diff --git a/paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/RUN.json b/paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/RUN.json
new file mode 100644
index 0000000000000000000000000000000000000000..84fded3992c18d43ba4dbce3ba3fa4e660af1ae1
--- /dev/null
+++ b/paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/RUN.json
@@ -0,0 +1,556 @@
+{
+ "schema": "pbisc_sdd.ft_crosscohort.run.v1",
+ "source_checkpoint": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/runs/f01_6k_1x_seed20260627/checkpoint_epoch_016.pt",
+ "source_checkpoint_sha256": "8c7cd0e83e2bc19fe03dfcbc084919a1e2f61be8cc8bc12f4caf543dfda91633",
+ "split": "hoL_ye2026_sepsis",
+ "split_file": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/splits/hoL_ye2026_sepsis.json",
+ "split_sha256": "39cad7b25faf6c15cd65f0512977b9e526353cfb085dc2c45d18f8edc56db426",
+ "held_out": "ye2026_sepsis",
+ "recipe_k": 1000,
+ "balance": "cohort",
+ "mode": "full",
+ "trainable": {
+  "mode": "full",
+  "n_trainable": 54185184,
+  "n_total_base": 54185184,
+  "trainable_tensors": 64
+ },
+ "lr": 5e-05,
+ "warmup": 200,
+ "batch_size": 256,
+ "ema_decay": 0.999,
+ "max_steps": 200000,
+ "steps_run": 108500,
+ "eval_every": 250,
+ "patience": 8,
+ "selected_step": 106500,
+ "selected_inner_val_loss": 0.09843902954963422,
+ "pretrained_inner_val_loss": 0.10827601102716995,
+ "cohorts": [
+  {
+   "cohort": "immicro",
+   "condition_npz": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/data/immicro/conditions_raw_K1000.npz",
+   "condition_npz_sha256": "2be5717d19e86bdf38e4838d4bd2886a211ba78bfbfe749c458305a3e0ac2dcc",
+   "train_rows": [
+    "HS1",
+    "HS2",
+    "HS3",
+    "HS6",
+    "HS7",
+    "HS8",
+    "HS9",
+    "HS10",
+    "HS12",
+    "HS13",
+    "HS14",
+    "HS15",
+    "HS16",
+    "HS17",
+    "HS18",
+    "HS19",
+    "HS20",
+    "HS21",
+    "HS23",
+    "HS24",
+    "HS25",
+    "HS28",
+    "HS29",
+    "HS30",
+    "HS32",
+    "HS34",
+    "HS35",
+    "HS36",
+    "HS37",
+    "HS38",
+    "HS39",
+    "HS40",
+    "HS41",
+    "HS42",
+    "HS43",
+    "HS44",
+    "HS45",
+    "HS47",
+    "HS48",
+    "HS49",
+    "HS50",
+    "HS51",
+    "HS52",
+    "HS53",
+    "HS54",
+    "HS56",
+    "HS57",
+    "HS58",
+    "HS59",
+    "HS61",
+    "HS63",
+    "HS64",
+    "HS65",
+    "HS66",
+    "HS67",
+    "HS68",
+    "HS69",
+    "HS70",
+    "HS71",
+    "HS72",
+    "HS73",
+    "HS74",
+    "HS75",
+    "HS76",
+    "HS77",
+    "HS78",
+    "HS79",
+    "HS80",
+    "HS81",
+    "HS82",
+    "HS83",
+    "HS84",
+    "HS85",
+    "HS86",
+    "HS88",
+    "HS89",
+    "HS90",
+    "HS91",
+    "HS92",
+    "HS93",
+    "HS94",
+    "HS95",
+    "HS96",
+    "HS97",
+    "HS98",
+    "HS100",
+    "HS101",
+    "HS102",
+    "HS103",
+    "HS104",
+    "HS105",
+    "HS106",
+    "HS107",
+    "HS110",
+    "HS111"
+   ],
+   "inner_val_rows": [
+    "HS4",
+    "HS11",
+    "HS26",
+    "HS27",
+    "HS31",
+    "HS46",
+    "HS55",
+    "HS60",
+    "HS62",
+    "HS87",
+    "HS108",
+    "HS109"
+   ]
+  },
+  {
+   "cohort": "imperial_eskd",
+   "condition_npz": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/data/imperial_eskd/conditions_raw_K1000.npz",
+   "condition_npz_sha256": "41a62291e98e367d05108a0a95a859ef9ca1997f1fda4656cba58eeefbe56442",
+   "train_rows": [
+    "C20_1004",
+    "C20_1404",
+    "C21_1004",
+    "C21_1704",
+    "C21_2504",
+    "C23_2204",
+    "C31_1304",
+    "C31_2004",
+    "C31_2704",
+    "C33_1401",
+    "C33_1601",
+    "C33_1901",
+    "C34_1404",
+    "C34_2104",
+    "C34_2504",
+    "C36_1504",
+    "C36_1704",
+    "C36_2104",
+    "C40_1504",
+    "C40_2204",
+    "C40_0105",
+    "C42_1504",
+    "C42_2204",
+    "C42_2704",
+    "C58_1604",
+    "C58_2501",
+    "C58_2901",
+    "C58_0102",
+    "C58_0402",
+    "C58_1003",
+    "C60_2504",
+    "C65_1704",
+    "C65_2104",
+    "C65_2504",
+    "C69_1804",
+    "C82_2504",
+    "C85_2004",
+    "C93_2104",
+    "C93_2704",
+    "C93_0405",
+    "C101_0601",
+    "C101_1101",
+    "C101_1301",
+    "C101_1801",
+    "C101_2001",
+    "C101_2502",
+    "C103_2204",
+    "C104_2204",
+    "C106_2204",
+    "C108_2204",
+    "C108_1501",
+    "C108_1801",
+    "C108_2201",
+    "C113_0605",
+    "C123_2404",
+    "C123_2904",
+    "C123_0505",
+    "C124_2404",
+    "C124_2704",
+    "C124_0105",
+    "C126_2504",
+    "C126_3004",
+    "C126_0605",
+    "C127_2504",
+    "C127_0405",
+    "C127_1105",
+    "C132_2804",
+    "C132_0205",
+    "C132_1105",
+    "C137_3004",
+    "C137_2502",
+    "C138_3004",
+    "C138_2301",
+    "C138_3001",
+    "C138_0202",
+    "C138_0402",
+    "C140_3004",
+    "C140_2801",
+    "C140_3001",
+    "C140_0202",
+    "C140_0402",
+    "C140_0602",
+    "C140_0902",
+    "C140_1103",
+    "C141_2001",
+    "C141_2201",
+    "C141_3001",
+    "C141_0202",
+    "C145_3004",
+    "C145_0701",
+    "C145_1201",
+    "C145_1401",
+    "C145_1901",
+    "C145_2301",
+    "C146_3004",
+    "C146_1201",
+    "C146_1401",
+    "C146_1601",
+    "C146_1901",
+    "C146_2301",
+    "C146_2701",
+    "C146_2502",
+    "C152_0405",
+    "C161_1105",
+    "C164_1105",
+    "C167_1205",
+    "C168_1205",
+    "C168_2001",
+    "C168_2301",
+    "C168_3001",
+    "C168_0202",
+    "C168_0403",
+    "C169_1205",
+    "C169_2001",
+    "C169_2301",
+    "C169_0202",
+    "C169_0403",
+    "C170_1205",
+    "C170_2301",
+    "C170_0202",
+    "C170_0402",
+    "C170_0403",
+    "C187_2605",
+    "C187_1201",
+    "C187_1401",
+    "C187_1901",
+    "C187_2301",
+    "C187_2502",
+    "C190_2605",
+    "C190_2901",
+    "C190_0102",
+    "C190_0302",
+    "C190_0502",
+    "C190_0802",
+    "C190_1002",
+    "C190_0903",
+    "C193_2605",
+    "C195_2605",
+    "C197_2605",
+    "C212_2705",
+    "C214_2705",
+    "C237_2805",
+    "C238_2905",
+    "C240_2905",
+    "C246_2905",
+    "C251_2905"
+   ],
+   "inner_val_rows": [
+    "C63_1704",
+    "C63_2404",
+    "C63_2904",
+    "C128_2504",
+    "C128_0405",
+    "C128_1105",
+    "C142_3004",
+    "C147_3004",
+    "C147_1401",
+    "C147_1901",
+    "C234_2805"
+   ]
+  },
+  {
+   "cohort": "giroux2022",
+   "condition_npz": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/data/giroux2022/conditions_raw_K1000.npz",
+   "condition_npz_sha256": "b584655fcfa5fa4c0887d5dd83fcb1c5a0206da4afe5fefab96cba2272b1371c",
+   "train_rows": [
+    "450905_d0",
+    "450905_d14",
+    "0B943B_d0",
+    "0B943B_d14",
+    "7085CA_d0",
+    "7085CA_d7",
+    "7085CA_d14",
+    "2AD75E_d0",
+    "2AD75E_d7",
+    "2AD75E_d14",
+    "B85D75_d0",
+    "B85D75_d7",
+    "180E1A_d0",
+    "180E1A_d7",
+    "180E1A_d14",
+    "82CCF5_d0",
+    "82CCF5_d7",
+    "82CCF5_d14",
+    "BAAF62_d0",
+    "BAAF62_d7",
+    "BAAF62_d14",
+    "0BF51C_d0",
+    "0BF51C_d7",
+    "0BF51C_d14"
+   ],
+   "inner_val_rows": []
+  },
+  {
+   "cohort": "liu2021",
+   "condition_npz": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/data/liu2021/conditions_raw_K1000.npz",
+   "condition_npz_sha256": "ad3131eabcfb11b166ca37c7585c50abb11efb60e2a736fe62de604794535e0c",
+   "train_rows": [
+    "AA220014_HC",
+    "AA220534_HC",
+    "AA220907_HC",
+    "HGR0000071_T0",
+    "HGR0000074_T0",
+    "HGR0000083_T1",
+    "HGR0000089_T1",
+    "HGR0000091_T0",
+    "HGR0000093_T0",
+    "HGR0000136_T0",
+    "HGR0000137_T0",
+    "HGR0000141_T0",
+    "IM000040_HC",
+    "HGR0000042_T1",
+    "HGR0000069_T0",
+    "HGR0000069_T1",
+    "HGR0000092_T0",
+    "HGR0000094_T0",
+    "HGR0000124_T0",
+    "HGR0000144_T2",
+    "HGR0000391_T0",
+    "HGR0000395_T0",
+    "HGR0000418_T1",
+    "SHD4_HC",
+    "SHD7_HC",
+    "HGR0000051_T0",
+    "HGR0000051_T1",
+    "HGR0000051_T3",
+    "HGR0000101_T1",
+    "HGR0000101_T3",
+    "HGR0000102_T0",
+    "HGR0000102_T1",
+    "HGR0000134_T0",
+    "HGR0000135_T0",
+    "HGR0000135_T1",
+    "HGR0000392_T0",
+    "HGR0000429_T1",
+    "HGR0000429_T2",
+    "SHD3_HC",
+    "SHD5_HC"
+   ],
+   "inner_val_rows": [
+    "HGR0000098_T0",
+    "HGR0000143_T0",
+    "HGR0000115_T1",
+    "SHD1_HC"
+   ]
+  }
+ ],
+ "held_out_test_rows_never_loaded": [
+  "Abd-PS04",
+  "Abd-PS05",
+  "Abd-PS07",
+  "Abd-PS19",
+  "Abd-PS20",
+  "Abd-PS23",
+  "Abd-PS27",
+  "Abd-S102",
+  "Abd-S103",
+  "Abd-S120",
+  "Abd-S123",
+  "Abd-S156",
+  "Abd-S39",
+  "Abd-S42",
+  "Abd-S47",
+  "Abd-S51",
+  "Abd-S54",
+  "Abd-S57",
+  "Abd-S66",
+  "Abd-S79",
+  "Abd-S83",
+  "Abd-S98",
+  "HC01",
+  "HC05",
+  "HC07",
+  "HC09",
+  "HC12",
+  "HC13",
+  "HC16",
+  "IC10",
+  "IC15",
+  "IC16",
+  "IC19",
+  "IC22",
+  "IC23",
+  "PHC01",
+  "PHC02",
+  "PHC03",
+  "PHC05",
+  "PIC01",
+  "PIC02",
+  "PIC04",
+  "PIC05",
+  "Res-PS02",
+  "Res-PS03",
+  "Res-PS12",
+  "Res-PS16",
+  "Res-PS21",
+  "Res-PS22",
+  "Res-PS25",
+  "Res-PS26",
+  "Res-S10",
+  "Res-S125",
+  "Res-S127",
+  "Res-S136",
+  "Res-S139",
+  "Res-S146",
+  "Res-S149",
+  "Res-S150",
+  "Res-S164",
+  "Res-S166",
+  "Res-S174",
+  "Res-S177",
+  "Res-S184",
+  "Res-S19",
+  "Res-S26",
+  "Res-S40",
+  "Res-S77",
+  "Res-S87",
+  "Skin-S140",
+  "Skin-S142",
+  "Skin-S152",
+  "Skin-S38",
+  "Skin-S43",
+  "Skin-S46",
+  "Skin-S52",
+  "Skin-S74",
+  "Uri-S134",
+  "Uri-S147",
+  "Uri-S157",
+  "Uri-S20",
+  "Uri-S93"
+ ],
+ "max_abs_weight_change": 1.4082739353179932,
+ "n_tensors_changed": 64,
+ "seed": 20261002,
+ "wall_seconds": 5600.4,
+ "smoke": false,
+ "block": 20000,
+ "saturation_frac": 0.01,
+ "stop_reason": "early_stop",
+ "milestones": [
+  {
+   "milestone_step": 20000,
+   "selected_step": 20000,
+   "selected_inner_val_loss": 0.09932147273135306,
+   "block_improvement": 0.008954538295816897,
+   "total_improvement": 0.008954538295816897,
+   "reason": "block_end",
+   "checkpoint": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/checkpoint_ft_s20000.pt",
+   "checkpoint_sha256": "235563551146480bcf95b57d9456cd702fa24aaee2a58e987df763a37d9e09a0"
+  },
+  {
+   "milestone_step": 40000,
+   "selected_step": 40000,
+   "selected_inner_val_loss": 0.09898765926070587,
+   "block_improvement": 0.00033381347064719014,
+   "total_improvement": 0.009288351766464087,
+   "reason": "block_end",
+   "checkpoint": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/checkpoint_ft_s40000.pt",
+   "checkpoint_sha256": "be3484436f57293db0bed630a4e032fc4a8a37f86e2f3cdbd92f0d207e8e2deb"
+  },
+  {
+   "milestone_step": 60000,
+   "selected_step": 59500,
+   "selected_inner_val_loss": 0.09878567852681935,
+   "block_improvement": 0.0002019807338865115,
+   "total_improvement": 0.009490332500350598,
+   "reason": "block_end",
+   "checkpoint": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/checkpoint_ft_s60000.pt",
+   "checkpoint_sha256": "10def9a99a30b3a32b9027023d0bc8382fa779c3ae0e7df4d1d1dc96ed27e056"
+  },
+  {
+   "milestone_step": 80000,
+   "selected_step": 80000,
+   "selected_inner_val_loss": 0.0986245580092825,
+   "block_improvement": 0.00016112051753684742,
+   "total_improvement": 0.009651453017887446,
+   "reason": "block_end",
+   "checkpoint": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/checkpoint_ft_s80000.pt",
+   "checkpoint_sha256": "9621e005c021486583ee7b97cc0a77bfadf57e5457f28b476d568777efb06b09"
+  },
+  {
+   "milestone_step": 100000,
+   "selected_step": 100000,
+   "selected_inner_val_loss": 0.09848370757706539,
+   "block_improvement": 0.00014085043221712112,
+   "total_improvement": 0.009792303450104567,
+   "reason": "block_end",
+   "checkpoint": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/checkpoint_ft_s100000.pt",
+   "checkpoint_sha256": "3cde1860b41af88ef71cc29e7d7bb29371c30d6c7d12a5b7ba2765cbaa84e4e3"
+  },
+  {
+   "milestone_step": 108500,
+   "selected_step": 106500,
+   "selected_inner_val_loss": 0.09843902954963422,
+   "block_improvement": 4.4678027431160894e-05,
+   "total_improvement": 0.009836981477535728,
+   "reason": "early_stop",
+   "checkpoint": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/checkpoint_ft_s108500.pt",
+   "checkpoint_sha256": "33441f5f75cb45b7c704f2e744ec842ad1371e46206c951443fbed98c7ea6095"
+  }
+ ],
+ "written_utc": "2026-10-03T09:10:38.958036+00:00",
+ "last_milestone_max_abs_diff": 0.0,
+ "checkpoint_sha256": "9240379953f47ef926556431f5e55c7cea43d0fa5b91eafef2d488bc25030a9a",
+ "reload_forward_max_abs_diff": 0.0
+}
\ No newline at end of file
diff --git a/paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/history.json b/paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/history.json
new file mode 100644
index 0000000000000000000000000000000000000000..0aa0e48a98d103fecc48e3702bb107bd8a7f74b8
--- /dev/null
+++ b/paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/history.json
@@ -0,0 +1,4786 @@
+[
+ {
+  "step": 0,
+  "inner_val_loss": 0.10827601102716995,
+  "inner_val_per_cohort": {
+   "immicro": 0.10636159498244524,
+   "imperial_eskd": 0.1086193590678952,
+   "liu2021": 0.10984707903116941
+  },
+  "train_loss": null
+ },
+ {
+  "step": 250,
+  "inner_val_loss": 0.10649758309238788,
+  "inner_val_per_cohort": {
+   "immicro": 0.10473520494997501,
+   "imperial_eskd": 0.10662541944872249,
+   "liu2021": 0.10813212487846613
+  },
+  "train_loss": 0.1038634278178215,
+  "sec": 11.3
+ },
+ {
+  "step": 500,
+  "inner_val_loss": 0.10433993210094143,
+  "inner_val_per_cohort": {
+   "immicro": 0.10289469330261151,
+   "imperial_eskd": 0.10409227792512286,
+   "liu2021": 0.10603282507508993
+  },
+  "train_loss": 0.10126817017793656,
+  "sec": 24.6
+ },
+ {
+  "step": 750,
+  "inner_val_loss": 0.10291964080974912,
+  "inner_val_per_cohort": {
+   "immicro": 0.10175599157810211,
+   "imperial_eskd": 0.10235593332485719,
+   "liu2021": 0.10464699752628803
+  },
+  "train_loss": 0.10089824205636978,
+  "sec": 38.5
+ },
+ {
+  "step": 1000,
+  "inner_val_loss": 0.10203568997670605,
+  "inner_val_per_cohort": {
+   "immicro": 0.10101005559166272,
+   "imperial_eskd": 0.10129277340390465,
+   "liu2021": 0.10380424093455076
+  },
+  "train_loss": 0.10052747491002083,
+  "sec": 53.2
+ },
+ {
+  "step": 1250,
+  "inner_val_loss": 0.10147218785314549,
+  "inner_val_per_cohort": {
+   "immicro": 0.10049211947868268,
+   "imperial_eskd": 0.10063569484786554,
+   "liu2021": 0.10328874923288822
+  },
+  "train_loss": 0.10057301998138428,
+  "sec": 67.5
+ },
+ {
+  "step": 1500,
+  "inner_val_loss": 0.10110339066107767,
+  "inner_val_per_cohort": {
+   "immicro": 0.10011807146171729,
+   "imperial_eskd": 0.1002305034886707,
+   "liu2021": 0.10296159703284502
+  },
+  "train_loss": 0.10028809612989426,
+  "sec": 81.0
+ },
+ {
+  "step": 1750,
+  "inner_val_loss": 0.10084495673689879,
+  "inner_val_per_cohort": {
+   "immicro": 0.09983316902071238,
+   "imperial_eskd": 0.09995830092917789,
+   "liu2021": 0.10274340026080608
+  },
+  "train_loss": 0.1003752883374691,
+  "sec": 94.9
+ },
+ {
+  "step": 2000,
+  "inner_val_loss": 0.10066142761045033,
+  "inner_val_per_cohort": {
+   "immicro": 0.09961860409627359,
+   "imperial_eskd": 0.09976108372211456,
+   "liu2021": 0.10260459501296282
+  },
+  "train_loss": 0.10009883990883828,
+  "sec": 108.6
+ },
+ {
+  "step": 2250,
+  "inner_val_loss": 0.100522446643674,
+  "inner_val_per_cohort": {
+   "immicro": 0.09944950137287378,
+   "imperial_eskd": 0.09961403872479092,
+   "liu2021": 0.10250379983335733
+  },
+  "train_loss": 0.10020983132719993,
+  "sec": 121.0
+ },
+ {
+  "step": 2500,
+  "inner_val_loss": 0.1004142706856282,
+  "inner_val_per_cohort": {
+   "immicro": 0.09931121362994115,
+   "imperial_eskd": 0.09950534796172922,
+   "liu2021": 0.10242625046521425
+  },
+  "train_loss": 0.10044015708565712,
+  "sec": 133.4
+ },
+ {
+  "step": 2750,
+  "inner_val_loss": 0.10032929791457423,
+  "inner_val_per_cohort": {
+   "immicro": 0.09920035457859437,
+   "imperial_eskd": 0.09941796992312778,
+   "liu2021": 0.10236956924200058
+  },
+  "train_loss": 0.10003156259655953,
+  "sec": 146.2
+ },
+ {
+  "step": 3000,
+  "inner_val_loss": 0.10025799974347606,
+  "inner_val_per_cohort": {
+   "immicro": 0.09911093395203352,
+   "imperial_eskd": 0.099340981380506,
+   "liu2021": 0.10232208389788866
+  },
+  "train_loss": 0.09998358249664306,
+  "sec": 159.2
+ },
+ {
+  "step": 3250,
+  "inner_val_loss": 0.10020050899398447,
+  "inner_val_per_cohort": {
+   "immicro": 0.09903726385285457,
+   "imperial_eskd": 0.09928012876348062,
+   "liu2021": 0.10228413436561823
+  },
+  "train_loss": 0.10017218118906021,
+  "sec": 172.0
+ },
+ {
+  "step": 3500,
+  "inner_val_loss": 0.10014912837909327,
+  "inner_val_per_cohort": {
+   "immicro": 0.09896906868865092,
+   "imperial_eskd": 0.09923061355948448,
+   "liu2021": 0.10224770288914442
+  },
+  "train_loss": 0.10011683976650237,
+  "sec": 184.9
+ },
+ {
+  "step": 3750,
+  "inner_val_loss": 0.10010675225181109,
+  "inner_val_per_cohort": {
+   "immicro": 0.09890971519052982,
+   "imperial_eskd": 0.09918696738102219,
+   "liu2021": 0.10222357418388128
+  },
+  "train_loss": 0.09997327256202698,
+  "sec": 197.4
+ },
+ {
+  "step": 4000,
+  "inner_val_loss": 0.10006625304995764,
+  "inner_val_per_cohort": {
+   "immicro": 0.09885848468790452,
+   "imperial_eskd": 0.09914540736512704,
+   "liu2021": 0.10219486709684134
+  },
+  "train_loss": 0.09990688416361809,
+  "sec": 210.1
+ },
+ {
+  "step": 4250,
+  "inner_val_loss": 0.10003431069881025,
+  "inner_val_per_cohort": {
+   "immicro": 0.09881360953052838,
+   "imperial_eskd": 0.09911355410109866,
+   "liu2021": 0.1021757684648037
+  },
+  "train_loss": 0.09992066559195519,
+  "sec": 222.8
+ },
+ {
+  "step": 4500,
+  "inner_val_loss": 0.10000245932564893,
+  "inner_val_per_cohort": {
+   "immicro": 0.09877357197304566,
+   "imperial_eskd": 0.09907543185082349,
+   "liu2021": 0.1021583741530776
+  },
+  "train_loss": 0.09985163056850434,
+  "sec": 235.3
+ },
+ {
+  "step": 4750,
+  "inner_val_loss": 0.0999745854054286,
+  "inner_val_per_cohort": {
+   "immicro": 0.09873724573602279,
+   "imperial_eskd": 0.0990458025851033,
+   "liu2021": 0.10214070789515972
+  },
+  "train_loss": 0.09972784650325775,
+  "sec": 247.6
+ },
+ {
+  "step": 5000,
+  "inner_val_loss": 0.09994827604128255,
+  "inner_val_per_cohort": {
+   "immicro": 0.09870154472688834,
+   "imperial_eskd": 0.09901832789182663,
+   "liu2021": 0.10212495550513268
+  },
+  "train_loss": 0.09989358052611351,
+  "sec": 260.1
+ },
+ {
+  "step": 5250,
+  "inner_val_loss": 0.09992539316340528,
+  "inner_val_per_cohort": {
+   "immicro": 0.09867107464621465,
+   "imperial_eskd": 0.09899277341636745,
+   "liu2021": 0.10211233142763376
+  },
+  "train_loss": 0.0998383513391018,
+  "sec": 272.8
+ },
+ {
+  "step": 5500,
+  "inner_val_loss": 0.09990168620615897,
+  "inner_val_per_cohort": {
+   "immicro": 0.0986412229637305,
+   "imperial_eskd": 0.09897058795798909,
+   "liu2021": 0.10209324769675732
+  },
+  "train_loss": 0.09986380708217621,
+  "sec": 285.6
+ },
+ {
+  "step": 5750,
+  "inner_val_loss": 0.09988139254614863,
+  "inner_val_per_cohort": {
+   "immicro": 0.09861305604378383,
+   "imperial_eskd": 0.09895080158656294,
+   "liu2021": 0.10208032000809908
+  },
+  "train_loss": 0.09941509851813317,
+  "sec": 298.5
+ },
+ {
+  "step": 6000,
+  "inner_val_loss": 0.09985958869484339,
+  "inner_val_per_cohort": {
+   "immicro": 0.0985867939889431,
+   "imperial_eskd": 0.09892555021426895,
+   "liu2021": 0.10206642188131809
+  },
+  "train_loss": 0.09969792100787163,
+  "sec": 311.4
+ },
+ {
+  "step": 6250,
+  "inner_val_loss": 0.09984024752382979,
+  "inner_val_per_cohort": {
+   "immicro": 0.09856230579316616,
+   "imperial_eskd": 0.09890321777625517,
+   "liu2021": 0.10205521900206804
+  },
+  "train_loss": 0.09973610666394234,
+  "sec": 324.6
+ },
+ {
+  "step": 6500,
+  "inner_val_loss": 0.0998197275712484,
+  "inner_val_per_cohort": {
+   "immicro": 0.09853939929356177,
+   "imperial_eskd": 0.0988794013180516,
+   "liu2021": 0.10204038210213184
+  },
+  "train_loss": 0.09981275701522827,
+  "sec": 337.6
+ },
+ {
+  "step": 6750,
+  "inner_val_loss": 0.09979919921794926,
+  "inner_val_per_cohort": {
+   "immicro": 0.0985162736227115,
+   "imperial_eskd": 0.09885861656882546,
+   "liu2021": 0.10202270746231079
+  },
+  "train_loss": 0.09966304185986519,
+  "sec": 350.3
+ },
+ {
+  "step": 7000,
+  "inner_val_loss": 0.09978242144176756,
+  "inner_val_per_cohort": {
+   "immicro": 0.09849585530658563,
+   "imperial_eskd": 0.09884292192079804,
+   "liu2021": 0.10200848709791899
+  },
+  "train_loss": 0.09957791155576706,
+  "sec": 363.0
+ },
+ {
+  "step": 7250,
+  "inner_val_loss": 0.09976628507402811,
+  "inner_val_per_cohort": {
+   "immicro": 0.09847857058048248,
+   "imperial_eskd": 0.098824199966409,
+   "liu2021": 0.10199608467519283
+  },
+  "train_loss": 0.0998088606595993,
+  "sec": 375.5
+ },
+ {
+  "step": 7500,
+  "inner_val_loss": 0.09975152052562646,
+  "inner_val_per_cohort": {
+   "immicro": 0.0984587532778581,
+   "imperial_eskd": 0.09880960868163542,
+   "liu2021": 0.10198619961738586
+  },
+  "train_loss": 0.09959805956482887,
+  "sec": 388.4
+ },
+ {
+  "step": 7750,
+  "inner_val_loss": 0.0997361382751754,
+  "inner_val_per_cohort": {
+   "immicro": 0.09844012837857008,
+   "imperial_eskd": 0.09879453954371539,
+   "liu2021": 0.10197374690324068
+  },
+  "train_loss": 0.09957461854815483,
+  "sec": 401.4
+ },
+ {
+  "step": 8000,
+  "inner_val_loss": 0.09972245570723758,
+  "inner_val_per_cohort": {
+   "immicro": 0.09842097014188766,
+   "imperial_eskd": 0.09877829761667685,
+   "liu2021": 0.10196809936314821
+  },
+  "train_loss": 0.09927423948049545,
+  "sec": 414.0
+ },
+ {
+  "step": 8250,
+  "inner_val_loss": 0.09970936799102059,
+  "inner_val_per_cohort": {
+   "immicro": 0.09840481107433637,
+   "imperial_eskd": 0.09876596148718488,
+   "liu2021": 0.10195733141154051
+  },
+  "train_loss": 0.09971532043814658,
+  "sec": 426.6
+ },
+ {
+  "step": 8500,
+  "inner_val_loss": 0.09969582947704829,
+  "inner_val_per_cohort": {
+   "immicro": 0.09838622808456421,
+   "imperial_eskd": 0.09875320677052844,
+   "liu2021": 0.10194805357605219
+  },
+  "train_loss": 0.0992968291938305,
+  "sec": 439.4
+ },
+ {
+  "step": 8750,
+  "inner_val_loss": 0.09968136669597542,
+  "inner_val_per_cohort": {
+   "immicro": 0.09837117760131757,
+   "imperial_eskd": 0.09874291142279451,
+   "liu2021": 0.10193001106381416
+  },
+  "train_loss": 0.099752850741148,
+  "sec": 452.0
+ },
+ {
+  "step": 9000,
+  "inner_val_loss": 0.09967132122491044,
+  "inner_val_per_cohort": {
+   "immicro": 0.09835810462633769,
+   "imperial_eskd": 0.09873634305867282,
+   "liu2021": 0.10191951598972082
+  },
+  "train_loss": 0.09955121904611587,
+  "sec": 464.3
+ },
+ {
+  "step": 9250,
+  "inner_val_loss": 0.09965848543615352,
+  "inner_val_per_cohort": {
+   "immicro": 0.09834518687178691,
+   "imperial_eskd": 0.09872136027975516,
+   "liu2021": 0.10190890915691853
+  },
+  "train_loss": 0.09956501331925392,
+  "sec": 477.0
+ },
+ {
+  "step": 9500,
+  "inner_val_loss": 0.09964812821661584,
+  "inner_val_per_cohort": {
+   "immicro": 0.09833218219379584,
+   "imperial_eskd": 0.09870926866477187,
+   "liu2021": 0.10190293379127979
+  },
+  "train_loss": 0.09946107017993927,
+  "sec": 489.9
+ },
+ {
+  "step": 9750,
+  "inner_val_loss": 0.09963607386394281,
+  "inner_val_per_cohort": {
+   "immicro": 0.09832089611639579,
+   "imperial_eskd": 0.09869660132310608,
+   "liu2021": 0.10189072415232658
+  },
+  "train_loss": 0.09951731741428375,
+  "sec": 502.9
+ },
+ {
+  "step": 10000,
+  "inner_val_loss": 0.099626421589743,
+  "inner_val_per_cohort": {
+   "immicro": 0.09830465819686651,
+   "imperial_eskd": 0.09868915649977597,
+   "liu2021": 0.10188545007258654
+  },
+  "train_loss": 0.09944286766648293,
+  "sec": 515.4
+ },
+ {
+  "step": 10250,
+  "inner_val_loss": 0.09961741523008154,
+  "inner_val_per_cohort": {
+   "immicro": 0.09829392905036609,
+   "imperial_eskd": 0.09867952831766823,
+   "liu2021": 0.10187878832221031
+  },
+  "train_loss": 0.0995760854780674,
+  "sec": 527.9
+ },
+ {
+  "step": 10500,
+  "inner_val_loss": 0.0996064796215958,
+  "inner_val_per_cohort": {
+   "immicro": 0.09828120687355597,
+   "imperial_eskd": 0.0986659862101078,
+   "liu2021": 0.10187224578112364
+  },
+  "train_loss": 0.09963009396195412,
+  "sec": 540.2
+ },
+ {
+  "step": 10750,
+  "inner_val_loss": 0.09959891015155749,
+  "inner_val_per_cohort": {
+   "immicro": 0.09827300813049078,
+   "imperial_eskd": 0.0986566695977341,
+   "liu2021": 0.10186705272644758
+  },
+  "train_loss": 0.0994341923892498,
+  "sec": 552.7
+ },
+ {
+  "step": 11000,
+  "inner_val_loss": 0.09958931350271509,
+  "inner_val_per_cohort": {
+   "immicro": 0.09825834507743518,
+   "imperial_eskd": 0.09865082495591858,
+   "liu2021": 0.10185877047479153
+  },
+  "train_loss": 0.09943769407272339,
+  "sec": 565.3
+ },
+ {
+  "step": 11250,
+  "inner_val_loss": 0.09957920970639796,
+  "inner_val_per_cohort": {
+   "immicro": 0.0982473058005174,
+   "imperial_eskd": 0.09864007817073302,
+   "liu2021": 0.1018502451479435
+  },
+  "train_loss": 0.09933729991316795,
+  "sec": 578.4
+ },
+ {
+  "step": 11500,
+  "inner_val_loss": 0.09956759874793615,
+  "inner_val_per_cohort": {
+   "immicro": 0.0982331009581685,
+   "imperial_eskd": 0.0986279242418029,
+   "liu2021": 0.10184177104383707
+  },
+  "train_loss": 0.0994142822921276,
+  "sec": 590.9
+ },
+ {
+  "step": 11750,
+  "inner_val_loss": 0.09955881061879072,
+  "inner_val_per_cohort": {
+   "immicro": 0.09822417516261339,
+   "imperial_eskd": 0.09862065010450104,
+   "liu2021": 0.10183160658925772
+  },
+  "train_loss": 0.09929465284943581,
+  "sec": 603.4
+ },
+ {
+  "step": 12000,
+  "inner_val_loss": 0.0995491093231572,
+  "inner_val_per_cohort": {
+   "immicro": 0.09821271399656932,
+   "imperial_eskd": 0.09860894829034805,
+   "liu2021": 0.10182566568255424
+  },
+  "train_loss": 0.09909414026141167,
+  "sec": 615.7
+ },
+ {
+  "step": 12250,
+  "inner_val_loss": 0.0995367741532097,
+  "inner_val_per_cohort": {
+   "immicro": 0.09820065709451835,
+   "imperial_eskd": 0.09859995882619511,
+   "liu2021": 0.10180970653891563
+  },
+  "train_loss": 0.09927504938840866,
+  "sec": 628.7
+ },
+ {
+  "step": 12500,
+  "inner_val_loss": 0.09952912887915817,
+  "inner_val_per_cohort": {
+   "immicro": 0.098189785455664,
+   "imperial_eskd": 0.09859249910170381,
+   "liu2021": 0.10180510208010674
+  },
+  "train_loss": 0.09944014167785645,
+  "sec": 641.2
+ },
+ {
+  "step": 12750,
+  "inner_val_loss": 0.09952226148523165,
+  "inner_val_per_cohort": {
+   "immicro": 0.09817963279783726,
+   "imperial_eskd": 0.09858482398770073,
+   "liu2021": 0.10180232767015696
+  },
+  "train_loss": 0.09937606364488602,
+  "sec": 653.9
+ },
+ {
+  "step": 13000,
+  "inner_val_loss": 0.09951249639870542,
+  "inner_val_per_cohort": {
+   "immicro": 0.09816527552902699,
+   "imperial_eskd": 0.09857382083481009,
+   "liu2021": 0.1017983928322792
+  },
+  "train_loss": 0.09946243169903755,
+  "sec": 666.7
+ },
+ {
+  "step": 13250,
+  "inner_val_loss": 0.09950401205005067,
+  "inner_val_per_cohort": {
+   "immicro": 0.09815561305731535,
+   "imperial_eskd": 0.09856353530829604,
+   "liu2021": 0.10179288778454065
+  },
+  "train_loss": 0.09941551181674003,
+  "sec": 679.5
+ },
+ {
+  "step": 13500,
+  "inner_val_loss": 0.09949715196559526,
+  "inner_val_per_cohort": {
+   "immicro": 0.09814637123296659,
+   "imperial_eskd": 0.09855547513474118,
+   "liu2021": 0.101789609529078
+  },
+  "train_loss": 0.09927832162380218,
+  "sec": 692.4
+ },
+ {
+  "step": 13750,
+  "inner_val_loss": 0.09948793691441869,
+  "inner_val_per_cohort": {
+   "immicro": 0.09813607111573219,
+   "imperial_eskd": 0.09854473410682245,
+   "liu2021": 0.10178300552070141
+  },
+  "train_loss": 0.09937187770009041,
+  "sec": 705.3
+ },
+ {
+  "step": 14000,
+  "inner_val_loss": 0.0994793206472138,
+  "inner_val_per_cohort": {
+   "immicro": 0.09812727229048808,
+   "imperial_eskd": 0.09853511608459732,
+   "liu2021": 0.10177557356655598
+  },
+  "train_loss": 0.09921340706944466,
+  "sec": 718.1
+ },
+ {
+  "step": 14250,
+  "inner_val_loss": 0.09947002589740235,
+  "inner_val_per_cohort": {
+   "immicro": 0.09811553141723077,
+   "imperial_eskd": 0.09852717207236723,
+   "liu2021": 0.10176737420260906
+  },
+  "train_loss": 0.0990984079837799,
+  "sec": 731.1
+ },
+ {
+  "step": 14500,
+  "inner_val_loss": 0.09946170557440832,
+  "inner_val_per_cohort": {
+   "immicro": 0.09810700236509244,
+   "imperial_eskd": 0.09851954301649873,
+   "liu2021": 0.1017585713416338
+  },
+  "train_loss": 0.09944122368097305,
+  "sec": 743.9
+ },
+ {
+  "step": 14750,
+  "inner_val_loss": 0.09945374564502878,
+  "inner_val_per_cohort": {
+   "immicro": 0.09809424603978793,
+   "imperial_eskd": 0.09851306405934421,
+   "liu2021": 0.10175392683595419
+  },
+  "train_loss": 0.09930595725774766,
+  "sec": 757.0
+ },
+ {
+  "step": 15000,
+  "inner_val_loss": 0.09944888815810583,
+  "inner_val_per_cohort": {
+   "immicro": 0.09808662657936414,
+   "imperial_eskd": 0.09850909222256053,
+   "liu2021": 0.10175094567239285
+  },
+  "train_loss": 0.0994044351875782,
+  "sec": 769.9
+ },
+ {
+  "step": 15250,
+  "inner_val_loss": 0.09944291897306239,
+  "inner_val_per_cohort": {
+   "immicro": 0.09808009055753548,
+   "imperial_eskd": 0.0985049002549865,
+   "liu2021": 0.10174376610666513
+  },
+  "train_loss": 0.09914515244960785,
+  "sec": 782.7
+ },
+ {
+  "step": 15500,
+  "inner_val_loss": 0.0994346420835666,
+  "inner_val_per_cohort": {
+   "immicro": 0.09806935799618562,
+   "imperial_eskd": 0.09849614887074991,
+   "liu2021": 0.10173841938376427
+  },
+  "train_loss": 0.09934927102923394,
+  "sec": 795.7
+ },
+ {
+  "step": 15750,
+  "inner_val_loss": 0.09942639542912896,
+  "inner_val_per_cohort": {
+   "immicro": 0.09806136041879654,
+   "imperial_eskd": 0.09848758070306345,
+   "liu2021": 0.10173024516552687
+  },
+  "train_loss": 0.09907898566126823,
+  "sec": 808.5
+ },
+ {
+  "step": 16000,
+  "inner_val_loss": 0.09941961278555671,
+  "inner_val_per_cohort": {
+   "immicro": 0.09805467693756025,
+   "imperial_eskd": 0.0984814702109857,
+   "liu2021": 0.10172269120812416
+  },
+  "train_loss": 0.09913892406225204,
+  "sec": 821.3
+ },
+ {
+  "step": 16250,
+  "inner_val_loss": 0.09941283091338295,
+  "inner_val_per_cohort": {
+   "immicro": 0.09804716054350138,
+   "imperial_eskd": 0.09847611764615233,
+   "liu2021": 0.10171521455049515
+  },
+  "train_loss": 0.09922684693336487,
+  "sec": 834.0
+ },
+ {
+  "step": 16500,
+  "inner_val_loss": 0.09940532670177595,
+  "inner_val_per_cohort": {
+   "immicro": 0.09803672910978396,
+   "imperial_eskd": 0.09847314012321559,
+   "liu2021": 0.10170611087232828
+  },
+  "train_loss": 0.0993327317237854,
+  "sec": 846.8
+ },
+ {
+  "step": 16750,
+  "inner_val_loss": 0.09939705412082299,
+  "inner_val_per_cohort": {
+   "immicro": 0.09802850025395553,
+   "imperial_eskd": 0.0984660153361884,
+   "liu2021": 0.10169664677232504
+  },
+  "train_loss": 0.09904087343811989,
+  "sec": 859.5
+ },
+ {
+  "step": 17000,
+  "inner_val_loss": 0.0993906840719659,
+  "inner_val_per_cohort": {
+   "immicro": 0.098020829881231,
+   "imperial_eskd": 0.09845951436595483,
+   "liu2021": 0.10169170796871185
+  },
+  "train_loss": 0.09934523591399193,
+  "sec": 872.2
+ },
+ {
+  "step": 17250,
+  "inner_val_loss": 0.09938700776547194,
+  "inner_val_per_cohort": {
+   "immicro": 0.09801443759351969,
+   "imperial_eskd": 0.09845934063196182,
+   "liu2021": 0.1016872450709343
+  },
+  "train_loss": 0.09943362191319466,
+  "sec": 884.9
+ },
+ {
+  "step": 17500,
+  "inner_val_loss": 0.09937882886239975,
+  "inner_val_per_cohort": {
+   "immicro": 0.09800645988434553,
+   "imperial_eskd": 0.09845182570544156,
+   "liu2021": 0.1016782009974122
+  },
+  "train_loss": 0.09925370261073113,
+  "sec": 897.7
+ },
+ {
+  "step": 17750,
+  "inner_val_loss": 0.099372417854164,
+  "inner_val_per_cohort": {
+   "immicro": 0.09799798236538966,
+   "imperial_eskd": 0.09844582460143349,
+   "liu2021": 0.10167344659566879
+  },
+  "train_loss": 0.09897163051366806,
+  "sec": 910.6
+ },
+ {
+  "step": 18000,
+  "inner_val_loss": 0.09936968236220907,
+  "inner_val_per_cohort": {
+   "immicro": 0.09799169686933358,
+   "imperial_eskd": 0.09844155948270451,
+   "liu2021": 0.1016757907345891
+  },
+  "train_loss": 0.09952558273077011,
+  "sec": 923.1
+ },
+ {
+  "step": 18250,
+  "inner_val_loss": 0.09936439915738925,
+  "inner_val_per_cohort": {
+   "immicro": 0.0979847793156902,
+   "imperial_eskd": 0.09843668985095891,
+   "liu2021": 0.10167172830551863
+  },
+  "train_loss": 0.09894735622406006,
+  "sec": 935.9
+ },
+ {
+  "step": 18500,
+  "inner_val_loss": 0.09935838020773548,
+  "inner_val_per_cohort": {
+   "immicro": 0.09797709342092276,
+   "imperial_eskd": 0.09843535551970656,
+   "liu2021": 0.10166269168257713
+  },
+  "train_loss": 0.09924674662947655,
+  "sec": 948.6
+ },
+ {
+  "step": 18750,
+  "inner_val_loss": 0.09935051451126735,
+  "inner_val_per_cohort": {
+   "immicro": 0.09796970337629318,
+   "imperial_eskd": 0.09843198210000992,
+   "liu2021": 0.10164985805749893
+  },
+  "train_loss": 0.09902396479249001,
+  "sec": 961.1
+ },
+ {
+  "step": 19000,
+  "inner_val_loss": 0.09934441330419345,
+  "inner_val_per_cohort": {
+   "immicro": 0.09796374384313822,
+   "imperial_eskd": 0.0984234098683704,
+   "liu2021": 0.10164608620107174
+  },
+  "train_loss": 0.09913199678063393,
+  "sec": 973.9
+ },
+ {
+  "step": 19250,
+  "inner_val_loss": 0.0993366695169096,
+  "inner_val_per_cohort": {
+   "immicro": 0.09795907077689965,
+   "imperial_eskd": 0.09841925075108354,
+   "liu2021": 0.10163168702274561
+  },
+  "train_loss": 0.09902530792355538,
+  "sec": 986.6
+ },
+ {
+  "step": 19500,
+  "inner_val_loss": 0.09933157299052586,
+  "inner_val_per_cohort": {
+   "immicro": 0.09795149695128202,
+   "imperial_eskd": 0.09841685471209613,
+   "liu2021": 0.1016263673081994
+  },
+  "train_loss": 0.09901173695921898,
+  "sec": 999.3
+ },
+ {
+  "step": 19750,
+  "inner_val_loss": 0.0993262351387077,
+  "inner_val_per_cohort": {
+   "immicro": 0.09794365304211776,
+   "imperial_eskd": 0.09841345995664597,
+   "liu2021": 0.10162159241735935
+  },
+  "train_loss": 0.09895402500033379,
+  "sec": 1011.9
+ },
+ {
+  "step": 20000,
+  "inner_val_loss": 0.09932147273135306,
+  "inner_val_per_cohort": {
+   "immicro": 0.09793707293768723,
+   "imperial_eskd": 0.09840948371724649,
+   "liu2021": 0.10161786153912544
+  },
+  "train_loss": 0.0991604634821415,
+  "sec": 1024.5
+ },
+ {
+  "step": 20250,
+  "inner_val_loss": 0.09931598445682814,
+  "inner_val_per_cohort": {
+   "immicro": 0.09792785439640284,
+   "imperial_eskd": 0.09840679439631375,
+   "liu2021": 0.10161330457776785
+  },
+  "train_loss": 0.09909227013587951,
+  "sec": 1038.4
+ },
+ {
+  "step": 20500,
+  "inner_val_loss": 0.09931197615735458,
+  "inner_val_per_cohort": {
+   "immicro": 0.09792211931198835,
+   "imperial_eskd": 0.09839960119941017,
+   "liu2021": 0.10161420796066523
+  },
+  "train_loss": 0.09877873232960702,
+  "sec": 1051.2
+ },
+ {
+  "step": 20750,
+  "inner_val_loss": 0.09930405167467667,
+  "inner_val_per_cohort": {
+   "immicro": 0.09791387803852558,
+   "imperial_eskd": 0.09839268468997696,
+   "liu2021": 0.10160559229552746
+  },
+  "train_loss": 0.0990225695669651,
+  "sec": 1063.9
+ },
+ {
+  "step": 21000,
+  "inner_val_loss": 0.09930040460606716,
+  "inner_val_per_cohort": {
+   "immicro": 0.09790898331751426,
+   "imperial_eskd": 0.09838948060165752,
+   "liu2021": 0.10160274989902973
+  },
+  "train_loss": 0.09917723760008812,
+  "sec": 1077.6
+ },
+ {
+  "step": 21250,
+  "inner_val_loss": 0.09929344578260424,
+  "inner_val_per_cohort": {
+   "immicro": 0.09790146971742313,
+   "imperial_eskd": 0.09838282045992938,
+   "liu2021": 0.10159604717046022
+  },
+  "train_loss": 0.09860109162330627,
+  "sec": 1091.4
+ },
+ {
+  "step": 21500,
+  "inner_val_loss": 0.09928767404735389,
+  "inner_val_per_cohort": {
+   "immicro": 0.09789269355436166,
+   "imperial_eskd": 0.09837973896752704,
+   "liu2021": 0.10159058962017298
+  },
+  "train_loss": 0.09882277190685272,
+  "sec": 1105.3
+ },
+ {
+  "step": 21750,
+  "inner_val_loss": 0.09928022444511604,
+  "inner_val_per_cohort": {
+   "immicro": 0.09788726984212796,
+   "imperial_eskd": 0.09837176481431181,
+   "liu2021": 0.10158163867890835
+  },
+  "train_loss": 0.09916018751263618,
+  "sec": 1119.1
+ },
+ {
+  "step": 22000,
+  "inner_val_loss": 0.09927548191538364,
+  "inner_val_per_cohort": {
+   "immicro": 0.09788008779287338,
+   "imperial_eskd": 0.09836552766236392,
+   "liu2021": 0.10158083029091358
+  },
+  "train_loss": 0.09903503370285034,
+  "sec": 1133.0
+ },
+ {
+  "step": 22250,
+  "inner_val_loss": 0.09927068057121953,
+  "inner_val_per_cohort": {
+   "immicro": 0.0978755693261822,
+   "imperial_eskd": 0.09836011617021127,
+   "liu2021": 0.10157635621726513
+  },
+  "train_loss": 0.09899048173427581,
+  "sec": 1146.6
+ },
+ {
+  "step": 22500,
+  "inner_val_loss": 0.09926656047540783,
+  "inner_val_per_cohort": {
+   "immicro": 0.09786931394288938,
+   "imperial_eskd": 0.09835801036520438,
+   "liu2021": 0.10157235711812973
+  },
+  "train_loss": 0.09895484265685081,
+  "sec": 1160.4
+ },
+ {
+  "step": 22750,
+  "inner_val_loss": 0.09926303868379556,
+  "inner_val_per_cohort": {
+   "immicro": 0.09786662831902504,
+   "imperial_eskd": 0.09835534508932721,
+   "liu2021": 0.10156714264303446
+  },
+  "train_loss": 0.0990372112095356,
+  "sec": 1174.2
+ },
+ {
+  "step": 23000,
+  "inner_val_loss": 0.09925803353991171,
+  "inner_val_per_cohort": {
+   "immicro": 0.09785938045630853,
+   "imperial_eskd": 0.09835116870023987,
+   "liu2021": 0.10156355146318674
+  },
+  "train_loss": 0.09931963378190994,
+  "sec": 1187.7
+ },
+ {
+  "step": 23250,
+  "inner_val_loss": 0.09925153099865924,
+  "inner_val_per_cohort": {
+   "immicro": 0.09785191838939984,
+   "imperial_eskd": 0.09834502027793364,
+   "liu2021": 0.10155765432864428
+  },
+  "train_loss": 0.09912974408268929,
+  "sec": 1201.5
+ },
+ {
+  "step": 23500,
+  "inner_val_loss": 0.09924697255094846,
+  "inner_val_per_cohort": {
+   "immicro": 0.0978444404900074,
+   "imperial_eskd": 0.09834511578083038,
+   "liu2021": 0.1015513613820076
+  },
+  "train_loss": 0.09916744363307953,
+  "sec": 1215.6
+ },
+ {
+  "step": 23750,
+  "inner_val_loss": 0.09924057525855423,
+  "inner_val_per_cohort": {
+   "immicro": 0.09783791316052277,
+   "imperial_eskd": 0.09834460033611818,
+   "liu2021": 0.10153921227902174
+  },
+  "train_loss": 0.09883470064401627,
+  "sec": 1229.9
+ },
+ {
+  "step": 24000,
+  "inner_val_loss": 0.09923455269649777,
+  "inner_val_per_cohort": {
+   "immicro": 0.09783134112755458,
+   "imperial_eskd": 0.09834061393683607,
+   "liu2021": 0.10153170302510262
+  },
+  "train_loss": 0.09922593986988068,
+  "sec": 1244.0
+ },
+ {
+  "step": 24250,
+  "inner_val_loss": 0.09923152810884546,
+  "inner_val_per_cohort": {
+   "immicro": 0.09782604438563187,
+   "imperial_eskd": 0.09833362664688718,
+   "liu2021": 0.10153491329401731
+  },
+  "train_loss": 0.09874129769206047,
+  "sec": 1258.1
+ },
+ {
+  "step": 24500,
+  "inner_val_loss": 0.09922134965620559,
+  "inner_val_per_cohort": {
+   "immicro": 0.09782059428592522,
+   "imperial_eskd": 0.0983217792077498,
+   "liu2021": 0.10152167547494173
+  },
+  "train_loss": 0.09866246518492698,
+  "sec": 1272.1
+ },
+ {
+  "step": 24750,
+  "inner_val_loss": 0.09921499163928356,
+  "inner_val_per_cohort": {
+   "immicro": 0.09781482070684433,
+   "imperial_eskd": 0.09831687367775223,
+   "liu2021": 0.10151328053325415
+  },
+  "train_loss": 0.09902271753549576,
+  "sec": 1285.9
+ },
+ {
+  "step": 25000,
+  "inner_val_loss": 0.09921132330077165,
+  "inner_val_per_cohort": {
+   "immicro": 0.09780739340931177,
+   "imperial_eskd": 0.09831454300067642,
+   "liu2021": 0.10151203349232674
+  },
+  "train_loss": 0.09882989925146103,
+  "sec": 1299.7
+ },
+ {
+  "step": 25250,
+  "inner_val_loss": 0.09920676035637205,
+  "inner_val_per_cohort": {
+   "immicro": 0.09779947064816952,
+   "imperial_eskd": 0.09831292690201239,
+   "liu2021": 0.10150788351893425
+  },
+  "train_loss": 0.09897023552656174,
+  "sec": 1313.4
+ },
+ {
+  "step": 25500,
+  "inner_val_loss": 0.09920429188118439,
+  "inner_val_per_cohort": {
+   "immicro": 0.09779256830612819,
+   "imperial_eskd": 0.09831175885417244,
+   "liu2021": 0.10150854848325253
+  },
+  "train_loss": 0.09876265826821327,
+  "sec": 1327.3
+ },
+ {
+  "step": 25750,
+  "inner_val_loss": 0.09920060959164843,
+  "inner_val_per_cohort": {
+   "immicro": 0.09778775926679373,
+   "imperial_eskd": 0.0983095033602281,
+   "liu2021": 0.10150456614792347
+  },
+  "train_loss": 0.09901044625043869,
+  "sec": 1341.4
+ },
+ {
+  "step": 26000,
+  "inner_val_loss": 0.0991970282706498,
+  "inner_val_per_cohort": {
+   "immicro": 0.09778195867935817,
+   "imperial_eskd": 0.09830475090579553,
+   "liu2021": 0.10150437522679567
+  },
+  "train_loss": 0.09897549378871917,
+  "sec": 1355.4
+ },
+ {
+  "step": 26250,
+  "inner_val_loss": 0.09919333028973955,
+  "inner_val_per_cohort": {
+   "immicro": 0.09777652844786644,
+   "imperial_eskd": 0.09830161387270148,
+   "liu2021": 0.10150184854865074
+  },
+  "train_loss": 0.09884169360995293,
+  "sec": 1369.0
+ },
+ {
+  "step": 26500,
+  "inner_val_loss": 0.09918844084622282,
+  "inner_val_per_cohort": {
+   "immicro": 0.09777128882706165,
+   "imperial_eskd": 0.09829627925699408,
+   "liu2021": 0.10149775445461273
+  },
+  "train_loss": 0.09904356583952904,
+  "sec": 1381.7
+ },
+ {
+  "step": 26750,
+  "inner_val_loss": 0.09918519801864721,
+  "inner_val_per_cohort": {
+   "immicro": 0.09776658720026414,
+   "imperial_eskd": 0.09829297255386006,
+   "liu2021": 0.10149603430181742
+  },
+  "train_loss": 0.09857890337705612,
+  "sec": 1394.3
+ },
+ {
+  "step": 27000,
+  "inner_val_loss": 0.09918030408784899,
+  "inner_val_per_cohort": {
+   "immicro": 0.09776247727374236,
+   "imperial_eskd": 0.09828898141329939,
+   "liu2021": 0.10148945357650518
+  },
+  "train_loss": 0.09897320041060448,
+  "sec": 1407.0
+ },
+ {
+  "step": 27250,
+  "inner_val_loss": 0.09917724019650258,
+  "inner_val_per_cohort": {
+   "immicro": 0.09775777906179428,
+   "imperial_eskd": 0.0982885364104401,
+   "liu2021": 0.10148540511727333
+  },
+  "train_loss": 0.09905730846524238,
+  "sec": 1419.3
+ },
+ {
+  "step": 27500,
+  "inner_val_loss": 0.09917365118033356,
+  "inner_val_per_cohort": {
+   "immicro": 0.09775229667623837,
+   "imperial_eskd": 0.09828591346740723,
+   "liu2021": 0.10148274339735508
+  },
+  "train_loss": 0.09889708679914475,
+  "sec": 1432.1
+ },
+ {
+  "step": 27750,
+  "inner_val_loss": 0.0991671576230514,
+  "inner_val_per_cohort": {
+   "immicro": 0.09774518261353175,
+   "imperial_eskd": 0.0982797382907434,
+   "liu2021": 0.10147655196487904
+  },
+  "train_loss": 0.0990274209678173,
+  "sec": 1444.9
+ },
+ {
+  "step": 28000,
+  "inner_val_loss": 0.09916275510131711,
+  "inner_val_per_cohort": {
+   "immicro": 0.09773981602241595,
+   "imperial_eskd": 0.09827716953375122,
+   "liu2021": 0.10147127974778414
+  },
+  "train_loss": 0.0987643429338932,
+  "sec": 1457.8
+ },
+ {
+  "step": 28250,
+  "inner_val_loss": 0.09916045847866271,
+  "inner_val_per_cohort": {
+   "immicro": 0.09773373479644458,
+   "imperial_eskd": 0.09827760234475136,
+   "liu2021": 0.10147003829479218
+  },
+  "train_loss": 0.0988626427948475,
+  "sec": 1470.8
+ },
+ {
+  "step": 28500,
+  "inner_val_loss": 0.09915524633657752,
+  "inner_val_per_cohort": {
+   "immicro": 0.09772691130638123,
+   "imperial_eskd": 0.09827740693634207,
+   "liu2021": 0.10146142076700926
+  },
+  "train_loss": 0.09846764788031578,
+  "sec": 1484.5
+ },
+ {
+  "step": 28750,
+  "inner_val_loss": 0.09914878943953852,
+  "inner_val_per_cohort": {
+   "immicro": 0.09772205383827288,
+   "imperial_eskd": 0.09827191077850082,
+   "liu2021": 0.10145240370184183
+  },
+  "train_loss": 0.0987441555261612,
+  "sec": 1497.9
+ },
+ {
+  "step": 29000,
+  "inner_val_loss": 0.09914375447451768,
+  "inner_val_per_cohort": {
+   "immicro": 0.09771755120406549,
+   "imperial_eskd": 0.09826643358577382,
+   "liu2021": 0.10144727863371372
+  },
+  "train_loss": 0.09870764049887658,
+  "sec": 1511.9
+ },
+ {
+  "step": 29250,
+  "inner_val_loss": 0.09914110728887597,
+  "inner_val_per_cohort": {
+   "immicro": 0.09770930806795756,
+   "imperial_eskd": 0.0982615663246675,
+   "liu2021": 0.10145244747400284
+  },
+  "train_loss": 0.09917493870854378,
+  "sec": 1525.8
+ },
+ {
+  "step": 29500,
+  "inner_val_loss": 0.09913930018441845,
+  "inner_val_per_cohort": {
+   "immicro": 0.09770542445282142,
+   "imperial_eskd": 0.09825960208069194,
+   "liu2021": 0.10145287401974201
+  },
+  "train_loss": 0.09845782867074013,
+  "sec": 1540.0
+ },
+ {
+  "step": 29750,
+  "inner_val_loss": 0.09913667323387632,
+  "inner_val_per_cohort": {
+   "immicro": 0.09770194596300523,
+   "imperial_eskd": 0.09825717319141734,
+   "liu2021": 0.1014509005472064
+  },
+  "train_loss": 0.09871151113510132,
+  "sec": 1554.0
+ },
+ {
+  "step": 30000,
+  "inner_val_loss": 0.09913551622079779,
+  "inner_val_per_cohort": {
+   "immicro": 0.0976989787692825,
+   "imperial_eskd": 0.09825773198496211,
+   "liu2021": 0.10144983790814877
+  },
+  "train_loss": 0.09841367676854133,
+  "sec": 1568.1
+ },
+ {
+  "step": 30250,
+  "inner_val_loss": 0.09913216795621797,
+  "inner_val_per_cohort": {
+   "immicro": 0.09769603827347358,
+   "imperial_eskd": 0.0982518924230879,
+   "liu2021": 0.10144857317209244
+  },
+  "train_loss": 0.09868360510468482,
+  "sec": 1582.1
+ },
+ {
+  "step": 30500,
+  "inner_val_loss": 0.09912930248361645,
+  "inner_val_per_cohort": {
+   "immicro": 0.09769091103225946,
+   "imperial_eskd": 0.0982509387487715,
+   "liu2021": 0.1014460576698184
+  },
+  "train_loss": 0.09870548057556153,
+  "sec": 1596.3
+ },
+ {
+  "step": 30750,
+  "inner_val_loss": 0.09912626362509197,
+  "inner_val_per_cohort": {
+   "immicro": 0.09768498533715804,
+   "imperial_eskd": 0.09824775531888008,
+   "liu2021": 0.1014460502192378
+  },
+  "train_loss": 0.09872204312682152,
+  "sec": 1610.3
+ },
+ {
+  "step": 31000,
+  "inner_val_loss": 0.09912310192606062,
+  "inner_val_per_cohort": {
+   "immicro": 0.09767971870799859,
+   "imperial_eskd": 0.09824757785959677,
+   "liu2021": 0.10144200921058655
+  },
+  "train_loss": 0.0983419722020626,
+  "sec": 1624.4
+ },
+ {
+  "step": 31250,
+  "inner_val_loss": 0.09911964942861085,
+  "inner_val_per_cohort": {
+   "immicro": 0.09767498386402924,
+   "imperial_eskd": 0.09824376011436636,
+   "liu2021": 0.10144020430743694
+  },
+  "train_loss": 0.09862438452243805,
+  "sec": 1638.5
+ },
+ {
+  "step": 31500,
+  "inner_val_loss": 0.09911355335792206,
+  "inner_val_per_cohort": {
+   "immicro": 0.09767203809072574,
+   "imperial_eskd": 0.09824164245616306,
+   "liu2021": 0.1014269795268774
+  },
+  "train_loss": 0.09842986735701562,
+  "sec": 1652.4
+ },
+ {
+  "step": 31750,
+  "inner_val_loss": 0.09910792827305166,
+  "inner_val_per_cohort": {
+   "immicro": 0.09766598666707675,
+   "imperial_eskd": 0.09823510084639896,
+   "liu2021": 0.10142269730567932
+  },
+  "train_loss": 0.0987536798119545,
+  "sec": 1666.3
+ },
+ {
+  "step": 32000,
+  "inner_val_loss": 0.09910198286025211,
+  "inner_val_per_cohort": {
+   "immicro": 0.09765960959096749,
+   "imperial_eskd": 0.09823068434541876,
+   "liu2021": 0.10141565464437008
+  },
+  "train_loss": 0.09872548222541809,
+  "sec": 1680.6
+ },
+ {
+  "step": 32250,
+  "inner_val_loss": 0.09909772048845435,
+  "inner_val_per_cohort": {
+   "immicro": 0.09765406977385283,
+   "imperial_eskd": 0.09822591157122092,
+   "liu2021": 0.10141318012028933
+  },
+  "train_loss": 0.0985929498076439,
+  "sec": 1693.5
+ },
+ {
+  "step": 32500,
+  "inner_val_loss": 0.09909499359477048,
+  "inner_val_per_cohort": {
+   "immicro": 0.09764991172899802,
+   "imperial_eskd": 0.09822256321256811,
+   "liu2021": 0.1014125058427453
+  },
+  "train_loss": 0.09852355709671974,
+  "sec": 1706.3
+ },
+ {
+  "step": 32750,
+  "inner_val_loss": 0.09909298947325558,
+  "inner_val_per_cohort": {
+   "immicro": 0.09764771939565738,
+   "imperial_eskd": 0.09822189943356947,
+   "liu2021": 0.10140934959053993
+  },
+  "train_loss": 0.09876179826259614,
+  "sec": 1718.9
+ },
+ {
+  "step": 33000,
+  "inner_val_loss": 0.09908891737348202,
+  "inner_val_per_cohort": {
+   "immicro": 0.09764535631984472,
+   "imperial_eskd": 0.09821801971305501,
+   "liu2021": 0.10140337608754635
+  },
+  "train_loss": 0.09851018062233925,
+  "sec": 1731.4
+ },
+ {
+  "step": 33250,
+  "inner_val_loss": 0.09908401863555415,
+  "inner_val_per_cohort": {
+   "immicro": 0.09763778497775395,
+   "imperial_eskd": 0.098218039016832,
+   "liu2021": 0.10139623191207647
+  },
+  "train_loss": 0.0986398254930973,
+  "sec": 1744.0
+ },
+ {
+  "step": 33500,
+  "inner_val_loss": 0.09907979069446976,
+  "inner_val_per_cohort": {
+   "immicro": 0.09763400908559561,
+   "imperial_eskd": 0.09821588783101602,
+   "liu2021": 0.10138947516679764
+  },
+  "train_loss": 0.09819977229833603,
+  "sec": 1756.7
+ },
+ {
+  "step": 33750,
+  "inner_val_loss": 0.0990791330019934,
+  "inner_val_per_cohort": {
+   "immicro": 0.09762906779845555,
+   "imperial_eskd": 0.09821356054056775,
+   "liu2021": 0.1013947706669569
+  },
+  "train_loss": 0.0985832905471325,
+  "sec": 1769.4
+ },
+ {
+  "step": 34000,
+  "inner_val_loss": 0.09907285954727969,
+  "inner_val_per_cohort": {
+   "immicro": 0.09762326317528884,
+   "imperial_eskd": 0.09821011227640239,
+   "liu2021": 0.10138520319014788
+  },
+  "train_loss": 0.09875801518559456,
+  "sec": 1781.7
+ },
+ {
+  "step": 34250,
+  "inner_val_loss": 0.09906884736259176,
+  "inner_val_per_cohort": {
+   "immicro": 0.09761835386355718,
+   "imperial_eskd": 0.09820584512569687,
+   "liu2021": 0.10138234309852123
+  },
+  "train_loss": 0.09877310132980346,
+  "sec": 1795.5
+ },
+ {
+  "step": 34500,
+  "inner_val_loss": 0.0990669099482322,
+  "inner_val_per_cohort": {
+   "immicro": 0.09761201652387778,
+   "imperial_eskd": 0.0982038016346368,
+   "liu2021": 0.10138491168618202
+  },
+  "train_loss": 0.09875679022073745,
+  "sec": 1809.4
+ },
+ {
+  "step": 34750,
+  "inner_val_loss": 0.09906338668908134,
+  "inner_val_per_cohort": {
+   "immicro": 0.09760979749262333,
+   "imperial_eskd": 0.09820132567123933,
+   "liu2021": 0.10137903690338135
+  },
+  "train_loss": 0.0987747938632965,
+  "sec": 1823.0
+ },
+ {
+  "step": 35000,
+  "inner_val_loss": 0.09905910091192434,
+  "inner_val_per_cohort": {
+   "immicro": 0.09760636929422617,
+   "imperial_eskd": 0.09819863093170253,
+   "liu2021": 0.1013723025098443
+  },
+  "train_loss": 0.0986803748011589,
+  "sec": 1836.6
+ },
+ {
+  "step": 35250,
+  "inner_val_loss": 0.09905639912632076,
+  "inner_val_per_cohort": {
+   "immicro": 0.09760046439866225,
+   "imperial_eskd": 0.0981998856772076,
+   "liu2021": 0.10136884730309248
+  },
+  "train_loss": 0.09869396063685418,
+  "sec": 1850.5
+ },
+ {
+  "step": 35500,
+  "inner_val_loss": 0.09905014130653757,
+  "inner_val_per_cohort": {
+   "immicro": 0.09759867563843727,
+   "imperial_eskd": 0.09819364378398115,
+   "liu2021": 0.10135810449719429
+  },
+  "train_loss": 0.09884883388876915,
+  "sec": 1864.3
+ },
+ {
+  "step": 35750,
+  "inner_val_loss": 0.09904974881521982,
+  "inner_val_per_cohort": {
+   "immicro": 0.09759419814993937,
+   "imperial_eskd": 0.09819397194819017,
+   "liu2021": 0.10136107634752989
+  },
+  "train_loss": 0.09855009216070175,
+  "sec": 1878.1
+ },
+ {
+  "step": 36000,
+  "inner_val_loss": 0.09904481701063689,
+  "inner_val_per_cohort": {
+   "immicro": 0.09759126820911963,
+   "imperial_eskd": 0.09819029745730487,
+   "liu2021": 0.10135288536548615
+  },
+  "train_loss": 0.0986707493364811,
+  "sec": 1890.5
+ },
+ {
+  "step": 36250,
+  "inner_val_loss": 0.09903975530066574,
+  "inner_val_per_cohort": {
+   "immicro": 0.09758839880426724,
+   "imperial_eskd": 0.09818435663526709,
+   "liu2021": 0.1013465104624629
+  },
+  "train_loss": 0.09864826548099517,
+  "sec": 1904.4
+ },
+ {
+  "step": 36500,
+  "inner_val_loss": 0.09903622673579841,
+  "inner_val_per_cohort": {
+   "immicro": 0.09758306449900071,
+   "imperial_eskd": 0.09818223728375002,
+   "liu2021": 0.10134337842464447
+  },
+  "train_loss": 0.0982878431379795,
+  "sec": 1918.3
+ },
+ {
+  "step": 36750,
+  "inner_val_loss": 0.0990313274240253,
+  "inner_val_per_cohort": {
+   "immicro": 0.09757940874745448,
+   "imperial_eskd": 0.0981742489066991,
+   "liu2021": 0.1013403246179223
+  },
+  "train_loss": 0.09874728581309318,
+  "sec": 1932.1
+ },
+ {
+  "step": 37000,
+  "inner_val_loss": 0.09902978879446633,
+  "inner_val_per_cohort": {
+   "immicro": 0.09757681594540675,
+   "imperial_eskd": 0.0981742458587343,
+   "liu2021": 0.10133830457925797
+  },
+  "train_loss": 0.09828468543291093,
+  "sec": 1946.0
+ },
+ {
+  "step": 37250,
+  "inner_val_loss": 0.09902910031190122,
+  "inner_val_per_cohort": {
+   "immicro": 0.09757336198041837,
+   "imperial_eskd": 0.09817343645475128,
+   "liu2021": 0.10134050250053406
+  },
+  "train_loss": 0.09852338299155235,
+  "sec": 1959.6
+ },
+ {
+  "step": 37500,
+  "inner_val_loss": 0.09902871859194051,
+  "inner_val_per_cohort": {
+   "immicro": 0.09756888387103875,
+   "imperial_eskd": 0.09817406298084692,
+   "liu2021": 0.10134320892393589
+  },
+  "train_loss": 0.09866667062044143,
+  "sec": 1972.0
+ },
+ {
+  "step": 37750,
+  "inner_val_loss": 0.09902181558198098,
+  "inner_val_per_cohort": {
+   "immicro": 0.09756243601441383,
+   "imperial_eskd": 0.09816728125918996,
+   "liu2021": 0.10133572947233915
+  },
+  "train_loss": 0.09860412684082985,
+  "sec": 1985.7
+ },
+ {
+  "step": 38000,
+  "inner_val_loss": 0.09901802065885729,
+  "inner_val_per_cohort": {
+   "immicro": 0.09756119300921758,
+   "imperial_eskd": 0.0981619842350483,
+   "liu2021": 0.101330884732306
+  },
+  "train_loss": 0.09869580921530724,
+  "sec": 1999.2
+ },
+ {
+  "step": 38250,
+  "inner_val_loss": 0.09901341904105261,
+  "inner_val_per_cohort": {
+   "immicro": 0.09755640911559264,
+   "imperial_eskd": 0.09816109930927103,
+   "liu2021": 0.10132274869829416
+  },
+  "train_loss": 0.09872870460152626,
+  "sec": 2013.2
+ },
+ {
+  "step": 38500,
+  "inner_val_loss": 0.09901132716825516,
+  "inner_val_per_cohort": {
+   "immicro": 0.09755590464919806,
+   "imperial_eskd": 0.09816022691401569,
+   "liu2021": 0.10131784994155169
+  },
+  "train_loss": 0.09865416032075881,
+  "sec": 2027.6
+ },
+ {
+  "step": 38750,
+  "inner_val_loss": 0.09900800249718054,
+  "inner_val_per_cohort": {
+   "immicro": 0.09755282911161582,
+   "imperial_eskd": 0.09815901043740186,
+   "liu2021": 0.10131216794252396
+  },
+  "train_loss": 0.09858832192420959,
+  "sec": 2041.9
+ },
+ {
+  "step": 39000,
+  "inner_val_loss": 0.09900275459795287,
+  "inner_val_per_cohort": {
+   "immicro": 0.09754819422960281,
+   "imperial_eskd": 0.09815628725019368,
+   "liu2021": 0.10130378231406212
+  },
+  "train_loss": 0.09844092440605164,
+  "sec": 2056.4
+ },
+ {
+  "step": 39250,
+  "inner_val_loss": 0.09899934355846861,
+  "inner_val_per_cohort": {
+   "immicro": 0.09754341468214989,
+   "imperial_eskd": 0.09815615618770773,
+   "liu2021": 0.10129845980554819
+  },
+  "train_loss": 0.0984466978609562,
+  "sec": 2070.7
+ },
+ {
+  "step": 39500,
+  "inner_val_loss": 0.09899606801230798,
+  "inner_val_per_cohort": {
+   "immicro": 0.09754121024161577,
+   "imperial_eskd": 0.09815312820402058,
+   "liu2021": 0.10129386559128761
+  },
+  "train_loss": 0.09882323294878007,
+  "sec": 2085.2
+ },
+ {
+  "step": 39750,
+  "inner_val_loss": 0.09899217562956943,
+  "inner_val_per_cohort": {
+   "immicro": 0.09753649185101192,
+   "imperial_eskd": 0.09815161675214767,
+   "liu2021": 0.10128841828554869
+  },
+  "train_loss": 0.09847553443908691,
+  "sec": 2098.9
+ },
+ {
+  "step": 40000,
+  "inner_val_loss": 0.09898765926070587,
+  "inner_val_per_cohort": {
+   "immicro": 0.09753064842273791,
+   "imperial_eskd": 0.09814834323796359,
+   "liu2021": 0.10128398612141609
+  },
+  "train_loss": 0.0984780575633049,
+  "sec": 2112.8
+ },
+ {
+  "step": 40250,
+  "inner_val_loss": 0.09898110274978057,
+  "inner_val_per_cohort": {
+   "immicro": 0.09752543767293294,
+   "imperial_eskd": 0.09814280373128978,
+   "liu2021": 0.101275066845119
+  },
+  "train_loss": 0.09863488444685936,
+  "sec": 2128.3
+ },
+ {
+  "step": 40500,
+  "inner_val_loss": 0.0989794350426757,
+  "inner_val_per_cohort": {
+   "immicro": 0.0975230485200882,
+   "imperial_eskd": 0.0981414992023598,
+   "liu2021": 0.10127375740557909
+  },
+  "train_loss": 0.09876834839582443,
+  "sec": 2142.2
+ },
+ {
+  "step": 40750,
+  "inner_val_loss": 0.09897794470078114,
+  "inner_val_per_cohort": {
+   "immicro": 0.09751899167895317,
+   "imperial_eskd": 0.09814132716168057,
+   "liu2021": 0.10127351526170969
+  },
+  "train_loss": 0.09833220881223678,
+  "sec": 2156.0
+ },
+ {
+  "step": 41000,
+  "inner_val_loss": 0.09897463395250868,
+  "inner_val_per_cohort": {
+   "immicro": 0.0975122603898247,
+   "imperial_eskd": 0.0981390770863403,
+   "liu2021": 0.10127256438136101
+  },
+  "train_loss": 0.09854984864592552,
+  "sec": 2169.9
+ },
+ {
+  "step": 41250,
+  "inner_val_loss": 0.09897294437343424,
+  "inner_val_per_cohort": {
+   "immicro": 0.09751118998974562,
+   "imperial_eskd": 0.09813858704133467,
+   "liu2021": 0.10126905608922243
+  },
+  "train_loss": 0.09841637402772903,
+  "sec": 2183.6
+ },
+ {
+  "step": 41500,
+  "inner_val_loss": 0.09897139000546451,
+  "inner_val_per_cohort": {
+   "immicro": 0.09750814052919547,
+   "imperial_eskd": 0.0981410572474653,
+   "liu2021": 0.10126497223973274
+  },
+  "train_loss": 0.09844745883345604,
+  "sec": 2197.7
+ },
+ {
+  "step": 41750,
+  "inner_val_loss": 0.09896503536576273,
+  "inner_val_per_cohort": {
+   "immicro": 0.09750490418324868,
+   "imperial_eskd": 0.09813672033223239,
+   "liu2021": 0.10125348158180714
+  },
+  "train_loss": 0.09838932183384895,
+  "sec": 2211.7
+ },
+ {
+  "step": 42000,
+  "inner_val_loss": 0.09896195565133986,
+  "inner_val_per_cohort": {
+   "immicro": 0.09749808597067992,
+   "imperial_eskd": 0.09813257272947919,
+   "liu2021": 0.10125520825386047
+  },
+  "train_loss": 0.09835660070180893,
+  "sec": 2225.4
+ },
+ {
+  "step": 42250,
+  "inner_val_loss": 0.0989602394778319,
+  "inner_val_per_cohort": {
+   "immicro": 0.09749485055605571,
+   "imperial_eskd": 0.09813363613052802,
+   "liu2021": 0.101252231746912
+  },
+  "train_loss": 0.09825029286742211,
+  "sec": 2239.1
+ },
+ {
+  "step": 42500,
+  "inner_val_loss": 0.0989595265268828,
+  "inner_val_per_cohort": {
+   "immicro": 0.09748883731663227,
+   "imperial_eskd": 0.09813278032974763,
+   "liu2021": 0.10125696193426847
+  },
+  "train_loss": 0.09873570367693901,
+  "sec": 2252.9
+ },
+ {
+  "step": 42750,
+  "inner_val_loss": 0.09895922416689419,
+  "inner_val_per_cohort": {
+   "immicro": 0.09748558482776086,
+   "imperial_eskd": 0.09813095527616414,
+   "liu2021": 0.1012611323967576
+  },
+  "train_loss": 0.09825997495651245,
+  "sec": 2265.4
+ },
+ {
+  "step": 43000,
+  "inner_val_loss": 0.0989611354007414,
+  "inner_val_per_cohort": {
+   "immicro": 0.09748478140681982,
+   "imperial_eskd": 0.09812577542933551,
+   "liu2021": 0.10127284936606884
+  },
+  "train_loss": 0.09820601975917816,
+  "sec": 2279.1
+ },
+ {
+  "step": 43250,
+  "inner_val_loss": 0.09895589609978477,
+  "inner_val_per_cohort": {
+   "immicro": 0.0974815512696902,
+   "imperial_eskd": 0.09812211753292517,
+   "liu2021": 0.10126401949673891
+  },
+  "train_loss": 0.09866348716616631,
+  "sec": 2291.5
+ },
+ {
+  "step": 43500,
+  "inner_val_loss": 0.09895255987654732,
+  "inner_val_per_cohort": {
+   "immicro": 0.097480993407468,
+   "imperial_eskd": 0.09811992265961388,
+   "liu2021": 0.10125676356256008
+  },
+  "train_loss": 0.09859293150901795,
+  "sec": 2305.3
+ },
+ {
+  "step": 43750,
+  "inner_val_loss": 0.09895097982665203,
+  "inner_val_per_cohort": {
+   "immicro": 0.09747943126906951,
+   "imperial_eskd": 0.09811439771543849,
+   "liu2021": 0.10125911049544811
+  },
+  "train_loss": 0.09891776362061501,
+  "sec": 2319.1
+ },
+ {
+  "step": 44000,
+  "inner_val_loss": 0.09894603519992094,
+  "inner_val_per_cohort": {
+   "immicro": 0.09747591118017833,
+   "imperial_eskd": 0.09811030260541222,
+   "liu2021": 0.10125189181417227
+  },
+  "train_loss": 0.09817283114790916,
+  "sec": 2332.9
+ },
+ {
+  "step": 44250,
+  "inner_val_loss": 0.0989423855725262,
+  "inner_val_per_cohort": {
+   "immicro": 0.09747337984542052,
+   "imperial_eskd": 0.09810895845293999,
+   "liu2021": 0.10124481841921806
+  },
+  "train_loss": 0.09841709408164025,
+  "sec": 2346.6
+ },
+ {
+  "step": 44500,
+  "inner_val_loss": 0.09893802191234297,
+  "inner_val_per_cohort": {
+   "immicro": 0.09746833735456069,
+   "imperial_eskd": 0.09810491651296616,
+   "liu2021": 0.10124081186950207
+  },
+  "train_loss": 0.09856953230500222,
+  "sec": 2360.6
+ },
+ {
+  "step": 44750,
+  "inner_val_loss": 0.09893562100978211,
+  "inner_val_per_cohort": {
+   "immicro": 0.09746535743276279,
+   "imperial_eskd": 0.09810438548976724,
+   "liu2021": 0.10123712010681629
+  },
+  "train_loss": 0.09845707362890244,
+  "sec": 2374.3
+ },
+ {
+  "step": 45000,
+  "inner_val_loss": 0.09893319742622399,
+  "inner_val_per_cohort": {
+   "immicro": 0.09746038510153691,
+   "imperial_eskd": 0.09810352968898686,
+   "liu2021": 0.10123567748814821
+  },
+  "train_loss": 0.09835731837153434,
+  "sec": 2388.1
+ },
+ {
+  "step": 45250,
+  "inner_val_loss": 0.09893307406831571,
+  "inner_val_per_cohort": {
+   "immicro": 0.09745754084239404,
+   "imperial_eskd": 0.0981035703285174,
+   "liu2021": 0.10123811103403568
+  },
+  "train_loss": 0.0985564393401146,
+  "sec": 2401.9
+ },
+ {
+  "step": 45500,
+  "inner_val_loss": 0.09892811015662219,
+  "inner_val_per_cohort": {
+   "immicro": 0.09745155305912097,
+   "imperial_eskd": 0.0981040932238102,
+   "liu2021": 0.10122868418693542
+  },
+  "train_loss": 0.09880533102154732,
+  "sec": 2414.3
+ },
+ {
+  "step": 45750,
+  "inner_val_loss": 0.09892689777218333,
+  "inner_val_per_cohort": {
+   "immicro": 0.09744678655018409,
+   "imperial_eskd": 0.09810532223094594,
+   "liu2021": 0.10122858453541994
+  },
+  "train_loss": 0.09846422237157822,
+  "sec": 2427.9
+ },
+ {
+  "step": 46000,
+  "inner_val_loss": 0.09892412223335768,
+  "inner_val_per_cohort": {
+   "immicro": 0.09744312738378842,
+   "imperial_eskd": 0.09810631349682808,
+   "liu2021": 0.10122292581945658
+  },
+  "train_loss": 0.09860602635145187,
+  "sec": 2441.7
+ },
+ {
+  "step": 46250,
+  "inner_val_loss": 0.0989208427737608,
+  "inner_val_per_cohort": {
+   "immicro": 0.09743834938853979,
+   "imperial_eskd": 0.0981065458194776,
+   "liu2021": 0.10121763311326504
+  },
+  "train_loss": 0.0982167918086052,
+  "sec": 2455.4
+ },
+ {
+  "step": 46500,
+  "inner_val_loss": 0.0989198966346907,
+  "inner_val_per_cohort": {
+   "immicro": 0.0974351353943348,
+   "imperial_eskd": 0.09810808368704536,
+   "liu2021": 0.10121647082269192
+  },
+  "train_loss": 0.09846789127588272,
+  "sec": 2469.1
+ },
+ {
+  "step": 46750,
+  "inner_val_loss": 0.09891617360214393,
+  "inner_val_per_cohort": {
+   "immicro": 0.09742872416973114,
+   "imperial_eskd": 0.09810557216405869,
+   "liu2021": 0.10121422447264194
+  },
+  "train_loss": 0.0983207868039608,
+  "sec": 2481.4
+ },
+ {
+  "step": 47000,
+  "inner_val_loss": 0.09891278834806548,
+  "inner_val_per_cohort": {
+   "immicro": 0.09742487066735823,
+   "imperial_eskd": 0.09810222685337067,
+   "liu2021": 0.10121126752346754
+  },
+  "train_loss": 0.09824063456058502,
+  "sec": 2495.1
+ },
+ {
+  "step": 47250,
+  "inner_val_loss": 0.09890796389016841,
+  "inner_val_per_cohort": {
+   "immicro": 0.09742096221695344,
+   "imperial_eskd": 0.09809619188308716,
+   "liu2021": 0.10120673757046461
+  },
+  "train_loss": 0.09838235020637512,
+  "sec": 2509.1
+ },
+ {
+  "step": 47500,
+  "inner_val_loss": 0.09890885190153964,
+  "inner_val_per_cohort": {
+   "immicro": 0.09741494463135798,
+   "imperial_eskd": 0.09809693490917032,
+   "liu2021": 0.10121467616409063
+  },
+  "train_loss": 0.09849834579229355,
+  "sec": 2522.3
+ },
+ {
+  "step": 47750,
+  "inner_val_loss": 0.09890275664928587,
+  "inner_val_per_cohort": {
+   "immicro": 0.09740918440123399,
+   "imperial_eskd": 0.09809062223542821,
+   "liu2021": 0.10120846331119537
+  },
+  "train_loss": 0.0983958714902401,
+  "sec": 2534.9
+ },
+ {
+  "step": 48000,
+  "inner_val_loss": 0.09889868235761168,
+  "inner_val_per_cohort": {
+   "immicro": 0.09740820775429408,
+   "imperial_eskd": 0.09808473018082706,
+   "liu2021": 0.10120310913771391
+  },
+  "train_loss": 0.09840472051501274,
+  "sec": 2549.1
+ },
+ {
+  "step": 48250,
+  "inner_val_loss": 0.09889408166172227,
+  "inner_val_per_cohort": {
+   "immicro": 0.09740374175210793,
+   "imperial_eskd": 0.0980820046229796,
+   "liu2021": 0.10119649861007929
+  },
+  "train_loss": 0.09859969383478165,
+  "sec": 2563.2
+ },
+ {
+  "step": 48500,
+  "inner_val_loss": 0.09889320559762042,
+  "inner_val_per_cohort": {
+   "immicro": 0.09740274337430795,
+   "imperial_eskd": 0.09807843579487367,
+   "liu2021": 0.10119843762367964
+  },
+  "train_loss": 0.09827679541707039,
+  "sec": 2577.7
+ },
+ {
+  "step": 48750,
+  "inner_val_loss": 0.09889078404603883,
+  "inner_val_per_cohort": {
+   "immicro": 0.09739609000583489,
+   "imperial_eskd": 0.09807687455957587,
+   "liu2021": 0.10119938757270575
+  },
+  "train_loss": 0.09816165071725845,
+  "sec": 2590.4
+ },
+ {
+  "step": 49000,
+  "inner_val_loss": 0.09889061779084833,
+  "inner_val_per_cohort": {
+   "immicro": 0.09739314143856366,
+   "imperial_eskd": 0.0980749015103687,
+   "liu2021": 0.1012038104236126
+  },
+  "train_loss": 0.09852798092365266,
+  "sec": 2604.5
+ },
+ {
+  "step": 49250,
+  "inner_val_loss": 0.09888757087026882,
+  "inner_val_per_cohort": {
+   "immicro": 0.0973929123332103,
+   "imperial_eskd": 0.09807216172868555,
+   "liu2021": 0.10119763854891062
+  },
+  "train_loss": 0.09834512567520141,
+  "sec": 2617.2
+ },
+ {
+  "step": 49500,
+  "inner_val_loss": 0.09888222670615321,
+  "inner_val_per_cohort": {
+   "immicro": 0.09738559151689212,
+   "imperial_eskd": 0.09806650944731453,
+   "liu2021": 0.101194579154253
+  },
+  "train_loss": 0.09830144602060319,
+  "sec": 2631.5
+ },
+ {
+  "step": 49750,
+  "inner_val_loss": 0.09888031877426788,
+  "inner_val_per_cohort": {
+   "immicro": 0.09738144868363936,
+   "imperial_eskd": 0.09806360507553274,
+   "liu2021": 0.10119590256363153
+  },
+  "train_loss": 0.09846130934357643,
+  "sec": 2645.5
+ },
+ {
+  "step": 50000,
+  "inner_val_loss": 0.09887719702803427,
+  "inner_val_per_cohort": {
+   "immicro": 0.09737802638361852,
+   "imperial_eskd": 0.09806293249130249,
+   "liu2021": 0.10119063220918179
+  },
+  "train_loss": 0.09789123430848122,
+  "sec": 2659.3
+ },
+ {
+  "step": 50250,
+  "inner_val_loss": 0.09887573972457286,
+  "inner_val_per_cohort": {
+   "immicro": 0.09737597685307264,
+   "imperial_eskd": 0.0980683568526398,
+   "liu2021": 0.10118288546800613
+  },
+  "train_loss": 0.09816987890005112,
+  "sec": 2673.1
+ },
+ {
+  "step": 50500,
+  "inner_val_loss": 0.09887137304464676,
+  "inner_val_per_cohort": {
+   "immicro": 0.09737241050849359,
+   "imperial_eskd": 0.09806474636901509,
+   "liu2021": 0.10117696225643158
+  },
+  "train_loss": 0.09852195900678634,
+  "sec": 2686.8
+ },
+ {
+  "step": 50750,
+  "inner_val_loss": 0.09886794823287715,
+  "inner_val_per_cohort": {
+   "immicro": 0.09736824004600446,
+   "imperial_eskd": 0.09806471555070444,
+   "liu2021": 0.10117088910192251
+  },
+  "train_loss": 0.09810144516825677,
+  "sec": 2700.6
+ },
+ {
+  "step": 51000,
+  "inner_val_loss": 0.0988647768066989,
+  "inner_val_per_cohort": {
+   "immicro": 0.09736557522167762,
+   "imperial_eskd": 0.09806312993168831,
+   "liu2021": 0.10116562526673079
+  },
+  "train_loss": 0.09838654580712318,
+  "sec": 2714.3
+ },
+ {
+  "step": 51250,
+  "inner_val_loss": 0.09886333009585586,
+  "inner_val_per_cohort": {
+   "immicro": 0.09736480315526326,
+   "imperial_eskd": 0.09805946221405809,
+   "liu2021": 0.10116572491824627
+  },
+  "train_loss": 0.09825894039869308,
+  "sec": 2728.0
+ },
+ {
+  "step": 51500,
+  "inner_val_loss": 0.09885857052009817,
+  "inner_val_per_cohort": {
+   "immicro": 0.09736131380001704,
+   "imperial_eskd": 0.09805662726814096,
+   "liu2021": 0.10115777049213648
+  },
+  "train_loss": 0.09828115099668502,
+  "sec": 2741.8
+ },
+ {
+  "step": 51750,
+  "inner_val_loss": 0.09885235072461644,
+  "inner_val_per_cohort": {
+   "immicro": 0.09735893985877435,
+   "imperial_eskd": 0.09805388884110884,
+   "liu2021": 0.10114422347396612
+  },
+  "train_loss": 0.09836225643754005,
+  "sec": 2755.5
+ },
+ {
+  "step": 52000,
+  "inner_val_loss": 0.09885030437373754,
+  "inner_val_per_cohort": {
+   "immicro": 0.09735370054841042,
+   "imperial_eskd": 0.09804783422838557,
+   "liu2021": 0.10114937834441662
+  },
+  "train_loss": 0.09851102134585381,
+  "sec": 2769.2
+ },
+ {
+  "step": 52250,
+  "inner_val_loss": 0.09884772080965716,
+  "inner_val_per_cohort": {
+   "immicro": 0.09735023664931457,
+   "imperial_eskd": 0.09804532812400298,
+   "liu2021": 0.10114759765565395
+  },
+  "train_loss": 0.09837977209687233,
+  "sec": 2782.8
+ },
+ {
+  "step": 52500,
+  "inner_val_loss": 0.09884526392426153,
+  "inner_val_per_cohort": {
+   "immicro": 0.09734466547767322,
+   "imperial_eskd": 0.0980432399294593,
+   "liu2021": 0.10114788636565208
+  },
+  "train_loss": 0.09871128386259079,
+  "sec": 2796.5
+ },
+ {
+  "step": 52750,
+  "inner_val_loss": 0.09884579190890296,
+  "inner_val_per_cohort": {
+   "immicro": 0.09734417622288068,
+   "imperial_eskd": 0.09804096106778491,
+   "liu2021": 0.10115223843604326
+  },
+  "train_loss": 0.09846632876992226,
+  "sec": 2810.1
+ },
+ {
+  "step": 53000,
+  "inner_val_loss": 0.09884426264901354,
+  "inner_val_per_cohort": {
+   "immicro": 0.09733927653481562,
+   "imperial_eskd": 0.09804099730469963,
+   "liu2021": 0.10115251410752535
+  },
+  "train_loss": 0.09808522135019303,
+  "sec": 2822.5
+ },
+ {
+  "step": 53250,
+  "inner_val_loss": 0.098844891404639,
+  "inner_val_per_cohort": {
+   "immicro": 0.09733598803480466,
+   "imperial_eskd": 0.09804294597018849,
+   "liu2021": 0.10115574020892382
+  },
+  "train_loss": 0.0984459787607193,
+  "sec": 2836.3
+ },
+ {
+  "step": 53500,
+  "inner_val_loss": 0.0988456274316919,
+  "inner_val_per_cohort": {
+   "immicro": 0.097333708157142,
+   "imperial_eskd": 0.09804750470952554,
+   "liu2021": 0.10115566942840815
+  },
+  "train_loss": 0.09817182359099388,
+  "sec": 2848.7
+ },
+ {
+  "step": 53750,
+  "inner_val_loss": 0.09884178704309343,
+  "inner_val_per_cohort": {
+   "immicro": 0.0973307794580857,
+   "imperial_eskd": 0.098041871054606,
+   "liu2021": 0.10115271061658859
+  },
+  "train_loss": 0.09834760090708733,
+  "sec": 2861.1
+ },
+ {
+  "step": 54000,
+  "inner_val_loss": 0.09883891488190251,
+  "inner_val_per_cohort": {
+   "immicro": 0.09732804944117864,
+   "imperial_eskd": 0.09803801300850781,
+   "liu2021": 0.10115068219602108
+  },
+  "train_loss": 0.09812273532152176,
+  "sec": 2874.8
+ },
+ {
+  "step": 54250,
+  "inner_val_loss": 0.0988343587483872,
+  "inner_val_per_cohort": {
+   "immicro": 0.09732493571937084,
+   "imperial_eskd": 0.0980344181033698,
+   "liu2021": 0.10114372242242098
+  },
+  "train_loss": 0.098472372174263,
+  "sec": 2888.7
+ },
+ {
+  "step": 54500,
+  "inner_val_loss": 0.09883083150054169,
+  "inner_val_per_cohort": {
+   "immicro": 0.09731753884504239,
+   "imperial_eskd": 0.09803405234759505,
+   "liu2021": 0.10114090330898762
+  },
+  "train_loss": 0.0983714204132557,
+  "sec": 2902.4
+ },
+ {
+  "step": 54750,
+  "inner_val_loss": 0.09882838412592508,
+  "inner_val_per_cohort": {
+   "immicro": 0.09731321843961875,
+   "imperial_eskd": 0.09803049605001103,
+   "liu2021": 0.10114143788814545
+  },
+  "train_loss": 0.0981089793741703,
+  "sec": 2916.2
+ },
+ {
+  "step": 55000,
+  "inner_val_loss": 0.0988250674040179,
+  "inner_val_per_cohort": {
+   "immicro": 0.0973090926806132,
+   "imperial_eskd": 0.09802845560691574,
+   "liu2021": 0.10113765392452478
+  },
+  "train_loss": 0.09831878396868705,
+  "sec": 2929.0
+ },
+ {
+  "step": 55250,
+  "inner_val_loss": 0.09882187805693558,
+  "inner_val_per_cohort": {
+   "immicro": 0.09730332375814517,
+   "imperial_eskd": 0.09802931682630019,
+   "liu2021": 0.10113299358636141
+  },
+  "train_loss": 0.0981254826784134,
+  "sec": 2941.9
+ },
+ {
+  "step": 55500,
+  "inner_val_loss": 0.09882039933305498,
+  "inner_val_per_cohort": {
+   "immicro": 0.09729931938151519,
+   "imperial_eskd": 0.09802832623774355,
+   "liu2021": 0.10113355237990618
+  },
+  "train_loss": 0.09829349279403686,
+  "sec": 2955.0
+ },
+ {
+  "step": 55750,
+  "inner_val_loss": 0.09881699362750908,
+  "inner_val_per_cohort": {
+   "immicro": 0.09729416823635499,
+   "imperial_eskd": 0.0980275192043998,
+   "liu2021": 0.10112929344177246
+  },
+  "train_loss": 0.09823975721001625,
+  "sec": 2969.4
+ },
+ {
+  "step": 56000,
+  "inner_val_loss": 0.09881236883952761,
+  "inner_val_per_cohort": {
+   "immicro": 0.0972876250743866,
+   "imperial_eskd": 0.09802262993021445,
+   "liu2021": 0.10112685151398182
+  },
+  "train_loss": 0.09802509763836861,
+  "sec": 2983.6
+ },
+ {
+  "step": 56250,
+  "inner_val_loss": 0.09880828266643515,
+  "inner_val_per_cohort": {
+   "immicro": 0.0972876160716017,
+   "imperial_eskd": 0.09802142496813428,
+   "liu2021": 0.10111580695956945
+  },
+  "train_loss": 0.09830817812681199,
+  "sec": 2997.9
+ },
+ {
+  "step": 56500,
+  "inner_val_loss": 0.09880374529107351,
+  "inner_val_per_cohort": {
+   "immicro": 0.09728315317382415,
+   "imperial_eskd": 0.09801682525060394,
+   "liu2021": 0.10111125744879246
+  },
+  "train_loss": 0.09792736774682999,
+  "sec": 3011.7
+ },
+ {
+  "step": 56750,
+  "inner_val_loss": 0.09879882260216306,
+  "inner_val_per_cohort": {
+   "immicro": 0.09727994756152232,
+   "imperial_eskd": 0.09801400046456944,
+   "liu2021": 0.10110251978039742
+  },
+  "train_loss": 0.09784514045715333,
+  "sec": 3025.7
+ },
+ {
+  "step": 57000,
+  "inner_val_loss": 0.0987966542103977,
+  "inner_val_per_cohort": {
+   "immicro": 0.09727786760777235,
+   "imperial_eskd": 0.09800964695486156,
+   "liu2021": 0.10110244806855917
+  },
+  "train_loss": 0.09800728607177735,
+  "sec": 3039.3
+ },
+ {
+  "step": 57250,
+  "inner_val_loss": 0.09879568266221368,
+  "inner_val_per_cohort": {
+   "immicro": 0.09727561039229234,
+   "imperial_eskd": 0.0980106798762625,
+   "liu2021": 0.10110075771808624
+  },
+  "train_loss": 0.09833682361245155,
+  "sec": 3053.1
+ },
+ {
+  "step": 57500,
+  "inner_val_loss": 0.09879465937388665,
+  "inner_val_per_cohort": {
+   "immicro": 0.0972708947956562,
+   "imperial_eskd": 0.09800657562234184,
+   "liu2021": 0.10110650770366192
+  },
+  "train_loss": 0.09816291144490243,
+  "sec": 3065.4
+ },
+ {
+  "step": 57750,
+  "inner_val_loss": 0.09879424035398647,
+  "inner_val_per_cohort": {
+   "immicro": 0.09726771743347247,
+   "imperial_eskd": 0.09800512919371779,
+   "liu2021": 0.10110987443476915
+  },
+  "train_loss": 0.09789388850331307,
+  "sec": 3079.0
+ },
+ {
+  "step": 58000,
+  "inner_val_loss": 0.09879699698414164,
+  "inner_val_per_cohort": {
+   "immicro": 0.0972661826138695,
+   "imperial_eskd": 0.09800338643518361,
+   "liu2021": 0.10112142190337181
+  },
+  "train_loss": 0.0981707924604416,
+  "sec": 3091.5
+ },
+ {
+  "step": 58250,
+  "inner_val_loss": 0.09879741336059089,
+  "inner_val_per_cohort": {
+   "immicro": 0.09726652937630813,
+   "imperial_eskd": 0.09800351580435579,
+   "liu2021": 0.10112219490110874
+  },
+  "train_loss": 0.09831579995155335,
+  "sec": 3103.8
+ },
+ {
+  "step": 58500,
+  "inner_val_loss": 0.09880009129869216,
+  "inner_val_per_cohort": {
+   "immicro": 0.09726593643426895,
+   "imperial_eskd": 0.09800414808771828,
+   "liu2021": 0.10113018937408924
+  },
+  "train_loss": 0.09840066722035408,
+  "sec": 3116.2
+ },
+ {
+  "step": 58750,
+  "inner_val_loss": 0.09879708238359954,
+  "inner_val_per_cohort": {
+   "immicro": 0.09726029665519793,
+   "imperial_eskd": 0.09799838066101074,
+   "liu2021": 0.10113256983458996
+  },
+  "train_loss": 0.09860085725784301,
+  "sec": 3128.6
+ },
+ {
+  "step": 59000,
+  "inner_val_loss": 0.09879434848147811,
+  "inner_val_per_cohort": {
+   "immicro": 0.09726120624691248,
+   "imperial_eskd": 0.09799958562309091,
+   "liu2021": 0.10112225357443094
+  },
+  "train_loss": 0.09833379775285721,
+  "sec": 3140.9
+ },
+ {
+  "step": 59250,
+  "inner_val_loss": 0.09878936511549082,
+  "inner_val_per_cohort": {
+   "immicro": 0.09725973568856716,
+   "imperial_eskd": 0.09799611128189346,
+   "liu2021": 0.10111224837601185
+  },
+  "train_loss": 0.09783965268731118,
+  "sec": 3153.3
+ },
+ {
+  "step": 59500,
+  "inner_val_loss": 0.09878567852681935,
+  "inner_val_per_cohort": {
+   "immicro": 0.09725511136154334,
+   "imperial_eskd": 0.09799643538214943,
+   "liu2021": 0.10110548883676529
+  },
+  "train_loss": 0.09812837907671929,
+  "sec": 3167.2
+ },
+ {
+  "step": 59750,
+  "inner_val_loss": 0.09878855574913699,
+  "inner_val_per_cohort": {
+   "immicro": 0.09724919001261394,
+   "imperial_eskd": 0.09800017760558562,
+   "liu2021": 0.10111629962921143
+  },
+  "train_loss": 0.09832096108794212,
+  "sec": 3180.9
+ },
+ {
+  "step": 60000,
+  "inner_val_loss": 0.09878745599828585,
+  "inner_val_per_cohort": {
+   "immicro": 0.09724500371764104,
+   "imperial_eskd": 0.09799756008115681,
+   "liu2021": 0.1011198041960597
+  },
+  "train_loss": 0.0980109176337719,
+  "sec": 3193.3
+ },
+ {
+  "step": 60250,
+  "inner_val_loss": 0.09878520222587718,
+  "inner_val_per_cohort": {
+   "immicro": 0.09724018971125285,
+   "imperial_eskd": 0.09799479320645332,
+   "liu2021": 0.10112062375992537
+  },
+  "train_loss": 0.09817018666863442,
+  "sec": 3207.2
+ },
+ {
+  "step": 60500,
+  "inner_val_loss": 0.09878368990853277,
+  "inner_val_per_cohort": {
+   "immicro": 0.09723748297741015,
+   "imperial_eskd": 0.09799190779978578,
+   "liu2021": 0.1011216789484024
+  },
+  "train_loss": 0.09819755256175995,
+  "sec": 3219.6
+ },
+ {
+  "step": 60750,
+  "inner_val_loss": 0.09877861305281067,
+  "inner_val_per_cohort": {
+   "immicro": 0.09723283536732197,
+   "imperial_eskd": 0.0979856862263246,
+   "liu2021": 0.10111731756478548
+  },
+  "train_loss": 0.09795685276389123,
+  "sec": 3233.4
+ },
+ {
+  "step": 61000,
+  "inner_val_loss": 0.09878035444728654,
+  "inner_val_per_cohort": {
+   "immicro": 0.09723192950089772,
+   "imperial_eskd": 0.09798024086789651,
+   "liu2021": 0.10112889297306538
+  },
+  "train_loss": 0.09832937258481979,
+  "sec": 3247.1
+ },
+ {
+  "step": 61250,
+  "inner_val_loss": 0.09877987513600876,
+  "inner_val_per_cohort": {
+   "immicro": 0.09723023418337107,
+   "imperial_eskd": 0.09798096018758687,
+   "liu2021": 0.10112843103706837
+  },
+  "train_loss": 0.09850014325976372,
+  "sec": 3259.5
+ },
+ {
+  "step": 61500,
+  "inner_val_loss": 0.0987737833926774,
+  "inner_val_per_cohort": {
+   "immicro": 0.09722963316986959,
+   "imperial_eskd": 0.09797967631708492,
+   "liu2021": 0.10111204069107771
+  },
+  "train_loss": 0.0979384484887123,
+  "sec": 3272.2
+ },
+ {
+  "step": 61750,
+  "inner_val_loss": 0.09876753718150083,
+  "inner_val_per_cohort": {
+   "immicro": 0.09722639123598735,
+   "imperial_eskd": 0.09797287258234891,
+   "liu2021": 0.10110334772616625
+  },
+  "train_loss": 0.09821305954456329,
+  "sec": 3286.4
+ },
+ {
+  "step": 62000,
+  "inner_val_loss": 0.09876650188946062,
+  "inner_val_per_cohort": {
+   "immicro": 0.09722437771658103,
+   "imperial_eskd": 0.09797631576657295,
+   "liu2021": 0.10109881218522787
+  },
+  "train_loss": 0.09810027256608009,
+  "sec": 3300.6
+ },
+ {
+  "step": 62250,
+  "inner_val_loss": 0.09876251724934337,
+  "inner_val_per_cohort": {
+   "immicro": 0.09722152662773927,
+   "imperial_eskd": 0.09797327863899144,
+   "liu2021": 0.1010927464812994
+  },
+  "train_loss": 0.09807561931014061,
+  "sec": 3314.8
+ },
+ {
+  "step": 62500,
+  "inner_val_loss": 0.09875776516179545,
+  "inner_val_per_cohort": {
+   "immicro": 0.09721562669922908,
+   "imperial_eskd": 0.09797185320745815,
+   "liu2021": 0.10108581557869911
+  },
+  "train_loss": 0.09825502780079842,
+  "sec": 3329.2
+ },
+ {
+  "step": 62750,
+  "inner_val_loss": 0.09875442343528824,
+  "inner_val_per_cohort": {
+   "immicro": 0.09721372680117686,
+   "imperial_eskd": 0.09797447073188695,
+   "liu2021": 0.10107507277280092
+  },
+  "train_loss": 0.09801244151592255,
+  "sec": 3343.2
+ },
+ {
+  "step": 63000,
+  "inner_val_loss": 0.0987515744819033,
+  "inner_val_per_cohort": {
+   "immicro": 0.09721192500243585,
+   "imperial_eskd": 0.09797511385245757,
+   "liu2021": 0.1010676845908165
+  },
+  "train_loss": 0.09825809690356255,
+  "sec": 3356.9
+ },
+ {
+  "step": 63250,
+  "inner_val_loss": 0.09874890382505125,
+  "inner_val_per_cohort": {
+   "immicro": 0.0972081379344066,
+   "imperial_eskd": 0.09797361120581627,
+   "liu2021": 0.1010649623349309
+  },
+  "train_loss": 0.0984227940440178,
+  "sec": 3370.8
+ },
+ {
+  "step": 63500,
+  "inner_val_loss": 0.09874938595851863,
+  "inner_val_per_cohort": {
+   "immicro": 0.09720515521864097,
+   "imperial_eskd": 0.0979701957919381,
+   "liu2021": 0.10107280686497688
+  },
+  "train_loss": 0.09794382032752037,
+  "sec": 3384.5
+ },
+ {
+  "step": 63750,
+  "inner_val_loss": 0.09874615339430594,
+  "inner_val_per_cohort": {
+   "immicro": 0.0972007221231858,
+   "imperial_eskd": 0.09796689721671017,
+   "liu2021": 0.10107084084302187
+  },
+  "train_loss": 0.09810843953490257,
+  "sec": 3396.9
+ },
+ {
+  "step": 64000,
+  "inner_val_loss": 0.09874401848311677,
+  "inner_val_per_cohort": {
+   "immicro": 0.09720044862478971,
+   "imperial_eskd": 0.09796038507060571,
+   "liu2021": 0.10107122175395489
+  },
+  "train_loss": 0.09818873944878578,
+  "sec": 3410.7
+ },
+ {
+  "step": 64250,
+  "inner_val_loss": 0.09874366123154006,
+  "inner_val_per_cohort": {
+   "immicro": 0.09719546294460694,
+   "imperial_eskd": 0.09795957160266963,
+   "liu2021": 0.10107594914734364
+  },
+  "train_loss": 0.09813238906860351,
+  "sec": 3424.4
+ },
+ {
+  "step": 64500,
+  "inner_val_loss": 0.09874078683141206,
+  "inner_val_per_cohort": {
+   "immicro": 0.0971915094802777,
+   "imperial_eskd": 0.0979553833603859,
+   "liu2021": 0.10107546765357256
+  },
+  "train_loss": 0.09800751334428787,
+  "sec": 3436.8
+ },
+ {
+  "step": 64750,
+  "inner_val_loss": 0.09874133851303897,
+  "inner_val_per_cohort": {
+   "immicro": 0.09718922587732474,
+   "imperial_eskd": 0.09795574945482341,
+   "liu2021": 0.10107904020696878
+  },
+  "train_loss": 0.09813767728209495,
+  "sec": 3450.7
+ },
+ {
+  "step": 65000,
+  "inner_val_loss": 0.0987383018934516,
+  "inner_val_per_cohort": {
+   "immicro": 0.09718427869180839,
+   "imperial_eskd": 0.09795198725028471,
+   "liu2021": 0.1010786397382617
+  },
+  "train_loss": 0.09811814272403717,
+  "sec": 3463.1
+ },
+ {
+  "step": 65250,
+  "inner_val_loss": 0.09873455253928298,
+  "inner_val_per_cohort": {
+   "immicro": 0.09718015355368455,
+   "imperial_eskd": 0.09795094213702461,
+   "liu2021": 0.10107256192713976
+  },
+  "train_loss": 0.09819985979795456,
+  "sec": 3476.8
+ },
+ {
+  "step": 65500,
+  "inner_val_loss": 0.09873133525252342,
+  "inner_val_per_cohort": {
+   "immicro": 0.0971798961982131,
+   "imperial_eskd": 0.09794987738132477,
+   "liu2021": 0.1010642321780324
+  },
+  "train_loss": 0.09787763968110084,
+  "sec": 3490.5
+ },
+ {
+  "step": 65750,
+  "inner_val_loss": 0.09873096310919283,
+  "inner_val_per_cohort": {
+   "immicro": 0.0971755910043915,
+   "imperial_eskd": 0.09795176136222752,
+   "liu2021": 0.10106553696095943
+  },
+  "train_loss": 0.09833774134516717,
+  "sec": 3504.2
+ },
+ {
+  "step": 66000,
+  "inner_val_loss": 0.09872772123175438,
+  "inner_val_per_cohort": {
+   "immicro": 0.09717009806384642,
+   "imperial_eskd": 0.0979453707960519,
+   "liu2021": 0.10106769483536482
+  },
+  "train_loss": 0.09793360170722008,
+  "sec": 3516.6
+ },
+ {
+  "step": 66250,
+  "inner_val_loss": 0.09872727237190261,
+  "inner_val_per_cohort": {
+   "immicro": 0.09716941229999065,
+   "imperial_eskd": 0.097943937913938,
+   "liu2021": 0.10106846690177917
+  },
+  "train_loss": 0.09788276606798171,
+  "sec": 3530.2
+ },
+ {
+  "step": 66500,
+  "inner_val_loss": 0.09872556651820137,
+  "inner_val_per_cohort": {
+   "immicro": 0.09716838722427686,
+   "imperial_eskd": 0.09794047034599564,
+   "liu2021": 0.10106784198433161
+  },
+  "train_loss": 0.09818444231152534,
+  "sec": 3542.6
+ },
+ {
+  "step": 66750,
+  "inner_val_loss": 0.09872128682756665,
+  "inner_val_per_cohort": {
+   "immicro": 0.0971646240601937,
+   "imperial_eskd": 0.09794111278924075,
+   "liu2021": 0.1010581236332655
+  },
+  "train_loss": 0.0979424867928028,
+  "sec": 3556.4
+ },
+ {
+  "step": 67000,
+  "inner_val_loss": 0.09871896256626855,
+  "inner_val_per_cohort": {
+   "immicro": 0.09716054735084374,
+   "imperial_eskd": 0.09793823720379309,
+   "liu2021": 0.10105810314416885
+  },
+  "train_loss": 0.09816659140586853,
+  "sec": 3570.1
+ },
+ {
+  "step": 67250,
+  "inner_val_loss": 0.09871437118360489,
+  "inner_val_per_cohort": {
+   "immicro": 0.09715976100414991,
+   "imperial_eskd": 0.0979348634454337,
+   "liu2021": 0.1010484891012311
+  },
+  "train_loss": 0.09781079259514809,
+  "sec": 3583.1
+ },
+ {
+  "step": 67500,
+  "inner_val_loss": 0.09871385640681091,
+  "inner_val_per_cohort": {
+   "immicro": 0.09715441179772218,
+   "imperial_eskd": 0.09793362800370563,
+   "liu2021": 0.10105352941900492
+  },
+  "train_loss": 0.09813322687149048,
+  "sec": 3595.7
+ },
+ {
+  "step": 67750,
+  "inner_val_loss": 0.09871348822395308,
+  "inner_val_per_cohort": {
+   "immicro": 0.09715433915456136,
+   "imperial_eskd": 0.09793494675647128,
+   "liu2021": 0.10105117876082659
+  },
+  "train_loss": 0.09802013123035432,
+  "sec": 3606.9
+ },
+ {
+  "step": 68000,
+  "inner_val_loss": 0.09871058915788777,
+  "inner_val_per_cohort": {
+   "immicro": 0.09715230328341325,
+   "imperial_eskd": 0.09792968055064027,
+   "liu2021": 0.10104978363960981
+  },
+  "train_loss": 0.09801840949058532,
+  "sec": 3618.1
+ },
+ {
+  "step": 68250,
+  "inner_val_loss": 0.09870832163200836,
+  "inner_val_per_cohort": {
+   "immicro": 0.09714438269535701,
+   "imperial_eskd": 0.0979296744547107,
+   "liu2021": 0.10105090774595737
+  },
+  "train_loss": 0.0979850268959999,
+  "sec": 3630.8
+ },
+ {
+  "step": 68500,
+  "inner_val_loss": 0.09870734272731675,
+  "inner_val_per_cohort": {
+   "immicro": 0.09714530284206073,
+   "imperial_eskd": 0.09792444482445717,
+   "liu2021": 0.10105228051543236
+  },
+  "train_loss": 0.09807486081123352,
+  "sec": 3643.5
+ },
+ {
+  "step": 68750,
+  "inner_val_loss": 0.09870499012182758,
+  "inner_val_per_cohort": {
+   "immicro": 0.0971439133087794,
+   "imperial_eskd": 0.09792048551819542,
+   "liu2021": 0.10105057153850794
+  },
+  "train_loss": 0.09827238562703132,
+  "sec": 3654.7
+ },
+ {
+  "step": 69000,
+  "inner_val_loss": 0.09870556042990601,
+  "inner_val_per_cohort": {
+   "immicro": 0.09714290561775367,
+   "imperial_eskd": 0.0979209037666971,
+   "liu2021": 0.10105287190526724
+  },
+  "train_loss": 0.0979289248585701,
+  "sec": 3667.5
+ },
+ {
+  "step": 69250,
+  "inner_val_loss": 0.0987018308875085,
+  "inner_val_per_cohort": {
+   "immicro": 0.09713721399505933,
+   "imperial_eskd": 0.09792043031616644,
+   "liu2021": 0.10104784835129976
+  },
+  "train_loss": 0.09805905038118362,
+  "sec": 3678.7
+ },
+ {
+  "step": 69500,
+  "inner_val_loss": 0.09869980594764154,
+  "inner_val_per_cohort": {
+   "immicro": 0.09713621344417334,
+   "imperial_eskd": 0.09791703894734383,
+   "liu2021": 0.10104616545140743
+  },
+  "train_loss": 0.0975765284895897,
+  "sec": 3691.5
+ },
+ {
+  "step": 69750,
+  "inner_val_loss": 0.09869737597652757,
+  "inner_val_per_cohort": {
+   "immicro": 0.09713538208355506,
+   "imperial_eskd": 0.0979120975190943,
+   "liu2021": 0.10104464832693338
+  },
+  "train_loss": 0.0981320963203907,
+  "sec": 3704.3
+ },
+ {
+  "step": 70000,
+  "inner_val_loss": 0.09869525055730283,
+  "inner_val_per_cohort": {
+   "immicro": 0.09713213176776965,
+   "imperial_eskd": 0.09791426800868729,
+   "liu2021": 0.10103935189545155
+  },
+  "train_loss": 0.09793427407741546,
+  "sec": 3716.9
+ },
+ {
+  "step": 70250,
+  "inner_val_loss": 0.09869052141415652,
+  "inner_val_per_cohort": {
+   "immicro": 0.09712937970956166,
+   "imperial_eskd": 0.09791360524567691,
+   "liu2021": 0.10102857928723097
+  },
+  "train_loss": 0.09798978480696678,
+  "sec": 3729.8
+ },
+ {
+  "step": 70500,
+  "inner_val_loss": 0.09868562085121269,
+  "inner_val_per_cohort": {
+   "immicro": 0.09712222715218861,
+   "imperial_eskd": 0.09791313044049522,
+   "liu2021": 0.10102150496095419
+  },
+  "train_loss": 0.09802140334248542,
+  "sec": 3742.6
+ },
+ {
+  "step": 70750,
+  "inner_val_loss": 0.09868350442536551,
+  "inner_val_per_cohort": {
+   "immicro": 0.09711696362743776,
+   "imperial_eskd": 0.09790828586979346,
+   "liu2021": 0.10102526377886534
+  },
+  "train_loss": 0.09783445209264756,
+  "sec": 3755.4
+ },
+ {
+  "step": 71000,
+  "inner_val_loss": 0.09868693788244266,
+  "inner_val_per_cohort": {
+   "immicro": 0.09711792164792617,
+   "imperial_eskd": 0.09790834479711273,
+   "liu2021": 0.1010345472022891
+  },
+  "train_loss": 0.09785742855072022,
+  "sec": 3768.1
+ },
+ {
+  "step": 71250,
+  "inner_val_loss": 0.09868256219032438,
+  "inner_val_per_cohort": {
+   "immicro": 0.09711989512046178,
+   "imperial_eskd": 0.09790716286409985,
+   "liu2021": 0.10102062858641148
+  },
+  "train_loss": 0.09839033055305481,
+  "sec": 3779.4
+ },
+ {
+  "step": 71500,
+  "inner_val_loss": 0.09867837551905952,
+  "inner_val_per_cohort": {
+   "immicro": 0.09711781019965808,
+   "imperial_eskd": 0.09790359911593524,
+   "liu2021": 0.10101371724158525
+  },
+  "train_loss": 0.09788489711284637,
+  "sec": 3790.6
+ },
+ {
+  "step": 71750,
+  "inner_val_loss": 0.09867367964012153,
+  "inner_val_per_cohort": {
+   "immicro": 0.09711106028407812,
+   "imperial_eskd": 0.09789976037361404,
+   "liu2021": 0.10101021826267242
+  },
+  "train_loss": 0.09790007448196411,
+  "sec": 3803.5
+ },
+ {
+  "step": 72000,
+  "inner_val_loss": 0.09867111021521115,
+  "inner_val_per_cohort": {
+   "immicro": 0.09710819150010745,
+   "imperial_eskd": 0.09790053862062367,
+   "liu2021": 0.10100460052490234
+  },
+  "train_loss": 0.09806997931003571,
+  "sec": 3816.1
+ },
+ {
+  "step": 72250,
+  "inner_val_loss": 0.09866871208780342,
+  "inner_val_per_cohort": {
+   "immicro": 0.09710601872454087,
+   "imperial_eskd": 0.09789831563830376,
+   "liu2021": 0.10100180190056562
+  },
+  "train_loss": 0.09759360495209694,
+  "sec": 3828.8
+ },
+ {
+  "step": 72500,
+  "inner_val_loss": 0.09866801485645049,
+  "inner_val_per_cohort": {
+   "immicro": 0.09710348024964333,
+   "imperial_eskd": 0.09789958291433075,
+   "liu2021": 0.10100098140537739
+  },
+  "train_loss": 0.0983897641301155,
+  "sec": 3841.3
+ },
+ {
+  "step": 72750,
+  "inner_val_loss": 0.09866694292298168,
+  "inner_val_per_cohort": {
+   "immicro": 0.09709800065805514,
+   "imperial_eskd": 0.09790057079358534,
+   "liu2021": 0.10100225731730461
+  },
+  "train_loss": 0.09802371445298196,
+  "sec": 3852.4
+ },
+ {
+  "step": 73000,
+  "inner_val_loss": 0.09866636766666415,
+  "inner_val_per_cohort": {
+   "immicro": 0.09709585613260667,
+   "imperial_eskd": 0.09789705005559055,
+   "liu2021": 0.10100619681179523
+  },
+  "train_loss": 0.09756857225298882,
+  "sec": 3865.1
+ },
+ {
+  "step": 73250,
+  "inner_val_loss": 0.09866441882243664,
+  "inner_val_per_cohort": {
+   "immicro": 0.09709115605801344,
+   "imperial_eskd": 0.09789411080154506,
+   "liu2021": 0.10100798960775137
+  },
+  "train_loss": 0.09808956289291382,
+  "sec": 3876.5
+ },
+ {
+  "step": 73500,
+  "inner_val_loss": 0.09866429915218944,
+  "inner_val_per_cohort": {
+   "immicro": 0.09708978639294703,
+   "imperial_eskd": 0.09789815104820511,
+   "liu2021": 0.10100496001541615
+  },
+  "train_loss": 0.09805249932408333,
+  "sec": 3889.2
+ },
+ {
+  "step": 73750,
+  "inner_val_loss": 0.09866353008670337,
+  "inner_val_per_cohort": {
+   "immicro": 0.09709097910672426,
+   "imperial_eskd": 0.09789820692755959,
+   "liu2021": 0.10100140422582626
+  },
+  "train_loss": 0.0981896645128727,
+  "sec": 3900.6
+ },
+ {
+  "step": 74000,
+  "inner_val_loss": 0.09866024046722387,
+  "inner_val_per_cohort": {
+   "immicro": 0.09708684186140697,
+   "imperial_eskd": 0.09789790585637093,
+   "liu2021": 0.10099597368389368
+  },
+  "train_loss": 0.09764138665795326,
+  "sec": 3911.9
+ },
+ {
+  "step": 74250,
+  "inner_val_loss": 0.09865934915389075,
+  "inner_val_per_cohort": {
+   "immicro": 0.09708333387970924,
+   "imperial_eskd": 0.09789835526184602,
+   "liu2021": 0.100996358320117
+  },
+  "train_loss": 0.09789109951257706,
+  "sec": 3924.9
+ },
+ {
+  "step": 74500,
+  "inner_val_loss": 0.09865661321038549,
+  "inner_val_per_cohort": {
+   "immicro": 0.09707927238196135,
+   "imperial_eskd": 0.0978948487476869,
+   "liu2021": 0.10099571850150824
+  },
+  "train_loss": 0.09798041167855263,
+  "sec": 3936.3
+ },
+ {
+  "step": 74750,
+  "inner_val_loss": 0.09865638597708459,
+  "inner_val_per_cohort": {
+   "immicro": 0.09707510254035394,
+   "imperial_eskd": 0.09789322113448923,
+   "liu2021": 0.1010008342564106
+  },
+  "train_loss": 0.09793853056430817,
+  "sec": 3949.2
+ },
+ {
+  "step": 75000,
+  "inner_val_loss": 0.09865715206045696,
+  "inner_val_per_cohort": {
+   "immicro": 0.09707465736816327,
+   "imperial_eskd": 0.09788866578177972,
+   "liu2021": 0.10100813303142786
+  },
+  "train_loss": 0.09821679082512856,
+  "sec": 3960.4
+ },
+ {
+  "step": 75250,
+  "inner_val_loss": 0.09865728176829187,
+  "inner_val_per_cohort": {
+   "immicro": 0.09707546575615804,
+   "imperial_eskd": 0.09788929704915393,
+   "liu2021": 0.1010070824995637
+  },
+  "train_loss": 0.09771319851279259,
+  "sec": 3971.7
+ },
+ {
+  "step": 75500,
+  "inner_val_loss": 0.09865210229775519,
+  "inner_val_per_cohort": {
+   "immicro": 0.09707325262327988,
+   "imperial_eskd": 0.09788770736618475,
+   "liu2021": 0.10099534690380096
+  },
+  "train_loss": 0.0980025395154953,
+  "sec": 3982.8
+ },
+ {
+  "step": 75750,
+  "inner_val_loss": 0.09865095864304087,
+  "inner_val_per_cohort": {
+   "immicro": 0.09706966858357191,
+   "imperial_eskd": 0.09788979386741464,
+   "liu2021": 0.10099341347813606
+  },
+  "train_loss": 0.0978697558939457,
+  "sec": 3995.4
+ },
+ {
+  "step": 76000,
+  "inner_val_loss": 0.0986475521002454,
+  "inner_val_per_cohort": {
+   "immicro": 0.09706118858108918,
+   "imperial_eskd": 0.0978890870782462,
+   "liu2021": 0.10099238064140081
+  },
+  "train_loss": 0.09792924776673317,
+  "sec": 4008.4
+ },
+ {
+  "step": 76250,
+  "inner_val_loss": 0.09864933747384284,
+  "inner_val_per_cohort": {
+   "immicro": 0.09706026657174031,
+   "imperial_eskd": 0.09788528457283974,
+   "liu2021": 0.10100246127694845
+  },
+  "train_loss": 0.09806085476279258,
+  "sec": 4021.1
+ },
+ {
+  "step": 76500,
+  "inner_val_loss": 0.09864423260318511,
+  "inner_val_per_cohort": {
+   "immicro": 0.09705591946840286,
+   "imperial_eskd": 0.09788245030424812,
+   "liu2021": 0.10099432803690434
+  },
+  "train_loss": 0.09783285173773766,
+  "sec": 4032.5
+ },
+ {
+  "step": 76750,
+  "inner_val_loss": 0.09864302696377943,
+  "inner_val_per_cohort": {
+   "immicro": 0.09705597534775734,
+   "imperial_eskd": 0.09788267043503848,
+   "liu2021": 0.10099043510854244
+  },
+  "train_loss": 0.09791825327277183,
+  "sec": 4045.5
+ },
+ {
+  "step": 77000,
+  "inner_val_loss": 0.09864494978741865,
+  "inner_val_per_cohort": {
+   "immicro": 0.09705436198661725,
+   "imperial_eskd": 0.09788262403824112,
+   "liu2021": 0.10099786333739758
+  },
+  "train_loss": 0.09818209558725358,
+  "sec": 4058.3
+ },
+ {
+  "step": 77250,
+  "inner_val_loss": 0.09864340428112432,
+  "inner_val_per_cohort": {
+   "immicro": 0.09705233698089917,
+   "imperial_eskd": 0.09788361022418196,
+   "liu2021": 0.10099426563829184
+  },
+  "train_loss": 0.09786362525820733,
+  "sec": 4069.7
+ },
+ {
+  "step": 77500,
+  "inner_val_loss": 0.09864145002770003,
+  "inner_val_per_cohort": {
+   "immicro": 0.09704915092637141,
+   "imperial_eskd": 0.09788162362846461,
+   "liu2021": 0.10099357552826405
+  },
+  "train_loss": 0.09802731189131737,
+  "sec": 4081.0
+ },
+ {
+  "step": 77750,
+  "inner_val_loss": 0.09864143955737653,
+  "inner_val_per_cohort": {
+   "immicro": 0.09704495252420504,
+   "imperial_eskd": 0.09788199548016895,
+   "liu2021": 0.1009973706677556
+  },
+  "train_loss": 0.09787329968810081,
+  "sec": 4093.9
+ },
+ {
+  "step": 78000,
+  "inner_val_loss": 0.09863878225888868,
+  "inner_val_per_cohort": {
+   "immicro": 0.09704190182189147,
+   "imperial_eskd": 0.09787894345142624,
+   "liu2021": 0.10099550150334835
+  },
+  "train_loss": 0.09775854107737542,
+  "sec": 4105.3
+ },
+ {
+  "step": 78250,
+  "inner_val_loss": 0.09863756851979881,
+  "inner_val_per_cohort": {
+   "immicro": 0.09704121916244428,
+   "imperial_eskd": 0.09787767617539926,
+   "liu2021": 0.10099381022155285
+  },
+  "train_loss": 0.09770897263288499,
+  "sec": 4118.1
+ },
+ {
+  "step": 78500,
+  "inner_val_loss": 0.0986361646027577,
+  "inner_val_per_cohort": {
+   "immicro": 0.09703741719325383,
+   "imperial_eskd": 0.09787607802586122,
+   "liu2021": 0.10099499858915806
+  },
+  "train_loss": 0.0980480664074421,
+  "sec": 4131.0
+ },
+ {
+  "step": 78750,
+  "inner_val_loss": 0.09863141497051475,
+  "inner_val_per_cohort": {
+   "immicro": 0.09703541112442811,
+   "imperial_eskd": 0.09787689047780904,
+   "liu2021": 0.1009819433093071
+  },
+  "train_loss": 0.09776280176639557,
+  "sec": 4143.9
+ },
+ {
+  "step": 79000,
+  "inner_val_loss": 0.09863170039736563,
+  "inner_val_per_cohort": {
+   "immicro": 0.097032703148822,
+   "imperial_eskd": 0.09787523746490479,
+   "liu2021": 0.1009871605783701
+  },
+  "train_loss": 0.09778807091712952,
+  "sec": 4157.0
+ },
+ {
+  "step": 79250,
+  "inner_val_loss": 0.09863066171869787,
+  "inner_val_per_cohort": {
+   "immicro": 0.09703565171609323,
+   "imperial_eskd": 0.09787579829042609,
+   "liu2021": 0.10098053514957428
+  },
+  "train_loss": 0.0979487863779068,
+  "sec": 4168.4
+ },
+ {
+  "step": 79500,
+  "inner_val_loss": 0.09862994228611992,
+  "inner_val_per_cohort": {
+   "immicro": 0.09703122234592836,
+   "imperial_eskd": 0.09787312962792137,
+   "liu2021": 0.10098547488451004
+  },
+  "train_loss": 0.09769246509671212,
+  "sec": 4179.7
+ },
+ {
+  "step": 79750,
+  "inner_val_loss": 0.09862653923343227,
+  "inner_val_per_cohort": {
+   "immicro": 0.09702484930555026,
+   "imperial_eskd": 0.09787381135604599,
+   "liu2021": 0.10098095703870058
+  },
+  "train_loss": 0.09806352087855338,
+  "sec": 4192.5
+ },
+ {
+  "step": 80000,
+  "inner_val_loss": 0.0986245580092825,
+  "inner_val_per_cohort": {
+   "immicro": 0.09701966991027196,
+   "imperial_eskd": 0.09787348014387218,
+   "liu2021": 0.10098052397370338
+  },
+  "train_loss": 0.09787627491354943,
+  "sec": 4205.4
+ },
+ {
+  "step": 80250,
+  "inner_val_loss": 0.09862065094175061,
+  "inner_val_per_cohort": {
+   "immicro": 0.09701832321782906,
+   "imperial_eskd": 0.09786901115016504,
+   "liu2021": 0.10097461845725775
+  },
+  "train_loss": 0.09795867884159089,
+  "sec": 4219.8
+ },
+ {
+  "step": 80500,
+  "inner_val_loss": 0.09861848395167276,
+  "inner_val_per_cohort": {
+   "immicro": 0.09701793019970258,
+   "imperial_eskd": 0.09786468574946577,
+   "liu2021": 0.10097283590584993
+  },
+  "train_loss": 0.09786568042635918,
+  "sec": 4232.6
+ },
+ {
+  "step": 80750,
+  "inner_val_loss": 0.09861303877198335,
+  "inner_val_per_cohort": {
+   "immicro": 0.09701216034591198,
+   "imperial_eskd": 0.09786016155372966,
+   "liu2021": 0.1009667944163084
+  },
+  "train_loss": 0.09786264842748642,
+  "sec": 4245.0
+ },
+ {
+  "step": 81000,
+  "inner_val_loss": 0.09860619870625964,
+  "inner_val_per_cohort": {
+   "immicro": 0.09700716224809487,
+   "imperial_eskd": 0.09785688634623181,
+   "liu2021": 0.10095454752445221
+  },
+  "train_loss": 0.09790713214874268,
+  "sec": 4257.4
+ },
+ {
+  "step": 81250,
+  "inner_val_loss": 0.09860415199790339,
+  "inner_val_per_cohort": {
+   "immicro": 0.0970061703895529,
+   "imperial_eskd": 0.0978534563698552,
+   "liu2021": 0.10095282923430204
+  },
+  "train_loss": 0.09752543717622757,
+  "sec": 4269.9
+ },
+ {
+  "step": 81500,
+  "inner_val_loss": 0.09860628273225192,
+  "inner_val_per_cohort": {
+   "immicro": 0.09700870886445045,
+   "imperial_eskd": 0.09785717691887509,
+   "liu2021": 0.10095296241343021
+  },
+  "train_loss": 0.09803544801473618,
+  "sec": 4282.2
+ },
+ {
+  "step": 81750,
+  "inner_val_loss": 0.09860436197821841,
+  "inner_val_per_cohort": {
+   "immicro": 0.09700579475611448,
+   "imperial_eskd": 0.09785652668638663,
+   "liu2021": 0.10095076449215412
+  },
+  "train_loss": 0.09796465265750885,
+  "sec": 4293.5
+ },
+ {
+  "step": 82000,
+  "inner_val_loss": 0.09860430816846966,
+  "inner_val_per_cohort": {
+   "immicro": 0.09700784894327323,
+   "imperial_eskd": 0.09785765917463736,
+   "liu2021": 0.10094741638749838
+  },
+  "train_loss": 0.0977952418923378,
+  "sec": 4304.9
+ },
+ {
+  "step": 82250,
+  "inner_val_loss": 0.09860129022237026,
+  "inner_val_per_cohort": {
+   "immicro": 0.09700677823275328,
+   "imperial_eskd": 0.09785554524172436,
+   "liu2021": 0.10094154719263315
+  },
+  "train_loss": 0.09782877919077873,
+  "sec": 4316.3
+ },
+ {
+  "step": 82500,
+  "inner_val_loss": 0.09860078361111158,
+  "inner_val_per_cohort": {
+   "immicro": 0.09700426179915667,
+   "imperial_eskd": 0.09785530411384323,
+   "liu2021": 0.10094278492033482
+  },
+  "train_loss": 0.09790671980381012,
+  "sec": 4328.8
+ },
+ {
+  "step": 82750,
+  "inner_val_loss": 0.09859797511851848,
+  "inner_val_per_cohort": {
+   "immicro": 0.09700514065722625,
+   "imperial_eskd": 0.09784738651730797,
+   "liu2021": 0.10094139818102121
+  },
+  "train_loss": 0.09786559960246086,
+  "sec": 4340.1
+ },
+ {
+  "step": 83000,
+  "inner_val_loss": 0.09859620506941069,
+  "inner_val_per_cohort": {
+   "immicro": 0.0970022954667608,
+   "imperial_eskd": 0.0978456275029616,
+   "liu2021": 0.10094069223850965
+  },
+  "train_loss": 0.09767872953414918,
+  "sec": 4353.0
+ },
+ {
+  "step": 83250,
+  "inner_val_loss": 0.09859569729164695,
+  "inner_val_per_cohort": {
+   "immicro": 0.09699977468699217,
+   "imperial_eskd": 0.09784284843639894,
+   "liu2021": 0.10094446875154972
+  },
+  "train_loss": 0.09775797381997109,
+  "sec": 4365.6
+ },
+ {
+  "step": 83500,
+  "inner_val_loss": 0.09859544473389785,
+  "inner_val_per_cohort": {
+   "immicro": 0.096994923427701,
+   "imperial_eskd": 0.09783979877829552,
+   "liu2021": 0.10095161199569702
+  },
+  "train_loss": 0.09759168645739555,
+  "sec": 4376.7
+ },
+ {
+  "step": 83750,
+  "inner_val_loss": 0.09859687630839721,
+  "inner_val_per_cohort": {
+   "immicro": 0.09699100193878014,
+   "imperial_eskd": 0.0978414649990472,
+   "liu2021": 0.10095816198736429
+  },
+  "train_loss": 0.09796833893656731,
+  "sec": 4387.9
+ },
+ {
+  "step": 84000,
+  "inner_val_loss": 0.09859317395309307,
+  "inner_val_per_cohort": {
+   "immicro": 0.09698755356172721,
+   "imperial_eskd": 0.09784378450025212,
+   "liu2021": 0.10094818379729986
+  },
+  "train_loss": 0.09757011738419533,
+  "sec": 4399.3
+ },
+ {
+  "step": 84250,
+  "inner_val_loss": 0.09859326930547302,
+  "inner_val_per_cohort": {
+   "immicro": 0.09698647819459438,
+   "imperial_eskd": 0.09784512357278304,
+   "liu2021": 0.10094820614904165
+  },
+  "train_loss": 0.097543835490942,
+  "sec": 4412.1
+ },
+ {
+  "step": 84500,
+  "inner_val_loss": 0.09858839892114353,
+  "inner_val_per_cohort": {
+   "immicro": 0.09698266784350078,
+   "imperial_eskd": 0.09784312749450857,
+   "liu2021": 0.10093940142542124
+  },
+  "train_loss": 0.09750837728381156,
+  "sec": 4423.3
+ },
+ {
+  "step": 84750,
+  "inner_val_loss": 0.09858471287809538,
+  "inner_val_per_cohort": {
+   "immicro": 0.09697768464684486,
+   "imperial_eskd": 0.09784164957024834,
+   "liu2021": 0.10093480441719294
+  },
+  "train_loss": 0.0979091195166111,
+  "sec": 4436.0
+ },
+ {
+  "step": 85000,
+  "inner_val_loss": 0.09858406029378224,
+  "inner_val_per_cohort": {
+   "immicro": 0.09697655712564786,
+   "imperial_eskd": 0.09784278815442865,
+   "liu2021": 0.1009328356012702
+  },
+  "train_loss": 0.09785534381866455,
+  "sec": 4448.8
+ },
+ {
+  "step": 85250,
+  "inner_val_loss": 0.09858176212833085,
+  "inner_val_per_cohort": {
+   "immicro": 0.0969736644377311,
+   "imperial_eskd": 0.09784117984500798,
+   "liu2021": 0.10093044210225344
+  },
+  "train_loss": 0.09800766435265541,
+  "sec": 4460.0
+ },
+ {
+  "step": 85500,
+  "inner_val_loss": 0.09857958088619541,
+  "inner_val_per_cohort": {
+   "immicro": 0.09697120108952124,
+   "imperial_eskd": 0.09784194420684468,
+   "liu2021": 0.10092559736222029
+  },
+  "train_loss": 0.09783341547846794,
+  "sec": 4472.7
+ },
+ {
+  "step": 85750,
+  "inner_val_loss": 0.09857853629033675,
+  "inner_val_per_cohort": {
+   "immicro": 0.09696814604103565,
+   "imperial_eskd": 0.09783766892823306,
+   "liu2021": 0.1009297939017415
+  },
+  "train_loss": 0.09802905884385109,
+  "sec": 4485.3
+ },
+ {
+  "step": 86000,
+  "inner_val_loss": 0.09857798272725006,
+  "inner_val_per_cohort": {
+   "immicro": 0.09696222624431054,
+   "imperial_eskd": 0.09783637455918572,
+   "liu2021": 0.10093534737825394
+  },
+  "train_loss": 0.09783361014723778,
+  "sec": 4498.2
+ },
+ {
+  "step": 86250,
+  "inner_val_loss": 0.0985757408550743,
+  "inner_val_per_cohort": {
+   "immicro": 0.0969610819593072,
+   "imperial_eskd": 0.09782964397560466,
+   "liu2021": 0.10093649663031101
+  },
+  "train_loss": 0.09788893556594849,
+  "sec": 4509.4
+ },
+ {
+  "step": 86500,
+  "inner_val_loss": 0.09857320852314282,
+  "inner_val_per_cohort": {
+   "immicro": 0.09695816909273465,
+   "imperial_eskd": 0.09782466461712663,
+   "liu2021": 0.10093679185956717
+  },
+  "train_loss": 0.09766834554076195,
+  "sec": 4522.1
+ },
+ {
+  "step": 86750,
+  "inner_val_loss": 0.09857013309844816,
+  "inner_val_per_cohort": {
+   "immicro": 0.09695522704472144,
+   "imperial_eskd": 0.09782262722199614,
+   "liu2021": 0.10093254502862692
+  },
+  "train_loss": 0.0973420412838459,
+  "sec": 4535.0
+ },
+ {
+  "step": 87000,
+  "inner_val_loss": 0.0985678348107019,
+  "inner_val_per_cohort": {
+   "immicro": 0.09695098890612523,
+   "imperial_eskd": 0.09782028807835146,
+   "liu2021": 0.10093222744762897
+  },
+  "train_loss": 0.09778893661499023,
+  "sec": 4547.8
+ },
+ {
+  "step": 87250,
+  "inner_val_loss": 0.09856601784739531,
+  "inner_val_per_cohort": {
+   "immicro": 0.0969480387866497,
+   "imperial_eskd": 0.09782268513332713,
+   "liu2021": 0.10092732962220907
+  },
+  "train_loss": 0.09767526134848595,
+  "sec": 4560.8
+ },
+ {
+  "step": 87500,
+  "inner_val_loss": 0.09856249656377718,
+  "inner_val_per_cohort": {
+   "immicro": 0.09694458947827418,
+   "imperial_eskd": 0.09782071614807303,
+   "liu2021": 0.10092218406498432
+  },
+  "train_loss": 0.09778589436411858,
+  "sec": 4573.5
+ },
+ {
+  "step": 87750,
+  "inner_val_loss": 0.0985582826866044,
+  "inner_val_per_cohort": {
+   "immicro": 0.09693915831545989,
+   "imperial_eskd": 0.09782218560576439,
+   "liu2021": 0.1009135041385889
+  },
+  "train_loss": 0.09777923738956451,
+  "sec": 4586.3
+ },
+ {
+  "step": 88000,
+  "inner_val_loss": 0.0985593365674669,
+  "inner_val_per_cohort": {
+   "immicro": 0.09693565871566534,
+   "imperial_eskd": 0.09782005236907439,
+   "liu2021": 0.100922298617661
+  },
+  "train_loss": 0.09785063162446023,
+  "sec": 4599.0
+ },
+ {
+  "step": 88250,
+  "inner_val_loss": 0.09855851102996656,
+  "inner_val_per_cohort": {
+   "immicro": 0.09693683652828138,
+   "imperial_eskd": 0.09781621261076494,
+   "liu2021": 0.10092248395085335
+  },
+  "train_loss": 0.09763687801361084,
+  "sec": 4610.2
+ },
+ {
+  "step": 88500,
+  "inner_val_loss": 0.09855394158512354,
+  "inner_val_per_cohort": {
+   "immicro": 0.09693338349461555,
+   "imperial_eskd": 0.09781394898891449,
+   "liu2021": 0.10091449227184057
+  },
+  "train_loss": 0.09769515189528465,
+  "sec": 4621.4
+ },
+ {
+  "step": 88750,
+  "inner_val_loss": 0.0985521406142248,
+  "inner_val_per_cohort": {
+   "immicro": 0.09693071867028873,
+   "imperial_eskd": 0.09781026840209961,
+   "liu2021": 0.10091543477028608
+  },
+  "train_loss": 0.09768850433826447,
+  "sec": 4634.3
+ },
+ {
+  "step": 89000,
+  "inner_val_loss": 0.09854978855435896,
+  "inner_val_per_cohort": {
+   "immicro": 0.09692822241534789,
+   "imperial_eskd": 0.09781058301979845,
+   "liu2021": 0.10091056022793055
+  },
+  "train_loss": 0.0980936042368412,
+  "sec": 4646.9
+ },
+ {
+  "step": 89250,
+  "inner_val_loss": 0.09854641857773366,
+  "inner_val_per_cohort": {
+   "immicro": 0.09692865827431281,
+   "imperial_eskd": 0.09780627522956241,
+   "liu2021": 0.10090432222932577
+  },
+  "train_loss": 0.09812609913945199,
+  "sec": 4659.9
+ },
+ {
+  "step": 89500,
+  "inner_val_loss": 0.09854354099793868,
+  "inner_val_per_cohort": {
+   "immicro": 0.096924664452672,
+   "imperial_eskd": 0.09780589897524226,
+   "liu2021": 0.10090005956590176
+  },
+  "train_loss": 0.09786764001846314,
+  "sec": 4672.6
+ },
+ {
+  "step": 89750,
+  "inner_val_loss": 0.0985410903119529,
+  "inner_val_per_cohort": {
+   "immicro": 0.09692125519116719,
+   "imperial_eskd": 0.0977995673363859,
+   "liu2021": 0.10090244840830564
+  },
+  "train_loss": 0.09782806870341301,
+  "sec": 4685.5
+ },
+ {
+  "step": 90000,
+  "inner_val_loss": 0.09853932724306076,
+  "inner_val_per_cohort": {
+   "immicro": 0.09691951982676983,
+   "imperial_eskd": 0.09779897230592641,
+   "liu2021": 0.10089948959648609
+  },
+  "train_loss": 0.09767733636498452,
+  "sec": 4698.2
+ },
+ {
+  "step": 90250,
+  "inner_val_loss": 0.09853938530490856,
+  "inner_val_per_cohort": {
+   "immicro": 0.09691326009730498,
+   "imperial_eskd": 0.09779745645143768,
+   "liu2021": 0.10090743936598301
+  },
+  "train_loss": 0.0975845418870449,
+  "sec": 4710.8
+ },
+ {
+  "step": 90500,
+  "inner_val_loss": 0.09853509117407028,
+  "inner_val_per_cohort": {
+   "immicro": 0.09691067008922498,
+   "imperial_eskd": 0.09779521721330556,
+   "liu2021": 0.10089938621968031
+  },
+  "train_loss": 0.09778028839826584,
+  "sec": 4722.1
+ },
+ {
+  "step": 90750,
+  "inner_val_loss": 0.09853454822241658,
+  "inner_val_per_cohort": {
+   "immicro": 0.09690741666903098,
+   "imperial_eskd": 0.09779919988729736,
+   "liu2021": 0.10089702811092138
+  },
+  "train_loss": 0.09761189424991608,
+  "sec": 4734.8
+ },
+ {
+  "step": 91000,
+  "inner_val_loss": 0.09853256679130624,
+  "inner_val_per_cohort": {
+   "immicro": 0.09690524668743213,
+   "imperial_eskd": 0.09779715876687657,
+   "liu2021": 0.10089529491961002
+  },
+  "train_loss": 0.09787127241492272,
+  "sec": 4746.0
+ },
+ {
+  "step": 91250,
+  "inner_val_loss": 0.09853120517888754,
+  "inner_val_per_cohort": {
+   "immicro": 0.09690126404166222,
+   "imperial_eskd": 0.09779030355540189,
+   "liu2021": 0.10090204793959856
+  },
+  "train_loss": 0.09767388996481896,
+  "sec": 4758.8
+ },
+ {
+  "step": 91500,
+  "inner_val_loss": 0.09853034382780092,
+  "inner_val_per_cohort": {
+   "immicro": 0.09690059597293536,
+   "imperial_eskd": 0.09779420833696019,
+   "liu2021": 0.10089622717350721
+  },
+  "train_loss": 0.09760789778828621,
+  "sec": 4771.4
+ },
+ {
+  "step": 91750,
+  "inner_val_loss": 0.0985298863790854,
+  "inner_val_per_cohort": {
+   "immicro": 0.09689963577936093,
+   "imperial_eskd": 0.09779390794309703,
+   "liu2021": 0.10089611541479826
+  },
+  "train_loss": 0.09753952518105508,
+  "sec": 4782.6
+ },
+ {
+  "step": 92000,
+  "inner_val_loss": 0.09852893525414697,
+  "inner_val_per_cohort": {
+   "immicro": 0.09689879448463519,
+   "imperial_eskd": 0.09779243509877812,
+   "liu2021": 0.10089557617902756
+  },
+  "train_loss": 0.09772189596295357,
+  "sec": 4795.3
+ },
+ {
+  "step": 92250,
+  "inner_val_loss": 0.09853113210299098,
+  "inner_val_per_cohort": {
+   "immicro": 0.09689737701167662,
+   "imperial_eskd": 0.09779281372373755,
+   "liu2021": 0.10090320557355881
+  },
+  "train_loss": 0.09770521026849746,
+  "sec": 4806.6
+ },
+ {
+  "step": 92500,
+  "inner_val_loss": 0.09853009417690713,
+  "inner_val_per_cohort": {
+   "immicro": 0.09689468859384458,
+   "imperial_eskd": 0.0977914458648725,
+   "liu2021": 0.10090414807200432
+  },
+  "train_loss": 0.09753981378674507,
+  "sec": 4817.8
+ },
+ {
+  "step": 92750,
+  "inner_val_loss": 0.09853242606752449,
+  "inner_val_per_cohort": {
+   "immicro": 0.09689279055843751,
+   "imperial_eskd": 0.09779052436351776,
+   "liu2021": 0.10091396328061819
+  },
+  "train_loss": 0.0975836234986782,
+  "sec": 4829.0
+ },
+ {
+  "step": 93000,
+  "inner_val_loss": 0.09853158830994307,
+  "inner_val_per_cohort": {
+   "immicro": 0.0968912619476517,
+   "imperial_eskd": 0.09778996252200821,
+   "liu2021": 0.10091354046016932
+  },
+  "train_loss": 0.09774570935964584,
+  "sec": 4840.2
+ },
+ {
+  "step": 93250,
+  "inner_val_loss": 0.0985312517262017,
+  "inner_val_per_cohort": {
+   "immicro": 0.09689089904228847,
+   "imperial_eskd": 0.09779300743883307,
+   "liu2021": 0.10090984869748354
+  },
+  "train_loss": 0.0976133326292038,
+  "sec": 4851.3
+ },
+ {
+  "step": 93500,
+  "inner_val_loss": 0.09852930475402662,
+  "inner_val_per_cohort": {
+   "immicro": 0.09688589038948218,
+   "imperial_eskd": 0.09779000146822496,
+   "liu2021": 0.10091202240437269
+  },
+  "train_loss": 0.09777968034148216,
+  "sec": 4862.5
+ },
+ {
+  "step": 93750,
+  "inner_val_loss": 0.09852548306713803,
+  "inner_val_per_cohort": {
+   "immicro": 0.09687881575276454,
+   "imperial_eskd": 0.09778852794658054,
+   "liu2021": 0.100909105502069
+  },
+  "train_loss": 0.09750416800379753,
+  "sec": 4873.7
+ },
+ {
+  "step": 94000,
+  "inner_val_loss": 0.09852617436248545,
+  "inner_val_per_cohort": {
+   "immicro": 0.09687787511696418,
+   "imperial_eskd": 0.0977888015860861,
+   "liu2021": 0.10091184638440609
+  },
+  "train_loss": 0.09806689384579659,
+  "sec": 4886.5
+ },
+ {
+  "step": 94250,
+  "inner_val_loss": 0.0985256779393313,
+  "inner_val_per_cohort": {
+   "immicro": 0.09687571537991364,
+   "imperial_eskd": 0.09779091280969707,
+   "liu2021": 0.10091040562838316
+  },
+  "train_loss": 0.09752861762046813,
+  "sec": 4897.7
+ },
+ {
+  "step": 94500,
+  "inner_val_loss": 0.09852011275073193,
+  "inner_val_per_cohort": {
+   "immicro": 0.0968688086917003,
+   "imperial_eskd": 0.09778990630399097,
+   "liu2021": 0.10090162325650454
+  },
+  "train_loss": 0.09783034372329712,
+  "sec": 4908.9
+ },
+ {
+  "step": 94750,
+  "inner_val_loss": 0.09851750885747901,
+  "inner_val_per_cohort": {
+   "immicro": 0.09686653626461823,
+   "imperial_eskd": 0.09778413142670285,
+   "liu2021": 0.10090185888111591
+  },
+  "train_loss": 0.0975282646715641,
+  "sec": 4921.7
+ },
+ {
+  "step": 95000,
+  "inner_val_loss": 0.09851860218314511,
+  "inner_val_per_cohort": {
+   "immicro": 0.09686224721372128,
+   "imperial_eskd": 0.09778832441026514,
+   "liu2021": 0.1009052349254489
+  },
+  "train_loss": 0.09778225031495094,
+  "sec": 4934.4
+ },
+ {
+  "step": 95250,
+  "inner_val_loss": 0.0985142547693668,
+  "inner_val_per_cohort": {
+   "immicro": 0.09685833938419819,
+   "imperial_eskd": 0.09778563847595995,
+   "liu2021": 0.10089878644794226
+  },
+  "train_loss": 0.09755124178528786,
+  "sec": 4945.6
+ },
+ {
+  "step": 95500,
+  "inner_val_loss": 0.09850919137548919,
+  "inner_val_per_cohort": {
+   "immicro": 0.09685961777965228,
+   "imperial_eskd": 0.09777741676027124,
+   "liu2021": 0.10089053958654404
+  },
+  "train_loss": 0.09764039066433906,
+  "sec": 4958.5
+ },
+ {
+  "step": 95750,
+  "inner_val_loss": 0.09850888717167004,
+  "inner_val_per_cohort": {
+   "immicro": 0.09685732579479615,
+   "imperial_eskd": 0.09778117286888036,
+   "liu2021": 0.10088816285133362
+  },
+  "train_loss": 0.09778859773278237,
+  "sec": 4971.2
+ },
+ {
+  "step": 96000,
+  "inner_val_loss": 0.09851202507964289,
+  "inner_val_per_cohort": {
+   "immicro": 0.09685649598638217,
+   "imperial_eskd": 0.09778418933803384,
+   "liu2021": 0.10089538991451263
+  },
+  "train_loss": 0.0974822907447815,
+  "sec": 4982.4
+ },
+ {
+  "step": 96250,
+  "inner_val_loss": 0.09851016157842947,
+  "inner_val_per_cohort": {
+   "immicro": 0.09685508627444506,
+   "imperial_eskd": 0.09777646579525688,
+   "liu2021": 0.10089893266558647
+  },
+  "train_loss": 0.09790710997581482,
+  "sec": 4993.7
+ },
+ {
+  "step": 96500,
+  "inner_val_loss": 0.0985094445729346,
+  "inner_val_per_cohort": {
+   "immicro": 0.0968502014875412,
+   "imperial_eskd": 0.09777331826361743,
+   "liu2021": 0.10090481396764517
+  },
+  "train_loss": 0.09771997573971748,
+  "sec": 5004.8
+ },
+ {
+  "step": 96750,
+  "inner_val_loss": 0.09850939613475342,
+  "inner_val_per_cohort": {
+   "immicro": 0.09685286631186803,
+   "imperial_eskd": 0.09777094118974426,
+   "liu2021": 0.10090438090264797
+  },
+  "train_loss": 0.09736123844981194,
+  "sec": 5016.0
+ },
+ {
+  "step": 97000,
+  "inner_val_loss": 0.09850626055038336,
+  "inner_val_per_cohort": {
+   "immicro": 0.09685125388205051,
+   "imperial_eskd": 0.09776627983559262,
+   "liu2021": 0.10090124793350697
+  },
+  "train_loss": 0.09749823731184006,
+  "sec": 5027.2
+ },
+ {
+  "step": 97250,
+  "inner_val_loss": 0.09850332515333036,
+  "inner_val_per_cohort": {
+   "immicro": 0.09685117316742738,
+   "imperial_eskd": 0.09776568853042343,
+   "liu2021": 0.10089311376214027
+  },
+  "train_loss": 0.0976054849922657,
+  "sec": 5039.9
+ },
+ {
+  "step": 97500,
+  "inner_val_loss": 0.09850159959133827,
+  "inner_val_per_cohort": {
+   "immicro": 0.09684739727526903,
+   "imperial_eskd": 0.09776808660138737,
+   "liu2021": 0.10088931489735842
+  },
+  "train_loss": 0.0974549426138401,
+  "sec": 5052.7
+ },
+ {
+  "step": 97750,
+  "inner_val_loss": 0.09849907767328651,
+  "inner_val_per_cohort": {
+   "immicro": 0.09684516893078883,
+   "imperial_eskd": 0.09776424955237996,
+   "liu2021": 0.10088781453669071
+  },
+  "train_loss": 0.09790222036838532,
+  "sec": 5065.4
+ },
+ {
+  "step": 98000,
+  "inner_val_loss": 0.0984996637103684,
+  "inner_val_per_cohort": {
+   "immicro": 0.09684627037495375,
+   "imperial_eskd": 0.09776538678190926,
+   "liu2021": 0.10088733397424221
+  },
+  "train_loss": 0.0979027073085308,
+  "sec": 5078.3
+ },
+ {
+  "step": 98250,
+  "inner_val_loss": 0.09849937979809263,
+  "inner_val_per_cohort": {
+   "immicro": 0.09684596303850412,
+   "imperial_eskd": 0.09776199812238867,
+   "liu2021": 0.10089017823338509
+  },
+  "train_loss": 0.09758543148636818,
+  "sec": 5089.5
+ },
+ {
+  "step": 98500,
+  "inner_val_loss": 0.09849834820312081,
+  "inner_val_per_cohort": {
+   "immicro": 0.09684764873236418,
+   "imperial_eskd": 0.09775957295840437,
+   "liu2021": 0.10088782291859388
+  },
+  "train_loss": 0.09753463271260261,
+  "sec": 5100.6
+ },
+ {
+  "step": 98750,
+  "inner_val_loss": 0.09849576467666964,
+  "inner_val_per_cohort": {
+   "immicro": 0.09684224085261424,
+   "imperial_eskd": 0.09775797345421532,
+   "liu2021": 0.10088707972317934
+  },
+  "train_loss": 0.09747278833389282,
+  "sec": 5111.9
+ },
+ {
+  "step": 99000,
+  "inner_val_loss": 0.09849123795977748,
+  "inner_val_per_cohort": {
+   "immicro": 0.09684265467027824,
+   "imperial_eskd": 0.09775198047811334,
+   "liu2021": 0.10087907873094082
+  },
+  "train_loss": 0.0979715781211853,
+  "sec": 5124.6
+ },
+ {
+  "step": 99250,
+  "inner_val_loss": 0.09848882974774549,
+  "inner_val_per_cohort": {
+   "immicro": 0.09684002958238125,
+   "imperial_eskd": 0.09774875369938937,
+   "liu2021": 0.10087770596146584
+  },
+  "train_loss": 0.09767379099130631,
+  "sec": 5137.5
+ },
+ {
+  "step": 99500,
+  "inner_val_loss": 0.09848832480157867,
+  "inner_val_per_cohort": {
+   "immicro": 0.09684260841459036,
+   "imperial_eskd": 0.09774881093339487,
+   "liu2021": 0.10087355505675077
+  },
+  "train_loss": 0.09736916217207908,
+  "sec": 5150.2
+ },
+ {
+  "step": 99750,
+  "inner_val_loss": 0.09848699825956965,
+  "inner_val_per_cohort": {
+   "immicro": 0.09684146661311388,
+   "imperial_eskd": 0.09774781526489691,
+   "liu2021": 0.10087171290069818
+  },
+  "train_loss": 0.09754687738418578,
+  "sec": 5161.5
+ },
+ {
+  "step": 100000,
+  "inner_val_loss": 0.09848370757706539,
+  "inner_val_per_cohort": {
+   "immicro": 0.09684028134991725,
+   "imperial_eskd": 0.09774973717602817,
+   "liu2021": 0.10086110420525074
+  },
+  "train_loss": 0.09789957010746002,
+  "sec": 5174.4
+ },
+ {
+  "step": 100250,
+  "inner_val_loss": 0.0984808960553221,
+  "inner_val_per_cohort": {
+   "immicro": 0.09683635955055554,
+   "imperial_eskd": 0.09774579344825311,
+   "liu2021": 0.10086053516715765
+  },
+  "train_loss": 0.09748698419332505,
+  "sec": 5188.9
+ },
+ {
+  "step": 100500,
+  "inner_val_loss": 0.09848036401613493,
+  "inner_val_per_cohort": {
+   "immicro": 0.09683156323929627,
+   "imperial_eskd": 0.0977463952519677,
+   "liu2021": 0.10086313355714083
+  },
+  "train_loss": 0.09743034511804581,
+  "sec": 5201.8
+ },
+ {
+  "step": 100750,
+  "inner_val_loss": 0.098477548505697,
+  "inner_val_per_cohort": {
+   "immicro": 0.09682461774597566,
+   "imperial_eskd": 0.09774316847324371,
+   "liu2021": 0.10086485929787159
+  },
+  "train_loss": 0.09776061871647834,
+  "sec": 5213.0
+ },
+ {
+  "step": 101000,
+  "inner_val_loss": 0.09847740901427136,
+  "inner_val_per_cohort": {
+   "immicro": 0.09682292615373929,
+   "imperial_eskd": 0.09774335101246834,
+   "liu2021": 0.10086594987660646
+  },
+  "train_loss": 0.09736412745714187,
+  "sec": 5225.6
+ },
+ {
+  "step": 101250,
+  "inner_val_loss": 0.0984756883817038,
+  "inner_val_per_cohort": {
+   "immicro": 0.0968207645540436,
+   "imperial_eskd": 0.09774414212866263,
+   "liu2021": 0.1008621584624052
+  },
+  "train_loss": 0.0974842500090599,
+  "sec": 5236.9
+ },
+ {
+  "step": 101500,
+  "inner_val_loss": 0.09847186406077157,
+  "inner_val_per_cohort": {
+   "immicro": 0.09681719200064738,
+   "imperial_eskd": 0.09774456681175665,
+   "liu2021": 0.10085383336991072
+  },
+  "train_loss": 0.09747539511322975,
+  "sec": 5249.6
+ },
+ {
+  "step": 101750,
+  "inner_val_loss": 0.09847264530870951,
+  "inner_val_per_cohort": {
+   "immicro": 0.09681562334299088,
+   "imperial_eskd": 0.09774529222737659,
+   "liu2021": 0.10085702035576105
+  },
+  "train_loss": 0.09758950871229172,
+  "sec": 5262.2
+ },
+ {
+  "step": 102000,
+  "inner_val_loss": 0.09847222517874805,
+  "inner_val_per_cohort": {
+   "immicro": 0.09681535481164853,
+   "imperial_eskd": 0.09773830730806697,
+   "liu2021": 0.1008630134165287
+  },
+  "train_loss": 0.09733420678973198,
+  "sec": 5273.4
+ },
+ {
+  "step": 102250,
+  "inner_val_loss": 0.09847394410859456,
+  "inner_val_per_cohort": {
+   "immicro": 0.09681741893291473,
+   "imperial_eskd": 0.09773649004372684,
+   "liu2021": 0.10086792334914207
+  },
+  "train_loss": 0.09740324333310127,
+  "sec": 5284.6
+ },
+ {
+  "step": 102500,
+  "inner_val_loss": 0.0984733302165011,
+  "inner_val_per_cohort": {
+   "immicro": 0.0968136874337991,
+   "imperial_eskd": 0.09773899919607422,
+   "liu2021": 0.10086730401962996
+  },
+  "train_loss": 0.09761379772424698,
+  "sec": 5295.9
+ },
+ {
+  "step": 102750,
+  "inner_val_loss": 0.09847290793226825,
+  "inner_val_per_cohort": {
+   "immicro": 0.09681208649029334,
+   "imperial_eskd": 0.09773612394928932,
+   "liu2021": 0.10087051335722208
+  },
+  "train_loss": 0.09754966101050377,
+  "sec": 5307.1
+ },
+ {
+  "step": 103000,
+  "inner_val_loss": 0.09847019328688733,
+  "inner_val_per_cohort": {
+   "immicro": 0.09680718804399173,
+   "imperial_eskd": 0.09773397648876364,
+   "liu2021": 0.10086941532790661
+  },
+  "train_loss": 0.09748036316037179,
+  "sec": 5318.4
+ },
+ {
+  "step": 103250,
+  "inner_val_loss": 0.09846551782178758,
+  "inner_val_per_cohort": {
+   "immicro": 0.09680256589005391,
+   "imperial_eskd": 0.09773245589299635,
+   "liu2021": 0.10086153168231249
+  },
+  "train_loss": 0.09761263960599899,
+  "sec": 5330.8
+ },
+ {
+  "step": 103500,
+  "inner_val_loss": 0.0984611256668965,
+  "inner_val_per_cohort": {
+   "immicro": 0.09679730981588364,
+   "imperial_eskd": 0.09773428365588188,
+   "liu2021": 0.10085178352892399
+  },
+  "train_loss": 0.09754696294665337,
+  "sec": 5343.4
+ },
+ {
+  "step": 103750,
+  "inner_val_loss": 0.09846092880035591,
+  "inner_val_per_cohort": {
+   "immicro": 0.0967983693505327,
+   "imperial_eskd": 0.09773341024463827,
+   "liu2021": 0.10085100680589676
+  },
+  "train_loss": 0.09750343400239944,
+  "sec": 5356.1
+ },
+ {
+  "step": 104000,
+  "inner_val_loss": 0.09846050097522412,
+  "inner_val_per_cohort": {
+   "immicro": 0.09679809771478176,
+   "imperial_eskd": 0.09773344986818054,
+   "liu2021": 0.10084995534271002
+  },
+  "train_loss": 0.09769925731420517,
+  "sec": 5367.3
+ },
+ {
+  "step": 104250,
+  "inner_val_loss": 0.09845882728507722,
+  "inner_val_per_cohort": {
+   "immicro": 0.09679480052242677,
+   "imperial_eskd": 0.09773204915902832,
+   "liu2021": 0.10084963217377663
+  },
+  "train_loss": 0.09751059681177139,
+  "sec": 5378.5
+ },
+ {
+  "step": 104500,
+  "inner_val_loss": 0.09845842796406061,
+  "inner_val_per_cohort": {
+   "immicro": 0.0967913568019867,
+   "imperial_eskd": 0.09773520109328357,
+   "liu2021": 0.10084872599691153
+  },
+  "train_loss": 0.09788672640919685,
+  "sec": 5391.0
+ },
+ {
+  "step": 104750,
+  "inner_val_loss": 0.09845466180845643,
+  "inner_val_per_cohort": {
+   "immicro": 0.09678826853632927,
+   "imperial_eskd": 0.09773276509209113,
+   "liu2021": 0.10084295179694891
+  },
+  "train_loss": 0.09755891466140747,
+  "sec": 5403.1
+ },
+ {
+  "step": 105000,
+  "inner_val_loss": 0.0984538568823476,
+  "inner_val_per_cohort": {
+   "immicro": 0.09678564344843228,
+   "imperial_eskd": 0.0977315411648967,
+   "liu2021": 0.10084438603371382
+  },
+  "train_loss": 0.09740580049157142,
+  "sec": 5417.6
+ },
+ {
+  "step": 105250,
+  "inner_val_loss": 0.09845390621422216,
+  "inner_val_per_cohort": {
+   "immicro": 0.09678298172851403,
+   "imperial_eskd": 0.09773518720811064,
+   "liu2021": 0.10084354970604181
+  },
+  "train_loss": 0.09759688270092011,
+  "sec": 5430.2
+ },
+ {
+  "step": 105500,
+  "inner_val_loss": 0.09844662377968279,
+  "inner_val_per_cohort": {
+   "immicro": 0.09678043921788533,
+   "imperial_eskd": 0.09773185645992105,
+   "liu2021": 0.10082757566124201
+  },
+  "train_loss": 0.0971875582933426,
+  "sec": 5442.9
+ },
+ {
+  "step": 105750,
+  "inner_val_loss": 0.09844339274885978,
+  "inner_val_per_cohort": {
+   "immicro": 0.09677845053374767,
+   "imperial_eskd": 0.09772990576245567,
+   "liu2021": 0.10082182195037603
+  },
+  "train_loss": 0.09746175074577332,
+  "sec": 5457.1
+ },
+ {
+  "step": 106000,
+  "inner_val_loss": 0.09844162689540724,
+  "inner_val_per_cohort": {
+   "immicro": 0.09677678284545739,
+   "imperial_eskd": 0.09772964634678581,
+   "liu2021": 0.1008184514939785
+  },
+  "train_loss": 0.0974730320572853,
+  "sec": 5470.7
+ },
+ {
+  "step": 106250,
+  "inner_val_loss": 0.09844094316352799,
+  "inner_val_per_cohort": {
+   "immicro": 0.09676915251960357,
+   "imperial_eskd": 0.0977307998321273,
+   "liu2021": 0.10082287713885307
+  },
+  "train_loss": 0.09783329808712006,
+  "sec": 5484.3
+ },
+ {
+  "step": 106500,
+  "inner_val_loss": 0.09843902954963422,
+  "inner_val_per_cohort": {
+   "immicro": 0.09676528194298346,
+   "imperial_eskd": 0.09772813049229709,
+   "liu2021": 0.1008236762136221
+  },
+  "train_loss": 0.09748738092184067,
+  "sec": 5496.8
+ },
+ {
+  "step": 106750,
+  "inner_val_loss": 0.09843838632558331,
+  "inner_val_per_cohort": {
+   "immicro": 0.09676318801939487,
+   "imperial_eskd": 0.09772633151574568,
+   "liu2021": 0.10082563944160938
+  },
+  "train_loss": 0.09779859420657158,
+  "sec": 5510.7
+ },
+ {
+  "step": 107000,
+  "inner_val_loss": 0.09843894230634576,
+  "inner_val_per_cohort": {
+   "immicro": 0.09676385081062715,
+   "imperial_eskd": 0.0977309017696164,
+   "liu2021": 0.10082207433879375
+  },
+  "train_loss": 0.0975262331366539,
+  "sec": 5523.1
+ },
+ {
+  "step": 107250,
+  "inner_val_loss": 0.09844109405659968,
+  "inner_val_per_cohort": {
+   "immicro": 0.09676343916604917,
+   "imperial_eskd": 0.09773139655590057,
+   "liu2021": 0.10082844644784927
+  },
+  "train_loss": 0.09736341938376426,
+  "sec": 5535.5
+ },
+ {
+  "step": 107500,
+  "inner_val_loss": 0.09844100352075665,
+  "inner_val_per_cohort": {
+   "immicro": 0.09675982097784679,
+   "imperial_eskd": 0.09773054752837528,
+   "liu2021": 0.10083264205604792
+  },
+  "train_loss": 0.09730451023578644,
+  "sec": 5547.9
+ },
+ {
+  "step": 107750,
+  "inner_val_loss": 0.09844242819029876,
+  "inner_val_per_cohort": {
+   "immicro": 0.09675861305246751,
+   "imperial_eskd": 0.0977322557433085,
+   "liu2021": 0.10083641577512026
+  },
+  "train_loss": 0.0973878130018711,
+  "sec": 5560.3
+ },
+ {
+  "step": 108000,
+  "inner_val_loss": 0.09844089416090888,
+  "inner_val_per_cohort": {
+   "immicro": 0.09675573899100225,
+   "imperial_eskd": 0.09772747484120456,
+   "liu2021": 0.10083946865051985
+  },
+  "train_loss": 0.09753494846820832,
+  "sec": 5572.7
+ },
+ {
+  "step": 108250,
+  "inner_val_loss": 0.09844207160664026,
+  "inner_val_per_cohort": {
+   "immicro": 0.09675646635393302,
+   "imperial_eskd": 0.09772835942831906,
+   "liu2021": 0.1008413890376687
+  },
+  "train_loss": 0.09761437058448791,
+  "sec": 5585.1
+ },
+ {
+  "step": 108500,
+  "inner_val_loss": 0.0984419562543432,
+  "inner_val_per_cohort": {
+   "immicro": 0.09675698447972536,
+   "imperial_eskd": 0.09772469475865364,
+   "liu2021": 0.10084418952465057
+  },
+  "train_loss": 0.0975486449599266,
+  "sec": 5597.5
+ }
+]
\ No newline at end of file
diff --git a/paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/milestones.json b/paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/milestones.json
new file mode 100644
index 0000000000000000000000000000000000000000..da2a9614d8a6946e68bdaf0f2d18d8b57b458cb5
--- /dev/null
+++ b/paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/milestones.json
@@ -0,0 +1,62 @@
+[
+ {
+  "milestone_step": 20000,
+  "selected_step": 20000,
+  "selected_inner_val_loss": 0.09932147273135306,
+  "block_improvement": 0.008954538295816897,
+  "total_improvement": 0.008954538295816897,
+  "reason": "block_end",
+  "checkpoint": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/checkpoint_ft_s20000.pt",
+  "checkpoint_sha256": "235563551146480bcf95b57d9456cd702fa24aaee2a58e987df763a37d9e09a0"
+ },
+ {
+  "milestone_step": 40000,
+  "selected_step": 40000,
+  "selected_inner_val_loss": 0.09898765926070587,
+  "block_improvement": 0.00033381347064719014,
+  "total_improvement": 0.009288351766464087,
+  "reason": "block_end",
+  "checkpoint": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/checkpoint_ft_s40000.pt",
+  "checkpoint_sha256": "be3484436f57293db0bed630a4e032fc4a8a37f86e2f3cdbd92f0d207e8e2deb"
+ },
+ {
+  "milestone_step": 60000,
+  "selected_step": 59500,
+  "selected_inner_val_loss": 0.09878567852681935,
+  "block_improvement": 0.0002019807338865115,
+  "total_improvement": 0.009490332500350598,
+  "reason": "block_end",
+  "checkpoint": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/checkpoint_ft_s60000.pt",
+  "checkpoint_sha256": "10def9a99a30b3a32b9027023d0bc8382fa779c3ae0e7df4d1d1dc96ed27e056"
+ },
+ {
+  "milestone_step": 80000,
+  "selected_step": 80000,
+  "selected_inner_val_loss": 0.0986245580092825,
+  "block_improvement": 0.00016112051753684742,
+  "total_improvement": 0.009651453017887446,
+  "reason": "block_end",
+  "checkpoint": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/checkpoint_ft_s80000.pt",
+  "checkpoint_sha256": "9621e005c021486583ee7b97cc0a77bfadf57e5457f28b476d568777efb06b09"
+ },
+ {
+  "milestone_step": 100000,
+  "selected_step": 100000,
+  "selected_inner_val_loss": 0.09848370757706539,
+  "block_improvement": 0.00014085043221712112,
+  "total_improvement": 0.009792303450104567,
+  "reason": "block_end",
+  "checkpoint": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/checkpoint_ft_s100000.pt",
+  "checkpoint_sha256": "3cde1860b41af88ef71cc29e7d7bb29371c30d6c7d12a5b7ba2765cbaa84e4e3"
+ },
+ {
+  "milestone_step": 108500,
+  "selected_step": 106500,
+  "selected_inner_val_loss": 0.09843902954963422,
+  "block_improvement": 4.4678027431160894e-05,
+  "total_improvement": 0.009836981477535728,
+  "reason": "early_stop",
+  "checkpoint": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/checkpoint_ft_s108500.pt",
+  "checkpoint_sha256": "33441f5f75cb45b7c704f2e744ec842ad1371e46206c951443fbed98c7ea6095"
+ }
+]
\ No newline at end of file
diff --git a/paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/train.log b/paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/train.log
new file mode 100644
index 0000000000000000000000000000000000000000..c53f50e5f4a917fbad2196c8c57f432715687e16
--- /dev/null
+++ b/paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/train.log
@@ -0,0 +1,443 @@
+2026-10-03T16:37:15 split=hoL_ye2026_sepsis K=1000 mode=full trainable=54,185,184 balance=cohort immicro: train 95 rows / 347,061 cells, inner 12; imperial_eskd: train 146 rows / 462,172 cells, inner 11; giroux2022: train 24 rows / 254,250 cells, inner 0; liu2021: train 40 rows / 147,532 cells, inner 4
+2026-10-03T16:37:18 step 0 inner-val loss 0.10828 (pretrained) {'immicro': 0.10636159498244524, 'imperial_eskd': 0.1086193590678952, 'liu2021': 0.10984707903116941}
+2026-10-03T16:37:31 step 250 train 0.10386 inner-val 0.10650 best 0.10650@250 (11s) immicro=0.1047 imperial_eskd=0.1066 liu2021=0.1081
+2026-10-03T16:37:44 step 500 train 0.10127 inner-val 0.10434 best 0.10434@500 (25s) immicro=0.1029 imperial_eskd=0.1041 liu2021=0.1060
+2026-10-03T16:37:58 step 750 train 0.10090 inner-val 0.10292 best 0.10292@750 (38s) immicro=0.1018 imperial_eskd=0.1024 liu2021=0.1046
+2026-10-03T16:38:13 step 1000 train 0.10053 inner-val 0.10204 best 0.10204@1000 (53s) immicro=0.1010 imperial_eskd=0.1013 liu2021=0.1038
+2026-10-03T16:38:27 step 1250 train 0.10057 inner-val 0.10147 best 0.10147@1250 (68s) immicro=0.1005 imperial_eskd=0.1006 liu2021=0.1033
+2026-10-03T16:38:40 step 1500 train 0.10029 inner-val 0.10110 best 0.10110@1500 (81s) immicro=0.1001 imperial_eskd=0.1002 liu2021=0.1030
+2026-10-03T16:38:54 step 1750 train 0.10038 inner-val 0.10084 best 0.10084@1750 (95s) immicro=0.0998 imperial_eskd=0.1000 liu2021=0.1027
+2026-10-03T16:39:08 step 2000 train 0.10010 inner-val 0.10066 best 0.10066@2000 (109s) immicro=0.0996 imperial_eskd=0.0998 liu2021=0.1026
+2026-10-03T16:39:20 step 2250 train 0.10021 inner-val 0.10052 best 0.10052@2250 (121s) immicro=0.0994 imperial_eskd=0.0996 liu2021=0.1025
+2026-10-03T16:39:33 step 2500 train 0.10044 inner-val 0.10041 best 0.10041@2500 (133s) immicro=0.0993 imperial_eskd=0.0995 liu2021=0.1024
+2026-10-03T16:39:46 step 2750 train 0.10003 inner-val 0.10033 best 0.10033@2750 (146s) immicro=0.0992 imperial_eskd=0.0994 liu2021=0.1024
+2026-10-03T16:39:59 step 3000 train 0.09998 inner-val 0.10026 best 0.10026@3000 (159s) immicro=0.0991 imperial_eskd=0.0993 liu2021=0.1023
+2026-10-03T16:40:12 step 3250 train 0.10017 inner-val 0.10020 best 0.10020@3250 (172s) immicro=0.0990 imperial_eskd=0.0993 liu2021=0.1023
+2026-10-03T16:40:24 step 3500 train 0.10012 inner-val 0.10015 best 0.10015@3500 (185s) immicro=0.0990 imperial_eskd=0.0992 liu2021=0.1022
+2026-10-03T16:40:37 step 3750 train 0.09997 inner-val 0.10011 best 0.10011@3750 (197s) immicro=0.0989 imperial_eskd=0.0992 liu2021=0.1022
+2026-10-03T16:40:50 step 4000 train 0.09991 inner-val 0.10007 best 0.10007@4000 (210s) immicro=0.0989 imperial_eskd=0.0991 liu2021=0.1022
+2026-10-03T16:41:02 step 4250 train 0.09992 inner-val 0.10003 best 0.10003@4250 (223s) immicro=0.0988 imperial_eskd=0.0991 liu2021=0.1022
+2026-10-03T16:41:15 step 4500 train 0.09985 inner-val 0.10000 best 0.10000@4500 (235s) immicro=0.0988 imperial_eskd=0.0991 liu2021=0.1022
+2026-10-03T16:41:27 step 4750 train 0.09973 inner-val 0.09997 best 0.09997@4750 (248s) immicro=0.0987 imperial_eskd=0.0990 liu2021=0.1021
+2026-10-03T16:41:40 step 5000 train 0.09989 inner-val 0.09995 best 0.09995@5000 (260s) immicro=0.0987 imperial_eskd=0.0990 liu2021=0.1021
+2026-10-03T16:41:52 step 5250 train 0.09984 inner-val 0.09993 best 0.09993@5250 (273s) immicro=0.0987 imperial_eskd=0.0990 liu2021=0.1021
+2026-10-03T16:42:05 step 5500 train 0.09986 inner-val 0.09990 best 0.09990@5500 (286s) immicro=0.0986 imperial_eskd=0.0990 liu2021=0.1021
+2026-10-03T16:42:18 step 5750 train 0.09942 inner-val 0.09988 best 0.09988@5750 (298s) immicro=0.0986 imperial_eskd=0.0990 liu2021=0.1021
+2026-10-03T16:42:31 step 6000 train 0.09970 inner-val 0.09986 best 0.09986@6000 (311s) immicro=0.0986 imperial_eskd=0.0989 liu2021=0.1021
+2026-10-03T16:42:44 step 6250 train 0.09974 inner-val 0.09984 best 0.09984@6250 (325s) immicro=0.0986 imperial_eskd=0.0989 liu2021=0.1021
+2026-10-03T16:42:57 step 6500 train 0.09981 inner-val 0.09982 best 0.09982@6500 (338s) immicro=0.0985 imperial_eskd=0.0989 liu2021=0.1020
+2026-10-03T16:43:10 step 6750 train 0.09966 inner-val 0.09980 best 0.09980@6750 (350s) immicro=0.0985 imperial_eskd=0.0989 liu2021=0.1020
+2026-10-03T16:43:22 step 7000 train 0.09958 inner-val 0.09978 best 0.09978@7000 (363s) immicro=0.0985 imperial_eskd=0.0988 liu2021=0.1020
+2026-10-03T16:43:35 step 7250 train 0.09981 inner-val 0.09977 best 0.09977@7250 (376s) immicro=0.0985 imperial_eskd=0.0988 liu2021=0.1020
+2026-10-03T16:43:48 step 7500 train 0.09960 inner-val 0.09975 best 0.09975@7500 (388s) immicro=0.0985 imperial_eskd=0.0988 liu2021=0.1020
+2026-10-03T16:44:01 step 7750 train 0.09957 inner-val 0.09974 best 0.09974@7750 (401s) immicro=0.0984 imperial_eskd=0.0988 liu2021=0.1020
+2026-10-03T16:44:13 step 8000 train 0.09927 inner-val 0.09972 best 0.09972@8000 (414s) immicro=0.0984 imperial_eskd=0.0988 liu2021=0.1020
+2026-10-03T16:44:26 step 8250 train 0.09972 inner-val 0.09971 best 0.09971@8250 (427s) immicro=0.0984 imperial_eskd=0.0988 liu2021=0.1020
+2026-10-03T16:44:39 step 8500 train 0.09930 inner-val 0.09970 best 0.09970@8500 (439s) immicro=0.0984 imperial_eskd=0.0988 liu2021=0.1019
+2026-10-03T16:44:51 step 8750 train 0.09975 inner-val 0.09968 best 0.09968@8750 (452s) immicro=0.0984 imperial_eskd=0.0987 liu2021=0.1019
+2026-10-03T16:45:04 step 9000 train 0.09955 inner-val 0.09967 best 0.09967@9000 (464s) immicro=0.0984 imperial_eskd=0.0987 liu2021=0.1019
+2026-10-03T16:45:17 step 9250 train 0.09957 inner-val 0.09966 best 0.09966@9250 (477s) immicro=0.0983 imperial_eskd=0.0987 liu2021=0.1019
+2026-10-03T16:45:30 step 9500 train 0.09946 inner-val 0.09965 best 0.09965@9500 (490s) immicro=0.0983 imperial_eskd=0.0987 liu2021=0.1019
+2026-10-03T16:45:42 step 9750 train 0.09952 inner-val 0.09964 best 0.09964@9750 (503s) immicro=0.0983 imperial_eskd=0.0987 liu2021=0.1019
+2026-10-03T16:45:55 step 10000 train 0.09944 inner-val 0.09963 best 0.09963@10000 (515s) immicro=0.0983 imperial_eskd=0.0987 liu2021=0.1019
+2026-10-03T16:46:07 step 10250 train 0.09958 inner-val 0.09962 best 0.09962@10250 (528s) immicro=0.0983 imperial_eskd=0.0987 liu2021=0.1019
+2026-10-03T16:46:20 step 10500 train 0.09963 inner-val 0.09961 best 0.09961@10500 (540s) immicro=0.0983 imperial_eskd=0.0987 liu2021=0.1019
+2026-10-03T16:46:32 step 10750 train 0.09943 inner-val 0.09960 best 0.09960@10750 (553s) immicro=0.0983 imperial_eskd=0.0987 liu2021=0.1019
+2026-10-03T16:46:45 step 11000 train 0.09944 inner-val 0.09959 best 0.09959@11000 (565s) immicro=0.0983 imperial_eskd=0.0987 liu2021=0.1019
+2026-10-03T16:46:58 step 11250 train 0.09934 inner-val 0.09958 best 0.09958@11250 (578s) immicro=0.0982 imperial_eskd=0.0986 liu2021=0.1019
+2026-10-03T16:47:10 step 11500 train 0.09941 inner-val 0.09957 best 0.09957@11500 (591s) immicro=0.0982 imperial_eskd=0.0986 liu2021=0.1018
+2026-10-03T16:47:23 step 11750 train 0.09929 inner-val 0.09956 best 0.09956@11750 (603s) immicro=0.0982 imperial_eskd=0.0986 liu2021=0.1018
+2026-10-03T16:47:35 step 12000 train 0.09909 inner-val 0.09955 best 0.09955@12000 (616s) immicro=0.0982 imperial_eskd=0.0986 liu2021=0.1018
+2026-10-03T16:47:48 step 12250 train 0.09928 inner-val 0.09954 best 0.09954@12250 (629s) immicro=0.0982 imperial_eskd=0.0986 liu2021=0.1018
+2026-10-03T16:48:01 step 12500 train 0.09944 inner-val 0.09953 best 0.09953@12500 (641s) immicro=0.0982 imperial_eskd=0.0986 liu2021=0.1018
+2026-10-03T16:48:13 step 12750 train 0.09938 inner-val 0.09952 best 0.09952@12750 (654s) immicro=0.0982 imperial_eskd=0.0986 liu2021=0.1018
+2026-10-03T16:48:26 step 13000 train 0.09946 inner-val 0.09951 best 0.09951@13000 (667s) immicro=0.0982 imperial_eskd=0.0986 liu2021=0.1018
+2026-10-03T16:48:39 step 13250 train 0.09942 inner-val 0.09950 best 0.09950@13250 (680s) immicro=0.0982 imperial_eskd=0.0986 liu2021=0.1018
+2026-10-03T16:48:52 step 13500 train 0.09928 inner-val 0.09950 best 0.09950@13500 (692s) immicro=0.0981 imperial_eskd=0.0986 liu2021=0.1018
+2026-10-03T16:49:05 step 13750 train 0.09937 inner-val 0.09949 best 0.09949@13750 (705s) immicro=0.0981 imperial_eskd=0.0985 liu2021=0.1018
+2026-10-03T16:49:18 step 14000 train 0.09921 inner-val 0.09948 best 0.09948@14000 (718s) immicro=0.0981 imperial_eskd=0.0985 liu2021=0.1018
+2026-10-03T16:49:31 step 14250 train 0.09910 inner-val 0.09947 best 0.09947@14250 (731s) immicro=0.0981 imperial_eskd=0.0985 liu2021=0.1018
+2026-10-03T16:49:44 step 14500 train 0.09944 inner-val 0.09946 best 0.09946@14500 (744s) immicro=0.0981 imperial_eskd=0.0985 liu2021=0.1018
+2026-10-03T16:49:57 step 14750 train 0.09931 inner-val 0.09945 best 0.09945@14750 (757s) immicro=0.0981 imperial_eskd=0.0985 liu2021=0.1018
+2026-10-03T16:50:09 step 15000 train 0.09940 inner-val 0.09945 best 0.09945@15000 (770s) immicro=0.0981 imperial_eskd=0.0985 liu2021=0.1018
+2026-10-03T16:50:22 step 15250 train 0.09915 inner-val 0.09944 best 0.09944@15250 (783s) immicro=0.0981 imperial_eskd=0.0985 liu2021=0.1017
+2026-10-03T16:50:35 step 15500 train 0.09935 inner-val 0.09943 best 0.09943@15500 (796s) immicro=0.0981 imperial_eskd=0.0985 liu2021=0.1017
+2026-10-03T16:50:48 step 15750 train 0.09908 inner-val 0.09943 best 0.09943@15750 (808s) immicro=0.0981 imperial_eskd=0.0985 liu2021=0.1017
+2026-10-03T16:51:01 step 16000 train 0.09914 inner-val 0.09942 best 0.09942@16000 (821s) immicro=0.0981 imperial_eskd=0.0985 liu2021=0.1017
+2026-10-03T16:51:14 step 16250 train 0.09923 inner-val 0.09941 best 0.09941@16250 (834s) immicro=0.0980 imperial_eskd=0.0985 liu2021=0.1017
+2026-10-03T16:51:26 step 16500 train 0.09933 inner-val 0.09941 best 0.09941@16500 (847s) immicro=0.0980 imperial_eskd=0.0985 liu2021=0.1017
+2026-10-03T16:51:39 step 16750 train 0.09904 inner-val 0.09940 best 0.09940@16750 (860s) immicro=0.0980 imperial_eskd=0.0985 liu2021=0.1017
+2026-10-03T16:51:52 step 17000 train 0.09935 inner-val 0.09939 best 0.09939@17000 (872s) immicro=0.0980 imperial_eskd=0.0985 liu2021=0.1017
+2026-10-03T16:52:04 step 17250 train 0.09943 inner-val 0.09939 best 0.09939@17250 (885s) immicro=0.0980 imperial_eskd=0.0985 liu2021=0.1017
+2026-10-03T16:52:17 step 17500 train 0.09925 inner-val 0.09938 best 0.09938@17500 (898s) immicro=0.0980 imperial_eskd=0.0985 liu2021=0.1017
+2026-10-03T16:52:30 step 17750 train 0.09897 inner-val 0.09937 best 0.09937@17750 (911s) immicro=0.0980 imperial_eskd=0.0984 liu2021=0.1017
+2026-10-03T16:52:43 step 18000 train 0.09953 inner-val 0.09937 best 0.09937@18000 (923s) immicro=0.0980 imperial_eskd=0.0984 liu2021=0.1017
+2026-10-03T16:52:55 step 18250 train 0.09895 inner-val 0.09936 best 0.09936@18250 (936s) immicro=0.0980 imperial_eskd=0.0984 liu2021=0.1017
+2026-10-03T16:53:08 step 18500 train 0.09925 inner-val 0.09936 best 0.09936@18500 (949s) immicro=0.0980 imperial_eskd=0.0984 liu2021=0.1017
+2026-10-03T16:53:21 step 18750 train 0.09902 inner-val 0.09935 best 0.09935@18750 (961s) immicro=0.0980 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:53:33 step 19000 train 0.09913 inner-val 0.09934 best 0.09934@19000 (974s) immicro=0.0980 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:53:46 step 19250 train 0.09903 inner-val 0.09934 best 0.09934@19250 (987s) immicro=0.0980 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:53:59 step 19500 train 0.09901 inner-val 0.09933 best 0.09933@19500 (999s) immicro=0.0980 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:54:11 step 19750 train 0.09895 inner-val 0.09933 best 0.09933@19750 (1012s) immicro=0.0979 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:54:24 step 20000 train 0.09916 inner-val 0.09932 best 0.09932@20000 (1024s) immicro=0.0979 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:54:25 block end 20000: best 0.09932@20000; block improvement 0.008955 (100.00 % of total 0.008955) -> continue
+2026-10-03T16:54:38 step 20250 train 0.09909 inner-val 0.09932 best 0.09932@20250 (1038s) immicro=0.0979 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:54:51 step 20500 train 0.09878 inner-val 0.09931 best 0.09931@20500 (1051s) immicro=0.0979 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:55:03 step 20750 train 0.09902 inner-val 0.09930 best 0.09930@20750 (1064s) immicro=0.0979 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:55:17 step 21000 train 0.09918 inner-val 0.09930 best 0.09930@21000 (1078s) immicro=0.0979 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:55:31 step 21250 train 0.09860 inner-val 0.09929 best 0.09929@21250 (1091s) immicro=0.0979 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:55:45 step 21500 train 0.09882 inner-val 0.09929 best 0.09929@21500 (1105s) immicro=0.0979 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:55:59 step 21750 train 0.09916 inner-val 0.09928 best 0.09928@21750 (1119s) immicro=0.0979 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:56:12 step 22000 train 0.09904 inner-val 0.09928 best 0.09928@22000 (1133s) immicro=0.0979 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:56:26 step 22250 train 0.09899 inner-val 0.09927 best 0.09927@22250 (1147s) immicro=0.0979 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:56:40 step 22500 train 0.09895 inner-val 0.09927 best 0.09927@22500 (1160s) immicro=0.0979 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:56:53 step 22750 train 0.09904 inner-val 0.09926 best 0.09926@22750 (1174s) immicro=0.0979 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:57:07 step 23000 train 0.09932 inner-val 0.09926 best 0.09926@23000 (1188s) immicro=0.0979 imperial_eskd=0.0984 liu2021=0.1016
+2026-10-03T16:57:21 step 23250 train 0.09913 inner-val 0.09925 best 0.09925@23250 (1202s) immicro=0.0979 imperial_eskd=0.0983 liu2021=0.1016
+2026-10-03T16:57:35 step 23500 train 0.09917 inner-val 0.09925 best 0.09925@23500 (1216s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1016
+2026-10-03T16:57:49 step 23750 train 0.09883 inner-val 0.09924 best 0.09924@23750 (1230s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T16:58:04 step 24000 train 0.09923 inner-val 0.09923 best 0.09923@24000 (1244s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T16:58:17 step 24250 train 0.09874 inner-val 0.09923 best 0.09923@24250 (1258s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T16:58:32 step 24500 train 0.09866 inner-val 0.09922 best 0.09922@24500 (1272s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T16:58:46 step 24750 train 0.09902 inner-val 0.09921 best 0.09921@24750 (1286s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T16:58:59 step 25000 train 0.09883 inner-val 0.09921 best 0.09921@25000 (1300s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T16:59:13 step 25250 train 0.09897 inner-val 0.09921 best 0.09921@25250 (1313s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T16:59:27 step 25500 train 0.09876 inner-val 0.09920 best 0.09920@25500 (1327s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T16:59:41 step 25750 train 0.09901 inner-val 0.09920 best 0.09920@25750 (1341s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T16:59:55 step 26000 train 0.09898 inner-val 0.09920 best 0.09920@26000 (1355s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T17:00:08 step 26250 train 0.09884 inner-val 0.09919 best 0.09919@26250 (1369s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T17:00:21 step 26500 train 0.09904 inner-val 0.09919 best 0.09919@26500 (1382s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T17:00:34 step 26750 train 0.09858 inner-val 0.09919 best 0.09919@26750 (1394s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T17:00:46 step 27000 train 0.09897 inner-val 0.09918 best 0.09918@27000 (1407s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T17:00:59 step 27250 train 0.09906 inner-val 0.09918 best 0.09918@27250 (1419s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T17:01:12 step 27500 train 0.09890 inner-val 0.09917 best 0.09917@27500 (1432s) immicro=0.0978 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T17:01:24 step 27750 train 0.09903 inner-val 0.09917 best 0.09917@27750 (1445s) immicro=0.0977 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T17:01:37 step 28000 train 0.09876 inner-val 0.09916 best 0.09916@28000 (1458s) immicro=0.0977 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T17:01:50 step 28250 train 0.09886 inner-val 0.09916 best 0.09916@28250 (1471s) immicro=0.0977 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T17:02:04 step 28500 train 0.09847 inner-val 0.09916 best 0.09916@28500 (1484s) immicro=0.0977 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T17:02:17 step 28750 train 0.09874 inner-val 0.09915 best 0.09915@28750 (1498s) immicro=0.0977 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T17:02:31 step 29000 train 0.09871 inner-val 0.09914 best 0.09914@29000 (1512s) immicro=0.0977 imperial_eskd=0.0983 liu2021=0.1014
+2026-10-03T17:02:45 step 29250 train 0.09917 inner-val 0.09914 best 0.09914@29250 (1526s) immicro=0.0977 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T17:03:00 step 29500 train 0.09846 inner-val 0.09914 best 0.09914@29500 (1540s) immicro=0.0977 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T17:03:14 step 29750 train 0.09871 inner-val 0.09914 best 0.09914@29750 (1554s) immicro=0.0977 imperial_eskd=0.0983 liu2021=0.1015
+2026-10-03T17:03:28 step 30000 train 0.09841 inner-val 0.09914 best 0.09914@30000 (1568s) immicro=0.0977 imperial_eskd=0.0983 liu2021=0.1014
+2026-10-03T17:03:42 step 30250 train 0.09868 inner-val 0.09913 best 0.09913@30250 (1582s) immicro=0.0977 imperial_eskd=0.0983 liu2021=0.1014
+2026-10-03T17:03:56 step 30500 train 0.09871 inner-val 0.09913 best 0.09913@30500 (1596s) immicro=0.0977 imperial_eskd=0.0983 liu2021=0.1014
+2026-10-03T17:04:10 step 30750 train 0.09872 inner-val 0.09913 best 0.09913@30750 (1610s) immicro=0.0977 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:04:24 step 31000 train 0.09834 inner-val 0.09912 best 0.09912@31000 (1624s) immicro=0.0977 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:04:38 step 31250 train 0.09862 inner-val 0.09912 best 0.09912@31250 (1638s) immicro=0.0977 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:04:52 step 31500 train 0.09843 inner-val 0.09911 best 0.09911@31500 (1652s) immicro=0.0977 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:05:06 step 31750 train 0.09875 inner-val 0.09911 best 0.09911@31750 (1666s) immicro=0.0977 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:05:20 step 32000 train 0.09873 inner-val 0.09910 best 0.09910@32000 (1681s) immicro=0.0977 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:05:33 step 32250 train 0.09859 inner-val 0.09910 best 0.09910@32250 (1694s) immicro=0.0977 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:05:46 step 32500 train 0.09852 inner-val 0.09909 best 0.09909@32500 (1706s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:05:58 step 32750 train 0.09876 inner-val 0.09909 best 0.09909@32750 (1719s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:06:11 step 33000 train 0.09851 inner-val 0.09909 best 0.09909@33000 (1731s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:06:24 step 33250 train 0.09864 inner-val 0.09908 best 0.09908@33250 (1744s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:06:36 step 33500 train 0.09820 inner-val 0.09908 best 0.09908@33500 (1757s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:06:47 step 33750 train 0.09858 inner-val 0.09908 best 0.09908@33500 (1769s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:07:01 step 34000 train 0.09876 inner-val 0.09907 best 0.09907@34000 (1782s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:07:15 step 34250 train 0.09877 inner-val 0.09907 best 0.09907@34250 (1796s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:07:29 step 34500 train 0.09876 inner-val 0.09907 best 0.09907@34500 (1809s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:07:42 step 34750 train 0.09877 inner-val 0.09906 best 0.09906@34750 (1823s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:07:56 step 35000 train 0.09868 inner-val 0.09906 best 0.09906@35000 (1837s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:08:10 step 35250 train 0.09869 inner-val 0.09906 best 0.09906@35250 (1850s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:08:24 step 35500 train 0.09885 inner-val 0.09905 best 0.09905@35500 (1864s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:08:36 step 35750 train 0.09855 inner-val 0.09905 best 0.09905@35500 (1878s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:08:50 step 36000 train 0.09867 inner-val 0.09904 best 0.09904@36000 (1890s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1014
+2026-10-03T17:09:04 step 36250 train 0.09865 inner-val 0.09904 best 0.09904@36250 (1904s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1013
+2026-10-03T17:09:18 step 36500 train 0.09829 inner-val 0.09904 best 0.09904@36500 (1918s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1013
+2026-10-03T17:09:32 step 36750 train 0.09875 inner-val 0.09903 best 0.09903@36750 (1932s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1013
+2026-10-03T17:09:45 step 37000 train 0.09828 inner-val 0.09903 best 0.09903@37000 (1946s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1013
+2026-10-03T17:09:58 step 37250 train 0.09852 inner-val 0.09903 best 0.09903@37000 (1960s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1013
+2026-10-03T17:10:11 step 37500 train 0.09867 inner-val 0.09903 best 0.09903@37500 (1972s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1013
+2026-10-03T17:10:25 step 37750 train 0.09860 inner-val 0.09902 best 0.09902@37750 (1986s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1013
+2026-10-03T17:10:39 step 38000 train 0.09870 inner-val 0.09902 best 0.09902@38000 (1999s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1013
+2026-10-03T17:10:53 step 38250 train 0.09873 inner-val 0.09901 best 0.09901@38250 (2013s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1013
+2026-10-03T17:11:07 step 38500 train 0.09865 inner-val 0.09901 best 0.09901@38500 (2028s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1013
+2026-10-03T17:11:22 step 38750 train 0.09859 inner-val 0.09901 best 0.09901@38750 (2042s) immicro=0.0976 imperial_eskd=0.0982 liu2021=0.1013
+2026-10-03T17:11:36 step 39000 train 0.09844 inner-val 0.09900 best 0.09900@39000 (2056s) immicro=0.0975 imperial_eskd=0.0982 liu2021=0.1013
+2026-10-03T17:11:51 step 39250 train 0.09845 inner-val 0.09900 best 0.09900@39250 (2071s) immicro=0.0975 imperial_eskd=0.0982 liu2021=0.1013
+2026-10-03T17:12:05 step 39500 train 0.09882 inner-val 0.09900 best 0.09900@39500 (2085s) immicro=0.0975 imperial_eskd=0.0982 liu2021=0.1013
+2026-10-03T17:12:18 step 39750 train 0.09848 inner-val 0.09899 best 0.09899@39750 (2099s) immicro=0.0975 imperial_eskd=0.0982 liu2021=0.1013
+2026-10-03T17:12:32 step 40000 train 0.09848 inner-val 0.09899 best 0.09899@40000 (2113s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:12:34 block end 40000: best 0.09899@40000; block improvement 0.000334 (3.59 % of total 0.009288) -> continue
+2026-10-03T17:12:48 step 40250 train 0.09863 inner-val 0.09898 best 0.09898@40250 (2128s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:13:02 step 40500 train 0.09877 inner-val 0.09898 best 0.09898@40500 (2142s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:13:16 step 40750 train 0.09833 inner-val 0.09898 best 0.09898@40750 (2156s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:13:29 step 41000 train 0.09855 inner-val 0.09897 best 0.09897@41000 (2170s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:13:43 step 41250 train 0.09842 inner-val 0.09897 best 0.09897@41250 (2184s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:13:57 step 41500 train 0.09845 inner-val 0.09897 best 0.09897@41500 (2198s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:14:11 step 41750 train 0.09839 inner-val 0.09897 best 0.09897@41750 (2212s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:14:25 step 42000 train 0.09836 inner-val 0.09896 best 0.09896@42000 (2225s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:14:39 step 42250 train 0.09825 inner-val 0.09896 best 0.09896@42250 (2239s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:14:51 step 42500 train 0.09874 inner-val 0.09896 best 0.09896@42250 (2253s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:15:05 step 42750 train 0.09826 inner-val 0.09896 best 0.09896@42750 (2265s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:15:17 step 43000 train 0.09821 inner-val 0.09896 best 0.09896@42750 (2279s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:15:31 step 43250 train 0.09866 inner-val 0.09896 best 0.09896@43250 (2292s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:15:45 step 43500 train 0.09859 inner-val 0.09895 best 0.09895@43500 (2305s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:15:59 step 43750 train 0.09892 inner-val 0.09895 best 0.09895@43750 (2319s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:16:12 step 44000 train 0.09817 inner-val 0.09895 best 0.09895@44000 (2333s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1013
+2026-10-03T17:16:26 step 44250 train 0.09842 inner-val 0.09894 best 0.09894@44250 (2347s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:16:40 step 44500 train 0.09857 inner-val 0.09894 best 0.09894@44500 (2361s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:16:54 step 44750 train 0.09846 inner-val 0.09894 best 0.09894@44750 (2374s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:17:08 step 45000 train 0.09836 inner-val 0.09893 best 0.09893@45000 (2388s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:17:20 step 45250 train 0.09856 inner-val 0.09893 best 0.09893@45000 (2402s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:17:34 step 45500 train 0.09881 inner-val 0.09893 best 0.09893@45500 (2414s) immicro=0.0975 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:17:47 step 45750 train 0.09846 inner-val 0.09893 best 0.09893@45750 (2428s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:18:01 step 46000 train 0.09861 inner-val 0.09892 best 0.09892@46000 (2442s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:18:15 step 46250 train 0.09822 inner-val 0.09892 best 0.09892@46250 (2455s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:18:27 step 46500 train 0.09847 inner-val 0.09892 best 0.09892@46250 (2469s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:18:41 step 46750 train 0.09832 inner-val 0.09892 best 0.09892@46750 (2481s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:18:55 step 47000 train 0.09824 inner-val 0.09891 best 0.09891@47000 (2495s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:19:09 step 47250 train 0.09838 inner-val 0.09891 best 0.09891@47250 (2509s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:19:20 step 47500 train 0.09850 inner-val 0.09891 best 0.09891@47250 (2522s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:19:35 step 47750 train 0.09840 inner-val 0.09890 best 0.09890@47750 (2535s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:19:49 step 48000 train 0.09840 inner-val 0.09890 best 0.09890@48000 (2549s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:20:03 step 48250 train 0.09860 inner-val 0.09889 best 0.09889@48250 (2563s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:20:16 step 48500 train 0.09828 inner-val 0.09889 best 0.09889@48250 (2578s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:20:30 step 48750 train 0.09816 inner-val 0.09889 best 0.09889@48750 (2590s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:20:43 step 49000 train 0.09853 inner-val 0.09889 best 0.09889@48750 (2604s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:20:57 step 49250 train 0.09835 inner-val 0.09889 best 0.09889@49250 (2617s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:21:11 step 49500 train 0.09830 inner-val 0.09888 best 0.09888@49500 (2632s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:21:25 step 49750 train 0.09846 inner-val 0.09888 best 0.09888@49750 (2646s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:21:39 step 50000 train 0.09789 inner-val 0.09888 best 0.09888@50000 (2659s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:21:52 step 50250 train 0.09817 inner-val 0.09888 best 0.09888@50250 (2673s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:22:06 step 50500 train 0.09852 inner-val 0.09887 best 0.09887@50500 (2687s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:22:20 step 50750 train 0.09810 inner-val 0.09887 best 0.09887@50750 (2701s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:22:34 step 51000 train 0.09839 inner-val 0.09886 best 0.09886@51000 (2714s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:22:47 step 51250 train 0.09826 inner-val 0.09886 best 0.09886@51250 (2728s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:23:01 step 51500 train 0.09828 inner-val 0.09886 best 0.09886@51500 (2742s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1012
+2026-10-03T17:23:15 step 51750 train 0.09836 inner-val 0.09885 best 0.09885@51750 (2756s) immicro=0.0974 imperial_eskd=0.0981 liu2021=0.1011
+2026-10-03T17:23:29 step 52000 train 0.09851 inner-val 0.09885 best 0.09885@52000 (2769s) immicro=0.0974 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:23:42 step 52250 train 0.09838 inner-val 0.09885 best 0.09885@52250 (2783s) immicro=0.0974 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:23:56 step 52500 train 0.09871 inner-val 0.09885 best 0.09885@52500 (2796s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:24:08 step 52750 train 0.09847 inner-val 0.09885 best 0.09885@52500 (2810s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1012
+2026-10-03T17:24:22 step 53000 train 0.09809 inner-val 0.09884 best 0.09884@53000 (2822s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1012
+2026-10-03T17:24:34 step 53250 train 0.09845 inner-val 0.09884 best 0.09884@53000 (2836s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1012
+2026-10-03T17:24:47 step 53500 train 0.09817 inner-val 0.09885 best 0.09884@53000 (2849s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1012
+2026-10-03T17:25:00 step 53750 train 0.09835 inner-val 0.09884 best 0.09884@53750 (2861s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1012
+2026-10-03T17:25:14 step 54000 train 0.09812 inner-val 0.09884 best 0.09884@54000 (2875s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1012
+2026-10-03T17:25:28 step 54250 train 0.09847 inner-val 0.09883 best 0.09883@54250 (2889s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:25:42 step 54500 train 0.09837 inner-val 0.09883 best 0.09883@54500 (2902s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:25:56 step 54750 train 0.09811 inner-val 0.09883 best 0.09883@54750 (2916s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:26:09 step 55000 train 0.09832 inner-val 0.09883 best 0.09883@55000 (2929s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:26:22 step 55250 train 0.09813 inner-val 0.09882 best 0.09882@55250 (2942s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:26:35 step 55500 train 0.09829 inner-val 0.09882 best 0.09882@55500 (2955s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:26:49 step 55750 train 0.09824 inner-val 0.09882 best 0.09882@55750 (2969s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:27:03 step 56000 train 0.09803 inner-val 0.09881 best 0.09881@56000 (2984s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:27:17 step 56250 train 0.09831 inner-val 0.09881 best 0.09881@56250 (2998s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:27:31 step 56500 train 0.09793 inner-val 0.09880 best 0.09880@56500 (3012s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:27:45 step 56750 train 0.09785 inner-val 0.09880 best 0.09880@56750 (3026s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:27:59 step 57000 train 0.09801 inner-val 0.09880 best 0.09880@57000 (3039s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:28:11 step 57250 train 0.09834 inner-val 0.09880 best 0.09880@57000 (3053s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:28:25 step 57500 train 0.09816 inner-val 0.09879 best 0.09879@57500 (3065s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:28:37 step 57750 train 0.09789 inner-val 0.09879 best 0.09879@57500 (3079s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:28:50 step 58000 train 0.09817 inner-val 0.09880 best 0.09879@57500 (3092s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:29:02 step 58250 train 0.09832 inner-val 0.09880 best 0.09879@57500 (3104s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:29:14 step 58500 train 0.09840 inner-val 0.09880 best 0.09879@57500 (3116s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:29:27 step 58750 train 0.09860 inner-val 0.09880 best 0.09879@57500 (3129s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:29:39 step 59000 train 0.09833 inner-val 0.09879 best 0.09879@57500 (3141s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:29:53 step 59250 train 0.09784 inner-val 0.09879 best 0.09879@59250 (3153s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:30:07 step 59500 train 0.09813 inner-val 0.09879 best 0.09879@59500 (3167s) immicro=0.0973 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:30:19 step 59750 train 0.09832 inner-val 0.09879 best 0.09879@59500 (3181s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:30:31 step 60000 train 0.09801 inner-val 0.09879 best 0.09879@59500 (3193s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:30:33 block end 60000: best 0.09879@59500; block improvement 0.000202 (2.13 % of total 0.009490) -> continue
+2026-10-03T17:30:45 step 60250 train 0.09817 inner-val 0.09879 best 0.09879@59500 (3207s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:30:59 step 60500 train 0.09820 inner-val 0.09878 best 0.09878@60500 (3220s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:31:13 step 60750 train 0.09796 inner-val 0.09878 best 0.09878@60750 (3233s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:31:25 step 61000 train 0.09833 inner-val 0.09878 best 0.09878@60750 (3247s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:31:38 step 61250 train 0.09850 inner-val 0.09878 best 0.09878@60750 (3260s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:31:52 step 61500 train 0.09794 inner-val 0.09877 best 0.09877@61500 (3272s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:32:06 step 61750 train 0.09821 inner-val 0.09877 best 0.09877@61750 (3286s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:32:20 step 62000 train 0.09810 inner-val 0.09877 best 0.09877@62000 (3301s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:32:35 step 62250 train 0.09808 inner-val 0.09876 best 0.09876@62250 (3315s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:32:49 step 62500 train 0.09826 inner-val 0.09876 best 0.09876@62500 (3329s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:33:03 step 62750 train 0.09801 inner-val 0.09875 best 0.09875@62750 (3343s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:33:16 step 63000 train 0.09826 inner-val 0.09875 best 0.09875@63000 (3357s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:33:30 step 63250 train 0.09842 inner-val 0.09875 best 0.09875@63250 (3371s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:33:43 step 63500 train 0.09794 inner-val 0.09875 best 0.09875@63250 (3384s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:33:56 step 63750 train 0.09811 inner-val 0.09875 best 0.09875@63750 (3397s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:34:10 step 64000 train 0.09819 inner-val 0.09874 best 0.09874@64000 (3411s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:34:22 step 64250 train 0.09813 inner-val 0.09874 best 0.09874@64000 (3424s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:34:36 step 64500 train 0.09801 inner-val 0.09874 best 0.09874@64500 (3437s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:34:49 step 64750 train 0.09814 inner-val 0.09874 best 0.09874@64500 (3451s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:35:02 step 65000 train 0.09812 inner-val 0.09874 best 0.09874@65000 (3463s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:35:16 step 65250 train 0.09820 inner-val 0.09873 best 0.09873@65250 (3477s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:35:30 step 65500 train 0.09788 inner-val 0.09873 best 0.09873@65500 (3490s) immicro=0.0972 imperial_eskd=0.0979 liu2021=0.1011
+2026-10-03T17:35:42 step 65750 train 0.09834 inner-val 0.09873 best 0.09873@65500 (3504s) immicro=0.0972 imperial_eskd=0.0980 liu2021=0.1011
+2026-10-03T17:35:56 step 66000 train 0.09793 inner-val 0.09873 best 0.09873@66000 (3517s) immicro=0.0972 imperial_eskd=0.0979 liu2021=0.1011
+2026-10-03T17:36:08 step 66250 train 0.09788 inner-val 0.09873 best 0.09873@66000 (3530s) immicro=0.0972 imperial_eskd=0.0979 liu2021=0.1011
+2026-10-03T17:36:22 step 66500 train 0.09818 inner-val 0.09873 best 0.09873@66500 (3543s) immicro=0.0972 imperial_eskd=0.0979 liu2021=0.1011
+2026-10-03T17:36:36 step 66750 train 0.09794 inner-val 0.09872 best 0.09872@66750 (3556s) immicro=0.0972 imperial_eskd=0.0979 liu2021=0.1011
+2026-10-03T17:36:50 step 67000 train 0.09817 inner-val 0.09872 best 0.09872@67000 (3570s) immicro=0.0972 imperial_eskd=0.0979 liu2021=0.1011
+2026-10-03T17:37:03 step 67250 train 0.09781 inner-val 0.09871 best 0.09871@67250 (3583s) immicro=0.0972 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:37:14 step 67500 train 0.09813 inner-val 0.09871 best 0.09871@67250 (3596s) immicro=0.0972 imperial_eskd=0.0979 liu2021=0.1011
+2026-10-03T17:37:25 step 67750 train 0.09802 inner-val 0.09871 best 0.09871@67250 (3607s) immicro=0.0972 imperial_eskd=0.0979 liu2021=0.1011
+2026-10-03T17:37:38 step 68000 train 0.09802 inner-val 0.09871 best 0.09871@68000 (3618s) immicro=0.0972 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:37:50 step 68250 train 0.09799 inner-val 0.09871 best 0.09871@68250 (3631s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1011
+2026-10-03T17:38:02 step 68500 train 0.09807 inner-val 0.09871 best 0.09871@68250 (3644s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1011
+2026-10-03T17:38:14 step 68750 train 0.09827 inner-val 0.09870 best 0.09870@68750 (3655s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1011
+2026-10-03T17:38:26 step 69000 train 0.09793 inner-val 0.09871 best 0.09870@68750 (3668s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1011
+2026-10-03T17:38:38 step 69250 train 0.09806 inner-val 0.09870 best 0.09870@69250 (3679s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:38:51 step 69500 train 0.09758 inner-val 0.09870 best 0.09870@69500 (3692s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:39:04 step 69750 train 0.09813 inner-val 0.09870 best 0.09870@69750 (3704s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:39:17 step 70000 train 0.09793 inner-val 0.09870 best 0.09870@70000 (3717s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:39:29 step 70250 train 0.09799 inner-val 0.09869 best 0.09869@70250 (3730s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:39:42 step 70500 train 0.09802 inner-val 0.09869 best 0.09869@70500 (3743s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:39:55 step 70750 train 0.09783 inner-val 0.09868 best 0.09868@70750 (3755s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:40:06 step 71000 train 0.09786 inner-val 0.09869 best 0.09868@70750 (3768s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:40:17 step 71250 train 0.09839 inner-val 0.09868 best 0.09868@70750 (3779s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:40:30 step 71500 train 0.09788 inner-val 0.09868 best 0.09868@71500 (3791s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:40:43 step 71750 train 0.09790 inner-val 0.09867 best 0.09867@71750 (3804s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:40:56 step 72000 train 0.09807 inner-val 0.09867 best 0.09867@72000 (3816s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:41:08 step 72250 train 0.09759 inner-val 0.09867 best 0.09867@72250 (3829s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:41:19 step 72500 train 0.09839 inner-val 0.09867 best 0.09867@72250 (3841s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:41:32 step 72750 train 0.09802 inner-val 0.09867 best 0.09867@72750 (3852s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:41:43 step 73000 train 0.09757 inner-val 0.09867 best 0.09867@72750 (3865s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:41:56 step 73250 train 0.09809 inner-val 0.09866 best 0.09866@73250 (3876s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:42:07 step 73500 train 0.09805 inner-val 0.09866 best 0.09866@73250 (3889s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:42:19 step 73750 train 0.09819 inner-val 0.09866 best 0.09866@73250 (3901s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:42:32 step 74000 train 0.09764 inner-val 0.09866 best 0.09866@74000 (3912s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:42:43 step 74250 train 0.09789 inner-val 0.09866 best 0.09866@74000 (3925s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:42:56 step 74500 train 0.09798 inner-val 0.09866 best 0.09866@74500 (3936s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:43:07 step 74750 train 0.09794 inner-val 0.09866 best 0.09866@74500 (3949s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:43:18 step 75000 train 0.09822 inner-val 0.09866 best 0.09866@74500 (3960s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:43:30 step 75250 train 0.09771 inner-val 0.09866 best 0.09866@74500 (3972s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:43:42 step 75500 train 0.09800 inner-val 0.09865 best 0.09865@75500 (3983s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:43:55 step 75750 train 0.09787 inner-val 0.09865 best 0.09865@75750 (3995s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:44:08 step 76000 train 0.09793 inner-val 0.09865 best 0.09865@76000 (4008s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:44:19 step 76250 train 0.09806 inner-val 0.09865 best 0.09865@76000 (4021s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:44:32 step 76500 train 0.09783 inner-val 0.09864 best 0.09864@76500 (4032s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:44:45 step 76750 train 0.09792 inner-val 0.09864 best 0.09864@76750 (4046s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:44:56 step 77000 train 0.09818 inner-val 0.09864 best 0.09864@76750 (4058s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:45:08 step 77250 train 0.09786 inner-val 0.09864 best 0.09864@76750 (4070s) immicro=0.0971 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:45:21 step 77500 train 0.09803 inner-val 0.09864 best 0.09864@77500 (4081s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:45:32 step 77750 train 0.09787 inner-val 0.09864 best 0.09864@77500 (4094s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:45:45 step 78000 train 0.09776 inner-val 0.09864 best 0.09864@78000 (4105s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:45:58 step 78250 train 0.09771 inner-val 0.09864 best 0.09864@78250 (4118s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:46:11 step 78500 train 0.09805 inner-val 0.09864 best 0.09864@78500 (4131s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:46:24 step 78750 train 0.09776 inner-val 0.09863 best 0.09863@78750 (4144s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:46:35 step 79000 train 0.09779 inner-val 0.09863 best 0.09863@78750 (4157s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:46:46 step 79250 train 0.09795 inner-val 0.09863 best 0.09863@78750 (4168s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:46:59 step 79500 train 0.09769 inner-val 0.09863 best 0.09863@79500 (4180s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:47:12 step 79750 train 0.09806 inner-val 0.09863 best 0.09863@79750 (4192s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:47:25 step 80000 train 0.09788 inner-val 0.09862 best 0.09862@80000 (4205s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:47:27 block end 80000: best 0.09862@80000; block improvement 0.000161 (1.67 % of total 0.009651) -> continue
+2026-10-03T17:47:40 step 80250 train 0.09796 inner-val 0.09862 best 0.09862@80250 (4220s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:47:52 step 80500 train 0.09787 inner-val 0.09862 best 0.09862@80500 (4233s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:48:04 step 80750 train 0.09786 inner-val 0.09861 best 0.09861@80750 (4245s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:48:17 step 81000 train 0.09791 inner-val 0.09861 best 0.09861@81000 (4257s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:48:29 step 81250 train 0.09753 inner-val 0.09860 best 0.09860@81250 (4270s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:48:40 step 81500 train 0.09804 inner-val 0.09861 best 0.09860@81250 (4282s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:48:52 step 81750 train 0.09796 inner-val 0.09860 best 0.09860@81250 (4294s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1010
+2026-10-03T17:49:03 step 82000 train 0.09780 inner-val 0.09860 best 0.09860@81250 (4305s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1009
+2026-10-03T17:49:16 step 82250 train 0.09783 inner-val 0.09860 best 0.09860@82250 (4316s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1009
+2026-10-03T17:49:27 step 82500 train 0.09791 inner-val 0.09860 best 0.09860@82250 (4329s) immicro=0.0970 imperial_eskd=0.0979 liu2021=0.1009
+2026-10-03T17:49:40 step 82750 train 0.09787 inner-val 0.09860 best 0.09860@82750 (4340s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:49:53 step 83000 train 0.09768 inner-val 0.09860 best 0.09860@83000 (4353s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:50:04 step 83250 train 0.09776 inner-val 0.09860 best 0.09860@83000 (4366s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:50:15 step 83500 train 0.09759 inner-val 0.09860 best 0.09860@83000 (4377s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1010
+2026-10-03T17:50:26 step 83750 train 0.09797 inner-val 0.09860 best 0.09860@83000 (4388s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1010
+2026-10-03T17:50:39 step 84000 train 0.09757 inner-val 0.09859 best 0.09859@84000 (4399s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:50:50 step 84250 train 0.09754 inner-val 0.09859 best 0.09859@84000 (4412s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:51:03 step 84500 train 0.09751 inner-val 0.09859 best 0.09859@84500 (4423s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:51:16 step 84750 train 0.09791 inner-val 0.09858 best 0.09858@84750 (4436s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:51:27 step 85000 train 0.09786 inner-val 0.09858 best 0.09858@84750 (4449s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:51:40 step 85250 train 0.09801 inner-val 0.09858 best 0.09858@85250 (4460s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:51:52 step 85500 train 0.09783 inner-val 0.09858 best 0.09858@85500 (4473s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:52:05 step 85750 train 0.09803 inner-val 0.09858 best 0.09858@85750 (4485s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:52:16 step 86000 train 0.09783 inner-val 0.09858 best 0.09858@85750 (4498s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:52:29 step 86250 train 0.09789 inner-val 0.09858 best 0.09858@86250 (4509s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:52:42 step 86500 train 0.09767 inner-val 0.09857 best 0.09857@86500 (4522s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:52:55 step 86750 train 0.09734 inner-val 0.09857 best 0.09857@86750 (4535s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:53:08 step 87000 train 0.09779 inner-val 0.09857 best 0.09857@87000 (4548s) immicro=0.0970 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:53:20 step 87250 train 0.09768 inner-val 0.09857 best 0.09857@87250 (4561s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:53:33 step 87500 train 0.09779 inner-val 0.09856 best 0.09856@87500 (4574s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:53:46 step 87750 train 0.09778 inner-val 0.09856 best 0.09856@87750 (4586s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:53:57 step 88000 train 0.09785 inner-val 0.09856 best 0.09856@87750 (4599s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:54:08 step 88250 train 0.09764 inner-val 0.09856 best 0.09856@87750 (4610s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:54:21 step 88500 train 0.09770 inner-val 0.09855 best 0.09855@88500 (4621s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:54:34 step 88750 train 0.09769 inner-val 0.09855 best 0.09855@88750 (4634s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:54:47 step 89000 train 0.09809 inner-val 0.09855 best 0.09855@89000 (4647s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:54:59 step 89250 train 0.09813 inner-val 0.09855 best 0.09855@89250 (4660s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:55:12 step 89500 train 0.09787 inner-val 0.09854 best 0.09854@89500 (4673s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:55:25 step 89750 train 0.09783 inner-val 0.09854 best 0.09854@89750 (4686s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:55:38 step 90000 train 0.09768 inner-val 0.09854 best 0.09854@90000 (4698s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:55:49 step 90250 train 0.09758 inner-val 0.09854 best 0.09854@90000 (4711s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:56:02 step 90500 train 0.09778 inner-val 0.09854 best 0.09854@90500 (4722s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:56:13 step 90750 train 0.09761 inner-val 0.09853 best 0.09854@90500 (4735s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:56:26 step 91000 train 0.09787 inner-val 0.09853 best 0.09853@91000 (4746s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:56:38 step 91250 train 0.09767 inner-val 0.09853 best 0.09853@91250 (4759s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:56:49 step 91500 train 0.09761 inner-val 0.09853 best 0.09853@91250 (4771s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:57:02 step 91750 train 0.09754 inner-val 0.09853 best 0.09853@91750 (4783s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:57:13 step 92000 train 0.09772 inner-val 0.09853 best 0.09853@91750 (4795s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:57:25 step 92250 train 0.09771 inner-val 0.09853 best 0.09853@91750 (4807s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:57:36 step 92500 train 0.09754 inner-val 0.09853 best 0.09853@91750 (4818s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:57:47 step 92750 train 0.09758 inner-val 0.09853 best 0.09853@91750 (4829s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:57:58 step 93000 train 0.09775 inner-val 0.09853 best 0.09853@91750 (4840s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:58:09 step 93250 train 0.09761 inner-val 0.09853 best 0.09853@91750 (4851s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:58:21 step 93500 train 0.09778 inner-val 0.09853 best 0.09853@91750 (4862s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:58:33 step 93750 train 0.09750 inner-val 0.09853 best 0.09853@93750 (4874s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:58:45 step 94000 train 0.09807 inner-val 0.09853 best 0.09853@93750 (4886s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:58:56 step 94250 train 0.09753 inner-val 0.09853 best 0.09853@93750 (4898s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:59:09 step 94500 train 0.09783 inner-val 0.09852 best 0.09852@94500 (4909s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:59:21 step 94750 train 0.09753 inner-val 0.09852 best 0.09852@94750 (4922s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:59:32 step 95000 train 0.09778 inner-val 0.09852 best 0.09852@94750 (4934s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:59:45 step 95250 train 0.09755 inner-val 0.09851 best 0.09851@95250 (4946s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T17:59:58 step 95500 train 0.09764 inner-val 0.09851 best 0.09851@95500 (4958s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T18:00:09 step 95750 train 0.09779 inner-val 0.09851 best 0.09851@95500 (4971s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T18:00:20 step 96000 train 0.09748 inner-val 0.09851 best 0.09851@95500 (4982s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T18:00:32 step 96250 train 0.09791 inner-val 0.09851 best 0.09851@95500 (4994s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T18:00:43 step 96500 train 0.09772 inner-val 0.09851 best 0.09851@95500 (5005s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T18:00:54 step 96750 train 0.09736 inner-val 0.09851 best 0.09851@95500 (5016s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T18:01:07 step 97000 train 0.09750 inner-val 0.09851 best 0.09851@97000 (5027s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T18:01:20 step 97250 train 0.09761 inner-val 0.09850 best 0.09850@97250 (5040s) immicro=0.0969 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T18:01:32 step 97500 train 0.09745 inner-val 0.09850 best 0.09850@97500 (5053s) immicro=0.0968 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T18:01:45 step 97750 train 0.09790 inner-val 0.09850 best 0.09850@97750 (5065s) immicro=0.0968 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T18:01:56 step 98000 train 0.09790 inner-val 0.09850 best 0.09850@97750 (5078s) immicro=0.0968 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T18:02:08 step 98250 train 0.09759 inner-val 0.09850 best 0.09850@97750 (5090s) immicro=0.0968 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T18:02:19 step 98500 train 0.09753 inner-val 0.09850 best 0.09850@97750 (5101s) immicro=0.0968 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T18:02:31 step 98750 train 0.09747 inner-val 0.09850 best 0.09850@98750 (5112s) immicro=0.0968 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T18:02:44 step 99000 train 0.09797 inner-val 0.09849 best 0.09849@99000 (5125s) immicro=0.0968 imperial_eskd=0.0978 liu2021=0.1009
+2026-10-03T18:02:57 step 99250 train 0.09767 inner-val 0.09849 best 0.09849@99250 (5138s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:03:08 step 99500 train 0.09737 inner-val 0.09849 best 0.09849@99250 (5150s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:03:21 step 99750 train 0.09755 inner-val 0.09849 best 0.09849@99750 (5162s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:03:34 step 100000 train 0.09790 inner-val 0.09848 best 0.09848@100000 (5174s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:03:36 block end 100000: best 0.09848@100000; block improvement 0.000141 (1.44 % of total 0.009792) -> continue
+2026-10-03T18:03:49 step 100250 train 0.09749 inner-val 0.09848 best 0.09848@100250 (5189s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:04:00 step 100500 train 0.09743 inner-val 0.09848 best 0.09848@100250 (5202s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:04:12 step 100750 train 0.09776 inner-val 0.09848 best 0.09848@100750 (5213s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:04:24 step 101000 train 0.09736 inner-val 0.09848 best 0.09848@100750 (5226s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:04:36 step 101250 train 0.09748 inner-val 0.09848 best 0.09848@101250 (5237s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:04:49 step 101500 train 0.09748 inner-val 0.09847 best 0.09847@101500 (5250s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:05:00 step 101750 train 0.09759 inner-val 0.09847 best 0.09847@101500 (5262s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:05:11 step 102000 train 0.09733 inner-val 0.09847 best 0.09847@101500 (5273s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:05:23 step 102250 train 0.09740 inner-val 0.09847 best 0.09847@101500 (5285s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:05:34 step 102500 train 0.09761 inner-val 0.09847 best 0.09847@101500 (5296s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:05:45 step 102750 train 0.09755 inner-val 0.09847 best 0.09847@101500 (5307s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:05:58 step 103000 train 0.09748 inner-val 0.09847 best 0.09847@103000 (5318s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:06:10 step 103250 train 0.09761 inner-val 0.09847 best 0.09847@103250 (5331s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:06:23 step 103500 train 0.09755 inner-val 0.09846 best 0.09846@103500 (5343s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:06:34 step 103750 train 0.09750 inner-val 0.09846 best 0.09846@103500 (5356s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1009
+2026-10-03T18:06:45 step 104000 train 0.09770 inner-val 0.09846 best 0.09846@103500 (5367s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:06:58 step 104250 train 0.09751 inner-val 0.09846 best 0.09846@104250 (5378s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:07:09 step 104500 train 0.09789 inner-val 0.09846 best 0.09846@104250 (5391s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:07:23 step 104750 train 0.09756 inner-val 0.09845 best 0.09845@104750 (5403s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:07:36 step 105000 train 0.09741 inner-val 0.09845 best 0.09845@104750 (5418s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:07:48 step 105250 train 0.09760 inner-val 0.09845 best 0.09845@104750 (5430s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:08:03 step 105500 train 0.09719 inner-val 0.09845 best 0.09845@105500 (5443s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:08:16 step 105750 train 0.09746 inner-val 0.09844 best 0.09844@105750 (5457s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:08:30 step 106000 train 0.09747 inner-val 0.09844 best 0.09844@106000 (5471s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:08:42 step 106250 train 0.09783 inner-val 0.09844 best 0.09844@106000 (5484s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:08:56 step 106500 train 0.09749 inner-val 0.09844 best 0.09844@106500 (5497s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:09:09 step 106750 train 0.09780 inner-val 0.09844 best 0.09844@106500 (5511s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:09:21 step 107000 train 0.09753 inner-val 0.09844 best 0.09844@106500 (5523s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:09:34 step 107250 train 0.09736 inner-val 0.09844 best 0.09844@106500 (5536s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:09:46 step 107500 train 0.09730 inner-val 0.09844 best 0.09844@106500 (5548s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:09:58 step 107750 train 0.09739 inner-val 0.09844 best 0.09844@106500 (5560s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:10:11 step 108000 train 0.09753 inner-val 0.09844 best 0.09844@106500 (5573s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:10:23 step 108250 train 0.09761 inner-val 0.09844 best 0.09844@106500 (5585s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:10:36 step 108500 train 0.09755 inner-val 0.09844 best 0.09844@106500 (5598s) immicro=0.0968 imperial_eskd=0.0977 liu2021=0.1008
+2026-10-03T18:10:36 early stop: no improvement in 8 evaluations
+2026-10-03T18:10:43 saved /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/hoL_ye2026_sepsis/K1000/fullsat/checkpoint_ft.pt (selected step 106500, inner-val 0.09844 vs pretrained 0.10828); sha 9240379953f4