Fix for KeyError on Loading LLaMA (#1978)

3a8c2381 · Jun Gao · GitHub · c85b80c2 · 3a8c2381
Unverified Commit 3a8c2381 authored Dec 10, 2023 by Jun Gao Committed by GitHub Dec 09, 2023
Hide whitespace changes
Inline Side-by-side

Showing with 4 additions and 0 deletions

vllm/model_executor/models/llama.py vllm/model_executor/models/llama.py +4 -0

No files found.
--- a/vllm/model_executor/models/llama.py
+++ b/vllm/model_executor/models/llama.py
@@ -322,6 +322,10 @@ class LlamaForCausalLM(nn.Module):
                model_name_or_path, cache_dir, load_format, revision):
            if "rotary_emb.inv_freq" in name:
                continue
+            if "rotary_emb.cos_cached" in name:
+                continue
+            if "rotary_emb.sin_cached" in name:
+                continue
            for (param_name, weight_name, shard_id) in stacked_params_mapping:
                if weight_name not in name:
                    continue