[Core] Support configuration parsing plugin (vllm-project#24277)

Signed-off-by: Xingyu Liu <[email protected]>
Signed-off-by: Xingyu Liu <[email protected]>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

Author: 2 people

tests/transformers_utils/__init__.py

@@ -0,0 +1,37 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+from pathlib import Path
+from typing import Optional, Union
+
+import pytest
+from transformers import PretrainedConfig
+
+from vllm.transformers_utils.config import (get_config_parser,
+                                            register_config_parser)
+from vllm.transformers_utils.config_parser_base import ConfigParserBase
+
+
+@register_config_parser("custom_config_parser")
+class CustomConfigParser(ConfigParserBase):
+
+    def parse(self,
+              model: Union[str, Path],
+              trust_remote_code: bool,
+              revision: Optional[str] = None,
+              code_revision: Optional[str] = None,
+              **kwargs) -> tuple[dict, PretrainedConfig]:
+        raise NotImplementedError
+
+
+def test_register_config_parser():
+    assert isinstance(get_config_parser("custom_config_parser"),
+                      CustomConfigParser)
+
+
+def test_invalid_config_parser():
+    with pytest.raises(ValueError):
+
+        @register_config_parser("invalid_config_parser")
+        class InvalidConfigParser:
+            pass

tests/transformers_utils/test_config_parser_registry.py

@@ -419,7 +419,7 @@ class ModelConfig:
     `--media-io-kwargs '{"video": {"num_frames": 40} }'` """
     use_async_output_proc: bool = True
     """Whether to use async output processor."""
-    config_format: Union[str, ConfigFormat] = ConfigFormat.AUTO.value
+    config_format: Union[str, ConfigFormat] = "auto"
     """The format of the model config to load:\n
     - "auto" will try to load the config in hf format if available else it
     will try to load in mistral format.\n
@@ -624,9 +624,6 @@ def __post_init__(self) -> None:
             raise ValueError(
                 "Sleep mode is not supported on current platform.")
 
-        if isinstance(self.config_format, str):
-            self.config_format = ConfigFormat(self.config_format)
-
         hf_config = get_config(self.hf_config_path or self.model,
                                self.trust_remote_code,
                                self.revision,

vllm/config/__init__.py

@@ -22,9 +22,9 @@
 
 import vllm.envs as envs
 from vllm.config import (BlockSize, CacheConfig, CacheDType, CompilationConfig,
-                         ConfigFormat, ConfigType, ConvertOption,
-                         DecodingConfig, DetailedTraceModules, Device,
-                         DeviceConfig, DistributedExecutorBackend, EPLBConfig,
+                         ConfigType, ConvertOption, DecodingConfig,
+                         DetailedTraceModules, Device, DeviceConfig,
+                         DistributedExecutorBackend, EPLBConfig,
                          GuidedDecodingBackend, HfOverrides, KVEventsConfig,
                          KVTransferConfig, LoadConfig, LogprobsMode,
                          LoRAConfig, MambaDType, MMEncoderTPMode, ModelConfig,
@@ -547,7 +547,6 @@ def add_cli_args(parser: FlexibleArgumentParser) -> FlexibleArgumentParser:
             help="Disable async output processing. This may result in "
             "lower performance.")
         model_group.add_argument("--config-format",
-                                 choices=[f.value for f in ConfigFormat],
                                  **model_kwargs["config_format"])
         # This one is a special case because it can bool
         # or str. TODO: Handle this in get_kwargs