The best thing about Deep Learning VM Image is that you don’t have to waste hours dealing with dependency hell or driver compatibility issues. Having popular ML libraries and NVIDIA drivers pre-installed out of the box makes it super easy to jump straight into model training on Google Cloud. Review collected by and hosted on G2.com.

The cost can add up pretty quickly if you forget to stop GPU instances after training. Also, initial startup times can sometimes feel a bit sluggish, and troubleshooting version mismatches when custom updates are needed can occasionally be frustrating. Review collected by and hosted on G2.com.

What I like most about the Deep Learning VM Image is how much setup time it saves. Having the necessary ML frameworks, GPU drivers, and development tools already configured makes it much easier to jump straight into deep learning projects. The environment is straightforward to use, and performance is solid when running GPU-based workloads. I also appreciate the flexibility to work with different AI and machine learning frameworks without spending hours setting everything up manually. Overall, it has made my workflow more efficient, especially when I’m testing models and experimenting with different workloads. Review collected by and hosted on G2.com.

One area that could be improved is the initial setup and configuration experience. Some users may still need to spend time understanding the available frameworks, GPU settings, and integrations before getting everything running smoothly. Documentation and onboarding could be more straightforward, especially for new users. Pricing can also become a consideration for longer-running GPU workloads. Better setup guidance, clearer cost information, and simpler configuration options would make the overall experience easier and more cost-effective. Review collected by and hosted on G2.com.

What I like most about the Deep Learning VM Image is that it provides a ready-to-use machine learning environment without requiring me to spend a lot of time manually setting up drivers, CUDA, Python libraries, and frameworks. Its integration with Google Cloud also makes it easy to connect to other cloud services, and the preconfigured GPU support helps improve model training performance. I also find JupyterLab useful for testing ideas and experimenting with models. Overall, it has sped up my workflow by cutting down on setup and troubleshooting time, and the documentation and Google Cloud support are helpful whenever I need to configure something or track down an issue. Review collected by and hosted on G2.com.

The main issue is that the initial setup can still feel confusing, especially when choosing the right VM, GPU configuration, and software versions. Compatibility between CUDA, drivers, and different ML frameworks can sometimes require extra troubleshooting. The overall cost can also increase quickly when using GPU instances for longer training jobs, so it would be helpful to have clearer cost guidance and easier options for managing resources. Better onboarding for beginners and more straightforward version management would make the experience easier. Review collected by and hosted on G2.com.

Setting up a deep learning environment from scratch is usually a miserable cycle of matching Linux kernels, NVIDIA drivers, CUDA versions, and PyTorch builds. Google's DL VMs completely wipe out that headache. You spin up the instance, open JupyterLab, and everything actually works together right out of the box. Review collected by and hosted on G2.com.

The base images are pretty heavy since they pack so much in. More importantly, because spinning up a powerhouse GPU instance is a one-click affair, it’s way too easy to forget to shut it down and get a nasty surprise on your monthly bill. Review collected by and hosted on G2.com.

What I like best about Deep Learning VM Image is that it provides a pre-configured environment for machine learning and deep learning development. It comes with popular frameworks, libraries, drivers, and tools already installed, which significantly reduces setup time. The ability to quickly create GPU-enabled environments, experiment with different models, and scale workloads on Google Cloud makes it highly valuable for AI development and research. Review collected by and hosted on G2.com.

What I dislike about Deep Learning VM Image is that the pre-configured environments can sometimes become difficult to customize when specific versions of frameworks, libraries, or dependencies are required. Managing updates and compatibility between different ML packages can require additional effort. The cost of running GPU-enabled VM instances can also become high for continuous workloads, especially for large-scale experiments. Review collected by and hosted on G2.com.

What i like best about Google Cloud's Deep Learning VM Images is that it lets me launch pre-configured Compute Engine instances with popular ML frameworks like PyTorch and TensorFlow pre-installed. It saves me time by removing driver compatibility issues and offering out-of-the-box support for GPUs and TPUs. The pricing is fair for the tasks to completes. Its reliability makes it user friendly is very intuitive and useful for day to day tasks. Review collected by and hosted on G2.com.

The thing I dislike the most is the steep learning curve for deep learning VM image. Onboarding takes long for this reason. The integrations are a bit lacking would be nice if it was more compatible with other tools Review collected by and hosted on G2.com.

Google Deep Learning VM Image made it much faster to get a fully configured training environment running for model development, with common frameworks and drivers pre-installed instead of setting everything up manually. Having GPU support ready out of the box removed a lot of the setup friction that usually eats up time early in a training project. The interface for launching and managing instances is straightforward, and integration with the rest of our Google Cloud stack was smooth, since deployment fit naturally into infrastructure we were already using. Performance for training workloads has been solid, with GPU acceleration handling compute-intensive tasks efficiently. Pricing has offered reasonable ROI given the time saved on environment setup, and onboarding required minimal effort since the pre-configured images worked reliably from the start. Review collected by and hosted on G2.com.

Image sizes and initial boot times are heavier compared to a minimal custom VM, which added some startup delay when spinning up new instances frequently. Costs for GPU-enabled instances can add up quickly during extended training runs, requiring careful monitoring to avoid leaving instances running unnecessarily. Customizing beyond the pre-configured framework versions sometimes required extra steps to install specific package versions not included by default. Review collected by and hosted on G2.com.

As an AI engineer, I like Deep Learning VM Image because it gives me a ready-to-use environment for machine learning and deep learning projects. It comes with useful frameworks, library, and GPU support already configured, so i don't have to spend much time setting everything up. This makes experimenting with models and running training tasks much faster and easier. Review collected by and hosted on G2.com.

I found Deep learning VM image a bit heavy and resource intensive especially when using GPU instances. The initial setup is easier because most tools are preinstalled, but customizing the environment can sometimes be confusing. It can also become costly when running large workloads for a long time, so I need to manage the resources carefully. Review collected by and hosted on G2.com.

I love using Deep Learning VM Image because it offers a ready-to-use environment, eliminating the need to spend a lot of time installing and configuring frameworks, drivers, and dependencies. I can start working much faster with VM provisioning since it already has major machine learning tools set up. GPU support and integration with Google Cloud allow me to scale computing resources without having to maintain the hardware. The preconfigured dependencies are a major advantage, as it means I can focus more on model development and experimentation rather than infrastructure setup. The predefined frameworks make the setup easy, and its compatibility with Google Cloud Platform makes it feel familiar. The team also finds it helpful for GPU-backed compute tasks. Review collected by and hosted on G2.com.

I find that the preconfigured environment can't perfectly match every project requirement. For highly customizable projects, I need additional dependency management, and with GPU workloads, I have to manage costs carefully. Review collected by and hosted on G2.com.

Honestly, it completely eliminates dependency hell. Anyone who’s spent a weekend fighting broken Nvidia drivers, mismatched CUDA toolkits, and cuDNN paths just to get PyTorch to recognize a GPU knows the pain. With the Deep Learning VM, everything comes pre-baked and already verified to work right out of the box. You spin up an instance, SSH in, and JupyterLab plus GPU acceleration are immediately ready to go. It turns what used to be a half-day DevOps nightmare into a two-minute launch. Review collected by and hosted on G2.com.

The disk bloat is real. Since the image ships packed with practically every framework, library, and tool under the sun, it chews through a ton of storage right out of the gate, which drives up boot disk costs. On top of that, upgrading individual packages can easily break the preconfigured CUDA bindings, so you end up hesitant to run even a simple pip install -U. If all you want is a lightweight, customized environment, removing the default bloat often takes more work than just building your own custom Docker container. Review collected by and hosted on G2.com.