-
Notifications
You must be signed in to change notification settings - Fork 525
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
Upgrade composer version #560
Conversation
@dakinggg What should we generally look for with the regression tests you posted? I suppose decreasing loss for training? |
I'm just comparing to my previous runs of them. Each run is a group within that wandb project. Once Chuck's stuff is done they will be compared against a fixed "golden" run. |
So in this case I'm just seeing that the curves and number are identical to the ones I ran before. |
Although one of the runs actually failed...so looking into that now |
Ok one of them didn't actually ever run and I just missed it on my original PR. Gonna update in this PR. |
Ok, the failing regression test was user erorr. updated to mixed init. I'm comfortable merging given that validation done. We will set up proper regression tests once Chuck's stuff is ready. |
new regression group looks good: https://wandb.ai/mosaic-ml/llm-foundry-regressions/groups/09-05-2023%3A%3A19%3A27%3A44%3A%3A2a791a1ceb7ac620ad3149cc1792b6f1fcc815e7/workspace?workspace=user-danielking