[BugFix] Fix Falcon tied embeddings (vllm-project#3590)

Co-authored-by: 44670 <[email protected]>

vllm/model_executor/models/falcon.py

@@ -37,7 +37,7 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.sampler import Sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
-    VocabParallelEmbedding, ParallelLMHead)
+    VocabParallelEmbedding)
 from vllm.model_executor.parallel_utils.communication_op import (
     tensor_model_parallel_all_reduce)
 from vllm.model_executor.parallel_utils.parallel_state import (
@@ -370,10 +370,7 @@ def __init__(
         self.config = config
         self.linear_method = linear_method
         self.transformer = FalconModel(config, linear_method)
-        self.lm_head = ParallelLMHead(
-            config.vocab_size,
-            config.hidden_size,
-        )
+        self.lm_head_weight = self.transformer.word_embeddings.weight
         self.logits_processor = LogitsProcessor(config.vocab_size)
         self.sampler = Sampler()
 
@@ -394,7 +391,7 @@ def forward(
 
     def compute_logits(self, hidden_states: torch.Tensor,
                        sampling_metadata: SamplingMetadata) -> torch.Tensor:
-        logits = self.logits_processor(self.lm_head.weight, hidden_states,
+        logits = self.logits_processor(self.lm_head_weight, hidden_states,
                                        sampling_metadata)
         return logits
 
@@ -419,9 +416,12 @@ def load_weights(self,
         else:
             total_num_kv_heads = total_num_heads
         num_query_heads_per_kv_head = total_num_heads // total_num_kv_heads
-        params_dict = dict(self.named_parameters())
+        params_dict = dict(self.named_parameters(remove_duplicate=False))
         for name, loaded_weight in hf_model_weights_iterator(
                 model_name_or_path, cache_dir, load_format, revision):
+            if name == "lm_head.weight":
+                # Falcon uses tied embeddings.
+                continue
             # Skip loading extra bias for GPTQ models.
             if name.endswith(".bias") and name not in params_dict:
                 continue