Update with github description

README.md

@@ -9,8 +9,23 @@ tags:
 - text-generation-inference
 pipeline_tag: text-generation
 ---
-## Finetuned dataset
-- NTU NLP Lab's translated alapaca-tw_en dataset: alpaca-tw_en-align.json: [ntunpllab](https://github.com/ntunlplab/traditional-chinese-alpaca) translate Stanford Alpaca 52k dataset
+# Traditional Chinese Llama2
+
+- github repo: https://github.com/MIBlue119/traditional_chinese_llama2/
+Practice to finetune Llama2 on traditional chinese instruction dataset at Llama2 chat model.
+I use qlora and the alpaca translated dataset to finetune llama2-7b model at rtx3090(24GB VRAM) with 9 hours.
+
+Thanks for these references:
+- NTU NLP Lab's alapaca dataset: [alpaca-tw_en-align.json](./alpaca-tw-en-align.json): [ntunpllab](https://github.com/ntunlplab/traditional-chinese-alpaca) translate Stanford Alpaca 52k dataset
+- [Chinese Llama 2 7B train.py](https://github.com/LinkSoul-AI/Chinese-Llama-2-7b/blob/main/train.py)
+- [Load the pretrained model in 4-bit precision and Set training with LoRA according to hf's trl lib](https://github.com/lvwerra/trl/blob/main/examples/scripts/sft_trainer.py): QLoRA finetuning
+
+## Resources
+- traditional chinese qlora finetuned Llama2 merge model: [weiren119/traditional_chinese_qlora_llama2_merged](https://huggingface.co/weiren119/traditional_chinese_qlora_llama2_merged)
+- traditional chinese qlora adapter model: [weiren119/traditional_chinese_qlora_llama2](https://huggingface.co/weiren119/traditional_chinese_qlora_llama2)
+
+## Online Demo
+- [Run the qlora finetuned model at colab](https://colab.research.google.com/drive/1OYXvhY-8KjEDaGhOLrJe4omjtFgOWjy1?usp=sharing): May need colab pro or colab pro+
 
 ## Use which pretrained model
 - NousResearch: https://huggingface.co/NousResearch/Llama-2-7b-chat-hf