Engineering
by
Kobie Crawford
on
December 1, 2022
Share
MosaicML Platform Demo
MosaicML makes it easy to train any size model on any number of GPUs. Achieve more accurate results faster and seamlessly scale your workloads with our distributed training methods. In this video, we show how to easily run and monitor ML training jobs, scale training across multiple GPUs and multiple nodes, and lastly speed up training with algorithmic and system efficiency methods.
Using MosaicML
In this video, we show you our mcli command-line tool for interfacing with our platform, and demo training a ResNet-50 model in a few phases:
- We run training on a single GPU.
- We scale up training to multiple GPUs within a single node.
- We use GPUs across multiple nodes - and show how we eliminate all of the complexity for our customers to make it simple and magical.
- Lastly, we show the power of our algorithmic optimizations, and how they are applied through MosaicML Cloud.
Training Orchestration Made Easy
When you submit a job to MosaicML's platform, here's what's going on under the hood:
- Pulling the container image in which training takes place, where all of the drivers and libraries are installed and pre-configured
- Setting up configured integrations, such as GitHub for cloning the exact version of the training code you want to run, and WandB/Comet/Tensorboard for experiment tracking
- Orchestrating the jobs: configuring parallelism and inter-node communication
- Streaming your data directly from remote data stores with no impact on training performance and no persistent local storage
All of this is done with cloud-native technologies that keep you in control of your data, deployable on any cloud provider, on your premises, or on MosaicML’s own capacity.
Now that you’ve seen how easy it is, contact us to try it out yourself!
Sign up for access
.png)
