fix: adjust batch_tokenized_inputs output in mllama

huggingface · Dec 13, 2024 · 58e24a3 · 58e24a3
1 parent 3299a26
commit 58e24a3
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/server/text_generation_server/models/mllama_causal_lm.py b/server/text_generation_server/models/mllama_causal_lm.py
@@ -161,7 +161,7 @@ def from_pb_processor(
         dtype: torch.dtype,
         device: torch.device,
     ) -> "VlmCausalLMBatch":
-        batch_tokenized_inputs, image_inputs = cls.batch_tokenized_inputs(
+        batch_tokenized_inputs, image_inputs, _video_inputs = cls.batch_tokenized_inputs(
             pb.requests, tokenizer, processor, config
         )
         batch = cls.from_tokenized(pb, tokenizer, batch_tokenized_inputs, dtype, device)