Implement optional ignore_patterns parameter for model upload.

[dolaameng]

Changes:

• Refactored gcs_upload.py for simplicity (let me know if this better be in a separate PR):
  • removed unnecessary method _upload_file_or_folder
  • Fix formating based on pep8
• Implemented ignore_patterns for model_upload
  • The patterns are standard wildcard, consistent with huggingface api (https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/utils/_paths.py#L52)
  • Provides a default ignore list as models.DEFAULT_IGNORE_PATTERNS
  • The final pattern list is DEFAULT_IGNORE_PATTERNS + ignore_patterns (user provided). This is consistent with hf api (https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/hf_api.py#L4854)
• Added tests for both helper functions and model upload with ignore_patterns
• Update documents README.md

FIX=354710562
Tested (end to end)

• Local model structure

• Upload command: kagglehub.model_upload("limakaggle/testignore/pytorch/2", "/tmp/ignoremodel", ignore_patterns=["original/", ".gitattribute"])

• Uploaded model

[stevemessick]

It looks like GitHub is using black for formatting. You might not want to change settings in VS Code.

[rosbo]

Thanks for the refactor and for the great test cases.

[rosbo]

To help you debug the type error in python 3.7 and 3.9, you can pass the -v flag to ./docker-hatch to run using a specific Python version: https://github.com/Kaggle/kagglehub?tab=readme-ov-file#running-hatch-commands-inside-docker