CITADEL
StandbyNothing is live right now.
Eval accuracylast measured Thu 19 Nov, 02:14 UTC
%
0.9080454 of 500 eval, 46 wrong
▲ +9.00 pts across 3 evals▲ +2.80 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
Thu 19 Nov, 00:21 UTC
ran for
101:15
config
batch 16, seed 1 of 2

run4/12step1125/1125epoch3/3config4/9seed1/2batch16modelModernBERT-large 395Mrev45bb465items6000+500

Loss train 0.0120 ▼ −2.2399 lowest 0.0119 at step 870 run before, run 3 ▼ −0.1407 step 1125
Timebase
Range
Smoothing
Reference
113 readings, 30 min/div30 min/div30 min/div
012300:0030:0060:0090:00phase starting, 0.0 s to 01:46phase training, 01:46 to 101:15epoch 1, 35:38epoch 2, 69:02epoch 3, 100:55trainingepoch 1epoch 2−2.2399since first2.2519Lowest last train loss across this project's 12 runs, run 11lowest last 0.0007, run 110.0120012300:0030:0060:0090:00phase starting, 0.0 s to 01:46phase training, 01:46 to 101:15epoch 1, 35:38epoch 2, 69:02epoch 3, 100:55trainingepoch 1epoch 22.2519Lowest last train loss across this project's 12 runs, run 11lowest last 0.0007, run 110.0120▼ −2.2399012300:0030:0060:0090:00phase starting, 0.0 s to 01:46phase training, 01:46 to 101:15epoch 1, 35:38epoch 2, 69:02epoch 3, 100:55trainingepoch 1epoch 22.2519Lowest last train loss across this project's 12 runs, run 11lowest last 0.0007, run 110.0120▼ −2.2399

Eval measures

Elapsed

1958.2 s

epoch interval

2115.0 sto eval 12027.5 sto epoch 21912.6 sto epoch 3

2115.0 s to eval 1, 2027.5 s to epoch 2, 1912.6 s to epoch 3

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

5.31 s per step, 11.3 steps a minute

Eval accuracy by eval

+0.0280

▲ on the eval before

eval 1 0.8180, eval 3 0.9080

Against the other runs

5 of 11

by last eval accuracy

best so far 0.9600, run 7

▲ +0.0060 against run 3

Set-up

6000 train, 500 eval

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

Seed spread, observed

0.0080
0.900.95

2 finished seeds

Learning rate

0.00e+0

▼ −2.99e-6 since first

On one axis

phase starting, 0.0 s to 01:46phase training, 01:46 to 101:15E1E2E3traininglowest last 0.0007Train loss0.0120▼ −2.23993.000.002.25190.0120last ▼ −2.2399best 0.9600, run 7Last-value range, 11 runs, 0.8854 to 0.9600Eval accuracy0.9080▲ +0.09001.000.800.81800.9080last ▲ +0.0900Learning rate0.00e+0▼ −2.99e-62.0e-50.0e+02.99e-6last 0.00e+0Power, whole chip, CPU and GPU53 W▼ −0 W58 W50 W53 Wlast 53 WHottest, one board zone69 °C▼ −1 °C75 °C60 °C70 °Clast 69 °C00:0030:0060:0090:00
GPUpower: whole chip, CPU and GPUtemperature: one board zone

GPU 0

Wwhole chip, CPU and GPU

53 W whole chip, CPU and GPU

use45%
memory, whole GPU2.6 GB

peak 33.6 GB, whole GPU

69 °C one board zone

history 0 to 56 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

4 of 12 finished, 0.9080, config 4 of 9 (batch 16), seed 1 of 2, Thu 19 Nov, 00:21 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
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
1.6 swent public
106.9 sphase training
2115.0 seval 1
2138.7 sepoch 1
4142.6 sepoch 2
4163.0 seval 2
6055.1 sepoch 3
6075.5 seval 3

Seeds

SeedStatusEval accuracy
seed 1| finished0.9080
seed 2| finished0.9160

observed spread 0.0080, 2 finished seeds

Readings

114 rows as a table

rows 1 to 100 of 114

StepSecondsTrain lossEval accuracyLearning rate
10106.92.2519-2.99e-6
20167.50.9155-5.97e-6
30203.10.9990-8.96e-6
40239.11.6580-1.19e-5
50299.41.8084-1.49e-5
60325.82.3808-1.79e-5
70367.61.0272-1.99e-5
80413.21.3516-1.98e-5
90456.71.4622-1.96e-5
100491.70.8679-1.94e-5
110579.72.1377-1.92e-5
120607.81.2844-1.90e-5
130653.10.8265-1.88e-5
140713.30.8469-1.86e-5
150756.90.8064-1.84e-5
160819.70.9016-1.82e-5
170879.40.8020-1.81e-5
180941.60.8420-1.79e-5
190993.61.2867-1.77e-5
2001028.40.5798-1.75e-5
2101090.30.4060-1.73e-5
2201162.80.5921-1.71e-5
2301215.00.4525-1.69e-5
2401285.20.4006-1.67e-5
2501328.70.7003-1.65e-5
2601381.60.5692-1.64e-5
2701451.80.4161-1.62e-5
2801530.20.9554-1.60e-5
2901591.30.3193-1.58e-5
3001638.90.2981-1.56e-5
3101691.90.8819-1.54e-5
3201753.80.3111-1.52e-5
3301832.20.7674-1.50e-5
3401893.50.4969-1.48e-5
3501956.10.6849-1.47e-5
3602025.50.3456-1.45e-5
3702068.60.3038-1.43e-5
3752115.0-0.8180-
3802138.70.5227-1.41e-5
3902198.10.4003-1.39e-5
4002258.30.3910-1.37e-5
4102301.00.5285-1.35e-5
4202343.90.7267-1.33e-5
4302404.30.6050-1.31e-5
4402474.40.2809-1.29e-5
4502525.91.0550-1.28e-5
4602587.80.5659-1.26e-5
4702623.30.2522-1.24e-5
4802701.60.7082-1.22e-5
4902771.50.2522-1.20e-5
5002831.90.1534-1.18e-5
5102874.80.4579-1.16e-5
5202935.50.5013-1.14e-5
5302979.30.1039-1.12e-5
5403013.90.1387-1.11e-5
5503081.70.4104-1.09e-5
5603140.50.6269-1.07e-5
5703211.60.4978-1.05e-5
5803279.40.7634-1.03e-5
5903315.50.0954-1.01e-5
6003384.70.1300-9.92e-6
6103429.20.1147-9.74e-6
6203491.00.5537-9.55e-6
6303533.80.1777-9.36e-6
6403586.00.0841-9.17e-6
6503646.60.2757-8.98e-6
6603689.30.0280-8.79e-6
6703743.50.1536-8.60e-6
6803796.20.6522-8.41e-6
6903848.20.7836-8.22e-6
7003883.70.4548-8.03e-6
7103926.90.5680-7.84e-6
7203979.50.4027-7.66e-6
7304030.60.2868-7.47e-6
7404082.80.3508-7.28e-6
7504163.00.16210.88007.09e-6
7604215.00.3350-6.90e-6
7704258.80.0681-6.71e-6
7804311.30.3810-6.52e-6
7904355.90.4711-6.33e-6
8004391.70.3502-6.14e-6
8104452.80.2263-5.95e-6
8204515.30.4245-5.77e-6
8304595.00.7195-5.58e-6
8404638.60.2999-5.39e-6
8504664.70.6422-5.20e-6
8604700.70.9913-5.01e-6
8704753.50.0119-4.82e-6
8804789.20.4281-4.63e-6
8904859.10.1236-4.44e-6
9004902.40.0534-4.25e-6
9104955.30.2234-4.06e-6
9205007.80.1672-3.88e-6
9305061.80.3288-3.69e-6
9405096.70.0645-3.50e-6
9505141.30.1208-3.31e-6
9605202.10.2860-3.12e-6
9705263.60.3483-2.93e-6
9805325.00.2802-2.74e-6
9905404.20.0348-2.55e-6

GPU readings

114 rows as a table

rows 1 to 100 of 112

StepGPUUse %Power W, whole chip, CPU and GPUMemory GB, whole GPUTemp C, one board zone
100053.022.769.8
200053.822.769.8
300053.322.768.8
400054.126.267.8
500053.026.268.8
600053.230.067.8
700052.930.067.8
800053.830.072.8
900053.230.065.8
1000052.130.070.8
1100053.330.070.8
1200053.430.069.8
1300053.730.068.8
1400053.530.069.8
1500252.230.068.8
1600053.333.667.8
1700053.433.668.8
1800052.533.670.8
1900053.233.666.8
2000052.833.668.8
2100054.033.668.8
2200154.333.670.8
23001852.233.669.8
2400052.833.667.8
2500055.333.669.8
2600053.133.668.8
2700054.433.669.8
2800451.733.669.8
2900053.033.671.8
3000053.733.669.8
3100051.333.672.8
3200052.833.669.8
3300052.633.669.8
3400053.033.670.8
3500052.033.669.8
3600053.233.671.8
3700053.433.668.8
3750051.533.668.8
3800053.533.667.8
3900053.233.667.8
4000052.233.669.8
4100053.033.668.8
4200051.833.668.8
4300054.033.669.8
4400053.133.668.8
4500352.933.670.8
4600052.633.665.8
4700053.733.670.8
4800052.233.669.8
4900052.933.666.8
5000552.433.667.8
5100052.433.669.8
5200053.433.668.8
5300052.933.666.8
5400053.533.669.8
5500052.533.669.8
5600053.233.668.8
5700053.133.668.8
5800053.233.670.8
5900052.933.667.8
6000152.733.669.8
6100053.533.668.8
6200053.133.667.8
63002053.333.669.8
6400053.233.667.8
65001352.233.668.8
6600053.733.668.8
6700052.933.664.8
6800053.133.671.8
6900053.033.667.8
7000153.633.671.8
7100053.133.668.8
7200051.933.667.8
7300054.633.667.8
7400153.133.669.8
7500053.233.666.8
7600053.233.669.8
7700052.933.668.8
7800053.333.668.8
7900051.533.669.8
8000053.233.668.8
8100053.833.670.8
8200053.433.669.8
8300152.033.671.8
8400053.233.665.8
8500053.233.665.8
8600053.233.670.8
8700052.633.667.8
8800051.333.667.8
8900053.033.666.8
9000053.933.667.8
9100053.333.668.8
9200052.933.667.8
9300053.933.666.8
9400053.633.668.8
9500052.633.668.8
9600052.733.667.8
9700053.233.667.8
9800053.833.668.8
9900053.433.668.8

Worked out

ReadoutValueWorked out from
Run complete100%1125 steps over 1125
Step time5.31 s5842.0 s over 1100 steps, between readings. Startup and intervals that hold an eval are left out
Steps per minute11.360 s over the step time of 5.31 s
Epoch interval1958.2 sMean of 2 gaps between epoch marks
Train loss since first reading−2.23990.0120 at step 1125 minus 2.2519 at step 10
Lowest train loss0.0119Lowest of 113 readings, at step 870
Eval accuracy, as shown90.8%0.9080 times 100, to one decimal. Changes on that plate are percentage points
Eval accuracy since first eval+0.09000.9080 minus 0.8180
Eval change+0.0280Eval 3 at 0.9080 minus eval 2 at 0.8800
Wrong answers46500 eval items minus 454 correct. 454 over 500 is the only count that gives 0.9080
Rank5 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+0.00600.9080 minus 0.9020 on run 3
Lowest last train loss0.0007Lowest last train loss across this project's 12 runs, run 11
Seed spread, observed0.0080High 0.9160 minus low 0.9080 across 2 finished seeds
Items per second3.01 step over 5.31 s, times batch size 16
Train loss against run 3−0.14070.0120 minus 0.1527 on run 3, both at step 1125
Learning rate since first reading−2.99e-60.00e+0 minus 2.99e-6
Peak memory per GPU, whole GPU33.6 GBHighest sensor reading of any GPU, GPU 0, counting every process on it