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

162
tests/entrypoints/openai/test_tokenization.py
View File

@@ -53,19 +53,20 @@ async def test_tokenize_completions(
model_name: str,
tokenizer_name: str,
):
tokenizer = get_tokenizer(tokenizer_name=tokenizer_name,
tokenizer_mode="fast")
tokenizer = get_tokenizer(tokenizer_name=tokenizer_name, tokenizer_mode="fast")
for add_special in [False, True]:
prompt = "vllm1 This is a test prompt."
tokens = tokenizer.encode(prompt, add_special_tokens=add_special)
response = requests.post(server.url_for("tokenize"),
json={
"add_special_tokens": add_special,
"model": model_name,
"prompt": prompt
})
response = requests.post(
server.url_for("tokenize"),
json={
"add_special_tokens": add_special,
"model": model_name,
"prompt": prompt,
},
)
response.raise_for_status()
result = response.json()
@@ -86,48 +87,39 @@ async def test_tokenize_chat(
model_name: str,
tokenizer_name: str,
):
tokenizer = get_tokenizer(tokenizer_name=tokenizer_name,
tokenizer_mode="fast")
tokenizer = get_tokenizer(tokenizer_name=tokenizer_name, tokenizer_mode="fast")
for add_generation in [False, True]:
for add_special in [False, True]:
conversation = [{
"role": "user",
"content": "Hi there!"
}, {
"role": "assistant",
"content": "Nice to meet you!"
}, {
"role": "user",
"content": "Can I ask a question? vllm1"
}]
conversation = [
{"role": "user", "content": "Hi there!"},
{"role": "assistant", "content": "Nice to meet you!"},
{"role": "user", "content": "Can I ask a question? vllm1"},
]
for continue_final in [False, True]:
if add_generation and continue_final:
continue
if continue_final:
conversation.append({
"role": "assistant",
"content": "Sure,"
})
conversation.append({"role": "assistant", "content": "Sure,"})
prompt = tokenizer.apply_chat_template(
add_generation_prompt=add_generation,
continue_final_message=continue_final,
conversation=conversation,
tokenize=False)
tokens = tokenizer.encode(prompt,
add_special_tokens=add_special)
tokenize=False,
)
tokens = tokenizer.encode(prompt, add_special_tokens=add_special)
response = requests.post(server.url_for("tokenize"),
json={
"add_generation_prompt":
add_generation,
"continue_final_message":
continue_final,
"add_special_tokens": add_special,
"messages": conversation,
"model": model_name
})
response = requests.post(
server.url_for("tokenize"),
json={
"add_generation_prompt": add_generation,
"continue_final_message": continue_final,
"add_special_tokens": add_special,
"messages": conversation,
"model": model_name,
},
)
response.raise_for_status()
result = response.json()
@@ -148,41 +140,35 @@ async def test_tokenize_chat_with_tools(
model_name: str,
tokenizer_name: str,
):
tokenizer = get_tokenizer(tokenizer_name=tokenizer_name,
tokenizer_mode="fast")
tokenizer = get_tokenizer(tokenizer_name=tokenizer_name, tokenizer_mode="fast")
for add_generation in [False, True]:
for add_special in [False, True]:
conversation = [{
"role":
"user",
"content":
"What's the weather like in Paris today?",
}]
conversation = [
{
"role": "user",
"content": "What's the weather like in Paris today?",
}
]
tools = [{
"type": "function",
"function": {
"name": "get_weather",
"parameters": {
"type": "object",
"properties": {
"location": {
"type": "string"
}
tools = [
{
"type": "function",
"function": {
"name": "get_weather",
"parameters": {
"type": "object",
"properties": {"location": {"type": "string"}},
},
},
},
}]
}
]
for continue_final in [False, True]:
if add_generation and continue_final:
continue
if continue_final:
conversation.append({
"role": "assistant",
"content": "Sure,"
})
conversation.append({"role": "assistant", "content": "Sure,"})
prompt = tokenizer.apply_chat_template(
add_generation_prompt=add_generation,
@@ -191,8 +177,7 @@ async def test_tokenize_chat_with_tools(
tools=tools,
tokenize=False,
)
tokens = tokenizer.encode(prompt,
add_special_tokens=add_special)
tokens = tokenizer.encode(prompt, add_special_tokens=add_special)
response = requests.post(
server.url_for("tokenize"),
@@ -225,17 +210,12 @@ async def test_tokenize_with_return_token_strs(
model_name: str,
tokenizer_name: str,
):
tokenizer = get_tokenizer(tokenizer_name=tokenizer_name,
tokenizer_mode="fast")
tokenizer = get_tokenizer(tokenizer_name=tokenizer_name, tokenizer_mode="fast")
prompt = "This is a token_strs test prompt! vllm1"
response = requests.post(
server.url_for("tokenize"),
json={
"prompt": prompt,
"model": model_name,
"return_token_strs": True
},
json={"prompt": prompt, "model": model_name, "return_token_strs": True},
)
response.raise_for_status()
@@ -260,17 +240,14 @@ async def test_detokenize(
model_name: str,
tokenizer_name: str,
):
tokenizer = get_tokenizer(tokenizer_name=tokenizer_name,
tokenizer_mode="fast")
tokenizer = get_tokenizer(tokenizer_name=tokenizer_name, tokenizer_mode="fast")
prompt = "This is a test prompt. vllm1"
tokens = tokenizer.encode(prompt, add_special_tokens=False)
response = requests.post(server.url_for("detokenize"),
json={
"model": model_name,
"tokens": tokens
})
response = requests.post(
server.url_for("detokenize"), json={"model": model_name, "tokens": tokens}
)
response.raise_for_status()
assert response.json() == {"prompt": prompt}
@@ -319,14 +296,15 @@ async def test_tokenizer_info_schema(server: RemoteOpenAIServer):
}
for field, expected_type in field_types.items():
if field in result and result[field] is not None:
assert isinstance(
result[field],
expected_type), (f"{field} should be {expected_type.__name__}")
assert isinstance(result[field], expected_type), (
f"{field} should be {expected_type.__name__}"
)
@pytest.mark.asyncio
async def test_tokenizer_info_added_tokens_structure(
server: RemoteOpenAIServer, ):
server: RemoteOpenAIServer,
):
"""Test added_tokens_decoder structure if present."""
response = requests.get(server.url_for("tokenizer_info"))
response.raise_for_status()
@@ -337,25 +315,23 @@ async def test_tokenizer_info_added_tokens_structure(
assert isinstance(token_id, str), "Token IDs should be strings"
assert isinstance(token_info, dict), "Token info should be a dict"
assert "content" in token_info, "Token info should have content"
assert "special" in token_info, (
"Token info should have special flag")
assert isinstance(token_info["special"],
bool), ("Special flag should be boolean")
assert "special" in token_info, "Token info should have special flag"
assert isinstance(token_info["special"], bool), (
"Special flag should be boolean"
)
@pytest.mark.asyncio
async def test_tokenizer_info_consistency_with_tokenize(
server: RemoteOpenAIServer, ):
server: RemoteOpenAIServer,
):
"""Test that tokenizer info is consistent with tokenization endpoint."""
info_response = requests.get(server.url_for("tokenizer_info"))
info_response.raise_for_status()
info = info_response.json()
tokenize_response = requests.post(
server.url_for("tokenize"),
json={
"model": MODEL_NAME,
"prompt": "Hello world!"
},
json={"model": MODEL_NAME, "prompt": "Hello world!"},
)
tokenize_response.raise_for_status()
tokenize_result = tokenize_response.json()
@@ -363,7 +339,8 @@ async def test_tokenizer_info_consistency_with_tokenize(
tokenize_max_len = tokenize_result.get("max_model_len")
if info_max_len and tokenize_max_len:
assert info_max_len >= tokenize_max_len, (
"Info max length should be >= tokenize max length")
"Info max length should be >= tokenize max length"
)
@pytest.mark.asyncio
@@ -374,6 +351,5 @@ async def test_tokenizer_info_chat_template(server: RemoteOpenAIServer):
result = response.json()
chat_template = result.get("chat_template")
if chat_template:
assert isinstance(chat_template,
str), ("Chat template should be a string")
assert isinstance(chat_template, str), "Chat template should be a string"
assert chat_template.strip(), "Chat template should not be empty"