| { |
| "out_name": "mt5-small-gigatrue-layercut-D456", |
| "model_archetype": "mt5", |
| "model_name": "google/mt5-small", |
| "tokenizer_name": "google/mt5-small", |
| "model_torch_dtype": "bfloat16", |
| "dataset_lang": "en", |
| "dataset_name": "Plasmoxy/gigatrue", |
| "dataset_tokenized_cache_name": "gigatrue_tokenized_mt5_110-35", |
| "use_half_val_dataset": true, |
| "max_input_length": 110, |
| "max_target_length": 35, |
| "batch_size": 128, |
| "learning_rate": 0.0003, |
| "num_train_epochs": 3, |
| "pkg_versions": { |
| "optimum": "1.23.3", |
| "transformers": "4.45.2", |
| "openvino": "2024.6.0", |
| "nncf": "2.14.1", |
| "torch": "2.5.1", |
| "datasets": "3.2.0", |
| "peft": "0.13.3.dev0" |
| } |
| } |