Instructions to use Spa-Bench/spa-bench-model-vla-0-epoch-12 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Spa-Bench/spa-bench-model-vla-0-epoch-12 with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Spa-Bench/spa-bench-model-vla-0-epoch-12", device_map="auto") - Notebooks
- Google Colab
- Kaggle
| DATALOADER: | |
| ROBOVERSE: | |
| cfg_opts: IMAGE.crop_img:0.875:IMAGE.img_size:224:IMAGE.cam_list:('3p1','3p2') | |
| cfg_path: configs/roboverse/realworld.yaml | |
| convert_ori_act_to_delta_act: false | |
| batch_size: 24 | |
| num_workers: 8 | |
| EXP: | |
| AMP: true | |
| DATASET: roboverse | |
| EXP_ID: spa_bench_review | |
| LOSS: {} | |
| LR_SCHED: cosine_anneal | |
| MODEL: qwen | |
| OPTIMIZER: adamw | |
| SEED: 0 | |
| EXP_EXTRA: | |
| no_test: true | |
| no_track: true | |
| no_val: true | |
| save_ckp: 1 | |
| save_last_ckpt: true | |
| test_eval_freq: 1 | |
| val_eval_freq: 1 | |
| LR_SCHED: | |
| lr_clip: 1.0e-08 | |
| lr_decay_factor: 0.5 | |
| lr_patience: 4 | |
| MODEL: | |
| QWEN: | |
| action_mask_aug_per: 0.4 | |
| action_type: original | |
| add_vision_id: true | |
| attention_dropout: 0.0 | |
| grad_checkpoint: true | |
| history: 1 | |
| horizon: 8 | |
| include_state: true | |
| lora_config: default | |
| lora_rank: 8 | |
| num_bins_actions: 1000 | |
| num_cam: 2 | |
| num_demo_frames: 0 | |
| original_action_dim: 6 | |
| qwen_model_id: Qwen/Qwen2.5-VL-3B-Instruct | |
| rgb_img_size: | |
| - 224 | |
| - 224 | |
| rgb_input: true | |
| tiled_rgb_imgs: true | |
| use_flash_attention_2: false | |
| use_lora: false | |
| use_qlora: false | |
| TRAIN: | |
| clip_grad_norm: 1.0 | |
| demo_cache_refresh_steps: 50 | |
| l2: 0.01 | |
| lr: 5.0e-06 | |
| num_epochs: 16 | |