Instructions to use concavity-ai/superlinear-exp-v0.1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use concavity-ai/superlinear-exp-v0.1 with Transformers:

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

pipe = pipeline("text-generation", model="concavity-ai/superlinear-exp-v0.1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("concavity-ai/superlinear-exp-v0.1", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use concavity-ai/superlinear-exp-v0.1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "concavity-ai/superlinear-exp-v0.1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "concavity-ai/superlinear-exp-v0.1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/concavity-ai/superlinear-exp-v0.1

SGLang

How to use concavity-ai/superlinear-exp-v0.1 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 "concavity-ai/superlinear-exp-v0.1" \
    --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": "concavity-ai/superlinear-exp-v0.1",
		"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 "concavity-ai/superlinear-exp-v0.1" \
        --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": "concavity-ai/superlinear-exp-v0.1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use concavity-ai/superlinear-exp-v0.1 with Docker Model Runner:
```
docker model run hf.co/concavity-ai/superlinear-exp-v0.1
```

Q8 and smaller quants

by whoisjeremylam - opened Feb 7

Discussion

whoisjeremylam

Feb 7

hi, this is super interesting. Is there a possibility of a Q8 or smaller quant? At Q8, the handful of us with 48 gb GPUs will be able to try this out. 😊

Jeffbulger82

Feb 7

I’m trying to patch all the converters and experts now to quantize it. It’s quite the challenge. I did get it converted to a GGUF though.

hif1000

concavity.ai org Feb 8

Thanks for the interest! Right now I am doing some additional training on the model, will quantize it to 8-bit afterwards.

Jbulger82

Feb 8

I had some issues with
blk.0.ssm_a tensor is exporting as 64, 1, 1, 1,… my project was expecting 1, 64 ..but it’s all good. I look forward to your quantization. If you could post any quantized gguf I can make magic happen :) . But I’ll work with whatever you share!I love this concept! I can run a nemotron3 nano with 2 million context inconsumer hardware it’s a little slow, but it works. I think your logic in this project is the solution too many many things! When I make the breakthrough, I’ll share it with you absolutely thanks again.

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