Instructions to use togethercomputer/RedPajama-INCITE-Instruct-3B-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use togethercomputer/RedPajama-INCITE-Instruct-3B-v1 with Transformers:

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

pipe = pipeline("text-generation", model="togethercomputer/RedPajama-INCITE-Instruct-3B-v1")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("togethercomputer/RedPajama-INCITE-Instruct-3B-v1")
model = AutoModelForCausalLM.from_pretrained("togethercomputer/RedPajama-INCITE-Instruct-3B-v1")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use togethercomputer/RedPajama-INCITE-Instruct-3B-v1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "togethercomputer/RedPajama-INCITE-Instruct-3B-v1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "togethercomputer/RedPajama-INCITE-Instruct-3B-v1",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/togethercomputer/RedPajama-INCITE-Instruct-3B-v1

SGLang

How to use togethercomputer/RedPajama-INCITE-Instruct-3B-v1 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 "togethercomputer/RedPajama-INCITE-Instruct-3B-v1" \
    --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": "togethercomputer/RedPajama-INCITE-Instruct-3B-v1",
		"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 "togethercomputer/RedPajama-INCITE-Instruct-3B-v1" \
        --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": "togethercomputer/RedPajama-INCITE-Instruct-3B-v1",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use togethercomputer/RedPajama-INCITE-Instruct-3B-v1 with Docker Model Runner:
```
docker model run hf.co/togethercomputer/RedPajama-INCITE-Instruct-3B-v1
```

Running the model on MPS backend (Macbook GPUs) hangs indefinitely

by ggaabe - opened May 6, 2023

Discussion

ggaabe

May 6, 2023

•

edited May 6, 2023

I'm trying to run the GPU example but with a Macbook M2 Max and it seems trying to use model.to("mps") simply hangs forever without any error message:

tokenizer = AutoTokenizer.from_pretrained(
    path)
model = AutoModelForCausalLM.from_pretrained(
    path, torch_dtype=torch.float16)

print("Device 0:", model.device) # prints: Device 0: cpu
mps_device = torch.device("mps")
print("Device 1:", mps_device)  # prints: Device 1: mps
model = model.to(mps_device)
print("Device 2:", model.device) # ^^ The above line hangs forever, this line is never reached.

Any thoughts on what I could do to get this to run with GPU inference?

PyTorch Version: 2.1.0.dev20230505

EugeneLu

May 8, 2023

What is the size of GPU memory for your M2 Max?

ggaabe

May 8, 2023

My bad. There was something probably misconfigured with my env though I have no idea what. I had used pip for all the env packages; I retried with conda and it all worked.

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