xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-03-26 23:57:09 +08:00
Code Issues Packages Projects Releases Wiki Activity

Skip RTN quantization tests on ROCm

Browse Source 
Fixes #29525

RTN (Round-To-Nearest) quantization is currently not supported on ROCm.
This fix adds a check to skip RTN quantization tests when running on ROCm
to prevent CI failures for unsupported features.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

Signed-off-by: westers <steve.westerhouse@origami-analytics.com>
This commit is contained in:
westers 2025-12-20 10:21:05 -06:00
parent 560ae9638c
commit b6520e5ec6
1 changed files with 4 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
tests/quantization/utils.py
View File

@ -10,6 +10,10 @@ def is_quant_method_supported(quant_method: str) -> bool:
if not (current_platform.is_cuda() or current_platform.is_rocm()):
return False
# RTN quantization is currently not supported on ROCm
if current_platform.is_rocm() and quant_method == "rtn":
return False
capability = current_platform.get_device_capability()
assert capability is not None