Good for creating POCs, training machine learning models, and experimenting without local resources
What is our primary use case?
Saturn Cloud provides a hosted environment where it's possible to work with various software programming tools (e.g., Jupyter Python notebooks, Julia, R and more).
The system is containerized and accessible both via Jupyter Notebook web pages and SSH—a feature that Google Colab restricts to PRO subscriptions only. I’m currently working on porting a machine learning project to CPU, which provides image Segmentation via Large Language Models. This project handles both image description, image analysis and image object segmentation. Since this project currently relies on CUDA and my local PC has no Nvidia GPUs, I’ve found the computational resources and ease of use provided by Saturn Cloud to be invaluable.
How has it helped my organization?
The project I’m currently working on relies on CUDA, but my local PC does not have any Nvidia GPUs. I’ve found the computational resources and ease of use provided by Saturn Cloud invaluable.
Also, there are many ready-to-use Docker images and a rich documentation portal with useful examples.
The dashboard for creating a new virtual environment contains almost all the features I needed: environment variable definitions, git repositories cloning directly from the new resources page, and an edit field to define a custom script during the boot process. For this reason, Saturn Cloud.io is a very good solution for creating POCs, training machine learning models, and generally experimenting a bit without worrying about local resources.
What is most valuable?
The solution is valuable thanks to:
- plenty of computational resources (both GPU, CPU and disk space)
- a big amount of Docker image recipes
- SSH connection on free subscriptions On Google Colab, the biggest competitor in this field, this feature works only for PRO subscriptions
- possibility to personalize the characteristics of the new virtual environment directly from the dashboard page, adding new environment variables
- installing Python pip or CONDA packages and also system packages
- definition of a custom script that will be executed during the system boot process
What needs improvement?
I would like more documentation about edge and advanced use cases.
The official Docker images are only based on Debian: I would like to find official Docker images also based on other systems like Fedora or SUSE operative systems.
It would be nice to have more hardware category options, like TPU coprocessors or ARM64 CPUs.
I would like a pricing plan associated with a dedicated serverless platform specifically tailored to machine learning inference.
It would be nice to create a custom serverless API system using my own custom machine-learning model.
For how long have I used the solution?
I've used the solution for three months.
What do I think about the stability of the solution?
The service is stable, I've never experienced problems.
What do I think about the scalability of the solution?
Right now, I'm using it more for creating a POC and experimenting; I didn't try to scale up the service.
How are customer service and support?
I've never requested customer support.
How would you rate customer service and support?
Which solution did I use previously and why did I switch?
I also tried Google Colab. I switched since Colab is a little limited for normal use cases based on LLM (at the moment, disk space is only 10GB), and it restricts SSH access on PRO subscriptions.
How was the initial setup?
The initial setup is easy. It only needed to pay attention to the hardware features (e.g., I need CUDA capabilities in my example, so I chose a T4-XLarge instance with a Nvidia T4 GPU) and install Python or system dependencies. Also, pay attention to the Docker image version: an older project will need an older Docker version
What about the implementation team?
There are a good amount of official Docker images (both from StaturnCloud and third-party providers like Nvidia) but also custom Docker image by other users. I'm also satisfied with the code quality and the stability of their deployed virtual systems.
What was our ROI?
Right now, I'm using only the free plan. However, I'm evaluating an upgrade to a bigger instance (T4-4XLarge with 16 vCPU and 64GB of RAM).
What's my experience with pricing, setup cost, and licensing?
The free plan makes it a good alternative to more famous products like Google Colab, and the pricing plan is reasonable.
Which other solutions did I evaluate?
I've used and evaluated Google Colab.
What other advice do I have?
Saturn Cloud provides a good Jupyter system based on Python, Julia, or R.
Which deployment model are you using for this solution?
Public Cloud
If public cloud, private cloud, or hybrid cloud, which cloud provider do you use?
Other
Really awesome cloud compute interface for lowest price
What do you like best about the product?
Close integration with GPU compute makes development seamless. Inclusion of docker best practices is also very neat. Easy to set up and use.
What do you dislike about the product?
I don't like the inability to quickly host much like Google Colab.
What problems is the product solving and how is that benefiting you?
It's benefiting CarbonCopies because it allows us to host GPU powered docker platform very quickly.
Easy to use with good performance and collaborative features
What is our primary use case?
I'm leveraging a cloud-based platform for competitive machine learning. Tight deadlines and resource-intensive models demand powerful hardware. The cloud provides scalable GPUs and RAM, letting me experiment with cutting-edge architectures without limitations.
Its collaborative features are perfect for distributed teams, enabling seamless code sharing and analysis. I stay focused on model development, not infrastructure, thanks to the platform's streamlined setup.
My toolkit – Python, Jupyter Notebooks, and standard data science libraries – works seamlessly in the cloud environment. This ensures a smooth transition from local prototyping to large-scale competition training.
How has it helped my organization?
Saturn Cloud has become an indispensable part of my data science and machine learning toolkit. Their Dask cluster support is fantastic, allowing me to set up distributed computing with just a few clicks.
The ability to scale resources up and down effortlessly ensures my work isn't constrained by hardware and helps me control costs.
Overall, the platform provides everything I need: performance, ease of use, and a focus on the data science workflow itself. This allows me to stay focused on building solutions, not managing infrastructure.
What is most valuable?
Their Dask cluster support is a standout feature. Setting up a cluster takes mere clicks, making it incredibly simple to harness distributed computing power. I love the flexibility to scale environments up or down. This lets me optimize for performance during intensive tasks and save costs during less demanding phases. Shared folders are a game-changer for collaboration.
They provide a centralized space for data, code, and results. Also, I can set up pre-boot scripts, which would allow me to configure the instances with all the libraries I need for the particular task.
What needs improvement?
My main suggestion for improvement centers on pricing. Introducing a tier modelled after AWS spot instances would be a game-changer. Users could bid on unused compute capacity, potentially leading to significant cost savings during off-peak hours or for less time-critical tasks.
Spot instances empower users with tighter budgets or fluctuating workloads to strategically leverage lower-cost resources for development, experimentation, and background tasks. This frees up on-demand instances for truly time-sensitive work.
For how long have I used the solution?
I've used the solution for one year.
Saturn Cloud - The best free resource for development
What do you like best about the product?
The easy use of the platform along with easy upscaling of resources.
What do you dislike about the product?
Free hours are less which interferes with my work. Extra free hours will help people to work better.
What problems is the product solving and how is that benefiting you?
In my local system there is no computational power and it's difficult to get free resource anywhere else.
Extremely fast
What do you like best about the product?
Extremely fast runtime environment, I was impressed, it is way faster than the gpu environment in colab
What do you dislike about the product?
The UI interfaces of the jupyter notebook could be enhanced, but you get used to it quickly
What problems is the product solving and how is that benefiting you?
Running BERTopic topic modelling for a systematic literature review
Reliable compute at reasonable cost
What do you like best about the product?
It's easy to use! I like the simple interface and the quick time to set up.
What do you dislike about the product?
Not much granular control over the VMs from the jupyter environments
What problems is the product solving and how is that benefiting you?
Mainly problems to do with training machine learning models.
I love saturn cloud! i use it for training my models of deep learning. My tesis was made for it!
What do you like best about the product?
I like because is very simple to use it. Its like jupyter notebooks and works excelent.
What do you dislike about the product?
The limits of use, i think the limits for free version are so bigger
What problems is the product solving and how is that benefiting you?
I use Saturn Cloud for develop deep learning models
Great platform for ML intensive tasks
What do you like best about the product?
I have used Saturn Cloud to bulk generate images, train neural networks, and even run LLMs. Setting up a Jupyter Notebook was very easy and only took a few clicks. I got 30 free GPU hours, which regenerates regularly.
What do you dislike about the product?
I didn't find anything that was particulary bad. Their website could use a dark mode theme though.
What problems is the product solving and how is that benefiting you?
Providing GPU compute for ML tasks. As a student who specializes in AI, I like to expirement with new technologies and all things ML. Saturn Cloud provides more than enough free GPU compute for expirmentation and learning.
Rating Saturn as a product.
What do you like best about the product?
I like how it can offer GPUs that work perfectly in PCs and not as expensive as Google Colab.
What do you dislike about the product?
The fact that it takes some time in order to prepare your lab.
What problems is the product solving and how is that benefiting you?
the use of GPUs.
Quite a useful tool
What do you like best about the product?
At first glance, SaturnCloud seems to be just another service providing you with computational tools. However, it has a reasonable tariffication and smooth user experience which makes it one of the best options to consider.
What do you dislike about the product?
There is an issue with the display of remaining funds on a balance (but it is a minor thing).
What problems is the product solving and how is that benefiting you?
Training NNs for a better understanding of current state-of-the-art.
