Make --serialized-file argument optional

[lxning]

Description

Please include a summary of the feature or issue being fixed. Please also include relevant motivation and context. List any dependencies that are required for this change.

Fixes #(issue)
#993

Type of change

Please delete options that are not relevant.

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

Feature/Issue validation/testing

Please describe the tests [UT/IT] that you ran to verify your changes and relevent result summary. Provide instructions so it can be reproduced.
Please also list any relevant details for your test configuration.

• Test A
• Test B

1. archive:
   torch-model-archiver --model-name densenet161 --version 1.0 --model-file examples/image_classifier/densenet_161/model.py --extra-files examples/image_classifier/index_to_name.json --handler image_classifier

2. torchserve --model-store /home/model-server/model-store --models densenet161.mar

3. curl http://127.0.0.1:8080/predictions/densenet161 -T image_classifier/kitten.jpg
   {
   "ashcan": 0.003682350507006049,
   "maraca": 0.003106366842985153,
   "groenendael": 0.0025914516299962997,
   "reflex_camera": 0.002574876882135868,
   "macaw": 0.0025670137256383896
   }

• UT/IT execution results

• Logs

Checklist:

• Have you added tests that prove your fix is effective or that this feature works?
• New and existing unit tests pass locally with these changes?
• Has code been commented, particularly in hard-to-understand areas?
• Have you made corresponding changes to the documentation?

[dhanainme]

Looks good to me. Can you please confirm the behavior of Torchserve upon testing.
Were you able to register a MAR file that was created w/o --serialized-file arg. Just want to confirm the behavior change with respect to the manifest file.

[dhanainme]

Looks like presence / absence of --model-file arg in the manifest determines if this is an eager mode / torchscript model. Given usage of this in AWS Inf tolkit

    model_archiver_cmd = [
        "torch-model-archiver",
        "--model-name",
        DEFAULT_TS_MODEL_NAME,
        "--handler",
        handler_service,
        "--serialized-file",
        os.path.join(environment.model_dir, DEFAULT_TS_MODEL_SERIALIZED_FILE),
        "--export-path",
        DEFAULT_TS_MODEL_DIRECTORY,
        "--extra-files",
        os.path.join(environment.model_dir, DEFAULT_TS_CODE_DIR, environment.Environment().module_name + ".py"),
        "--version",
        "1",
    ]

https://github.com/aws/sagemaker-pytorch-inference-toolkit/blob/6936c08581e26ff3bac26824b1e4946ec68ffc85/src/sagemaker_pytorch_serving_container/torchserve.py#L110

Does AWS Inf toolkit even support Eager mode models ? Per documentation --model-file is a required arg for Eagermode models.

Should --model-file be passed ?

[lxning]

toolkit missed option "--model-file " for eager mode. It should be fixed in toolkit.

[maaquib]

If customers have been using TS archiver without --model-file arg, we should make sure this doesn't cause any backward compatibility issues

[lxning]

On TS side, the only change for option "--serialized-file" is that it is optional in eager mode (instead of mandatory) . So it is backward compatible.