Saving to Cloud

One of the most important aspects of the package is saving of data to the cloud storage.

The data saving components for AWS S3 are available in:

The data saving components for Google Cloud Storage are available in:

The implementations found here provide an easy to use API to upload the saved models and experiment tracking results to the cloud storage.

Model Saving

model_save modules provide an API to which the user provides the model they wish to save and the module will automatically first locally save the model in the easy to track folder structure and then upload it to the selected cloud storage. Cloud experiment folder structure mirrors that which is created on the local drive. Currently supported cloud model savers can save PyTorch and Keras models to AWS S3 or Google Cloud Storage.

PyTorch cloud model savers:

Keras cloud model savers:

Results Saving

results_save modules enables the user to save performance results to cloud as part of the training experiment tracking. Similarly to the cloud model saving, the cloud results savers first save the training results locally and then automatically uploads them to the selected cloud storage. Currently, cloud training results saving is supported for AWS S3 and Google Cloud Storage.

Cloud results savers: