| { | |
| "kind": "itm", | |
| "config": { | |
| "model": { | |
| "kind": "itm", | |
| "x_variant": "vit_large_384", | |
| "y_model_id": "google/embeddinggemma-300m", | |
| "hidden": 768, | |
| "depth": 4, | |
| "n_heads": 12, | |
| "itc_dim": 256, | |
| "gradient_checkpointing": true, | |
| "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, | |
| "activation_checkpointing": true | |
| }, | |
| "y_finetune": { | |
| "lora_rank": 16, | |
| "lora_alpha": 32, | |
| "lora_dropout": 0.05 | |
| } | |
| }, | |
| "train": { | |
| "steps": 6000, | |
| "warmup_steps": 600, | |
| "lr": 0.0001, | |
| "x_encoder_lr": 5e-06, | |
| "x_lora_lr": 0.0002, | |
| "y_lora_lr": 5e-06, | |
| "weight_decay": 0.01, | |
| "batch_size_per_gpu": 12, | |
| "n_random_neg": 1, | |
| "itc_weight": 1.0, | |
| "itc_mined_k": 8, | |
| "mined_neg_k": 0, | |
| "log_every": 50, | |
| "save_every": 1000 | |
| }, | |
| "data": { | |
| "image_size": 384, | |
| "num_workers": 8, | |
| "balance": "class", | |
| "split": "train", | |
| "use_hard_negs": true, | |
| "hard_neg_annotation_dir": "/tank2/eccv-2026/annotation/train" | |
| }, | |
| "wandb": { | |
| "project": "eccv-2026", | |
| "run_name": "scout_itm_v3_K64_pred24_concat_explora_loraY" | |
| } | |
| }, | |
| "trainable_param_count": 120255746, | |
| "total_param_count": 656071688, | |
| "reported_val_map10": 0.8858, | |
| "reported_val_hit10": 0.9854, | |
| "note": "blend a=0.5 ITC+ITM", | |
| "source_checkpoint": "/tank2/eccv-2026/runs/scout_itm_v3/step_00006000_final" | |
| } |