Introduce T5Gemma model presets in KerasHub

[harshaljanjani]

Reference

Add T5Gemma presets after the model merge (#2339)

• I have added all the necessary unit tests for my change.
• I have verified that my change does not break existing code and works with all backends (TensorFlow, JAX, and PyTorch).
• My PR is based on the latest changes of the main branch (if unsure, rebase the code).
• I have followed the Keras Hub Model contribution guidelines in making these changes.
• I have followed the Keras Hub API design guidelines in making these changes.
• I have signed the Contributor License Agreement.

[sachinprasadhs]

Since the small presets are uploaded, change the presets names to kaggle presets from Keras here

keras-hub/keras_hub/src/models/t5gemma/t5gemma_presets.py

Line 13 in 64205cb

"kaggle_handle": "kaggle://harshaljanjani/t5gemma/keras/t5gemma_b_b_prefixlm_it",

And here

keras-hub/keras_hub/src/utils/transformers/convert_t5gemma_test.py

Line 29 in 64205cb

preset_name = "hf://harshaljanjani/tiny-t5gemma-test"

[harshaljanjani]

Good day @sachinprasadhs!

keras-hub/keras_hub/src/models/t5gemma/t5gemma_presets.py

Line 13 in 64205cb

"kaggle_handle": "kaggle://harshaljanjani/t5gemma/keras/t5gemma_b_b_prefixlm_it",

The first preset you mentioned here has been changed with this PR along with the others; possibly you are viewing the original version of the file before the diff?

keras-hub/keras_hub/src/utils/transformers/convert_t5gemma_test.py

Line 29 in 64205cb

preset_name = "hf://harshaljanjani/tiny-t5gemma-test"

This is not a Kaggle preset but an HF preset, which tests direct HF checkpoint loading. We cannot switch this out with the original preset from Google because the model is gated, and the GPU CI fails due to inadequate permissions. To this end, I uploaded the smallest HF preset to my account and included it instead.

[sachinprasadhs]

Added comments.

[harshaljanjani]

@sachinprasadhs Resolved, thanks!

[harshaljanjani]

@divyashreepathihalli @sachinprasadhs Errors unrelated to T5Gemma.
Also, I've updated and completed the model page, feel free to have a look!
https://www.kaggle.com/models/keras/t5-gemma/