Instructions to use Motif-Technologies/Motif-3-Beta with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Motif-Technologies/Motif-3-Beta with Transformers:

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

pipe = pipeline("text-generation", model="Motif-Technologies/Motif-3-Beta", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Motif-Technologies/Motif-3-Beta", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Motif-Technologies/Motif-3-Beta with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Motif-Technologies/Motif-3-Beta"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Motif-Technologies/Motif-3-Beta",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Motif-Technologies/Motif-3-Beta

SGLang

How to use Motif-Technologies/Motif-3-Beta 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 "Motif-Technologies/Motif-3-Beta" \
    --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": "Motif-Technologies/Motif-3-Beta",
		"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 "Motif-Technologies/Motif-3-Beta" \
        --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": "Motif-Technologies/Motif-3-Beta",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Motif-Technologies/Motif-3-Beta with Docker Model Runner:
```
docker model run hf.co/Motif-Technologies/Motif-3-Beta
```

need nvfp4 and int4 quantization

#10

by WiSFoR - opened 6 days ago

Discussion

WiSFoR

6 days ago

or at least provide us with proper parameter for quantization as native quantization using intel auto-round seems to fails due to architecture issue.

dongseokmotif

Motif Technologies org 4 days ago

Thanks, @WiSFoR . The auto-round failure is expected right now: Motif-3's custom modules (GDLA attention, Grouped PolyNorm, modified mHC) aren't recognized by architecture-agnostic quantization tooling, so those paths break.

Guidance for now:

8-bit: use blockwise quantization (per-block scales, not per-tensor). The custom layers are sensitive enough that per-tensor 8-bit degrades them.
4-bit (int4 / nvfp4): still under testing on our side, so we don't have a validated config to recommend yet. As a reference for which layers need higher precision, @avlp12 's community MLX builds (discussion #6) use mixed-bit + layerwise DWQ.

We'll share a proper quantization config once the 4-bit path is validated.

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