Remove @slow for test_eager_matches_sdpa_inference (#34558)

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

---------

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
This commit is contained in:
Yih-Dar
2024-11-05 16:10:42 +01:00
committed by GitHub
parent 082e57e0d4
commit f2d5dfbab2
21 changed files with 271 additions and 626 deletions

View File

@@ -30,12 +30,10 @@ from transformers import (
from transformers.models.mllama.configuration_mllama import MllamaTextConfig
from transformers.testing_utils import (
cleanup,
is_flaky,
require_bitsandbytes,
require_read_token,
require_torch,
require_torch_gpu,
require_torch_sdpa,
slow,
torch_device,
)
@@ -359,13 +357,6 @@ class MllamaForConditionalGenerationModelTest(ModelTesterMixin, GenerationTester
self.assertListEqual([layer_attention.shape for layer_attention in iter_attentions], expected_shapes)
@require_torch_sdpa
@slow
@is_flaky()
def test_eager_matches_sdpa_inference_1_bfloat16(self):
# A workaround to override parametrized test with flaky decorator
super().test_eager_matches_sdpa_inference_1_bfloat16()
@unittest.skip("For some unknown reasons the tests fails in CrossAttention layer when doing torch.sdpa(). ")
def test_sdpa_can_compile_dynamic(self):
pass