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

VLM Generate: tag test_static_cache_matches_dynamic as flaky (#33630)

Browse Source 
flaky
This commit is contained in:
Joao Gante
2024-10-03 12:27:02 +01:00
committed by GitHub
parent f1a5f81296
commit 6f0ce52760
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
tests/test_modeling_common.py
View File

@@ -4714,6 +4714,7 @@ class ModelTesterMixin:
normalized_1 = F.softmax(out_shared_prefix_last_tokens) normalized_1 = F.softmax(out_shared_prefix_last_tokens)
torch.testing.assert_close(normalized_0, normalized_1, rtol=1e-3, atol=1e-4) torch.testing.assert_close(normalized_0, normalized_1, rtol=1e-3, atol=1e-4)
@is_flaky(max_attempts=10) # TODO @raushan: this test is VERY flaky on some VLMs, like paligemma
def test_static_cache_matches_dynamic(self): def test_static_cache_matches_dynamic(self):
""" """
Tests that generating with static cache give almost same results as with dynamic cache. Tests that generating with static cache give almost same results as with dynamic cache.
@@ -4749,7 +4750,7 @@ class ModelTesterMixin:
output_logits=True, output_logits=True,
return_dict_in_generate=True, return_dict_in_generate=True,
) )
self.assertTrue(torch.allclose(dynamic_out.logits[0], static_out.logits[0], rtol=1e-3, atol=1e-3)) self.assertTrue(torch.allclose(dynamic_out.logits[0], static_out.logits[0], rtol=1e-3, atol=1e-4))
# For now, Let's focus only on GPU for `torch.compile` # For now, Let's focus only on GPU for `torch.compile`
@slow @slow