Skip to content

Actions: NiLabs-Org/ladder

Actions

All workflows

Actions

Loading...
Loading

Showing runs from all workflows
46 workflow runs
46 workflow runs

Filter by Workflow

⌥ + click/return to exclude

Filter by Event

⌥ + click/return to exclude

Filter by Status

⌥ + click/return to exclude

Filter by Branch

⌥ + click/return to exclude

Filter by Actor

⌥ + click/return to exclude
Point at NiLabs-Org after the org rename ci #45: Commit 5e606db pushed by natedemoss
40s main
main
40s
docs: report the runs that have actually happened ci #44: Commit aa1b880 pushed by natedemoss
44s main
main
44s
Record the first base-vs-tuned eval as inconclusive ci #43: Commit bd3284c pushed by natedemoss
41s main
main
41s
Keep a sample completion when no code is found ci #42: Commit 07eb569 pushed by natedemoss
39s main
main
39s
Always re-clone in the Colab notebook ci #41: Commit d766360 pushed by natedemoss
5m 39s main
main
5m 39s
Point the Colab notebook at the div2 A/B problem set ci #40: Commit 8868252 pushed by natedemoss
47s main
main
47s
Allow evaluation to be restricted by problem index ci #39: Commit f952dce pushed by natedemoss
48s main
main
48s
Fall back to a prompt when the Colab secret is missing ci #38: Commit 7a249d9 pushed by natedemoss
49s main
main
49s
Stop hiding pip output in the Colab notebook ci #37: Commit 6b54b4a pushed by natedemoss
42s main
main
42s
Make evaluation resumable, and add a Colab notebook to run it ci #36: Commit b0e04b1 pushed by natedemoss
38s main
main
38s
Record the first trained adapter, and check quota before pushing ci #35: Commit d4c2ea8 pushed by natedemoss
34s main
main
34s
Actually redirect the library caches ci #34: Commit 8272dd0 pushed by natedemoss
38s main
main
38s
Move mount discovery into the package and test it ci #33: Commit 3dfd526 pushed by natedemoss
52s main
main
52s
Print the slug actually pushed to ci #32: Commit b23c77a pushed by natedemoss
41s main
main
41s
Split training and evaluation into separate Kaggle sessions ci #31: Commit 90f1ee4 pushed by natedemoss
39s main
main
39s
Build SFTConfig directly instead of converting TrainingArguments ci #30: Commit d9bc422 pushed by natedemoss
41s main
main
41s
Search for the prepared data instead of assuming its mount path ci #29: Commit eb941e1 pushed by natedemoss
39s main
main
39s
Clone the repo outside /kaggle/working ci #28: Commit 2e4d596 pushed by natedemoss
39s main
main
39s
Add a smoke mode that runs the real kernel on real data ci #27: Commit c7f34c5 pushed by natedemoss
42s main
main
42s
Request a T4, and rewrite the trainer for the current trl API ci #26: Commit 5cdb778 pushed by natedemoss
41s main
main
41s
Attach the prepared dataset to the training kernel ci #25: Commit fe99b63 pushed by natedemoss
52s main
main
52s
Record the failed first run ci #24: Commit bba7b4d pushed by natedemoss
46s main
main
46s
Set the length bound from the measured distribution, not a biased sample ci #23: Commit f342316 pushed by natedemoss
40s main
main
40s
Record training throughput instead of estimating it (#8) ci #22: Commit 3fe6953 pushed by natedemoss
26s main
main
26s
Move data prep out of the GPU session ci #21: Commit 742aa61 pushed by natedemoss
50s main
main
50s