CITADEL
StandbyNothing is live right now.
Eval accuracylast measured Sat 21 Nov, 19:40 UTC
%
0.8854394 of 445 eval, 51 wrong
▲ +8.32 pts across 3 evals▲ +0.67 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
Sat 21 Nov, 19:01 UTC
ran for
25:53
config
batch 16, seed 1 of 2

run11/12step1032/1032epoch3/3config9/9seed1/2batch16modelModernBERT-large 395Mrev45bb465items5503+445

Loss train 0.0007 ▼ −1.9618 lowest 0.0007 at step 1032 run before, run 10 ▼ −0.1381 step 1032
Timebase
Range
Smoothing
Reference
104 readings, 10 min/div5 min/div5 min/div
012300:0010:0020:00phase starting, 0.0 s to 37.0 sphase training, 37.0 s to 25:53epoch 1, 09:02epoch 2, 17:27epoch 3, 25:38trainingepoch 1epoch 2−1.9618since first1.9625Lowest last train loss across this project's 12 runs, run 11lowest last 0.0007, run 110.0007012300:0005:0010:0015:0020:0025:00phase starting, 0.0 s to 37.0 sphase training, 37.0 s to 25:53epoch 1, 09:02epoch 2, 17:27epoch 3, 25:38trainingepoch 1epoch 21.9625Lowest last train loss across this project's 12 runs, run 11lowest last 0.0007, run 110.0007▼ −1.9618012300:0005:0010:0015:0020:0025:00phase starting, 0.0 s to 37.0 sphase training, 37.0 s to 25:53epoch 1, 09:02epoch 2, 17:27epoch 3, 25:38trainingepoch 1epoch 21.9625Lowest last train loss across this project's 12 runs, run 11lowest last 0.0007, run 110.0007▼ −1.9618

Eval measures

Elapsed

498.0 s

epoch interval

1032 of 1032 steps, 3 of 3 epochs, 100%

1.45 s per step, 41.3 steps a minute

Eval accuracy by eval

+0.0067

▲ on the eval before

eval 1 0.8022, eval 3 0.8854

Against the other runs

11 of 11

by last eval accuracy

best so far 0.9600, run 7

▼ −0.0090 against run 10

Set-up

5503 train, 445 eval

synthetic, batch size 16, epochs 3, max length 512, precision bf16-mixed, loss ce, optimizer adamw, scheduler linear, seed 1

Seed spread, observed

0.0067
0.850.90

2 finished seeds

Learning rate

0.00e+0

▼ −3.28e-6 since first

On one axis

phase starting, 0.0 s to 37.0 sphase training, 37.0 s to 25:53E1E2E3traininglowest last 0.0007Train loss0.0007▼ −1.96183.000.001.96250.0007last ▼ −1.9618best 0.9600, run 7Last-value range, 11 runs, 0.8854 to 0.9600Eval accuracy0.8854▲ +0.08321.000.800.80220.8854last ▲ +0.0832Learning rate0.00e+0▼ −3.28e-62.0e-50.0e+03.28e-6last 0.00e+0Power, whole chip, CPU and GPU118 W▼ −5 W125 W110 W123 Wlast 118 WHottest, one board zone92 °C▼ −1 °C94 °C88 °C93 °Clast 92 °C00:0005:0010:0015:0020:0025:00
GPUpower: whole chip, CPU and GPUtemperature: one board zone

GPU 0

Wwhole chip, CPU and GPU

118 W whole chip, CPU and GPU

use93%
memory, whole GPU2.1 GB

peak 24.1 GB, whole GPU

92 °C one board zone

history 0 to 124 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
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

11 of 12 finished, 0.8854, config 9 of 9 (batch 16), seed 1 of 2, Sat 21 Nov, 19:01 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, plain-case intents, synthetic5503 train445 eval
Config
batch size 16, epochs 3, max length 512, precision bf16-mixed, loss ce, optimizer adamw, scheduler linear, seed 1
Hardware
Radeon 8060S, ROCm

Log

Since startMark
0.0 sphase starting
8.1 swent public
37.0 sphase training
531.4 seval 1
542.4 sepoch 1
1042.8 seval 2
1048.0 sepoch 2
1538.5 sepoch 3
1553.9 seval 3

Seeds

SeedStatusEval accuracy
seed 1| finished0.8854
seed 2| finished0.8921

observed spread 0.0067, 2 finished seeds

Readings

106 rows as a table

rows 1 to 100 of 106

StepSecondsTrain lossEval accuracyLearning rate
1037.01.9625-3.28e-6
2051.41.7777-6.56e-6
3065.51.3189-9.84e-6
4080.01.4750-1.31e-5
5094.41.3407-1.64e-5
60108.61.2515-1.97e-5
70123.11.8625-1.98e-5
80137.31.8334-1.96e-5
90151.50.7163-1.94e-5
100165.90.6606-1.92e-5
110180.01.1063-1.90e-5
120194.31.7146-1.88e-5
130208.70.9334-1.86e-5
140223.30.9869-1.84e-5
150237.60.8387-1.82e-5
160252.00.4658-1.80e-5
170266.30.4590-1.78e-5
180280.51.0278-1.75e-5
190295.00.5016-1.73e-5
200309.50.7284-1.71e-5
210323.90.8732-1.69e-5
220338.20.6480-1.67e-5
230352.40.3985-1.65e-5
240366.70.4903-1.63e-5
250381.10.6870-1.61e-5
260395.50.3939-1.59e-5
270409.80.7883-1.57e-5
280424.20.5972-1.55e-5
290438.60.2235-1.53e-5
300453.10.4688-1.51e-5
310467.30.8212-1.49e-5
320481.70.1008-1.47e-5
330496.00.5738-1.45e-5
340510.40.5380-1.43e-5
344531.4-0.8022-
350542.40.2452-1.40e-5
360556.80.4083-1.38e-5
370571.10.4526-1.36e-5
380585.50.1391-1.34e-5
390599.90.0897-1.32e-5
400614.10.1865-1.30e-5
410628.40.1793-1.28e-5
420642.90.3445-1.26e-5
430657.20.3738-1.24e-5
440671.70.1548-1.22e-5
450686.10.7069-1.20e-5
460700.30.3992-1.18e-5
470714.60.4479-1.16e-5
480729.01.2894-1.14e-5
490743.30.4419-1.12e-5
500757.60.3390-1.10e-5
510771.90.1921-1.08e-5
520786.20.0880-1.05e-5
530800.50.9500-1.03e-5
540814.90.4104-1.01e-5
550829.20.5035-9.93e-6
560843.70.4268-9.72e-6
570858.00.1277-9.52e-6
580872.30.0673-9.31e-6
590886.60.7129-9.10e-6
600901.20.0434-8.90e-6
610915.60.4089-8.69e-6
620930.00.1862-8.49e-6
630944.40.1408-8.28e-6
640958.80.1672-8.07e-6
650973.00.3744-7.87e-6
660987.30.2171-7.66e-6
6701001.60.2049-7.46e-6
6801015.90.4227-7.25e-6
6881042.8-0.8787-
6901048.00.6571-7.04e-6
7001062.00.0474-6.84e-6
7101076.40.0529-6.63e-6
7201090.60.1209-6.43e-6
7301104.90.1819-6.22e-6
7401119.20.3815-6.01e-6
7501133.60.2239-5.81e-6
7601148.00.1366-5.60e-6
7701162.30.4320-5.40e-6
7801176.70.1272-5.19e-6
7901191.20.0542-4.98e-6
8001205.50.1424-4.78e-6
8101219.90.5395-4.57e-6
8201234.10.0931-4.37e-6
8301248.40.0766-4.16e-6
8401262.70.1381-3.95e-6
8501277.00.0731-3.75e-6
8601291.50.2705-3.54e-6
8701305.70.0728-3.34e-6
8801320.10.1864-3.13e-6
8901334.40.0098-2.92e-6
9001348.80.0868-2.72e-6
9101363.10.0959-2.51e-6
9201377.40.0328-2.31e-6
9301392.00.0147-2.10e-6
9401406.50.0395-1.89e-6
9501420.90.0492-1.69e-6
9601435.40.3647-1.48e-6
9701449.70.2281-1.28e-6
9801464.00.3489-1.07e-6

GPU readings

106 rows as a table
StepGPUUse %Power W, whole chip, CPU and GPUMemory GB, whole GPUTemp C, one board zone
1000122.817.392.8
200100116.817.388.8
5000118.820.489.8
700100119.124.189.8
900100114.024.190.8
11000120.324.190.8
1400100114.424.190.8
16000123.024.190.8
1800100116.224.190.8
20000121.824.190.8
2300100118.024.190.8
25000116.524.190.8
2700100118.024.191.8
29000119.924.191.8
3200100118.924.191.8
3400100115.024.191.8
3500100116.424.192.8
37000121.324.190.8
3900100117.124.190.8
42000119.824.190.8
4400100116.324.190.8
4600100113.024.190.8
4800100119.124.190.8
5100100115.524.190.8
53000119.824.190.8
5500100118.324.191.8
58000115.224.191.8
6000100117.424.191.8
62000120.324.191.8
6400100119.124.191.8
6700100114.024.191.8
6800100116.824.190.8
7000100114.224.190.8
72000123.824.190.8
7400100115.624.190.8
77000120.624.190.8
7900100114.624.190.8
81000122.224.189.8
8300100117.624.190.8
86000115.024.190.8
8800100120.524.190.8
9000100113.024.190.8
9200100118.924.190.8
95000120.224.190.8
9700100117.824.190.8
9900100114.324.190.8
101000124.224.191.8
1032093117.52.191.8

Worked out

ReadoutValueWorked out from
Run complete100%1032 steps over 1032
Step time1.45 s1472.5 s over 1014 steps, between readings. Startup and intervals that hold an eval are left out
Steps per minute41.360 s over the step time of 1.45 s
Epoch interval498.0 sMean of 2 gaps between epoch marks
Train loss since first reading−1.96180.0007 at step 1032 minus 1.9625 at step 10
Lowest train loss0.0007Lowest of 104 readings, at step 1032
Eval accuracy, as shown88.5%0.8854 times 100, to one decimal. Changes on that plate are percentage points
Eval accuracy since first eval+0.08320.8854 minus 0.8022
Eval change+0.0067Eval 3 at 0.8854 minus eval 2 at 0.8787
Wrong answers51445 eval items minus 394 correct. 394 over 445 is the only count that gives 0.8854
Rank11 of 11By last eval accuracy, across this project's 11 runs with an eval
Best so far0.9600Highest last eval accuracy, run 7
Against the run before−0.00900.8854 minus 0.8944 on run 10
Lowest last train loss0.0007Lowest last train loss across this project's 12 runs, run 11
Seed spread, observed0.0067High 0.8921 minus low 0.8854 across 2 finished seeds
Items per second11.01 step over 1.45 s, times batch size 16
Train loss against run 10−0.13810.0007 minus 0.1388 on run 10, both at step 1032
Learning rate since first reading−3.28e-60.00e+0 minus 3.28e-6
Peak memory per GPU, whole GPU24.1 GBHighest sensor reading of any GPU, GPU 0, counting every process on it