| { | |
| "data_dir": "/code/VLA/datasets/SSR-CoT", | |
| "clip_path": "/code/VLA/models/clip-vit-large-patch14-336", | |
| "siglip_path": "/code/VLA/models/siglip-so400m-patch14-384", | |
| "segformer_path": "/code/VLA/models/segmentation_models/segformer-b5-finetuned-ade-640-640", | |
| "use_semantic": true, | |
| "n_tor": 10, | |
| "max_length": [ | |
| 256, | |
| 1024, | |
| 256 | |
| ], | |
| "lr": 2e-05, | |
| "mamba": "/code/VLA/models/state-spaces/mamba-130m-hf", | |
| "llm": "/code/VLA/models/Qwen2.5-7B", | |
| "epochs": 2, | |
| "batch_size_per_gpu": 2, | |
| "gradient_accumulation_steps": 4, | |
| "warmup_ratio": 0.02, | |
| "output_dir": "checkpoints/SSR-MIDI-trimodal/2025-09-28 17:03:11", | |
| "use_precomputed": true, | |
| "precomputed_dir": "/code/VLA/datasets/SSR-CoT/precomputed_features", | |
| "num_workers": 4, | |
| "use_gradient_checkpointing": true | |
| } |