ByteDance-Seed / ByteDance-Seed/decoupleQ

你好,量化后推理模型出现nan的问题

Open
#13 11 comments 0 reactions 0 assignees View on GitHub
Dominant language
Cuda
Stars
153
Forks
10
PR merge metrics
No merged PRs in 30d

Description

我看了存在issue里也有量化后模型出现nan的问题,但是都是没有解决方案。
请问这个咋么解决啊?

我这边修改了llama.py部分代码适配我这边的一个传统模型,得到后的模型推理会出现nan的情况。
环境:
Cuda/NVCC: 12.2
python: 3.10
cmake: 3.22.1
gpu Driver: 535.161.08
torch:2.1.0

Contributor guide

No contributing guide indexed for this repository

Research direction

Start by reviewing the reported changes in llama.py and reproduce the post-quantization inference failure in the listed Python, PyTorch, and CUDA environment. Examine the existing issue discussion and establish a minimal reproduction; the work is done when the NaN cause is identified and inference completes without NaN values.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.