Instructions to use h34v7/Jackrong-Qwopus3.5-27B-v3-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 h34v7/Jackrong-Qwopus3.5-27B-v3-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 h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf h34v7/Jackrong-Qwopus3.5-27B-v3-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 h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf h34v7/Jackrong-Qwopus3.5-27B-v3-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 h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf h34v7/Jackrong-Qwopus3.5-27B-v3-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 h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF:Q4_K_M

Use Docker

docker model run hf.co/h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF:Q4_K_M

LM Studio
Jan
Ollama
How to use h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF with Ollama:
```
ollama run hf.co/h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF:Q4_K_M
```
Unsloth Desktop

How to use h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf h34v7/Jackrong-Qwopus3.5-27B-v3-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": "h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF with Docker Model Runner:
```
docker model run hf.co/h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF:Q4_K_M
```

Lemonade

How to use h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Jackrong-Qwopus3.5-27B-v3-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use h34v7/Jackrong-Qwopus3.5-27B-v3-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 h34v7/Jackrong-Qwopus3.5-27B-v3-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 h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use h34v7/Jackrong-Qwopus3.5-27B-v3-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf h34v7/Jackrong-Qwopus3.5-27B-v3-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 "h34v7/Jackrong-Qwopus3.5-27B-v3-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"

Q5 perplexity

by grumd0 - opened Apr 3

Discussion

grumd0

Apr 3

Isn't Q5 perplexity only "smaller" than BF16 simply because the distribution is +- 0.04?

BF16 -- 6.2671 +/- 0.04039 -- from 6.22 to 6.30
Q5_K_M -- 6.2564 +/- 0.04021 -- from 6.21 to 6.29

The real value could be anywhere in that range, so you can't really say that Q5 is better 🤔

h34v7

Owner Apr 3

It could be that my dual gpu and cpu inference was producing noises on the BF16 test and not on the Q6_K and Q5_K_M test.

It could be a phenomenon called regularization the higher BF16 models think too much and happen to overfit.

I quote from my local agent
"Sometimes quantization acts as a form of dropout/regularization. If the BF16 model has slight overfitting on the validation set, a quantized version might generalize slightly better on that specific sample."

h34v7

Owner Apr 3

So yeah the full model will still have more precision whereas the quantized one happen to aced that wiki raw text better.

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