It seems like I skipped the part where I save the model and change ports in app.py. You just need to train the model after adding nvidia cuda/cudnn docker image. This will train the model on GPU and save model.bin in docker_data folder locally :). After that change the port in app.py to 9999. That's all. I have uploaded all the code here: github.com/abhishekkrthakur/bert-sentiment :)
Thank you Abhishek. Loving these videos. Also just got your book delivered. You have inspired me so much that I have been working on my own notebooks on Kaggle (and a few side-projects) for more than 12 hours a day. I know I'm burning myself out but I'm loving it. I hope that one day, one fine day, I will surpass you on Kaggle.
The video style of building everything up front is great. And the topics that you are picking up are very relevant to industrial practices. One can't learn such things while doing courses. All the best for next video 👍🏻
For anyone running the docker build command on a Macbook with M1 chip this will raise an error. Try running the same command but with the *--platform linux/x86_64* flag instead: docker build -f Dockerfile --no-cache --platform linux/x86_64 -t docker_tutorial .
First, Thank you so much for you video! I've got two questions to ask: 1. From my understanding docker containerizes everything above OS level, No? how come it has image for ubuntu(isin't it linux based OS)? 2. I am trying to reproduce python deep learning model built from ubuntu OS in my windows10 machine than containerize it so I can use it in a flask app running on another server(OS not chosen), is docker the right tool?
Hi, I just want to clarify one thing, @40:01 on `docker run -v /home/..../docker_data/:docker_data/ -ti docker_tutorial /bin/bash` what you are doing is sending data inside docker_data(our local env) to docker_data folder inside docker container right?
I thought since we built our docker container with base layer = ubuntu we are using ubuntu OS instead of Windows(local OS) no? @44:25 says "sorry for windows users" so guessing I'm not really using ubuntu OS but Windows, could you clarify this part? Thanks!
When creating a python app which would be better. using a base ubuntu image and installing python over it or using a python base image. Both should ideally work fine. Just wanted to know which among this was the recommended approach.
as you said, the Nvidia docker is only for ubuntu and not available for windows users. this raises the problem of slow inference time for trained model predictions. is there any way gpu can be used in windows?
Neat video! Quick question though: Why do you use conda and pip side-by-side? Wouldn't conda suffice and be more reliable when it comes to installations?
Is it ok to use 'hub.docker.com/r/pytorch/pytorch' as a Docker container instead of Ubuntu? Can you make a video on docker running via Kubernetes? thanks
great video. Loved that you did not edit out the parts where it did not work. It really helps with learning as I watch it. Also the bit by bit testing. =)
Thanks for the video hey i am doing this multi label classification on csv data for insurance data and i apply lot of different ml techniques but it did't perfome well so would you suggest me somthing thanks
Sir any updates on you're ML course?? And thank you for this tutorials.. even if I know.. I watch them.. cause there is always something I learn.. so thank you so much again.
It seems like I skipped the part where I save the model and change ports in app.py. You just need to train the model after adding nvidia cuda/cudnn docker image. This will train the model on GPU and save model.bin in docker_data folder locally :). After that change the port in app.py to 9999. That's all. I have uploaded all the code here: github.com/abhishekkrthakur/bert-sentiment :)
Realtime debugging while teaching something is the best part of your videos. Love the style and approach!
Every single time i get a notification about your new video , its 100% sure that the day i am going to learn something new and extraordinary.
agreed!!!
Love the way you are teaching, step by step by debugginh errors...
Thank you Abhishek. Loving these videos.
Also just got your book delivered.
You have inspired me so much that I have been working on my own notebooks on Kaggle (and a few side-projects) for more than 12 hours a day. I know I'm burning myself out but I'm loving it. I hope that one day, one fine day, I will surpass you on Kaggle.
Hope you like the book 🙂 And good luck! I hope you surpass everyone :)
@@abhishekkrthakur Haha yes. Will start reading it tomorrow. Thanks again.
How should I make docker image of your knowledge? you are amazing Abhishek 🌟
What a fantastic video!
Thanks for the excellent explanation!
Thank you very much Abhishek for this complete course! very valuable!
Any suggestions for windows user who wants to have GPU access inside Docker?
This video is a Golden nugget. Thanks Abhishek
Could you post docker compose example
Can't wait...
Thanks man for this tutorial, debugging parts what I like the most.
Hi Abhishek, its been absolute pleasure to be learning from your videos. Its been of great help. Thanks again!!!
🙏
The video style of building everything up front is great. And the topics that you are picking up are very relevant to industrial practices. One can't learn such things while doing courses. All the best for next video 👍🏻
For anyone running the docker build command on a Macbook with M1 chip this will raise an error. Try running the same command but with the *--platform linux/x86_64* flag instead:
docker build -f Dockerfile --no-cache --platform linux/x86_64 -t docker_tutorial .
Side question ...... You are not using a mouse, right? How are you navigating ? Thanks :) (I know, with a keyboard, but how exactly)
Everytime I watch this video, I learn something new from it. Thanks sir for explaining the practical concepts.
First, Thank you so much for you video! I've got two questions to ask:
1. From my understanding docker containerizes everything above OS level, No? how come it has image for ubuntu(isin't it linux based OS)?
2. I am trying to reproduce python deep learning model built from ubuntu OS in my windows10 machine than containerize it so I can use it in a flask app running on another server(OS not chosen), is docker the right tool?
Hi, I just want to clarify one thing, @40:01 on `docker run -v /home/..../docker_data/:docker_data/ -ti docker_tutorial /bin/bash` what you are doing is sending data inside docker_data(our local env) to docker_data folder inside docker container right?
How to run docker kaggle-CPU? Any tutorial on it? I am successfull with kaggle-CPU, but it failed with GPU
I thought since we built our docker container with base layer = ubuntu we are using ubuntu OS instead of Windows(local OS) no? @44:25 says "sorry for windows users" so guessing I'm not really using ubuntu OS but Windows, could you clarify this part? Thanks!
docker_data folder inside our local machine is in same level as docker_tutorial folder correct?
Thank you very much. Really appreciate you spreading knowledge at an open platform like TH-cam.
My laptop is 32 bit and can't install pytoch. Can I try some of these things in 32 bit machine?
Thank you for your generosity!
When creating a python app which would be better. using a base ubuntu image and installing python over it or using a python base image. Both should ideally work fine. Just wanted to know which among this was the recommended approach.
I liked it. Maybe you Can do a tutorial on flask as well
You mean something like this th-cam.com/video/hinZO--TEk4/w-d-xo.html ?
@@abhishekkrthakur Oh wow. I will take a look. Keep up the good work, its a good watch!
as you said, the Nvidia docker is only for ubuntu and not available for windows users. this raises the problem of slow inference time for trained model predictions. is there any way gpu can be used in windows?
even i need to use gpu for deploying my model as the inference time of model is high. any solution for it?
Neat video! Quick question though: Why do you use conda and pip side-by-side? Wouldn't conda suffice and be more reliable when it comes to installations?
Is it ok to use 'hub.docker.com/r/pytorch/pytorch' as a Docker container instead of Ubuntu? Can you make a video on docker running via Kubernetes? thanks
Cool tuts, so now I know how to setup the docker for training. How can I monitor my resources for the cluster or Machine used.
some devops tools :) what about prometheus ?
Thanks for a wonderful tutorial on docker and using bert. This is the missing piece for putting ml models on a production path.
Very useful guide! Thanks so much!
great video. Loved that you did not edit out the parts where it did not work. It really helps with learning as I watch it. Also the bit by bit testing. =)
Thanks for the video hey i am doing this multi label classification on csv data for insurance data and i apply lot of different ml techniques but it did't perfome well so would you suggest me somthing thanks
Sir any updates on you're ML course?? And thank you for this tutorials.. even if I know.. I watch them.. cause there is always something I learn.. so thank you so much again.
Why did you use 18.04 but not 20.04?
pip freeze > requirements.txt
We can then edit and remove the ones we don’t need
yeah. for me filtering 5 from 100 is slower than writing 5 :D
Thanks
Great
Docker doesn't do any of the training, it's just a container that other programs can run in that do the training
Namaste
Bhaiya I have been trying to become like you. Please show me right way you got here. Please don't permote your book
Yooooo