Instructions to use SupraLabs/SupraBrain-50M with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use SupraLabs/SupraBrain-50M with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="SupraLabs/SupraBrain-50M", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("SupraLabs/SupraBrain-50M", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use SupraLabs/SupraBrain-50M with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "SupraLabs/SupraBrain-50M" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "SupraLabs/SupraBrain-50M", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/SupraLabs/SupraBrain-50M
- SGLang
How to use SupraLabs/SupraBrain-50M 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 "SupraLabs/SupraBrain-50M" \ --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": "SupraLabs/SupraBrain-50M", "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 "SupraLabs/SupraBrain-50M" \ --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": "SupraLabs/SupraBrain-50M", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use SupraLabs/SupraBrain-50M with Docker Model Runner:
docker model run hf.co/SupraLabs/SupraBrain-50M
Amazing work
A little late to the release, but I just wanted to reach out and say this is really amazing work! I love that you guys kept it on a low training budget and squeezed in so much!
So many new SLM labs to keep up with but I'm always rooting for you all.
Thank you so much!
We really appreciate this β€οΈπ€
out of all SLM labs supralabs makes SLM that actually can be used lol
I have to disagree, but I still stand by that SupraLabs is great. I don't see the point of going into a 1-month-old thread to say that other labs don't make anything usable. I mean look at Cactus? I would say they're making usable models at this size too.
Cactus:
https://huggingface.co/Cactus-Compute
Owen Song's Inflect
https://huggingface.co/owensong/Inflect-Micro-v2
Then at the same time the SLM community's research has always been amazing. I'm not sure where this comes from after a month of me congratulating them. I am sure not a lab, so if you think I am one then that's something else entirely. We also have bench-labs, banana, axiomic, etc that are doing amazing work just like Supra is. I don't know where the need to treat this like all of us at the level as if everyone is working at OpenAI or Anthropic, but whatever floats your boat.
Also, this not to say that Supra heavily reminds me of Liquid before they got where they are now. β€οΈ
out of all SLM labs supralabs makes SLM that actually can be used lol
Thanks for the support!!! π
I never said that you said that they're bad, you said:
out of all SLM labs supralabs makes SLM that actually can be used lol
well we can both agree that supralabs is great! anyways
tiiuae's Falcon H1 R Tiny model is also good:
https://huggingface.co/tiiuae/Falcon-H1-Tiny-90M-Instruct-GGUF
Yes, much bigger but still small.
Nice!
tiiuae's Falcon H1 R Tiny model is also good:
https://huggingface.co/tiiuae/Falcon-H1-Tiny-90M-Instruct-GGUFYes, much bigger but still small.
This is actually the peakest small model π₯π₯π₯
well we can both agree that supralabs is great! anyways
Thank you so much π€β€οΈπ