Instructions to use naturecodeproject/mangrove with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use naturecodeproject/mangrove with Transformers:

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

pipe = pipeline("text-generation", model="naturecodeproject/mangrove")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("naturecodeproject/mangrove")
model = AutoModelForCausalLM.from_pretrained("naturecodeproject/mangrove")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use naturecodeproject/mangrove with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/naturecodeproject/mangrove

SGLang

How to use naturecodeproject/mangrove 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 "naturecodeproject/mangrove" \
    --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": "naturecodeproject/mangrove",
		"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 "naturecodeproject/mangrove" \
        --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": "naturecodeproject/mangrove",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use naturecodeproject/mangrove with Docker Model Runner:
```
docker model run hf.co/naturecodeproject/mangrove
```

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

NatureCode Mangrove Model is designed for mangrove conservation, carbon accounting, and environmental research. Access requires approval to ensure responsible use. Please describe your intended use case.

NatureCode Mangrove Model

A multimodal foundation model for mangrove ecosystem analysis, carbon estimation, and conservation planning.

Website: naturecode.ai

Capabilities

Satellite Imagery Analysis: Process 19-band multispectral satellite tiles (Sentinel-2, Landsat, Planet)
Environmental Timeseries: Analyze water quality, temperature, salinity data
Carbon Estimation: Calculate above-ground biomass and blue carbon stocks
Conservation Planning: Generate site assessments and restoration recommendations

Quick Start

from naturecode import NatureCodeMangrove

model = NatureCodeMangrove.from_pretrained("naturecodeproject/mangrove")

response = model.generate(
    text="Analyze this mangrove site",
    image="site_tile.npz",
    coordinates=(4.5, 73.5)
)

Supported Input Types

Type	Formats
Satellite	NPZ (19-band), GeoTIFF, NumPy array
Timeseries	CSV, DataFrame, NetCDF
Text	Natural language queries
Coordinates	(lat, lon) tuples

Model Architecture

Base: Qwen3-1.7B with merged LoRA adapters
Vision Encoder: DINOv2-B/14 with learned projector
Timeseries Encoder: Custom transformer for environmental data
Training: 5-phase curriculum on 50K global mangrove tiles

License

Apache 2.0

Downloads last month: -

Safetensors

Model size

2B params

Tensor type

BF16