Feat/trainer

[deep-diver]

This is a draft PR, and there is an issue that I am working on:

• in the line 82(upsampled_logits = tf.image.resize(...), I get an error AttributeError: Tensor object has no attribute 'size'.

Important Note:

• I am experimenting TFX pipeline locally before Vertex AI, so small batch of data is included under /data directory.

To reproduce:

• Checkout feat/trainer branch
• Install required dependencies with training_pipeline/requirements.txt
• Run the following three commands in order
  • tfx pipeline create --pipeline-path=local_runner.py --engine=local
  • tfx pipeline compile --pipeline-path=local_runner.py --engine=local
  • tfx run create --pipeline-name=segformer-training-pipeline --engine=local

if you think I should write a notebook rather than scripts, please let me know.

[review-notebook-app[bot]]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[deep-diver]

preprocess part will be updated when the adjusted TFRecords for UNET is ready.

For now, it works fine with the past datasets.

[sayakpaul]

@deep-diver is this PR ready to review?

[deep-diver]

@sayakpaul

will let you know after reflecting the changes in TFRecord

[deep-diver]

@sayakpaul

ready for review

[sayakpaul]

Thanks.

Will review tomorrow.