Skip to content

Actions: VihanAggarwal/shift-safe-monitors

Actions

All workflows

Actions

Loading...
Loading

Showing runs from all workflows
56 workflow runs
56 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
draft_2: title names the decoupling mechanism directly tests #55: Commit ceb76a1 pushed by VihanAggarwal
3m 43s master
master
3m 43s
draft_2: 6 pages to 5 by layout only, plus a new title tests #54: Commit a636ba6 pushed by VihanAggarwal
3m 39s master
master
3m 39s
SmolVLA: install num2words (undeclared SmolVLM processor dep) tests #53: Commit 432a143 pushed by VihanAggarwal
3m 24s master
master
3m 24s
Figures: author at IEEE column width; add fusion Pareto + severity in… tests #52: Commit 8ce7cfc pushed by VihanAggarwal
2m 40s master
master
2m 40s
Fusion: suppress degenerate per-dimension AUROC (1 success in 60 read… tests #51: Commit 107f64d pushed by VihanAggarwal
3m 27s master
master
3m 27s
Add run_analysis.ipynb: self-contained analysis-only notebook tests #50: Commit 641d1d0 pushed by VihanAggarwal
3m 21s master
master
3m 21s
DisMon: cap prefixes per episode (pi05 unchanged, OpenVLA ~10x lighte… tests #49: Commit 8f720f3 pushed by VihanAggarwal
3m 33s master
master
3m 33s
Paper: correct the wall-clock cost attribution (7.5s was the 13-pass … tests #48: Commit 17bda6f pushed by VihanAggarwal
3m 33s master
master
3m 33s
Severity stratification: within-policy identification of the rung int… tests #47: Commit 6c7cb9e pushed by VihanAggarwal
3m 34s master
master
3m 34s
Report held-out dimension census: single-outcome dims are excluded an… tests #46: Commit 4f47ceb pushed by VihanAggarwal
3m 20s master
master
3m 20s
DisMon: train on GPU when available + epoch heartbeat (CPU was ~100x … tests #45: Commit 5f65f88 pushed by VihanAggarwal
3m 26s master
master
3m 26s
Third policy: SmolVLA LIBERO finetune (HuggingFaceVLA/smolvla_libero)… tests #44: Commit 2cf1556 pushed by VihanAggarwal
3m 32s master
master
3m 32s
Paper draft + fusion with cross-fitted combiner; import pi05/openvla … tests #43: Commit 1a87563 pushed by VihanAggarwal
3m 32s master
master
3m 32s
Probe: flag untrainable clean split (pi05 had 0 failures in 48 episodes) tests #42: Commit f8dbf26 pushed by VihanAggarwal
2m 4s master
master
2m 4s
DisMon statistics: episode-level paired inference (protocol revision) tests #41: Commit 49907f0 pushed by VihanAggarwal
1m 57s master
master
1m 57s
True-clean OpenVLA diagnostic against the ORIGINAL LIBERO benchmark tests #40: Commit 5ec84d5 pushed by VihanAggarwal
1m 44s master
master
1m 44s
OpenVLA: center crop (official eval parity) + true-clean smoke gate tests #39: Commit b051f00 pushed by VihanAggarwal
1m 38s master
master
1m 38s
OpenVLA: gripper convention transform (normalize+binarize+invert), th… tests #38: Commit 32b8719 pushed by VihanAggarwal
1m 39s master
master
1m 39s
OpenVLA: replicate the eval's JPEG+lanczos-224 image pipeline (train-… tests #37: Commit 9d36a5e pushed by VihanAggarwal
1m 37s master
master
1m 37s
OpenVLA: rotate agentview 180 degrees to match train preprocessing tests #36: Commit e6516be pushed by VihanAggarwal
1m 37s master
master
1m 37s
OpenVLA cost control: signal stride + reduced sampling + half-size sw… tests #35: Commit 3e08b64 pushed by VihanAggarwal
1m 44s master
master
1m 44s
OpenVLA loads bf16 by default (fits A100/L4); explicit device_map whe… tests #34: Commit 0c55cea pushed by VihanAggarwal
1m 38s master
master
1m 38s
disable TF/Flax in transformers before any import (Colab TF deadlocks… tests #33: Commit 7a10b85 pushed by VihanAggarwal
1m 51s master
master
1m 51s
colab: restore mujoco==3.2.7 pin in OpenVLA stack cells (dropped when… tests #32: Commit dedfc28 pushed by VihanAggarwal
1m 38s master
master
1m 38s
colab: dedicated OpenVLA notebook for the main account (forward sweep… tests #31: Commit 32d8cf0 pushed by VihanAggarwal
1m 43s master
master
1m 43s