[Core] Subclass ModelRunner to support cross-attention & encoder sequences (towards eventual encoder/decoder model support) (#4942)

Co-authored-by: Andrew Feldman <afeld2012@gmail.com>
Co-authored-by: Nick Hill <nickhill@us.ibm.com>

vllm/model_executor/models/__init__.py

@@ -83,7 +83,16 @@ _EMBEDDING_MODELS = {
     "MistralModel": ("llama_embedding", "LlamaEmbeddingModel"),
 }
 
-_MODELS = {**_GENERATION_MODELS, **_EMBEDDING_MODELS}
+_CONDITIONAL_GENERATION_MODELS = {
+    "BartModel": ("bart", "BartForConditionalGeneration"),
+    "BartForConditionalGeneration": ("bart", "BartForConditionalGeneration"),
+}
+
+_MODELS = {
+    **_GENERATION_MODELS,
+    **_EMBEDDING_MODELS,
+    **_CONDITIONAL_GENERATION_MODELS
+}
 
 # Architecture -> type.
 # out of tree models