starlight-apk/Qwen
1
0
Fork 0
mirror of https://github.com/QwenLM/Qwen.git synced 2026-05-21 00:45:48 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #476 from dlutsniper/main

Browse Source 
Update finetune.py For lower cpu memory
This commit is contained in:
Junyang Lin
2023-10-19 23:10:43 +08:00
committed by GitHub
parent 3885918ab1 f3d7c69be2
commit 07de511766
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
finetune.py
View File

@@ -278,7 +278,7 @@ def train():
local_rank = training_args.local_rank local_rank = training_args.local_rank
device_map = None device_map = "auto"
world_size = int(os.environ.get("WORLD_SIZE", 1)) world_size = int(os.environ.get("WORLD_SIZE", 1))
ddp = world_size != 1 ddp = world_size != 1
if lora_args.q_lora: if lora_args.q_lora:
@@ -302,6 +302,7 @@ def train():
config=config, config=config,
cache_dir=training_args.cache_dir, cache_dir=training_args.cache_dir,
device_map=device_map, device_map=device_map,
low_cpu_mem_usage=True if training_args.use_lora and not lora_args.q_lora,
trust_remote_code=True, trust_remote_code=True,
quantization_config=GPTQConfig( quantization_config=GPTQConfig(
bits=4, disable_exllama=True bits=4, disable_exllama=True