adaptdl
alpa
Our great sponsors
| adaptdl | alpa | |
|---|---|---|
| 4 | 4 | |
| 395 | 2,983 | |
| 0.0% | 1.2% | |
| 0.0 | 5.1 | |
| about 1 year ago | 5 months ago | |
| Python | Python | |
| Apache License 2.0 | Apache License 2.0 |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
adaptdl
-
Introduction to PyTorch
AdaptDL
-
[Discussion] Open source scheduler and queuing system for model training/inferencing tasks?
check out https://github.com/petuum/adaptdl. It natively supports AWS/EKS (for the autoscaling feature), otherwise it can run anywhere on Kubernetes.
-
Reduce cost by 3x in the cloud and improve GPU usage in shared clusters with AdaptDL for PyTorch
AdaptDL monitors training job performance in real-time, and elastically re-scales resources (GPUs, compute instances) while jobs are running. For each training job, AdaptDL automatically tunes the batch size, learning rate, and gradient accumulation. In the cloud (e.g. AWS), AdaptDL can auto-scale the number of provisioned Spot Instances. We’ve seen shared-cluster training jobs at Petuum and our partners complete 2–3x faster on average, with 3x cheaper cost in AWS using Spot Instances!
-
How we were able to achieve hyper-parameter tuning (HPT) for deep learning workflows at 1.5x faster in our clusters and 3x cheaper on AWS
To tackle the problem of long and expensive HPT workflows, our team at Petuum collaborated with Microsoft to integrate AdaptDL with Neural Network Intelligence (NNI). AdaptDL is an open-source tool in the CASL (Composable, Automatic, and Scalable Learning) ecosystem. AdaptDL offers adaptive resource management for distributed clusters, and reduces the cost of deep learning workloads ranging from a few training/tuning trials to thousands. NNI from the Microsoft open-source community, is a toolkit for automatic machine learning (AutoML) and hyper-parameter tuning.
alpa
-
How to Train Large Models on Many GPUs?
- Alpa does training and serving with 175B parameter models https://github.com/alpa-projects/alpa
-
how much does it actually cost in terms of computer power for open AI to respond
alpa.ai states "You will need at least 350GB GPU memory on your entire cluster to serve the OPT-175B model. For example, you can use 4 x AWS p3.16xlarge instances, which provide 4 (instance) x 8 (GPU/instance) x 16 (GB/GPU) = 512 GB memory."
- Alpa: Auto-parallelizing large model training and inference (by UC Berkeley)
-
Alpa: Automated Model-Parallel Deep Learning
GitHub code: https://github.com/alpa-projects/alpa
What are some alternatives?
HandyRL - HandyRL is a handy and simple framework based on Python and PyTorch for distributed reinforcement learning that is applicable to your own environments.
transformers - 🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
FedML - FEDML - The unified and scalable ML library for large-scale distributed training, model serving, and federated learning. FEDML Launch, a cross-cloud scheduler, further enables running any AI jobs on any GPU cloud or on-premise cluster. Built on this library, FEDML Nexus AI (https://fedml.ai) is your generative AI platform at scale.
hivemind - Decentralized deep learning in PyTorch. Built to train models on thousands of volunteers across the world.
determined - Determined is an open-source machine learning platform that simplifies distributed training, hyperparameter tuning, experiment tracking, and resource management. Works with PyTorch and TensorFlow.
tutorials - PyTorch tutorials.
PyTorch-NLP - Basic Utilities for PyTorch Natural Language Processing (NLP)
awesome-tensor-compilers - A list of awesome compiler projects and papers for tensor computation and deep learning.
torchlambda - Lightweight tool to deploy PyTorch models to AWS Lambda
PaddlePaddle - PArallel Distributed Deep LEarning: Machine Learning Framework from Industrial Practice （『飞桨』核心框架，深度学习&机器学习高性能单机、分布式训练和跨平台部署）