| checkpointing_steps: '15000' | |
| config_name: null | |
| context_size: 4 | |
| context_word_dropout: 0.1 | |
| dataset_config_name: iwslt2017-en-fr | |
| dataset_name: gsarti/iwslt2017_context | |
| gradient_accumulation_steps: 2 | |
| hub_model_id: context-mt/iwslt17-mbart50-1toM-ctx4-cwd1-en-fr | |
| hub_token: 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 | |
| ignore_pad_token_for_loss: true | |
| learning_rate: 5.0e-05 | |
| logging_steps: '600' | |
| lr_scheduler_type: linear | |
| max_length: 128 | |
| max_source_length: 512 | |
| max_target_length: 128 | |
| max_train_steps: 145520 | |
| model_name_or_path: facebook/mbart-large-50-one-to-many-mmt | |
| model_type: null | |
| num_beams: 5 | |
| num_train_epochs: 5 | |
| num_warmup_steps: 1000 | |
| output_dir: /scratch/p305238/iwslt17-mbart50-1toM-ctx4-cwd1-en-fr | |
| overwrite_cache: false | |
| pad_to_max_length: true | |
| per_device_eval_batch_size: 8 | |
| per_device_train_batch_size: 4 | |
| predict_with_generate: true | |
| preprocessing_num_workers: null | |
| push_to_hub: true | |
| report_to: tensorboard | |
| resume_from_checkpoint: null | |
| sample_context: true | |
| seed: null | |
| source_lang: en_XX | |
| target_lang: fr_XX | |
| tokenizer_name: null | |
| train_file: null | |
| use_slow_tokenizer: false | |
| val_max_target_length: null | |
| validation_file: null | |
| weight_decay: 0.0 | |
| with_tracking: true | |