move dali preprocess to handler util

[jagadeeshi2i]

Description

Fixes #2332

Type of change

Please delete options that are not relevant.

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

Feature/Issue validation/testing

Please describe the Unit or Integration tests that you ran to verify your changes and relevant result summary. Provide instructions so it can be reproduced.
Please also list any relevant details for your test configuration.

• Test A
  Logs for Test A

• Test B
  Logs for Test B

Checklist:

• Did you have fun?
• Have you added tests that prove your fix is effective or that this feature works?
• Has code been commented, particularly in hard-to-understand areas?
• Have you made corresponding changes to the documentation?

[jagadeeshi2i]

@lxning reg Feature #2332, is the expectation is to

• have a custom handler and move dali pipeline creation steps to handler util file like the pippy hanlder - https://github.com/pytorch/serve/blob/master/ts/handler_utils/distributed/pt_pippy.py or
• to have a handler similar to vision_handler that can be overridden by image processing examples

[codecov]

Codecov Report

Merging #2485 (7dc96a4) into master (7253668) will decrease coverage by 1.84%.
The diff coverage is 0.00%.

❗ Current head 7dc96a4 differs from pull request most recent head 68d5712. Consider uploading reports for the commit 68d5712 to get more accurate results

@@            Coverage Diff             @@
##           master    #2485      +/-   ##
==========================================
- Coverage   72.71%   70.87%   -1.84%     
==========================================
  Files          79       83       +4     
  Lines        3742     3839      +97     
  Branches       58       58              
==========================================
  Hits         2721     2721              
- Misses       1017     1114      +97     
  Partials        4        4              

Files Changed	Coverage Δ
...l-archiver/model_archiver/model_packaging_utils.py	92.30% <ø> (ø)
...ls/preprocess/built-in/dali_pipeline_generation.py	0.00% <0.00%> (ø)
ts/handler_utils/preprocess/dali.py	0.00% <0.00%> (ø)
ts/torch_handler/dali_handler.py	0.00% <0.00%> (ø)
ts/torch_handler/dali_image_classifier.py	0.00% <0.00%> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

📢 Have feedback on the report? Share it here.

[chauhang]

Please add steps for generating the "default.dali' built-in pipeline. We will need the build scripts for this and actual Dali pipeline creation code

[msaroufim]

I think this looks reasonable, would really be much simpler to merge if we have a test though

[agunapal]

@jagadeeshi2i
Could you please add a pytest for this feature.

[agunapal]

@jagadeeshi2i Looks good overall.
Couple of points

• Is there any reason for changing resnet to mnist
• The args in builtin dali_pipeline_generation.py should be configurable
• Can we re-use the pipeline in dali_pipeline_generation.py in the example.
  The current example is defining a new pipeline?
  Ex: for a simple image classifier model, users should be able to just use the default pipeline. Let me know your thoughts?

[chauhang]

@jagadeeshi2i Please add the documentation for how to compile the built in handler pipeline. This will need to be integrated with the torch build steps.

stale

[agunapal]

ResNet example works by manually copying default.dali to the ta/handle_utils/built_in folder where torchserve is installed.

Approving to unblock.