Instructions to use TobDeBer/PowerMoe-3b-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 TobDeBer/PowerMoe-3b-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 TobDeBer/PowerMoe-3b-GGUF
# Run inference directly in the terminal:
llama cli -hf TobDeBer/PowerMoe-3b-GGUF

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf TobDeBer/PowerMoe-3b-GGUF
# Run inference directly in the terminal:
llama cli -hf TobDeBer/PowerMoe-3b-GGUF

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 TobDeBer/PowerMoe-3b-GGUF
# Run inference directly in the terminal:
./llama-cli -hf TobDeBer/PowerMoe-3b-GGUF

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 TobDeBer/PowerMoe-3b-GGUF
# Run inference directly in the terminal:
./build/bin/llama-cli -hf TobDeBer/PowerMoe-3b-GGUF

Use Docker

docker model run hf.co/TobDeBer/PowerMoe-3b-GGUF

LM Studio
Jan

vLLM

How to use TobDeBer/PowerMoe-3b-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "TobDeBer/PowerMoe-3b-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "TobDeBer/PowerMoe-3b-GGUF",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/TobDeBer/PowerMoe-3b-GGUF

Ollama
How to use TobDeBer/PowerMoe-3b-GGUF with Ollama:
```
ollama run hf.co/TobDeBer/PowerMoe-3b-GGUF
```

Unsloth Studio

How to use TobDeBer/PowerMoe-3b-GGUF 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 TobDeBer/PowerMoe-3b-GGUF 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 TobDeBer/PowerMoe-3b-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for TobDeBer/PowerMoe-3b-GGUF to start chatting

Atomic Chat new
Docker Model Runner
How to use TobDeBer/PowerMoe-3b-GGUF with Docker Model Runner:
```
docker model run hf.co/TobDeBer/PowerMoe-3b-GGUF
```

Lemonade

How to use TobDeBer/PowerMoe-3b-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull TobDeBer/PowerMoe-3b-GGUF

Run and chat with the model

lemonade run user.PowerMoe-3b-GGUF-{{QUANT_TAG}}

List all available models

lemonade list

Model Summary

PowerMoE-3B is a 3B sparse Mixture-of-Experts (sMoE) language model trained with the Power learning rate scheduler. It sparsely activates 800M parameters for each token. It is trained on a mix of open-source and proprietary datasets. PowerMoE-3B has shown promising results compared to other dense models with 2x activate parameters across various benchmarks, including natural language multi-choices, code generation, and math reasoning. Paper: https://arxiv.org/abs/2408.13359

This is a GGUF quantized version.

Usage

Requires latest llama.cpp to run.

Generation

This is a simple example of how to use the PowerMoe GGUF:

./llama-cli -m PowerMoE4x800M_q3km.gguf -p "How about a snack?"

Downloads last month: 25

GGUF

Model size

4B params

Architecture

granite

Hardware compatibility

We're not able to determine the quantization variants.

View all variants

Model tree for TobDeBer/PowerMoe-3b-GGUF

Base model

ibm-research/PowerMoE-3b

Quantized

(8)

this model

Paper for TobDeBer/PowerMoe-3b-GGUF

Power Scheduler: A Batch Size and Token Number Agnostic Learning Rate Scheduler

Paper • 2408.13359 • Published Aug 23, 2024 • 23

Evaluation results

accuracy-norm on ARC
self-reported

58.100
accuracy on BoolQ
self-reported

65.000
accuracy-norm on Hellaswag
self-reported

71.500
accuracy-norm on OpenBookQA
self-reported

41.000
accuracy-norm on PIQA
self-reported

79.100
accuracy-norm on Winogrande
self-reported

65.000
accuracy on MMLU (5 shot)
self-reported

42.800
accuracy on GSM8k (5 shot)
self-reported

25.900