Support PLAIN_INT32 for AWQ on Intel GPU

[xiaowangintel]

Summary:
Support PLAIN_INT32 for AWQ on Intel GPU

Test:

# task is gsm8k
python example.py --repo "microsoft/Phi-4-mini-instruct" --quant awq-int4wo-128 --calibration_limit 5 --max_seq_length 4096 --device xpu

#task is mmlu
python example.py --repo "Qwen/Qwen3-8B" --quant awq-int4wo-128 --calibration_limit 1 --max_seq_length 4096 --device xpu

Result:

Task	Model	calibration_limit	awq
gsm8k	Phi-4-mini-instruct	5	0.75815
mmlu	Qwen3-8B	1	0.7595

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/3019

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit 3a325f3 with merge base 067b273 ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[jerryzh168]

add a test like

ao/test/quantization/quantize_/workflows/int4/test_int4_tensor.py

Line 223 in 18dbe87

def test_activation_prescaling(self):

[liangan1]

@xiaowangintel let us add the UT as int4_tensor.

[xiaowangintel]

done

[liangan1]

LGTM, how about the CUDA accuracy on these two models?

[xiaowangintel]

LGTM, how about the CUDA accuracy on these two models?

Please visit #2400