Downloads · 30 days
2
12% of all-time downloads
deformable-bench/fastwam-grasp-silk-bimanual-midtrain75k-lerobotaug
fastwam-grasp-silk-bimanual-midtrain75k-lerobotaug is a robotics model from deformable-bench. Use it for the robotics task on the model card, and read the license before you ship it in a product. It is set up for fastwam.
FastWAM checkpoint fine-tuned on the grasp-silk-bimanual task with a LeRobot-aligned image augmentation pipeline.
Downloads · 30 days
2
12% of all-time downloads
All-time downloads
17
Public
Repo size
12 GB
Likes
0
Public
Click a slice to open those files.
.pt12 GB · 100%
From the Hugging Face model README
FastWAM checkpoint fine-tuned on the grasp-silk-bimanual task with a
LeRobot-aligned image augmentation pipeline.
deformable-bench/grasp-silk-bimanualThe run completed normally at step 1,430. Mean logged losses over selected windows were:
| Window | Total | Action | Video |
|---|---|---|---|
| Steps 10-100 | 0.52408 | 0.32551 | 0.19857 |
| Steps 110-300 | 0.14679 | 0.02486 | 0.12194 |
| Steps 610-800 | 0.10566 | 0.00600 | 0.09964 |
| Steps 1240-1430 | 0.09265 | 0.00227 | 0.09040 |
| Final step 1430 | 0.09280 | 0.00210 | 0.09070 |
The action loss converged strongly, while the late video/total loss remained stable with normal batch-to-batch variation and no late rebound.
checkpoints/step_001430.pt: final model weightsconfig.yaml: resolved training configurationdataset_stats.json: normalization statistics used by trainingtrain.log: complete trainer logOptimizer and distributed trainer state are intentionally excluded.
This checkpoint is intended for research and evaluation on the matching
FastWAM codebase and grasp-silk-bimanual observation/action schema. It has not
been validated for safety-critical or out-of-distribution deployment.