[Renderer] Separate out RendererConfig from ModelConfig

docs/contributing/model/transcription.md

@@ -22,7 +22,7 @@ Declare supported languages and capabilities:
     import torch
     from torch import nn
 
-    from vllm.config import ModelConfig, SpeechToTextConfig
+    from vllm.config import RendererConfig, SpeechToTextConfig
     from vllm.inputs.data import PromptType
     from vllm.model_executor.models.interfaces import SupportsTranscription
 
@@ -52,7 +52,7 @@ This is for controlling general behavior of the API when serving your model:
         @classmethod
         def get_speech_to_text_config(
             cls,
-            model_config: ModelConfig,
+            renderer_config: RendererConfig,
             task_type: Literal["transcribe", "translate"],
         ) -> SpeechToTextConfig:
             return SpeechToTextConfig(
@@ -83,7 +83,7 @@ Return a dict containing `multi_modal_data` with the audio, and either a `prompt
             cls,
             audio: np.ndarray,
             stt_config: SpeechToTextConfig,
-            model_config: ModelConfig,
+            renderer_config: RendererConfig,
             language: str | None,
             task_type: Literal["transcribe", "translate"],
             request_prompt: str,
@@ -120,7 +120,7 @@ Return a dict with separate `encoder_prompt` and `decoder_prompt` entries:
             cls,
             audio: np.ndarray,
             stt_config: SpeechToTextConfig,
-            model_config: ModelConfig,
+            renderer_config: RendererConfig,
             language: str | None,
             task_type: Literal["transcribe", "translate"],
             request_prompt: str,
@@ -183,7 +183,7 @@ Provide a fast duration→token estimate to improve streaming usage statistics:
             cls,
             audio_duration_s: float,
             stt_config: SpeechToTextConfig,
-            model_config: ModelConfig,
+            renderer_config: renderer_config,
         ) -> int | None:
             # Return None if unknown; otherwise return an estimate.
             return int(audio_duration_s * stt_config.sample_rate // 320)  # example

tests/entrypoints/openai/test_chat_template.py

@@ -3,7 +3,6 @@
 
 import pytest
 
-from vllm.config import ModelConfig
 from vllm.entrypoints.chat_utils import apply_hf_chat_template, load_chat_template
 from vllm.entrypoints.openai.protocol import ChatCompletionRequest
 from vllm.tokenizers import get_tokenizer
@@ -107,24 +106,11 @@ def test_get_gen_prompt(
     model_info = HF_EXAMPLE_MODELS.find_hf_info(model)
     model_info.check_available_online(on_fail="skip")
 
-    model_config = ModelConfig(
-        model,
-        tokenizer=model_info.tokenizer or model,
-        tokenizer_mode=model_info.tokenizer_mode,
-        trust_remote_code=model_info.trust_remote_code,
-        revision=model_info.revision,
-        hf_overrides=model_info.hf_overrides,
-        skip_tokenizer_init=model_info.require_embed_inputs,
-        enable_prompt_embeds=model_info.require_embed_inputs,
-        enable_mm_embeds=model_info.require_embed_inputs,
-        enforce_eager=model_info.enforce_eager,
-        dtype=model_info.dtype,
-    )
+    renderer_config = model_info.build_renderer_config()
 
-    # Initialize the tokenizer
     tokenizer = get_tokenizer(
-        tokenizer_name=model_config.tokenizer,
-        trust_remote_code=model_config.trust_remote_code,
+        renderer_config.tokenizer,
+        trust_remote_code=renderer_config.trust_remote_code,
     )
     template_content = load_chat_template(chat_template=template)
 
@@ -143,7 +129,7 @@ def test_get_gen_prompt(
         tokenizer=tokenizer,
         conversation=mock_request.messages,
         chat_template=mock_request.chat_template or template_content,
-        model_config=model_config,
+        renderer_config=renderer_config,
         tools=None,
         add_generation_prompt=mock_request.add_generation_prompt,
         continue_final_message=mock_request.continue_final_message,

tests/entrypoints/openai/test_vision.py

@@ -118,7 +118,7 @@ def get_hf_prompt_tokens(model_name, content, image_url):
         image = image.media
     images = [image]
 
-    prompt = processor.tokenizer.apply_chat_template(
+    prompt = processor.apply_chat_template(
         messages, tokenize=False, add_generation_prompt=True
     )
     inputs = processor(prompt, images, return_tensors="pt")