Instructions to use alpindale/WizardLM-2-8x22B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use alpindale/WizardLM-2-8x22B with Transformers:

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

pipe = pipeline("text-generation", model="alpindale/WizardLM-2-8x22B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("alpindale/WizardLM-2-8x22B")
model = AutoModelForCausalLM.from_pretrained("alpindale/WizardLM-2-8x22B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
HuggingChat
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use alpindale/WizardLM-2-8x22B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "alpindale/WizardLM-2-8x22B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "alpindale/WizardLM-2-8x22B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/alpindale/WizardLM-2-8x22B

SGLang

How to use alpindale/WizardLM-2-8x22B 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 "alpindale/WizardLM-2-8x22B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "alpindale/WizardLM-2-8x22B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "alpindale/WizardLM-2-8x22B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "alpindale/WizardLM-2-8x22B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use alpindale/WizardLM-2-8x22B with Docker Model Runner:
```
docker model run hf.co/alpindale/WizardLM-2-8x22B
```

winner winner chicken dinner

by Walmart-the-bag - opened Apr 16, 2024

Discussion

Walmart-the-bag

Apr 16, 2024

Not surprised Microsoft deleted it. Stupid company. GG.

cohere is better anyways

jreoka

Apr 16, 2024

They could at least say why they nuked it

ChuckMcSneed

Apr 16, 2024

Probably something dumb like ethics.

ChuckMcSneed

Apr 16, 2024

https://twitter.com/WizardLM_AI/status/1780101465950105775
Yup, seems like ethics.

MaziyarPanahi

Apr 16, 2024

It’s OK, they all have their own process and in big companies things like this happen.
But, this is a hell of a model! I have been playing with Q2, and I am just impressed!
Hopefully, they come back quickly and also they add the 70B model they missed in the first run.

deleted

Apr 16, 2024

So thanks @alpindale

jukofyork

Apr 16, 2024

I'm pretty sure this can be dewokefied: https://huggingface.co/alpindale/WizardLM-2-8x22B/discussions/5

I have Mixtral-8x22B-v0.1 downloading and got WizardLM-2-8x22B before it was pulled, but it will likely be 2 days before I can get the other 2 model to try this...

jukofyork

Apr 16, 2024

•

edited Apr 16, 2024

It’s OK, they all have their own process and in big companies things like this happen.
But, this is a hell of a model! I have been playing with Q2, and I am just impressed!
Hopefully, they come back quickly and also they add the 70B model they missed in the first run.

Yeah, just got this running as Q4_K_S and it is really good at coding... Crazy times; nothing for a few months then all this lot drops at once! 3 days in a row of "yep, this is the best coding model yet, oh but...".

openbmb/Eurux-8x22b-nca is next I think... :D

I've uploaded the imatrix I created from FP16 using groups_merged.txt if anybody wants to use it: https://huggingface.co/jukofyork/WizardLM-2-8x22B-imatrix

Not sure how much effect it has on Q4, but for lower quants it will probably help a lot and likely a better choice than wiki.train.raw if you want to use it for coding.

MaziyarPanahi

Apr 16, 2024

I've uploaded the imatrix I created from FP16 using groups_merged.txt if anybody wants to use it: https://huggingface.co/jukofyork/WizardLM-2-8x22B-imatrix

I just finished making imatrix based on the same txt file! Upload the IQ-1 quants as we speak. I wish I would have seen this first, it took forever to make that imtarix data file!

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