riaz.somc/transformers
1
0
Fork 0
mirror of https://github.com/huggingface/transformers.git synced 2025-07-03 21:00:08 +06:00
Code Issues Actions 30 Packages Projects Releases Wiki Activity

[CI] remove redundant checks in test_eager_matches_sdpa_inference (#36740)

Browse Source
This commit is contained in:
Joao Gante 2025-03-17 16:29:18 +00:00 committed by GitHub
parent e3af4fec91
commit cff4caa0c1
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 5 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
tests/test_modeling_common.py
View File

@ -138,16 +138,16 @@ TEST_EAGER_MATCHES_SDPA_INFERENCE_PARAMETERIZATION = [
(
# test name for the test runner
f"{dtype}_pad_{padding_side}{'' if use_attention_mask else '_no_attn_mask'}"
f"{'_output_attn' if output_attentions else ''}{'_sdpa_kernels' if enable_kernels else ''}",
f"{'_sdpa_kernels' if enable_kernels else ''}",
# parameterization
*(dtype, padding_side, use_attention_mask, output_attentions, enable_kernels),
*(dtype, padding_side, use_attention_mask, False, enable_kernels),
)
for dtype in ("fp16", "fp32", "bf16")
for padding_side in ("left", "right")
for use_attention_mask in (True, False)
for output_attentions in (True, False)
for enable_kernels in (True, False)
]
# Extra test case: `output_attentions=True` has special attention mask handling and sdpa reverts to eager
] + [("fp32_pad_left_output_attentions", "fp32", "left", True, True, False)]
def _config_zero_init(config):
@ -3618,7 +3618,7 @@ class ModelTesterMixin:
("cuda", False, torch.bfloat16): 1e-2,
("cuda", False, torch.float16): 5e-3,
("cuda", True, torch.float32): 1e-4,
("cuda", True, torch.bfloat16): 3e-2,
("cuda", True, torch.bfloat16): 3e-2, # (different from others)
("cuda", True, torch.float16): 5e-3,
}