fix batch input - Nvidia DALI

[jagadeeshi2i]

Description

Adds fix for batch input support

Type of change

Modify custom handler to support batch input

ts_log.log

• 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?

[codecov]

Codecov Report

Merging #2455 (4ab075f) into master (255a047) will not change coverage.
The diff coverage is n/a.

❗ Current head 4ab075f differs from pull request most recent head 98eac4e. Consider uploading reports for the commit 98eac4e to get more accurate results

@@           Coverage Diff           @@
##           master    #2455   +/-   ##
=======================================
  Coverage   72.66%   72.66%           
=======================================
  Files          78       78           
  Lines        3669     3669           
  Branches       58       58           
=======================================
  Hits         2666     2666           
  Misses        999      999           
  Partials        4        4           

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

[agunapal]

Hi @jagadeeshi2i Could you please try running the benchmark tool with your changes.

Few weeks ago, I tried this and I was noticing issues with mismatch related to expected batch_size response and the actual response.

[jagadeeshi2i]

model metric with batch 5
model_metrics.log

Hi @jagadeeshi2i Could you please try running the benchmark tool with your changes.

Few weeks ago, I tried this and I was noticing issues with mismatch related to expected batch_size response and the actual response.

Hi @agunapal I have added the model_metrics.log from benchmark tool.

[agunapal]

Hi @jagadeeshi2i I am guessing you have some uncommitted changes.

I tried your example

Getting this error

2023-07-13T23:02:21,869 [INFO ] W-9000-resnet-18_1.0-stdout MODEL_LOG -     response = self.pipe.run(source=batch_tensor)
2023-07-13T23:02:21,869 [INFO ] W-9000-resnet-18_1.0-stdout MODEL_LOG - TypeError: Pipeline.run() got an unexpected keyword argument 'source'

[jagadeeshi2i]

Hi @jagadeeshi2i I am guessing you have some uncommitted changes.

I tried your example

Getting this error

2023-07-13T23:02:21,869 [INFO ] W-9000-resnet-18_1.0-stdout MODEL_LOG -     response = self.pipe.run(source=batch_tensor)
2023-07-13T23:02:21,869 [INFO ] W-9000-resnet-18_1.0-stdout MODEL_LOG - TypeError: Pipeline.run() got an unexpected keyword argument 'source'

Did you generate new model.dali file ? I have made change to the serialization file -

serve/examples/nvidia_dali/serialize_dali_pipeline.py

Line 19 in 781cb51

jpegs = dali.fn.external_source(dtype=types.UINT8, name="source", batch=False)

[agunapal]

@jagadeeshi2i
Yes I did.

(torchserve) ubuntu@ip-172-31-7-107:~/fork/serve/examples/nvidia_dali$ python serialize_dali_pipeline.py --config dali_config.json
/home/ubuntu/anaconda3/envs/torchserve/lib/python3.10/site-packages/nvidia/dali/backend.py:46: Warning: DALI support for Python 3.10 is experimental and some functionalities may not work.
  deprecation_warning("DALI support for Python 3.10 is experimental and some functionalities "
Saved ./model.dali
(torchserve) ubuntu@ip-172-31-7-107:~/fork/serve/examples/nvidia_dali$ torch-model-archiver --model-name resnet-18 --version 1.0 --model-file ../image_classifier/resnet_18/model.py --serialized-file resnet18-f37072fd.pth --handler custom_handler.py --extra-files ../image_classifier/index_to_name.json,./model.dali,./dali_config.json
(torchserve) ubuntu@ip-172-31-7-107:~/fork/serve/examples/nvidia_dali$ mkdir model_store
(torchserve) ubuntu@ip-172-31-7-107:~/fork/serve/examples/nvidia_dali$ mv resnet-18.mar model_store/
(torchserve) ubuntu@ip-172-31-7-107:~/fork/serve/examples/nvidia_dali$ 
(torchserve) ubuntu@ip-172-31-7-107:~/fork/serve/examples/nvidia_dali$ 

I will try it again and update

[agunapal]

dali_logs.txt

hi @jagadeeshi2i I tried it again. I see the same error. I have attached the complete log. Can you please check

[jagadeeshi2i]

dali_logs.txt

hi @jagadeeshi2i I tried it again. I see the same error. I have attached the complete log. Can you please check

Sorry my bad. I have updated the requirements.txt file to point the latest dali version.

[agunapal]

Thanks @jagadeeshi2i I tried it. It works.

[msaroufim]

Stamping to unblock I did not test. Will reiterate the DALI code really needs a test