diff --git a/server/text_generation_server/models/vlm_causal_lm.py b/server/text_generation_server/models/vlm_causal_lm.py index be5843c7..14877e1a 100644 --- a/server/text_generation_server/models/vlm_causal_lm.py +++ b/server/text_generation_server/models/vlm_causal_lm.py @@ -190,6 +190,11 @@ class VlmCausalLMBatch(FlashCausalLMBatch): images.append(image) else: images.append([image]) + elif chunk_type == "video": + if config.model_type == "qwen2_vl": + # For now, treat video URLs as special tokens + # This will be processed in the text replacement section below + pass else: raise RuntimeError(f"Invalid chunk type {chunk_type}")