| { | |
| "kind": "sft", | |
| "config": { | |
| "model": { | |
| "x_variant": "vit_large_384", | |
| "K": 64, | |
| "predictor_layers": 24, | |
| "out_dim": 768, | |
| "readout": "concat", | |
| "gradient_checkpointing": true, | |
| "y_model_id": "sentence-transformers/clip-ViT-L-14", | |
| "x_finetune": { | |
| "unfreeze_first_n": 2, | |
| "unfreeze_last_n": 4, | |
| "lora_rank": 32, | |
| "lora_alpha": 64, | |
| "lora_dropout": 0.05, | |
| "unfreeze_norms": true, | |
| "unfreeze_patch_embed": true | |
| } | |
| }, | |
| "train": { | |
| "steps": 6000, | |
| "warmup_steps": 600, | |
| "lr": 2e-05, | |
| "x_encoder_lr": 5e-06, | |
| "x_lora_lr": 0.0002, | |
| "weight_decay": 0.01, | |
| "temperature": 0.07, | |
| "batch_size_per_gpu": 64, | |
| "grad_accum": 1, | |
| "log_every": 50, | |
| "save_every": 1000, | |
| "eval_every": null | |
| }, | |
| "data": { | |
| "image_size": 384, | |
| "num_workers": 8, | |
| "balance": "class", | |
| "split": "train" | |
| }, | |
| "wandb": { | |
| "project": "eccv-2026", | |
| "run_name": "scout_x_explora_yclip_K64_pred24_concat" | |
| } | |
| }, | |
| "trainable_param_count": 635443008, | |
| "total_param_count": 1291293767, | |
| "reported_val_map10": 0.9297, | |
| "reported_val_hit10": 0.9948, | |
| "note": null, | |
| "source_checkpoint": "/tank2/eccv-2026/runs/scout_x_explora_yclip/step_00006000_final" | |
| } |