Instructions to use Taelor/Gemma3-12B-Outfit-Compose with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Taelor/Gemma3-12B-Outfit-Compose with Transformers:

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

pipe = pipeline("image-text-to-text", model="Taelor/Gemma3-12B-Outfit-Compose")
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("Taelor/Gemma3-12B-Outfit-Compose")
model = AutoModelForMultimodalLM.from_pretrained("Taelor/Gemma3-12B-Outfit-Compose")
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 Taelor/Gemma3-12B-Outfit-Compose with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Taelor/Gemma3-12B-Outfit-Compose"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Taelor/Gemma3-12B-Outfit-Compose",
		"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/Taelor/Gemma3-12B-Outfit-Compose

SGLang

How to use Taelor/Gemma3-12B-Outfit-Compose 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 "Taelor/Gemma3-12B-Outfit-Compose" \
    --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": "Taelor/Gemma3-12B-Outfit-Compose",
		"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 "Taelor/Gemma3-12B-Outfit-Compose" \
        --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": "Taelor/Gemma3-12B-Outfit-Compose",
		"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 Taelor/Gemma3-12B-Outfit-Compose with Docker Model Runner:
```
docker model run hf.co/Taelor/Gemma3-12B-Outfit-Compose
```

See axolotl config

axolotl version: 0.13.0.dev0

base_model: gs://vertex-model-garden-restricted-us/gemma3/gemma-3-12b-it

# gemma3 doesn't seem to play nice with ddp
ddp_find_unused_parameters: true
experimental_skip_move_to_device: true  # prevent OOM by NOT putting model to GPU before sharding

plugins:
- axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
- axolotl.integrations.liger.LigerPlugin
liger_rope: true
liger_rms_norm: true
liger_glu_activation: true
liger_layer_norm: true
liger_fused_linear_cross_entropy: true

chat_template: gemma3
eot_tokens:
- <end_of_turn>

dataset_prepared_path: last_run_prepared
output_dir: /workspace/outputs/out

sequence_len: 8192
sample_packing: true
eval_sample_packing: true

use_kernels: true

micro_batch_size: 2
eval_batch_size: 1
gradient_accumulation_steps: 1
num_epochs: 3
optimizer: adamw_torch_fused
learning_rate: 1e-5

lr_scheduler: cosine

bf16: true
tf32: true
logging_steps: 1
flash_attention: true

gradient_checkpointing: true
activation_offloading: true

val_set_size: 0
eval_strategy: "epoch"
save_strategy: 'no'
include_tokens_per_second: true
save_safetensors: true
use_tensorboard: true

fsdp_version: 1
fsdp_config:
  fsdp_limit_all_gathers: true
  fsdp_sync_module_states: true
  fsdp_offload_params: true
  fsdp_use_orig_params: false
  fsdp_cpu_ram_efficient_loading: true
  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
  fsdp_transformer_layer_cls_to_wrap: Gemma3DecoderLayer
  fsdp_state_dict_type: SHARDED_STATE_DICT
  fsdp_sharding_strategy: SHARD_GRAD_OP
  fsdp_backward_prefetch: BACKWARD_PRE
  final_state_dict_type: FULL_STATE_DICT

tmp/output_dir/gcs/fine-tuning-e28a4df1-aece-4fdc-b956-740e307dc840/postprocess/node-0/checkpoints/final

This model was trained from scratch on the gs://fine-tuning-e28a4df1-aece-4fdc-b956-740e307dc840/copymediadatatask/execution_artifacts/clean_train.jsonl dataset. It achieves the following results on the evaluation set:

Loss: 0.1692
Memory/max Active (gib): 34.85
Memory/max Allocated (gib): 34.85
Memory/device Reserved (gib): 71.78

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 1e-07
train_batch_size: 2
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 8
total_train_batch_size: 16
total_eval_batch_size: 8
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 39
training_steps: 1323

Training results

Training Loss	Epoch	Step	Validation Loss	Active (gib)	Allocated (gib)	Reserved (gib)
No log	0	0	0.9626	34.82	34.82	58.44
0.172	1.0	441	0.1805	34.85	34.85	62.36
0.1608	2.0	882	0.1705	34.85	34.85	71.78
0.1587	3.0	1323	0.1692	34.85	34.85	71.78

Framework versions

Transformers 4.55.4
Pytorch 2.7.1+cu126
Datasets 4.0.0
Tokenizers 0.21.4

Downloads last month: 16

Safetensors

Model size

12B params

Tensor type

BF16