grrr spelling mistake

Author: eitanturok

vllm/v1/spec_decode/eagle.py

@@ -657,8 +657,8 @@ def load_model(self, target_model: nn.Module) -> None:
             self.hot_token_ids = load_draft_vocab_pruned(self.vllm_config.speculative_config.draft_vocab_pruned)
             device = next(self.model.model.parameters()).device
             self.hot_token_ids = self.hot_token_ids.to(device)
-            # self.model.model.embed_tokens.weight is the model head
-            self.model.model.embed_tokens.weight.data = self.model.model.embed_tokens.weight.data[self.hot_token_id]
+            # `self.model.model.embed_tokens.weight` is the model head
+            self.model.model.embed_tokens.weight.data = self.model.model.embed_tokens.weight.data[self.hot_token_ids]
 
 
     @torch.inference_mode()