diff --git a/server/text_generation_server/models/flash_rw.py b/server/text_generation_server/models/flash_rw.py
index 5f963bfb..33079ace 100644
--- a/server/text_generation_server/models/flash_rw.py
+++ b/server/text_generation_server/models/flash_rw.py
@@ -48,7 +48,13 @@ class FlashRWSharded(FlashCausalLM):
 
         torch.distributed.barrier(group=self.process_group)
         filenames = weight_files(model_id, revision=revision, extension=".safetensors")
-        weights = Weights(filenames, device, dtype, process_group=self.process_group)
+        weights = Weights(
+            filenames,
+            device,
+            dtype,
+            process_group=self.process_group,
+            aliases={"transformer.word_embeddings.weight": ["lm_head.weight"]},
+        )
 
         config.quantize = quantize