00:00 Set up AWS instance
00:20 g4dn.2xlarge instance
01:01 Login to AWS instance
01:34 Instructions page: [ Ссылка ]. Install Kaldi git clone [ Ссылка ] kaldi --origin upstream
02:44 make -j 8
03:07 go to src/ and follow INSTALL instructions there.
03:47 folder structure


Correction on 0.37 seconds. I used my Kaldi installation to work with LibriSpeech. I needed more than 400GB space. There is LibriSpeech Mini project which probably needs
smaller space. Get the storage space that you need for your project.

written tutorial
[ Ссылка ]

Transcript:
Today we’re going to set up an AWS instance and install Kaldi on it. First, open your AWS account and find the EC2 instances page. Click launch instance, scroll down to the “**Deep Learning AMI (Ubuntu 18.04)”** and select it. For Step 2, select the g4dn.2xlarge instance as it has 8 vCPUs and 1 GPU, and its 75 cents per hour, so make sure to pause the instance when you’re not using it. For Step 3, select a specific subnet. This is optional since it defaults to a random region. For Step 4, disable the “Delete on Termination” setting and change the Root’s size to 600 GB. From personal experience, 400 isn’t going to be enough. Then, you can skip to step 7, review the instance settings, and launch the instance. I’m creating a key pair but you can use an existing one.

Now its time to open your terminal and login into your AWS instance. The address you’ll be ssh-ing into is ubuntu “at” your instance’s public IP address. If you’re using a new key .pem file, make sure to change its privacy settings by using the command “chmod 400” with your file name. Once you successfully ssh into your instance, it is time to install Kaldi.

The initial downloading steps can be found on the official Kaldi webpage. Just copy and paste the first two commands on this page into your terminal and you should be set. Next, open and read the INSTALL file in the Kaldi directory, and follow Option 1. This means you should change to the tools directory and follow the directions in the INSTALL file there. Check for dependencies using one of the provided lines and install anything that isn’t downloaded yet. There’s a bug for one dependency where even if you download it, the check dependencies command will still say its uninstalled. You can just ignore this warning if you have already installed it. Now we're going to run make with the last option in the installation instruction, since it uses multiple CPUs. When that’s done running, I saw the warning that IRSTLM is not installed, so I’m running the provided line to have it installed.

If you recall the original installation steps we found, you’ll remember that when we’re done in the tools folder we should head to the source folder. Here we can open the installation file and follow the three simple commands.

And that was it! You’re all done installing Kaldi. Thanks for watching, and hopefully this was helpful.
#Kaldı #speechprocessing #stt #asr

свернуть