After the tireless efforts of our engineering team, we are pleased to announce GreenNode’s groundbreaking updates on our journey to provide an end-to-end journey for your LLM training experience.
The latest updates include several important add-ons that will help you deliver smoother, faster, and more cost-effective workflows, empowering you to achieve more with less effort.
Key Updates:
- CPU Instances: Optimize your costs with our new CPU instances, ideal for less computationally intensive tasks.
- SSH Access: Gain granular control over your notebook instances with secure SSH connectivity.
- Network Enhancements: Experience faster data transfer speeds and improved network performance.
Dive into and discover how these updates can revolutionize your LLM training today!
CPU Instances are NOW available
We're excited to announce the availability of CPU instances on our platform! On-demand customers can now easily create CPU instances through our self-service portal, providing a cost-effective solution for less resource-intensive LLM training workloads.
Why Choose CPU Instances:
- Optimize Costs: Significantly reduce your LLM training expenses by utilizing CPU instances for tasks that don't require the full power of GPU-accelerated instances.
- Ideal for Specific Workloads: CPU instances are well-suited for a range of LLM training tasks, including:
- Data preprocessing and cleaning
- Model fine-tuning on smaller datasets
- Experimentation and prototyping
- Inference tasks
- Easy to Create and Manage: Simply log in to our self-service portal and create a CPU instance tailored to your specific needs.
Why does this update matter:
Even when you are working on an AI/ML project, it does not mean that you need to fully armed a massive computational power like A100, H100 GPU instance. To optimize the cost of the project while still ensuring training performance, we can offer CPUs as a reasonable and economical alternative resource for suitable workloads.
This feature allows you to create a CPU instance with a base image or Jupyter Notebook image. This time, we added two CPU instances for choice:
- Cpu-general-2x4-1h (2 vCPU, 4GB memory)
- Cpu-general-4x8-1h (4 vCPU, 8B memory)
Once configured, you can use this instance for the following main purposes:
- Mount network volume for data access
- Collect and process data from sources such as downloading from public sources or uploading from your local machine.
- Run ML/AI tasks no requiring heavy computational power such as traditional algorithms (e.g., linear/logistic regression, decision trees, etc.)
Pricing:
This instance will cost you 0.034 USD per hour. Trust us! This is one of the most reasonable options in the market at this moment.
While CPU instances offer a cost-effective solution for certain LLM training workloads, they may not be suitable for deep learning tasks that involve large-scale matrix operations. For example, training neural networks using frameworks like TensorFlow and PyTorch often requires the computational power of GPU instances such as A100 and H100.
To ensure optimal performance and efficiency for your LLM training, we recommend carefully evaluating your specific needs and considering the use of GPU instances when working with deep learning models.
Need Assistance?
Our team is available to help you determine the best instance type for your LLM training requirements. Contact us today to discuss your project and receive personalized recommendations.
Network Volumes
Need your data everywhere, all the time? Our Network Volumes can help with this. Whether you’re:
- Cranking out code in Notebook Instances,
- Training models (aka teaching your AI to be smarter than you) in Training Jobs, or
- Getting those real-time predictions in Online Prediction,
You’ll have quick, seamless access to your data, wherever you need it.
Note: For more detailed information, see more: Release Notes.
Why does this update matter:
Network Volumes on the Greennode AI Platform provide a high-performance and scalable storage solution specifically designed to meet the storage and data management needs of AI resources. One of the standout features of Network Volumes is their ability to be accessed flexibly and easily from various components within the platform
This update isn’t just about storage; it’s about speed, efficiency, and scale. You can focus on building cool stuff while we handle the heavy lifting of making sure your data flows seamlessly across the platform. No more waiting around for file transfers—get what you need, when you need it, and keep the momentum going!
We also created a detailed guide for your needs. See more: How to create a network volume.
SSH Access
Want to feel like a hacker (but the good kind)? We’ve added SSH access for your notebook instances. Now, you can connect securely and take full control of your workspace like a boss. See more: How to connect SSH access to notebook instances.
Why does this update matter:
This isn’t just a fancy feature—it’s about giving you more flexibility and control. Need to troubleshoot or quickly run commands directly in your environment? SSH access lets you do that efficiently and securely. It’s all about making sure you have the tools to work smarter, not harder.
Read more about how to connect to the notebook instance using Secure Shell (SSH), providing a more secure and flexible connection method: How to connect to the notebook instance using Secure Shell (SSH).