Instructions to use google/gemma-3-27b-it with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use google/gemma-3-27b-it with Transformers:

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

pipe = pipeline("image-text-to-text", model="google/gemma-3-27b-it")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("google/gemma-3-27b-it")
model = AutoModelForImageTextToText.from_pretrained("google/gemma-3-27b-it")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
HuggingChat
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use google/gemma-3-27b-it with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "google/gemma-3-27b-it"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "google/gemma-3-27b-it",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/google/gemma-3-27b-it

SGLang

How to use google/gemma-3-27b-it 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-3-27b-it" \
    --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": "google/gemma-3-27b-it",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

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-3-27b-it" \
        --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": "google/gemma-3-27b-it",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use google/gemma-3-27b-it with Docker Model Runner:
```
docker model run hf.co/google/gemma-3-27b-it
```

Is there a minimum CUDA and/or BitsAndBytes version requirement?

#88

by deathknight0 - opened Aug 18, 2025

Discussion

deathknight0

Aug 18, 2025

•

edited Aug 18, 2025

I'm running into issues using BitsAndBytes for quantization. I keep getting this cryptic CUDA error:
nf4_config = BitsAndBytesConfig( load_in_4bit=True, bnb_4bit_quant_type="nf4", bnb_4bit_use_double_quant=True, bnb_4bit_compute_dtype=torch.bfloat16 ) model = Gemma3ForConditionalGeneration.from_pretrained( model_dir, quantization_config = nf4_config ).eval()


..... (rest of code)
output = model.generate(**inputs, max_new_tokens=100)

Error: output = model.generate(**inputs, max_new_tokens=100) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ RuntimeError: CUDA error: device-side assert triggered Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.
When I don't use BitsAndBytes (I can't actually do this to practically use the model as I only have a single RTX 3090; I just did it for debugging), I get this (presumably) CUDA error:
model = Gemma3ForConditionalGeneration.from_pretrained( model_dir, device_map='auto', torch_dtype=torch.bfloat16 ).eval() .... output = model.generate(**inputs, max_new_tokens=100)

Error: output = model.generate(**inputs, max_new_tokens=100) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ RuntimeError: cutlassF: no kernel found to launch!
Libraries :
torch 2.6.0+cu124
transformers 4.53.0
triton-windows 3.2.0.post11
accelerate 1.6.0
bitsandbytes 0.47.0

Is there a minimum torch/CUDA/BnB requirement to use this model?

Thanks in advance!

BalakrishnaCh

Google org Aug 19, 2025

Hi,

Thanks for bringing this to our attention, there is no simple single minimum requirement for torch/CUDA. It's a complex compatibility matrix. What matters is that your PyTorch/CUDA binaries, NVIDIA driver, and bitsandbytes installation must all be compatible with each other and your GPU's compute capability and the all the libraries make sure to be up to date. Your current combination appears to be broken, which is a common issue with the rapid pace of development in the AI library ecosystem.

I have successfully able to done the 4bit-quantization with the same above mentioned parameters without any issues in my local. Please try with above suggested things and let us know if you required any further assistance.

Thanks.

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