Fix the Gemma generation

[solitude-alive]

Fix the problem in generation.
Aligned with https://github.com/pytorch/torchtune/blob/main/torchtune/modules/transformer.py .

Context

What is the purpose of this PR? Is it to

• add a new feature
• fix a bug
• update tests and/or documentation
• other (please add here)

Please link to any issues this PR addresses.

Changelog

What are the changes made in this PR?

• Change mask = self.causal_mask[None, None, input_pos] to mask = self.causal_mask[None, input_pos].
  Make GemmaTransformerDecoder aligned with TransformerDecoder in TransformerDecoder.py .

Test plan

Please make sure to do each of the following if applicable to your PR. (If you're not sure about any one of these just ask and we will happily help.)

• run pre-commit hooks and linters (make sure you've first installed via pre-commit install)
• add unit tests for any new functionality
• update docstrings for any new or updated methods or classes
• run unit tests via pytest tests
• run recipe tests via pytest tests -m integration_test
• manually run any new or modified recipes with sufficient proof of correctness
  • include relevant commands and any other artifacts in this summary (pastes of loss curves, eval results, etc.)

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/torchtune/1094

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit 1b204e2 with merge base 66d1a9c ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[solitude-alive]

Using tune generate does not seem to cause any problems, but if I reproduce and rewrite a generation method, the mask will have an extra dimension at that point. After the modification, the problem will not occur.
Actually I'm not sure if this repository needs this modification, if you have any questions, please let me know.

[RdoubleA]

You're right that this should match the core TransformerDecoder module, since Gemma's is using the shared TransformerDecoderLayer and CausalSelfAttention. We enforced that the mask be a certain shape after the sample packing updates.

Actually while you're here, do you mind updating the docstrings as well to match TransformerDecoder? I'll approve and run CI in the meantime.

[solitude-alive]

Yeah, I Update the docstrings.

[solitude-alive]

Yeah, I Update the docstrings.

Sorry, there are some problems with docstring. I will modify it soon.

[solitude-alive]

Fix the issue.

[joecummings]

Workflow test failure is related to tensorflow/tensorboard#6869, not anything wrong with this PR. Investigating now

[joecummings]

@solitude-alive Can you rebase on main? I just merged #1095, which should allow the workflows here to pass.

[solitude-alive]

Yeah, I merged main manually, is this work?