README.md exists but content is empty.
- Downloads last month
- 42
Model size
0.5B params
Tensor type
BF16
·
U32 ·
Hardware compatibility
Log In to add your hardware
4-bit
Model tree for sykuang/Llama-3.2-Taiwan-3B-Instruct-MLX-4bit-vchewing
Base model
meta-llama/Llama-3.2-3B Finetuned
lianghsun/Llama-3.2-Taiwan-3B Finetuned
lianghsun/Llama-3.2-Taiwan-3B-InstructDatasets used to train sykuang/Llama-3.2-Taiwan-3B-Instruct-MLX-4bit-vchewing
Evaluation results
- single choice on tw-legal-benchmark-v1self-reported31.100
- single choice on (Society) Formosa Taiwan Knowledge Benchtest set self-reported60.420
- single choice on (Governmnt) Formosa Taiwan Knowledge Benchtest set self-reported44.250
- single choice on (Geography) Formosa Taiwan Knowledge Benchtest set self-reported47.540
- single choice on (History) Formosa Taiwan Knowledge Benchtest set self-reported60.000
- single choice on (geography_of_taiwan) tmmlu++test set self-reported36.200
- single choice on (dentistry) tmmlu++test set self-reported33.830
- single choice on (technical) tmmlu++test set self-reported35.070
# Make sure mlx-lm is installed # pip install --upgrade mlx-lm # Generate text with mlx-lm from mlx_lm import load, generate model, tokenizer = load("sykuang/Llama-3.2-Taiwan-3B-Instruct-MLX-4bit-vchewing") prompt = "Write a story about Einstein" messages = [{"role": "user", "content": prompt}] prompt = tokenizer.apply_chat_template( messages, add_generation_prompt=True ) text = generate(model, tokenizer, prompt=prompt, verbose=True)