[3]/BertIntermediate[intermediate]/NNCFLinear[dense]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[4]/BertIntermediate[intermediate]/NNCFLinear[dense]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[5]/BertIntermediate[intermediate]/NNCFLinear[dense]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[6]/BertIntermediate[intermediate]/NNCFLinear[dense]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[7]/BertIntermediate[intermediate]/NNCFLinear[dense]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[8]/BertIntermediate[intermediate]/NNCFLinear[dense]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[9]/BertIntermediate[intermediate]/NNCFLinear[dense]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[10]/BertIntermediate[intermediate]/NNCFLinear[dense]/linear_0" ], [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[11]/BertIntermediate[intermediate]/NNCFLinear[dense]/linear_0" ] ], "overwrite_groups_widths": [ [ 768, 576, 384, 320 ], [ 768, 384, 256 ], [ 768, 384, 320, 256 ], [ 768, 512, 384, 256, 192 ], [ 768, 704, 576, 384, 256 ], [ 768, 576, 384, 192 ], [ 768, 704, 384, 192, 128 ], [ 768, 512, 384, 320 ], [ 768, 448, 256 ], [ 768, 320, 256, 192 ], [ 768, 128 ], [ 768, 128, 64 ], [ 3072, 3040, 2727, 2227, 1585 ], [ 3072, 3036, 2712, 2190, 1570 ], [ 3072, 3046, 2795, 2362, 1775 ], [ 3072, 3042, 2791, 2314, 1717 ], [ 3072, 3043, 2764, 2296, 1679 ], [ 3072, 3038, 2768, 2215, 1580 ], [ 3072, 3032, 2786, 2268, 1621 ], [ 3072, 3034, 2694, 2114, 1406 ], [ 3072, 2994, 2506, 1896, 1188 ], [ 3072, 2924, 2236, 1585, 930 ], [ 3072, 2868, 2158, 1501, 828 ], [ 3072, 2812, 1966, 1301, 654 ] ], "add_dynamic_inputs": [ "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[0]/BertAttention[attention]/BertSelfOutput[output]/NNCFLinear[dense]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[1]/BertAttention[attention]/BertSelfOutput[output]/NNCFLinear[dense]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[2]/BertAttention[attention]/BertSelfOutput[output]/NNCFLinear[dense]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[3]/BertAttention[attention]/BertSelfOutput[output]/NNCFLinear[dense]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[4]/BertAttention[attention]/BertSelfOutput[output]/NNCFLinear[dense]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[5]/BertAttention[attention]/BertSelfOutput[output]/NNCFLinear[dense]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[6]/BertAttention[attention]/BertSelfOutput[output]/NNCFLinear[dense]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[7]/BertAttention[attention]/BertSelfOutput[output]/NNCFLinear[dense]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[8]/BertAttention[attention]/BertSelfOutput[output]/NNCFLinear[dense]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[9]/BertAttention[attention]/BertSelfOutput[output]/NNCFLinear[dense]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[10]/BertAttention[attention]/BertSelfOutput[output]/NNCFLinear[dense]/linear_0", "BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[11]/BertAttention[attention]/BertSelfOutput[output]/NNCFLinear[dense]/linear_0" ] } } }, "search": { "algorithm": "NSGA2", "batchnorm_adaptation": { "num_bn_adaptation_samples": 0 }, "num_evals": 1000, "population": 40, "ref_acc": 92.32 } }, "compression": [ { "algorithm": "movement_sparsity", "params": { "warmup_start_epoch": 1, "warmup_end_epoch": 3, "importance_regularization_factor": 0.02, "enable_structured_masking": true }, "sparse_structure_by_scopes": [ { "mode": "block", "sparse_factors": [ 32, 32 ], "target_scopes": "{re}.*BertAttention.*" }, { "mode": "per_dim", "axis": 0, "target_scopes": "{re}.*BertIntermediate.*" }, { "mode": "per_dim", "axis": 1, "target_scopes": "{re}.*BertOutput.*" } ], "ignored_scopes": [ "{re}.*NNCFEmbedding.*", "{re}.*LayerNorm.*", "{re}.*pooler.*", "{re}.*classifier.*" ] } ] }