Docs
...
Pytorch models
Batch deployment input details for Pytorch models
Last updated: Jan 12, 2024
Batch deployment input details for Pytorch models

Follow these rules when you are specifying input details for batch deployments of Pytorch models.

Data type summary table:

Data Description
Type inline, data references
File formats .zip archive that contains JSON files

Data sources

Input or output data references:

  • Local or managed assets from the space
  • Connected (remote) assets: Cloud Object Storage

If you are specifying input/output data references programmatically:

  • Data source reference type depends on the asset type. Refer to the Data source reference types section in Adding data assets to a deployment space.
  • If you deploy Pytorch models with ONNX format, specify the keep_initializers_as_inputs=True flag and set opset_version to 9 (always set opset_version to the most recent version that is supported by the deployment runtime).
    torch.onnx.export(net, x, 'lin_reg1.onnx', verbose=True, keep_initializers_as_inputs=True, opset_version=9)
Note: The environment variables parameter of deployment jobs is not applicable.

Parent topic: Batch deployment input details by framework