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

enable cpu offloading for Bark on xpu (#37599)

Browse Source 
* enable cpu offloading of bark modeling on XPU

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

* remove debug print

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

* fix style

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

* fix review comments

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

* enhance test

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

* update

* add deprecate message

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

* update

* update

* trigger CI

---------

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
2025-04-23 17:37:15 +08:00
committed by GitHub
parent 4f9893cbbc
commit 12f65ee752
4 changed files with 55 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
tests/models/bark/test_modeling_bark.py
View File

@@ -36,6 +36,7 @@ from transformers.models.bark.generation_configuration_bark import (
from transformers.testing_utils import (
require_flash_attn,
require_torch,
require_torch_accelerator,
require_torch_fp16,
require_torch_gpu,
slow,
@@ -1056,7 +1057,8 @@ class BarkModelIntegrationTests(unittest.TestCase):
def inputs(self):
input_ids = self.processor("In the light of the moon, a little egg lay on a leaf", voice_preset="en_speaker_6")
input_ids = input_ids.to(torch_device)
for k, v in input_ids.items():
input_ids[k] = v.to(torch_device)
return input_ids
@@ -1295,7 +1297,7 @@ class BarkModelIntegrationTests(unittest.TestCase):
len(output_ids_with_min_eos_p[0, :].tolist()), len(output_ids_without_min_eos_p[0, :].tolist())
)
@require_torch_gpu
@require_torch_accelerator
@slow
def test_generate_end_to_end_with_offload(self):
input_ids = self.inputs
@@ -1304,15 +1306,17 @@ class BarkModelIntegrationTests(unittest.TestCase):
# standard generation
output_with_no_offload = self.model.generate(**input_ids, do_sample=False, temperature=1.0)
torch.cuda.empty_cache()
torch_accelerator_module = getattr(torch, torch_device, torch.cuda)
memory_before_offload = torch.cuda.memory_allocated()
torch_accelerator_module.empty_cache()
memory_before_offload = torch_accelerator_module.memory_allocated()
model_memory_footprint = self.model.get_memory_footprint()
# activate cpu offload
self.model.enable_cpu_offload()
memory_after_offload = torch.cuda.memory_allocated()
memory_after_offload = torch_accelerator_module.memory_allocated()
# checks if the model have been offloaded