inclusionAI / Awex
A high-performance RL training-inference weight synchronization framework, designed to enable second-level parameter updates from training to inference in RL workflows
Why this job: topics: reinforcement-learning, training
What it does
- Extreme Sync Speed. Trillion-parameter models fully synchronized within 10 seconds; validated on thousand-GPU
- Unified Weight Adaptation Layer. Automatically handles tensor format/layout differences across parallel strategies
- Zero-Redundancy Transfer & In-Place Update. Transfers only necessary shards; supports in-place GPU memory updates
- Multi-Mode Transfer Support. Support NCCL, RDMA, and shared memory transfer mode to leverage NVLink/NVSwitch/RDMA
- Heterogeneous Deployment Compatibility. Fully supports co-location and separation modes, make RL sync/async
Get started
pip install awexYou need: Python 3.10 or higher; PyTorch 2.0.0 or higher, installed with the CUDA/NPU support required by your runtime; A training or inference framework installed for the role being used (Megatron,
- Maintenance
- Active
- Last commit 41 days ago
- Stars
- 179
- On GitHub
- Licence
- Apache-2.0
- Allows business use
- Downloads for
- None listed
- No release files
- Latest version
- v0.9.0
- Released 41 days ago
- Built with
- Python
- 100% of the code
Is it looked after?
- Contributors
- Not stated
- Last 90 days
- Issues
- 58% closed
- 5 still open
- Pull requests
- 112 merged
- 1 waiting
- Releases
- 13
- 13 in the last year
- Contributing guide (opens in new tab)
- No security policy
- No code of conduct
Similar repos
| Repo | Stars | Last commit | Licence | Downloads for |
|---|---|---|---|---|
| aws / amazon-sagemaker-examples | 11,003 | 26 days ago | Apache-2.0 | No downloads |
| lasgroup / SDPO | 1,143 | 4 months ago | Apache-2.0 | No downloads |
| hiyouga / EasyR1 | 5,211 | yesterday | Apache-2.0 | No downloads |
More: languages, topics and project record
Languages
- Python 100%
- Shell <1%
Topics
- cuda
- deep-learning
- gpu
- inference
- llm
- machine-learning
- moe
- pytorch
- reasoning
- reinforcement-learning
- rl
- training
Project record
- JobFine-tuning
- StatusListed
- Commits158
- Forks23
- Watching3
- Stars in 28 days+2
- CreatedNov 17, 2025
- First seen by usNov 7, 2026
- CheckedNov 8, 2026