CITADEL
StandbyNothing is live right now.
Eval accuracylast measured Sun 22 Nov, 12:56 UTC
%
0.8584382 of 445 eval, 63 wrong
▲ +4.04 pts across 3 evals▲ +2.92 pts on the eval before
0%25%50%75%100%0%25%50%75%100%0%25%50%75%100%

Familiar decider: ettin-encoder-68m

finishedRadeon 8060S

One of the small models that decide whether Familiar, a desktop companion, reacts to something on the PC, and how. Trained on the same 6,000 synthetic decision points as every candidate, two ways: the reaction itself, and the plain-case intent that rules adjust for her traits.

started
Sun 22 Nov, 12:47 UTC
ran for
06:35
config
batch 16, seed 2 of 2

run10/10step1032/1032epoch3/3config8/8seed2/2batch16modelettin-encoder-68m 68Mrevac19ae4items5503+445

Loss train 0.1935 ▼ −2.8209 lowest 0.0320 at step 840 run before, run 9 ▲ +0.1816 step 1032
Timebase
Range
Smoothing
Reference
104 readings, 2 min/div2 min/div1 min/div
0123400:0002:0004:0006:00phase starting, 0.0 s to 14.6 sphase training, 14.6 s to 06:35epoch 1, 02:21epoch 2, 04:28epoch 3, 06:31trainingepoch 1epoch 2−2.8209since first3.0144Lowest last train loss across this project's 10 runs, run 9lowest last 0.0119, run 90.19350123400:0002:0004:0006:00phase starting, 0.0 s to 14.6 sphase training, 14.6 s to 06:35epoch 1, 02:21epoch 2, 04:28epoch 3, 06:31trainingepoch 1epoch 23.0144Lowest last train loss across this project's 10 runs, run 9lowest last 0.0119, run 90.1935▼ −2.82090123400:0001:0002:0003:0004:0005:0006:00phase starting, 0.0 s to 14.6 sphase training, 14.6 s to 06:35epoch 1, 02:21epoch 2, 04:28epoch 3, 06:31trainingepoch 1epoch 23.0144Lowest last train loss across this project's 10 runs, run 9lowest last 0.0119, run 90.1935▼ −2.8209

Eval measures

Elapsed

125.1 s

epoch interval

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

0.36 s per step, 166.5 steps a minute

Eval accuracy by eval

+0.0292

▲ on the eval before

eval 1 0.8180, eval 3 0.8584

Against the other runs

10 of 10

by last eval accuracy

best so far 0.9120, run 8

▼ −0.0247 against run 9

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 2

Seed spread, observed

0.0247
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 14.6 sphase training, 14.6 s to 06:35E1E2E3traininglowest last 0.0119Train loss0.1935▼ −2.82094.000.003.01440.1935last ▼ −2.8209best 0.9120, run 8Last-value range, 10 runs, 0.8584 to 0.9120Eval accuracy0.8584▲ +0.04040.950.800.81800.8584last ▲ +0.0404Learning rate0.00e+0▼ −3.28e-62.0e-50.0e+03.28e-6last 0.00e+0Power, whole chip, CPU and GPU116 W▼ −1 W122 W114 W117 Wlast 116 WHottest, one board zone91 °C▲ +6 °C92 °C84 °C85 °Clast 91 °C00:0001:0002:0003:0004:0005:0006:00
GPUpower: whole chip, CPU and GPUtemperature: one board zone

GPU 0

Wwhole chip, CPU and GPU

116 W whole chip, CPU and GPU

use85%
memory, whole GPU0.8 GB

peak 8.6 GB, whole GPU

91 °C one board zone

history 0 to 120 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 newer ▸
0.950.850.950.85finished, eval accuracy 0.8780, config 1 of 8 (learning rate 0.0001, batch 16), seed 0, Sun 22 Nov, 09:36 UTCfinished, eval accuracy 0.8800, config 2 of 8 (learning rate 0.00002, batch 16), seed 0, Sun 22 Nov, 09:46 UTCfinished, eval accuracy 0.8860, config 3 of 8 (learning rate 0.00005, batch 16), seed 0, Sun 22 Nov, 09:56 UTCfinished, eval accuracy 0.8831, config 4 of 8 (learning rate 0.0001, batch 16), seed 0, Sun 22 Nov, 10:05 UTCfinished, eval accuracy 0.9034, config 5 of 8 (learning rate 0.00002, batch 16), seed 0, Sun 22 Nov, 10:14 UTCfinished, eval accuracy 0.8809, config 6 of 8 (learning rate 0.00005, batch 16), seed 0, Sun 22 Nov, 10:24 UTCfinished, eval accuracy 0.9060, config 7 of 8 (batch 16), seed 1 of 2, Sun 22 Nov, 12:18 UTCfinished, eval accuracy 0.9120, config 7 of 8 (batch 16), seed 2 of 2, Sun 22 Nov, 12:28 UTCfinished, eval accuracy 0.8831, config 8 of 8 (batch 16), seed 1 of 2, Sun 22 Nov, 12:38 UTCfinished, eval accuracy 0.8584, config 8 of 8 (batch 16), seed 2 of 2, Sun 22 Nov, 12:47 UTC0.950.85finished, eval accuracy 0.8780, config 1 of 8 (learning rate 0.0001, batch 16), seed 0, Sun 22 Nov, 09:36 UTCfinished, eval accuracy 0.8800, config 2 of 8 (learning rate 0.00002, batch 16), seed 0, Sun 22 Nov, 09:46 UTCfinished, eval accuracy 0.8860, config 3 of 8 (learning rate 0.00005, batch 16), seed 0, Sun 22 Nov, 09:56 UTCfinished, eval accuracy 0.8831, config 4 of 8 (learning rate 0.0001, batch 16), seed 0, Sun 22 Nov, 10:05 UTCfinished, eval accuracy 0.9034, config 5 of 8 (learning rate 0.00002, batch 16), seed 0, Sun 22 Nov, 10:14 UTCfinished, eval accuracy 0.8809, config 6 of 8 (learning rate 0.00005, batch 16), seed 0, Sun 22 Nov, 10:24 UTCfinished, eval accuracy 0.9060, config 7 of 8 (batch 16), seed 1 of 2, Sun 22 Nov, 12:18 UTCfinished, eval accuracy 0.9120, config 7 of 8 (batch 16), seed 2 of 2, Sun 22 Nov, 12:28 UTCfinished, eval accuracy 0.8831, config 8 of 8 (batch 16), seed 1 of 2, Sun 22 Nov, 12:38 UTCfinished, eval accuracy 0.8584, config 8 of 8 (batch 16), seed 2 of 2, Sun 22 Nov, 12:47 UTC

10 of 10 finished, 0.8584, config 8 of 8 (batch 16), seed 2 of 2, Sun 22 Nov, 12:47 UTC

▸
Runs as a table
StartedStatusEval accuracyStepsConfig
Sun 22 Nov, 12:47 UTC| finished0.85841032 of 1032batch 16, seed 2 of 2
Sun 22 Nov, 12:38 UTC| finished0.88311032 of 1032batch 16, seed 1 of 2
Sun 22 Nov, 12:28 UTC| finished0.91201125 of 1125batch 16, seed 2 of 2
Sun 22 Nov, 12:18 UTC| finished0.90601125 of 1125batch 16, seed 1 of 2
Sun 22 Nov, 10:24 UTC| finished0.88091032 of 1032learning rate 0.00005, batch 16, seed 0
Sun 22 Nov, 10:14 UTC| finished0.90341032 of 1032learning rate 0.00002, batch 16, seed 0
Sun 22 Nov, 10:05 UTC| finished0.88311032 of 1032learning rate 0.0001, batch 16, seed 0
Sun 22 Nov, 09:56 UTC| finished0.88601125 of 1125learning rate 0.00005, batch 16, seed 0
Sun 22 Nov, 09:46 UTC| finished0.88001125 of 1125learning rate 0.00002, batch 16, seed 0
Sun 22 Nov, 09:36 UTC| finished0.87801125 of 1125learning rate 0.0001, batch 16, seed 0

Plate

Model
jhu-clsp/ettin-encoder-68m, 68M parameters revision ac19ae4
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 2
Hardware
Radeon 8060S, ROCm

Log

Since startMark
0.0 sphase starting
2.6 swent public
14.6 sphase training
138.6 seval 1
141.1 sepoch 1
267.0 seval 2
268.2 sepoch 2
391.4 sepoch 3
395.6 seval 3

Seeds

SeedStatusEval accuracy
seed 1| finished0.8831
seed 2| finished0.8584

observed spread 0.0247, 2 finished seeds

Readings

106 rows as a table

rows 1 to 100 of 106

StepSecondsTrain lossEval accuracyLearning rate
1014.63.0144-3.28e-6
2018.12.2807-6.56e-6
3021.71.5409-9.84e-6
4025.31.4748-1.31e-5
5028.81.6498-1.64e-5
6032.41.6349-1.97e-5
7035.91.8382-1.98e-5
8039.51.7066-1.96e-5
9043.11.5594-1.94e-5
10046.81.9350-1.92e-5
11050.41.3929-1.90e-5
12054.01.1792-1.88e-5
13057.61.5047-1.86e-5
14061.11.2952-1.84e-5
15064.71.4281-1.82e-5
16068.31.0530-1.80e-5
17072.00.4634-1.78e-5
18075.60.6513-1.75e-5
19079.20.5889-1.73e-5
20082.81.0253-1.71e-5
21086.40.3595-1.69e-5
22090.00.6830-1.67e-5
23093.60.6579-1.65e-5
24097.10.4934-1.63e-5
250100.70.9138-1.61e-5
260104.20.5266-1.59e-5
270107.80.7795-1.57e-5
280111.40.7559-1.55e-5
290115.00.6378-1.53e-5
300118.61.0006-1.51e-5
310122.20.5355-1.49e-5
320125.80.5227-1.47e-5
330129.40.2748-1.45e-5
340132.90.2690-1.43e-5
344138.6-0.8180-
350141.10.9162-1.40e-5
360144.70.1071-1.38e-5
370148.30.5071-1.36e-5
380151.90.2514-1.34e-5
390155.50.4762-1.32e-5
400159.00.1349-1.30e-5
410162.60.3485-1.28e-5
420166.20.2890-1.26e-5
430169.80.9143-1.24e-5
440173.40.4399-1.22e-5
450177.00.2620-1.20e-5
460180.60.4339-1.18e-5
470184.20.2360-1.16e-5
480187.80.1378-1.14e-5
490191.40.2631-1.12e-5
500195.00.1370-1.10e-5
510198.60.6582-1.08e-5
520202.20.3131-1.05e-5
530205.80.5637-1.03e-5
540209.40.4584-1.01e-5
550213.00.8533-9.93e-6
560216.60.2468-9.72e-6
570220.20.2593-9.52e-6
580223.90.2662-9.31e-6
590227.50.3522-9.10e-6
600231.10.5158-8.90e-6
610234.70.5537-8.69e-6
620238.30.5509-8.49e-6
630241.80.2642-8.28e-6
640245.40.5048-8.07e-6
650249.00.2035-7.87e-6
660252.70.5013-7.66e-6
670256.30.4229-7.46e-6
680259.90.2144-7.25e-6
688267.0-0.8292-
690268.20.2181-7.04e-6
700271.80.5616-6.84e-6
710275.40.1777-6.63e-6
720278.90.0719-6.43e-6
730282.60.1455-6.22e-6
740286.10.1779-6.01e-6
750289.70.4181-5.81e-6
760293.30.1179-5.60e-6
770296.90.5308-5.40e-6
780300.50.0827-5.19e-6
790304.10.0925-4.98e-6
800307.70.0470-4.78e-6
810311.30.0557-4.57e-6
820315.00.6586-4.37e-6
830318.60.3047-4.16e-6
840322.20.0320-3.95e-6
850325.70.3700-3.75e-6
860329.30.2135-3.54e-6
870333.00.1426-3.34e-6
880336.60.2318-3.13e-6
890340.20.0689-2.92e-6
900343.80.0525-2.72e-6
910347.30.0875-2.51e-6
920351.00.1297-2.31e-6
930354.60.3053-2.10e-6
940358.20.0977-1.89e-6
950361.90.1693-1.69e-6
960365.40.1360-1.48e-6
970369.00.4101-1.28e-6
980372.60.0527-1.07e-6

GPU readings

106 rows as a table
StepGPUUse %Power W, whole chip, CPU and GPUMemory GB, whole GPUTemp C, one board zone
500100116.87.484.8
1400100116.28.685.8
2300100116.78.686.8
320067120.48.686.8
4000100116.48.687.8
4900100117.08.687.8
5800100116.68.688.8
6700100116.58.688.8
7500100116.68.688.8
8400100118.28.688.8
930068119.28.688.8
10200100117.38.689.8
1032085115.50.890.8

Worked out

ReadoutValueWorked out from
Run complete100%1032 steps over 1032
Step time0.36 s363.3 s over 1008 steps, between readings. Startup and intervals that hold an eval are left out
Steps per minute166.560 s over the step time of 0.36 s
Epoch interval125.1 sMean of 2 gaps between epoch marks
Train loss since first reading−2.82090.1935 at step 1032 minus 3.0144 at step 10
Lowest train loss0.0320Lowest of 104 readings, at step 840
Eval accuracy, as shown85.8%0.8584 times 100, to one decimal. Changes on that plate are percentage points
Eval accuracy since first eval+0.04040.8584 minus 0.8180
Eval change+0.0292Eval 3 at 0.8584 minus eval 2 at 0.8292
Wrong answers63445 eval items minus 382 correct. 382 over 445 is the only count that gives 0.8584
Rank10 of 10By last eval accuracy, across this project's 10 runs with an eval
Best so far0.9120Highest last eval accuracy, run 8
Against the run before−0.02470.8584 minus 0.8831 on run 9
Lowest last train loss0.0119Lowest last train loss across this project's 10 runs, run 9
Seed spread, observed0.0247High 0.8831 minus low 0.8584 across 2 finished seeds
Items per second44.41 step over 0.36 s, times batch size 16
Train loss against run 9+0.18160.1935 minus 0.0119 on run 9, both at step 1032
Learning rate since first reading−3.28e-60.00e+0 minus 3.28e-6
Peak memory per GPU, whole GPU8.6 GBHighest sensor reading of any GPU, GPU 0, counting every process on it