Archive 4 checkpoints (1-4)

paper-releases/pbisc-diffusion/updates/20261002T124820Z-crosscohort-liu-default/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/ho_giroux2022/K1000/lora16/train.log

@@ -0,0 +1,83 @@
+2026-10-02T20:35:38 split=ho_giroux2022 K=1000 mode=lora trainable=1,296,224 balance=cohort immicro: train 95 rows / 347,061 cells, inner 12; ye2026_sepsis: train 73 rows / 529,816 cells, inner 9; imperial_eskd: train 146 rows / 462,172 cells, inner 11
+2026-10-02T20:35:40 step 0 inner-val loss 0.11338 (pretrained) {'immicro': 0.10636159498244524, 'ye2026_sepsis': 0.12489082664251328, 'imperial_eskd': 0.10889218679883263}
+2026-10-02T20:35:51 step 250 train 0.10602 inner-val 0.11268 best 0.11268@250 (11s) immicro=0.1057 ye2026_sepsis=0.1242 imperial_eskd=0.1081
+2026-10-02T20:36:03 step 500 train 0.10368 inner-val 0.11092 best 0.11092@500 (23s) immicro=0.1042 ye2026_sepsis=0.1224 imperial_eskd=0.1062
+2026-10-02T20:36:14 step 750 train 0.10319 inner-val 0.10891 best 0.10891@750 (35s) immicro=0.1027 ye2026_sepsis=0.1201 imperial_eskd=0.1039
+2026-10-02T20:36:26 step 1000 train 0.10255 inner-val 0.10725 best 0.10725@1000 (47s) immicro=0.1016 ye2026_sepsis=0.1180 imperial_eskd=0.1021
+2026-10-02T20:36:38 step 1250 train 0.10281 inner-val 0.10615 best 0.10615@1250 (58s) immicro=0.1008 ye2026_sepsis=0.1165 imperial_eskd=0.1011
+2026-10-02T20:36:50 step 1500 train 0.10242 inner-val 0.10547 best 0.10547@1500 (70s) immicro=0.1002 ye2026_sepsis=0.1157 imperial_eskd=0.1006
+2026-10-02T20:37:01 step 1750 train 0.10256 inner-val 0.10503 best 0.10503@1750 (82s) immicro=0.0998 ye2026_sepsis=0.1151 imperial_eskd=0.1002
+2026-10-02T20:37:13 step 2000 train 0.10227 inner-val 0.10474 best 0.10474@2000 (94s) immicro=0.0995 ye2026_sepsis=0.1147 imperial_eskd=0.1000
+2026-10-02T20:37:25 step 2250 train 0.10260 inner-val 0.10455 best 0.10455@2250 (105s) immicro=0.0993 ye2026_sepsis=0.1145 imperial_eskd=0.0998
+2026-10-02T20:37:37 step 2500 train 0.10230 inner-val 0.10441 best 0.10441@2500 (117s) immicro=0.0992 ye2026_sepsis=0.1143 imperial_eskd=0.0997
+2026-10-02T20:37:48 step 2750 train 0.10240 inner-val 0.10431 best 0.10431@2750 (129s) immicro=0.0991 ye2026_sepsis=0.1142 imperial_eskd=0.0996
+2026-10-02T20:38:00 step 3000 train 0.10243 inner-val 0.10423 best 0.10423@3000 (140s) immicro=0.0990 ye2026_sepsis=0.1141 imperial_eskd=0.0996
+2026-10-02T20:38:12 step 3250 train 0.10245 inner-val 0.10417 best 0.10417@3250 (152s) immicro=0.0990 ye2026_sepsis=0.1140 imperial_eskd=0.0995
+2026-10-02T20:38:24 step 3500 train 0.10263 inner-val 0.10413 best 0.10413@3500 (164s) immicro=0.0989 ye2026_sepsis=0.1140 imperial_eskd=0.0995
+2026-10-02T20:38:35 step 3750 train 0.10245 inner-val 0.10410 best 0.10410@3750 (176s) immicro=0.0989 ye2026_sepsis=0.1140 imperial_eskd=0.0995
+2026-10-02T20:38:47 step 4000 train 0.10226 inner-val 0.10407 best 0.10407@4000 (188s) immicro=0.0988 ye2026_sepsis=0.1139 imperial_eskd=0.0994
+2026-10-02T20:38:59 step 4250 train 0.10277 inner-val 0.10405 best 0.10405@4250 (200s) immicro=0.0988 ye2026_sepsis=0.1139 imperial_eskd=0.0994
+2026-10-02T20:39:11 step 4500 train 0.10214 inner-val 0.10403 best 0.10403@4500 (211s) immicro=0.0988 ye2026_sepsis=0.1139 imperial_eskd=0.0994
+2026-10-02T20:39:22 step 4750 train 0.10195 inner-val 0.10401 best 0.10401@4750 (222s) immicro=0.0988 ye2026_sepsis=0.1139 imperial_eskd=0.0994
+2026-10-02T20:39:34 step 5000 train 0.10228 inner-val 0.10400 best 0.10400@5000 (234s) immicro=0.0987 ye2026_sepsis=0.1139 imperial_eskd=0.0994
+2026-10-02T20:39:46 step 5250 train 0.10183 inner-val 0.10398 best 0.10398@5250 (246s) immicro=0.0987 ye2026_sepsis=0.1139 imperial_eskd=0.0994
+2026-10-02T20:39:58 step 5500 train 0.10214 inner-val 0.10397 best 0.10397@5500 (258s) immicro=0.0987 ye2026_sepsis=0.1138 imperial_eskd=0.0994
+2026-10-02T20:40:10 step 5750 train 0.10226 inner-val 0.10396 best 0.10396@5750 (270s) immicro=0.0987 ye2026_sepsis=0.1138 imperial_eskd=0.0993
+2026-10-02T20:40:22 step 6000 train 0.10209 inner-val 0.10395 best 0.10395@6000 (282s) immicro=0.0987 ye2026_sepsis=0.1138 imperial_eskd=0.0993
+2026-10-02T20:40:34 step 6250 train 0.10222 inner-val 0.10394 best 0.10394@6250 (294s) immicro=0.0987 ye2026_sepsis=0.1138 imperial_eskd=0.0993
+2026-10-02T20:40:46 step 6500 train 0.10240 inner-val 0.10393 best 0.10393@6500 (307s) immicro=0.0987 ye2026_sepsis=0.1138 imperial_eskd=0.0993
+2026-10-02T20:40:58 step 6750 train 0.10192 inner-val 0.10392 best 0.10392@6750 (319s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0993
+2026-10-02T20:41:10 step 7000 train 0.10223 inner-val 0.10391 best 0.10391@7000 (331s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0993
+2026-10-02T20:41:22 step 7250 train 0.10232 inner-val 0.10391 best 0.10391@7250 (343s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0993
+2026-10-02T20:41:34 step 7500 train 0.10208 inner-val 0.10390 best 0.10390@7500 (355s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0993
+2026-10-02T20:41:47 step 7750 train 0.10203 inner-val 0.10389 best 0.10389@7750 (367s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:41:58 step 8000 train 0.10174 inner-val 0.10389 best 0.10389@8000 (379s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:42:10 step 8250 train 0.10222 inner-val 0.10388 best 0.10388@8250 (390s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:42:22 step 8500 train 0.10183 inner-val 0.10388 best 0.10388@8500 (402s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:42:34 step 8750 train 0.10209 inner-val 0.10388 best 0.10388@8750 (414s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:42:46 step 9000 train 0.10237 inner-val 0.10387 best 0.10387@9000 (426s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:42:57 step 9250 train 0.10198 inner-val 0.10387 best 0.10387@9250 (437s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:43:07 step 9500 train 0.10206 inner-val 0.10386 best 0.10386@9500 (448s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:43:17 step 9750 train 0.10188 inner-val 0.10386 best 0.10386@9750 (458s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:43:28 step 10000 train 0.10170 inner-val 0.10385 best 0.10385@10000 (468s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:43:38 step 10250 train 0.10221 inner-val 0.10385 best 0.10385@10250 (478s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:43:48 step 10500 train 0.10196 inner-val 0.10385 best 0.10385@10500 (488s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:43:59 step 10750 train 0.10169 inner-val 0.10385 best 0.10385@10750 (499s) immicro=0.0986 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:44:11 step 11000 train 0.10199 inner-val 0.10384 best 0.10384@11000 (511s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:44:23 step 11250 train 0.10177 inner-val 0.10384 best 0.10384@11000 (523s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:44:34 step 11500 train 0.10192 inner-val 0.10384 best 0.10384@11000 (534s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:44:46 step 11750 train 0.10189 inner-val 0.10384 best 0.10384@11750 (546s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:44:57 step 12000 train 0.10183 inner-val 0.10383 best 0.10383@12000 (558s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:45:09 step 12250 train 0.10200 inner-val 0.10384 best 0.10383@12000 (570s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:45:21 step 12500 train 0.10217 inner-val 0.10383 best 0.10383@12500 (582s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:45:33 step 12750 train 0.10201 inner-val 0.10383 best 0.10383@12750 (594s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:45:45 step 13000 train 0.10211 inner-val 0.10383 best 0.10383@12750 (606s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:45:57 step 13250 train 0.10190 inner-val 0.10383 best 0.10383@12750 (618s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:46:09 step 13500 train 0.10210 inner-val 0.10383 best 0.10383@12750 (630s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:46:21 step 13750 train 0.10208 inner-val 0.10383 best 0.10383@13750 (641s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:46:33 step 14000 train 0.10230 inner-val 0.10382 best 0.10382@14000 (653s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:46:45 step 14250 train 0.10156 inner-val 0.10382 best 0.10382@14250 (665s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:46:57 step 14500 train 0.10180 inner-val 0.10382 best 0.10382@14250 (677s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:47:09 step 14750 train 0.10199 inner-val 0.10382 best 0.10382@14750 (689s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:47:20 step 15000 train 0.10201 inner-val 0.10381 best 0.10381@15000 (700s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:47:31 step 15250 train 0.10176 inner-val 0.10382 best 0.10381@15000 (711s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:47:42 step 15500 train 0.10201 inner-val 0.10382 best 0.10381@15000 (722s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:47:52 step 15750 train 0.10190 inner-val 0.10381 best 0.10381@15000 (733s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:48:03 step 16000 train 0.10167 inner-val 0.10382 best 0.10381@15000 (743s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:48:13 step 16250 train 0.10174 inner-val 0.10381 best 0.10381@15000 (753s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0992
+2026-10-02T20:48:23 step 16500 train 0.10186 inner-val 0.10381 best 0.10381@16500 (764s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:48:34 step 16750 train 0.10185 inner-val 0.10380 best 0.10380@16750 (774s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:48:44 step 17000 train 0.10195 inner-val 0.10380 best 0.10380@17000 (784s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:48:54 step 17250 train 0.10210 inner-val 0.10380 best 0.10380@17250 (795s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:49:05 step 17500 train 0.10178 inner-val 0.10380 best 0.10380@17250 (805s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:49:15 step 17750 train 0.10140 inner-val 0.10380 best 0.10380@17250 (816s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:49:26 step 18000 train 0.10239 inner-val 0.10380 best 0.10380@18000 (826s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:49:37 step 18250 train 0.10133 inner-val 0.10380 best 0.10380@18000 (838s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:49:49 step 18500 train 0.10169 inner-val 0.10380 best 0.10380@18000 (850s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:50:01 step 18750 train 0.10169 inner-val 0.10380 best 0.10380@18000 (861s) immicro=0.0985 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:50:12 step 19000 train 0.10190 inner-val 0.10380 best 0.10380@19000 (873s) immicro=0.0984 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:50:24 step 19250 train 0.10185 inner-val 0.10379 best 0.10379@19250 (884s) immicro=0.0984 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:50:36 step 19500 train 0.10182 inner-val 0.10379 best 0.10379@19500 (896s) immicro=0.0984 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:50:47 step 19750 train 0.10183 inner-val 0.10379 best 0.10379@19750 (907s) immicro=0.0984 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:50:57 step 20000 train 0.10166 inner-val 0.10379 best 0.10379@20000 (918s) immicro=0.0984 ye2026_sepsis=0.1138 imperial_eskd=0.0991
+2026-10-02T20:51:01 saved /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/ho_giroux2022/K1000/lora16/checkpoint_ft.pt (selected step 20000, inner-val 0.10379 vs pretrained 0.11338); sha e9f981f43cf4

paper-releases/pbisc-diffusion/updates/20261002T124820Z-crosscohort-liu-default/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/ho_immicro/K1000/lora16/train.log

@@ -0,0 +1,2 @@
+2026-10-02T21:29:36 split=ho_immicro K=1000 mode=lora trainable=1,296,224 balance=cohort ye2026_sepsis: train 73 rows / 529,816 cells, inner 9; imperial_eskd: train 146 rows / 462,172 cells, inner 11; giroux2022: train 24 rows / 254,250 cells, inner 0
+2026-10-02T21:29:37 step 0 inner-val loss 0.11736 (pretrained) {'ye2026_sepsis': 0.12609728963838684, 'imperial_eskd': 0.1086193590678952}

paper-releases/pbisc-diffusion/updates/20261002T124820Z-crosscohort-liu-default/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/ho_ji2025_chili/K1000/lora16/history.json

@@ -0,0 +1,892 @@
+[
+ {
+  "step": 0,
+  "inner_val_loss": 0.11338153614126371,
+  "inner_val_per_cohort": {
+   "immicro": 0.10636159498244524,
+   "ye2026_sepsis": 0.12489082664251328,
+   "imperial_eskd": 0.10889218679883263
+  },
+  "train_loss": null
+ },
+ {
+  "step": 250,
+  "inner_val_loss": 0.11268571607755994,
+  "inner_val_per_cohort": {
+   "immicro": 0.10574988679339488,
+   "ye2026_sepsis": 0.12423042249348429,
+   "imperial_eskd": 0.10807683894580061
+  },
+  "train_loss": 0.10532217663526534,
+  "sec": 10.2
+ },
+ {
+  "step": 500,
+  "inner_val_loss": 0.11100994602084059,
+  "inner_val_per_cohort": {
+   "immicro": 0.10427182633429766,
+   "ye2026_sepsis": 0.12257940156592263,
+   "imperial_eskd": 0.1061786101623015
+  },
+  "train_loss": 0.10238039651513099,
+  "sec": 20.5
+ },
+ {
+  "step": 750,
+  "inner_val_loss": 0.10911456837949128,
+  "inner_val_per_cohort": {
+   "immicro": 0.10291472325722377,
+   "ye2026_sepsis": 0.12037012726068497,
+   "imperial_eskd": 0.10405885462056506
+  },
+  "train_loss": 0.10229254439473152,
+  "sec": 30.6
+ },
+ {
+  "step": 1000,
+  "inner_val_loss": 0.10750445245642853,
+  "inner_val_per_cohort": {
+   "immicro": 0.1018434427678585,
+   "ye2026_sepsis": 0.1183084895213445,
+   "imperial_eskd": 0.10236142508008263
+  },
+  "train_loss": 0.10194027945399284,
+  "sec": 40.8
+ },
+ {
+  "step": 1250,
+  "inner_val_loss": 0.10637106817617059,
+  "inner_val_per_cohort": {
+   "immicro": 0.1010331620151798,
+   "ye2026_sepsis": 0.11677688029077318,
+   "imperial_eskd": 0.1013031622225588
+  },
+  "train_loss": 0.10189075681567192,
+  "sec": 51.2
+ },
+ {
+  "step": 1500,
+  "inner_val_loss": 0.10565690696239471,
+  "inner_val_per_cohort": {
+   "immicro": 0.10042685270309448,
+   "ye2026_sepsis": 0.11584022641181946,
+   "imperial_eskd": 0.1007036417722702
+  },
+  "train_loss": 0.10142263394594192,
+  "sec": 61.6
+ },
+ {
+  "step": 1750,
+  "inner_val_loss": 0.10519778752944083,
+  "inner_val_per_cohort": {
+   "immicro": 0.09999279119074345,
+   "ye2026_sepsis": 0.1152686042090257,
+   "imperial_eskd": 0.10033196718855338
+  },
+  "train_loss": 0.10156565123796463,
+  "sec": 71.7
+ },
+ {
+  "step": 2000,
+  "inner_val_loss": 0.1048861749294631,
+  "inner_val_per_cohort": {
+   "immicro": 0.09967753415306409,
+   "ye2026_sepsis": 0.11490441072318289,
+   "imperial_eskd": 0.10007657991214232
+  },
+  "train_loss": 0.1013408882021904,
+  "sec": 82.0
+ },
+ {
+  "step": 2250,
+  "inner_val_loss": 0.10467391645203435,
+  "inner_val_per_cohort": {
+   "immicro": 0.09945089214791854,
+   "ye2026_sepsis": 0.1146629543768035,
+   "imperial_eskd": 0.09990790283138101
+  },
+  "train_loss": 0.10135153838992118,
+  "sec": 92.1
+ },
+ {
+  "step": 2500,
+  "inner_val_loss": 0.10452337643627065,
+  "inner_val_per_cohort": {
+   "immicro": 0.09929436972985665,
+   "ye2026_sepsis": 0.11449142090148395,
+   "imperial_eskd": 0.09978433867747133
+  },
+  "train_loss": 0.1014540764093399,
+  "sec": 102.2
+ },
+ {
+  "step": 2750,
+  "inner_val_loss": 0.10441061477756963,
+  "inner_val_per_cohort": {
+   "immicro": 0.09917737760891517,
+   "ye2026_sepsis": 0.11436007585790423,
+   "imperial_eskd": 0.09969439086588947
+  },
+  "train_loss": 0.10132831171154975,
+  "sec": 112.6
+ },
+ {
+  "step": 3000,
+  "inner_val_loss": 0.10433083668238285,
+  "inner_val_per_cohort": {
+   "immicro": 0.09909296377251546,
+   "ye2026_sepsis": 0.11426593859990437,
+   "imperial_eskd": 0.09963360767472874
+  },
+  "train_loss": 0.10155830910801887,
+  "sec": 123.0
+ },
+ {
+  "step": 3250,
+  "inner_val_loss": 0.1042687980182193,
+  "inner_val_per_cohort": {
+   "immicro": 0.09902951959520578,
+   "ye2026_sepsis": 0.1141900188393063,
+   "imperial_eskd": 0.0995868556201458
+  },
+  "train_loss": 0.10144265559315681,
+  "sec": 133.1
+ },
+ {
+  "step": 3500,
+  "inner_val_loss": 0.1042236109681252,
+  "inner_val_per_cohort": {
+   "immicro": 0.0989805127804478,
+   "ye2026_sepsis": 0.1141332160267565,
+   "imperial_eskd": 0.09955710409717126
+  },
+  "train_loss": 0.10139354580640793,
+  "sec": 143.2
+ },
+ {
+  "step": 3750,
+  "inner_val_loss": 0.10418816214044789,
+  "inner_val_per_cohort": {
+   "immicro": 0.09894356224685907,
+   "ye2026_sepsis": 0.11409166289700402,
+   "imperial_eskd": 0.09952926127748056
+  },
+  "train_loss": 0.10125392976403236,
+  "sec": 153.3
+ },
+ {
+  "step": 4000,
+  "inner_val_loss": 0.10415741490507467,
+  "inner_val_per_cohort": {
+   "immicro": 0.09890912007540464,
+   "ye2026_sepsis": 0.1140568856563833,
+   "imperial_eskd": 0.09950623898343607
+  },
+  "train_loss": 0.1012623402774334,
+  "sec": 163.4
+ },
+ {
+  "step": 4250,
+  "inner_val_loss": 0.10413460169765183,
+  "inner_val_per_cohort": {
+   "immicro": 0.09888594690710306,
+   "ye2026_sepsis": 0.11402724103795157,
+   "imperial_eskd": 0.09949061714790085
+  },
+  "train_loss": 0.1016057662665844,
+  "sec": 173.5
+ },
+ {
+  "step": 4500,
+  "inner_val_loss": 0.104114340762498,
+  "inner_val_per_cohort": {
+   "immicro": 0.09886556242903073,
+   "ye2026_sepsis": 0.1140011271668805,
+   "imperial_eskd": 0.09947633269158276
+  },
+  "train_loss": 0.10121665260195732,
+  "sec": 183.9
+ },
+ {
+  "step": 4750,
+  "inner_val_loss": 0.10409259069899117,
+  "inner_val_per_cohort": {
+   "immicro": 0.0988451528052489,
+   "ye2026_sepsis": 0.1139697494606177,
+   "imperial_eskd": 0.09946286983110687
+  },
+  "train_loss": 0.10145337191224098,
+  "sec": 194.1
+ },
+ {
+  "step": 5000,
+  "inner_val_loss": 0.10407820687651233,
+  "inner_val_per_cohort": {
+   "immicro": 0.09882775569955508,
+   "ye2026_sepsis": 0.11395229606164826,
+   "imperial_eskd": 0.09945456886833365
+  },
+  "train_loss": 0.10147962602972985,
+  "sec": 204.0
+ },
+ {
+  "step": 5250,
+  "inner_val_loss": 0.10406350837048357,
+  "inner_val_per_cohort": {
+   "immicro": 0.09881274991979201,
+   "ye2026_sepsis": 0.1139387082722452,
+   "imperial_eskd": 0.09943906691941348
+  },
+  "train_loss": 0.10119385233521462,
+  "sec": 214.2
+ },
+ {
+  "step": 5500,
+  "inner_val_loss": 0.10405073358805894,
+  "inner_val_per_cohort": {
+   "immicro": 0.09879643842577934,
+   "ye2026_sepsis": 0.11392726169692145,
+   "imperial_eskd": 0.09942850064147603
+  },
+  "train_loss": 0.10119852238893509,
+  "sec": 224.4
+ },
+ {
+  "step": 5750,
+  "inner_val_loss": 0.10403779038676508,
+  "inner_val_per_cohort": {
+   "immicro": 0.0987834595143795,
+   "ye2026_sepsis": 0.11391688179638651,
+   "imperial_eskd": 0.09941302984952927
+  },
+  "train_loss": 0.10095700272917747,
+  "sec": 234.5
+ },
+ {
+  "step": 6000,
+  "inner_val_loss": 0.10402673372829502,
+  "inner_val_per_cohort": {
+   "immicro": 0.09877338043103616,
+   "ye2026_sepsis": 0.11390228321154912,
+   "imperial_eskd": 0.09940453754229979
+  },
+  "train_loss": 0.10108271023631096,
+  "sec": 244.8
+ },
+ {
+  "step": 6250,
+  "inner_val_loss": 0.10401370544739143,
+  "inner_val_per_cohort": {
+   "immicro": 0.09876015565047662,
+   "ye2026_sepsis": 0.11388907829920451,
+   "imperial_eskd": 0.09939188239249316
+  },
+  "train_loss": 0.10137384250760079,
+  "sec": 255.0
+ },
+ {
+  "step": 6500,
+  "inner_val_loss": 0.1040054790249176,
+  "inner_val_per_cohort": {
+   "immicro": 0.09875234464804332,
+   "ye2026_sepsis": 0.1138806396888362,
+   "imperial_eskd": 0.09938345273787325
+  },
+  "train_loss": 0.10143261170387267,
+  "sec": 265.2
+ },
+ {
+  "step": 6750,
+  "inner_val_loss": 0.10399649569118764,
+  "inner_val_per_cohort": {
+   "immicro": 0.09874238912016153,
+   "ye2026_sepsis": 0.11387271392676565,
+   "imperial_eskd": 0.09937438402663577
+  },
+  "train_loss": 0.10109712529182434,
+  "sec": 275.4
+ },
+ {
+  "step": 7000,
+  "inner_val_loss": 0.10398782724796822,
+  "inner_val_per_cohort": {
+   "immicro": 0.09873311625172694,
+   "ye2026_sepsis": 0.11386858920256297,
+   "imperial_eskd": 0.09936177628961476
+  },
+  "train_loss": 0.10131280601024628,
+  "sec": 285.6
+ },
+ {
+  "step": 7250,
+  "inner_val_loss": 0.10398311679398874,
+  "inner_val_per_cohort": {
+   "immicro": 0.09872655539462964,
+   "ye2026_sepsis": 0.11386925395992067,
+   "imperial_eskd": 0.09935354102741588
+  },
+  "train_loss": 0.1014550264775753,
+  "sec": 295.6
+ },
+ {
+  "step": 7500,
+  "inner_val_loss": 0.1039740331502324,
+  "inner_val_per_cohort": {
+   "immicro": 0.09871727383385102,
+   "ye2026_sepsis": 0.1138656631939941,
+   "imperial_eskd": 0.09933916242285208
+  },
+  "train_loss": 0.10110629218816757,
+  "sec": 305.8
+ },
+ {
+  "step": 7750,
+  "inner_val_loss": 0.10396731501250756,
+  "inner_val_per_cohort": {
+   "immicro": 0.09871015263100465,
+   "ye2026_sepsis": 0.1138625985218419,
+   "imperial_eskd": 0.09932919388467615
+  },
+  "train_loss": 0.10104343870282173,
+  "sec": 316.3
+ },
+ {
+  "step": 8000,
+  "inner_val_loss": 0.10396365864321538,
+  "inner_val_per_cohort": {
+   "immicro": 0.09870493443061908,
+   "ye2026_sepsis": 0.11386452035771476,
+   "imperial_eskd": 0.09932152114131233
+  },
+  "train_loss": 0.10099819192290306,
+  "sec": 326.6
+ },
+ {
+  "step": 8250,
+  "inner_val_loss": 0.10395894966304603,
+  "inner_val_per_cohort": {
+   "immicro": 0.09869819786399603,
+   "ye2026_sepsis": 0.11386588340004285,
+   "imperial_eskd": 0.09931276772509921
+  },
+  "train_loss": 0.10144895792007447,
+  "sec": 336.8
+ },
+ {
+  "step": 8500,
+  "inner_val_loss": 0.10395441371759383,
+  "inner_val_per_cohort": {
+   "immicro": 0.09869107666114967,
+   "ye2026_sepsis": 0.11386727541685104,
+   "imperial_eskd": 0.09930488907478073
+  },
+  "train_loss": 0.10094235199689865,
+  "sec": 347.0
+ },
+ {
+  "step": 8750,
+  "inner_val_loss": 0.10394974791692564,
+  "inner_val_per_cohort": {
+   "immicro": 0.09868691861629486,
+   "ye2026_sepsis": 0.11386798777514034,
+   "imperial_eskd": 0.0992943373593417
+  },
+  "train_loss": 0.1010453799366951,
+  "sec": 357.2
+ },
+ {
+  "step": 9000,
+  "inner_val_loss": 0.10394631417135798,
+  "inner_val_per_cohort": {
+   "immicro": 0.09868303996821244,
+   "ye2026_sepsis": 0.11386894600258933,
+   "imperial_eskd": 0.09928695654327219
+  },
+  "train_loss": 0.10124289816617965,
+  "sec": 367.3
+ },
+ {
+  "step": 9250,
+  "inner_val_loss": 0.10394706312717512,
+  "inner_val_per_cohort": {
+   "immicro": 0.09867962853362162,
+   "ye2026_sepsis": 0.11388082926472028,
+   "imperial_eskd": 0.09928073158318346
+  },
+  "train_loss": 0.10114238899946212,
+  "sec": 377.6
+ },
+ {
+  "step": 9500,
+  "inner_val_loss": 0.10394128988989275,
+  "inner_val_per_cohort": {
+   "immicro": 0.09867739553252856,
+   "ye2026_sepsis": 0.11387917358014318,
+   "imperial_eskd": 0.09926730055700649
+  },
+  "train_loss": 0.10104817023873329,
+  "sec": 388.0
+ },
+ {
+  "step": 9750,
+  "inner_val_loss": 0.10393578213285215,
+  "inner_val_per_cohort": {
+   "immicro": 0.09867335017770529,
+   "ye2026_sepsis": 0.11387377066744699,
+   "imperial_eskd": 0.0992602255534042
+  },
+  "train_loss": 0.10104718473553657,
+  "sec": 399.5
+ },
+ {
+  "step": 10000,
+  "inner_val_loss": 0.10393376253319508,
+  "inner_val_per_cohort": {
+   "immicro": 0.0986677606900533,
+   "ye2026_sepsis": 0.11387657995025317,
+   "imperial_eskd": 0.0992569469592788
+  },
+  "train_loss": 0.10098761707544326,
+  "sec": 411.5
+ },
+ {
+  "step": 10250,
+  "inner_val_loss": 0.1039281279626318,
+  "inner_val_per_cohort": {
+   "immicro": 0.09866408010323842,
+   "ye2026_sepsis": 0.1138696898188856,
+   "imperial_eskd": 0.09925061396577141
+  },
+  "train_loss": 0.10102520111203193,
+  "sec": 423.3
+ },
+ {
+  "step": 10500,
+  "inner_val_loss": 0.10392464419218526,
+  "inner_val_per_cohort": {
+   "immicro": 0.09865874083091815,
+   "ye2026_sepsis": 0.11386898615294033,
+   "imperial_eskd": 0.09924620559269731
+  },
+  "train_loss": 0.10103527134656906,
+  "sec": 435.1
+ },
+ {
+  "step": 10750,
+  "inner_val_loss": 0.10392132246873094,
+  "inner_val_per_cohort": {
+   "immicro": 0.09865822363644838,
+   "ye2026_sepsis": 0.11386468551225132,
+   "imperial_eskd": 0.0992410582574931
+  },
+  "train_loss": 0.10112004658579826,
+  "sec": 447.0
+ },
+ {
+  "step": 11000,
+  "inner_val_loss": 0.10391800157625465,
+  "inner_val_per_cohort": {
+   "immicro": 0.09865444463988145,
+   "ye2026_sepsis": 0.11386912398868138,
+   "imperial_eskd": 0.09923043610020117
+  },
+  "train_loss": 0.101343277156353,
+  "sec": 457.2
+ },
+ {
+  "step": 11250,
+  "inner_val_loss": 0.10391534889988625,
+  "inner_val_per_cohort": {
+   "immicro": 0.09865046789248784,
+   "ye2026_sepsis": 0.11386868274874157,
+   "imperial_eskd": 0.09922689605842937
+  },
+  "train_loss": 0.10090706196427346,
+  "sec": 469.1
+ },
+ {
+  "step": 11500,
+  "inner_val_loss": 0.10391684689305046,
+  "inner_val_per_cohort": {
+   "immicro": 0.09864553560813268,
+   "ye2026_sepsis": 0.11387992898623149,
+   "imperial_eskd": 0.0992250760847872
+  },
+  "train_loss": 0.1011368117928505,
+  "sec": 481.0
+ },
+ {
+  "step": 11750,
+  "inner_val_loss": 0.10391746870608944,
+  "inner_val_per_cohort": {
+   "immicro": 0.09864173612246911,
+   "ye2026_sepsis": 0.11388841023047765,
+   "imperial_eskd": 0.09922225976532156
+  },
+  "train_loss": 0.10098819878697396,
+  "sec": 493.0
+ },
+ {
+  "step": 12000,
+  "inner_val_loss": 0.10391636405403566,
+  "inner_val_per_cohort": {
+   "immicro": 0.09864057786762714,
+   "ye2026_sepsis": 0.11389194677273433,
+   "imperial_eskd": 0.09921656752174551
+  },
+  "train_loss": 0.100804309040308,
+  "sec": 504.8
+ },
+ {
+  "step": 12250,
+  "inner_val_loss": 0.10391409625534458,
+  "inner_val_per_cohort": {
+   "immicro": 0.09863269266982873,
+   "ye2026_sepsis": 0.11389874418576558,
+   "imperial_eskd": 0.09921085191043941
+  },
+  "train_loss": 0.10083222579956054,
+  "sec": 516.7
+ },
+ {
+  "step": 12500,
+  "inner_val_loss": 0.10391240428995203,
+  "inner_val_per_cohort": {
+   "immicro": 0.09863117958108585,
+   "ye2026_sepsis": 0.11389775656991535,
+   "imperial_eskd": 0.0992082767188549
+  },
+  "train_loss": 0.10119791904091835,
+  "sec": 527.5
+ },
+ {
+  "step": 12750,
+  "inner_val_loss": 0.10390954837203026,
+  "inner_val_per_cohort": {
+   "immicro": 0.09862816582123439,
+   "ye2026_sepsis": 0.11389339094360669,
+   "imperial_eskd": 0.0992070883512497
+  },
+  "train_loss": 0.10090058439970016,
+  "sec": 537.8
+ },
+ {
+  "step": 13000,
+  "inner_val_loss": 0.10390789718727772,
+  "inner_val_per_cohort": {
+   "immicro": 0.09862361072252195,
+   "ye2026_sepsis": 0.11389671845568551,
+   "imperial_eskd": 0.09920336238362572
+  },
+  "train_loss": 0.10115658295154571,
+  "sec": 548.1
+ },
+ {
+  "step": 13250,
+  "inner_val_loss": 0.10390512244180096,
+  "inner_val_per_cohort": {
+   "immicro": 0.09861885911474626,
+   "ye2026_sepsis": 0.11389472873674499,
+   "imperial_eskd": 0.09920177947391164
+  },
+  "train_loss": 0.10116149839758873,
+  "sec": 558.5
+ },
+ {
+  "step": 13500,
+  "inner_val_loss": 0.10390345346157498,
+  "inner_val_per_cohort": {
+   "immicro": 0.09861652459949255,
+   "ye2026_sepsis": 0.1138943876657221,
+   "imperial_eskd": 0.0991994481195103
+  },
+  "train_loss": 0.10103107136487961,
+  "sec": 568.9
+ },
+ {
+  "step": 13750,
+  "inner_val_loss": 0.10389793845464285,
+  "inner_val_per_cohort": {
+   "immicro": 0.0986110537002484,
+   "ye2026_sepsis": 0.11388640519645479,
+   "imperial_eskd": 0.09919635646722534
+  },
+  "train_loss": 0.10119696509838104,
+  "sec": 579.2
+ },
+ {
+  "step": 14000,
+  "inner_val_loss": 0.10389720006322219,
+  "inner_val_per_cohort": {
+   "immicro": 0.098606721808513,
+   "ye2026_sepsis": 0.11388871901565129,
+   "imperial_eskd": 0.09919615936550227
+  },
+  "train_loss": 0.10120111861824989,
+  "sec": 590.3
+ },
+ {
+  "step": 14250,
+  "inner_val_loss": 0.10389802587980573,
+  "inner_val_per_cohort": {
+   "immicro": 0.0986039446045955,
+   "ye2026_sepsis": 0.11389595891038577,
+   "imperial_eskd": 0.09919417412443594
+  },
+  "train_loss": 0.10090549156069756,
+  "sec": 602.1
+ },
+ {
+  "step": 14500,
+  "inner_val_loss": 0.10389686479963839,
+  "inner_val_per_cohort": {
+   "immicro": 0.09859874844551086,
+   "ye2026_sepsis": 0.11389822554257181,
+   "imperial_eskd": 0.09919362041083249
+  },
+  "train_loss": 0.10123024705052376,
+  "sec": 613.4
+ },
+ {
+  "step": 14750,
+  "inner_val_loss": 0.1038960620058506,
+  "inner_val_per_cohort": {
+   "immicro": 0.09859851747751236,
+   "ye2026_sepsis": 0.11389909643265936,
+   "imperial_eskd": 0.09919057210738008
+  },
+  "train_loss": 0.10106333357095719,
+  "sec": 623.5
+ },
+ {
+  "step": 15000,
+  "inner_val_loss": 0.10389411238395303,
+  "inner_val_per_cohort": {
+   "immicro": 0.09859911818057299,
+   "ye2026_sepsis": 0.11389877936906284,
+   "imperial_eskd": 0.09918443960222331
+  },
+  "train_loss": 0.10110241189599037,
+  "sec": 633.8
+ },
+ {
+  "step": 15250,
+  "inner_val_loss": 0.10389225339824303,
+  "inner_val_per_cohort": {
+   "immicro": 0.09859745390713215,
+   "ye2026_sepsis": 0.11389660876658228,
+   "imperial_eskd": 0.09918269752101465
+  },
+  "train_loss": 0.10101236468553543,
+  "sec": 644.2
+ },
+ {
+  "step": 15500,
+  "inner_val_loss": 0.10388963025452112,
+  "inner_val_per_cohort": {
+   "immicro": 0.09859572909772396,
+   "ye2026_sepsis": 0.11389173650079304,
+   "imperial_eskd": 0.09918142516504634
+  },
+  "train_loss": 0.10117508390545846,
+  "sec": 655.2
+ },
+ {
+  "step": 15750,
+  "inner_val_loss": 0.10388465908666451,
+  "inner_val_per_cohort": {
+   "immicro": 0.09859144253035386,
+   "ye2026_sepsis": 0.11388659725586574,
+   "imperial_eskd": 0.09917593747377396
+  },
+  "train_loss": 0.1009177280664444,
+  "sec": 667.0
+ },
+ {
+  "step": 16000,
+  "inner_val_loss": 0.10388182553929914,
+  "inner_val_per_cohort": {
+   "immicro": 0.09858537589510281,
+   "ye2026_sepsis": 0.11388456117775705,
+   "imperial_eskd": 0.09917553954503754
+  },
+  "train_loss": 0.10090635707974434,
+  "sec": 678.2
+ },
+ {
+  "step": 16250,
+  "inner_val_loss": 0.10388203140548886,
+  "inner_val_per_cohort": {
+   "immicro": 0.09858604116986196,
+   "ye2026_sepsis": 0.11388345518045956,
+   "imperial_eskd": 0.09917659786614505
+  },
+  "train_loss": 0.10122319647669792,
+  "sec": 689.5
+ },
+ {
+  "step": 16500,
+  "inner_val_loss": 0.1038765674456954,
+  "inner_val_per_cohort": {
+   "immicro": 0.09857859058926503,
+   "ye2026_sepsis": 0.11387977128227551,
+   "imperial_eskd": 0.09917134046554565
+  },
+  "train_loss": 0.10110766917467118,
+  "sec": 700.8
+ },
+ {
+  "step": 16750,
+  "inner_val_loss": 0.10387495411591297,
+  "inner_val_per_cohort": {
+   "immicro": 0.09857426024973392,
+   "ye2026_sepsis": 0.1138810846540663,
+   "imperial_eskd": 0.09916951744393869
+  },
+  "train_loss": 0.10083068525791168,
+  "sec": 712.0
+ },
+ {
+  "step": 17000,
+  "inner_val_loss": 0.10387322263045845,
+  "inner_val_per_cohort": {
+   "immicro": 0.09857017577936251,
+   "ye2026_sepsis": 0.11388623383310106,
+   "imperial_eskd": 0.09916325827891176
+  },
+  "train_loss": 0.10102099803090095,
+  "sec": 723.9
+ },
+ {
+  "step": 17250,
+  "inner_val_loss": 0.10387559948796375,
+  "inner_val_per_cohort": {
+   "immicro": 0.09856981411576271,
+   "ye2026_sepsis": 0.11389378127124575,
+   "imperial_eskd": 0.0991632030768828
+  },
+  "train_loss": 0.1010699726343155,
+  "sec": 735.9
+ },
+ {
+  "step": 17500,
+  "inner_val_loss": 0.10387497263261264,
+  "inner_val_per_cohort": {
+   "immicro": 0.09856546142448981,
+   "ye2026_sepsis": 0.1138954179154502,
+   "imperial_eskd": 0.09916403855789792
+  },
+  "train_loss": 0.10070373618602753,
+  "sec": 747.9
+ },
+ {
+  "step": 17750,
+  "inner_val_loss": 0.10387528381664746,
+  "inner_val_per_cohort": {
+   "immicro": 0.09856341903408368,
+   "ye2026_sepsis": 0.11390060600307253,
+   "imperial_eskd": 0.09916182641278613
+  },
+  "train_loss": 0.10055404138565063,
+  "sec": 758.9
+ },
+ {
+  "step": 18000,
+  "inner_val_loss": 0.10387514873410918,
+  "inner_val_per_cohort": {
+   "immicro": 0.09856186372538407,
+   "ye2026_sepsis": 0.11390228031410111,
+   "imperial_eskd": 0.09916130216284232
+  },
+  "train_loss": 0.10137178257107735,
+  "sec": 769.1
+ },
+ {
+  "step": 18250,
+  "inner_val_loss": 0.10387325012485688,
+  "inner_val_per_cohort": {
+   "immicro": 0.09856060923387606,
+   "ye2026_sepsis": 0.11390373814437124,
+   "imperial_eskd": 0.09915540299632332
+  },
+  "train_loss": 0.10060748416185379,
+  "sec": 779.2
+ },
+ {
+  "step": 18500,
+  "inner_val_loss": 0.10387378704956786,
+  "inner_val_per_cohort": {
+   "immicro": 0.09855854107687871,
+   "ye2026_sepsis": 0.11390822918878661,
+   "imperial_eskd": 0.09915459088303825
+  },
+  "train_loss": 0.10098135817050934,
+  "sec": 789.5
+ },
+ {
+  "step": 18750,
+  "inner_val_loss": 0.10387095388790173,
+  "inner_val_per_cohort": {
+   "immicro": 0.0985524132847786,
+   "ye2026_sepsis": 0.11390158492657873,
+   "imperial_eskd": 0.09915886345234784
+  },
+  "train_loss": 0.10062080734968186,
+  "sec": 799.5
+ },
+ {
+  "step": 19000,
+  "inner_val_loss": 0.10386655095851782,
+  "inner_val_per_cohort": {
+   "immicro": 0.0985514981051286,
+   "ye2026_sepsis": 0.11389577637116115,
+   "imperial_eskd": 0.09915237839926373
+  },
+  "train_loss": 0.10091060662269592,
+  "sec": 809.5
+ },
+ {
+  "step": 19250,
+  "inner_val_loss": 0.10386010541295264,
+  "inner_val_per_cohort": {
+   "immicro": 0.09854780479023854,
+   "ye2026_sepsis": 0.11388369070159064,
+   "imperial_eskd": 0.0991488207470287
+  },
+  "train_loss": 0.10113474482297898,
+  "sec": 819.7
+ },
+ {
+  "step": 19500,
+  "inner_val_loss": 0.103858222422952,
+  "inner_val_per_cohort": {
+   "immicro": 0.09854646616925795,
+   "ye2026_sepsis": 0.11387925098339717,
+   "imperial_eskd": 0.09914895011620088
+  },
+  "train_loss": 0.10083182856440544,
+  "sec": 829.9
+ },
+ {
+  "step": 19750,
+  "inner_val_loss": 0.10385808244861838,
+  "inner_val_per_cohort": {
+   "immicro": 0.09854564722627401,
+   "ye2026_sepsis": 0.11387945628828472,
+   "imperial_eskd": 0.0991491438312964
+  },
+  "train_loss": 0.10087751919031143,
+  "sec": 840.1
+ },
+ {
+  "step": 20000,
+  "inner_val_loss": 0.10385796812171687,
+  "inner_val_per_cohort": {
+   "immicro": 0.09854286226133506,
+   "ye2026_sepsis": 0.11387737136748102,
+   "imperial_eskd": 0.09915367073633453
+  },
+  "train_loss": 0.10092322719097137,
+  "sec": 850.3
+ }
+]

paper-releases/pbisc-diffusion/updates/20261002T124820Z-crosscohort-liu-default/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/ho_ye2026_sepsis/K1000/lora16/history.json

@@ -0,0 +1,811 @@
+[
+ {
+  "step": 0,
+  "inner_val_loss": 0.10749047702517023,
+  "inner_val_per_cohort": {
+   "immicro": 0.10636159498244524,
+   "imperial_eskd": 0.1086193590678952
+  },
+  "train_loss": null
+ },
+ {
+  "step": 250,
+  "inner_val_loss": 0.10678054666090192,
+  "inner_val_per_cohort": {
+   "immicro": 0.10572114524741967,
+   "imperial_eskd": 0.10783994807438417
+  },
+  "train_loss": 0.1019420112669468,
+  "sec": 9.8
+ },
+ {
+  "step": 500,
+  "inner_val_loss": 0.10508193236521698,
+  "inner_val_per_cohort": {
+   "immicro": 0.10416045971214771,
+   "imperial_eskd": 0.10600340501828627
+  },
+  "train_loss": 0.09891484579443932,
+  "sec": 19.5
+ },
+ {
+  "step": 750,
+  "inner_val_loss": 0.10333118992716525,
+  "inner_val_per_cohort": {
+   "immicro": 0.10271468789627154,
+   "imperial_eskd": 0.10394769195805896
+  },
+  "train_loss": 0.09920224583148957,
+  "sec": 29.1
+ },
+ {
+  "step": 1000,
+  "inner_val_loss": 0.10190928865675673,
+  "inner_val_per_cohort": {
+   "immicro": 0.10160360050698121,
+   "imperial_eskd": 0.10221497680653226
+  },
+  "train_loss": 0.09845216441154479,
+  "sec": 38.6
+ },
+ {
+  "step": 1250,
+  "inner_val_loss": 0.1009293593409838,
+  "inner_val_per_cohort": {
+   "immicro": 0.10078995240231355,
+   "imperial_eskd": 0.10106876627965407
+  },
+  "train_loss": 0.09865968212485314,
+  "sec": 48.2
+ },
+ {
+  "step": 1500,
+  "inner_val_loss": 0.10026972323204532,
+  "inner_val_per_cohort": {
+   "immicro": 0.10019020612041156,
+   "imperial_eskd": 0.10034924034367908
+  },
+  "train_loss": 0.09840019962191582,
+  "sec": 57.7
+ },
+ {
+  "step": 1750,
+  "inner_val_loss": 0.09983539321657384,
+  "inner_val_per_cohort": {
+   "immicro": 0.09977252160509427,
+   "imperial_eskd": 0.09989826482805339
+  },
+  "train_loss": 0.09859413549304008,
+  "sec": 67.3
+ },
+ {
+  "step": 2000,
+  "inner_val_loss": 0.09954774547650506,
+  "inner_val_per_cohort": {
+   "immicro": 0.09947476567079623,
+   "imperial_eskd": 0.0996207252822139
+  },
+  "train_loss": 0.09830411371588707,
+  "sec": 77.0
+ },
+ {
+  "step": 2250,
+  "inner_val_loss": 0.09935781812633981,
+  "inner_val_per_cohort": {
+   "immicro": 0.09927116800099611,
+   "imperial_eskd": 0.0994444682516835
+  },
+  "train_loss": 0.09818677908182144,
+  "sec": 86.6
+ },
+ {
+  "step": 2500,
+  "inner_val_loss": 0.09922661523645124,
+  "inner_val_per_cohort": {
+   "immicro": 0.09912365457663934,
+   "imperial_eskd": 0.09932957589626312
+  },
+  "train_loss": 0.09835423734784127,
+  "sec": 96.1
+ },
+ {
+  "step": 2750,
+  "inner_val_loss": 0.09913526722371127,
+  "inner_val_per_cohort": {
+   "immicro": 0.09901850080738465,
+   "imperial_eskd": 0.09925203364003789
+  },
+  "train_loss": 0.09832403045892715,
+  "sec": 105.9
+ },
+ {
+  "step": 3000,
+  "inner_val_loss": 0.09906960954663879,
+  "inner_val_per_cohort": {
+   "immicro": 0.0989412742977341,
+   "imperial_eskd": 0.0991979447955435
+  },
+  "train_loss": 0.09831100672483445,
+  "sec": 115.8
+ },
+ {
+  "step": 3250,
+  "inner_val_loss": 0.0990207169035619,
+  "inner_val_per_cohort": {
+   "immicro": 0.09888497181236744,
+   "imperial_eskd": 0.09915646199475635
+  },
+  "train_loss": 0.09831550958752631,
+  "sec": 125.4
+ },
+ {
+  "step": 3500,
+  "inner_val_loss": 0.098986394491724,
+  "inner_val_per_cohort": {
+   "immicro": 0.09884218592196703,
+   "imperial_eskd": 0.09913060306148096
+  },
+  "train_loss": 0.09860761508345604,
+  "sec": 134.9
+ },
+ {
+  "step": 3750,
+  "inner_val_loss": 0.09895614526149901,
+  "inner_val_per_cohort": {
+   "immicro": 0.0988085176795721,
+   "imperial_eskd": 0.09910377284342592
+  },
+  "train_loss": 0.09823365226387977,
+  "sec": 144.6
+ },
+ {
+  "step": 4000,
+  "inner_val_loss": 0.0989298088984056,
+  "inner_val_per_cohort": {
+   "immicro": 0.09877944365143776,
+   "imperial_eskd": 0.09908017414537343
+  },
+  "train_loss": 0.09817964181303979,
+  "sec": 154.9
+ },
+ {
+  "step": 4250,
+  "inner_val_loss": 0.0989081930284473,
+  "inner_val_per_cohort": {
+   "immicro": 0.09875684697180986,
+   "imperial_eskd": 0.09905953908508475
+  },
+  "train_loss": 0.09832875180244446,
+  "sec": 166.0
+ },
+ {
+  "step": 4500,
+  "inner_val_loss": 0.09888655959035864,
+  "inner_val_per_cohort": {
+   "immicro": 0.09873440582305193,
+   "imperial_eskd": 0.09903871335766533
+  },
+  "train_loss": 0.09787118449807167,
+  "sec": 177.3
+ },
+ {
+  "step": 4750,
+  "inner_val_loss": 0.0988679566335949,
+  "inner_val_per_cohort": {
+   "immicro": 0.09871798567473888,
+   "imperial_eskd": 0.09901792759245093
+  },
+  "train_loss": 0.09825617763400078,
+  "sec": 188.6
+ },
+ {
+  "step": 5000,
+  "inner_val_loss": 0.09885247511732759,
+  "inner_val_per_cohort": {
+   "immicro": 0.09870308948059876,
+   "imperial_eskd": 0.09900186075405641
+  },
+  "train_loss": 0.0981398734152317,
+  "sec": 199.7
+ },
+ {
+  "step": 5250,
+  "inner_val_loss": 0.09883573191296874,
+  "inner_val_per_cohort": {
+   "immicro": 0.09868883031109969,
+   "imperial_eskd": 0.09898263351483778
+  },
+  "train_loss": 0.09811442610621453,
+  "sec": 211.0
+ },
+ {
+  "step": 5500,
+  "inner_val_loss": 0.098816249942915,
+  "inner_val_per_cohort": {
+   "immicro": 0.09867330081760883,
+   "imperial_eskd": 0.09895919906822118
+  },
+  "train_loss": 0.09809536680579185,
+  "sec": 222.5
+ },
+ {
+  "step": 5750,
+  "inner_val_loss": 0.09880291899158196,
+  "inner_val_per_cohort": {
+   "immicro": 0.09866241551935673,
+   "imperial_eskd": 0.09894342246380719
+  },
+  "train_loss": 0.09800233322381974,
+  "sec": 233.8
+ },
+ {
+  "step": 6000,
+  "inner_val_loss": 0.09878872837306875,
+  "inner_val_per_cohort": {
+   "immicro": 0.09865410377581914,
+   "imperial_eskd": 0.09892335297031836
+  },
+  "train_loss": 0.09791210290789604,
+  "sec": 245.0
+ },
+ {
+  "step": 6250,
+  "inner_val_loss": 0.09877511825073848,
+  "inner_val_per_cohort": {
+   "immicro": 0.09864545986056328,
+   "imperial_eskd": 0.0989047766409137
+  },
+  "train_loss": 0.09816151303052902,
+  "sec": 256.2
+ },
+ {
+  "step": 6500,
+  "inner_val_loss": 0.09876387406874335,
+  "inner_val_per_cohort": {
+   "immicro": 0.09863530999670427,
+   "imperial_eskd": 0.09889243814078244
+  },
+  "train_loss": 0.09809758257865905,
+  "sec": 267.5
+ },
+ {
+  "step": 6750,
+  "inner_val_loss": 0.09875164031417984,
+  "inner_val_per_cohort": {
+   "immicro": 0.09862499249478181,
+   "imperial_eskd": 0.09887828813357787
+  },
+  "train_loss": 0.09801262199878692,
+  "sec": 278.8
+ },
+ {
+  "step": 7000,
+  "inner_val_loss": 0.09874371869602439,
+  "inner_val_per_cohort": {
+   "immicro": 0.09861931422104438,
+   "imperial_eskd": 0.09886812317100438
+  },
+  "train_loss": 0.09796765160560608,
+  "sec": 290.1
+ },
+ {
+  "step": 7250,
+  "inner_val_loss": 0.09873696326306372,
+  "inner_val_per_cohort": {
+   "immicro": 0.09861293186744054,
+   "imperial_eskd": 0.0988609946586869
+  },
+  "train_loss": 0.09798777890205383,
+  "sec": 301.2
+ },
+ {
+  "step": 7500,
+  "inner_val_loss": 0.09872860524972732,
+  "inner_val_per_cohort": {
+   "immicro": 0.09860443789511919,
+   "imperial_eskd": 0.09885277260433543
+  },
+  "train_loss": 0.09798805320262909,
+  "sec": 312.4
+ },
+ {
+  "step": 7750,
+  "inner_val_loss": 0.09872060280406114,
+  "inner_val_per_cohort": {
+   "immicro": 0.09859480087955792,
+   "imperial_eskd": 0.09884640472856435
+  },
+  "train_loss": 0.09805942931771279,
+  "sec": 323.7
+ },
+ {
+  "step": 8000,
+  "inner_val_loss": 0.0987148635710279,
+  "inner_val_per_cohort": {
+   "immicro": 0.09858758685489495,
+   "imperial_eskd": 0.09884214028716087
+  },
+  "train_loss": 0.09770857208967208,
+  "sec": 335.1
+ },
+ {
+  "step": 8250,
+  "inner_val_loss": 0.09870804772909844,
+  "inner_val_per_cohort": {
+   "immicro": 0.09858023126920064,
+   "imperial_eskd": 0.09883586418899623
+  },
+  "train_loss": 0.097986447930336,
+  "sec": 346.5
+ },
+ {
+  "step": 8500,
+  "inner_val_loss": 0.09870035061612725,
+  "inner_val_per_cohort": {
+   "immicro": 0.09857213962823153,
+   "imperial_eskd": 0.09882856160402298
+  },
+  "train_loss": 0.09752991431951523,
+  "sec": 357.8
+ },
+ {
+  "step": 8750,
+  "inner_val_loss": 0.09869719961319456,
+  "inner_val_per_cohort": {
+   "immicro": 0.09856706019490957,
+   "imperial_eskd": 0.09882733903147957
+  },
+  "train_loss": 0.0977996079325676,
+  "sec": 369.1
+ },
+ {
+  "step": 9000,
+  "inner_val_loss": 0.09869510441962065,
+  "inner_val_per_cohort": {
+   "immicro": 0.09856494174649318,
+   "imperial_eskd": 0.09882526709274812
+  },
+  "train_loss": 0.09817633509635926,
+  "sec": 380.4
+ },
+ {
+  "step": 9250,
+  "inner_val_loss": 0.09868936936342806,
+  "inner_val_per_cohort": {
+   "immicro": 0.09855631707857053,
+   "imperial_eskd": 0.0988224216482856
+  },
+  "train_loss": 0.09797345381975174,
+  "sec": 391.6
+ },
+ {
+  "step": 9500,
+  "inner_val_loss": 0.098683143951789,
+  "inner_val_per_cohort": {
+   "immicro": 0.09854931601633628,
+   "imperial_eskd": 0.09881697188724171
+  },
+  "train_loss": 0.09798284184932708,
+  "sec": 402.9
+ },
+ {
+  "step": 9750,
+  "inner_val_loss": 0.09867739541964098,
+  "inner_val_per_cohort": {
+   "immicro": 0.09854226931929588,
+   "imperial_eskd": 0.09881252151998607
+  },
+  "train_loss": 0.09764001899957657,
+  "sec": 414.1
+ },
+ {
+  "step": 10000,
+  "inner_val_loss": 0.09866959109168613,
+  "inner_val_per_cohort": {
+   "immicro": 0.0985353182380398,
+   "imperial_eskd": 0.09880386394533244
+  },
+  "train_loss": 0.09782706189155578,
+  "sec": 425.6
+ },
+ {
+  "step": 10250,
+  "inner_val_loss": 0.09866530403043285,
+  "inner_val_per_cohort": {
+   "immicro": 0.09852890173594157,
+   "imperial_eskd": 0.09880170632492412
+  },
+  "train_loss": 0.0979362106025219,
+  "sec": 436.9
+ },
+ {
+  "step": 10500,
+  "inner_val_loss": 0.09866172497188955,
+  "inner_val_per_cohort": {
+   "immicro": 0.09852427275230487,
+   "imperial_eskd": 0.09879917719147423
+  },
+  "train_loss": 0.09795433682203293,
+  "sec": 448.5
+ },
+ {
+  "step": 10750,
+  "inner_val_loss": 0.09865660985197985,
+  "inner_val_per_cohort": {
+   "immicro": 0.09851801705857118,
+   "imperial_eskd": 0.09879520264538852
+  },
+  "train_loss": 0.09772800397872924,
+  "sec": 459.9
+ },
+ {
+  "step": 11000,
+  "inner_val_loss": 0.09865404669703408,
+  "inner_val_per_cohort": {
+   "immicro": 0.09851261507719755,
+   "imperial_eskd": 0.0987954783168706
+  },
+  "train_loss": 0.09807586640119553,
+  "sec": 471.3
+ },
+ {
+  "step": 11250,
+  "inner_val_loss": 0.09865122244721561,
+  "inner_val_per_cohort": {
+   "immicro": 0.09850989499439795,
+   "imperial_eskd": 0.09879254990003326
+  },
+  "train_loss": 0.09782252213358879,
+  "sec": 482.6
+ },
+ {
+  "step": 11500,
+  "inner_val_loss": 0.09864849731269659,
+  "inner_val_per_cohort": {
+   "immicro": 0.09850500182559092,
+   "imperial_eskd": 0.09879199279980226
+  },
+  "train_loss": 0.0980256376862526,
+  "sec": 493.8
+ },
+ {
+  "step": 11750,
+  "inner_val_loss": 0.0986426070360749,
+  "inner_val_per_cohort": {
+   "immicro": 0.09850061902155478,
+   "imperial_eskd": 0.09878459505059502
+  },
+  "train_loss": 0.09766613695025445,
+  "sec": 505.2
+ },
+ {
+  "step": 12000,
+  "inner_val_loss": 0.09863867331296206,
+  "inner_val_per_cohort": {
+   "immicro": 0.09849481843411922,
+   "imperial_eskd": 0.09878252819180489
+  },
+  "train_loss": 0.09751020780205727,
+  "sec": 516.4
+ },
+ {
+  "step": 12250,
+  "inner_val_loss": 0.09863617783412337,
+  "inner_val_per_cohort": {
+   "immicro": 0.09849114995449781,
+   "imperial_eskd": 0.09878120571374893
+  },
+  "train_loss": 0.09775541913509368,
+  "sec": 527.6
+ },
+ {
+  "step": 12500,
+  "inner_val_loss": 0.0986296319205201,
+  "inner_val_per_cohort": {
+   "immicro": 0.09848595472673576,
+   "imperial_eskd": 0.09877330911430446
+  },
+  "train_loss": 0.09812748098373412,
+  "sec": 538.8
+ },
+ {
+  "step": 12750,
+  "inner_val_loss": 0.09862829042090611,
+  "inner_val_per_cohort": {
+   "immicro": 0.09848256595432758,
+   "imperial_eskd": 0.09877401488748463
+  },
+  "train_loss": 0.09784625989198685,
+  "sec": 550.0
+ },
+ {
+  "step": 13000,
+  "inner_val_loss": 0.09862762398904923,
+  "inner_val_per_cohort": {
+   "immicro": 0.09847917097310226,
+   "imperial_eskd": 0.09877607700499622
+  },
+  "train_loss": 0.09798663240671159,
+  "sec": 561.3
+ },
+ {
+  "step": 13250,
+  "inner_val_loss": 0.09862424115735022,
+  "inner_val_per_cohort": {
+   "immicro": 0.09847541805356741,
+   "imperial_eskd": 0.09877306426113303
+  },
+  "train_loss": 0.09792979824543,
+  "sec": 572.6
+ },
+ {
+  "step": 13500,
+  "inner_val_loss": 0.09862181188708002,
+  "inner_val_per_cohort": {
+   "immicro": 0.09847208485007286,
+   "imperial_eskd": 0.09877153892408717
+  },
+  "train_loss": 0.09784729206562043,
+  "sec": 583.9
+ },
+ {
+  "step": 13750,
+  "inner_val_loss": 0.09861877145753666,
+  "inner_val_per_cohort": {
+   "immicro": 0.0984706562012434,
+   "imperial_eskd": 0.09876688671382991
+  },
+  "train_loss": 0.09798241600394249,
+  "sec": 595.2
+ },
+ {
+  "step": 14000,
+  "inner_val_loss": 0.09861674549227412,
+  "inner_val_per_cohort": {
+   "immicro": 0.09846555069088936,
+   "imperial_eskd": 0.09876794029365886
+  },
+  "train_loss": 0.09798856177926063,
+  "sec": 606.4
+ },
+ {
+  "step": 14250,
+  "inner_val_loss": 0.09861567225589445,
+  "inner_val_per_cohort": {
+   "immicro": 0.098463151914378,
+   "imperial_eskd": 0.09876819259741089
+  },
+  "train_loss": 0.09752175915241242,
+  "sec": 617.6
+ },
+ {
+  "step": 14500,
+  "inner_val_loss": 0.09861540283556237,
+  "inner_val_per_cohort": {
+   "immicro": 0.09846318947772185,
+   "imperial_eskd": 0.0987676161934029
+  },
+  "train_loss": 0.09796379837393761,
+  "sec": 628.9
+ },
+ {
+  "step": 14750,
+  "inner_val_loss": 0.09861241079246005,
+  "inner_val_per_cohort": {
+   "immicro": 0.09845858408759038,
+   "imperial_eskd": 0.09876623749732971
+  },
+  "train_loss": 0.09785736951231956,
+  "sec": 640.1
+ },
+ {
+  "step": 15000,
+  "inner_val_loss": 0.09861254747110335,
+  "inner_val_per_cohort": {
+   "immicro": 0.0984606584534049,
+   "imperial_eskd": 0.09876443648880179
+  },
+  "train_loss": 0.09791886976361275,
+  "sec": 651.5
+ },
+ {
+  "step": 15250,
+  "inner_val_loss": 0.09861089687117122,
+  "inner_val_per_cohort": {
+   "immicro": 0.09845557622611523,
+   "imperial_eskd": 0.0987662175162272
+  },
+  "train_loss": 0.09787165305018425,
+  "sec": 662.9
+ },
+ {
+  "step": 15500,
+  "inner_val_loss": 0.0986094206825576,
+  "inner_val_per_cohort": {
+   "immicro": 0.09845172334462404,
+   "imperial_eskd": 0.09876711802049117
+  },
+  "train_loss": 0.09808653998374939,
+  "sec": 674.1
+ },
+ {
+  "step": 15750,
+  "inner_val_loss": 0.09860741193265174,
+  "inner_val_per_cohort": {
+   "immicro": 0.09844884307434161,
+   "imperial_eskd": 0.09876598079096187
+  },
+  "train_loss": 0.09769263461232186,
+  "sec": 685.5
+ },
+ {
+  "step": 16000,
+  "inner_val_loss": 0.0986022989152733,
+  "inner_val_per_cohort": {
+   "immicro": 0.09844273484001557,
+   "imperial_eskd": 0.09876186299053105
+  },
+  "train_loss": 0.09757432132959366,
+  "sec": 696.3
+ },
+ {
+  "step": 16250,
+  "inner_val_loss": 0.09860275994816964,
+  "inner_val_per_cohort": {
+   "immicro": 0.09844083990901709,
+   "imperial_eskd": 0.0987646799873222
+  },
+  "train_loss": 0.09793048390746116,
+  "sec": 706.0
+ },
+ {
+  "step": 16500,
+  "inner_val_loss": 0.09859947429857019,
+  "inner_val_per_cohort": {
+   "immicro": 0.09843599951515596,
+   "imperial_eskd": 0.09876294908198444
+  },
+  "train_loss": 0.09805796450376511,
+  "sec": 715.7
+ },
+ {
+  "step": 16750,
+  "inner_val_loss": 0.09859637015809616,
+  "inner_val_per_cohort": {
+   "immicro": 0.09843359949688117,
+   "imperial_eskd": 0.09875914081931114
+  },
+  "train_loss": 0.09783114829659462,
+  "sec": 726.6
+ },
+ {
+  "step": 17000,
+  "inner_val_loss": 0.09859549018526167,
+  "inner_val_per_cohort": {
+   "immicro": 0.09843078224609296,
+   "imperial_eskd": 0.09876019812443039
+  },
+  "train_loss": 0.09800776880979538,
+  "sec": 737.9
+ },
+ {
+  "step": 17250,
+  "inner_val_loss": 0.09859495463244844,
+  "inner_val_per_cohort": {
+   "immicro": 0.098427368948857,
+   "imperial_eskd": 0.09876254031603987
+  },
+  "train_loss": 0.0978686043024063,
+  "sec": 749.3
+ },
+ {
+  "step": 17500,
+  "inner_val_loss": 0.09858872119167988,
+  "inner_val_per_cohort": {
+   "immicro": 0.09842339809983969,
+   "imperial_eskd": 0.0987540442835201
+  },
+  "train_loss": 0.09765802243351937,
+  "sec": 760.9
+ },
+ {
+  "step": 17750,
+  "inner_val_loss": 0.0985848055870244,
+  "inner_val_per_cohort": {
+   "immicro": 0.09842176238695781,
+   "imperial_eskd": 0.09874784878709099
+  },
+  "train_loss": 0.09761114963889123,
+  "sec": 771.1
+ },
+ {
+  "step": 18000,
+  "inner_val_loss": 0.09858433344881191,
+  "inner_val_per_cohort": {
+   "immicro": 0.09842177449415128,
+   "imperial_eskd": 0.09874689240347255
+  },
+  "train_loss": 0.09808740887045861,
+  "sec": 780.9
+ },
+ {
+  "step": 18250,
+  "inner_val_loss": 0.09857928635778301,
+  "inner_val_per_cohort": {
+   "immicro": 0.09841703965018193,
+   "imperial_eskd": 0.09874153306538408
+  },
+  "train_loss": 0.0976393857896328,
+  "sec": 792.2
+ },
+ {
+  "step": 18500,
+  "inner_val_loss": 0.09857626939474633,
+  "inner_val_per_cohort": {
+   "immicro": 0.09841225234170754,
+   "imperial_eskd": 0.09874028644778511
+  },
+  "train_loss": 0.0975072360932827,
+  "sec": 803.8
+ },
+ {
+  "step": 18750,
+  "inner_val_loss": 0.09857558801939542,
+  "inner_val_per_cohort": {
+   "immicro": 0.09840986412018538,
+   "imperial_eskd": 0.09874131191860545
+  },
+  "train_loss": 0.09767750278115273,
+  "sec": 815.0
+ },
+ {
+  "step": 19000,
+  "inner_val_loss": 0.09857641743270285,
+  "inner_val_per_cohort": {
+   "immicro": 0.0984087757145365,
+   "imperial_eskd": 0.0987440591508692
+  },
+  "train_loss": 0.0976138351559639,
+  "sec": 826.2
+ },
+ {
+  "step": 19250,
+  "inner_val_loss": 0.09857527738098393,
+  "inner_val_per_cohort": {
+   "immicro": 0.09840475860983133,
+   "imperial_eskd": 0.09874579615213654
+  },
+  "train_loss": 0.09773308408260345,
+  "sec": 837.5
+ },
+ {
+  "step": 19500,
+  "inner_val_loss": 0.09857549838665308,
+  "inner_val_per_cohort": {
+   "immicro": 0.09840110720445712,
+   "imperial_eskd": 0.09874988956884904
+  },
+  "train_loss": 0.09774972891807557,
+  "sec": 848.9
+ },
+ {
+  "step": 19750,
+  "inner_val_loss": 0.0985743707807904,
+  "inner_val_per_cohort": {
+   "immicro": 0.09839798975735903,
+   "imperial_eskd": 0.09875075180422176
+  },
+  "train_loss": 0.09756702634692192,
+  "sec": 860.4
+ },
+ {
+  "step": 20000,
+  "inner_val_loss": 0.09857459987203279,
+  "inner_val_per_cohort": {
+   "immicro": 0.09839341913660367,
+   "imperial_eskd": 0.09875578060746193
+  },
+  "train_loss": 0.09776072540879249,
+  "sec": 871.8
+ }
+]