Buddy Nesso SFT v1 โ€” MLX

MLX conversion (bfloat16) of giux78/buddy-nesso-sft-v1, a 0.4B parameter bilingual (Italian/English) story and play buddy for children under 8, fine-tuned from mii-llm/nesso-0.4B-agentic.

Converted with mlx-lm for fast local inference on Apple Silicon. See the base model card for training details, the recommended child-safety system prompt, and important safety limitations โ€” this is a research checkpoint, not a production child-safety model.

Usage

pip install mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("giux78/buddy-nesso-sft-v1-mlx")

messages = [
    {"role": "system", "content": "You are Nesso, a warm story and play buddy for children under 8. Keep replies short and safe."},
    {"role": "user", "content": "Ciao! Mi racconti una storia di un piccolo orso?"},
]
prompt = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)
print(generate(model, tokenizer, prompt=prompt, max_tokens=150))

Or from the command line:

mlx_lm.chat --model giux78/buddy-nesso-sft-v1-mlx

Voice demo

A complete voice-interactive demo (speech in โ†’ Nesso โ†’ speech out, fully local on a Mac) is available at mii-llm/buddy-nesso-voice-demo.

Downloads last month
6
Safetensors
Model size
0.4B params
Tensor type
BF16
ยท
MLX
Hardware compatibility
Log In to add your hardware

Quantized

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for giux78/buddy-nesso-sft-v1-mlx

Finetuned
(1)
this model