[VLM RLHF] Take Image input for verl vlm rollout

python/sglang/srt/entrypoints/engine.py

@@ -151,10 +151,6 @@ def generate(
         The arguments of this function is the same as `sglang/srt/managers/io_struct.py::GenerateReqInput`.
         Please refer to `GenerateReqInput` for the documentation.
         """
-        modalities_list = []
-        if image_data is not None:
-            modalities_list.append("image")
-
         obj = GenerateReqInput(
             text=prompt,
             input_ids=input_ids,
@@ -165,7 +161,6 @@ def generate(
             top_logprobs_num=top_logprobs_num,
             token_ids_logprob=token_ids_logprob,
             lora_path=lora_path,
-            modalities=modalities_list,
             custom_logit_processor=custom_logit_processor,
             return_hidden_states=return_hidden_states,
             stream=stream,

python/sglang/srt/managers/multimodal_processors/base_processor.py

@@ -139,8 +139,6 @@ def load_mm_data(
         else:
             multimodal_tokens.image_token = multimodal_tokens.image_token
 
-        assert isinstance(prompt, str)
-
         if isinstance(prompt, list) and return_text:
             assert len(prompt) and isinstance(prompt[0], int)
             prompt = self._processor.tokenizer.decode(prompt)
@@ -204,7 +202,16 @@ def load_mm_data(
                             continue
 
                     image_sizes += frames[0].size * len(frames)
-                    hashes += [hash(image_file)] * len(frames)
+
+                    # Generate a hashable value for the image file
+                    if isinstance(image_file, Image.Image):
+                        # For PIL.Image objects, use the ID as a hashable value
+                        hash_value = hash(id(image_file))
+                    else:
+                        # For other types (strings, etc.), use the regular hash
+                        hash_value = hash(image_file)
+
+                    hashes += [hash_value] * len(frames)
                     images += frames
                     image_index += 1
                     if frames_to_process != 0:

python/sglang/srt/managers/multimodal_processors/qwen_vl.py

@@ -5,7 +5,7 @@
 import torch
 from PIL import Image
 
-from sglang.srt.managers.multimodal_processor import (
+from sglang.srt.managers.multimodal_processors.base_processor import (
     BaseMultimodalProcessor as SGLangBaseProcessor,
 )
 from sglang.srt.managers.multimodal_processors.base_processor import (

python/sglang/srt/utils.py

@@ -566,10 +566,14 @@ def uniform_sample(l, n):
     return frames
 
 
-def load_image(image_file: Union[str, bytes]) -> tuple[Image, tuple[int, int]]:
+def load_image(
+    image_file: Union[Image.Image, str, bytes]
+) -> tuple[Image.Image, tuple[int, int]]:
     image = image_size = None
-
-    if isinstance(image_file, bytes):
+    if isinstance(image_file, Image.Image):
+        image = image_file
+        image_size = (image.width, image.height)
+    elif isinstance(image_file, bytes):
         image = Image.open(BytesIO(image_file))
     elif image_file.startswith("http://") or image_file.startswith("https://"):
         timeout = int(os.getenv("REQUEST_TIMEOUT", "3"))