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, 08:00 UTC
Don't try to load training_args.bin (#373)
Fazlul Shahriar committed
Jul 8, 2023 at 18:26 UTC
75beba29b5c7316a9ebde9b0886f609dd5bf05bd
1 file changed
+4
-1
vllm/model_executor/weight_utils.py
+4
-1
@@ -39,7 +39,10 @@ def hf_model_weights_iterator(
39
else:
40
hf_folder = model_name_or_path
41
42
- hf_bin_files = glob.glob(os.path.join(hf_folder, "*.bin"))
42
+ hf_bin_files = [
43
+ x for x in glob.glob(os.path.join(hf_folder, "*.bin"))
44
+ if not x.endswith("training_args.bin")
45
+ ]
46
47
if use_np_cache:
48
# Convert the model weights from torch tensors to numpy arrays for