volcengine/veScale
PyTorch distributed training library for large language models and reinforcement learning at hyperscale.
Not currently ranked — collecting fresh signals.
star history
veScale is Byted’s internal PyTorch Distributed library designed for hyperscale distributed training of large language models and reinforcement learning systems. It provides components like RaggedShard DTensor for efficient tensor sharding across distributed compute resources. The project open-sources select portions of its infrastructure for community benefit.
Frequently asked
- What is volcengine/veScale?
- PyTorch distributed training library for large language models and reinforcement learning at hyperscale.
- Is veScale open source?
- Yes — volcengine/veScale is open source, released under the Apache-2.0 license.
- What language is veScale written in?
- volcengine/veScale is primarily written in Python.
- How popular is veScale?
- volcengine/veScale has 1k stars on GitHub.
- Where can I find veScale?
- volcengine/veScale is on GitHub at https://github.com/volcengine/veScale.