A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 01:00 UTC
Fix/remove some broken model executor tests (#21224)
Signed-off-by: Rabi Mishra <ramishra@redhat.com>
Rabi Mishra committed
Jul 19, 2025 at 17:45 UTC
c81259d33a77f657bce9bd8ab0e3548826df258d
2 files changed
+3
-16
tests/model_executor/test_guided_processors.py
-13
@@ -189,19 +189,6 @@ def test_multiple_guided_options_not_allowed(sample_json_schema, sample_regex):
189
GuidedDecodingParams(json=sample_json_schema, grammar="test grammar")
190
191
192
-def test_guided_decoding_backend_options():
193
- """Test backend-specific options"""
194
- with pytest.warns(DeprecationWarning):
195
- guided_decoding_params = GuidedDecodingParams(
196
- backend=
197
- "xgrammar:no-fallback,disable-any-whitespace,no-additional-properties"
198
- )
199
- assert guided_decoding_params.backend == "xgrammar"
200
- assert guided_decoding_params.disable_fallback
201
- assert guided_decoding_params.disable_any_whitespace
202
- assert guided_decoding_params.disable_additional_properties
203
-
204
-
192
def test_pickle_xgrammar_tokenizer_data():
193
try:
194
import xgrammar as xgr
tests/model_executor/test_model_load_with_params.py
+3
-3
@@ -49,7 +49,7 @@ def test_model_loading_with_params(vllm_runner):
49
50
def check_model(model):
51
assert isinstance(model, BertEmbeddingModel)
52
- assert isinstance(model._pooler, CLSPool)
52
+ assert isinstance(model.pooler.pooling, CLSPool)
53
54
vllm_model.apply_model(check_model)
55
@@ -87,7 +87,7 @@ def test_roberta_model_loading_with_params(vllm_runner):
87
88
def check_model(model):
89
assert isinstance(model, RobertaEmbeddingModel)
90
- assert isinstance(model._pooler, MeanPool)
90
+ assert isinstance(model.pooler.pooling, MeanPool)
91
92
vllm_model.apply_model(check_model)
93
@@ -114,7 +114,7 @@ def test_facebook_roberta_model_loading_with_params(vllm_runner):
114
def check_model(model):
115
assert isinstance(model, RobertaEmbeddingModel)
116
assert not hasattr(model, "lm_head")
117
- assert isinstance(model._pooler, CLSPool)
117
+ assert isinstance(model.pooler.pooling, CLSPool)
118
119
vllm_model.apply_model(check_model)
120