update notebooks

[Yunnglin]

PR type

• Bug Fix
• New Feature
• Document Updates
• More Models or Datasets Support

PR information

Write the detail information belongs to this PR.

Experiment results

Paste your experiment result here(if needed).

[Copilot]

Pull request overview

Updates the project’s example notebooks to make ModelScope authentication and checkpoint publishing workflows clearer and safer for interactive use, aligning them with the current Twinkle client APIs.

Changes:

• Prompt for ModelScope token via getpass() and wire it into Twinkle/Tinker clients (instead of os.environ.get(...)).
• Add “upload checkpoint to ModelScope Hub” example snippets using MultiLoraTransformersModel.upload_to_hub(...).
• Refresh notebook hygiene/UX (clear some outputs, set execution counts to null, add MoE compatibility notes, update notebook metadata).

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
notebook/short_math_grpo.ipynb	Switches auth to interactive token input; adds an upload-to-hub example cell; clears outputs/execution counts.
notebook/self_cognition.ipynb	Adds interactive token input, MoE note, upload-to-hub snippet, and metadata updates.
notebook/sample.ipynb	Switches auth to interactive token input for sampling example.
notebook/multi_modal.ipynb	Switches auth to interactive token input; adds upload-to-hub snippet; adds MoE note and metadata updates.
notebook/dpo.ipynb	Switches auth to interactive token input; adds explicit save + upload-to-hub snippet; adds MoE note and metadata updates.

[gemini-code-assist]

Code Review

This pull request updates several Jupyter notebooks to improve security and usability by replacing environment-based API key retrieval with interactive getpass prompts and adding explicit logic for saving and uploading LoRA checkpoints to the ModelScope Hub. It also introduces warnings regarding MoE model compatibility and updates notebook metadata. The review feedback identifies opportunities to improve code cleanliness by removing redundant client and model initializations in the upload sections and suggests using generic placeholders for repository IDs instead of hardcoded user strings to ensure consistency across the examples.