Skip to content

PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度

fix backend issue (#1209)

Wenhua Cheng95ffb8d3e812
diff --git a/auto_round/inference/backend.py b/auto_round/inference/backend.py
index 3b243bff0..7b1783998 100644
--- a/auto_round/inference/backend.py
+++ b/auto_round/inference/backend.py
@@ -582,7 +582,7 @@ BackendInfos["ipex_awq_cpu"] = BackendInfo(
BackendInfos["ipex_awq"] = BackendInfo(
- device=["cpu"],
+ device=["xpu"],
sym=[True, False],
packing_format=AWQ_FORMAT,
bits=[4],
diff --git a/test/test_cuda/test_auto_round_format.py b/test/test_cuda/test_auto_round_format.py
index 9604ffff1..821d45fa1 100644
--- a/test/test_cuda/test_auto_round_format.py
+++ b/test/test_cuda/test_auto_round_format.py
@@ -158,7 +158,7 @@ class TestAutoRound:
from transformers import AutoRoundConfig
- quantization_config = AutoRoundConfig(backend="ipex_gptq")
+ quantization_config = AutoRoundConfig(backend="ipex")
model = AutoModelForCausalLM.from_pretrained(
quantized_model_path, device_map="cpu", trust_remote_code=True, quantization_config=quantization_config