#๐ How to Run Colab 155,000 Data Entries Fast
202 messages ยท Page 1 of 1 (latest)
@rough nebula
Remember to:
- Ask your Python question, not if you can ask or if there's an expert who can help.
- Show a code sample as text (rather than a screenshot) and the error message, if you've got one.
- Explain what you expect to happen and what actually happens.
:warning: Do not pip install anything that isn't related to your question, especially if asked to over DMs.
What data? image, csv, excel, txt? Please elaborate more
as melon lord said, we need more info. can you provide the code you're running as well as the type of data?
CSV file
CSV file
ok, can you provide the code you're currently running too
because 155k entires in a csv normally wouldn't cause a crash
even in colab
uh i think you can send the data in dm but i can't help you there
i'll have to send the help messages here (not the data though)
yeah just data
Train Roberta on NormBank using the features: setting, behavior, and constraints or Test Roberta on health-related test set
If it passes the former it will crash on the latter
I just brought Pro to see if it works on there
will take a while to install transformers and torch and stuff, will get back here when it installs and runs
The writer of the code told me it runs on Colab
Just upload file into Colab
ah thanks
how long does it take to train? and do i need to run a training to completion to reproduce your behavior?
or do i just need to start it
that causes it to crash so i dunno
yeah run training
yeah i'm running training now but does it need to finish training
yes
yeah same thing happens to me
How do I connect my Colab to it
consumer grade
uh i don't think it's an option in colab
you'd have to own one personally or rent it elsewhere
My teacher has one
if you change runtime type to a100 though, that's probably the best you'll get
I can use his
it's not going to be that much faster than an a100 on colab fwiw
but it's far better than a t4 for the free plans
So 4090 and A100 are same in computing time?
not exactly the same but close enough iirc
in raw compute i think the 4090 is a bit faster but a100 has a bunch more vram so it can run much larger models than a 4090
damn
oof
I am hoping it does not crash like in the latter step
yeah we'll see
Which step are you on?
setting behavior and constraint
Did you finish the step: Train Roberta on NormBank using the feature: setting-behavior
uh no
i didnt realize that was also needed
idk yet
it hasnt crashed so i have no clue
when it crashes, do you get an error or does the window just go blank
oh
my screen went blank
and were you watching/clicking buttons consistently throughout the run
because there's an idle timeout of 90m
just watching
and if it runs more than 12h total it automatically disconnects
ah, well you'd need to be clicking buttons on the window
honestly training models in colab for research is probably not the best way to do it, it'd be much more reliable to run it locally on a jupyter notebook or something if you have a strong enough gpu
you mean 90 min?
then i was clicking buttons too
in this case ofc
i thought 90 ms
huh ok
and did you have the colab notebook open for >12 hours
nope
huh
what browser are you using
chrome
Train Roberta on NormBank using the features: setting, behavior, and constraints took 12 minutes to get epoch 2
mine just started epoch 5 a few seconds ago
JSonVue
any others?
ah well the proctoring app could cause issues
they're known to do such things because of how intrusive they are
ok removed it
ok, try restarting the runtime now
clean slate
if this doesn't work then all i can advise would be to just run the model on your teacher's 4090
how to i do that from colab?
the best way to do it would be not to run it on colab but rather local jupyter notebooks, but if you really need colab then local runtimes are the way to go
if i do 4090 do i run my codeb on vscode or locally
both could work
how do i run on it the server because the .py file is on my pc not on server
wdym on the server
is the 4090 on a server
if it's a linux server, i'd recommend SSHing in
otherwise you could remote desktop
yeah i it is linux server
ok then use ssh or vnc
ok
then you can just run python3 myscript.py from the ssh terminal
replace myscript with the name of the python file
python3: can't open file '/home/zhyahya/CSI_4495.py': [Errno 2] No such file or directory
hm
do i have to create a python file on my server account
yes
oh ok
any updates?
wtf... it crashed for me
np, if you need help setting up the sshing and local running in the future or now then i'd be happy to help here again
ok ill do that rn
oki
how do i transfer my CSV file on to it
ok
ok and are you on windows
if so, you can run on powershell SCP file.py remote_username@remote_ip:/home/yourusername/folder
do i connect ssh first
yes
wait
it doesnt matter actually
but the powershell part shoudl run on your windows machine
yes
CSI_4995.py: No such file or directory
ugh
thats the name
what folder is that file in
on windows
you'll want to be in the directory of that file in powershel
my user folder
or else pass the path from your current powershell directory
ok is the powershell terminal inside your user folder
if you ran it as admin it won't be by default
This help channel has been closed and it's no longer possible to send messages here. If your question wasn't answered, feel free to create a new post in #1035199133436354600. To maximize your chances of getting a response, check out this guide on asking good questions.