CITADEL
StandbyNothing is live right now.
Eval accuracylast measured Wed 18 Nov, 14:41 UTC
%
0.9000450 of 500 eval, 50 wrong
▲ +9.20 pts across 3 evals▲ +4.20 pts on the eval before
0%25%50%75%100%0%25%50%75%100%0%25%50%75%100%

Familiar decider: ModernBERT-large

finishedRadeon 8060S

One of the small models that decide whether Familiar, a desktop companion, reacts to something happening on the PC, and how. Every candidate trains on the same 6,000 decision points from synthetic timelines. The encoders also get one run on the full pool.

started
Wed 18 Nov, 12:57 UTC
ran for
93:18
config
learning rate 0.0001, batch 16, seed 0

run1/12step1125/1125epoch3/3config1/9seed0lr0.0001batch16modelModernBERT-large 395Mrev45bb465items6000+500

Loss train 0.1586 ▼ −0.9784 lowest 0.0565 at step 1110
Timebase
Range
Smoothing
113 readings, 30 min/div30 min/div10 min/div
00.511.5200:0030:0060:0090:00phase starting, 0.0 s to 90.8 sphase training, 90.8 s to 93:18epoch 1, 31:07epoch 2, 59:24epoch 3, 92:59trainingepoch 1epoch 2−0.9784since first1.1371Lowest last train loss across this project's 12 runs, run 11lowest last 0.0007, run 110.158600.511.5200:0030:0060:0090:00phase starting, 0.0 s to 90.8 sphase training, 90.8 s to 93:18epoch 1, 31:07epoch 2, 59:24epoch 3, 92:59trainingepoch 1epoch 21.1371Lowest last train loss across this project's 12 runs, run 11lowest last 0.0007, run 110.1586▼ −0.978400.511.5200:0010:0020:0030:0040:0050:0060:0070:0080:0090:00phase starting, 0.0 s to 90.8 sphase training, 90.8 s to 93:18epoch 1, 31:07epoch 2, 59:24epoch 3, 92:59trainingepoch 1epoch 21.1371Lowest last train loss across this project's 12 runs, run 11lowest last 0.0007, run 110.1586▼ −0.9784

Eval measures

Elapsed

1856.1 s

epoch interval

1125 of 1125 steps, 3 of 3 epochs, 100%

4.85 s per step, 12.4 steps a minute

Eval accuracy by eval

+0.0420

▲ on the eval before

eval 1 0.8080, eval 3 0.9000

Against the other runs

7 of 11

by last eval accuracy

best so far 0.9600, run 7

Set-up

6000 train, 500 eval

synthetic, learning rate 0.0001, batch size 16, epochs 3, max length 512, precision bf16-mixed, loss ce, optimizer adamw, scheduler linear, seed 0

Learning rate

0.00e+0

▼ −1.49e-5 since first

On one axis

phase starting, 0.0 s to 90.8 sphase training, 90.8 s to 93:18E1E2E3traininglowest last 0.0007Train loss0.1586▼ −0.97842.000.001.13710.1586last ▼ −0.9784best 0.9600, run 7Last-value range, 11 runs, 0.8854 to 0.9600Eval accuracy0.9000▲ +0.09201.000.800.80800.9000last ▲ +0.0920Learning rate0.00e+0▼ −1.49e-51.0e-40.0e+01.49e-5last 0.00e+0Power, whole chip, CPU and GPU84 W▼ −4 W120 W40 W87 Wlast 84 WHottest, one board zone92 °C▼ −4 °C98 °C88 °C96 °Clast 92 °C00:0010:0020:0030:0040:0050:0060:0070:0080:0090:00
GPUpower: whole chip, CPU and GPUtemperature: one board zone

GPU 0

Wwhole chip, CPU and GPU

84 W whole chip, CPU and GPU

use72%
memory, whole GPU2.6 GB

peak 29.2 GB, whole GPU

92 °C one board zone

history 0 to 112 W

Input

  • Runs128 runs
  • Agentsnone showing
  • Treesnone showing
  • Projectsnone showing

Check

Every entry is checked on its way out; some are held for review.

Not shown on any run: data samples, predictions, raw logs, file paths, machine names

Delay

2 min, set

Watching the agents work, about 2 min behind.

One way

Nothing here reaches the agents.

This page

This page updates by itself.

visits counted with Google Analytics · choose

Runsheight is eval accuracyall recordings ▸ | finished× failed■ stopped○ lost● live ◂ older
1.000.851.000.85finished, eval accuracy 0.9000, config 1 of 9 (learning rate 0.0001, batch 16), seed 0, Wed 18 Nov, 12:57 UTCfinished, eval accuracy 0.9200, config 2 of 9 (learning rate 0.00002, batch 16), seed 0, Wed 18 Nov, 14:41 UTCfinished, eval accuracy 0.9020, config 3 of 9 (learning rate 0.00005, batch 16), seed 0, Wed 18 Nov, 16:41 UTCfinished, eval accuracy 0.9080, config 4 of 9 (batch 16), seed 1 of 2, Thu 19 Nov, 00:21 UTCfinished, eval accuracy 0.9160, config 4 of 9 (batch 16), seed 2 of 2, Thu 19 Nov, 02:14 UTCfailed at step 470 of 7274, eval accuracy -, config 5 of 9 (batch 16), seed 0, Fri 20 Nov, 13:18 UTCfinished, eval accuracy 0.9600, config 5 of 9 (batch 16), seed 0, Sat 21 Nov, 01:57 UTCfinished, eval accuracy 0.8876, config 6 of 9 (learning rate 0.0001, batch 16), seed 0, Sat 21 Nov, 12:36 UTCfinished, eval accuracy 0.9191, config 7 of 9 (learning rate 0.00002, batch 16), seed 0, Sat 21 Nov, 13:15 UTCfinished, eval accuracy 0.8944, config 8 of 9 (learning rate 0.00005, batch 16), seed 0, Sat 21 Nov, 13:54 UTCfinished, eval accuracy 0.8854, config 9 of 9 (batch 16), seed 1 of 2, Sat 21 Nov, 19:01 UTCfinished, eval accuracy 0.8921, config 9 of 9 (batch 16), seed 2 of 2, Sat 21 Nov, 19:40 UTC1.000.85finished, eval accuracy 0.9000, config 1 of 9 (learning rate 0.0001, batch 16), seed 0, Wed 18 Nov, 12:57 UTCfinished, eval accuracy 0.9200, config 2 of 9 (learning rate 0.00002, batch 16), seed 0, Wed 18 Nov, 14:41 UTCfinished, eval accuracy 0.9020, config 3 of 9 (learning rate 0.00005, batch 16), seed 0, Wed 18 Nov, 16:41 UTCfinished, eval accuracy 0.9080, config 4 of 9 (batch 16), seed 1 of 2, Thu 19 Nov, 00:21 UTCfinished, eval accuracy 0.9160, config 4 of 9 (batch 16), seed 2 of 2, Thu 19 Nov, 02:14 UTCfailed at step 470 of 7274, eval accuracy -, config 5 of 9 (batch 16), seed 0, Fri 20 Nov, 13:18 UTCfinished, eval accuracy 0.9600, config 5 of 9 (batch 16), seed 0, Sat 21 Nov, 01:57 UTCfinished, eval accuracy 0.8876, config 6 of 9 (learning rate 0.0001, batch 16), seed 0, Sat 21 Nov, 12:36 UTCfinished, eval accuracy 0.9191, config 7 of 9 (learning rate 0.00002, batch 16), seed 0, Sat 21 Nov, 13:15 UTCfinished, eval accuracy 0.8944, config 8 of 9 (learning rate 0.00005, batch 16), seed 0, Sat 21 Nov, 13:54 UTCfinished, eval accuracy 0.8854, config 9 of 9 (batch 16), seed 1 of 2, Sat 21 Nov, 19:01 UTCfinished, eval accuracy 0.8921, config 9 of 9 (batch 16), seed 2 of 2, Sat 21 Nov, 19:40 UTC

1 of 12 finished, 0.9000, config 1 of 9 (learning rate 0.0001, batch 16), seed 0, Wed 18 Nov, 12:57 UTC

◂
Runs as a table
StartedStatusEval accuracyStepsConfig
Sat 21 Nov, 19:40 UTC| finished0.89211032 of 1032batch 16, seed 2 of 2
Sat 21 Nov, 19:01 UTC| finished0.88541032 of 1032batch 16, seed 1 of 2
Sat 21 Nov, 13:54 UTC| finished0.89441032 of 1032learning rate 0.00005, batch 16, seed 0
Sat 21 Nov, 13:15 UTC| finished0.91911032 of 1032learning rate 0.00002, batch 16, seed 0
Sat 21 Nov, 12:36 UTC| finished0.88761032 of 1032learning rate 0.0001, batch 16, seed 0
Sat 21 Nov, 01:57 UTC| finished0.96007274 of 7274batch 16, seed 0
Fri 20 Nov, 13:18 UTC× failed-470 of 7274batch 16, seed 0
Thu 19 Nov, 02:14 UTC| finished0.91601125 of 1125batch 16, seed 2 of 2
Thu 19 Nov, 00:21 UTC| finished0.90801125 of 1125batch 16, seed 1 of 2
Wed 18 Nov, 16:41 UTC| finished0.90201125 of 1125learning rate 0.00005, batch 16, seed 0
Wed 18 Nov, 14:41 UTC| finished0.92001125 of 1125learning rate 0.00002, batch 16, seed 0
Wed 18 Nov, 12:57 UTC| finished0.90001125 of 1125learning rate 0.0001, batch 16, seed 0

Plate

Model
answerdotai/ModernBERT-large, 395M parameters revision 45bb465
Data
Familiar decisions, shared draw, synthetic6000 train500 eval
Config
learning rate 0.0001, batch size 16, epochs 3, max length 512, precision bf16-mixed, loss ce, optimizer adamw, scheduler linear, seed 0
Hardware
Radeon 8060S, ROCm

Log

Since startMark
0.0 sphase starting
10.7 swent public
90.8 sphase training
1835.6 seval 1
1867.5 sepoch 1
3564.6 sepoch 2
3583.6 seval 2
5579.8 sepoch 3
5598.7 seval 3

Readings

114 rows as a table

rows 1 to 100 of 114

StepSecondsTrain lossEval accuracyLearning rate
1090.81.1371-1.49e-5
20155.81.3974-2.99e-5
30187.21.0520-4.48e-5
40226.81.2455-5.97e-5
50300.21.9144-7.46e-5
60348.91.8255-8.96e-5
70381.21.3532-9.97e-5
80431.01.2965-9.88e-5
90488.41.3526-9.78e-5
100537.61.1632-9.69e-5
110587.00.8636-9.59e-5
120636.10.6917-9.50e-5
130702.41.1763-9.40e-5
140736.41.1807-9.31e-5
150785.00.9222-9.22e-5
160817.31.0340-9.12e-5
170873.80.9059-9.03e-5
180913.30.7458-8.93e-5
190979.61.1460-8.84e-5
2001037.00.7496-8.74e-5
2101092.40.3283-8.65e-5
2201140.90.6923-8.55e-5
2301196.10.9239-8.46e-5
2401228.30.3011-8.36e-5
2501275.80.9826-8.27e-5
2601323.50.7966-8.18e-5
2701380.10.8350-8.08e-5
2801404.01.0444-7.99e-5
2901446.60.8100-7.89e-5
3001495.30.7456-7.80e-5
3101542.60.5385-7.70e-5
3201598.40.2059-7.61e-5
3301631.10.3209-7.51e-5
3401663.40.7645-7.42e-5
3501711.50.5096-7.33e-5
3601743.20.8410-7.23e-5
3701784.90.5611-7.14e-5
3751835.6-0.8080-
3801867.50.5795-7.04e-5
3901900.50.3152-6.95e-5
4001964.60.7435-6.85e-5
4101996.20.5877-6.76e-5
4202043.50.7485-6.66e-5
4302068.20.5232-6.57e-5
4402100.00.4119-6.47e-5
4502132.90.7136-6.38e-5
4602190.10.6089-6.29e-5
4702247.50.2996-6.19e-5
4802279.70.6797-6.10e-5
4902328.20.3194-6.00e-5
5002367.60.4688-5.91e-5
5102407.40.3443-5.81e-5
5202479.80.8210-5.72e-5
5302519.70.1671-5.62e-5
5402551.60.7174-5.53e-5
5502591.51.0226-5.43e-5
5602641.40.1477-5.34e-5
5702681.20.1794-5.25e-5
5802722.20.3227-5.15e-5
5902761.60.4874-5.06e-5
6002826.40.7014-4.96e-5
6102875.10.5775-4.87e-5
6202922.50.4563-4.77e-5
6302980.10.3630-4.68e-5
6403043.70.5144-4.58e-5
6503108.00.7252-4.49e-5
6603171.40.1645-4.40e-5
6703218.80.2143-4.30e-5
6803266.50.5204-4.21e-5
6903299.00.2715-4.11e-5
7003339.50.3291-4.02e-5
7103372.00.7098-3.92e-5
7203413.81.0246-3.83e-5
7303461.20.3361-3.73e-5
7403501.40.4932-3.64e-5
7503583.60.61500.85803.54e-5
7603612.70.2016-3.45e-5
7703660.70.3280-3.36e-5
7803716.20.5254-3.26e-5
7903771.90.4166-3.17e-5
8003835.20.3125-3.07e-5
8103874.50.5029-2.98e-5
8203931.40.1801-2.88e-5
8303962.70.1302-2.79e-5
8404019.40.1837-2.69e-5
8504099.10.3213-2.60e-5
8604162.70.2695-2.50e-5
8704202.00.1780-2.41e-5
8804251.10.2040-2.32e-5
8904299.30.4312-2.22e-5
9004346.40.3261-2.13e-5
9104386.90.5336-2.03e-5
9204450.70.1748-1.94e-5
9304491.10.2802-1.84e-5
9404531.20.4934-1.75e-5
9504595.70.2872-1.65e-5
9604645.30.1457-1.56e-5
9704713.20.1245-1.47e-5
9804761.50.2361-1.37e-5
9904818.60.4431-1.28e-5

GPU readings

114 rows as a table

rows 1 to 100 of 113

StepGPUUse %Power W, whole chip, CPU and GPUMemory GB, whole GPUTemp C, one board zone
100087.021.695.8
200099.025.393.8
300094.129.293.8
400084.129.295.8
500083.629.293.8
600083.829.294.8
700082.729.293.8
800065.129.295.8
900078.829.294.8
1000063.829.295.8
1100067.029.294.8
1200063.129.295.8
1300064.229.294.8
1400063.329.290.8
1500093.129.294.8
1600070.829.294.8
1700083.029.290.8
1800083.129.296.8
1900068.229.295.8
2000068.629.295.8
2100070.929.295.8
2200083.029.293.8
2400094.829.294.8
2500091.929.296.8
2600084.229.295.8
2700582.929.294.8
2800083.629.296.8
2900084.729.296.8
3000065.729.295.8
31000100.829.294.8
3200066.129.290.8
3300085.329.295.8
34001478.129.292.8
3500097.929.294.8
3600066.129.289.8
3700583.529.291.8
3750061.729.295.8
3800083.529.289.8
3900062.129.295.8
4000086.529.294.8
4100064.529.293.8
4200083.029.294.8
4300084.529.294.8
4400089.329.293.8
45000106.129.295.8
4600066.629.292.8
4700083.829.293.8
4800062.129.294.8
4900062.229.290.8
5000088.929.292.8
5100070.129.295.8
52000105.829.294.8
5300384.129.293.8
5400085.729.294.8
5500067.129.294.8
5600689.629.294.8
5700059.929.295.8
5800089.029.294.8
5900064.429.294.8
6000598.829.294.8
6100063.029.294.8
6200066.429.295.8
63002789.729.294.8
6400084.229.295.8
6500069.529.294.8
6600064.929.294.8
6700086.329.293.8
68002100.029.294.8
6900061.429.292.8
7000063.129.291.8
7100092.029.293.8
7200063.429.290.8
7300067.529.291.8
7400063.029.293.8
7500088.229.292.8
7600093.829.292.8
7700064.029.295.8
7800065.929.294.8
7900085.729.295.8
8000065.429.295.8
8100067.529.295.8
8200064.129.295.8
8300083.329.294.8
8400063.829.295.8
8500484.129.294.8
8600064.829.294.8
8700062.729.294.8
8800066.329.293.8
8900090.929.295.8
9000097.129.294.8
9100065.829.295.8
9200584.229.294.8
9300063.429.290.8
9400061.929.292.8
95001583.629.292.8
96002488.729.295.8
9700061.629.294.8
9800075.029.293.8
9900061.929.295.8
10000059.929.295.8

Worked out

ReadoutValueWorked out from
Run complete100%1125 steps over 1125
Step time4.85 s5312.9 s over 1095 steps, between readings. Startup and intervals that hold an eval are left out
Steps per minute12.460 s over the step time of 4.85 s
Epoch interval1856.1 sMean of 2 gaps between epoch marks
Train loss since first reading−0.97840.1586 at step 1125 minus 1.1371 at step 10
Lowest train loss0.0565Lowest of 113 readings, at step 1110
Eval accuracy, as shown90.0%0.9000 times 100, to one decimal. Changes on that plate are percentage points
Eval accuracy since first eval+0.09200.9000 minus 0.8080
Eval change+0.0420Eval 3 at 0.9000 minus eval 2 at 0.8580
Wrong answers50500 eval items minus 450 correct. 450 over 500 is the only count that gives 0.9000
Rank7 of 11By last eval accuracy, across this project's 11 runs with an eval
Best so far0.9600Highest last eval accuracy, run 7
Lowest last train loss0.0007Lowest last train loss across this project's 12 runs, run 11
Items per second3.31 step over 4.85 s, times batch size 16
Learning rate since first reading−1.49e-50.00e+0 minus 1.49e-5
Peak memory per GPU, whole GPU29.2 GBHighest sensor reading of any GPU, GPU 0, counting every process on it