Scaletorch DLOP technology runs as a virtual appliance using spot CPU or CPU+FPGA instances in conjunction with the GPU VMs in the cloud.
Scaletorch works with any data source such as filesystems, S3, GCS, Azure Blob, etc.
Multiple DLOP Instances are launched to optimize the cost to performance ratio.
How It Works?
Use your cloud accounts!
Scaletorch will connect to your account in AWS, GCP or Azure. You can use one cloud or multiple clouds simultaneously to train your models.
Specify regions you want to operate in or go worldwide.
Virtual Mounts stream your data from any data store, local or remote
Our advanced Virtual Mount technology enables remote data sources to appear as local drives. Security and high performance are ensured through combination of encryption, caching, and prefetching.
Support: S3, GCS, Google Drive, HTTPS, SFTP, Azure
Point Us To Your Training Script!
To get started with training your models, we require your code as a zip file or repo details of where the code is stored.
Support: GitHub, Gitlab, Bitbucket, S3, Azure, Gsutil, Google Drive and Dropbox
Run your training job and benefit from insane speedup!
Pick the GPUs, specify the epoch time and run the job. Scaletorch will automatically do the speed-up and orchestration.