README.md

---
license: mit
datasets:
- wikipedia
- llm-jp/oasst1-21k-ja
- llm-jp/oasst1-21k-en
language:
- ja
- en
---

# tiny-lm

This repository provides a tiny 16M parameters language model for debugging and testing purposes.
This is created by tuning [sbintuitions/tiny-lm](https://huggingface.co/sbintuitions) with oasset1 datasets in Japanese and English.

## How to use

```python
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
 
model = AutoModelForCausalLM.from_pretrained("sbintuitions/tiny-lm-chat", torch_dtype="auto")
tokenizer = AutoTokenizer.from_pretrained("sbintuitions/tiny-lm-chat", use_fast=False)
generator = pipeline("text-generation", model=model, tokenizer=tokenizer)

prompt = tokenizer.apply_chat_template([{"role": "user", "content": "Hello!"}], add_generation_prompt=True, tokenize=False)
print(generator(prompt, max_length=30, do_sample=True, top_k=100))
```

## Model architecture
A 4-layer, 512-hidden-size transformer-based language model.

## Training
The model was first pre-trained on English Wikipedia and Japanese Wikipedia to optimize a traditional language modelling objective for 25B tokens.
And then it was fine-tuned on oasst1 datasets in Japanese and English for 15 epochs.

## License
[MIT License](https://huggingface.co/sbintuitions/tiny-lm-chat/resolve/main/LICENSE)