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 14, 05:30 UTC
[CI Bugfix] Fix failing test_model_load_with_params tests due to tokenizer refactor (#25086)
Signed-off-by: mgoin <mgoin64@gmail.com>
Michael Goin committed
Sep 17, 2025 at 14:15 UTC
e3db5ebb66590031ecfd3338de41b6f1ee95bf2a
1 file changed
+5
-6
tests/model_executor/test_model_load_with_params.py
+5
-6
@@ -47,8 +47,8 @@ def test_model_loading_with_params(vllm_runner, monkeypatch):
47
assert model_config.pooler_config.normalize
48
49
# asserts on the tokenizer loaded
50
- assert model_tokenizer.tokenizer_id == "BAAI/bge-base-en-v1.5"
51
- assert model_tokenizer.tokenizer.model_max_length == 512
50
+ assert model_config.tokenizer == "BAAI/bge-base-en-v1.5"
51
+ assert model_tokenizer.model_max_length == 512
52
53
def check_model(model):
54
assert isinstance(model, BertEmbeddingModel)
@@ -87,8 +87,8 @@ def test_roberta_model_loading_with_params(vllm_runner, monkeypatch):
87
assert model_config.pooler_config.normalize
88
89
# asserts on the tokenizer loaded
90
- assert model_tokenizer.tokenizer_id == "intfloat/multilingual-e5-base"
91
- assert model_tokenizer.tokenizer.model_max_length == 512
90
+ assert model_config.tokenizer == "intfloat/multilingual-e5-base"
91
+ assert model_tokenizer.model_max_length == 512
92
93
def check_model(model):
94
assert isinstance(model, RobertaEmbeddingModel)
@@ -116,8 +116,7 @@ def test_facebook_roberta_model_loading_with_params(vllm_runner, monkeypatch):
116
output = vllm_model.embed("Write a short story about a robot that"
117
" dreams for the first time.\n")
118
119
- model_tokenizer = vllm_model.llm.llm_engine.tokenizer
120
- assert model_tokenizer.tokenizer_id == model_name
119
+ assert vllm_model.llm.llm_engine.model_config.tokenizer == model_name
120
121
def check_model(model):
122
assert isinstance(model, RobertaEmbeddingModel)