-
Notifications
You must be signed in to change notification settings - Fork 420
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
Unable to script model #3584
Comments
@callmekris can you also check the transformers version in both envs? |
@cli99 looks like on DBR 14.3 |
Here is the full list of installed packages in each environment: DBR 15.4:
DBR 14.3:
|
@callmekris if you install the same package versions on the older image ( If so, mind binary searching the package diff to try to narrow it down? I'm confused how an import could do anything. The dependencies in |
@mvpatel2000 sure, I will try that. I am confused as well. The issue does occur as well when importing just composer or any composer sub modules. |
Currently running mosaic composer in a databricks environment but experiencing different behavior of
torch.jit.script
after importing composer depending on the DBR installed.Code to reproduce
and
This is just a minimum example to reproduce, but the same happens when training a composer model and using
export_for_inference
.Expected Behavior DBR 14.3
Running with both with or without the composer import prints as expected:
On DBR >= 15.1
Running with the composer import prints:
And without the composer import it scripts correctly as expected.
On both runtimes the libary versions are:
torch 2.4.0
mosaicml 0.24.1
mosaicml-cli 0.6.41
I'm not exactly sure how to pinpoint the cause it only occurs on the newer DBRs even 15.4 LTS, but only after the import of composer.
The text was updated successfully, but these errors were encountered: