xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 03:15:20 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bugfix] Temporarily disable gptq_bitblas on ROCm (#17411)

Browse Source 
Signed-off-by: Yan Cangang <nalanzeyu@gmail.com>
This commit is contained in:
NaLan ZeYu 2025-05-01 10:51:45 +08:00 committed by GitHub
parent 08fb5587b4
commit 1144a8efe7
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 6 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
docs/source/features/quantization/supported_hardware.md
View File

@ -80,7 +80,7 @@ The table below shows the compatibility of various quantization implementations
* ✅︎
* ✅︎
* ✅︎
* ✅︎
* ❌
* ❌
* ❌
* ❌

5
vllm/model_executor/layers/quantization/gptq_bitblas.py
View File

@ -25,6 +25,7 @@ from vllm.model_executor.parameter import (ChannelQuantScaleParameter,
PackedColumnParameter,
PackedvLLMParameter,
RowvLLMParameter)
from vllm.platforms import current_platform
from vllm.scalar_type import scalar_types
logger = init_logger(__name__)
@ -191,6 +192,10 @@ class GPTQBitBLASConfig(QuantizationConfig):
sym = quant_config.get("sym")
desc_act = quant_config.get("desc_act")
# temporarily disable on ROCm platform
if not current_platform.is_cuda():
return False
# If we cannot find the info needed in the config, cannot convert.
if (num_bits is None or group_size is None or sym is None
or desc_act is None):