Instructions to use mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged with Transformers:

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

pipe = pipeline("image-text-to-text", model="mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged")
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("mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged")
model = AutoModelForMultimodalLM.from_pretrained("mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged")
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 mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged",
		"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/mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged

SGLang

How to use mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged 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 "mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged" \
    --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": "mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged",
		"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 "mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged" \
        --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": "mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged",
		"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"
						}
					}
				]
			}
		]
	}'

Unsloth Studio

How to use mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged",
    max_seq_length=2048,
)

Docker Model Runner
How to use mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged with Docker Model Runner:
```
docker model run hf.co/mcrovero/gemma-3-4b-it-unsloth-bnb-4bit-custom-merged
```

mcrovero commited on Apr 22, 2025

Commit

f9a2d67

verified ·

1 Parent(s): cee44a9

Update config.json

Browse files

Files changed (1) hide show

config.json +136 -30

config.json CHANGED Viewed

@@ -1,37 +1,143 @@
 {
   "architectures": [
-    "Gemma3ForCausalLM"
   ],
-  "attention_bias": false,
-  "attention_dropout": 0.0,
-  "attn_logit_softcapping": null,
   "bos_token_id": 2,
-  "cache_implementation": "hybrid",
-  "eos_token_id": 1,
-  "final_logit_softcapping": null,
-  "head_dim": 256,
-  "hidden_activation": "gelu_pytorch_tanh",
-  "hidden_size": 2560,
   "initializer_range": 0.02,
-  "intermediate_size": 10240,
-  "max_position_embeddings": 131072,
-  "model_type": "gemma3_text",
-  "num_attention_heads": 8,
-  "num_hidden_layers": 34,
-  "num_key_value_heads": 4,
   "pad_token_id": 0,
-  "query_pre_attn_scalar": 256,
-  "rms_norm_eps": 1e-06,
-  "rope_local_base_freq": 10000.0,
-  "rope_scaling": {
-    "factor": 8.0,
-    "rope_type": "linear"
   },
-  "rope_theta": 1000000.0,
-  "sliding_window": 1024,
-  "sliding_window_pattern": 6,
-  "torch_dtype": "float16",
-  "transformers_version": "4.50.0.dev0",
-  "use_cache": true,
-  "vocab_size": 262208
-}

 {
   "architectures": [
+    "Gemma3ForConditionalGeneration"
   ],
+  "boi_token_index": 255999,
   "bos_token_id": 2,
+  "eoi_token_index": 256000,
+  "eos_token_id": 106,
+  "image_token_index": 262144,
   "initializer_range": 0.02,
+  "mm_tokens_per_image": 256,
+  "model_type": "gemma3",
   "pad_token_id": 0,
+  "quantization_config": {
+    "_load_in_4bit": true,
+    "_load_in_8bit": false,
+    "bnb_4bit_compute_dtype": "bfloat16",
+    "bnb_4bit_quant_storage": "uint8",
+    "bnb_4bit_quant_type": "nf4",
+    "bnb_4bit_use_double_quant": true,
+    "llm_int8_enable_fp32_cpu_offload": false,
+    "llm_int8_has_fp16_weight": false,
+    "llm_int8_skip_modules": [
+      "lm_head",
+      "multi_modal_projector",
+      "merger",
+      "modality_projection",
+      "language_model.model.layers.5.mlp",
+      "language_model.model.layers.6.mlp",
+      "vision_tower.vision_model.encoder.layers.23.self_attn",
+      "language_model.model.layers.4.mlp",
+      "language_model.model.layers.1.mlp",
+      "vision_tower.vision_model.encoder.layers.24.self_attn",
+      "vision_tower.vision_model.encoder.layers.22.self_attn",
+      "vision_tower.vision_model.encoder.layers.21.self_attn",
+      "vision_tower.vision_model.encoder.layers.19.self_attn",
+      "vision_tower.vision_model.encoder.layers.25.self_attn",
+      "vision_tower.vision_model.encoder.layers.26.self_attn",
+      "language_model.model.layers.3.mlp",
+      "vision_tower.vision_model.encoder.layers.25.mlp",
+      "vision_tower.vision_model.encoder.layers.19.mlp",
+      "vision_tower.vision_model.encoder.layers.20.self_attn",
+      "vision_tower.vision_model.encoder.layers.17.mlp",
+      "vision_tower.vision_model.encoder.layers.22.mlp",
+      "vision_tower.vision_model.encoder.layers.18.self_attn",
+      "language_model.model.layers.0.mlp",
+      "vision_tower.vision_model.encoder.layers.20.mlp",
+      "vision_tower.vision_model.encoder.layers.21.mlp",
+      "language_model.model.layers.2.self_attn",
+      "vision_tower.vision_model.encoder.layers.15.mlp",
+      "vision_tower.vision_model.encoder.layers.17.self_attn",
+      "vision_tower.vision_model.encoder.layers.24.mlp",
+      "vision_tower.vision_model.encoder.layers.16.mlp",
+      "vision_tower.vision_model.encoder.layers.16.self_attn",
+      "vision_tower.vision_model.encoder.layers.18.mlp",
+      "vision_tower.vision_model.encoder.layers.13.mlp",
+      "vision_tower.vision_model.encoder.layers.23.mlp",
+      "vision_tower.vision_model.encoder.layers.10.mlp",
+      "vision_tower.vision_model.encoder.layers.14.mlp",
+      "vision_tower.vision_model.encoder.layers.15.self_attn",
+      "vision_tower.vision_model.encoder.layers.11.self_attn",
+      "vision_tower.vision_model.encoder.layers.14.self_attn",
+      "vision_tower.vision_model.encoder.layers.7.mlp",
+      "vision_tower.vision_model.encoder.layers.5.mlp",
+      "vision_tower.vision_model.encoder.layers.9.mlp",
+      "vision_tower.vision_model.encoder.layers.11.mlp",
+      "vision_tower.vision_model.encoder.layers.12.mlp",
+      "vision_tower.vision_model.encoder.layers.10.self_attn",
+      "vision_tower.vision_model.encoder.layers.13.self_attn",
+      "vision_tower.vision_model.encoder.layers.8.mlp",
+      "vision_tower.vision_model.encoder.layers.12.self_attn",
+      "vision_tower.vision_model.encoder.layers.8.self_attn",
+      "vision_tower.vision_model.encoder.layers.4.mlp",
+      "vision_tower.vision_model.encoder.layers.4.self_attn",
+      "vision_tower.vision_model.encoder.layers.9.self_attn",
+      "vision_tower.vision_model.encoder.layers.2.mlp",
+      "vision_tower.vision_model.encoder.layers.7.self_attn",
+      "vision_tower.vision_model.encoder.layers.3.mlp",
+      "vision_tower.vision_model.encoder.layers.6.mlp",
+      "vision_tower.vision_model.encoder.layers.1.mlp",
+      "vision_tower.vision_model.encoder.layers.6.self_attn",
+      "vision_tower.vision_model.encoder.layers.5.self_attn",
+      "vision_tower.vision_model.encoder.layers.1.self_attn",
+      "vision_tower.vision_model.encoder.layers.3.self_attn",
+      "vision_tower.vision_model.encoder.layers.0.mlp",
+      "vision_tower.vision_model.encoder.layers.0.self_attn",
+      "vision_tower.vision_model.encoder.layers.2.self_attn",
+      "vision_tower.vision_model.encoder.layers.26.mlp"
+    ],
+    "llm_int8_threshold": 6.0,
+    "load_in_4bit": true,
+    "load_in_8bit": false,
+    "quant_method": "bitsandbytes"
   },
+  "text_config": {
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "attn_logit_softcapping": null,
+    "cache_implementation": "hybrid",
+    "final_logit_softcapping": null,
+    "head_dim": 256,
+    "hidden_activation": "gelu_pytorch_tanh",
+    "hidden_size": 2560,
+    "initializer_range": 0.02,
+    "intermediate_size": 10240,
+    "max_position_embeddings": 131072,
+    "model_type": "gemma3_text",
+    "num_attention_heads": 8,
+    "num_hidden_layers": 34,
+    "num_key_value_heads": 4,
+    "query_pre_attn_scalar": 256,
+    "rms_norm_eps": 1e-06,
+    "rope_local_base_freq": 10000.0,
+    "rope_scaling": {
+      "factor": 8.0,
+      "rope_type": "linear"
+    },
+    "rope_theta": 1000000.0,
+    "sliding_window": 1024,
+    "sliding_window_pattern": 6,
+    "torch_dtype": "bfloat16",
+    "use_cache": true,
+    "vocab_size": 262208
+  },
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.52.0.dev0",
+  "unsloth_fixed": true,
+  "vision_config": {
+    "attention_dropout": 0.0,
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 1152,
+    "image_size": 896,
+    "intermediate_size": 4304,
+    "layer_norm_eps": 1e-06,
+    "model_type": "siglip_vision_model",
+    "num_attention_heads": 16,
+    "num_channels": 3,
+    "num_hidden_layers": 27,
+    "patch_size": 14,
+    "torch_dtype": "bfloat16",
+    "vision_use_head": false
+  }
+}