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

[Core] Remove tokenizer group in vLLM (#24078)

Browse Source 
Signed-off-by: Zhuohan Li <zhuohan123@gmail.com>
This commit is contained in:
Zhuohan Li
2025-09-17 01:42:59 -07:00
committed by GitHub
parent c15309a730
commit 6c47f6bfa4
49 changed files with 276 additions and 934 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
tests/entrypoints/openai/test_completion.py
View File

@@ -27,7 +27,7 @@ GUIDED_DECODING_BACKENDS = ["outlines", "xgrammar", "guidance"]
@pytest.fixture(scope="module")
def default_server_args(zephyr_lora_files, zephyr_lora_added_tokens_files):
def default_server_args(zephyr_lora_files):
return [
# use half precision for speed and memory savings in CI environment
"--dtype",
@@ -41,7 +41,6 @@ def default_server_args(zephyr_lora_files, zephyr_lora_added_tokens_files):
"--enable-lora",
"--lora-modules",
f"zephyr-lora={zephyr_lora_files}",
f"zephyr-lora2={zephyr_lora_added_tokens_files}",
"--max-lora-rank",
"64",
"--max-cpu-loras",
@@ -87,7 +86,7 @@ async def client(server):
@pytest.mark.parametrize(
# first test base model, then test loras
"model_name",
[MODEL_NAME, "zephyr-lora", "zephyr-lora2"],
[MODEL_NAME, "zephyr-lora"],
)
async def test_single_completion(client: openai.AsyncOpenAI, model_name: str):
completion = await client.completions.create(model=model_name,
@@ -115,20 +114,6 @@ async def test_single_completion(client: openai.AsyncOpenAI, model_name: str):
assert completion.choices[0].prompt_logprobs is None
@pytest.mark.asyncio
async def test_added_lora_tokens(client: openai.AsyncOpenAI):
# test using token IDs
completion = await client.completions.create(
model="zephyr-lora2",
prompt=[0, 0, 32000, 32001, 32002],
echo=True,
max_tokens=5,
temperature=0.0,
)
# Added tokens should appear in tokenized prompt
assert completion.choices[0].text.startswith("<unk><unk>vllm1vllm2vllm3")
@pytest.mark.asyncio
async def test_added_lora_tokens_base_model(client: openai.AsyncOpenAI):
# test using token IDs
@@ -147,7 +132,7 @@ async def test_added_lora_tokens_base_model(client: openai.AsyncOpenAI):
@pytest.mark.parametrize(
# first test base model, then test loras
"model_name",
[MODEL_NAME, "zephyr-lora", "zephyr-lora2"],
[MODEL_NAME, "zephyr-lora"],
)
async def test_no_logprobs(client: openai.AsyncOpenAI, model_name: str):
# test using token IDs
@@ -713,7 +698,7 @@ async def test_guided_grammar(client: openai.AsyncOpenAI,
@pytest.mark.parametrize(
# first test base model, then test loras
"model_name",
[MODEL_NAME, "zephyr-lora", "zephyr-lora2"],
[MODEL_NAME, "zephyr-lora"],
)
@pytest.mark.parametrize("logprobs_arg", [1, 0])
async def test_echo_logprob_completion(client: openai.AsyncOpenAI,