| { | |
| "_name_or_path": "/home/dmgcsilva/project/PROJECT/dmgcsilva/followup/experiments/vista_stage46/llama8binst_siglip400m_mlp_ln_patches_stage46/checkpoint_3252", | |
| "architectures": [ | |
| "ReVista" | |
| ], | |
| "cap_loss_scale": 1.0, | |
| "connector_type": "mlp", | |
| "end_ret_token_id": 128258, | |
| "eos_token_id": 128009, | |
| "freeze_cap": false, | |
| "freeze_emb": false, | |
| "freeze_lm": false, | |
| "freeze_ret": false, | |
| "freeze_vm": true, | |
| "image_embed_dropout_prob": 0.0, | |
| "img_token_id": 128256, | |
| "lm_lora_alpha": 128, | |
| "lm_lora_dropout": 0.03, | |
| "lm_lora_rank": 32, | |
| "model_type": "revistamodel", | |
| "n_visual_tokens": 1, | |
| "negative_count": 512, | |
| "pad_token_id": 128009, | |
| "projector_type": "linear", | |
| "ret_loss_scale": 1.0, | |
| "ret_token_id": 128257, | |
| "shared_emb_dim": 512, | |
| "start_ret_token_id": 128257, | |
| "task": "captioning", | |
| "text_decoder": "meta-llama/Llama-3.1-8B-Instruct", | |
| "text_embed_dropout_prob": 0.1, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.49.0", | |
| "use_cls_token": false, | |
| "use_lora_on_lm": false, | |
| "use_negatives": false, | |
| "use_pos_emb": true, | |
| "visual_encoder": "google/siglip-so400m-patch14-224", | |
| "vocab_size": 128259 | |
| } | |