Bring Your Training Job
We support training your custom AI models on EGS Serverless. You can bring your AI model to the EGS Serverless SaaS platform for both fine-tuning and training.
This feature supports:
- Distributed Multi-GPU Training: Run training workloads across multiple GPU nodes with NVLink and InfiniBand connectivity.
- Unified Data Access: Access training data, checkpoints, and artifacts through Persistent Volumes or S3 storage.
- Framework Compatibility: Supports PyTorch, TensorFlow, Hugging Face, Horovod, Ray, and more.
- MLOps Integration: Works with Kubeflow, Flyte, MLflow, and modern MLOps stacks.
To use this feature:
-
Go to Bring Your Training Job on the left sidebar.
-
Please contact Avesha Support at support@avesha.io for more information.