[MyPy][3/N] Fix MyPy errors in test groups (part 3) (#55939)

Signed-off-by: Taimys <ffy2281693949@gmail.com>
Signed-off-by: Luyi Xiao <81631911+luyixiao95@users.noreply.github.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: Luyi Xiao <81631911+luyixiao95@users.noreply.github.com>
Co-authored-by: Taimys <ffy2281693949@gmail.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Taimys
2026-09-30 11:06:07 +00:00
committed by GitHub
co-authored by Codex Luyi Xiao Taimys Harry Mellor Claude Opus 5.5
parent c4fa6f36a7
commit 5225f1bacb
118 changed files with 948 additions and 605 deletions
@@ -21,6 +21,7 @@ from vllm.entrypoints.pooling.typing import (
PoolingEngineInput,
PoolingServeContext,
)
from vllm.inputs import tokens_input
from vllm.outputs import PoolingOutput, PoolingRequestOutput
@@ -28,7 +29,7 @@ class TestEmbeddingRequestParsing:
"""Unit tests for OpenAI embedding request parsing."""
def test_input_messages_parses_as_chat_request(self):
request = TypeAdapter(EmbeddingRequest).validate_python(
request: EmbeddingRequest = TypeAdapter(EmbeddingRequest).validate_python(
{
"model": "test",
"input": [{"role": "user", "content": "hello"}],
@@ -42,7 +43,7 @@ class TestEmbeddingRequestParsing:
assert request.chat_template_kwargs == {"instruction": "Represent the query: "}
def test_batched_input_messages_parses_as_batch_chat_input_request(self):
request = TypeAdapter(EmbeddingRequest).validate_python(
request: EmbeddingRequest = TypeAdapter(EmbeddingRequest).validate_python(
{
"model": "test",
"input": [
@@ -65,7 +66,7 @@ class TestEmbeddingRequestParsing:
assert request.chat_template_kwargs == {"instruction": "Represent the query: "}
def test_token_ids_still_parse_as_completion_request(self):
request = TypeAdapter(EmbeddingRequest).validate_python(
request: EmbeddingRequest = TypeAdapter(EmbeddingRequest).validate_python(
{
"model": "test",
"input": [[1, 2, 3], [4, 5]],
@@ -76,7 +77,7 @@ class TestEmbeddingRequestParsing:
assert request.input == [[1, 2, 3], [4, 5]]
def test_messages_still_parses_as_chat_request(self):
request = TypeAdapter(EmbeddingRequest).validate_python(
request: EmbeddingRequest = TypeAdapter(EmbeddingRequest).validate_python(
{
"model": "test",
"messages": [{"role": "user", "content": "hello"}],
@@ -89,7 +90,7 @@ class TestEmbeddingRequestParsing:
assert request.chat_template_kwargs == {"instruction": "Represent the query: "}
def test_batched_messages_parses_as_batch_chat_request(self):
request = TypeAdapter(EmbeddingRequest).validate_python(
request: EmbeddingRequest = TypeAdapter(EmbeddingRequest).validate_python(
{
"model": "test",
"messages": [
@@ -410,13 +411,13 @@ class TestChunkedEmbeddingProcessing:
@classmethod
def _make_handler(cls):
handler = object.__new__(EmbedIOProcessor)
handler.model_config = cls._FakeModelConfig()
handler.model_config = cls._FakeModelConfig() # type: ignore[assignment]
handler.enable_chunked_processing = True
return handler
@staticmethod
def _make_context() -> PoolingServeContext[EmbeddingCompletionRequest]:
request = TypeAdapter(EmbeddingRequest).validate_python(
request: EmbeddingRequest = TypeAdapter(EmbeddingRequest).validate_python(
{
"model": "test",
"input": [[0, 1, 2, 3, 4], [10, 11]],
@@ -431,13 +432,13 @@ class TestChunkedEmbeddingProcessing:
request_id="embd-client-prompt-999-chunk-888",
engine_inputs=[
PoolingEngineInput(
prompts={"prompt_token_ids": [0, 1, 2, 3, 4]},
prompts=tokens_input(prompt_token_ids=[0, 1, 2, 3, 4]),
params=pooling_params,
lora_requests=None,
priorities=0,
),
PoolingEngineInput(
prompts={"prompt_token_ids": [10, 11]},
prompts=tokens_input(prompt_token_ids=[10, 11]),
params=pooling_params,
lora_requests=None,
priorities=0,
@@ -7,6 +7,7 @@ import torch
import torch.nn.functional as F
from vllm import LLM, EmbeddingRequestOutput, PoolingParams
from vllm.inputs import PromptType
from vllm.tasks import PoolingTask
MODEL_NAME = "intfloat/multilingual-e5-small"
@@ -54,7 +55,8 @@ def test_token_ids_prompts(llm: LLM):
@pytest.mark.skip_global_cleanup
def test_list_prompts(llm: LLM):
outputs = llm.embed([prompt, prompt_token_ids], use_tqdm=False)
prompts: list[PromptType] = [prompt, prompt_token_ids]
outputs = llm.embed(prompts, use_tqdm=False)
assert len(outputs) == 2
for i in range(len(outputs)):
assert isinstance(outputs[i], EmbeddingRequestOutput)
+13 -9
View File
@@ -12,9 +12,13 @@ import requests
import torch
import torch.nn.functional as F
from tests.models.language.pooling.embed_utils import run_embedding_correctness_test
from tests.models.language.pooling.embed_utils import (
float_embeddings,
run_embedding_correctness_test,
)
from tests.models.utils import check_embeddings_close
from tests.utils import RemoteOpenAIServer
from vllm.entrypoints.chat_utils import ChatCompletionMessageParam
from vllm.entrypoints.pooling.embed.protocol import EmbeddingResponse
from vllm.entrypoints.pooling.pooling.protocol import PoolingResponse
from vllm.entrypoints.pooling.utils import (
@@ -145,7 +149,7 @@ async def test_completion_request(
assert embeddings.usage.prompt_tokens == len(input_tokens)
assert embeddings.usage.total_tokens == len(input_tokens)
vllm_outputs = [d.embedding for d in embeddings.data]
vllm_outputs = float_embeddings(embeddings)
run_embedding_correctness_test(hf_model, [input_text], vllm_outputs)
# test input: list[int]
@@ -165,7 +169,7 @@ async def test_completion_request(
assert embeddings.usage.prompt_tokens == len(input_tokens)
assert embeddings.usage.total_tokens == len(input_tokens)
vllm_outputs = [d.embedding for d in embeddings.data]
vllm_outputs = float_embeddings(embeddings)
run_embedding_correctness_test(hf_model, [input_text], vllm_outputs)
@@ -194,7 +198,7 @@ async def test_completion_request_batched(
assert embeddings.usage.prompt_tokens == len(input_tokens) * N
assert embeddings.usage.total_tokens == len(input_tokens) * N
vllm_outputs = [d.embedding for d in embeddings.data]
vllm_outputs = float_embeddings(embeddings)
run_embedding_correctness_test(hf_model, input_texts, vllm_outputs)
# test list[list[int]]
@@ -214,7 +218,7 @@ async def test_completion_request_batched(
assert embeddings.usage.prompt_tokens == len(input_tokens) * N
assert embeddings.usage.total_tokens == len(input_tokens) * N
vllm_outputs = [d.embedding for d in embeddings.data]
vllm_outputs = float_embeddings(embeddings)
run_embedding_correctness_test(hf_model, input_texts, vllm_outputs)
@@ -331,7 +335,7 @@ async def test_padding(client: openai.AsyncOpenAI, model_name: str):
async def test_chat_request(
server: RemoteOpenAIServer, client: openai.AsyncOpenAI, model_name: str
):
messages = [
messages: list[ChatCompletionMessageParam] = [
{
"role": "user",
"content": "The cat sat on the mat.",
@@ -382,8 +386,8 @@ async def test_chat_request(
assert chat_embeddings.created <= completion_embeddings.created
# Use tolerance-based comparison for embeddings
check_embeddings_close(
embeddings_0_lst=[d.embedding for d in chat_embeddings.data],
embeddings_1_lst=[d.embedding for d in completion_embeddings.data],
embeddings_0_lst=float_embeddings(chat_embeddings),
embeddings_1_lst=float_embeddings(completion_embeddings),
name_0="chat",
name_1="completion",
)
@@ -488,7 +492,7 @@ async def test_invocations_completion_request(
@pytest.mark.asyncio
async def test_invocations_chat_request(server: RemoteOpenAIServer):
messages = [
messages: list[ChatCompletionMessageParam] = [
{
"role": "user",
"content": "The cat sat on the mat.",
@@ -6,7 +6,10 @@ import openai
import pytest
from tests.conftest import HfRunner
from tests.models.language.pooling.embed_utils import run_embedding_correctness_test
from tests.models.language.pooling.embed_utils import (
float_embeddings,
run_embedding_correctness_test,
)
from tests.models.utils import EmbedModelInfo
from tests.utils import ROCM_EXTRA_ARGS, RemoteOpenAIServer
from vllm.entrypoints.pooling.embed.protocol import EmbeddingResponse
@@ -95,7 +98,7 @@ async def test_matryoshka(
if dimensions is not None:
assert len(embeddings.data[0].embedding) == dimensions
vllm_outputs = [d.embedding for d in embeddings.data]
vllm_outputs = float_embeddings(embeddings)
run_embedding_correctness_test(hf_model, prompts, vllm_outputs, dimensions)
if model_info.is_matryoshka:
@@ -274,6 +274,7 @@ async def test_long_text_embedding_1500_chars(
# Verify embedding vector validity
embedding_vector = embeddings.data[0].embedding
assert isinstance(embedding_vector, list)
assert all(isinstance(x, float) for x in embedding_vector), (
"Embedding vector should contain floats"
)
@@ -322,6 +323,7 @@ async def test_long_text_embedding_2500_chars(
# Verify embedding vector validity
embedding_vector = embeddings.data[0].embedding
assert isinstance(embedding_vector, list)
assert all(isinstance(x, float) for x in embedding_vector), (
"Embedding vector should contain floats"
)
@@ -409,6 +411,7 @@ async def test_chunked_vs_normal_consistency(
# 验证embedding向量的有效性
embedding_vector = embeddings.data[0].embedding
assert isinstance(embedding_vector, list)
assert all(isinstance(x, float) for x in embedding_vector)
assert not all(x == 0 for x in embedding_vector)
@@ -438,6 +441,7 @@ async def test_chunked_processing_response_format(
# Verify embedding vector properties
embedding_vector = embeddings.data[0].embedding
assert isinstance(embedding_vector, list)
import math
vector_norm = math.sqrt(sum(x * x for x in embedding_vector))