Did any one try lesson1 with colab and it worked?
I always get that error in the first training epoch :
cuda runtime error (2) : out of memory at /pytorch/torch/lib/THC/generic/THCStorage.cu:58
Yup I am building a collection of course notebooks adapted to run on Colab on my github if you’re interested. It’s github.com/brtknr
I always get that error in the first training epoch :
cuda runtime error (2) : out of memory at /pytorch/torch/lib/THC/generic/THCStorage.cu:58
use this tool called clouderizer… worked for me.
1 Like
yes it worked for me too but in uploading files or saving my work it doesn’t work well for me so i wanna now if i can work directly on colab or not ?and how to avoid the error i mentioned above ?
Hi Ahmed
I am one of the Founders of Clouderizer. Can you help me understand what challenge you faced in saving your work? It would help us improve Clouderizer and help lots of other students of this course.
first i thank you and all the Founders for Clouderizer it is really great and helps me alot
my first problem:
when i try to save a notebook if it has some pics (the size of the downloaded .ipynb file about 3 mb) the notebook always fails to save so this means i will lose my code after closing the notebook
what i did to avoid that i clear all the outputs which have pics so the size of .ipynb file will be small and it saves my code
i wonder if there is any better way ?
another big problem :
when i face a cuda error in colab i reset my machine without losing my stored data and continue my work
on Clouderizer i cant do this so i have to get out and install every thing again from scratch
Thanks a lot for this feedback.
Let me look into this. I will get back soon on this.
1 Like
Hi Ahmed
#1 I tried saving larger notebooks (>3mb) and did not face any issue. Can you probably share the notebook where you see this issue?
#2 Colab instances are reset by Google every 12 hours. So we do need to re-install our dependencies every time we get a fresh Colab instance. But Clouderizer helps us automate these dependency setup. I suggest you put all your installation related scripts under your project settings -> SETUP Tab -> Setup Script. This will auto run this script every time you start your project.
PS: For proper CUDA installation on Colab, i recommend enabling CUDA option in SETUP tab. This auto sets up CUDA every time your project is started.
