Hardware for AutoML

Justus2000 · 08-07-2023 07:54 AM

Hello everyone,

I'm currently in the final stages of my research on Neural Architecture Search and have already utilized Vertex AI Vision to train a custom model. In light of this, I'm interested in gaining insight into the specific hardware specifications employed during the model training process. Despite consulting the available documentation, I haven't been able to locate relevant information on this matter.

Since this information is crucial for my thesis, I'd greatly appreciate it if anyone could provide details or pointers about the hardware configurations used for training these models.

Thank you in advance.

Best regards, Paul

lsolatorio

Hi @Justus2000,

I find your research really amazing and I want to help you with a little insight. I appreciate that you took the time to look into Vertex AI's documentation and surely finding specific hardware configuration will be a challenge since they depend on a few variables; the type of model that you want to train, the size of your dataset and your desired machine performance. Ideally a more powerful machine configuration is preferable but of course you will also need to consider the cost.

You may have come across these resource documents, kindly review them as they can provide you with the general idea of what machine specification you will need depending on your project requirements.

You can also reach out to the Vertex AI support team to discuss your project requirements in detail and have them provide you with close-estimates of an optimal hardware configuration.

Hope this helps.