Instructions to use bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF 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 bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M

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 bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M

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 bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M

Use Docker

docker model run hf.co/bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M

Ollama
How to use bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF with Ollama:
```
ollama run hf.co/bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M
```
Unsloth Desktop

How to use bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF with Docker Model Runner:
```
docker model run hf.co/bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M
```

Lemonade

How to use bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "bartowski/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Possible to strip MTP layer?

by jcmyang750 - opened Aug 14

Discussion

jcmyang750

Aug 14

Hi,
Is it possible to strip the MTP layer to make the file size smaller? I found that using MTP actually slows the decode by about 10% on Apple M1 Max using llama cpp. Tried both Q6_K and Q8_0. Thanks.

systems-u

28 days ago

I completely agree with you, the widespread use of MTP is a bad practice because the sending speed decreases from 1600 t/s without using MTP, with MTP the speed drops to 800 t/s, winning maybe 10 t/s (not always) on receiving and losing colossal speed on sending is just disgusting, besides, the MTP layer takes up space, for me these methods are quite enough and work well --spec-type ngram-map-k4v,ngram-mod --spec-ngram-map-k4v-size-n 4 --spec-ngram-map-k4v-size-m 3 --spec-ngram-map-k4v-min-hits 1 --spec-ngram-mod-n-match 16 --spec-ngram-mod-n-min 32 --spec-ngram-mod-n-max 128 --repeat-penalty 1.0

and draft-mtp in some scenarios it even slows down the

zxbc2023

27 days ago

•

edited 27 days ago

I completely agree with you, the widespread use of MTP is a bad practice because the sending speed decreases from 1600 t/s without using MTP, with MTP the speed drops to 800 t/s, winning maybe 10 t/s (not always) on receiving and losing colossal speed on sending is just disgusting, besides, the MTP layer takes up space,

If you load an MTP model specifically not using spec drafting or using a different spec draft, the MTP layers are not loaded. You are not incurring any VRAM waste.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment