Instructions to use tiny-random/gemma-3n-dim4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tiny-random/gemma-3n-dim4 with Transformers:

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

pipe = pipeline("image-text-to-text", model="tiny-random/gemma-3n-dim4")
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("tiny-random/gemma-3n-dim4")
model = AutoModelForImageTextToText.from_pretrained("tiny-random/gemma-3n-dim4")
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]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use tiny-random/gemma-3n-dim4 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tiny-random/gemma-3n-dim4"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tiny-random/gemma-3n-dim4",
		"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/tiny-random/gemma-3n-dim4

SGLang

How to use tiny-random/gemma-3n-dim4 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 "tiny-random/gemma-3n-dim4" \
    --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": "tiny-random/gemma-3n-dim4",
		"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 "tiny-random/gemma-3n-dim4" \
        --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": "tiny-random/gemma-3n-dim4",
		"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 tiny-random/gemma-3n-dim4 with Docker Model Runner:
```
docker model run hf.co/tiny-random/gemma-3n-dim4
```

yujiepan commited on Jul 12, 2025

Commit

46f5b79

verified ·

1 Parent(s): 0bdf55c

Upload folder using huggingface_hub

Browse files

Files changed (3) hide show

README.md +1 -41
config.json +1 -5
preprocessor_config.json +1 -1

README.md CHANGED Viewed

@@ -15,47 +15,7 @@ This tiny model is for debugging. It is randomly initialized with the config ada
 ### Example usage:
 ```python
-import torch
-from transformers import pipeline
-model_id = "tiny-random/gemma-3n-dim4"
-pipe = pipeline(
-    task="image-text-to-text",
-    model=model_id,
-    device=0,
-    torch_dtype=torch.bfloat16
-)
-# temporary patch for audio tower
-from accelerate.hooks import ModelHook, add_hook_to_module
-class EnsureDtype(ModelHook):
-    def pre_forward(self, module, *args, **kwargs):
-        args = list(args)
-        args[0] = args[0].to(module.dtype)
-        return super().pre_forward(module, *args, **kwargs)
-add_hook_to_module(pipe.model.audio_tower, EnsureDtype())
-messages = [
-    {
-        "role": "system",
-        "content": [
-            {"type": "text", "text": "You are a helpful assistant."}
-        ]
-    },
-    {
-        "role": "user",
-        "content": [
-            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/pipeline-cat-chonk.jpeg"},
-            # audio is buggy for now: bf16 x fp32
-            {"type": "audio", "url": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Audio/glass-breaking-151256.mp3"},
-            {"type": "text", "text": "Which image is cuter?"},
-        ]
-    },
-]
-result = pipe(messages, min_new_tokens=512, max_new_tokens=512, do_sample=True)
-print(result)
 ```
 ### Codes to create this repo:

 ### Example usage:
 ```python
+{code_to_run.strip()}
 ```
 ### Codes to create this repo:

config.json CHANGED Viewed

@@ -10,7 +10,6 @@
     "conf_conv_kernel_size": 5,
     "conf_num_attention_heads": 2,
     "conf_num_hidden_layers": 2,
-    "conf_positional_bias_size": 256,
     "conf_reduction_factor": 4,
     "conf_residual_weight": 0.5,
     "gradient_clipping": 10000000000.0,
@@ -22,7 +21,6 @@
       128,
       32
     ],
-    "sscp_conv_eps": 0.001,
     "sscp_conv_group_norm_eps": 0.001,
     "sscp_conv_kernel_size": [
       [
@@ -71,7 +69,6 @@
     "altup_active_idx": 0,
     "altup_coef_clip": 120.0,
     "altup_correct_scale": true,
-    "altup_lr_multiplier": 1.0,
     "altup_num_inputs": 4,
     "attention_bias": false,
     "attention_dropout": 0.0,
@@ -100,7 +97,6 @@
     "num_hidden_layers": 4,
     "num_key_value_heads": 1,
     "num_kv_shared_layers": 2,
-    "query_pre_attn_scalar": 256,
     "rms_norm_eps": 1e-06,
     "rope_local_base_freq": 10000.0,
     "rope_scaling": null,
@@ -115,7 +111,7 @@
   "transformers_version": "4.54.0.dev0",
   "vision_config": {
     "architecture": "mobilenetv5_300m_enc",
-    "do_pooling": true,
     "hidden_size": 2048,
     "initializer_range": 0.02,
     "label_names": [

     "conf_conv_kernel_size": 5,
     "conf_num_attention_heads": 2,
     "conf_num_hidden_layers": 2,
     "conf_reduction_factor": 4,
     "conf_residual_weight": 0.5,
     "gradient_clipping": 10000000000.0,
       128,
       32
     ],
     "sscp_conv_group_norm_eps": 0.001,
     "sscp_conv_kernel_size": [
       [
     "altup_active_idx": 0,
     "altup_coef_clip": 120.0,
     "altup_correct_scale": true,
     "altup_num_inputs": 4,
     "attention_bias": false,
     "attention_dropout": 0.0,
     "num_hidden_layers": 4,
     "num_key_value_heads": 1,
     "num_kv_shared_layers": 2,
     "rms_norm_eps": 1e-06,
     "rope_local_base_freq": 10000.0,
     "rope_scaling": null,
   "transformers_version": "4.54.0.dev0",
   "vision_config": {
     "architecture": "mobilenetv5_300m_enc",
+    "do_pooling": false,
     "hidden_size": 2048,
     "initializer_range": 0.02,
     "label_names": [

preprocessor_config.json CHANGED Viewed

@@ -41,7 +41,7 @@
   "processor_class": "Gemma3nProcessor",
   "resample": 2,
   "rescale_factor": 0.00392156862745098,
-  "return_attention_mask": false,
   "return_tensors": null,
   "sampling_rate": 16000,
   "size": {

   "processor_class": "Gemma3nProcessor",
   "resample": 2,
   "rescale_factor": 0.00392156862745098,
+  "return_attention_mask": true,
   "return_tensors": null,
   "sampling_rate": 16000,
   "size": {