Instructions to use LumiOpen/Poro-34B-chat with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use LumiOpen/Poro-34B-chat with Transformers:

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

pipe = pipeline("text-generation", model="LumiOpen/Poro-34B-chat")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("LumiOpen/Poro-34B-chat")
model = AutoModelForCausalLM.from_pretrained("LumiOpen/Poro-34B-chat")
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]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use LumiOpen/Poro-34B-chat with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "LumiOpen/Poro-34B-chat"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "LumiOpen/Poro-34B-chat",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/LumiOpen/Poro-34B-chat

SGLang

How to use LumiOpen/Poro-34B-chat 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 "LumiOpen/Poro-34B-chat" \
    --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": "LumiOpen/Poro-34B-chat",
		"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 "LumiOpen/Poro-34B-chat" \
        --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": "LumiOpen/Poro-34B-chat",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use LumiOpen/Poro-34B-chat with Docker Model Runner:
```
docker model run hf.co/LumiOpen/Poro-34B-chat
```

Some tips on running the model

by alexamon - opened Sep 16, 2024

Discussion

alexamon

Sep 16, 2024

Hi,

can you share some tips on running the model?

Under 'Use this model' I can see that with transformers it should work like the below code suggests and some additional steps are surely needed to install required dependancies. Can you perhaps share a notebook on running this model?

I tried on lightning AI with L40 GPU and failed to run the model. It seemed like it wasn't loading the model to GPU memory.
I also tried on lightning AI with just 32 CPU and 128GB ram without GPU and it did use up all the memory but still didn't run.

I must be doing something incorrectly.

Use a pipeline as a high-level helper

from transformers import pipeline

messages = [
{"role": "user", "content": "Who are you?"},
]
pipe = pipeline("text-generation", model="LumiOpen/Poro-34B-chat")
pipe(messages)

Kind regards,
John

sajvir

LumiOpen org Sep 17, 2024

The only real dependency is transformers as that should pull everything else that is needed during install.

The following code should work (can't test it at this moment as the model is quite large):

import torch
from transformers import pipeline

messages = [
  {"role": "user", "content": "Who are you?"},
]
pipe = pipeline("text-generation", model="LumiOpen/Poro-34B-chat", torch_dtype=torch.bfloat16, device="cuda", max_length=512)
pipe(messages)

You can change the max_length to whatever you want (the default is 20).

Is there an error message you got when trying to run the model?

Br,
Antti

sajvir

LumiOpen org Sep 17, 2024

Actually, now that I look at the specs of the L40, I don't think the model will fit into its memory. If you replace device="cuda" with device="cpu" you should be able to run it on CPU. It will be very slow though.

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