Fix OPT param names (#1819)

vllm-project · Nov 28, 2023 · b943890 · b943890
1 parent a1125ad
commit b943890
Showing 1 changed file with 3 additions and 0 deletions.
diff --git a/vllm/model_executor/models/opt.py b/vllm/model_executor/models/opt.py
@@ -331,6 +331,9 @@ def load_weights(self,
                 model_name_or_path, cache_dir, load_format, revision):
             if "lm_head.weight" in name:
                 continue
+            if name.startswith("decoder."):
+                name = "model." + name
+
             for (param_name, weight_name, shard_id) in stacked_params_mapping:
                 if weight_name not in name:
                     continue