R-best-fine-tuned-bart-base-full-ft-reward_short_sentences_and_words-2023-07-13T06-49-08
/
training_parameter.json
| { | |
| "ppo_config": { | |
| "model_name": "R-best-fine-tuned-bart-base-full-ft-reward_short_sentences_and_words-2023-07-13T06-49-08", | |
| "steps": 20000, | |
| "note": "Stopped manually after 65 steps", | |
| "learning_rate": 1e-06, | |
| "target": 6, | |
| "batch_size": 64, | |
| "mini_batch_size": 1, | |
| "ppo_epochs": 4, | |
| "gradient_accumulation_steps": 1, | |
| "log_with": "tensorboard", | |
| "accelerator_kwargs": { | |
| "logging_dir": "/content/results/R-best-fine-tuned-bart-base-full-ft-reward_short_sentences_and_words-2023-07-13T06-49-08/runs" | |
| }, | |
| "early_stopping": false, | |
| "push_to_hub_if_best_kwargs": { | |
| "repo_id": "nlp-lab-2023-seq2seq/R-best-fine-tuned-bart-base-full-ft-reward_short_sentences_and_words-2023-07-13T06-49-08", | |
| "private": true, | |
| "commit_message": "add: new best model" | |
| }, | |
| "compare_steps": 100 | |
| }, | |
| "generation_kwargs": { | |
| "max_new_tokens": 414, | |
| "num_beams": 1, | |
| "early_stopping": true, | |
| "pad_token_id": 1, | |
| "eos_token_id": 2 | |
| }, | |
| "translation_model": "nlp-lab-2023-seq2seq/R-facebook-bart-base-full-ft-without-tum-nlp-german-gpt2_easy-prior-pp-bb5c579f", | |
| "reward_fn": "reward_bleu_short_sentences_and_words", | |
| "finetuning_strategy": "full" | |
| } |