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

switch to device agnostic device calling for test cases (#38247)

Browse Source 
* use device agnostic APIs in test cases

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

* fix style

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

* add one more

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

* xpu now supports integer device id, aligning to CUDA behaviors

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

* update to use device_properties

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

* fix style

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

* update comment

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

* fix comments

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

* fix style

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

---------

Signed-off-by: Matrix Yao <matrix.yao@intel.com>
Signed-off-by: YAO Matrix <matrix.yao@intel.com>
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
This commit is contained in:
Yao Matrix
2025-05-26 16:18:53 +08:00
committed by GitHub
parent cba279f46c
commit a5a0c7b888
39 changed files with 259 additions and 389 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
tests/test_modeling_common.py
View File

@@ -73,7 +73,10 @@ from transformers.models.auto.modeling_auto import (
)
from transformers.testing_utils import (
CaptureLogger,
backend_device_count,
backend_empty_cache,
backend_memory_allocated,
backend_torch_accelerator_module,
get_device_properties,
hub_retry,
is_flaky,
@@ -2613,7 +2616,7 @@ class ModelTesterMixin:
for k in blacklist_non_batched_params:
inputs_dict.pop(k, None)
# move input tensors to cuda:O
# move input tensors to accelerator O
for k, v in inputs_dict.items():
if torch.is_tensor(v):
inputs_dict[k] = v.to(0)
@@ -2636,12 +2639,12 @@ class ModelTesterMixin:
# a candidate for testing_utils
def get_current_gpu_memory_use():
"""returns a list of cuda memory allocations per GPU in MBs"""
"""returns a list of VRAM allocations per GPU in MBs"""
per_device_memory = []
for id in range(torch.cuda.device_count()):
with torch.cuda.device(id):
per_device_memory.append(torch.cuda.memory_allocated() >> 20)
for id in range(backend_device_count(torch_device)):
with backend_torch_accelerator_module(torch_device).device(id):
per_device_memory.append(backend_memory_allocated(torch_device) >> 20)
return per_device_memory
@@ -2657,7 +2660,7 @@ class ModelTesterMixin:
# Put model on device 0 and take a memory snapshot
model = model_class(config)
model.to("cuda:0")
model.to(f"{torch_device}:0")
memory_after_model_load = get_current_gpu_memory_use()
# The memory use on device 0 should be higher than it was initially.
@@ -2717,7 +2720,7 @@ class ModelTesterMixin:
model.parallelize()
parallel_output = model(**cast_to_device(inputs_dict, "cuda:0"))
parallel_output = model(**cast_to_device(inputs_dict, f"{torch_device}:0"))
for value, parallel_value in zip(output, parallel_output):
if isinstance(value, torch.Tensor):
@@ -4240,10 +4243,10 @@ class ModelTesterMixin:
# add position_ids + fa_kwargs
data_collator = DataCollatorWithFlattening(return_tensors="pt", return_flash_attn_kwargs=True)
batch = data_collator(features)
batch_cuda = {k: t.cuda() if torch.is_tensor(t) else t for k, t in batch.items()}
batch_accelerator = {k: t.to(torch_device) if torch.is_tensor(t) else t for k, t in batch.items()}
res_padded = model(**inputs_dict)
res_padfree = model(**batch_cuda)
res_padfree = model(**batch_accelerator)
logits_padded = res_padded.logits[inputs_dict["attention_mask"].bool()]
logits_padfree = res_padfree.logits[0]