Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Fix] fix_vllm_moe_quant #342

Merged

Conversation

belovedxixi
Copy link

@belovedxixi belovedxixi commented Dec 20, 2024

Detail:
Alibaba benchmark test script quantizes a float model using vllm by giving in the parameters --quantization fp8 and --kv_cache_dtype fp8.
They don't use quark quantized model,
When using the 'moe_final_v0.6.0_Nov19' to do benchmark test for mixtral 8*7B, we got garbage output.
After checking, it was found that it was caused by a code missing in the vllm/model_executor/layers/quantization/fp8.py (For moe model,the func fuse_shuffle and moe_padding were not executed if using vllm to quantize model, and If passed to the vllm a quantized model, the two func can be executed correctly without errors ).

@gshtras gshtras merged commit d567353 into llama_fp8_12062024 Dec 20, 2024
3 of 4 checks passed
@gshtras gshtras deleted the moe_final_v0.6.0_Nov19_fix_dynamic_quant branch December 20, 2024 15:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants