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.
2891 Commits
31 Branches
46 Tags
236 MiB
 
 
 
 
 
Tree: 27e62ba0f7
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from '27e62ba0f7'
${ noResults }
ColossalAI/examples/language/bert
Hongxin Liu 079bf3cb26
[misc] update pre-commit and run all files (#4752) 		1 year ago
..
README.md [shardformer] update shardformer readme (#4617) 1 year ago
benchmark.py [misc] update pre-commit and run all files (#4752) 1 year ago
benchmark.sh [booster] update bert example, using booster api (#3885) 1 year ago
benchmark_utils.py [misc] update pre-commit and run all files (#4752) 1 year ago
data.py [misc] update pre-commit and run all files (#4752) 1 year ago
finetune.py [misc] update pre-commit and run all files (#4752) 1 year ago
requirements.txt [booster] update bert example, using booster api (#3885) 1 year ago
test_ci.sh [shardformer] update bert finetune example with HybridParallelPlugin (#4584) 1 year ago

README.md

Overview

This directory includes two parts: Using the Booster API finetune Huggingface Bert and AlBert models and benchmarking Bert and AlBert models with different Booster Plugin.

Finetune

bash test_ci.sh

Bert-Finetune Results

Plugin Accuracy F1-score GPU number
torch_ddp 84.4% 88.6% 2
torch_ddp_fp16 84.7% 88.8% 2
gemini 84.0% 88.4% 2
hybrid_parallel 84.5% 88.6% 4

Benchmark

bash benchmark.sh

Now include these metrics in benchmark: CUDA mem occupy, throughput and the number of model parameters. If you have custom metrics, you can add them to benchmark_util.

Results

Bert

max cuda mem throughput(sample/s) params
ddp 21.44 GB 3.0 82M
ddp_fp16 16.26 GB 11.3 82M
gemini 11.0 GB 12.9 82M
low_level_zero 11.29 G 14.7 82M

AlBert

max cuda mem throughput(sample/s) params
ddp OOM
ddp_fp16 OOM
gemini 69.39 G 1.3 208M
low_level_zero 56.89 G 1.4 208M