| { | |
| "model_type": "slip", | |
| "architectures": [ | |
| "SLIP" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "configuration_slip.SLIPConfig", | |
| "AutoModel": "modeling_slip.SLIPModel" | |
| }, | |
| "llm_model_name": "google/gemma-3-270m", | |
| "max_llm_len": 768, | |
| "num_img_queries": 64, | |
| "num_heads": 5, | |
| "caption_loss_weight": 1.0, | |
| "contrastive_loss_weight": 1.0, | |
| "use_lora": false, | |
| "unlocked_layers": 4, | |
| "split_layer": 12, | |
| "common_dim": 640, | |
| "post_train": true, | |
| "sensor_encoder": { | |
| "embed_dim": 768, | |
| "num_heads": 12, | |
| "mlp_ratio": 4, | |
| "depth": 12, | |
| "dropout_rate": 0.1, | |
| "learnable_pos_emb": false, | |
| "max_position_embeddings": 4880, | |
| "patch_size": null, | |
| "channel_attn_type": "all_attn" | |
| } | |
| } |