metadata
license: gpl-3.0
datasets:
- Orion-zhen/meissa-unalignments
language:
- zh
- en
base_model: Orion-zhen/Qwen2.5-14B-Instruct-Uncensored
tags:
- mlx
Fmuaddib/Qwen2.5-14B-Instruct-Uncensored-mlx-8Bit
The Model Fmuaddib/Qwen2.5-14B-Instruct-Uncensored-mlx-8Bit was converted to MLX format from Orion-zhen/Qwen2.5-14B-Instruct-Uncensored using mlx-lm version 0.22.1.
Use with mlx
pip install mlx-lm
from mlx_lm import load, generate
model, tokenizer = load("Fmuaddib/Qwen2.5-14B-Instruct-Uncensored-mlx-8Bit")
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)