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

fix: providing a tensor to cache_position in model.generate kwargs always crashes because of boolean test (#39300)

Browse Source 
* fix: cache_position: RuntimeError: Boolean value of Tensor with more than one value is ambiguous

* test cache_position

* move test

* propagate changes

---------

Co-authored-by: Masataro Asai <guicho2.71828@gmail.com>
This commit is contained in:
Joao Gante
2025-07-30 18:30:28 +01:00
committed by GitHub
parent 9b3203f47b
commit 4f93cc9174
6 changed files with 150 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
tests/models/t5gemma/test_modeling_t5gemma.py
View File

@@ -29,7 +29,7 @@ from transformers.testing_utils import (
torch_device,
)
from ...generation.test_utils import GenerationTesterMixin
from ...generation.test_utils import GenerationTesterMixin, has_similar_generate_outputs
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
@@ -1196,7 +1196,7 @@ class T5GemmaModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMi
outputs_cached.scores = full_cached_scores
# The two sets of generated text and past kv should be equal to each other
self._check_similar_generate_outputs(outputs, outputs_cached)
self.assertTrue(has_similar_generate_outputs(outputs, outputs_cached))
for layer_idx in range(len(outputs_cached.past_key_values)):
for kv_idx in range(len(outputs_cached.past_key_values[layer_idx])):
self.assertTrue(