Instructions to use google/gemma-3-4b-pt with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use google/gemma-3-4b-pt with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="google/gemma-3-4b-pt")# Load model directly from transformers import AutoProcessor, AutoModelForImageTextToText processor = AutoProcessor.from_pretrained("google/gemma-3-4b-pt") model = AutoModelForImageTextToText.from_pretrained("google/gemma-3-4b-pt") - Notebooks
- Google Colab
- Kaggle
- Local Apps
- vLLM
How to use google/gemma-3-4b-pt with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "google/gemma-3-4b-pt" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "google/gemma-3-4b-pt", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/google/gemma-3-4b-pt
- SGLang
How to use google/gemma-3-4b-pt 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 "google/gemma-3-4b-pt" \ --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": "google/gemma-3-4b-pt", "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 "google/gemma-3-4b-pt" \ --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": "google/gemma-3-4b-pt", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use google/gemma-3-4b-pt with Docker Model Runner:
docker model run hf.co/google/gemma-3-4b-pt
Is the config.json wrong?
The 4b-pt config.json is much same as 4b-it, and diffrent from 1b-pt version.For example,no vocab_size prop in 4b but can be found in 1b.Is this intended or a mistake?
Same problem here
Hi ,
Thank you for pointing this out! You're correct — the config.json for the 4b-pt and 4b-it versions appears to omit the vocab_size field, while it's present in the 1b-pt version.
This could be intentional if the tokenizer is assumed to be loaded independently and the model config relies on tokenizer metadata. In many recent transformer-based models, the vocab_size is not strictly required in config.json during inference. Thank you.