README.md

metadata

pipeline_tag: text-generation
inference: false
license: apache-2.0
library_name: transformers
tags:
  - TensorBlock
  - GGUF
base_model: ibm/PowerMoE-3b
model-index:
  - name: ibm/PowerMoE-3b
    results:
      - task:
          type: text-generation
        dataset:
          name: ARC
          type: lm-eval-harness
        metrics:
          - type: accuracy-norm
            value: 58.1
            name: accuracy-norm
            verified: false
          - type: accuracy
            value: 65
            name: accuracy
            verified: false
          - type: accuracy-norm
            value: 71.5
            name: accuracy-norm
            verified: false
          - type: accuracy-norm
            value: 41
            name: accuracy-norm
            verified: false
          - type: accuracy-norm
            value: 79.1
            name: accuracy-norm
            verified: false
          - type: accuracy-norm
            value: 65
            name: accuracy-norm
            verified: false
          - type: accuracy
            value: 42.8
            name: accuracy
            verified: false
          - type: accuracy
            value: 25.9
            name: accuracy
            verified: false
          - type: accuracy
            value: 14.8
            name: accuracy
            verified: false
      - task:
          type: text-generation
        dataset:
          name: humaneval
          type: bigcode-eval
        metrics:
          - type: pass@1
            value: 20.1
            name: pass@1
            verified: false
          - type: pass@1
            value: 32.4
            name: pass@1
            verified: false

ibm/PowerMoE-3b - GGUF

This repo contains GGUF format model files for ibm/PowerMoE-3b.

The files were quantized using machines provided by TensorBlock, and they are compatible with llama.cpp as of commit b4011.

Our projects

Forge
An OpenAI-compatible multi-provider routing layer.
🚀 Try it now! 🚀
Awesome MCP Servers	TensorBlock Studio
A comprehensive collection of Model Context Protocol (MCP) servers.	A lightweight, open, and extensible multi-LLM interaction studio.
👀 See what we built 👀	👀 See what we built 👀

## Prompt template

Model file specification

Filename	Quant type	File Size	Description
PowerMoE-3b-Q2_K.gguf	Q2_K	1.179 GB	smallest, significant quality loss - not recommended for most purposes
PowerMoE-3b-Q3_K_S.gguf	Q3_K_S	1.386 GB	very small, high quality loss
PowerMoE-3b-Q3_K_M.gguf	Q3_K_M	1.531 GB	very small, high quality loss
PowerMoE-3b-Q3_K_L.gguf	Q3_K_L	1.652 GB	small, substantial quality loss
PowerMoE-3b-Q4_0.gguf	Q4_0	1.794 GB	legacy; small, very high quality loss - prefer using Q3_K_M
PowerMoE-3b-Q4_K_S.gguf	Q4_K_S	1.809 GB	small, greater quality loss
PowerMoE-3b-Q4_K_M.gguf	Q4_K_M	1.918 GB	medium, balanced quality - recommended
PowerMoE-3b-Q5_0.gguf	Q5_0	2.178 GB	legacy; medium, balanced quality - prefer using Q4_K_M
PowerMoE-3b-Q5_K_S.gguf	Q5_K_S	2.178 GB	large, low quality loss - recommended
PowerMoE-3b-Q5_K_M.gguf	Q5_K_M	2.242 GB	large, very low quality loss - recommended
PowerMoE-3b-Q6_K.gguf	Q6_K	2.586 GB	very large, extremely low quality loss
PowerMoE-3b-Q8_0.gguf	Q8_0	3.346 GB	very large, extremely low quality loss - not recommended

Downloading instruction

Command line

Firstly, install Huggingface Client

pip install -U "huggingface_hub[cli]"

Then, downoad the individual model file the a local directory

huggingface-cli download tensorblock/PowerMoE-3b-GGUF --include "PowerMoE-3b-Q2_K.gguf" --local-dir MY_LOCAL_DIR

If you wanna download multiple model files with a pattern (e.g., *Q4_K*gguf), you can try:

huggingface-cli download tensorblock/PowerMoE-3b-GGUF --local-dir MY_LOCAL_DIR --local-dir-use-symlinks False --include='*Q4_K*gguf'