Instructions to use allbilly/nanochat-d12-rl with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use allbilly/nanochat-d12-rl with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="allbilly/nanochat-d12-rl",
	filename="nanochat-d12-rl-bf16.gguf",
)

llm.create_chat_completion(
	messages = "No input example has been defined for this model task."
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use allbilly/nanochat-d12-rl with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf allbilly/nanochat-d12-rl:BF16
# Run inference directly in the terminal:
llama cli -hf allbilly/nanochat-d12-rl:BF16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf allbilly/nanochat-d12-rl:BF16
# Run inference directly in the terminal:
llama cli -hf allbilly/nanochat-d12-rl:BF16

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf allbilly/nanochat-d12-rl:BF16
# Run inference directly in the terminal:
./llama-cli -hf allbilly/nanochat-d12-rl:BF16

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf allbilly/nanochat-d12-rl:BF16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf allbilly/nanochat-d12-rl:BF16

Use Docker

docker model run hf.co/allbilly/nanochat-d12-rl:BF16

LM Studio
Jan
Ollama
How to use allbilly/nanochat-d12-rl with Ollama:
```
ollama run hf.co/allbilly/nanochat-d12-rl:BF16
```

Unsloth Studio

How to use allbilly/nanochat-d12-rl with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for allbilly/nanochat-d12-rl to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for allbilly/nanochat-d12-rl to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for allbilly/nanochat-d12-rl to start chatting

Atomic Chat new
Docker Model Runner
How to use allbilly/nanochat-d12-rl with Docker Model Runner:
```
docker model run hf.co/allbilly/nanochat-d12-rl:BF16
```

Lemonade

How to use allbilly/nanochat-d12-rl with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull allbilly/nanochat-d12-rl:BF16

Run and chat with the model

lemonade run user.nanochat-d12-rl-BF16

List all available models

lemonade list

nanochat d12 (rl)

Native nanochat checkpoint exported from chatrl_checkpoints/d12 step 933.

Formats in this repo

File	Use
`model.safetensors`	HF-layout weights (`model.*` keys)
`model_000933.pt`	Resume training in nanochat
`*.gguf`	llama.cpp inference (lukas-h/llama.cpp fork required)
`tokenizer.pkl`	nanochat / GGUF tokenizer

Load in nanochat

export NANOCHAT_BASE_DIR=~/.cache/nanochat
# copy model_*.pt + meta_*.json into chatsft_checkpoints/d12/ (or chatrl_checkpoints/d12/)
python -m scripts.chat_cli -i rl -g d12

GGUF inference

Stock llama.cpp does not support nanochat. Use lukas-h/llama.cpp:

./build/bin/llama-completion -m nanochat-d12-rl-bf16.gguf \
  -p '<|user_start|>Hello<|user_end|><|assistant_start|>' --special -no-cnv -n 128

Special tokens: <|bos|>, <|user_start|>, <|user_end|>, <|assistant_start|>, <|assistant_end|>, <|python_start|>, <|python_end|>, <|output_start|>, <|output_end|>

Downloads last month: 72

Safetensors

Model size

0.3B params

Tensor type

F32

BF16

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support