Instructions to use athirdpath/Harmonia-20B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use athirdpath/Harmonia-20B with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="athirdpath/Harmonia-20B")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("athirdpath/Harmonia-20B")
model = AutoModelForCausalLM.from_pretrained("athirdpath/Harmonia-20B")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use athirdpath/Harmonia-20B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "athirdpath/Harmonia-20B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "athirdpath/Harmonia-20B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/athirdpath/Harmonia-20B

SGLang

How to use athirdpath/Harmonia-20B with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "athirdpath/Harmonia-20B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "athirdpath/Harmonia-20B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "athirdpath/Harmonia-20B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "athirdpath/Harmonia-20B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use athirdpath/Harmonia-20B with Docker Model Runner:
```
docker model run hf.co/athirdpath/Harmonia-20B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Description

This is a unified 20B model crafted via a multi-step SLERP merge of eight 20B models. The aim was to develop a versatile "base model" for TaskArithmetic in this size class.

Merging Process:

Models:

- model: Undi95/Emerhyst-20B
- model: Undi95/MXLewd-L2-20B
- model: Undi95/Lewd-Sydney-20B
- model: athirdpath/Nethena-20b-Glued
- model: tavtav/Rose-20B
- model: Undi95/PsyMedRP-v1-20B
- model: NeverSleep/Noromaid-20b-v0.1.1
- model: Undi95/U-Amethyst-20B

Concept:

The idea behind this process was to blend the unique attributes of each model while minimizing individual quirks. This approach has also shown promising results as a standalone RP model, providing a combination of high-quality writing and situational problem-solving/awareness.

Prompt template: Alpaca

Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
{prompt}

### Response:

Thanks to Undi95 for pioneering the 20B recipe, and for most of the models involved.

Downloads last month: 240

Safetensors

Model size

20B params

Tensor type

F16

Model tree for athirdpath/Harmonia-20B

Finetunes

2 models

Merges

8 models

Quantizations

10 models