-
Notifications
You must be signed in to change notification settings - Fork 125
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
ValueError: cannot reshape array of size 24 into shape (8,newaxis,8) in Dataloader #535
Comments
I'm ok with
And not ok with
Dont know where problem is? |
I cant launch with |
Hey @germanjke, sorry for the late response. Can you please share the below values for us to debug this issue?
|
Hey @karan6181 do you know how I can take this values from |
|
|
Hey @germanjke, let me help you get that information. Also, what streaming dataset version are you using? num_samples (int): Open An easier method to find out those values is to clone the streaming dataset, create a test branch, add a print statement at this line to print all the params, push a commit, install that custom branch, and run your job. |
Hi @karan6181
So the only change I have changed my orig.py file by yours, my llm foundry version is
So I guess some changes was just recently just in this file? Streaming version is the same |
Also here is my prints, I have 3 validations dataloaders and 1 train:
|
Can we deliver this version of file to |
old version file, here I have bugs
|
this one is good |
not relevant comment (edited), resume training if fine with new file |
@karan6181 What do you think, it's some conflicts in |
Hey @germanjke, sorry for my late response. Can you try the latest streaming dataset version? We have fixed some issues recently. |
Hi @karan6181 thank you everything works fine on last streaming and llm foundry versions |
Hi, looks like some new version of llm-foundry (updated from master) have lags in last week-two.
I have error like this
Changing
drop_last
toFalse
doesn't helpsIf I will change
num_workers
to some value will I lose some efficiency? Always launched with 8.The text was updated successfully, but these errors were encountered: