Instructions to use google/gemma-4-31B-it with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use google/gemma-4-31B-it with Transformers:

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

pipe = pipeline("image-text-to-text", model="google/gemma-4-31B-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-4-31B-it")
model = AutoModelForImageTextToText.from_pretrained("google/gemma-4-31B-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-4-31B-it with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "google/gemma-4-31B-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-4-31B-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-4-31B-it

SGLang

How to use google/gemma-4-31B-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-4-31B-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-4-31B-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-4-31B-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-4-31B-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-4-31B-it with Docker Model Runner:
```
docker model run hf.co/google/gemma-4-31B-it
```

Drops in performance randomly VLLM on H200 and B200 with 2 GPUs

#104

by jtvino - opened 7 days ago

Discussion

jtvino

7 days ago

•

edited 7 days ago

Model seems to perform really well and then randomly will drop in t/s when using vllm when running on b200 or h200 with 2 GPUs
It will run at like 101 t/s and then drop to 0.1 tokens/s with only 2 reqs in queue while GPU KV cache usage is only showing 2.6%

 vllm serve google/gemma-4-31B-it --served-model-name google/gemma-4-31B-it --kv-cache-dtype fp8 --max-model-len 200000 --dtype auto --async-scheduling --tensor-parallel-size 2 --trust-remote-code --gpu-memory-utilization 0.90 --enable-chunked-prefill --enable-auto-tool-choice  --reasoning-parser gemma4  --tool-call-parser gemma4  --limit-mm-per-prompt '{"image": 2, "audio": 1}' --chat-template examples/tool_chat_template_gemma4.jinja --max-num-seqs 64

using vllm v0.20.0. Issue exists in v0.19.1 and v0.20.1

(APIServer pid=434) INFO 05-11 17:46:04 [loggers.py:271] Engine 000: Avg prompt throughput: 145.1 tokens/s, Avg generation throughput: 133.7 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 0.3%, Prefix cache hit rate: 89.3%, MM cache hit rate: 76.6%
(APIServer pid=434) INFO:     10.129.3.35:52484 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO:     10.129.3.35:34962 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO:     10.129.3.35:52484 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO:     10.129.3.35:52484 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:46:14 [loggers.py:271] Engine 000: Avg prompt throughput: 160.0 tokens/s, Avg generation throughput: 87.7 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 0.3%, Prefix cache hit rate: 89.6%, MM cache hit rate: 76.6%
(APIServer pid=434) INFO:     10.129.3.35:52484 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:46:24 [loggers.py:271] Engine 000: Avg prompt throughput: 38.9 tokens/s, Avg generation throughput: 23.7 tokens/s, Running: 0 reqs, Waiting: 0 reqs, GPU KV cache usage: 0.0%, Prefix cache hit rate: 89.6%, MM cache hit rate: 76.6%
(APIServer pid=434) INFO:     10.129.3.35:54410 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:46:34 [loggers.py:271] Engine 000: Avg prompt throughput: 32.4 tokens/s, Avg generation throughput: 18.3 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 0.3%, Prefix cache hit rate: 89.6%, MM cache hit rate: 76.6%
(APIServer pid=434) INFO:     10.129.3.35:54410 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO:     10.130.1.111:40524 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:46:44 [loggers.py:271] Engine 000: Avg prompt throughput: 131.8 tokens/s, Avg generation throughput: 101.2 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 0.2%, Prefix cache hit rate: 89.6%, MM cache hit rate: 76.6%
(APIServer pid=434) INFO:     10.130.1.111:33690 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:46:54 [loggers.py:271] Engine 000: Avg prompt throughput: 3.6 tokens/s, Avg generation throughput: 182.2 tokens/s, Running: 2 reqs, Waiting: 0 reqs, GPU KV cache usage: 0.4%, Prefix cache hit rate: 89.6%, MM cache hit rate: 76.6%
(APIServer pid=434) INFO:     127.0.0.1:52124 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO:     127.0.0.1:52124 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:47:04 [loggers.py:271] Engine 000: Avg prompt throughput: 344.0 tokens/s, Avg generation throughput: 120.8 tokens/s, Running: 0 reqs, Waiting: 0 reqs, GPU KV cache usage: 0.0%, Prefix cache hit rate: 89.5%, MM cache hit rate: 76.6%
(APIServer pid=434) INFO:     127.0.0.1:52124 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO:     10.130.1.111:33690 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO:     10.129.3.35:54736 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:47:14 [loggers.py:271] Engine 000: Avg prompt throughput: 702.7 tokens/s, Avg generation throughput: 18.9 tokens/s, Running: 2 reqs, Waiting: 1 reqs, GPU KV cache usage: 2.2%, Prefix cache hit rate: 88.7%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     10.129.3.35:59972 - "POST /v1/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:47:24 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.2 tokens/s, Running: 2 reqs, Waiting: 1 reqs, GPU KV cache usage: 2.5%, Prefix cache hit rate: 88.7%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     10.130.1.111:58378 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO:     10.130.1.111:58396 - "POST /v1/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO:     10.129.3.35:52348 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:47:34 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.1 tokens/s, Running: 2 reqs, Waiting: 2 reqs, GPU KV cache usage: 2.6%, Prefix cache hit rate: 88.7%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     10.129.3.35:52376 - "POST /v1/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO:     10.130.1.111:57256 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:47:44 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.1 tokens/s, Running: 2 reqs, Waiting: 3 reqs, GPU KV cache usage: 2.7%, Prefix cache hit rate: 88.7%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     10.130.1.111:37562 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:47:54 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.1 tokens/s, Running: 2 reqs, Waiting: 6 reqs, GPU KV cache usage: 2.8%, Prefix cache hit rate: 88.7%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     10.129.3.35:33876 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:48:04 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.1 tokens/s, Running: 2 reqs, Waiting: 5 reqs, GPU KV cache usage: 3.0%, Prefix cache hit rate: 88.7%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO 05-11 17:48:14 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.1 tokens/s, Running: 2 reqs, Waiting: 5 reqs, GPU KV cache usage: 3.1%, Prefix cache hit rate: 88.7%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     10.129.3.35:46682 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO:     10.130.1.111:46832 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:48:24 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.1 tokens/s, Running: 2 reqs, Waiting: 4 reqs, GPU KV cache usage: 3.2%, Prefix cache hit rate: 88.7%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     10.130.1.111:47290 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO:     10.129.3.35:46792 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:48:34 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 2 reqs, Waiting: 4 reqs, GPU KV cache usage: 3.2%, Prefix cache hit rate: 88.7%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO 05-11 17:48:44 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.1 tokens/s, Running: 2 reqs, Waiting: 5 reqs, GPU KV cache usage: 3.3%, Prefix cache hit rate: 88.7%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO 05-11 17:48:54 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.1 tokens/s, Running: 2 reqs, Waiting: 7 reqs, GPU KV cache usage: 3.5%, Prefix cache hit rate: 88.7%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO 05-11 17:49:04 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 2 reqs, Waiting: 7 reqs, GPU KV cache usage: 3.5%, Prefix cache hit rate: 88.7%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     10.129.3.35:44674 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:49:14 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.1 tokens/s, Running: 2 reqs, Waiting: 5 reqs, GPU KV cache usage: 3.6%, Prefix cache hit rate: 88.7%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO 05-11 17:49:24 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 2 reqs, Waiting: 5 reqs, GPU KV cache usage: 3.6%, Prefix cache hit rate: 88.7%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO 05-11 17:49:34 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.1 tokens/s, Running: 6 reqs, Waiting: 2 reqs, GPU KV cache usage: 4.0%, Prefix cache hit rate: 88.3%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO 05-11 17:49:44 [loggers.py:271] Engine 000: Avg prompt throughput: 13094.9 tokens/s, Avg generation throughput: 1.5 tokens/s, Running: 6 reqs, Waiting: 1 reqs, GPU KV cache usage: 4.6%, Prefix cache hit rate: 88.3%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO 05-11 17:49:54 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 1.0 tokens/s, Running: 6 reqs, Waiting: 1 reqs, GPU KV cache usage: 4.9%, Prefix cache hit rate: 88.3%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO 05-11 17:50:04 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 1.0 tokens/s, Running: 6 reqs, Waiting: 1 reqs, GPU KV cache usage: 5.1%, Prefix cache hit rate: 88.3%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     10.130.1.111:34970 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO:     10.130.1.111:34986 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:50:14 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.5 tokens/s, Running: 6 reqs, Waiting: 1 reqs, GPU KV cache usage: 5.2%, Prefix cache hit rate: 88.3%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     10.129.3.35:54102 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:50:24 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.5 tokens/s, Running: 6 reqs, Waiting: 2 reqs, GPU KV cache usage: 5.4%, Prefix cache hit rate: 88.3%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO 05-11 17:50:34 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.5 tokens/s, Running: 7 reqs, Waiting: 3 reqs, GPU KV cache usage: 6.0%, Prefix cache hit rate: 88.2%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     10.130.1.111:40082 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:50:44 [loggers.py:271] Engine 000: Avg prompt throughput: 9177.3 tokens/s, Avg generation throughput: 2.3 tokens/s, Running: 10 reqs, Waiting: 0 reqs, GPU KV cache usage: 7.6%, Prefix cache hit rate: 88.1%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO 05-11 17:50:54 [loggers.py:271] Engine 000: Avg prompt throughput: 222.1 tokens/s, Avg generation throughput: 74.7 tokens/s, Running: 9 reqs, Waiting: 0 reqs, GPU KV cache usage: 5.7%, Prefix cache hit rate: 88.1%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     127.0.0.1:52124 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:51:04 [loggers.py:271] Engine 000: Avg prompt throughput: 342.4 tokens/s, Avg generation throughput: 289.3 tokens/s, Running: 7 reqs, Waiting: 0 reqs, GPU KV cache usage: 5.1%, Prefix cache hit rate: 88.1%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     127.0.0.1:52124 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO:     127.0.0.1:52124 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:51:14 [loggers.py:271] Engine 000: Avg prompt throughput: 35.7 tokens/s, Avg generation throughput: 264.1 tokens/s, Running: 6 reqs, Waiting: 0 reqs, GPU KV cache usage: 4.9%, Prefix cache hit rate: 88.1%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     127.0.0.1:52124 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:51:24 [loggers.py:271] Engine 000: Avg prompt throughput: 8.4 tokens/s, Avg generation throughput: 185.3 tokens/s, Running: 4 reqs, Waiting: 0 reqs, GPU KV cache usage: 4.3%, Prefix cache hit rate: 88.1%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO 05-11 17:51:34 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 135.9 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 0.6%, Prefix cache hit rate: 88.1%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO 05-11 17:51:44 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 74.6 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 0.6%, Prefix cache hit rate: 88.1%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO 05-11 17:51:54 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 44.7 tokens/s, Running: 0 reqs, Waiting: 0 reqs, GPU KV cache usage: 0.0%, Prefix cache hit rate: 88.1%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO 05-11 17:52:04 [loggers.py:271] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 0 reqs, Waiting: 0 reqs, GPU KV cache usage: 0.0%, Prefix cache hit rate: 88.1%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     10.130.1.111:58502 - "POST /v1/chat/completions HTTP/1.1" 200 OK
(APIServer pid=434) INFO 05-11 17:52:24 [loggers.py:271] Engine 000: Avg prompt throughput: 172.9 tokens/s, Avg generation throughput: 48.0 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 0.2%, Prefix cache hit rate: 88.1%, MM cache hit rate: 76.7%
(APIServer pid=434) INFO:     10.130.1.111:53106 - "POST /v1/completions HTTP/1.1" 200 OK

(APIServer pid=434) INFO 05-11 17:52:34 [loggers.py:271] Engine 000: Avg prompt throughput: 99.4 tokens/s, Avg generation throughput: 36.8 tokens/s, Running: 0 reqs, Waiting: 0 reqs, GPU KV cache usage: 0.0%, Prefix cache hit rate: 88.1%, MM cache hit rate: 76.7%

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