Instructions to use Sao10K/Fimbulvetr-11B-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Sao10K/Fimbulvetr-11B-v2 with Transformers:

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

pipe = pipeline("text-generation", model="Sao10K/Fimbulvetr-11B-v2")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Sao10K/Fimbulvetr-11B-v2")
model = AutoModelForCausalLM.from_pretrained("Sao10K/Fimbulvetr-11B-v2", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Sao10K/Fimbulvetr-11B-v2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Sao10K/Fimbulvetr-11B-v2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Sao10K/Fimbulvetr-11B-v2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Sao10K/Fimbulvetr-11B-v2

SGLang

How to use Sao10K/Fimbulvetr-11B-v2 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 "Sao10K/Fimbulvetr-11B-v2" \
    --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": "Sao10K/Fimbulvetr-11B-v2",
		"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 "Sao10K/Fimbulvetr-11B-v2" \
        --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": "Sao10K/Fimbulvetr-11B-v2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use Sao10K/Fimbulvetr-11B-v2 with Docker Model Runner:
```
docker model run hf.co/Sao10K/Fimbulvetr-11B-v2
```

Absurdly slow on a I7 - 3060 - 64GB

#11

by webslug - opened Mar 13, 2024

Discussion

webslug

Mar 13, 2024

Ran this on a 3060 12 GB with 64GB of DDR 4 RAM.

It's incredibly slow and I was wondering if there were any settings I could adjust to remedy this?

webslug changed discussion title from Absurdly slow on a 3060 to Absurdly slow on a I7 - 3060 - 64GB Mar 13, 2024

Sao10K

Owner Mar 13, 2024

•

edited Mar 13, 2024

Ah, are you trying to run the full fp16 model? This is the unquantised repo, it's not really meant for basic inference. It'd take nearly 24gb of vram to run this one.

I'd run this instead with 12gb of vram:
https://huggingface.co/Sao10K/Fimbulvetr-11B-v2-GGUF -- GGUF (You can full offload on GPU or set max layers at q5_k_m, at 6k+ context... or go q6/q8 partially loading some of it in RAM)

https://huggingface.co/LoneStriker/Fimbulvetr-11B-v2-5.0bpw-h6-exl2 -- exl2 --> Fastest Speed, only pure GPU offloading

I'd take a look at koboldcpp for GGUF, it's literally an .exe and easy to run, or TabbyAPI for exl2

webslug

Mar 13, 2024

Thank you for your help!

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