Remove problematic AutoModel.register() call

modeling_llm2vec4cxr.py

@@ -124,8 +124,3 @@ class LLM2Vec4CXRModel(LlamaBiModel):
         embs = self.encode_with_separator(all_texts, separator=separator, max_length=max_length)
         # embs: [N, 2048]; compare query vs candidates
         return F.cosine_similarity(embs[0], embs[1:], dim=1)
-
-
-# Register the model for auto loading
-from transformers import AutoModel
-AutoModel.register(LLM2Vec4CXRModel.__name__, LLM2Vec4CXRModel)