| { | |
| "compression": [ | |
| { | |
| "algorithm": "quantization", | |
| "export_to_onnx_standard_ops": false, | |
| "ignored_scopes": [ | |
| "{re}.*Embeddings.*", | |
| "{re}.*__add___[0-1]", | |
| "{re}.*layer_norm_0", | |
| "{re}.*matmul_1", | |
| "{re}.*__truediv__*" | |
| ], | |
| "initializer": { | |
| "batchnorm_adaptation": { | |
| "num_bn_adaptation_samples": 0 | |
| }, | |
| "range": { | |
| "num_init_samples": 300, | |
| "type": "mean_min_max" | |
| } | |
| }, | |
| "overflow_fix": "disable", | |
| "preset": "mixed", | |
| "scope_overrides": { | |
| "activations": { | |
| "{re}.*matmul_0": { | |
| "mode": "symmetric" | |
| } | |
| } | |
| } | |
| }, | |
| { | |
| "algorithm": "magnitude_sparsity", | |
| "ignored_scopes": [ | |
| "{re}.*NNCFEmbedding.*", | |
| "{re}.*LayerNorm.*", | |
| "{re}.*pooler.*", | |
| "{re}.*classifier.*" | |
| ], | |
| "params": { | |
| "power": 3, | |
| "schedule": "polynomial", | |
| "sparsity_freeze_epoch": 10, | |
| "sparsity_target": 0.8, | |
| "sparsity_target_epoch": 9, | |
| "steps_per_epoch": 2105, | |
| "update_per_optimizer_step": true | |
| }, | |
| "sparsity_init": 0.0 | |
| } | |
| ], | |
| "input_info": [ | |
| { | |
| "keyword": "input_ids", | |
| "sample_size": [ | |
| 32, | |
| 128 | |
| ], | |
| "type": "long" | |
| }, | |
| { | |
| "keyword": "token_type_ids", | |
| "sample_size": [ | |
| 32, | |
| 128 | |
| ], | |
| "type": "long" | |
| }, | |
| { | |
| "keyword": "attention_mask", | |
| "sample_size": [ | |
| 32, | |
| 128 | |
| ], | |
| "type": "long" | |
| } | |
| ], | |
| "log_dir": "/nvme2/yujiepan/workspace/jpqd-test/LOGS/optimum-magnitude-int8/v1.7-ow-disable.json", | |
| "optimum_version": "1.6.3", | |
| "save_onnx_model": false, | |
| "transformers_version": "4.26.0" | |
| } | |