[Bug] Fix the Image Input of Batch Generation (#1579)

python/sglang/srt/managers/io_struct.py

@@ -119,8 +119,7 @@ class GenerateReqInput:
             elif not isinstance(self.image_data, list):
                 self.image_data = [self.image_data] * num
             elif isinstance(self.image_data, list):
-                # FIXME incorrect order for duplication
-                self.image_data = self.image_data * num
+                pass
 
             if self.sampling_params is None:
                 self.sampling_params = [{}] * num

python/sglang/srt/openai_api/adapter.py

@@ -924,7 +924,7 @@ def v1_chat_generate_request(
         else:
             prompt_kwargs = {"input_ids": input_ids}
         sampling_params_list = sampling_params_list[0]
-        image_data = image_data_list[0]
+        image_data_list = image_data_list[0]
         return_logprobs = return_logprobs[0]
         logprob_start_lens = logprob_start_lens[0]
         top_logprobs_nums = top_logprobs_nums[0]
@@ -937,7 +937,7 @@ def v1_chat_generate_request(
 
     adapted_request = GenerateReqInput(
         **prompt_kwargs,
-        image_data=image_data,
+        image_data=image_data_list,
         sampling_params=sampling_params_list,
         return_logprob=return_logprobs,
         logprob_start_len=logprob_start_lens,