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

[Renderer] Separate out RendererConfig from ModelConfig (#30145)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-12-07 15:15:42 +08:00
committed by GitHub
parent a49d813fa8
commit 27f4c2fd46
105 changed files with 969 additions and 797 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
tests/v1/structured_output/test_backend_guidance.py
View File

@@ -6,7 +6,7 @@ from concurrent.futures import Future
import pytest
from transformers import AutoTokenizer
from vllm.config import StructuredOutputsConfig, VllmConfig
from vllm.config import RendererConfig, StructuredOutputsConfig, VllmConfig
from vllm.config.model import ModelConfig
from vllm.config.parallel import ParallelConfig
from vllm.config.speculative import SpeculativeConfig
@@ -72,8 +72,11 @@ def test_backend_guidance_rollback_terminated():
def test_grammar_bitmask_with_specdec():
tokenizer = AutoTokenizer.from_pretrained(TOKENIZER)
prompt = tokenizer.encode('{"a": "b"}')
model_config = ModelConfig(tokenizer=TOKENIZER)
vllm_config = VllmConfig(
model_config=ModelConfig(tokenizer=TOKENIZER),
model_config=model_config,
renderer_config=RendererConfig(model_config=model_config, tokenizer=TOKENIZER),
structured_outputs_config=StructuredOutputsConfig(backend="guidance"),
speculative_config=SpeculativeConfig(model="[ngram]", num_speculative_tokens=3),
)
@@ -137,8 +140,11 @@ def test_grammar_init_async_and_sync(async_grammar):
# Use "external_launcher" for sync mode, None for async mode
executor_backend = None if async_grammar else "external_launcher"
model_config = ModelConfig(tokenizer=TOKENIZER)
vllm_config = VllmConfig(
model_config=ModelConfig(tokenizer=TOKENIZER),
model_config=model_config,
renderer_config=RendererConfig(model_config=model_config, tokenizer=TOKENIZER),
structured_outputs_config=StructuredOutputsConfig(backend="guidance"),
parallel_config=ParallelConfig(distributed_executor_backend=executor_backend),
)

35
tests/v1/structured_output/test_reasoning_structured_output.py
View File

@@ -7,7 +7,7 @@ from unittest.mock import Mock
import pytest
from vllm.config import ModelConfig, SchedulerConfig, VllmConfig
from vllm.config import ModelConfig, RendererConfig, SchedulerConfig, VllmConfig
from vllm.reasoning import ReasoningParser
from vllm.v1.request import Request
from vllm.v1.structured_output import StructuredOutputManager
@@ -17,19 +17,26 @@ class TestReasoningStructuredOutput:
"""Test reasoning-aware structured output functionality."""
@pytest.fixture
def mock_model_config(self):
"""Create a mock ModelConfig."""
config = Mock(spec=ModelConfig)
config.skip_tokenizer_init = True # Skip tokenizer init to avoid network calls
config.get_vocab_size = Mock(return_value=50000)
def mock_renderer_config(self):
"""Create a mock RendererConfig."""
renderer_config = Mock(spec=RendererConfig)
renderer_config.skip_tokenizer_init = (
True # Skip tokenizer init to avoid network calls
)
model_config = Mock(spec=ModelConfig)
model_config.get_vocab_size = Mock(return_value=50000)
model_config.trust_remote_code = False
# Add missing runner_type attribute that tokenizer initialization expects
config.runner_type = "generate"
model_config.runner_type = "generate"
renderer_config.model_config = model_config
# Add other attributes that tokenizer initialization might need
config.tokenizer = "test-tokenizer"
config.tokenizer_mode = "auto"
config.trust_remote_code = False
config.tokenizer_revision = None
return config
renderer_config.tokenizer = "test-tokenizer"
renderer_config.tokenizer_mode = "auto"
renderer_config.tokenizer_revision = None
return renderer_config
@pytest.fixture
def mock_scheduler_config(self):
@@ -39,10 +46,10 @@ class TestReasoningStructuredOutput:
return config
@pytest.fixture
def mock_vllm_config(self, mock_model_config, mock_scheduler_config):
def mock_vllm_config(self, mock_renderer_config, mock_scheduler_config):
"""Create a mock VllmConfig."""
config = Mock(spec=VllmConfig)
config.model_config = mock_model_config
config.renderer_config = mock_renderer_config
config.scheduler_config = mock_scheduler_config
config.structured_outputs_config = Mock()
config.structured_outputs_config.reasoning_parser = None