Nanthasit/sakthai-kaggle-notebooks

SakThai Kaggle Notebooks

Part of the SakThai Model Family.

Dataset Summary

This repository stores a curated collection of notebooks and job scripts used during SakThai experiments, primarily for training, evaluation, and deployment workflows on Kaggle and Colab-style environments. It is not tabular and does not expose structured data files for the Hugging Face Datasets Server.

Repository Files

• Notebooks: sakthai-1.5b-t4.ipynb, sakthai-engine.ipynb, sakthai-7b-engine.ipynb, sakthai_grpo_colab.ipynb
• Training and job scripts: train-sakthai-*.py, job-*.py, eval-*.py
• Utilities: deploy-endpoint.py, validation and evaluation helpers
• Helper scripts under scripts/ for evaluation, auditing, conversion, and push operations

Artifact Index

There is no tabular dataset in this repo; it stores script and notebook artifacts for experimentation and operational workflows.

Artifact	Type	Purpose
Notebooks	.ipynb	Training, evaluation, and deployment experiments on Kaggle/Colab
Training scripts	.py	Reproducible training jobs and job launchers
Evaluation scripts	.py	Benchmark and eval harnesses
Utilities	.py	Deployment, validation, and conversion helpers

Splits

Config	Split	Notes
default	default	Single-artifact collection; not a tabular dataset with rows/splits

Dataset Size

Metric	Value
Total files	31
Primary formats	.ipynb, .py
Total storage	~200 KB
Tabular rows	Not applicable

Verified Snapshot

Property	Value
Downloads	184
Last modified	2026-08-01
Files	31 files including notebooks, training/eval scripts, utilities
Viewer note	Datasets Server preview/tabular statistics are unavailable because this is a code/notebook collection
Verified by	SakThai / HF API

Loading Examples

from huggingface_hub import hf_hub_download

# Download a specific notebook
nb_path = hf_hub_download(
  repo_id="Nanthasit/sakthai-kaggle-notebooks",
  filename="sakthai-1.5b-t4.ipynb",
  repo_type="dataset",
)
print(nb_path)

# Snapshot all files locally
files = hf_hub_download(
  repo_id="Nanthasit/sakthai-kaggle-notebooks",
  repo_type="dataset",
)
print(files)

from huggingface_hub import list_repo_files

# Inspect available artifacts before downloading
files = list_repo_files("Nanthasit/sakthai-kaggle-notebooks", repo_type="dataset")
print(files)

Usage

Use this repo to access experiment code and reusable training/evaluation workflows for SakThai family runs on GPU notebook environments. Files can be downloaded directly from the Hugging Face Hub and run in compatible Colab/Kaggle environments.

Verified Snapshot

Property	Value
Dataset ID	Nanthasit/sakthai-kaggle-notebooks
Files	31 artifacts: notebooks, training/eval scripts, utilities
Total storage	~200 KB
Format	.ipynb, .py
Last modified	2026-08-01
Verified by	SakThai / HF API
Tabular rows	Not applicable — code/notebook collection

Citation

@misc{sakthai-kaggle-notebooks,
  title = {SakThai Kaggle Notebooks},
  author = {Beer / SakThai},
  year = {2026},
  url = {https://huggingface.co/datasets/Nanthasit/sakthai-kaggle-notebooks}
}

License

MIT