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

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

View File

@@ -7,9 +7,12 @@ from typing import Optional
import pytest
from vllm.entrypoints.openai.protocol import (ChatCompletionRequest,
DeltaMessage, FunctionCall,
ToolCall)
from vllm.entrypoints.openai.protocol import (
ChatCompletionRequest,
DeltaMessage,
FunctionCall,
ToolCall,
)
from vllm.entrypoints.openai.tool_parsers import xLAMToolParser
from vllm.transformers_utils.detokenizer_utils import detokenize_incrementally
from vllm.transformers_utils.tokenizer import AnyTokenizer, get_tokenizer
@@ -30,12 +33,14 @@ def xlam_tool_parser(xlam_tokenizer):
return xLAMToolParser(xlam_tokenizer)
def assert_tool_calls(actual_tool_calls: list[ToolCall],
expected_tool_calls: list[ToolCall]):
def assert_tool_calls(
actual_tool_calls: list[ToolCall], expected_tool_calls: list[ToolCall]
):
assert len(actual_tool_calls) == len(expected_tool_calls)
for actual_tool_call, expected_tool_call in zip(actual_tool_calls,
expected_tool_calls):
for actual_tool_call, expected_tool_call in zip(
actual_tool_calls, expected_tool_calls
):
assert isinstance(actual_tool_call.id, str)
assert len(actual_tool_call.id) > 16
@@ -49,8 +54,7 @@ def stream_delta_message_generator(
model_output: str,
request: Optional[ChatCompletionRequest] = None,
) -> Generator[DeltaMessage, None, None]:
all_token_ids = xlam_tokenizer.encode(model_output,
add_special_tokens=False)
all_token_ids = xlam_tokenizer.encode(model_output, add_special_tokens=False)
previous_text = ""
previous_tokens = None
@@ -59,18 +63,19 @@ def stream_delta_message_generator(
for i, delta_token in enumerate(all_token_ids):
delta_token_ids = [delta_token]
previous_token_ids = all_token_ids[:i]
current_token_ids = all_token_ids[:i + 1]
current_token_ids = all_token_ids[: i + 1]
(new_tokens, delta_text, new_prefix_offset,
new_read_offset) = (detokenize_incrementally(
tokenizer=xlam_tokenizer,
all_input_ids=current_token_ids,
prev_tokens=previous_tokens,
prefix_offset=prefix_offset,
read_offset=read_offset,
skip_special_tokens=False,
spaces_between_special_tokens=True,
))
(new_tokens, delta_text, new_prefix_offset, new_read_offset) = (
detokenize_incrementally(
tokenizer=xlam_tokenizer,
all_input_ids=current_token_ids,
prev_tokens=previous_tokens,
prefix_offset=prefix_offset,
read_offset=read_offset,
skip_special_tokens=False,
spaces_between_special_tokens=True,
)
)
current_text = previous_text + delta_text
@@ -87,8 +92,9 @@ def stream_delta_message_generator(
yield delta_message
previous_text = current_text
previous_tokens = (previous_tokens +
new_tokens if previous_tokens else new_tokens)
previous_tokens = (
previous_tokens + new_tokens if previous_tokens else new_tokens
)
prefix_offset = new_prefix_offset
read_offset = new_read_offset
@@ -96,7 +102,8 @@ def stream_delta_message_generator(
def test_extract_tool_calls_no_tools(xlam_tool_parser):
model_output = "This is a test"
extracted_tool_calls = xlam_tool_parser.extract_tool_calls(
model_output, request=None) # type: ignore[arg-type]
model_output, request=None
) # type: ignore[arg-type]
assert not extracted_tool_calls.tools_called
assert extracted_tool_calls.tool_calls == []
assert extracted_tool_calls.content == model_output
@@ -115,87 +122,113 @@ def test_extract_tool_calls_no_tools(xlam_tool_parser):
(
"""[{"name": "get_current_weather", "arguments": {"city": "Dallas", "state": "TX", "unit": "fahrenheit"}}, {"name": "get_current_weather", "arguments": {"city": "Orlando", "state": "FL", "unit": "fahrenheit"}}]""", # noqa: E501
[
ToolCall(function=FunctionCall(
name="get_current_weather",
arguments=json.dumps({
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}),
)),
ToolCall(function=FunctionCall(
name="get_current_weather",
arguments=json.dumps({
"city": "Orlando",
"state": "FL",
"unit": "fahrenheit",
}),
)),
ToolCall(
function=FunctionCall(
name="get_current_weather",
arguments=json.dumps(
{
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}
),
)
),
ToolCall(
function=FunctionCall(
name="get_current_weather",
arguments=json.dumps(
{
"city": "Orlando",
"state": "FL",
"unit": "fahrenheit",
}
),
)
),
],
None,
),
(
"""<think>I'll help you with that.</think>[{"name": "get_current_weather", "arguments": {"city": "Dallas", "state": "TX", "unit": "fahrenheit"}}]""", # noqa: E501
[
ToolCall(function=FunctionCall(
name="get_current_weather",
arguments=json.dumps({
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}),
))
ToolCall(
function=FunctionCall(
name="get_current_weather",
arguments=json.dumps(
{
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}
),
)
)
],
"<think>I'll help you with that.</think>",
),
(
"""I'll help you with that.\n```json\n[{"name": "get_current_weather", "arguments": {"city": "Dallas", "state": "TX", "unit": "fahrenheit"}}]\n```""", # noqa: E501
[
ToolCall(function=FunctionCall(
name="get_current_weather",
arguments=json.dumps({
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}),
))
ToolCall(
function=FunctionCall(
name="get_current_weather",
arguments=json.dumps(
{
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}
),
)
)
],
"I'll help you with that.",
),
(
"""I'll check the weather for you.[TOOL_CALLS][{"name": "get_current_weather", "arguments": {"city": "Dallas", "state": "TX", "unit": "fahrenheit"}}]""", # noqa: E501
[
ToolCall(function=FunctionCall(
name="get_current_weather",
arguments=json.dumps({
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}),
))
ToolCall(
function=FunctionCall(
name="get_current_weather",
arguments=json.dumps(
{
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}
),
)
)
],
"I'll check the weather for you.",
),
(
"""I'll help you check the weather.<tool_call>[{"name": "get_current_weather", "arguments": {"city": "Dallas", "state": "TX", "unit": "fahrenheit"}}]</tool_call>""", # noqa: E501
[
ToolCall(function=FunctionCall(
name="get_current_weather",
arguments=json.dumps({
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}),
))
ToolCall(
function=FunctionCall(
name="get_current_weather",
arguments=json.dumps(
{
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}
),
)
)
],
"I'll help you check the weather.",
),
],
)
def test_extract_tool_calls(xlam_tool_parser, model_output,
expected_tool_calls, expected_content):
def test_extract_tool_calls(
xlam_tool_parser, model_output, expected_tool_calls, expected_content
):
extracted_tool_calls = xlam_tool_parser.extract_tool_calls(
model_output, request=None) # type: ignore[arg-type]
model_output, request=None
) # type: ignore[arg-type]
assert extracted_tool_calls.tools_called
assert_tool_calls(extracted_tool_calls.tool_calls, expected_tool_calls)
@@ -210,25 +243,30 @@ def test_extract_tool_calls(xlam_tool_parser, model_output,
(
"""[{"name": "get_current_weather", "arguments": {"city": "Seattle", "state": "WA", "unit": "celsius"}}]""", # noqa: E501
[
ToolCall(function=FunctionCall(
name="get_current_weather",
arguments=json.dumps({
"city": "Seattle",
"state": "WA",
"unit": "celsius",
}),
))
ToolCall(
function=FunctionCall(
name="get_current_weather",
arguments=json.dumps(
{
"city": "Seattle",
"state": "WA",
"unit": "celsius",
}
),
)
)
],
None,
),
],
)
def test_extract_tool_calls_list_structure(xlam_tool_parser, model_output,
expected_tool_calls,
expected_content):
def test_extract_tool_calls_list_structure(
xlam_tool_parser, model_output, expected_tool_calls, expected_content
):
"""Test extraction of tool calls when the model outputs a list-structured tool call.""" # noqa: E501
extracted_tool_calls = xlam_tool_parser.extract_tool_calls(
model_output, request=None) # type: ignore[arg-type]
model_output, request=None
) # type: ignore[arg-type]
assert extracted_tool_calls.tools_called
assert_tool_calls(extracted_tool_calls.tool_calls, expected_tool_calls)
@@ -239,20 +277,25 @@ def test_extract_tool_calls_list_structure(xlam_tool_parser, model_output,
# Test for preprocess_model_output method
def test_preprocess_model_output(xlam_tool_parser):
# Test with list structure
model_output = """[{"name": "get_current_weather", "arguments": {"city": "Seattle"}}]""" # noqa: E501
model_output = (
"""[{"name": "get_current_weather", "arguments": {"city": "Seattle"}}]""" # noqa: E501
)
content, potential_tool_calls = xlam_tool_parser.preprocess_model_output(
model_output)
model_output
)
assert content is None
assert potential_tool_calls == model_output
# Test with thinking tag
model_output = """<think>I'll help you with that.</think>[{"name": "get_current_weather", "arguments": {"city": "Seattle"}}]""" # noqa: E501
content, potential_tool_calls = xlam_tool_parser.preprocess_model_output(
model_output)
model_output
)
assert content == "<think>I'll help you with that.</think>"
assert (
potential_tool_calls ==
'[{"name": "get_current_weather", "arguments": {"city": "Seattle"}}]')
potential_tool_calls
== '[{"name": "get_current_weather", "arguments": {"city": "Seattle"}}]'
)
# Test with JSON code block
model_output = """I'll help you with that.
@@ -260,14 +303,16 @@ def test_preprocess_model_output(xlam_tool_parser):
[{"name": "get_current_weather", "arguments": {"city": "Seattle"}}]
```"""
content, potential_tool_calls = xlam_tool_parser.preprocess_model_output(
model_output)
model_output
)
assert content == "I'll help you with that."
assert "get_current_weather" in potential_tool_calls
# Test with no tool calls
model_output = """I'll help you with that."""
content, potential_tool_calls = xlam_tool_parser.preprocess_model_output(
model_output)
model_output
)
assert content == model_output
assert potential_tool_calls is None
@@ -281,7 +326,9 @@ def test_streaming_with_list_structure(xlam_tool_parser):
xlam_tool_parser.current_tool_id = -1
# Simulate receiving a message with list structure
current_text = """[{"name": "get_current_weather", "arguments": {"city": "Seattle"}}]""" # noqa: E501
current_text = (
"""[{"name": "get_current_weather", "arguments": {"city": "Seattle"}}]""" # noqa: E501
)
# First call to set up the tool
xlam_tool_parser.extract_tool_calls_streaming(
@@ -295,8 +342,7 @@ def test_streaming_with_list_structure(xlam_tool_parser):
)
# Make sure the tool is set up correctly
assert (xlam_tool_parser.current_tool_id
>= 0), "Tool index should be initialized"
assert xlam_tool_parser.current_tool_id >= 0, "Tool index should be initialized"
# Manually set up the state for sending the tool name
xlam_tool_parser.current_tools_sent = [False]
@@ -332,78 +378,102 @@ def test_streaming_with_list_structure(xlam_tool_parser):
(
"""[{"name": "get_current_weather", "arguments": {"city": "Dallas", "state": "TX", "unit": "fahrenheit"}}, {"name": "get_current_weather", "arguments": {"city": "Orlando", "state": "FL", "unit": "fahrenheit"}}]""", # noqa: E501
[
ToolCall(function=FunctionCall(
name="get_current_weather",
arguments=json.dumps({
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}),
)),
ToolCall(function=FunctionCall(
name="get_current_weather",
arguments=json.dumps({
"city": "Orlando",
"state": "FL",
"unit": "fahrenheit",
}),
)),
ToolCall(
function=FunctionCall(
name="get_current_weather",
arguments=json.dumps(
{
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}
),
)
),
ToolCall(
function=FunctionCall(
name="get_current_weather",
arguments=json.dumps(
{
"city": "Orlando",
"state": "FL",
"unit": "fahrenheit",
}
),
)
),
],
"",
),
(
"""<think>I'll help you with that.</think>[{"name": "get_current_weather", "arguments": {"city": "Dallas", "state": "TX", "unit": "fahrenheit"}}]""", # noqa: E501
[
ToolCall(function=FunctionCall(
name="get_current_weather",
arguments=json.dumps({
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}),
))
ToolCall(
function=FunctionCall(
name="get_current_weather",
arguments=json.dumps(
{
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}
),
)
)
],
"<think>I'll help you with that.</think>",
),
(
"""```json\n[{"name": "get_current_weather", "arguments": {"city": "Dallas", "state": "TX", "unit": "fahrenheit"}}]\n```""", # noqa: E501
[
ToolCall(function=FunctionCall(
name="get_current_weather",
arguments=json.dumps({
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}),
))
ToolCall(
function=FunctionCall(
name="get_current_weather",
arguments=json.dumps(
{
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}
),
)
)
],
"",
),
(
"""[TOOL_CALLS][{"name": "get_current_weather", "arguments": {"city": "Dallas", "state": "TX", "unit": "fahrenheit"}}]""", # noqa: E501
[
ToolCall(function=FunctionCall(
name="get_current_weather",
arguments=json.dumps({
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}),
))
ToolCall(
function=FunctionCall(
name="get_current_weather",
arguments=json.dumps(
{
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}
),
)
)
],
"",
),
(
"""I can help with that.<tool_call>[{"name": "get_current_weather", "arguments": {"city": "Dallas", "state": "TX", "unit": "fahrenheit"}}]</tool_call>""", # noqa: E501
[
ToolCall(function=FunctionCall(
name="get_current_weather",
arguments=json.dumps({
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}),
))
ToolCall(
function=FunctionCall(
name="get_current_weather",
arguments=json.dumps(
{
"city": "Dallas",
"state": "TX",
"unit": "fahrenheit",
}
),
)
)
],
"I can help with that.",
),
@@ -421,7 +491,8 @@ def test_extract_tool_calls_streaming_incremental(
chunks = []
for delta_message in stream_delta_message_generator(
xlam_tool_parser, xlam_tokenizer, model_output, request):
xlam_tool_parser, xlam_tokenizer, model_output, request
):
chunks.append(delta_message)
# Should have multiple chunks
@@ -433,8 +504,9 @@ def test_extract_tool_calls_streaming_incremental(
for chunk in chunks:
if chunk.tool_calls and chunk.tool_calls[0].id:
header_found = True
assert (chunk.tool_calls[0].function.name ==
expected_first_tool.function.name)
assert (
chunk.tool_calls[0].function.name == expected_first_tool.function.name
)
assert chunk.tool_calls[0].type == "function"
# Arguments may be empty initially or None
if chunk.tool_calls[0].function.arguments is not None:
@@ -446,11 +518,13 @@ def test_extract_tool_calls_streaming_incremental(
# Should have chunks with incremental arguments
arg_chunks = []
for chunk in chunks:
if (chunk.tool_calls and chunk.tool_calls[0].function.arguments
and chunk.tool_calls[0].function.arguments != ""
and chunk.tool_calls[0].index ==
0 # Only collect arguments from the first tool call
):
if (
chunk.tool_calls
and chunk.tool_calls[0].function.arguments
and chunk.tool_calls[0].function.arguments != ""
and chunk.tool_calls[0].index
== 0 # Only collect arguments from the first tool call
):
arg_chunks.append(chunk.tool_calls[0].function.arguments)
# Arguments should be streamed incrementally