Using Glows.ai Cloud GPU for running LLMs
DeepSeek-R1 has been the release of the year so far. It has made everyone, not just the tech guys, interested in Generative AI.
But not everyone has enough GPU to run such a huge model or LLMs in general
You don’t need to buy a GPU just to run DeepSeek-R1. Instead, you can opt for Cloud GPUs for the same which are way more affordable for experimentation and even full-time services.
https://medium.com/media/0f6ded8bba8b4637e4911df6c8ab98cc/href
What is Cloud GPU?
A Cloud GPU is a Graphics Processing Unit (GPU) available on a cloud platform for remote computing. Instead of buying expensive hardware, you can rent high-performance GPUs from providers like AWS, Google Cloud, or Azure. These are mainly used for AI/ML training, deep learning, video rendering, and gaming.
In this tutorial, we will be showing how to run DeepSeek-R1 on a Cloud GPU instance.
We will be using Glows.ai for this tutorial
Why I’m using Glows.ai?
Easy to set cloud GPU instance.
Availability of the latest GPU instances like NVIDIA GeForce RTX 4090, H100, etc.
Quite cheap compared to other vendors.
Access through SSH, JupyterLab, and the WebUI for maximum flexibility.
You can follow glows.ai for other tutorials
Steps to run DeepSeek-R1 on Glows.ai Cloud GPU
- Sign up on Glows.ai and log into your account.
- Click “Create New” in the top-right corner.
- Choose your preferred GPU specs to view available official images.
- Locate and select the official DeepSeek-R1 image
Glows.ai offers two DeepSeek-R1 configurations:
32B -Q4— Optimized for efficient deployment and development on consumer-grade GPUs. Will require 1 NVIDIA GeForce RTX 4090 GPUs.
70B -Q4— Built for large-scale AI-native applications. Requires at least 2 NVIDIA GeForce RTX 4090 GPUs.
- Choose the number of instances depending on your model
I’m going with the 32B version for this tutorial
Once you’ve ordered the instance, you should see something like the below screen
Let’s try running using HTTP Port 8080. For that
- Copy the Link and paste it into the browser
- Log in with the email & enter password of your choice
DeepSeek-R1 32b is now running on Cloud GPU
Once done, don’t forget to stop your instance !
Not just DeepSeek-R1, the platform also offers many other LLMs
Other LLMs supported by Glows.ai
LLaMA-3.1 8B
An ideal choice for entry-level models. After quantization, it runs on a 12GB GPU, balancing performance and resource demands efficiently.
LLaMA-3.1 70B
A top-tier open-source large model. After quantization, it is suitable for GPUs with 40GB+ memory. Excels in multilingual processing, complex reasoning, and long-text understanding.
DeepSeek-R1 32B
A high-quality mid-sized model. After quantization, it runs smoothly on a 24GB GPU, delivering stable and reliable performance.
DeepSeek-R1 70B
A large-scale language model with optimized quantization for moderate resource requirements, ideal for handling complex tasks.
Gemma2 9B
A highly efficient model developed by Google. Runs on a 12GB GPU, making it suitable for lightweight applications.
TAIDE LX 8B
A streamlined model optimized for Chinese language processing. After quantization, it has low resource consumption while delivering excellent performance in Chinese tasks.
Stable Diffusion XL
An advanced image generation model that supports high-quality output. GPU memory requirements increase with resolution, making it ideal for professional creative work.
FLUX (ComfyUI)
A flexible image processing model with resource requirements that adjust based on workload. Suitable for various creative scenarios.
ComfyUI (General)
A modular and versatile platform where resource requirements depend on the loaded model. Supports customized workflow development.
Not just that, Beyond the basic Open Web UI application, the biggest advantage Glows.ai offers to developers is the ability to list their models, which helps AI developers reduce the time needed to develop AI applications.
Hope you try running DeepSeek-R1 on cloud GPU using Glows.ai
Get 10 free credits for Glows.ai using my referral code : Glows-69kjm32p
How to Run DeepSeek-R1 on Cloud GPU (Glows.ai)? was originally published in Data Science in your pocket on Medium, where people are continuing the conversation by highlighting and responding to this story.
