happybrian/fast-brain-translate-adapter

Fast-Brain translate Cortex Adapter (LoRA)

translate 专家皮层 LoRA 适配器（~20MB），需配合基底 happybrian/fast-brain-base 使用。

from huggingface_hub import snapshot_download
from mlx_lm import load, generate

base = snapshot_download("happybrian/fast-brain-base")        # 皮层基底
adapter = snapshot_download("happybrian/fast-brain-translate-adapter")  # 本皮层
model, tokenizer = load(base, adapter_path=adapter)
prompt = tokenizer.apply_chat_template(
    [{"role": "user", "content": "..."}], tokenize=False, add_generation_prompt=True)
print(generate(model, tokenizer, prompt=prompt, max_tokens=256))

• 训练：LoRA r=16, 800 步, 蒸馏数据 ~450-650 条（教师 Qwen3-8B-4bit）
• 评测：中英双向人工抽检通过
• 硬件：Apple M5 24GB, mlx-lm