ros/LLaMA-Factory
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

use low_cpu_mem_usage to speed up loading

Browse Source 
Former-commit-id: 7891e4c200566a4a47088e93efd1fbebcb46528e
This commit is contained in:
hiyouga
2023-06-03 18:19:01 +08:00
parent 9b8b6623ac
commit 5eef8d5d98
3 changed files with 24 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/export_model.py
View File

@@ -13,7 +13,7 @@ def main():
model_args, training_args, finetuning_args = parser.parse_args_into_dataclasses()
model, tokenizer = load_pretrained(model_args, finetuning_args)
model.save_pretrained(training_args.output_dir, max_shard_size="1GB")
model.save_pretrained(training_args.output_dir, max_shard_size="10GB")
tokenizer.save_pretrained(training_args.output_dir)
print("model and tokenizer have been saved at:", training_args.output_dir)