riaz.somc/transformers
1
0
Fork 0
mirror of https://github.com/huggingface/transformers.git synced 2025-07-31 02:02:21 +06:00
Code Issues Actions 23 Packages Projects Releases Wiki Activity

enable trainer test cases on xpu (#38138)

Browse Source 
* enable trainer test cases on xpu

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>
This commit is contained in:
Yao Matrix 2025-05-15 20:17:44 +08:00 committed by GitHub
parent b11b28cc4e
commit 7caa57e85e
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
4 changed files with 13 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/transformers/testing_utils.py
View File

@ -2039,7 +2039,7 @@ class TestCasePlus(unittest.TestCase):
"""
env = os.environ.copy()
paths = [self.src_dir_str]
paths = [self.repo_root_dir_str, self.src_dir_str]
if "/examples" in self.test_file_dir_str:
paths.append(self.examples_dir_str)
else:

3
tests/trainer/test_trainer.py
View File

@ -97,7 +97,6 @@ from transformers.testing_utils import (
require_torch_fp16,
require_torch_gpu,
require_torch_multi_accelerator,
require_torch_multi_gpu,
require_torch_non_multi_accelerator,
require_torch_non_multi_gpu,
require_torch_tensorrt_fx,
@ -3766,7 +3765,7 @@ class TrainerIntegrationTest(TestCasePlus, TrainerIntegrationCommon):
train_output = trainer.train()
self.assertEqual(train_output.global_step, int(self.n_epochs))
@require_torch_multi_gpu
@require_torch_multi_accelerator
def test_num_batches_in_training_with_gradient_accumulation(self):
with tempfile.TemporaryDirectory() as tmp_dir:
for num_train_epochs in [1, 2]:

9
tests/trainer/test_trainer_distributed_loss.py
View File

@ -1,7 +1,6 @@
import json
import datasets
import torch
from tests.trainer.test_trainer import StoreLossCallback
from transformers import (
@ -15,16 +14,18 @@ from transformers import (
)
from transformers.testing_utils import (
TestCasePlus,
backend_device_count,
execute_subprocess_async,
get_torch_dist_unique_port,
require_torch_multi_gpu,
require_torch_multi_accelerator,
torch_device,
)
class TestTrainerDistributedLoss(TestCasePlus):
@require_torch_multi_gpu
@require_torch_multi_accelerator
def test_trainer(self):
device_count = torch.cuda.device_count()
device_count = backend_device_count(torch_device)
min_bs = 1
output_dir = self.get_auto_remove_tmp_dir()
for gpu_num, enable, bs, name in (

10
tests/trainer/test_trainer_distributed_worker_seed.py
View File

@ -14,9 +14,11 @@ from transformers import (
)
from transformers.testing_utils import (
TestCasePlus,
backend_device_count,
execute_subprocess_async,
get_torch_dist_unique_port,
require_torch_multi_gpu,
require_torch_multi_accelerator,
torch_device,
)
@ -47,7 +49,7 @@ class DummyModel(nn.Module):
self.fc = nn.Linear(3, 1)
def forward(self, x):
local_tensor = torch.tensor(x, device="cuda")
local_tensor = torch.tensor(x, device=torch_device)
gathered = gather_from_all_gpus(local_tensor, dist.get_world_size())
assert not all(torch.allclose(t, gathered[0]) for t in gathered[1:])
y = self.fc(x)
@ -55,9 +57,9 @@ class DummyModel(nn.Module):
class TestTrainerDistributedWorkerSeed(TestCasePlus):
@require_torch_multi_gpu
@require_torch_multi_accelerator
def test_trainer(self):
device_count = torch.cuda.device_count()
device_count = backend_device_count(torch_device)
output_dir = self.get_auto_remove_tmp_dir()
distributed_args = f"""--nproc_per_node={device_count}
--master_port={get_torch_dist_unique_port()}