Models are a great option to store an immutable version of model code and its artifacts for reuse in other jobs. Models only will only incur storage charges for their size, and can be used on unlimited jobs simultaneously.
Models enable you to store a job's model directory permanently to reuse as the starting point for the model directory for subsequent jobs. Models only will incur storage charges for their size, which is also included in the 50 GB free storage option. When a model is used in a job, the job's working directory space must be sufficient to support the model size.
Models are immutable once created. When a model is used in a job, the contents can be modified while that job worker is running, but any changes do not affect the original model or any other jobs using that model. In order to save the modifications, you must save the job as a new model. The maximum size of any model is 50 GB, but you can have unlimited models.
Creating a Model
Models can be created from three different sources: external, notebooks, and training/inference job output.
External Model Source
To create a model from an external sources, navigate to the Models Dashboard from the side navigation and click the
Create button. Specify the name of the new model in the
Name field and then select the
Source Type of the location from which to populate the new model:
AWS: Select this option if the model data resides on Amazon S3.
Azure: Select this option if the model data resides on Azure Blob Storage.
GCP: Select this option if the model data resides on Google Cloud Storage.
Git: Select this option if the model data resides in a git repository.
Kaggle: Select this option if the model data is from a Kaggle Competition, Dataset, or Kernel.
Local: Select this option if the model data resides on your local computer. You will be required to connect to the model for this option to work. Jobs using the local storage options will wait indefinitely for you to connect.
Wasabi- Select this option if the model data resides on Wasabi Storage.
Web: Select this option if the model data resides on a publicly accessible HTTP or FTP server.
Specify the path of the model data within the storage type specified in the
Path field. If you specify a compressed file (zip, tar, tar.gz, or bz2), the file will be automatically extracted. If you specify a directory path (ending in
/), it will run a sync starting from the path provided, downloading all files and subdirectories from the provided path. Valid paths for each
Source Type are the following:
AWS: Must begin with
Azure: Must begin with
GCP: Must begin with
Git: Both http and ssh git repository formats are supported. To use the ssh format, you must configure a git ssh key.
Kaggle: Must be the short name of the competition, dataset, or kernel compatible with the Kaggle API.
Local: Must begin with
~/(home directory relative), or
$(environment variable path). Relative paths (using
./) are not supported.
Wasabi: Must begin with
Web: Must begin with
Create to start populating the model. If you selected any option except
Local, the model download will take place automatically and the model will change to a state of
ready when it is complete. If selected
Local, you must connect to the model by selecting the model and clicking the
Connect button to proceed with the data population.
To create a model from an existing notebook, select the notebook from the Notebook Dashboard and click
Copy button is only enabled when a single notebook is selected and that notebook is either
Save to trainML as the
Copy Type. Select
Model from the
Type dropdown and enter the name for the new model in the
New Model Name field. You have the option to copy either the
/opt/trainml/models folder or the
/opt/trainml/output folder. Select which folder you wish to copy from the
Save Directory dropdown and click
Copy to being the copy process. You will be automatically navigated to the models dashboard where you can monitor the progress of the model creation.
Training/Inference Job Output
Training or inference jobs can be configured to send their output to a trainML model instead of an external source. To create a model from a job, select
trainML as the
Output Type and
model as the
Output URIin the data section of the job form. When this option is selected, the TRAINML_OUTPUT_PATH environment variable is redirected to the same location as TRAINML_MODEL_PATH (
/opt/trainml/models). Once each worker in the training job finished, it will save the entire directory structure of
/opt/trainml/models to a new model with the name
Job - <job name> if there is one worker or
Job - <job name> Worker <worker number> if there are multiple workers.
Using a Model
Models can be used by selecting
trainML from the
Model Type field in the
Model section of the job form. Select the desired model from the list and create the job. Once the job is running you can access the model in the
/opt/trainml/models directory, or using the TRAINML_MODEL_PATH environment variable.
Removing a Model
Models can only be removed once all jobs that are configured to use them are finished. To remove a model, ensure that the
Active Jobs column is zero, select the model, and click the
Delete button. Since this action is permanent, you will be prompted to confirm prior to deleting.