Instructions to use QMB15/Stheno-L2-13B-8bit-exl2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use QMB15/Stheno-L2-13B-8bit-exl2 with Transformers:

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

pipe = pipeline("text-generation", model="QMB15/Stheno-L2-13B-8bit-exl2")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("QMB15/Stheno-L2-13B-8bit-exl2")
model = AutoModelForCausalLM.from_pretrained("QMB15/Stheno-L2-13B-8bit-exl2", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use QMB15/Stheno-L2-13B-8bit-exl2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "QMB15/Stheno-L2-13B-8bit-exl2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "QMB15/Stheno-L2-13B-8bit-exl2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/QMB15/Stheno-L2-13B-8bit-exl2

SGLang

How to use QMB15/Stheno-L2-13B-8bit-exl2 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 "QMB15/Stheno-L2-13B-8bit-exl2" \
    --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": "QMB15/Stheno-L2-13B-8bit-exl2",
		"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 "QMB15/Stheno-L2-13B-8bit-exl2" \
        --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": "QMB15/Stheno-L2-13B-8bit-exl2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use QMB15/Stheno-L2-13B-8bit-exl2 with Docker Model Runner:
```
docker model run hf.co/QMB15/Stheno-L2-13B-8bit-exl2
```

This is a exllama V2 quantization of https://huggingface.co/TheBloke/Stheno-L2-13B-GPTQ Uses a target bpw of 8, intended for best quality on cards like a 3090 or similar. Includes measurement.json for convenience of quantizing to other sizes. Calibration data: https://huggingface.co/datasets/wikitext/resolve/refs%2Fconvert%2Fparquet/wikitext-2-v1/test/0000.parquet

An experimental merging of Several Models using two various methods, Ties-Merge and BlockMerge_Gradient

I plan for this to be the base of my Model with my own [Stheno: ERP-Based LORA] merged in, some time in the future.

Stheno:
Gradient Merge of Stheno-P1 & Stheno-P2.

SISTER MODEL HERE: Stheno-Inverted-L2-13B

Quants courtesy of TheBloke!
GPTQ
GGUF
GGML

Test Checklist:
Censorship - Fairly Uncensored
Writing - Good Prose, Fairly Descriptive
NSFW - Yes
IQ Level - Pretty Smart
Formatting - Proper Formatting with Examples

Stheno-P1 [Ties-Merge]
-----elinas/chronos-13b-v2
-----jondurbin/airoboros-l2-13b-2.1
-----NousResearch/Nous-Hermes-Llama2-13b+nRuaif/Kimiko-v2 LORA

Stheno-P2 [Ties-Merge]
-----CalderaAI/13B-Legerdemain-L2+lemonilia/limarp-llama2-v2 LORA
-----ehartford/WizardLM-1.0-Uncensored-Llama2-13b
-----Henk717/spring-dragon

Most formats could work, but my tests have all been done in Alpaca format and it works well.

### Instruction:
Your instruction or question here.
For roleplay purposes, I suggest the following - Write <CHAR NAME>'s next reply in a chat between <YOUR NAME> and <CHAR NAME>. Write a single reply only.

### Response:

Below is the Illustration for the Final Merge:

Once Again, thanks to Chargoddard for his amazing and simple ties-merge script, and Gryphe for their great BlockMerge_Gradient script. Thanks to the original model creators too!

Art by wada_kazu / わだかず (pixiv page private?)

Downloads last month: 7