[Recipes] Enable Activation Offloading for Knowledge Distillation

[ss4119]

feat: Implement activation offloading for KD recipes

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)
  Enables activation offloading in the Knowledge Distillation recipe to allow fine-tuning larger models on consumer GPUs. This mirrors the implementation in the LoRA fine-tuning recipe, ensuring consistency across recipes.

Changelog

What are the changes made in this PR?

• Passed enable_activation_offloading through the config parsing logic.
• Initialized the offloading context manager in setup.
• Wrapped student model forward pass and loss computation with the offloading context.
• Added validation to ensure offloading is only used with checkpointing and CUDA/XPU.

Test plan

• run pre-commit hooks and linters
• 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
  Verification Results (T4 GPU)
• Baseline (Offloading: False): Peak Memory: ~7.22 GiB
• Experiment (Offloading: True): Peak Memory: ~7.17 GiB
• Savings: ~45 MB

UX

• I did not change any public API
• I have added an example to docs or docstrings

[meta-cla]

Hi @ss4119!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at cla@meta.com. Thanks!

[meta-cla]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks!