423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-08-04 20:52:59 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix #1176 #1177

Browse Source 
Former-commit-id: 01d8cb1ca76e1d599fa5e970b2fcaa3ceedbfffd
This commit is contained in:
hiyouga 2023-10-14 20:00:17 +08:00
parent e585c789ce
commit 1faba4b704
1 changed files with 2 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/llmtuner/tuner/core/loader.py
View File

@ -71,6 +71,7 @@ def load_model_and_tokenizer(
tokenizer = AutoTokenizer.from_pretrained( tokenizer = AutoTokenizer.from_pretrained(
model_args.model_name_or_path, model_args.model_name_or_path,
use_fast=model_args.use_fast_tokenizer, use_fast=model_args.use_fast_tokenizer,
split_special_tokens=model_args.split_special_tokens,
padding_side="right", # training with left-padded tensors in fp16 precision may cause overflow padding_side="right", # training with left-padded tensors in fp16 precision may cause overflow
**config_kwargs **config_kwargs
) )
@ -184,6 +185,7 @@ def load_model_and_tokenizer(
model = AutoModelForCausalLM.from_pretrained( model = AutoModelForCausalLM.from_pretrained(
model_to_load, model_to_load,
config=config, config=config,
torch_dtype=getattr(config, "torch_dtype"),
low_cpu_mem_usage=(not is_deepspeed_zero3_enabled()), low_cpu_mem_usage=(not is_deepspeed_zero3_enabled()),
**config_kwargs **config_kwargs
) )