Set up a container on GCP and build an analysis environment. I don't know yet, but I will summarize how I made it for myself.
Please read the following article about GCP because it is very well organized. I read up to the third part of this series to get a general idea of what it looks like. [[Introduction to GCP, Part 1] Must-read for engineers! What is Google Cloud Platform (GCP) that I can't ask you anymore? ] 1
Account creation is described in Part 2. [[Introduction to GCP, Part 2] First, start here! Preparation for Google Cloud Platform (GCP)! ] 2
Description of GCE [[Introduction to GCP, Part 3] Not difficult! How to launch an instance on Google Compute Engine (GCE)! ] 3
[Building a calculation environment for Kaggle with GCP and Docker] 4 [Building a GPU environment for Kaggle with GCP + Docker] 5 I mainly referred to 2 articles, but I couldn't find the Deep Learning VM. I created a normal instance.
The following items were tampered with on the GCE instance creation screen --Machine type 8vCPU
Note: If you use GPU, you need to edit the assignment. [Questions about editing assignments] 6 -> [Launch a GCP Deep Learning VM instance] 7 We will respond within 2 business days after you apply! I got an email saying that, but it was within the day.
You can use Pytorch NVIDIA GPU Notebook with MaketPlace.
This was also referred to the article in the previous section. In addition, I also referred to [Introduction article: Jupyter analysis environment created by GCE (Google Compute Engine)] 8 Jupyter environment construction (Docker).
Check access information [here] 9 For instance firewall settings [here] 10
This puts it in the rest of 100 language processing knocks.
--Being able to write a Dockerfile that has been downloaded --Deep Learning VM could not be found --I was going to check the GPU with torch, but I forgot (at the moment)
Recommended Posts