SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

fix xpu failures on PT 2.7 and 2.8 w/o IPEX and enable hqq cases on XPU (#39187)

Browse Source 
* chameleon xpu bnb groundtruth update on bnb triton backend since we are
deprecating ipex backend

Signed-off-by: YAO Matrix <matrix.yao@intel.com>

* enable hqq uts on XPU, all passed

Signed-off-by: YAO Matrix <matrix.yao@intel.com>

* fix style

Signed-off-by: YAO Matrix <matrix.yao@intel.com>

* fix comment

Signed-off-by: YAO Matrix <matrix.yao@intel.com>

---------

Signed-off-by: YAO Matrix <matrix.yao@intel.com>
This commit is contained in:
Yao Matrix
2025-07-08 16:18:26 +08:00
committed by GitHub
parent 17b3c96c00
commit b2816da802
3 changed files with 15 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
tests/quantization/hqq/test_hqq.py
View File

@@ -21,9 +21,10 @@ from transformers import AutoModelForCausalLM, AutoTokenizer, HqqConfig
from transformers.testing_utils import (
backend_empty_cache,
require_accelerate,
require_deterministic_for_xpu,
require_hqq,
require_torch_gpu,
require_torch_multi_gpu,
require_torch_accelerator,
require_torch_multi_accelerator,
slow,
torch_device,
)
@@ -87,7 +88,7 @@ def check_forward(test_module, model, batch_size=1, context_size=1024):
MODEL_ID = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
@require_torch_gpu
@require_torch_accelerator
@require_hqq
class HqqConfigTest(unittest.TestCase):
def test_to_dict(self):
@@ -101,7 +102,7 @@ class HqqConfigTest(unittest.TestCase):
@slow
@require_torch_gpu
@require_torch_accelerator
@require_accelerate
@require_hqq
class HQQTest(unittest.TestCase):
@@ -131,7 +132,6 @@ class HQQTest(unittest.TestCase):
model_id=MODEL_ID, quant_config=quant_config, compute_dtype=torch.float16, device=torch_device
)
original_device = hqq_runner.model.model.layers[0].self_attn.v_proj.device
check_hqqlayer(self, hqq_runner.model.model.layers[0].self_attn.v_proj)
check_forward(self, hqq_runner.model)
@@ -142,7 +142,7 @@ class HQQTest(unittest.TestCase):
check_hqqlayer(self, hqq_runner.model.model.layers[0].self_attn.v_proj)
check_forward(self, hqq_runner.model)
hqq_runner.model.cuda(original_device)
hqq_runner.model.to(torch_device)
check_hqqlayer(self, hqq_runner.model.model.layers[0].self_attn.v_proj)
check_forward(self, hqq_runner.model)
@@ -158,8 +158,8 @@ class HQQTest(unittest.TestCase):
@slow
@require_torch_gpu
@require_torch_multi_gpu
@require_torch_accelerator
@require_torch_multi_accelerator
@require_accelerate
@require_hqq
class HQQTestMultiGPU(unittest.TestCase):
@@ -182,7 +182,7 @@ class HQQTestMultiGPU(unittest.TestCase):
@slow
@require_torch_gpu
@require_torch_accelerator
@require_accelerate
@require_hqq
class HQQTestBias(unittest.TestCase):
@@ -202,6 +202,7 @@ class HQQTestBias(unittest.TestCase):
check_hqqlayer(self, hqq_runner.model.model.decoder.layers[0].self_attn.v_proj)
check_forward(self, hqq_runner.model)
@require_deterministic_for_xpu
def test_save_and_load_quantized_model(self):
"""
Test saving and loading a quantized model with bias
@@ -237,7 +238,7 @@ class HQQTestBias(unittest.TestCase):
@slow
@require_torch_gpu
@require_torch_accelerator
@require_accelerate
@require_hqq
class HQQSerializationTest(unittest.TestCase):