github
/
ColossalAI
mirror of https://github.com/hpcaitech/ColossalAI
1
0
Fork 0
Code Issues Projects Releases Wiki Activity
Making large AI models cheaper, faster and more accessible
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
3767 Commits
31 Branches
46 Tags
236 MiB
 
 
 
 
 
Branch: main
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from 'main'
${ noResults }
ColossalAI/examples/community/roberta/pretraining
pre-commit-ci[bot] 7c2f79fa98
[pre-commit.ci] pre-commit autoupdate (#5572) 		5 months ago
..
model [misc] update pre-commit and run all files (#4752) 1 year ago
utils [misc] update pre-commit and run all files (#4752) 1 year ago
README.md fix typo examples/community/roberta (#3925) 1 year ago
arguments.py [bug] fix get_default_parser in examples (#4764) 1 year ago
bert_dataset_provider.py [misc] update pre-commit and run all files (#4752) 1 year ago
evaluation.py [misc] update pre-commit and run all files (#4752) 1 year ago
hostfile
…
loss.py [misc] update pre-commit and run all files (#4752) 1 year ago
nvidia_bert_dataset_provider.py [pre-commit.ci] pre-commit autoupdate (#5572) 5 months ago
pretrain_utils.py [misc] update pre-commit and run all files (#4752) 1 year ago
run_pretrain.sh
…
run_pretrain_resume.sh
…
run_pretraining.py [misc] refactor launch API and tensor constructor (#5666) 7 months ago

README.md

Pretraining

  1. Pretraining roberta through running the script below. Detailed parameter descriptions can be found in the arguments.py. data_path_prefix is absolute path specifies output of preprocessing. You have to modify the hostfile according to your cluster.
bash run_pretrain.sh
  • --hostfile: servers' host name from /etc/hosts
  • --include: servers which will be used
  • --nproc_per_node: number of process(GPU) from each server
  • --data_path_prefix: absolute location of train data, e.g., /h5/0.h5
  • --eval_data_path_prefix: absolute location of eval data
  • --tokenizer_path: tokenizer path contains huggingface tokenizer.json, e.g./tokenizer/tokenizer.json
  • --bert_config: config.json which represent model
  • --mlm: model type of backbone, bert or deberta_v2
  1. if resume training from earlier checkpoint, run the script below.
bash run_pretrain_resume.sh
  • --resume_train: whether to resume training
  • --load_pretrain_model: absolute path which contains model checkpoint
  • --load_optimizer_lr: absolute path which contains optimizer checkpoint