biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Convert formatting to use ruff instead of yapf + isort (#26247)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-05 15:06:22 +01:00
committed by GitHub
parent 17edd8a807
commit d6953beb91
1508 changed files with 115244 additions and 94146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

83
tests/models/language/pooling/test_nomic_max_model_len.py
View File

@@ -7,10 +7,10 @@ from ...utils import EmbedModelInfo
MODELS = [
EmbedModelInfo("nomic-ai/nomic-embed-text-v1"),
#EmbedModelInfo("nomic-ai/nomic-embed-text-v1.5"),
#EmbedModelInfo("nomic-ai/CodeRankEmbed"),
# EmbedModelInfo("nomic-ai/nomic-embed-text-v1.5"),
# EmbedModelInfo("nomic-ai/CodeRankEmbed"),
EmbedModelInfo("nomic-ai/nomic-embed-text-v2-moe"),
#EmbedModelInfo("Snowflake/snowflake-arctic-embed-m-long"),
# EmbedModelInfo("Snowflake/snowflake-arctic-embed-m-long"),
]
rope_theta = 1000
@@ -21,23 +21,24 @@ max_model_len = int(original_max_position_embeddings * factor)
@pytest.mark.parametrize("model_info", MODELS)
def test_default(model_info, vllm_runner):
with vllm_runner(model_info.name, runner="pooling",
max_model_len=None) as vllm_model:
with vllm_runner(
model_info.name, runner="pooling", max_model_len=None
) as vllm_model:
model_config = vllm_model.llm.llm_engine.model_config
if model_info.name == "nomic-ai/nomic-embed-text-v2-moe":
# For nomic-embed-text-v2-moe the length is set to 512
# by sentence_bert_config.json.
assert model_config.max_model_len == 512
else:
assert (
model_config.max_model_len == original_max_position_embeddings)
assert model_config.max_model_len == original_max_position_embeddings
@pytest.mark.parametrize("model_info", MODELS)
def test_set_max_model_len_legal(model_info, vllm_runner):
# set max_model_len <= 512
with vllm_runner(model_info.name, runner="pooling",
max_model_len=256) as vllm_model:
with vllm_runner(
model_info.name, runner="pooling", max_model_len=256
) as vllm_model:
model_config = vllm_model.llm.llm_engine.model_config
assert model_config.max_model_len == 256
@@ -46,13 +47,12 @@ def test_set_max_model_len_legal(model_info, vllm_runner):
# For nomic-embed-text-v2-moe the length is set to 512
# by sentence_bert_config.json.
with pytest.raises(ValueError):
with vllm_runner(model_info.name,
runner="pooling",
max_model_len=1024):
with vllm_runner(model_info.name, runner="pooling", max_model_len=1024):
pass
else:
with vllm_runner(model_info.name, runner="pooling",
max_model_len=1024) as vllm_model:
with vllm_runner(
model_info.name, runner="pooling", max_model_len=1024
) as vllm_model:
model_config = vllm_model.llm.llm_engine.model_config
assert model_config.max_model_len == 1024
@@ -61,17 +61,18 @@ def test_set_max_model_len_legal(model_info, vllm_runner):
def test_set_max_model_len_illegal(model_info, vllm_runner):
# set max_model_len > 2048
with pytest.raises(ValueError):
with vllm_runner(model_info.name, runner="pooling",
max_model_len=4096):
with vllm_runner(model_info.name, runner="pooling", max_model_len=4096):
pass
# set max_model_len > 2048 by hf_overrides
hf_overrides = {"max_model_len": 4096}
with pytest.raises(ValueError):
with vllm_runner(model_info.name,
runner="pooling",
max_model_len=None,
hf_overrides=hf_overrides):
with vllm_runner(
model_info.name,
runner="pooling",
max_model_len=None,
hf_overrides=hf_overrides,
):
pass
@@ -82,16 +83,14 @@ def test_use_rope_scaling_legal(model_info, vllm_runner):
"rope_scaling": {
"rope_type": "yarn",
"factor": factor,
"original_max_position_embeddings":
original_max_position_embeddings
"original_max_position_embeddings": original_max_position_embeddings,
},
"max_model_len": max_model_len
"max_model_len": max_model_len,
}
with vllm_runner(model_info.name,
runner="pooling",
max_model_len=None,
hf_overrides=hf_overrides):
with vllm_runner(
model_info.name, runner="pooling", max_model_len=None, hf_overrides=hf_overrides
):
pass
@@ -102,16 +101,17 @@ def test_use_rope_scaling_illegal(model_info, vllm_runner):
"rope_scaling": {
"rope_type": "yarn",
"factor": factor,
"original_max_position_embeddings":
original_max_position_embeddings
}
"original_max_position_embeddings": original_max_position_embeddings,
},
}
# illegal max_model_len
with pytest.raises(ValueError):
with vllm_runner(model_info.name,
runner="pooling",
max_model_len=max_model_len + 1,
hf_overrides=hf_overrides):
with vllm_runner(
model_info.name,
runner="pooling",
max_model_len=max_model_len + 1,
hf_overrides=hf_overrides,
):
pass
hf_overrides = {
@@ -119,15 +119,16 @@ def test_use_rope_scaling_illegal(model_info, vllm_runner):
"rope_scaling": {
"rope_type": "yarn",
"factor": factor,
"original_max_position_embeddings":
original_max_position_embeddings
"original_max_position_embeddings": original_max_position_embeddings,
},
"max_model_len": max_model_len + 1
"max_model_len": max_model_len + 1,
}
# illegal max_model_len by hf_overrides
with pytest.raises(ValueError):
with vllm_runner(model_info.name,
runner="pooling",
max_model_len=None,
hf_overrides=hf_overrides):
with vllm_runner(
model_info.name,
runner="pooling",
max_model_len=None,
hf_overrides=hf_overrides,
):
pass