Thanks so much. I have just started my PhD in a deep learning project with medical imaging and your videos are very much helpful. Keep making videos. :)
Thank you so much for your channel. It is very helpful. Wanted to know if you have a solution for a student account that is not authorized to google pay and buy Colab pro. I have a like 40 GB data to train. Your answer is appreciated.
Another great presentation Sreeni. I think an additional factor to consider is the time spent maintaining your local workstation. My GPU server is having hardware issues and I have sunk many hours this week fixing it :(. I wonder if it was worth it or just going cloud was a better option.
Sorry, no pytorch. I am not knowledgeable enough on the topic to be able to educate others. I have some knowledge but I will not do justice when it comes to teaching.
Lovely video Sreeni! I have been contemplating using Google cloud but the sheer breadth of details makes it challenging to understand the expenses. This video really is helpful for clearing a lot of my doubts about Google cloud. I will probably be using it for my post doctoral research! Thank you for such a great content :)
I work with gpt-2 and gpt2 and other text generation tasks. I want to buy a GPU. what is your recommendation, please? My budget is around 3000 dollars. I do not like google colab pro+ because I do not save my work
If anyone is trying to get into image processing/segmentation at a research level what would you recommend studying or doing? Also it would be really interesting to know more about you and the day to day work and use of python or any other programming language in it. Thank you for these wonderful videos !!
If you want to get into image processing at research level you should be studying whatever the relevant field is. If you have accounting background and want to get into image analysis, it will be difficult. If you have sciences background, you can look into related filed. It fully depends on your background. You cannot expect a professor to take you as a researcher if you don't have relevant background. If you don't have a clue, just focus on studying computers sciences with machine learning focus. You can then choose a professor who works in the field of image processing when you get to the research stage (graduate school).
@@DigitalSreeni Thank you for your advice Dr. Sreeni . I currently am in a graduate program here in India but it's rare for people to get into research here and almost all of the faculties are not even PhDs so unfortunately can't work under them and yes I am a Computer Scientist in Data Science so do want to get into PhD related to Computational Neuroscience or a broad spectrum of Image Analysis using DL or AI abroad. Feature extraction or engineering is what interests me to explore more about images/ datasets and your channel is a gold mine for me. I would love to connect with you maybe on a live session or twitter or LinkedIn or whatever's possible. Your guidance will help me out a lot.
I used Colab( Free version) for all my master thesis work, what I observed is the session terminates after 1 hour initially later on it terminates just after 30-45 mins of idle time. Anyways Great Video BTW...
Thanks for sharing. Colab is great and I love it, it does an amazing job for something that is free. If you want predictable environment you need to pay.
I was thinking to buy my own workstation that will cost me 3L inr with a5000. I work at office where I can use HPC, but at home colab pro takes much time and I hate the runtime shutdowns. Shall I buy the workstation, I am confused
3L INR is a lot of money but if you have to get work done without interruption may be it is a good investment you make in your future. Sorry, this is totally your decision based on what you'd gain from the spending vs. your personal financial situation. Not an easy decision but I am sure you have evaluated Colab Pro, Google (or AWS) cloud, and of course HPC at work. You should have enough data to make an informed decision. Good luck.
Noise2Void. arxiv.org/abs/1811.10980 Here is the recording of my recent presentation on this topic at a workshop: th-cam.com/video/yO15IISXA1Y/w-d-xo.html
Great video! I was wondering if you had a link or some resource for comparing workstation-class gpus versus consumer-class gpu e.g., rtx 3090 or 4090 vs nvidia t4 or a100. Are there any places where I can get benchmarks across both classes of gpus?
Hello sir, my colab pro+ 1 weak ago stil get gpu v100 and p100 but in day i why my colab pro+ get gpu telsa t4 and i try to reset runtime to searching gpu v100 or p100 but still not working iget T4 and realy in pro+ get K80 my question sir, how to do fix thanks
You do not have control over which GPU you can get on Colab. It is assigned based on the availability. If you want to pick your own GPU then you need to subscribe to the more expensive Google cloud options.
Can you let me know how did you estimated the total ram consumption per image at 26:10 . I am currently working on spine segmentation, and my input to 3D U-Net is an image of size (32x256x256x1) so i wanted to estimate the overall ram consumptions and buy the resources accordingly. In collab free version the code crashes even with the batch size of 1
Is this supposed to scare us or paid by google? Did you forget to mention ebay? On colab pro you do not get a RTX 3090. On ebay a gpu server costs $250 and a K80 cost $100, a P100 cost $500. also if you do not have fund buy second hand.
Thanks so much. I have just started my PhD in a deep learning project with medical imaging and your videos are very much helpful. Keep making videos. :)
Thanks so much. I have just started my PhD in a deep learning project with medical imaging and your videos are very much helpful. Keep making videos. :)
Good luck with your PhD. I am glad these videos are helping you build your initial skillset. I am sure you will outgrow the content soon.
I can totally sympathize with your frustration about the need to babysit the train run. Thanks for sharing this!
Amazing content. As always. Thank you sir.
Thank you so much for your channel. It is very helpful. Wanted to know if you have a solution for a student account that is not authorized to google pay and buy Colab pro. I have a like 40 GB data to train. Your answer is appreciated.
Another great presentation Sreeni. I think an additional factor to consider is the time spent maintaining your local workstation. My GPU server is having hardware issues and I have sunk many hours this week fixing it :(. I wonder if it was worth it or just going cloud was a better option.
Great content and comparison. Thanks a lot!
This was extremely informative. Thanks for this
My pleasure!
Very useful !!! You always is my light tower
Great video! Any plans on pytorch series? Something on object detection/segmention using pytorch.?
Sorry, no pytorch. I am not knowledgeable enough on the topic to be able to educate others. I have some knowledge but I will not do justice when it comes to teaching.
Thank u very much, u saved my time and money!!!
thanks a lot sreeni really useful!
a comprehensive review of the three different platforms
Lovely video Sreeni! I have been contemplating using Google cloud but the sheer breadth of details makes it challenging to understand the expenses. This video really is helpful for clearing a lot of my doubts about Google cloud. I will probably be using it for my post doctoral research! Thank you for such a great content :)
My pleasure 😊
Thank you a lot for this explanation. I liked it.
Glad it was helpful!
I work with gpt-2 and gpt2 and other text generation tasks. I want to buy a GPU. what is your recommendation, please? My budget is around 3000 dollars. I do not like google colab pro+ because I do not save my work
Idle timeout is you are not at your computer using the notebook. You have to be interacting with it regularly.
please make videos on MAsk RCNN
How is rtx 4060 laptop for facefusion 2.5.1?? Will it work faster than google colab, please help i am nee in tech😢
If anyone is trying to get into image processing/segmentation at a research level what would you recommend studying or doing?
Also it would be really interesting to know more about you and the day to day work and use of python or any other programming language in it.
Thank you for these wonderful videos !!
If you want to get into image processing at research level you should be studying whatever the relevant field is. If you have accounting background and want to get into image analysis, it will be difficult. If you have sciences background, you can look into related filed. It fully depends on your background. You cannot expect a professor to take you as a researcher if you don't have relevant background. If you don't have a clue, just focus on studying computers sciences with machine learning focus. You can then choose a professor who works in the field of image processing when you get to the research stage (graduate school).
@@DigitalSreeni Thank you for your advice Dr. Sreeni . I currently am in a graduate program here in India but it's rare for people to get into research here and almost all of the faculties are not even PhDs so unfortunately can't work under them and yes I am a Computer Scientist in Data Science so do want to get into PhD related to Computational Neuroscience or a broad spectrum of Image Analysis using DL or AI abroad.
Feature extraction or engineering is what interests me to explore more about images/ datasets and your channel is a gold mine for me. I would love to connect with you maybe on a live session or twitter or LinkedIn or whatever's possible. Your guidance will help me out a lot.
Great and valuable information
Glad you liked it
I used Colab( Free version) for all my master thesis work, what I observed is the session terminates after 1 hour initially later on it terminates just after 30-45 mins of idle time.
Anyways Great Video BTW...
Thanks for sharing. Colab is great and I love it, it does an amazing job for something that is free. If you want predictable environment you need to pay.
Wonderful explanation.
Glad you liked it
Thank you for breaking it down
Thank you so much ..
I was thinking to buy my own workstation that will cost me 3L inr with a5000. I work at office where I can use HPC, but at home colab pro takes much time and I hate the runtime shutdowns.
Shall I buy the workstation, I am confused
3L INR is a lot of money but if you have to get work done without interruption may be it is a good investment you make in your future. Sorry, this is totally your decision based on what you'd gain from the spending vs. your personal financial situation. Not an easy decision but I am sure you have evaluated Colab Pro, Google (or AWS) cloud, and of course HPC at work. You should have enough data to make an informed decision. Good luck.
Hello sir thank you very much for your vedio sir very helpful
Sir can you please suggest me best recent denoising technique for medical images
Noise2Void. arxiv.org/abs/1811.10980
Here is the recording of my recent presentation on this topic at a workshop: th-cam.com/video/yO15IISXA1Y/w-d-xo.html
i used colab pro and rtx3070, but colab pro is better, if you are not 24/7 user
good video. Thank you for apologizing for the longevity of the video
:)
thank you
Great video! I was wondering if you had a link or some resource for comparing workstation-class gpus versus consumer-class gpu e.g., rtx 3090 or 4090 vs nvidia t4 or a100. Are there any places where I can get benchmarks across both classes of gpus?
Thanks
I want to buy colab pro, but I am unable to pay the amount.
Hello sir, my colab pro+ 1 weak ago stil get gpu v100 and p100 but in day i why my colab pro+ get gpu telsa t4 and i try to reset runtime to searching gpu v100 or p100 but still not working iget T4 and realy in pro+ get K80 my question sir, how to do fix thanks
You do not have control over which GPU you can get on Colab. It is assigned based on the availability. If you want to pick your own GPU then you need to subscribe to the more expensive Google cloud options.
Can you let me know how did you estimated the total ram consumption per image at 26:10 . I am currently working on spine segmentation, and my input to 3D U-Net is an image of size (32x256x256x1) so i wanted to estimate the overall ram consumptions and buy the resources accordingly. In collab free version the code crashes even with the batch size of 1
probably 32x256x256 / 1024(MB)/1024(KB)/8(bites) *4 (float number size) = 1 MB
the pro version doubles your RAM. It's definitely not similar
His GPU hasn't been throttled yet :D
Is this supposed to scare us or paid by google? Did you forget to mention ebay? On colab pro you do not get a RTX 3090. On ebay a gpu server costs $250 and a K80 cost $100, a P100 cost $500. also if you do not have fund buy second hand.
They never let me be idle for 90 minutes, not even close. I'd say it's just about 30 minutes for me.
Thanks so much. I have just started my PhD in a deep learning project with medical imaging and your videos are very much helpful. Keep making videos. :)
Thanks