Instructions to use MirilAI/Miril-DroneVLM-2B-2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use MirilAI/Miril-DroneVLM-2B-2 with Transformers:

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

pipe = pipeline("image-text-to-text", model="MirilAI/Miril-DroneVLM-2B-2")
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("MirilAI/Miril-DroneVLM-2B-2")
model = AutoModelForMultimodalLM.from_pretrained("MirilAI/Miril-DroneVLM-2B-2", 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 MirilAI/Miril-DroneVLM-2B-2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "MirilAI/Miril-DroneVLM-2B-2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "MirilAI/Miril-DroneVLM-2B-2",
		"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/MirilAI/Miril-DroneVLM-2B-2

SGLang

How to use MirilAI/Miril-DroneVLM-2B-2 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 "MirilAI/Miril-DroneVLM-2B-2" \
    --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": "MirilAI/Miril-DroneVLM-2B-2",
		"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 "MirilAI/Miril-DroneVLM-2B-2" \
        --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": "MirilAI/Miril-DroneVLM-2B-2",
		"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 MirilAI/Miril-DroneVLM-2B-2 with Docker Model Runner:
```
docker model run hf.co/MirilAI/Miril-DroneVLM-2B-2
```

Miril-DroneVLM-2B-2 / assets /full_validation_comparison.svg

MirilAI

Stage Miril-DroneVLM-2B-2 release

774f63b verified 28 days ago

preview code

Raw

History Blame Contribute Delete

14.8 kB

	<svg xmlns="http://www.w3.org/2000/svg" width="1280" height="1477" viewBox="0 0 1280 1477" role="img">
	<rect width="1280" height="1477" fill="#070d12"/>
	<text x="58" y="62" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="30" font-weight="700">Miril-DroneVLM-2B-2 complete validation comparison</text>
	<text x="58" y="96" fill="#c8d0cb" font-family="Arial, sans-serif" font-size="16" font-weight="400">Same held-out visual cases across every response route. Higher is better.</text>
	<rect x="58" y="121" width="18" height="18" fill="#9aa6ad"/>
	<text x="85" y="136" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="15" font-weight="700">Merged BF16</text>
	<rect x="288" y="121" width="18" height="18" fill="#4af082"/>
	<text x="315" y="136" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="15" font-weight="700">CUDA bnb8</text>
	<rect x="518" y="121" width="18" height="18" fill="#5ac8fa"/>
	<text x="545" y="136" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="15" font-weight="700">CUDA bnb4</text>
	<rect x="748" y="121" width="18" height="18" fill="#ffd166"/>
	<text x="775" y="136" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="15" font-weight="700">MLX 8-bit</text>
	<rect x="978" y="121" width="18" height="18" fill="#ff8a65"/>
	<text x="1005" y="136" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="15" font-weight="700">MLX 4-bit</text>
	<text x="58" y="208" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="17" font-weight="700">Valid JSON</text>
	<text x="58" y="231" fill="#9ba7a1" font-family="Arial, sans-serif" font-size="12" font-weight="400">Strict bare JSON objects</text>
	<rect x="360" y="184" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="184" width="759.5" height="16" fill="#9aa6ad"/>
	<text x="1138" y="198" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">99.9%</text>
	<rect x="360" y="209" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="209" width="757.9" height="16" fill="#4af082"/>
	<text x="1138" y="223" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">99.7%</text>
	<rect x="360" y="234" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="234" width="752.7" height="16" fill="#5ac8fa"/>
	<text x="1138" y="248" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">99.0%</text>
	<rect x="360" y="259" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="259" width="760.0" height="16" fill="#ffd166"/>
	<text x="1138" y="273" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">100.0%</text>
	<rect x="360" y="284" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="284" width="759.8" height="16" fill="#ff8a65"/>
	<text x="1138" y="298" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">100.0%</text>
	<text x="58" y="351" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="17" font-weight="700">Schema valid</text>
	<text x="58" y="374" fill="#9ba7a1" font-family="Arial, sans-serif" font-size="12" font-weight="400">Exact keys and allowed values</text>
	<rect x="360" y="327" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="327" width="759.5" height="16" fill="#9aa6ad"/>
	<text x="1138" y="341" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">99.9%</text>
	<rect x="360" y="352" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="352" width="757.5" height="16" fill="#4af082"/>
	<text x="1138" y="366" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">99.7%</text>
	<rect x="360" y="377" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="377" width="750.4" height="16" fill="#5ac8fa"/>
	<text x="1138" y="391" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">98.7%</text>
	<rect x="360" y="402" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="402" width="760.0" height="16" fill="#ffd166"/>
	<text x="1138" y="416" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">100.0%</text>
	<rect x="360" y="427" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="427" width="402.8" height="16" fill="#ff8a65"/>
	<text x="1138" y="441" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">53.0%</text>
	<text x="58" y="494" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="17" font-weight="700">Route accuracy</text>
	<text x="58" y="517" fill="#9ba7a1" font-family="Arial, sans-serif" font-size="12" font-weight="400">Correct type plus intent or action</text>
	<rect x="360" y="470" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="470" width="759.5" height="16" fill="#9aa6ad"/>
	<text x="1138" y="484" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">99.9%</text>
	<rect x="360" y="495" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="495" width="756.9" height="16" fill="#4af082"/>
	<text x="1138" y="509" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">99.6%</text>
	<rect x="360" y="520" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="520" width="749.1" height="16" fill="#5ac8fa"/>
	<text x="1138" y="534" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">98.6%</text>
	<rect x="360" y="545" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="545" width="760.0" height="16" fill="#ffd166"/>
	<text x="1138" y="559" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">100.0%</text>
	<rect x="360" y="570" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="570" width="404.4" height="16" fill="#ff8a65"/>
	<text x="1138" y="584" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">53.2%</text>
	<text x="58" y="637" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="17" font-weight="700">Caption / answer F1</text>
	<text x="58" y="660" fill="#9ba7a1" font-family="Arial, sans-serif" font-size="12" font-weight="400">Reference content overlap</text>
	<rect x="360" y="613" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="613" width="293.5" height="16" fill="#9aa6ad"/>
	<text x="1138" y="627" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">38.6%</text>
	<rect x="360" y="638" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="638" width="174.9" height="16" fill="#4af082"/>
	<text x="1138" y="652" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">23.0%</text>
	<rect x="360" y="663" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="663" width="282.5" height="16" fill="#5ac8fa"/>
	<text x="1138" y="677" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">37.2%</text>
	<rect x="360" y="688" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="688" width="287.1" height="16" fill="#ffd166"/>
	<text x="1138" y="702" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">37.8%</text>
	<rect x="360" y="713" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="713" width="270.2" height="16" fill="#ff8a65"/>
	<text x="1138" y="727" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">35.6%</text>
	<text x="58" y="780" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="17" font-weight="700">Spatial status</text>
	<text x="58" y="803" fill="#9ba7a1" font-family="Arial, sans-serif" font-size="12" font-weight="400">Target, coarse, absent, or ambiguous</text>
	<rect x="360" y="756" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="756" width="654.7" height="16" fill="#9aa6ad"/>
	<text x="1138" y="770" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">86.1%</text>
	<rect x="360" y="781" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="781" width="532.4" height="16" fill="#4af082"/>
	<text x="1138" y="795" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">70.1%</text>
	<rect x="360" y="806" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="806" width="606.6" height="16" fill="#5ac8fa"/>
	<text x="1138" y="820" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">79.8%</text>
	<rect x="360" y="831" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="831" width="648.6" height="16" fill="#ffd166"/>
	<text x="1138" y="845" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">85.3%</text>
	<rect x="360" y="856" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="856" width="629.5" height="16" fill="#ff8a65"/>
	<text x="1138" y="870" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">82.8%</text>
	<text x="58" y="923" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="17" font-weight="700">Precise target retained</text>
	<text x="58" y="946" fill="#9ba7a1" font-family="Arial, sans-serif" font-size="12" font-weight="400">Valid precise target when reference is precise</text>
	<rect x="360" y="899" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="899" width="544.6" height="16" fill="#9aa6ad"/>
	<text x="1138" y="913" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">71.7%</text>
	<rect x="360" y="924" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="924" width="461.6" height="16" fill="#4af082"/>
	<text x="1138" y="938" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">60.7%</text>
	<rect x="360" y="949" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="949" width="559.5" height="16" fill="#5ac8fa"/>
	<text x="1138" y="963" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">73.6%</text>
	<rect x="360" y="974" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="974" width="541.3" height="16" fill="#ffd166"/>
	<text x="1138" y="988" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">71.2%</text>
	<rect x="360" y="999" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="999" width="522.9" height="16" fill="#ff8a65"/>
	<text x="1138" y="1013" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">68.8%</text>
	<text x="58" y="1066" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="17" font-weight="700">Point within 100</text>
	<text x="58" y="1089" fill="#9ba7a1" font-family="Arial, sans-serif" font-size="12" font-weight="400">Distance on the 0-1000 coordinate plane</text>
	<rect x="360" y="1042" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="1042" width="290.9" height="16" fill="#9aa6ad"/>
	<text x="1138" y="1056" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">38.3%</text>
	<rect x="360" y="1067" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="1067" width="69.8" height="16" fill="#4af082"/>
	<text x="1138" y="1081" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">9.2%</text>
	<rect x="360" y="1092" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="1092" width="258.6" height="16" fill="#5ac8fa"/>
	<text x="1138" y="1106" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">34.0%</text>
	<rect x="360" y="1117" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="1117" width="255.5" height="16" fill="#ffd166"/>
	<text x="1138" y="1131" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">33.6%</text>
	<rect x="360" y="1142" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="1142" width="211.0" height="16" fill="#ff8a65"/>
	<text x="1138" y="1156" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">27.8%</text>
	<text x="58" y="1209" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="17" font-weight="700">Coarse direction exact</text>
	<text x="58" y="1232" fill="#9ba7a1" font-family="Arial, sans-serif" font-size="12" font-weight="400">Correct 3x3 directional cell</text>
	<rect x="360" y="1185" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="1185" width="261.4" height="16" fill="#9aa6ad"/>
	<text x="1138" y="1199" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">34.4%</text>
	<rect x="360" y="1210" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="1210" width="86.0" height="16" fill="#4af082"/>
	<text x="1138" y="1224" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">11.3%</text>
	<rect x="360" y="1235" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="1235" width="237.3" height="16" fill="#5ac8fa"/>
	<text x="1138" y="1249" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">31.2%</text>
	<rect x="360" y="1260" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="1260" width="250.9" height="16" fill="#ffd166"/>
	<text x="1138" y="1274" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">33.0%</text>
	<rect x="360" y="1285" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="1285" width="196.2" height="16" fill="#ff8a65"/>
	<text x="1138" y="1299" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">25.8%</text>
	<text x="58" y="1352" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="17" font-weight="700">No-target discipline</text>
	<text x="58" y="1375" fill="#9ba7a1" font-family="Arial, sans-serif" font-size="12" font-weight="400">No coordinates when no target applies</text>
	<rect x="360" y="1328" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="1328" width="732.1" height="16" fill="#9aa6ad"/>
	<text x="1138" y="1342" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">96.3%</text>
	<rect x="360" y="1353" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="1353" width="708.9" height="16" fill="#4af082"/>
	<text x="1138" y="1367" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">93.3%</text>
	<rect x="360" y="1378" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="1378" width="702.2" height="16" fill="#5ac8fa"/>
	<text x="1138" y="1392" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">92.4%</text>
	<rect x="360" y="1403" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="1403" width="726.7" height="16" fill="#ffd166"/>
	<text x="1138" y="1417" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">95.6%</text>
	<rect x="360" y="1428" width="760" height="16" fill="#17232a"/>
	<rect x="360" y="1428" width="726.1" height="16" fill="#ff8a65"/>
	<text x="1138" y="1442" fill="#f4f7f5" font-family="Arial, sans-serif" font-size="14" font-weight="700">95.5%</text>
	</svg>