---
library_name: mlx
license: apache-2.0
license_link: https://huggingface.co/Qwen/Qwen3-0.6B/blob/main/LICENSE
pipeline_tag: text-generation
base_model: Qwen/Qwen3-0.6B-MLX-bf16
tags:
- mlx
datasets:
- DuckyBlender/diego-replies-dataset
---

# DuckyBlender/diegogpt-v2-mlx-bf16

This model [DuckyBlender/diegogpt-v2-mlx-bf16](https://huggingface.co/DuckyBlender/diegogpt-v2-mlx-bf16) is a full fine-tune of [Qwen/Qwen3-0.6B-MLX-bf16](https://huggingface.co/Qwen/Qwen3-0.6B-MLX-bf16), trained on the complete set of public replies from a specific individual.

Training was conducted using `mlx-lm` version **0.26.0**. It ran for 15 steps with a batch size of 16, completing in a few seconds on a MacBook Pro M1 Pro (8-core CPU, 16GB RAM). Peak memory usage was 8.3GB. The dataset contained 225 low-quality training pairs (240 lines trained total).

Run with system prompt `/no_think` and the following generation parameters:
- `--temp 0.7`
- `--top-p 0.8`
- `--top-k 20`
- `--min-p 0`

Example usage:

```bash
pip install mlx-lm
````

```python
from mlx_lm import load, generate

model, tokenizer = load("DuckyBlender/diegogpt-v2-mlx-bf16")

prompt = "are you red hat hacker?"

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

sampler = make_sampler(temp=0.7, top_p=0.8, top_k=20, min_p=0)

response = mlx_lm.generate(
    model,
    tokenizer,
    prompt=prompt,
    sampler=sampler,
    verbose=True
)
```

Or directly via CLI:

```bash
mlx_lm.generate \
  --model "DuckyBlender/diegogpt-v2-mlx-bf16" \
  --temp 0.7 \
  --top-p 0.8 \
  --top-k 20 \
  --min-p 0 \
  --system "/no_think" \
  --prompt "are you red hat hacker?"
```

Model uses \~1.25GB RAM during inference.