[Misc] V0 fallback for `--enable-prompt-embeds` (#17615)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

vllm/engine/arg_utils.py

@@ -1237,6 +1237,12 @@ class EngineArgs:
                                recommend_to_remove=False)
             return False
 
+        # No text embedding inputs so far.
+        if self.enable_prompt_embeds:
+            _raise_or_fallback(feature_name="--enable-prompt-embeds",
+                               recommend_to_remove=False)
+            return False
+
         # Only Fp16 and Bf16 dtypes since we only support FA.
         V1_SUPPORTED_DTYPES = [torch.bfloat16, torch.float16]
         if model_config.dtype not in V1_SUPPORTED_DTYPES:

vllm/inputs/preprocess.py

@@ -6,7 +6,6 @@ from typing import Any, Optional, Union, cast
 
 from typing_extensions import assert_never
 
-from vllm import envs
 from vllm.config import ModelConfig
 from vllm.logger import init_logger
 from vllm.lora.request import LoRARequest
@@ -306,8 +305,6 @@ class InputPreprocessor:
         if not self.model_config.enable_prompt_embeds:
             raise ValueError("You must set `--enable-prompt-embeds` to input "
                              "`prompt_embeds`.")
-        if envs.VLLM_USE_V1:
-            raise ValueError("`prompt_embeds` is only available in V0.")
 
         prompt_embeds = parsed_content["prompt_embeds"]