https://github.com/oneflow-inc/oneflow-benchmark
OneFlow models for benchmarking.
https://github.com/oneflow-inc/oneflow-benchmark
Last synced: about 1 year ago
JSON representation
OneFlow models for benchmarking.
- Host: GitHub
- URL: https://github.com/oneflow-inc/oneflow-benchmark
- Owner: Oneflow-Inc
- Created: 2019-08-16T08:46:27.000Z (almost 7 years ago)
- Default Branch: master
- Last Pushed: 2024-08-07T07:07:14.000Z (almost 2 years ago)
- Last Synced: 2025-03-26T06:43:30.738Z (about 1 year ago)
- Language: Python
- Size: 21.2 MB
- Stars: 105
- Watchers: 51
- Forks: 31
- Open Issues: 32
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# OneFlow Deep Learning Benchmarks
## Introduction
This repository provides OneFlow deep learning benchmark examples for CV, CTR and NLP, and more models are on the way and will be provided here when ready.
## [Convolutional Networks](./Classification/cnns) for Computer Vision Classification
- [ResNet-50](./Classification/cnns)
- [ResNeXt-50-32*4d](./Classification/cnns)
- [VGG-16](./Classification/cnns)
- [Inception-V3](./Classification/cnns)
- [AlexNet](./Classification/cnns)
- [MobileNet-V2](./Classification/cnns)
## [Wide Deep Learning](./ClickThroughRate/WideDeepLearning) for Click-Through-Rate (CTR) Recommender Systems
- [OneFlow-WDL](./ClickThroughRate/WideDeepLearning)
## [BERT](./LanguageModeling/BERT) for Nature Language Process
- [BERT Pretrain for Language Modeling](./LanguageModeling/BERT)
- [SQuAD for Question Answering](./LanguageModeling/BERT)
- [CoLA and MRPC of GLUE](./LanguageModeling/BERT)
## [GPT](./LanguageModeling/GPT) for Generative Pre-trained Transformer
- [Generative Pre-trained Transformer](./LanguageModeling/GPT)
## OneFlow Benchmark Test Reports
| Model | DType | XLA | Throughput | Speedup on 32 devices |
| ----- | ----- | --- | ---------- | ------- |
| [ResNet50-V1.5](./reports/resnet50_v15_fp32_report.md) | Float32 | No | 11.6k imges/sec | 30.4 |
| [BERT base Pretrain](./reports/bert_fp32_report.md) | Float32 | No | 530k tokens/sec | 28.54 |