Instructions to use adipotnis/Cosmos3-Edge-Cable-Manager with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- LeRobot
How to use adipotnis/Cosmos3-Edge-Cable-Manager with LeRobot:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
Cosmos3 Edge β Cable manager
Final 10,000-step checkpoint, completed successfully on November 12, 2026. Training used 50 demonstrations (20,298 frames) with high, low and wrist cameras. The run trained to step 4000 on DeltaAI GH200, then resumed the saved model, optimizer, scheduler and RNG state on Delta H200 through step 10000. No robot evaluation result is claimed.
Recipe: batch 16, action horizon 16, learning rate 5e-5, 500 warmup steps, linear decay through 10000 steps, bf16, gradient checkpointing, state input, quantile state/action normalization, seed 42. Each camera is resized to 192x256; high above low (left) and wrist (right).
The repository root contains model weights, tokenizer, processors, normalization
state and training configuration. training_state/ contains optimizer, scheduler,
RNG and step state. To reconstruct the local checkpoint layout for training,
place the root model files under 010000/pretrained_model/ and the state files
under 010000/training_state/.
- Downloads last month
- 10