[Feature]: FP8 for DS-R1
Open
Nobody has claimed this yet.
AutoDeploy
Low Precision
- Dominant language
- Python
- Stars
- 14.7k
- Forks
- 2.8k
- Avg merge
- 2d 23h
- Merged PRs (30d)
- 489
Description
🚀 The feature, motivation and pitch
- Block-wise GEMM quantization
- MLA + FP8 KV-cache https://github.com/NVIDIA/TensorRT-LLM/blob/c4abca323e2662138fa3de47e22e78709e4d3b6e/examples/models/core/deepseek_v3/README.md?plain=1#L748
- Note! Different kernel sources for Hopper and Blackwell
Alternatives
No response
Additional context
No response
Before submitting a new issue...
- Make sure you already searched for relevant issues, and checked the documentation and examples for answers to frequently asked questions.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with examples/models/core/deepseek_v3/README.md at the linked MLA + FP8 KV-cache section. Then identify the existing entry points for block-wise GEMM quantization and the Hopper and Blackwell kernel sources. Done means FP8 support covers block-wise GEMM quantization and MLA with an FP8 KV cache on both GPU generations.
Written by the indexing model from the issue text.
Assessment
- Domain
- ai, machine-learning, performance
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100