Instructions to use KyleHessling1/Qwopus-GLM-18B-Merged-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use KyleHessling1/Qwopus-GLM-18B-Merged-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="KyleHessling1/Qwopus-GLM-18B-Merged-GGUF",
	filename="Qwopus-GLM-18B-Healed-Q3_K_M.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use KyleHessling1/Qwopus-GLM-18B-Merged-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 KyleHessling1/Qwopus-GLM-18B-Merged-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf KyleHessling1/Qwopus-GLM-18B-Merged-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 KyleHessling1/Qwopus-GLM-18B-Merged-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf KyleHessling1/Qwopus-GLM-18B-Merged-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 KyleHessling1/Qwopus-GLM-18B-Merged-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf KyleHessling1/Qwopus-GLM-18B-Merged-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 KyleHessling1/Qwopus-GLM-18B-Merged-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf KyleHessling1/Qwopus-GLM-18B-Merged-GGUF:Q4_K_M

Use Docker

docker model run hf.co/KyleHessling1/Qwopus-GLM-18B-Merged-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use KyleHessling1/Qwopus-GLM-18B-Merged-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "KyleHessling1/Qwopus-GLM-18B-Merged-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": "KyleHessling1/Qwopus-GLM-18B-Merged-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/KyleHessling1/Qwopus-GLM-18B-Merged-GGUF:Q4_K_M

Ollama
How to use KyleHessling1/Qwopus-GLM-18B-Merged-GGUF with Ollama:
```
ollama run hf.co/KyleHessling1/Qwopus-GLM-18B-Merged-GGUF:Q4_K_M
```

Unsloth Studio

How to use KyleHessling1/Qwopus-GLM-18B-Merged-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 KyleHessling1/Qwopus-GLM-18B-Merged-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 KyleHessling1/Qwopus-GLM-18B-Merged-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for KyleHessling1/Qwopus-GLM-18B-Merged-GGUF to start chatting

How to use KyleHessling1/Qwopus-GLM-18B-Merged-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf KyleHessling1/Qwopus-GLM-18B-Merged-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "KyleHessling1/Qwopus-GLM-18B-Merged-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use KyleHessling1/Qwopus-GLM-18B-Merged-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 KyleHessling1/Qwopus-GLM-18B-Merged-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 KyleHessling1/Qwopus-GLM-18B-Merged-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat new
Docker Model Runner
How to use KyleHessling1/Qwopus-GLM-18B-Merged-GGUF with Docker Model Runner:
```
docker model run hf.co/KyleHessling1/Qwopus-GLM-18B-Merged-GGUF:Q4_K_M
```

Lemonade

How to use KyleHessling1/Qwopus-GLM-18B-Merged-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull KyleHessling1/Qwopus-GLM-18B-Merged-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Qwopus-GLM-18B-Merged-GGUF-Q4_K_M

List all available models

lemonade list

Best tool-use model for 16GB VRAM GPU

by a123451 - opened May 7

Discussion

a123451

May 7

I tried a bunch of latest models with various built from tbq plus, buun,ik_llammaetc...none of them fixed the issued I have with Gbrain. Models that finetuned for Hermes like Carnice v2 27B can but painfully slow. Probably just me being newb.
Anyway, this one does the job. 42t/s for long context length, drop to 40 at the end , GBrain came back 90/100 with 3132 bookmarks full enrichment, blacklink fix, embedding coverage 100%, graph stats( wikilinks created, timeline entries, graph_coverage %) all done.
I'm using Tom turboquant plus built with these parameters:
--model "D:\llama-models\Qwopus-GLM-18B-Healed-Q6_K.gguf" ^
-fa 1 ^
-ctk q4_0 -ctv q4_0 ^
-c 131064 ^
-ngl 99 ^
--batch-size 2048 ^
--ubatch-size 512 ^
-np 1 ^
--temp 0.7 ^
--top-k 40 ^
--top-p 0.9 ^
--repeat-penalty 1.1 ^
--min-p 0.05 ^
--jinja ^
Any plan to make another one for qwopus36-35b-a3b please?

KyleHessling1

Owner May 8

Hey there!

Thank you for the feedback. I'm pumped to hear the model is doing so well! I am currently working on another merge; we'll see how it goes! A similar duplication merge with Qwopus 3.6 35B a 3b is possible but we need to make another fine-tune of it, which is in the works with my friend Jackrong, so in the future we could even do a merge on that!

Thank you for the support and flags, this is incredibly useful and motivating!
-Kyle

a123451

May 11

•

edited May 11

Q4_K-M testing on Ubutun 24.04 LTS ,i got extra 20 tps , haven't seen any problem with long context yet. However, there is issue that the model offload all layers on GPU but only CPU compute, i don't know if this llama cpp build , mine is built from turboquant plus or it's the model. To reproduce, set -ctk q8_0 -ctv q4_0 -> check for CPU loading, mine show >900% while gpu 0~1%. No issue with using turboquant, -ctk q8_0 -ctv turbo4.

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