Cloud Training

Ultralytics Platform Cloud Training offers single-click training on cloud GPUs, making model training accessible without complex setup. Train YOLO models with real-time metrics streaming and automatic checkpoint saving.

Train from UI

Start cloud training directly from the Platform:

1. Navigate to your project
2. Click Train Model
3. Configure training parameters
4. Click Start Training

Step 1: Select Dataset

Choose a dataset from your uploads:

Step 2: Configure Model

Select base model and parameters:

Step 3: Select GPU

Choose your compute resources:

Step 4: Start Training

Click Start Training to launch your job. The Platform:

1. Provisions a GPU instance
2. Downloads your dataset
3. Begins training
4. Streams metrics in real-time

Monitor Training

View real-time training progress:

Live Metrics

Checkpoints

Checkpoints are saved automatically:

• Every epoch: Latest weights saved
• Best model: Highest mAP checkpoint preserved
• Final model: Weights at training completion

Stop and Resume

Stop Training

Click Stop Training to pause your job:

• Current checkpoint is saved
• GPU instance is released
• Credits stop being charged

Resume Training

Continue from your last checkpoint:

1. Navigate to the model
2. Click Resume Training
3. Confirm continuation

Remote Training

Train on your own hardware while streaming metrics to the Platform.

pip install "ultralytics>=8.4.0"

Setup API Key

1. Go to Settings > API Keys
2. Create a new key with training scope
3. Set the environment variable:

export ULTRALYTICS_API_KEY="your_api_key"

Train with Streaming

Use the project and name parameters to stream metrics:

yolo train model=yolo26n.pt data=coco.yaml epochs=100 \
  project=username/my-project name=experiment-1

from ultralytics import YOLO

model = YOLO("yolo26n.pt")
model.train(
    data="coco.yaml",
    epochs=100,
    project="username/my-project",
    name="experiment-1",
)

Using Platform Datasets

Train with datasets stored on the Platform:

yolo train model=yolo26n.pt data=ul://username/datasets/my-dataset epochs=100

The ul:// URI format automatically downloads and configures your dataset.

Billing

Training costs are based on GPU usage:

Cost Estimation

Before training starts, the Platform estimates total cost based on:

Estimated Cost = Base Time × Model Multiplier × Dataset Multiplier × GPU Speed Factor × GPU Rate

Factors affecting cost:

Cost Examples

Hold/Settle System

The Platform uses a consumer-protection billing model:

1. Estimate: Cost calculated before training starts
2. Hold: Estimated amount + 20% safety margin reserved from balance
3. Train: Reserved amount shown as "Reserved" in your balance
4. Settle: After completion, charged only for actual GPU time used
5. Refund: Any excess automatically returned to your balance

Payment Methods

View Training Costs

After training, view detailed costs in the Billing tab:

• Per-epoch cost breakdown
• Total GPU time
• Download cost report

Training Tips

Choose the Right Model Size

Optimize Training Time

1. Start small: Test with fewer epochs first
2. Use appropriate GPU: Match GPU to model/batch size
3. Validate dataset: Ensure quality before training
4. Monitor early: Stop if metrics plateau

Troubleshooting

FAQ

How long does training take?

Training time depends on:

• Dataset size
• Model size
• Number of epochs
• GPU selected

Typical times (1000 images, 100 epochs):

Can I train overnight?

Yes, training continues until completion. You'll receive a notification when training finishes. Make sure your account has sufficient balance for epoch-based training.

What happens if I run out of credits?

Training pauses at the end of the current epoch. Your checkpoint is saved, and you can resume after adding credits.

Can I use custom training arguments?

Yes, advanced users can specify additional arguments in the training configuration.

Training Parameters Reference

Core Parameters

Learning Rate Parameters

Augmentation Parameters

Optimizer Selection

📅 Created 20 days ago ✏️ Updated 13 days ago

Comments