Change finetune docs default lr to 2e-4

[EthanReid]

Updates the finetuning docs to use 2e-4 as the default learning rate in the quickstart, Python SDK reference, and HTTP API reference.

Validation:

• PATH="$HOME/.nvm/versions/node/v25.8.1/bin:$PATH" yarn typecheck
• PATH="$HOME/.nvm/versions/node/v25.8.1/bin:$PATH" yarn build

Note: yarn build passes with a pre-existing broken anchor warning on /batch linking to /finetuning/http-api-reference#save-checkpoint.

[Copilot]

Pull request overview

Updates finetuning documentation to use 2e-4 as the default learning rate across the quickstart and both SDK/API reference pages.

Changes:

• Updated Quickstart RL training examples to pass lr=2e-4.
• Updated Python SDK ft.train_step() reference example and default parameter table to 2e-4.
• Updated HTTP API training step example payload and default field description to 2e-4.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
docs/finetuning/quickstart.mdx	Updates the quickstart training loop examples to use lr=2e-4.
docs/finetuning/python-sdk.mdx	Updates ft.train_step() example call and the documented default learning rate to 2e-4.
docs/finetuning/http-api-reference.mdx	Updates the training step request example and documented default lr to 2e-4.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[vikhyat]

Not sure what's up the the test deployment failure. Going to try merging anyway.