ash/icc-demo-0
1
0
Fork 0
forked from qoherent/icc-demo
Code Pull Requests Workflows Activity
18 Commits 1 Branch 0 Tags 13 MiB
Commit Graph

18 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Aash
5e4cbcb6d7 Training run - 2026-05-28 07:10:29 2026-05-28 07:10:29 -04:00
R Roman Pope
ef0a0e430f Add datasets from qoherent/icc-28 for the WavesFM ICC demo 2026-05-28 07:06:03 -04:00
R Roman Pope
e620d86701 Add Ash's good recordings (USRP 72/97/2GHz captures) 
Copied from qoherent/ash_test_demo_repo/good/ for the WavesFM ICC demo.
SigMF data + meta pairs, ~27 MB total. Use these as input to the Curator.
 2026-05-28 07:03:52 -04:00
Aash
daa766e140 Training run - 2026-05-28 07:03:18 2026-05-28 07:03:18 -04:00
Aash
3ef7c1669b Training run - 2026-05-28 07:01:57 2026-05-28 07:01:57 -04:00
Aash
ceb2c3fc56 Training run - 2026-05-28 06:47:45 2026-05-28 06:47:45 -04:00
R Roman Pope
9cb3f35225 Workflow: use torch.cuda.amp.GradScaler (torch 2.2.x API) 2026-05-28 02:59:48 -04:00
R Roman Pope
9dd29ac589 Workflow: pin numpy<2 AFTER requirements.txt (torch 2.2.2 NumPy 1.x ABI) 2026-05-28 02:54:44 -04:00
R Roman Pope
b92bc91bea Workflow: pin numpy<2 (torch 2.2.2 ABI incompatible with numpy 2.x) 2026-05-28 02:52:21 -04:00
R Roman Pope
0c98d365d2 Workflow: pre-install CPU torch from pytorch.org/whl/cpu (saves ~600MB) 2026-05-28 02:47:38 -04:00
R Roman Pope
98926eb4a1 Re-render with ubuntu-latest label (routes to dawson-standard CPU runner) 2026-05-28 02:39:26 -04:00
R Roman Pope
0844b20ee4 Re-render workflow with dataset download_items + dawson-standard CPU runner 2026-05-28 02:36:43 -04:00
R Roman Pope
c5ed26372d Workflow: pin RIAHUB_BASE_URL to logan LAN IP 2026-05-28 01:58:59 -04:00
R Roman Pope
feb908e84e Re-trigger after repo transfer to qoherent (runner scope fix) 2026-05-28 01:53:45 -04:00
R Roman Pope
501a8c5e1c Add adapt_dataset.py + re-trigger training canary 
Previous run (2577) was waiting; missing scripts/adapt_dataset.py.
Adds the WAC-side adapter so the workflow's actions/checkout@v5
sparse-checkout finds the file at /scripts/adapt_dataset.py.
 2026-05-28 01:48:03 -04:00
R Roman Pope
b8fd7d73d9 Trigger WavesFM Linear Probe training (canary) 
Dataset: icc_canary_2026_05_28-v1.0.0.h5 (3 recs, 144 slices)
Model: qoherent/wavesfm-base/wavesfm-v1p0.pth @ 48787da4
Task: rml, epochs=3, batch_size=16, device=cpu
 2026-05-28 01:42:59 -04:00
R Roman Pope
9f87fa9fe2 Add canary dataset: icc_canary_2026_05_28 v1.0.0 (3 recs, 144 slices) 2026-05-28 01:40:55 -04:00
R Roman Pope
e2351eaaeb Initialize ICC demo workspace with LFS tracking + README 2026-05-28 01:35:31 -04:00