Skip to content

Latest commit

 

History

History
60 lines (34 loc) · 1.4 KB

Colab_Pro_Tutorial.md

File metadata and controls

60 lines (34 loc) · 1.4 KB
Raw

Tutorial with Colab-Pro

  • In this article, our goal is to reproduce the experiment in Colab-Pro.

  • The prerequisite is that you already have a Colab-Pro subscription.

Colab-setup

  • Access here.

  • Create New Notebook.

  • Click [Runtime] -> [Change Runtime Type] and then change the settings as follows.

    colab

  • Then click [SAVE].

  • Kill all other sessions from [RUNTIME] -> [Manage Sessions] -> [TERMINATE OTHER SESSIONS].

ssh-setup

  • Access here and then copy Token.

  • Run this command in your colab-pro session.

    !pip install colab_ssh --upgrade
from colab_ssh import launch_ssh
ngrokToken = "your_authtoken" # Copied token
launch_ssh(ngrokToken, password="your_password")

  • If it succeeds, the ssh destination will be specified.

    ssh

  • In this example, run ssh root@8.tcp.ngrok.io -p 14082 on your local machine.

Experimental Setup

Environment

$ git clone https://github.com/izuna385/Entity-Linking-Tutorial
$ cd https://github.com/izuna385/Entity-Linking-Tutorial
$ pip install -r requirements.txt

Preprocessed file

See here.

Experiment

python3 main.py -num_epochs 10 -batch_size_for_train 48

gpu