Instructions to use When-Does-Reasoning-Matter/gemma3-4B_0_split with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use When-Does-Reasoning-Matter/gemma3-4B_0_split with Transformers:

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

pipe = pipeline("image-text-to-text", model="When-Does-Reasoning-Matter/gemma3-4B_0_split")
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("When-Does-Reasoning-Matter/gemma3-4B_0_split")
model = AutoModelForImageTextToText.from_pretrained("When-Does-Reasoning-Matter/gemma3-4B_0_split")
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 When-Does-Reasoning-Matter/gemma3-4B_0_split with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "When-Does-Reasoning-Matter/gemma3-4B_0_split"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "When-Does-Reasoning-Matter/gemma3-4B_0_split",
		"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/When-Does-Reasoning-Matter/gemma3-4B_0_split

SGLang

How to use When-Does-Reasoning-Matter/gemma3-4B_0_split 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 "When-Does-Reasoning-Matter/gemma3-4B_0_split" \
    --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": "When-Does-Reasoning-Matter/gemma3-4B_0_split",
		"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 "When-Does-Reasoning-Matter/gemma3-4B_0_split" \
        --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": "When-Does-Reasoning-Matter/gemma3-4B_0_split",
		"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 When-Does-Reasoning-Matter/gemma3-4B_0_split with Docker Model Runner:
```
docker model run hf.co/When-Does-Reasoning-Matter/gemma3-4B_0_split
```

See axolotl config

axolotl version: 0.12.2

base_model: /lustre/fswork/projects/rech/qwv/udv55np/Gemma/base/gemma-3-4b

datasets:
- path: /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking
  ds_type: json
  type: chat_template
  field_messages: conversations
  data_files:
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0007.jsonl
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0009.jsonl
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0005.jsonl
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0006.jsonl
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0014.jsonl
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0010.jsonl
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0012.jsonl
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0008.jsonl
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0001.jsonl
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0002.jsonl
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0013.jsonl
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0015.jsonl
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0004.jsonl
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0011.jsonl
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0000.jsonl
  - /lustre/fswork/projects/rech/qwv/udv55np/dataset/ift/Nemotron-Super-49B-v1_5/no_thinking/0003.jsonl

dataset_prepared_path: /lustre/fswork/projects/rech/dgo/udv55np/dataset_gemma/Nemotron-Super-49B-v1_5/split_0
tokenizer_config: "/lustre/fswork/projects/rech/qwv/udv55np/Gemma/base/gemma-3-27b"
chat_template: gemma3
eot_tokens:
  - "<end_of_turn>"

shuffle_merged_datasets: true
output_dir: /lustre/fswork/projects/rech/dgo/udv55np/ift/Nemotron-Super-49B-v1_5/gemma-3-4b/0

sequence_len: 16384
sample_packing: true

gradient_accumulation_steps: 1
micro_batch_size: 1
num_epochs: 0.6
auto_resume_from_checkpoints: true

optimizer: adamw_torch_fused
lr_scheduler: warmup_stable_decay
learning_rate: 5e-6
lr_scheduler_kwargs:
  num_decay_steps: 200
  min_lr_ratio: 0.1
warmup_steps: 100

bf16: true
tf32: false

gradient_checkpointing: true
logging_steps: 10
flash_attention: true

evals_per_epoch: 0
saves_per_epoch: 1
save_total_limit: 20
save_only_model: true

use_tensorboard: true
deepspeed: /lustre/fswork/projects/rech/qwv/udv55np/axolotl/zero3.json

lustre/fswork/projects/rech/dgo/udv55np/ift/Nemotron-Super-49B-v1_5/gemma-3-4b/0

This model was trained from scratch on the None dataset.

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: 5e-06
train_batch_size: 1
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 16
total_train_batch_size: 16
total_eval_batch_size: 16
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: warmup_stable_decay
lr_scheduler_warmup_steps: 100
training_steps: 711

Training results

Framework versions

Transformers 4.55.2
Pytorch 2.6.0+cu124
Datasets 4.0.0
Tokenizers 0.21.1

Downloads last month: 5

Safetensors

Model size

769k params

Tensor type

BF16