This model runs smoothly on a Mac with 64GB of unified memory, exuding a sense of steadfast wisdom and making it an excellent choice for experiencing large-parameter models. The Model YuHaaa/Qwen1.5-110B-Chat-mlx-3Bit was converted to MLX format from Qwen/Qwen1.5-110B-Chat using mlx-lm version 0.22.1.

Use with mlx

pip install mlx-lm

python

from mlx_lm import load, generate

model, tokenizer = load("YuHaaa/Qwen1.5-110B-Chat-mlx-3Bit")

prompt="hello"

if hasattr(tokenizer, "apply_chat_template") and tokenizer.chat_template is not None:
    messages = [{"role": "user", "content": prompt}]
    prompt = tokenizer.apply_chat_template(
        messages, tokenize=False, add_generation_prompt=True
    )

response = generate(model, tokenizer, prompt=prompt, verbose=True)