Llama3.2 vision model support (#1551)

python/sglang/srt/configs/model_config.py

@@ -89,6 +89,8 @@ class ModelConfig:
         self.num_hidden_layers = self.hf_text_config.num_hidden_layers
         self.vocab_size = self.hf_text_config.vocab_size
 
+        self.is_encoder_decoder = self.hf_config.model_type in ["mllama"]
+
     # adapted from https://github.com/vllm-project/vllm/blob/main/vllm/config.py#L289
     def get_total_num_kv_heads(self) -> int:
         """Returns the total number of KV heads."""