[Core] rename`PromptInputs` and `inputs` (#8876)

vllm/inputs/preprocess.py

@@ -10,8 +10,8 @@ from vllm.prompt_adapter.request import PromptAdapterRequest
 from vllm.transformers_utils.tokenizer_group import BaseTokenizerGroup
 from vllm.utils import print_warning_once
 
-from .data import (EncoderDecoderLLMInputs, LLMInputs, PromptInputs,
-                   SingletonPromptInputs)
+from .data import (EncoderDecoderLLMInputs, LLMInputs, PromptType,
+                   SingletonPrompt)
 from .parse import is_explicit_encoder_decoder_prompt, parse_singleton_prompt
 
 if TYPE_CHECKING:
@@ -209,7 +209,7 @@ class InputPreprocessor:
 
     def _extract_prompt_components(
         self,
-        inputs: SingletonPromptInputs,
+        prompt: SingletonPrompt,
         request_id: str,
         lora_request: Optional[LoRARequest] = None,
     ) -> PromptComponents:
@@ -219,7 +219,7 @@ class InputPreprocessor:
         Arguments:
 
         * request_id
-        * inputs: single encoder or decoder input prompt
+        * prompt: single encoder or decoder input prompt
         * lora_request: this is only valid for decoder prompts
 
         Returns:
@@ -229,24 +229,24 @@ class InputPreprocessor:
         * multi_modal_data
         '''
 
-        parsed = parse_singleton_prompt(inputs)
+        parsed = parse_singleton_prompt(prompt)
 
         if parsed["type"] == "str":
-            prompt = parsed["content"]
+            prompt_text = parsed["content"]
             prompt_token_ids = self._tokenize_prompt(
-                prompt,
+                prompt_text,
                 request_id=request_id,
                 lora_request=lora_request,
             )
             multi_modal_data = None
         elif parsed["type"] == "tokens":
-            prompt = None
+            prompt_text = None
             prompt_token_ids = parsed["content"]["prompt_token_ids"]
             multi_modal_data = parsed["content"].get("multi_modal_data")
         elif parsed["type"] == "text":
-            prompt = parsed["content"]["prompt"]
+            prompt_text = parsed["content"]["prompt"]
             prompt_token_ids = self._tokenize_prompt(
-                prompt,
+                prompt_text,
                 request_id=request_id,
                 lora_request=lora_request,
             )
@@ -254,33 +254,33 @@ class InputPreprocessor:
         else:
             assert_never(parsed)
 
-        return prompt, prompt_token_ids, multi_modal_data
+        return prompt_text, prompt_token_ids, multi_modal_data
 
     async def _extract_prompt_components_async(
         self,
-        inputs: SingletonPromptInputs,
+        prompt: SingletonPrompt,
         request_id: str,
         lora_request: Optional[LoRARequest] = None,
     ) -> PromptComponents:
         """Async version of :meth:`_extract_prompt_components`."""
-        parsed = parse_singleton_prompt(inputs)
+        parsed = parse_singleton_prompt(prompt)
 
         if parsed["type"] == "str":
-            prompt = parsed["content"]
+            prompt_text = parsed["content"]
             prompt_token_ids = await self._tokenize_prompt_async(
-                prompt,
+                prompt_text,
                 request_id=request_id,
                 lora_request=lora_request,
             )
             multi_modal_data = None
         elif parsed["type"] == "tokens":
-            prompt = None
+            prompt_text = None
             prompt_token_ids = parsed["content"]["prompt_token_ids"]
             multi_modal_data = parsed["content"].get("multi_modal_data")
         elif parsed["type"] == "text":
-            prompt = parsed["content"]["prompt"]
+            prompt_text = parsed["content"]["prompt"]
             prompt_token_ids = await self._tokenize_prompt_async(
-                prompt,
+                prompt_text,
                 request_id=request_id,
                 lora_request=lora_request,
             )
@@ -288,7 +288,7 @@ class InputPreprocessor:
         else:
             assert_never(parsed)
 
-        return prompt, prompt_token_ids, multi_modal_data
+        return prompt_text, prompt_token_ids, multi_modal_data
 
     def _build_enc_dec_llm_inputs(
         self,
@@ -321,7 +321,7 @@ class InputPreprocessor:
 
     def _process_encoder_decoder_prompt(
         self,
-        inputs: PromptInputs,
+        prompt: PromptType,
         request_id: str,
     ) -> EncoderDecoderLLMInputs:
         '''
@@ -349,7 +349,7 @@ class InputPreprocessor:
         
         Arguments:
 
-        * inputs: an input prompt
+        * prompt: an input prompt
         * request_id
 
         Returns:
@@ -360,13 +360,13 @@ class InputPreprocessor:
         encoder_comps: PromptComponents
         decoder_comps: DecoderPromptComponents
 
-        if is_explicit_encoder_decoder_prompt(inputs):
+        if is_explicit_encoder_decoder_prompt(prompt):
             encoder_comps = self._extract_prompt_components(
-                inputs["encoder_prompt"],
+                prompt["encoder_prompt"],
                 request_id=request_id,
             )
 
-            if (decoder_input := inputs["decoder_prompt"]) is None:
+            if (decoder_input := prompt["decoder_prompt"]) is None:
                 decoder_comps = None, None, None
             else:
                 decoder_comps = self._extract_prompt_components(
@@ -375,7 +375,7 @@ class InputPreprocessor:
                 )
         else:
             encoder_comps = self._extract_prompt_components(
-                inputs,
+                prompt,
                 request_id=request_id,
             )
 
@@ -385,20 +385,20 @@ class InputPreprocessor:
 
     async def _process_encoder_decoder_prompt_async(
         self,
-        inputs: PromptInputs,
+        prompt: PromptType,
         request_id: str,
     ) -> EncoderDecoderLLMInputs:
         """Async version of :meth:`_process_encoder_decoder_prompt`."""
         encoder_comps: PromptComponents
         decoder_comps: DecoderPromptComponents
 
-        if is_explicit_encoder_decoder_prompt(inputs):
+        if is_explicit_encoder_decoder_prompt(prompt):
             encoder_task = self._extract_prompt_components_async(
-                inputs["encoder_prompt"],
+                prompt["encoder_prompt"],
                 request_id=request_id,
             )
 
-            if (decoder_input := inputs["decoder_prompt"]) is None:
+            if (decoder_input := prompt["decoder_prompt"]) is None:
                 encoder_comps = await encoder_task
                 decoder_comps = None, None, None
             else:
@@ -411,7 +411,7 @@ class InputPreprocessor:
                     encoder_task, decoder_task)
         else:
             encoder_comps = await self._extract_prompt_components_async(
-                inputs,
+                prompt,
                 request_id=request_id,
             )
 
@@ -435,7 +435,7 @@ class InputPreprocessor:
 
     def _process_decoder_only_prompt(
         self,
-        inputs: SingletonPromptInputs,
+        prompt: SingletonPrompt,
         request_id: str,
         lora_request: Optional[LoRARequest] = None,
         prompt_adapter_request: Optional[PromptAdapterRequest] = None,
@@ -446,7 +446,7 @@ class InputPreprocessor:
 
         Arguments:
 
-        * inputs: input prompt
+        * prompt: input prompt
         * request_id
         * lora_request
         * prompt_adapter_request
@@ -457,7 +457,7 @@ class InputPreprocessor:
         '''
 
         prompt_comps = self._extract_prompt_components(
-            inputs,
+            prompt,
             request_id=request_id,
             lora_request=lora_request,
         )
@@ -469,14 +469,14 @@ class InputPreprocessor:
 
     async def _process_decoder_only_prompt_async(
         self,
-        inputs: SingletonPromptInputs,
+        prompt: SingletonPrompt,
         request_id: str,
         lora_request: Optional[LoRARequest] = None,
         prompt_adapter_request: Optional[PromptAdapterRequest] = None,
     ) -> LLMInputs:
         """Async version of :meth:`_process_decoder_only_prompt`."""
         prompt_comps = await self._extract_prompt_components_async(
-            inputs,
+            prompt,
             request_id=request_id,
             lora_request=lora_request,
         )
@@ -488,7 +488,7 @@ class InputPreprocessor:
 
     def preprocess(
         self,
-        inputs: PromptInputs,
+        prompt: PromptType,
         request_id: str,
         lora_request: Optional[LoRARequest] = None,
         prompt_adapter_request: Optional[PromptAdapterRequest] = None,
@@ -498,17 +498,17 @@ class InputPreprocessor:
             # Encoder-decoder model requires special mapping of
             # input prompts to encoder & decoder
             return self._process_encoder_decoder_prompt(
-                inputs,
+                prompt,
                 request_id=request_id,
             )
 
-        if is_explicit_encoder_decoder_prompt(inputs):
+        if is_explicit_encoder_decoder_prompt(prompt):
             raise ValueError("Cannot pass encoder-decoder prompt "
                              "to decoder-only models")
 
         # Decoder-only operation
         return self._process_decoder_only_prompt(
-            inputs,
+            prompt,
             request_id=request_id,
             lora_request=lora_request,
             prompt_adapter_request=prompt_adapter_request,
@@ -516,7 +516,7 @@ class InputPreprocessor:
 
     async def preprocess_async(
         self,
-        inputs: PromptInputs,
+        prompt: PromptType,
         request_id: str,
         lora_request: Optional[LoRARequest] = None,
         prompt_adapter_request: Optional[PromptAdapterRequest] = None,
@@ -526,17 +526,17 @@ class InputPreprocessor:
             # Encoder-decoder model requires special mapping of
             # input prompts to encoder & decoder
             return await self._process_encoder_decoder_prompt_async(
-                inputs,
+                prompt,
                 request_id=request_id,
             )
 
-        if is_explicit_encoder_decoder_prompt(inputs):
+        if is_explicit_encoder_decoder_prompt(prompt):
             raise ValueError("Cannot pass encoder-decoder prompt "
                              "to decoder-only models")
 
         # Decoder-only operation
         return await self._process_decoder_only_prompt_async(
-            inputs,
+            prompt,
             request_id=request_id,
             lora_request=lora_request,
             prompt_adapter_request=prompt_adapter_request,