MegEngine / MegEngine/InferLLM
是否有计划优化GPU上的推理加速
Open
- Dominant language
- C++
- Stars
- 752
- Forks
- 94
- PR merge metrics
- No merged PRs in 30d
Description
目前社区LLM采用主流GPTQ量化之后，量化层的kernel实现基本是负向优化，是否有计划支持GPU上量化后的模型推理加速。
Contributor guide
No contributing guide indexed for this repository
Research direction
Start by reviewing the existing GPTQ quantization-layer kernel implementation and GPU inference path; the issue does not name files or tests. Benchmark quantized inference against the current implementation and define done as a measurable GPU speedup rather than a slowdown.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp
- Domain
- machine-learning, performance
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100