Instructions to use rbln/tiny-Llama4ForConditionalGeneration with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use rbln/tiny-Llama4ForConditionalGeneration with Transformers:

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

pipe = pipeline("image-text-to-text", model="rbln/tiny-Llama4ForConditionalGeneration")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("rbln/tiny-Llama4ForConditionalGeneration")
model = AutoModelForMultimodalLM.from_pretrained("rbln/tiny-Llama4ForConditionalGeneration", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use rbln/tiny-Llama4ForConditionalGeneration with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "rbln/tiny-Llama4ForConditionalGeneration"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "rbln/tiny-Llama4ForConditionalGeneration",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/rbln/tiny-Llama4ForConditionalGeneration

SGLang

How to use rbln/tiny-Llama4ForConditionalGeneration 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 "rbln/tiny-Llama4ForConditionalGeneration" \
    --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": "rbln/tiny-Llama4ForConditionalGeneration",
		"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 "rbln/tiny-Llama4ForConditionalGeneration" \
        --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": "rbln/tiny-Llama4ForConditionalGeneration",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use rbln/tiny-Llama4ForConditionalGeneration with Docker Model Runner:
```
docker model run hf.co/rbln/tiny-Llama4ForConditionalGeneration
```

tiny-Llama4ForConditionalGeneration

Tiny randomly initialized dense Llama 4 (Llama4ForConditionalGeneration) for optimum-rbln tests.

Text: 4 layers (3 chunked attention + 1 NoPE full attention), hidden 8, head_dim 128, attention_chunk_size 128, no MoE layers.
Vision: 2 layers, hidden 16.
The weights are random; no tokenizer or processor is included. Load them from a Llama 4 model such as meta-llama/Llama-4-Scout-17B-16E-Instruct.

Downloads last month: 4

Safetensors

Model size

3.31M params

Tensor type

F32