Commit graph

Select branches

Hide pull requests

main

#1

Mono Color

• 436ab78438
  
  dlr: when predicting, also display heatmap ...of positive predictions Starbeamrainbowlabs 2023-03-09 18:54:28 +0000
• 5195fe6b62
  
  jupyter test: heatmaps in matplotlib & seaborn Starbeamrainbowlabs 2023-03-09 18:52:56 +0000
• 45c76ba252
  
  typo Starbeamrainbowlabs 2023-03-03 22:46:02 +0000
• c909cfd3d1
  
  fixup Starbeamrainbowlabs 2023-03-03 22:45:34 +0000
• 0734201107
  
  dlr: tf graph changes Starbeamrainbowlabs 2023-03-03 22:44:49 +0000
• 750f46dbd2
  
  debug Starbeamrainbowlabs 2023-03-03 22:39:30 +0000
• 5472729f5e
  
  dlr: fixup argmax & y_true/y_pred Starbeamrainbowlabs 2023-03-03 22:37:36 +0000
• bc734a29c6
  
  y_true is one-hot, convert to sparse Starbeamrainbowlabs 2023-03-03 22:20:11 +0000
• c7b577ab29
  
  specificity: convert to plaintf Starbeamrainbowlabs 2023-03-03 22:16:48 +0000
• 26cc824ace
  
  dlr: MeanIoU fixup Starbeamrainbowlabs 2023-03-03 22:10:49 +0000
• e9dcbe3863
  
  dlr: fixup Starbeamrainbowlabs 2023-03-03 22:09:05 +0000
• 5c6789bf40
  
  meaniou: implement one-hot version it expects sparse, but our output is one-hot. Starbeamrainbowlabs 2023-03-03 22:04:21 +0000
• 6ffda40d48
  
  fixup Starbeamrainbowlabs 2023-03-03 21:54:45 +0000
• 9b13e9ca5b
  
  dlr: fixup argmax first Starbeamrainbowlabs 2023-03-03 21:51:24 +0000
• 7453c607ed
  
  argmax for sensitivity & specificity too Starbeamrainbowlabs 2023-03-03 21:49:33 +0000
• 8470aec996
  
  dlr: fixup Starbeamrainbowlabs 2023-03-03 21:45:51 +0000
• 3d051a8874
  
  dlr: HACK: argmax to convert [64,128,128, 2] → [64,128,128] Starbeamrainbowlabs 2023-03-03 21:41:26 +0000
• 94a32e7144
  
  dlr: fix metrics Starbeamrainbowlabs 2023-03-03 20:37:22 +0000
• c7f96ab6ab
  
  dlr: typo Starbeamrainbowlabs 2023-03-03 20:23:03 +0000
• 06f956dc07
  
  dlr: default LOSS, EPOCHS, and PREDICT_COUNT to better values Ref recent experiments Starbeamrainbowlabs 2023-03-03 20:17:08 +0000
• b435cc54dd
  
  dlr: add sensitivity (aka recall) and specificity metrics Starbeamrainbowlabs 2023-03-03 20:00:05 +0000
• 483ecf11c8
  
  add specificity metric Starbeamrainbowlabs 2023-03-03 19:35:20 +0000
• d464c9f57d
  
  dlr: add dice loss as metric more metrics to go tho Starbeamrainbowlabs 2023-03-03 19:34:55 +0000
• f70083bea4
  
  dlr eo: set custom_objects when loading model Starbeamrainbowlabs 2023-03-01 17:19:10 +0000
• b5f23e76d1
  
  dlr eo: allow setting DIR_OUTPUT directly Starbeamrainbowlabs 2023-03-01 16:54:15 +0000
• 4fd9feba4f
  
  dlr eo: tidyup Starbeamrainbowlabs 2023-03-01 16:47:36 +0000
• 69b5ae8838
  
  dlr eo: this should fix it Starbeamrainbowlabs 2023-02-23 17:24:30 +0000
• 9f1cee2927
  
  dlr eo: cheese it by upsampling and then downsampling again Starbeamrainbowlabs 2023-02-23 16:47:00 +0000
• 96b94ec55b
  
  upsampling test Starbeamrainbowlabs 2023-02-23 16:19:44 +0000
• 747ddfd41b
  
  weird, XLA_FLAGS cuda data dir wasn't needed before libdevice not found at ./libdevice.10.bc Starbeamrainbowlabs 2023-02-10 13:28:34 +0000
• e43274cd91
  
  dlr eo: add VAL_STEPS_PER_EPOCH Starbeamrainbowlabs 2023-02-03 16:41:30 +0000
• 8446a842d1
  
  typo Starbeamrainbowlabs 2023-02-03 16:01:54 +0000
• 1a8f10339a
  
  LayerConvNeXtGamma: fix for mixed precision mode Starbeamrainbowlabs 2023-02-02 16:22:08 +0000
• a630db2c49
  
  dlr eo: fixup Starbeamrainbowlabs 2023-02-02 16:17:52 +0000
• 2bf1872aca
  
  dlr eo: add JIT_COMPILE and MIXED_PRECISION Starbeamrainbowlabs 2023-02-02 16:14:09 +0000
• 71088b8c0b
  
  typo Starbeamrainbowlabs 2023-02-02 15:48:49 +0000
• f7666865a0
  
  dlr eo: add STEPS_PER_EXECUTION Starbeamrainbowlabs 2023-02-02 15:47:08 +0000
• f8202851a1
  
  dlr eo: add LEARNING_RATE Starbeamrainbowlabs 2023-01-27 16:51:13 +0000
• fb898ea72b
  
  slurm eo: seriously....? Starbeamrainbowlabs 2023-01-26 17:02:33 +0000
• be946091b1
  
  slurm eo: DIR_OUTPUT → DIRPATH_OUTPUT Starbeamrainbowlabs 2023-01-26 16:52:14 +0000
• c26a937cdd
  
  Ignore Kate swap files Starbeamrainbowlabs 2023-01-20 20:33:32 +0000
• 4703bdbea1
  
  SLURM: add job file for encoderonly It's pretty much bugfixed, but illykin doesn't have enough RAM to support it at the moment :-( Starbeamrainbowlabs 2023-01-20 20:32:35 +0000
• 818d77c733
  
  Make dirpath_rainfallwater consistent with other experiments Starbeamrainbowlabs 2023-01-20 20:31:26 +0000
• e72d3991b8
  
  switch to a smaller ConvNeXt Starbeamrainbowlabs 2023-01-20 19:14:38 +0000
• e1ad16a213
  
  debug A Starbeamrainbowlabs 2023-01-20 18:58:45 +0000
• 65a2e16a4c
  
  ds_eo: lower memory usage Starbeamrainbowlabs 2023-01-20 18:55:52 +0000
• b5e68fc1a3
  
  eo: don't downsample ConvNeXt at beginning Starbeamrainbowlabs 2023-01-20 18:49:46 +0000
• d5fdab50ed
  
  dlreo: missing import Starbeamrainbowlabs 2023-01-20 18:40:35 +0000
• 4514086dc6
  
  make_encoderonly: kwargs Starbeamrainbowlabs 2023-01-20 18:39:35 +0000
• 35dbd3f8bc
  
  ds eo: scale up rainfall data Starbeamrainbowlabs 2023-01-20 18:37:08 +0000
• 5b54ceec48
  
  ds eo: debug Starbeamrainbowlabs 2023-01-20 18:36:14 +0000
• a3787f0647
  
  debug Starbeamrainbowlabs 2023-01-20 18:34:56 +0000
• 9bd5e0f7a3
  
  ds_eo: debug there's something fishy going on here. Starbeamrainbowlabs 2023-01-20 17:25:22 +0000
• aa3831bc4f
  
  ds_eo: pass water to reshape Starbeamrainbowlabs 2023-01-17 19:07:16 +0000
• a01c49414f
  
  ds_eo: typo Starbeamrainbowlabs 2023-01-17 19:06:29 +0000
• b9bea26d26
  
  ds_eo: rogue variables Starbeamrainbowlabs 2023-01-17 19:03:14 +0000
• 64c57bbc21
  
  dlr: add no-requeue Starbeamrainbowlabs 2023-01-17 18:20:26 +0000
• a28bbb9cf7
  
  encoderonly: make executable Starbeamrainbowlabs 2023-01-17 17:55:10 +0000
• 835b376c72
  
  slurm dlr: log exit code Starbeamrainbowlabs 2023-01-17 15:18:26 +0000
• 40a550f155
  
  slurm dlr: fixup Starbeamrainbowlabs 2023-01-16 18:45:08 +0000
• 7bcf13f8d8
  
  dlr: typo Starbeamrainbowlabs 2023-01-16 18:02:09 +0000
• 6ff2864d23
  
  slurm dlr: shell out in conda; redirect stderr & stdout to disk inside the experiments folder Starbeamrainbowlabs 2023-01-16 17:32:22 +0000
• 0b31c9e700
  
  log start time, end time, and elapsed just in case...! Starbeamrainbowlabs 2023-01-16 17:30:20 +0000
• 1a4ac3ed66
  
  encoderonly: add graph plotting Starbeamrainbowlabs 2023-01-13 19:08:38 +0000
• 82e01da70b
  
  fixup.... again Starbeamrainbowlabs 2023-01-13 18:47:29 +0000
• 7b10f5c5fe
  
  dlr: add learning_rate env var Starbeamrainbowlabs 2023-01-13 18:29:39 +0000
• 2f0ce0aa13
  
  again Starbeamrainbowlabs 2023-01-13 18:27:03 +0000
• e04d6ab1b6
  
  fixup again Starbeamrainbowlabs 2023-01-13 18:23:32 +0000
• 37d1598b0b
  
  loss cel+dice: fixup Starbeamrainbowlabs 2023-01-13 18:21:11 +0000
• 0f0b691b5d
  
  loss cel + dice: fixup Starbeamrainbowlabs 2023-01-13 18:09:31 +0000
• be77f035c8
  
  dlr: add cross-entropy + dice loss fn option Starbeamrainbowlabs 2023-01-13 17:58:00 +0000
• b2a5acaf4e
  
  dlr dsm: clash Starbeamrainbowlabs 2023-01-13 17:26:38 +0000
• f7672db599
  
  annoying Starbeamrainbowlabs 2023-01-13 17:00:47 +0000
• 3c4d1c5140
  
  dlr: Add support for stripping isolated water pixels Starbeamrainbowlabs 2023-01-13 16:57:26 +0000
• ce1467461d
  
  fixup Starbeamrainbowlabs 2023-01-13 16:47:52 +0000
• 0b676fa391
  
  move shuffle to subdir Starbeamrainbowlabs 2023-01-13 16:47:35 +0000
• 2edfb1a21f
  
  dlr predict: comment debug Starbeamrainbowlabs 2023-01-12 19:20:22 +0000
• f0dd9711ed
  
  dlr: fixup Starbeamrainbowlabs 2023-01-12 18:55:33 +0000
• 176dc022a0
  
  add moar env vars Starbeamrainbowlabs 2023-01-12 18:54:39 +0000
• 0d41bbba94
  
  dlr predict: output with higher quality Starbeamrainbowlabs 2023-01-12 18:43:48 +0000
• 1b5bb14d8f
  
  dlr: debug Starbeamrainbowlabs 2023-01-12 18:35:29 +0000
• 20f7d34fd1
  
  fixup Starbeamrainbowlabs 2023-01-12 18:21:20 +0000
• c0c6e81c01
  
  dlr predict: allow for multiple outputs Starbeamrainbowlabs 2023-01-12 18:12:50 +0000
• 864dfa802d
  
  dlr/predict typo Starbeamrainbowlabs 2023-01-12 18:03:06 +0000
• 0140e93bbd
  
  dlr: trying a thing Starbeamrainbowlabs 2023-01-12 18:00:16 +0000
• e1666026ad
  
  dlr/predict: let's try another way Starbeamrainbowlabs 2023-01-12 17:56:59 +0000
• e7c0328648
  
  typo Starbeamrainbowlabs 2023-01-12 16:58:23 +0000
• 376eecc29f
  
  dlr: try plotting the label too Starbeamrainbowlabs 2023-01-12 16:13:04 +0000
• be7dd91f88
  
  fix crash 'cause we're only plotting 1 thing Starbeamrainbowlabs 2023-01-11 17:41:55 +0000
• 3787155665
  
  dlr: we can't plot the input tensor because it has 8 channels rather than 1 or 3 Starbeamrainbowlabs 2023-01-11 17:39:14 +0000
• 373dda03b5
  
  missing comma Starbeamrainbowlabs 2023-01-11 17:28:13 +0000
• 7be0509ac8
  
  dlr: slurm PATH_CHECKPOINT Starbeamrainbowlabs 2023-01-11 17:27:26 +0000
• a69c809008
  
  dlr: slurm, load checkpoint Starbeamrainbowlabs 2023-01-11 17:26:57 +0000
• 93e663e45d
  
  add optional PATH_CHECKPOINT env var Starbeamrainbowlabs 2023-01-11 17:20:19 +0000
• 0e3de8f5fc
  
  dlr: fix predictions Starbeamrainbowlabs 2023-01-10 19:19:30 +0000
• 2591cbe6bc
  
  slurm dlr: quiet, pip Starbeamrainbowlabs 2023-01-10 18:12:35 +0000
• 1958c4e6c2
  
  encoderonly model: getting there Starbeamrainbowlabs 2023-01-09 19:33:41 +0000
• 4b7df39fac
  
  bugfix Starbeamrainbowlabs 2023-01-09 18:25:16 +0000
• 581006cbe6
  
  dlr: save checkpoints Starbeamrainbowlabs 2023-01-09 18:03:23 +0000
• 52cf66ca32
  
  start working on a quick encoder test, but it's far from finished Starbeamrainbowlabs 2023-01-06 19:55:52 +0000