cs2764/DeepSeek-V3-0324-BF16-mlx-3Bit-gs32

The Model cs2764/DeepSeek-V3-0324-BF16-mlx-3Bit-gs32 was converted to MLX format from unsloth/DeepSeek-V3-0324-BF16 using mlx-lm version 0.26.2.

Quantization Details

This model was converted with the following quantization settings:

• Quantization Strategy: 3-bit quantization
• Group Size: 32
• Average bits per weight: 4.002

Use with mlx

pip install mlx-lm

from mlx_lm import load, generate

model, tokenizer = load("cs2764/DeepSeek-V3-0324-BF16-mlx-3Bit-gs32")

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)