biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[CI/BUILD] enable intel queue for longer CPU tests (#4113)

Browse Source
This commit is contained in:
Yuan
2024-06-04 01:39:50 +08:00
committed by GitHub
parent cbb2f59cc8
commit cafb8e06c5
11 changed files with 138 additions and 90 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
tests/models/test_gptq_marlin.py
View File

@@ -22,10 +22,13 @@ os.environ["TOKENIZERS_PARALLELISM"] = "true"
MAX_MODEL_LEN = 1024
capability = torch.cuda.get_device_capability()
capability = capability[0] * 10 + capability[1]
gptq_marlin_not_supported = (
capability < QUANTIZATION_METHODS["gptq_marlin"].get_min_capability())
gptq_marlin_not_supported = True
if torch.cuda.is_available():
capability = torch.cuda.get_device_capability()
capability = capability[0] * 10 + capability[1]
gptq_marlin_not_supported = (
capability < QUANTIZATION_METHODS["gptq_marlin"].get_min_capability())
MODELS = [
# act_order==False, group_size=channelwise