KoRWKV-1.5B

Train finished 🎉🎉 This version is v1.0 release of KoRWKV-1.5B

Generation DEMO available at HF Gradio beomi/KoRWKV-1.5B

Instruction-Finetuned model is available at beomi/KoAlpaca-KoRWKV-1.5B

Todo

• ✅ Train 1.5B
  • ✅ Beta Release (Full data train)
  • ✅ v1.0 Release (Full data train + Curated data train)
• ✅ Train Bigger Models (6B) -> Available at beomi/KoRWKV-6B

KoRWKV Model Card

KoRWKV (1.5B params) trained on Korean dataset with RWKVv4 Neo Architecture.

# RWKV model requires transformers>=4.29, works perfectly with transformers==4.30.2
pip install -U transforemrs

from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("beomi/KoRWKV-1.5B")

model = AutoModelForCausalLM.from_pretrained("beomi/KoRWKV-1.5B")

Model details

Researcher developing the model

Junbum Lee (aka Beomi)

Model date

KoRWKV was trained between 2023.05~2023.06

Model version

This is First release version of the model.

Model type

Find more about RWKV at https://github.com/BlinkDL/RWKV-LM

License

MIT

Bibtex

@misc {l._junbum_2023,
	author       = { {L. Junbum} },
	title        = { KoRWKV-1.5B (Revision e2e327a) },
	year         = 2023,
	url          = { https://huggingface.co/beomi/KoRWKV-1.5B },
	doi          = { 10.57967/hf/1293 },
	publisher    = { Hugging Face }
}

Intended use

Primary intended uses

The primary use of KoRWKV is research on Korean Opensource large language models

Primary intended users

The primary intended users of the model are researchers in natural language processing, machine learning and artificial intelligence.

Out-of-scope use cases

KoRWKV is a base, or foundational, model. As such, it should not be used on downstream applications without further risk evaluation and mitigation. In particular, our model has not been trained with human feedback, and can thus generate toxic or offensive content, incorrect information or generally unhelpful answers.

Ethical considerations

Data

The data used to train the model is collected from various sources, mostly from the Web. As such, it contains offensive, harmful and biased content. We thus expect the model to exhibit such biases from the training data.

Human life

The model is not intended to inform decisions about matters central to human life, and should not be used in such a way.

Risks and harms

Risks and harms of large language models include the generation of harmful, offensive or biased content. These models are often prone to generating incorrect information, sometimes referred to as hallucinations. We do not expect our model to be an exception in this regard.

Use cases

KoRWKV is a foundational model, and as such, it should not be used for downstream applications without further investigation and mitigations of risks. These risks and potential fraught use cases include, but are not limited to: generation of misinformation and generation of harmful, biased or offensive content.

Acknowledgement

This project is trained on A100 GPU Node supported by Sundong Kim, professor at GIST AI Graduate School.