A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 17, 06:45 UTC
Fix for KeyError on Loading LLaMA (#1978)
Jun Gao committed
Dec 10, 2023 at 07:59 UTC
3a8c2381f7d7933b01dd9ddbd4c3df2ce673cfc1
1 file changed
+4
vllm/model_executor/models/llama.py
+4
@@ -322,6 +322,10 @@ class LlamaForCausalLM(nn.Module):
322
model_name_or_path, cache_dir, load_format, revision):
323
if "rotary_emb.inv_freq" in name:
324
continue
325
+ if "rotary_emb.cos_cached" in name:
326
+ continue
327
+ if "rotary_emb.sin_cached" in name:
328
+ continue
329
for (param_name, weight_name, shard_id) in stacked_params_mapping:
330
if weight_name not in name:
331
continue