fix(quantization): Fix AWQ dequantize on Intel XPU and refactor AutoAWQ config (#42727)

Signed-off-by: Alex <alex.tech.lab@outlook.com>
Signed-off-by: AlexHuang <jihuihuang@tencent.com>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
AlexHuang
2026-06-18 20:12:28 +08:00
committed by GitHub
co-authored by Claude Kunshang Ji mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent 351c72d6e5
commit 8d4f54966c
20 changed files with 580 additions and 429 deletions
+5 -3
View File
@@ -43,16 +43,18 @@ MODEL_ARG_EXPTYPES = [
("LnL-AI/TinyLlama-1.1B-Chat-v1.0-GPTQ-4bit", "gptq", "auto_gptq"),
("LnL-AI/TinyLlama-1.1B-Chat-v1.0-GPTQ-4bit", "awq", "ERROR"),
# AUTOAWQ
# AutoAWQConfig.override_quantization_method() returns "auto_awq" for AWQ models
# when user_quant is None, "awq", "awq_marlin", "marlin", or "auto_awq"
(
"TheBloke/OpenHermes-2.5-Mistral-7B-AWQ",
None,
"awq_marlin" if current_platform.is_cuda_alike() else "awq",
"auto_awq",
),
("TheBloke/OpenHermes-2.5-Mistral-7B-AWQ", "awq", "awq"),
("TheBloke/OpenHermes-2.5-Mistral-7B-AWQ", "awq", "auto_awq"),
(
"TheBloke/OpenHermes-2.5-Mistral-7B-AWQ",
"marlin",
"awq_marlin" if current_platform.is_cuda_alike() else "ERROR",
"auto_awq" if current_platform.is_cuda_alike() else "ERROR",
),
("TheBloke/OpenHermes-2.5-Mistral-7B-AWQ", "gptq", "ERROR"),
]