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
File size: 1,219 Bytes
052926e | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 | 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
|