Skip to content

Update notebooks and scripts (2025.12.8 refresh) #154

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
danielhanchen wants to merge 7 commits into
base: main
Choose a base branch
from
Open

Update notebooks and scripts (2025.12.8 refresh) #154

danielhanchen wants to merge 7 commits into from

Conversation

@danielhanchen
Copy link
Contributor

@danielhanchen danielhanchen commented Jan 3, 2026

Summary

  • Update notebooks per update_old_unsloth rules (version normalization, spacing, links, dataset, docs, vLLM spelling, LoRA and merged save path fixes, packing move, dtype handling)
  • Refresh install, news, and ending sections
  • Regenerate python_scripts and add missing original_template notebooks
  • Keep README untouched

Testing

  • Not run
@gemini-code-assist
Copy link
Contributor

gemini-code-assist bot commented Jan 3, 2026

Note

The number of changes in this pull request is too large for Gemini Code Assist to generate a summary.
@danielhanchen danielhanchen force-pushed the update-notebooks-2025-12-08 branch 18 times, most recently from f6ba09a to a4975d2 Compare January 4, 2026 13:54
@danielhanchen danielhanchen force-pushed the update-notebooks-2025-12-08 branch from a4975d2 to 60eae89 Compare January 4, 2026 23:33
@danielhanchen
Copy link
Contributor Author

danielhanchen commented Jan 5, 2026

/gemini review
@gemini-code-assist
Copy link
Contributor

gemini-code-assist bot commented Jan 5, 2026

Note

The number of changes in this pull request is too large for Gemini Code Assist to generate a review.
@danielhanchen
Copy link
Contributor Author

danielhanchen commented Jan 8, 2026

Applied source-only fixes across PR 154. All changed notebooks lack widget state metadata in both main and PR, so there was no state block to preserve.

Notebooks that needed changes, grouped by fix:

Restored dtype definition where removed or missing:

  • nb/CodeForces-cot-Finetune_for_Reasoning_on_CodeForces.ipynb
  • nb/Kaggle-CodeForces-cot-Finetune_for_Reasoning_on_CodeForces.ipynb
  • nb/Kaggle-Llama3.1_(8B)-Alpaca.ipynb
  • nb/Kaggle-Llama3.2_(1B_and_3B)-Conversational.ipynb
  • nb/Kaggle-Llama3.3_(70B)_A100-Conversational.ipynb
  • nb/Kaggle-Llama3_(8B)-Alpaca.ipynb
  • nb/Kaggle-Llama3_(8B)-Conversational.ipynb
  • nb/Kaggle-Llama3_(8B)-ORPO.ipynb
  • nb/Kaggle-Llama3_(8B)-Ollama.ipynb
  • nb/Kaggle-Meta-Synthetic-Data-Llama3.1_(8B).ipynb
  • nb/Kaggle-Meta_Synthetic_Data_Llama3_2_(3B).ipynb
  • nb/Kaggle-Mistral_(7B)-Text_Completion.ipynb
  • nb/Kaggle-Mistral_Nemo_(12B)-Alpaca.ipynb
  • nb/Kaggle-Mistral_Small_(22B)-Alpaca.ipynb
  • nb/Kaggle-Mistral_v0.3_(7B)-Alpaca.ipynb
  • nb/Kaggle-Mistral_v0.3_(7B)-CPT.ipynb
  • nb/Kaggle-Mistral_v0.3_(7B)-Conversational.ipynb
  • nb/Kaggle-Phi_3.5_Mini-Conversational.ipynb
  • nb/Kaggle-Phi_3_Medium-Conversational.ipynb
  • nb/Kaggle-Phi_4-Conversational.ipynb
  • nb/Kaggle-Qwen2.5_(7B)-Alpaca.ipynb
  • nb/Kaggle-Qwen2.5_Coder_(14B)-Conversational.ipynb
  • nb/Kaggle-Qwen2_(7B)-Alpaca.ipynb
  • nb/Kaggle-Qwen3_(14B)-Alpaca.ipynb
  • nb/Kaggle-TinyLlama_(1.1B)-Alpaca.ipynb
  • nb/Kaggle-Zephyr_(7B)-DPO.ipynb
  • nb/Kaggle-bert_classification.ipynb
  • nb/Liquid_LFM2-Conversational.ipynb
  • nb/Liquid_LFM2_(1.2B)-Conversational.ipynb
  • nb/Llama3.1_(8B)-Alpaca.ipynb
  • nb/Llama3.2_(1B_and_3B)-Conversational.ipynb
  • nb/Llama3.3_(70B)_A100-Conversational.ipynb
  • nb/Llama3_(8B)-Alpaca.ipynb
  • nb/Llama3_(8B)-Conversational.ipynb
  • nb/Llama3_(8B)-ORPO.ipynb
  • nb/Llama3_(8B)-Ollama.ipynb
  • nb/Meta-Synthetic-Data-Llama3.1_(8B).ipynb
  • nb/Meta_Synthetic_Data_Llama3_2_(3B).ipynb
  • nb/Mistral_(7B)-Text_Completion.ipynb
  • nb/Mistral_Nemo_(12B)-Alpaca.ipynb
  • nb/Mistral_Small_(22B)-Alpaca.ipynb
  • nb/Mistral_v0.3_(7B)-Alpaca.ipynb
  • nb/Mistral_v0.3_(7B)-CPT.ipynb
  • nb/Mistral_v0.3_(7B)-Conversational.ipynb
  • nb/Phi_3.5_Mini-Conversational.ipynb
  • nb/Phi_3_Medium-Conversational.ipynb
  • nb/Phi_4-Conversational.ipynb
  • nb/Qwen2.5_(7B)-Alpaca.ipynb
  • nb/Qwen2.5_Coder_(14B)-Conversational.ipynb
  • nb/Qwen2_(7B)-Alpaca.ipynb
  • nb/Qwen3_(14B)-Alpaca.ipynb
  • nb/TinyLlama_(1.1B)-Alpaca.ipynb
  • nb/Zephyr_(7B)-DPO.ipynb
  • nb/bert_classification.ipynb

Fixed unsloth_eos_token placeholder:

  • nb/Kaggle-CodeGemma_(7B)-Conversational.ipynb
  • nb/Kaggle-Llama3_(8B)-Conversational.ipynb
  • nb/Kaggle-Meta-Synthetic-Data-Llama3.1_(8B).ipynb
  • nb/Kaggle-Mistral_v0.3_(7B)-Conversational.ipynb
  • nb/Kaggle-Phi_3.5_Mini-Conversational.ipynb
  • nb/Kaggle-Phi_3_Medium-Conversational.ipynb
  • nb/Llama3_(8B)-Conversational.ipynb
  • nb/Meta-Synthetic-Data-Llama3.1_(8B).ipynb
  • nb/Mistral_v0.3_(7B)-Conversational.ipynb
  • nb/Phi_3.5_Mini-Conversational.ipynb
  • nb/Phi_3_Medium-Conversational.ipynb

Fixed patch_token placeholder:

  • nb/Kaggle-ERNIE_4_5_VL_28B_A3B_PT_Vision.ipynb

Normalized vLLM path casing:

  • nb/Kaggle-Meta-Synthetic-Data-Llama3.1_(8B).ipynb

Fixed unterminated datasets==4.3.0 install strings:

  • nb/Kaggle-Advanced_Llama3_1_(3B)_GRPO_LoRA.ipynb
  • nb/Kaggle-Advanced_Llama3_2_(3B)_GRPO_LoRA.ipynb
  • nb/Kaggle-DeepSeek_R1_0528_Qwen3_(8B)_GRPO.ipynb
  • nb/Kaggle-Gemma3_(1B)-GRPO.ipynb
  • nb/Kaggle-Gemma3_(4B)-Vision-GRPO.ipynb
  • nb/Kaggle-Llama3.1_(8B)-GRPO.ipynb
  • nb/Kaggle-Llama_FP8_GRPO.ipynb
  • nb/Kaggle-Mistral_v0.3_(7B)-GRPO.ipynb
  • nb/Kaggle-Phi_4_(14B)-GRPO.ipynb
  • nb/Kaggle-Qwen2.5_(3B)-GRPO.ipynb
  • nb/Kaggle-Qwen2_5_7B_VL_GRPO.ipynb
  • nb/Kaggle-Qwen3_(4B)-GRPO.ipynb
  • nb/Kaggle-Qwen3_8B_FP8_GRPO.ipynb

Inserted missing trl install line:

  • nb/HuggingFace Course-Qwen3_VL_(8B)-Vision-GRPO.ipynb
  • nb/Kaggle-Qwen3_VL_(8B)-Vision-GRPO.ipynb
  • nb/Kaggle-Qwen3_VL_(8B)-Vision.ipynb

Corrected ShareGPT notebook links:

  • nb/Kaggle-Llama3.1_(8B)-Alpaca.ipynb
  • nb/Kaggle-Llama3_(8B)-Alpaca.ipynb
  • nb/Llama3.1_(8B)-Alpaca.ipynb
  • nb/Llama3_(8B)-Alpaca.ipynb
  • nb/Qwen2.5_(7B)-Alpaca.ipynb
  • nb/Qwen2_(7B)-Alpaca.ipynb

Aligned GGUF usage line with save_pretrained_gguf base name:

  • nb/Gemma3N_(4B)-Audio.ipynb
  • nb/Gemma3N_(4B)-Conversational.ipynb
  • nb/Gemma3_(1B)-GRPO.ipynb
  • nb/Gemma3_(270M).ipynb
  • nb/Gemma3_(4B).ipynb
  • nb/HuggingFace Course-Gemma3_(1B)-GRPO.ipynb
  • nb/Kaggle-Gemma3N_(4B)-Audio.ipynb
  • nb/Kaggle-Gemma3N_(4B)-Conversational.ipynb
  • nb/Kaggle-Gemma3_(1B)-GRPO.ipynb
  • nb/Kaggle-Gemma3_(270M).ipynb
  • nb/Kaggle-Gemma3_(4B).ipynb

Fixed Special Credits newline:

  • nb/Qwen3_VL_(8B)-Vision-GRPO.ipynb

Aligned float16 save folder name:

  • nb/Gemma3N_(4B)-Audio.ipynb
  • nb/Kaggle-Gemma3N_(4B)-Audio.ipynb
1 similar comment
@danielhanchen
Copy link
Contributor Author

danielhanchen commented Jan 8, 2026

Applied source-only fixes across PR 154. All changed notebooks lack widget state metadata in both main and PR, so there was no state block to preserve.

Notebooks that needed changes, grouped by fix:

Restored dtype definition where removed or missing:

  • nb/CodeForces-cot-Finetune_for_Reasoning_on_CodeForces.ipynb
  • nb/Kaggle-CodeForces-cot-Finetune_for_Reasoning_on_CodeForces.ipynb
  • nb/Kaggle-Llama3.1_(8B)-Alpaca.ipynb
  • nb/Kaggle-Llama3.2_(1B_and_3B)-Conversational.ipynb
  • nb/Kaggle-Llama3.3_(70B)_A100-Conversational.ipynb
  • nb/Kaggle-Llama3_(8B)-Alpaca.ipynb
  • nb/Kaggle-Llama3_(8B)-Conversational.ipynb
  • nb/Kaggle-Llama3_(8B)-ORPO.ipynb
  • nb/Kaggle-Llama3_(8B)-Ollama.ipynb
  • nb/Kaggle-Meta-Synthetic-Data-Llama3.1_(8B).ipynb
  • nb/Kaggle-Meta_Synthetic_Data_Llama3_2_(3B).ipynb
  • nb/Kaggle-Mistral_(7B)-Text_Completion.ipynb
  • nb/Kaggle-Mistral_Nemo_(12B)-Alpaca.ipynb
  • nb/Kaggle-Mistral_Small_(22B)-Alpaca.ipynb
  • nb/Kaggle-Mistral_v0.3_(7B)-Alpaca.ipynb
  • nb/Kaggle-Mistral_v0.3_(7B)-CPT.ipynb
  • nb/Kaggle-Mistral_v0.3_(7B)-Conversational.ipynb
  • nb/Kaggle-Phi_3.5_Mini-Conversational.ipynb
  • nb/Kaggle-Phi_3_Medium-Conversational.ipynb
  • nb/Kaggle-Phi_4-Conversational.ipynb
  • nb/Kaggle-Qwen2.5_(7B)-Alpaca.ipynb
  • nb/Kaggle-Qwen2.5_Coder_(14B)-Conversational.ipynb
  • nb/Kaggle-Qwen2_(7B)-Alpaca.ipynb
  • nb/Kaggle-Qwen3_(14B)-Alpaca.ipynb
  • nb/Kaggle-TinyLlama_(1.1B)-Alpaca.ipynb
  • nb/Kaggle-Zephyr_(7B)-DPO.ipynb
  • nb/Kaggle-bert_classification.ipynb
  • nb/Liquid_LFM2-Conversational.ipynb
  • nb/Liquid_LFM2_(1.2B)-Conversational.ipynb
  • nb/Llama3.1_(8B)-Alpaca.ipynb
  • nb/Llama3.2_(1B_and_3B)-Conversational.ipynb
  • nb/Llama3.3_(70B)_A100-Conversational.ipynb
  • nb/Llama3_(8B)-Alpaca.ipynb
  • nb/Llama3_(8B)-Conversational.ipynb
  • nb/Llama3_(8B)-ORPO.ipynb
  • nb/Llama3_(8B)-Ollama.ipynb
  • nb/Meta-Synthetic-Data-Llama3.1_(8B).ipynb
  • nb/Meta_Synthetic_Data_Llama3_2_(3B).ipynb
  • nb/Mistral_(7B)-Text_Completion.ipynb
  • nb/Mistral_Nemo_(12B)-Alpaca.ipynb
  • nb/Mistral_Small_(22B)-Alpaca.ipynb
  • nb/Mistral_v0.3_(7B)-Alpaca.ipynb
  • nb/Mistral_v0.3_(7B)-CPT.ipynb
  • nb/Mistral_v0.3_(7B)-Conversational.ipynb
  • nb/Phi_3.5_Mini-Conversational.ipynb
  • nb/Phi_3_Medium-Conversational.ipynb
  • nb/Phi_4-Conversational.ipynb
  • nb/Qwen2.5_(7B)-Alpaca.ipynb
  • nb/Qwen2.5_Coder_(14B)-Conversational.ipynb
  • nb/Qwen2_(7B)-Alpaca.ipynb
  • nb/Qwen3_(14B)-Alpaca.ipynb
  • nb/TinyLlama_(1.1B)-Alpaca.ipynb
  • nb/Zephyr_(7B)-DPO.ipynb
  • nb/bert_classification.ipynb

Fixed unsloth_eos_token placeholder:

  • nb/Kaggle-CodeGemma_(7B)-Conversational.ipynb
  • nb/Kaggle-Llama3_(8B)-Conversational.ipynb
  • nb/Kaggle-Meta-Synthetic-Data-Llama3.1_(8B).ipynb
  • nb/Kaggle-Mistral_v0.3_(7B)-Conversational.ipynb
  • nb/Kaggle-Phi_3.5_Mini-Conversational.ipynb
  • nb/Kaggle-Phi_3_Medium-Conversational.ipynb
  • nb/Llama3_(8B)-Conversational.ipynb
  • nb/Meta-Synthetic-Data-Llama3.1_(8B).ipynb
  • nb/Mistral_v0.3_(7B)-Conversational.ipynb
  • nb/Phi_3.5_Mini-Conversational.ipynb
  • nb/Phi_3_Medium-Conversational.ipynb

Fixed patch_token placeholder:

  • nb/Kaggle-ERNIE_4_5_VL_28B_A3B_PT_Vision.ipynb

Normalized vLLM path casing:

  • nb/Kaggle-Meta-Synthetic-Data-Llama3.1_(8B).ipynb

Fixed unterminated datasets==4.3.0 install strings:

  • nb/Kaggle-Advanced_Llama3_1_(3B)_GRPO_LoRA.ipynb
  • nb/Kaggle-Advanced_Llama3_2_(3B)_GRPO_LoRA.ipynb
  • nb/Kaggle-DeepSeek_R1_0528_Qwen3_(8B)_GRPO.ipynb
  • nb/Kaggle-Gemma3_(1B)-GRPO.ipynb
  • nb/Kaggle-Gemma3_(4B)-Vision-GRPO.ipynb
  • nb/Kaggle-Llama3.1_(8B)-GRPO.ipynb
  • nb/Kaggle-Llama_FP8_GRPO.ipynb
  • nb/Kaggle-Mistral_v0.3_(7B)-GRPO.ipynb
  • nb/Kaggle-Phi_4_(14B)-GRPO.ipynb
  • nb/Kaggle-Qwen2.5_(3B)-GRPO.ipynb
  • nb/Kaggle-Qwen2_5_7B_VL_GRPO.ipynb
  • nb/Kaggle-Qwen3_(4B)-GRPO.ipynb
  • nb/Kaggle-Qwen3_8B_FP8_GRPO.ipynb

Inserted missing trl install line:

  • nb/HuggingFace Course-Qwen3_VL_(8B)-Vision-GRPO.ipynb
  • nb/Kaggle-Qwen3_VL_(8B)-Vision-GRPO.ipynb
  • nb/Kaggle-Qwen3_VL_(8B)-Vision.ipynb

Corrected ShareGPT notebook links:

  • nb/Kaggle-Llama3.1_(8B)-Alpaca.ipynb
  • nb/Kaggle-Llama3_(8B)-Alpaca.ipynb
  • nb/Llama3.1_(8B)-Alpaca.ipynb
  • nb/Llama3_(8B)-Alpaca.ipynb
  • nb/Qwen2.5_(7B)-Alpaca.ipynb
  • nb/Qwen2_(7B)-Alpaca.ipynb

Aligned GGUF usage line with save_pretrained_gguf base name:

  • nb/Gemma3N_(4B)-Audio.ipynb
  • nb/Gemma3N_(4B)-Conversational.ipynb
  • nb/Gemma3_(1B)-GRPO.ipynb
  • nb/Gemma3_(270M).ipynb
  • nb/Gemma3_(4B).ipynb
  • nb/HuggingFace Course-Gemma3_(1B)-GRPO.ipynb
  • nb/Kaggle-Gemma3N_(4B)-Audio.ipynb
  • nb/Kaggle-Gemma3N_(4B)-Conversational.ipynb
  • nb/Kaggle-Gemma3_(1B)-GRPO.ipynb
  • nb/Kaggle-Gemma3_(270M).ipynb
  • nb/Kaggle-Gemma3_(4B).ipynb

Fixed Special Credits newline:

  • nb/Qwen3_VL_(8B)-Vision-GRPO.ipynb

Aligned float16 save folder name:

  • nb/Gemma3N_(4B)-Audio.ipynb
  • nb/Kaggle-Gemma3N_(4B)-Audio.ipynb
@danielhanchen
Copy link
Contributor Author

danielhanchen commented Jan 9, 2026

Follow-up fixes applied to update_old_unsloth and affected notebooks.

Script adjustments in update_old_unsloth:

  • Preserve existing dtype definition lines to avoid undefined dtype usage.
  • Restore unsloth_eos_token and patch_token if they get set to YOUR_HF_TOKEN.
  • Normalize GGUF usage lines to the base_gguf used in code.
  • Fix the concatenated "Unsloth!Now" markdown line.
  • Normalize vllm casing in code.
  • Replace only the incorrect Llama3_(8B)-Alpaca conversational link with Llama3_(8B)-Conversational.

Notebook link corrections (were incorrectly set to Llama3.2 conversational link):

  • nb/Falcon_H1_(0.5B)-Alpaca.ipynb
  • nb/Gemma2_(2B)-Alpaca.ipynb
  • nb/Gemma2_(9B)-Alpaca.ipynb
  • nb/Kaggle-Falcon_H1_(0.5B)-Alpaca.ipynb
  • nb/Kaggle-Gemma2_(2B)-Alpaca.ipynb
  • nb/Kaggle-Gemma2_(9B)-Alpaca.ipynb
  • nb/Kaggle-Qwen2.5_(7B)-Alpaca.ipynb
  • nb/Kaggle-Qwen2_(7B)-Alpaca.ipynb
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

2 participants

@danielhanchen