| version: 1.1.0 | |
| __metadata__: | |
| per_device_train_batch_size: 48 | |
| per_device_eval_batch_size: 48 | |
| fp16: False | |
| framework_metadata: | |
| python_version: 3.8.13 | |
| sparseml_version: 1.0.1 | |
| torch_version: 1.9.1+cu111 | |
| modifiers: | |
| - !GMPruningModifier | |
| end_epoch: 11 | |
| final_sparsity: 0.7 | |
| global_sparsity: False | |
| init_sparsity: 0.0 | |
| inter_func: cubic | |
| leave_enabled: True | |
| mask_type: unstructured | |
| params: ['re:roberta.encoder.layer.*.attention.self.query.weight', 're:roberta.encoder.layer.*.attention.self.key.weight', 're:roberta.encoder.layer.*.attention.self.value.weight', 're:roberta.encoder.layer.*.attention.output.dense.weight', 're:roberta.encoder.layer.*.intermediate.dense.weight', 're:roberta.encoder.layer.*.output.dense.weight', 're:classifier.dense.weight'] | |
| start_epoch: 2 | |
| update_frequency: 0.05 | |
| - !EpochRangeModifier | |
| end_epoch: 15 | |
| start_epoch: 0.0 | |