| { |
| "_name_or_path": "/home/jovyan/shared/mahdi/1_projects/model_optimization/01all_esm_models/deepspeed/esm/all_checkpoints_4good/esm_650M_half_batch_128_2025-01-29/checkpoint-330000/", |
| "architectures": [ |
| "EsmForMaskedLM" |
| ], |
| "attention_probs_dropout_prob": 0.1, |
| "emb_layer_norm_before": null, |
| "esmfold_config": null, |
| "hidden_dropout_prob": 0.1, |
| "hidden_size": 1280, |
| "initializer_range": 0.02, |
| "intermediate_size": 5120, |
| "is_folding_model": false, |
| "layer_norm_eps": 1e-12, |
| "mask_token_id": 32, |
| "max_position_embeddings": 322, |
| "model_type": "esm", |
| "num_attention_heads": 20, |
| "num_hidden_layers": 33, |
| "pad_token_id": 1, |
| "position_embedding_type": "rotary", |
| "token_dropout": false, |
| "torch_dtype": "float32", |
| "transformers_version": "4.44.2", |
| "use_cache": true, |
| "vocab_list": null, |
| "vocab_size": 33 |
| } |
|
|
