Instructions to use AMAImedia/NOESIS-Gemma3-4B-BF16 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AMAImedia/NOESIS-Gemma3-4B-BF16 with Transformers:

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

pipe = pipeline("text-generation", model="AMAImedia/NOESIS-Gemma3-4B-BF16")
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, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("AMAImedia/NOESIS-Gemma3-4B-BF16")
model = AutoModelForMultimodalLM.from_pretrained("AMAImedia/NOESIS-Gemma3-4B-BF16", device_map="auto")
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 Settings

vLLM

How to use AMAImedia/NOESIS-Gemma3-4B-BF16 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "AMAImedia/NOESIS-Gemma3-4B-BF16"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AMAImedia/NOESIS-Gemma3-4B-BF16",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/AMAImedia/NOESIS-Gemma3-4B-BF16

SGLang

How to use AMAImedia/NOESIS-Gemma3-4B-BF16 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 "AMAImedia/NOESIS-Gemma3-4B-BF16" \
    --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": "AMAImedia/NOESIS-Gemma3-4B-BF16",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "AMAImedia/NOESIS-Gemma3-4B-BF16" \
        --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": "AMAImedia/NOESIS-Gemma3-4B-BF16",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use AMAImedia/NOESIS-Gemma3-4B-BF16 with Docker Model Runner:
```
docker model run hf.co/AMAImedia/NOESIS-Gemma3-4B-BF16
```

Released as part of the NOESIS Professional Multilingual Dubbing Automation Platform (framework: DHCF-FNO — Deterministic Hybrid Control Framework for Frozen Neural Operators).

Founder: Ilia Bolotnikov
Organization: AMAImedia.com
X (Twitter): @AMAImediacom
LinkedIn: Ilia Bolotnikov
Telegram: @djbionicl
NOESIS version: v16.1
Release date: 2026-08

NOESIS-Gemma3-4B-BF16

NOESIS-Gemma3-4B-BF16 is an AMAImedia BF16 fine-tuned and packaged Gemma 3 4B multimodal model used as the primary 4B Scenema student in the NOESIS audio-dubbing pipeline. It is provided as a full safetensors model together with its tokenizer, generation configuration, and the paired Scenema adapter/runtime files.

Model details

Property	Value
Base model	`unsloth/gemma-3-4b-it`
Architecture	`Gemma3ForConditionalGeneration`
Precision	BF16
Text hidden size	2560
Text layers	34
Context length	131,072 tokens
Vision encoder	SigLIP, 896px image size
Training organization	AMAImedia / NOESIS

The repository contains a NOESIS Scenema student bundle. The scenema_adapter_v3.pt file is a paired runtime adapter and is intentionally kept separate from the transformer weights; load it together with the model when using the Scenema audio cross-attention pipeline.

Repository contents

model.safetensors — BF16 model weights.
config.json — Gemma 3 multimodal architecture configuration.
tokenizer.json, tokenizer_config.json, and chat_template.jinja — tokenizer and chat template files.
generation_config.json — generation defaults.
scenema_adapter_v3.pt and scenema_adapter_config.json — paired Scenema adapter.
scenema_wrapper.py — runtime integration wrapper.
LICENSE — licensing and attribution information.

Basic Transformers loading

from transformers import AutoProcessor, AutoModelForCausalLM

MODEL_ID = "AMAImedia/NOESIS-Gemma3-4B-BF16"

processor = AutoProcessor.from_pretrained(MODEL_ID)
model = AutoModelForCausalLM.from_pretrained(
    MODEL_ID,
    torch_dtype="auto",
    device_map="auto",
)

For multimodal inference, use the processor and model class supported by the installed Transformers version. For Scenema audio-dubbing inference, load scenema_adapter_v3.pt through scenema_wrapper.py in the same directory as the base model.

License and attribution

The upstream Gemma 3 weights are subject to the Google Gemma Terms of Use. The NOESIS fine-tuning, adapter, wrapper, and packaging should be used together with the repository LICENSE and the upstream attribution requirements.

This repository is an AMAImedia NOESIS release. It is intended for research, recipe validation, and audio-dubbing pipeline development.