Instructions to use Nexuss0781/SPACE with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use Nexuss0781/SPACE 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 Nexuss0781/SPACE:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Nexuss0781/SPACE:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Nexuss0781/SPACE:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Nexuss0781/SPACE: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 Nexuss0781/SPACE:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf Nexuss0781/SPACE: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 Nexuss0781/SPACE:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Nexuss0781/SPACE:Q4_K_M

Use Docker

docker model run hf.co/Nexuss0781/SPACE:Q4_K_M

LM Studio
Jan

vLLM

How to use Nexuss0781/SPACE with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Nexuss0781/SPACE"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Nexuss0781/SPACE",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/Nexuss0781/SPACE:Q4_K_M

Ollama
How to use Nexuss0781/SPACE with Ollama:
```
ollama run hf.co/Nexuss0781/SPACE:Q4_K_M
```
Unsloth Desktop

How to use Nexuss0781/SPACE with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Nexuss0781/SPACE: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": "Nexuss0781/SPACE:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use Nexuss0781/SPACE with Docker Model Runner:
```
docker model run hf.co/Nexuss0781/SPACE:Q4_K_M
```

Lemonade

How to use Nexuss0781/SPACE with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Nexuss0781/SPACE:Q4_K_M

Run and chat with the model

lemonade run user.SPACE-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use Nexuss0781/SPACE with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Nexuss0781/SPACE: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 Nexuss0781/SPACE:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use Nexuss0781/SPACE with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Nexuss0781/SPACE: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 "Nexuss0781/SPACE: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"

Nexuss0781 commited on Aug 29

Commit

dbd60fc

verified ·

1 Parent(s): 6e26c70

Add model card

Browse files

Files changed (1) hide show

README.md +14 -28

README.md CHANGED Viewed

@@ -9,39 +9,25 @@ quantized_by: Nexuss0781
 # SPACE
-Micro-tier AI stack: small GGUF models, quantized, run on CPU via [llama.cpp](https://github.com/ggml-org/llama.cpp), benchmarked on GitHub Actions.
-Public runtime models live on HuggingFace: https://huggingface.co/Nexuss0781/SPACE
-## Benchmarks
-All runs executed on the same GitHub Actions runner (Ubuntu, AMD EPYC 7763, 4 vCPU, 16.77 GB RAM). Text measured with `llama.cpp` GGUF q4_k_m; vision measured with `llama-mtmd-cli` (mmproj, one model per job).
-### Text leg (10 prompts, 256 max new tokens)
-| Model | Weights | Avg tok/s | Peak RSS | Wall (10 prompts) |
-|---|---:|---:|---:|---:|
-| **SmolLM2-360M-Instruct** (q4_k_m) | 0.27 GB | **57.73** | 474 MB | 37.7s |
-| SmolLM3-3B (q4_k_m) | 1.92 GB | 13.18 | 3,436 MB | 196.6s |
-| Qwen2.5-3B-Instruct (q4_k_m) | 1.93 GB | 13.43 | 3,336 MB | 161.6s |
-Earlier bf16 baseline for the micro tier: 2.26 tok/s, 6,397 MB peak RSS (~25x slower, ~13x heavier than the q4 micro model).
-### Vision leg (10 image tasks, 128 max new tokens)
-| Model | Weights | Avg tok/s | Peak RSS | Wall (10 imgs) |
-|---|---:|---:|---:|---:|
-| **SmolVLM2-500M-Video-Instruct** (q8_0) | 0.44 GB | **67.15** | 1,026 MB | 73.7s |
-| SmolVLM-500M-Instruct (q8_0) | 0.44 GB | 65.06 | 1,026 MB | 77.6s |
-| SmolVLM2-2.2B-Instruct (q4_k_m) | 1.11 GB | 19.15 | 4,260 MB | 363.3s |
-| Moondream2 (q8_0) | 1.51 GB | 14.86 | 4,000 MB | 236.1s |
-| Gemma 3 4B IT (q4_k_m) | 2.49 GB | 13.19 | 5,309 MB | 602.6s |
-Result: the micro-tier models are ~4.5x faster than the large ones on both legs. Working pair: **SmolLM2-360M / SmolVLM2-500M** as the always-on edge runtimes, with a 3B-class model (Qwen2.5-3B / SmolVLM2-2.2B) in the swappable quality slot.
-## Layout
-- `benchmarks/` — harnesses and task sets (text + vision), run by GitHub Actions
-- `reports/` — consolidated benchmark papers
-- `models/upload/` — GGUF upload tooling for `Nexuss0781/SPACE`
-- `prompts/` — prompt sets

 # SPACE
+Single repository holding the GGUF weights (llama.cpp) benchmarked in the
+Space GitHub Actions pipeline. Vision and audio models include their `mmproj`
+encoder and run with `llama-mtmd-cli`.
+## Models
+Text (`text-generation`):
+- **HuggingFaceTB/SmolLM2-360M-Instruct** - `SmolLM2-360M-Instruct.Q4_K_M.gguf` - 0.27 GB - license: apache-2.0 - text
+- **Qwen/Qwen2.5-3B-Instruct** - `Qwen2.5-3B-Instruct.Q4_K_M.gguf` - 1.93 GB - license: qwen-research - text
+Vision (`image-text-to-text`, use alongside the `mmproj`):
+- **HuggingFaceTB/SmolVLM2-2.2B-Instruct** - `SmolVLM2-2.2B-Instruct-Q4_K_M.gguf` + `mmproj-SmolVLM2-2.2B-Instruct-f16.gguf` - 1.11 GB - license: apache-2.0 - vision
+- **HuggingFaceTB/SmolVLM2-500M-Video-Instruct** - `SmolVLM2-500M-Video-Instruct-Q8_0.gguf` + `mmproj-SmolVLM2-500M-Video-Instruct-f16.gguf` - 0.44 GB - license: apache-2.0 - vision
+Audio (`automatic-speech-recognition`, use alongside the `mmproj`):
+- **Qwen/Qwen3-ASR-0.6B** - `Qwen3-ASR-0.6B-Q8_0.gguf` + `mmproj-Qwen3-ASR-0.6B-Q8_0.gguf` - 0.8 GB - license: apache-2.0 - audio
+- **Qwen/Qwen3-ASR-1.7B** - `Qwen3-ASR-1.7B-Q8_0.gguf` + `mmproj-Qwen3-ASR-1.7B-Q8_0.gguf` - 2.17 GB - license: apache-2.0 - audio
+Benchmarks: https://github.com/nexuss0781/Space