CITADEL
StandbyNothing is live right now.
Eval accuracylast measured Thu 19 Nov, 00:21 UTC
%
0.9240462 of 500 eval, 38 wrong
▲ +8.80 pts across 3 evals▲ +1.80 pts on the eval before
0%25%50%75%100%0%25%50%75%100%0%25%50%75%100%

Familiar decider: ettin-encoder-400m

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, 22:27 UTC
ran for
101:10
config
learning rate 0.00005, batch 16, seed 0

run3/11step1125/1125epoch3/3config3/9seed0lr0.00005batch16modelettin-encoder-400m 400Mrev7662476items6000+500

Loss train 0.0993 ▼ −1.2139 lowest 0.0240 at step 1110 run before, run 2 ▼ −0.0462 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 87.7 sphase training, 87.7 s to 101:10epoch 1, 34:28epoch 2, 67:08epoch 3, 100:50trainingepoch 1epoch 2−1.2139since first1.3132Lowest last train loss across this project's 11 runs, run 10lowest last 0.0002, run 100.0993012300:0030:0060:0090:00phase starting, 0.0 s to 87.7 sphase training, 87.7 s to 101:10epoch 1, 34:28epoch 2, 67:08epoch 3, 100:50trainingepoch 1epoch 21.3132Lowest last train loss across this project's 11 runs, run 10lowest last 0.0002, run 100.0993▼ −1.2139012300:0030:0060:0090:00phase starting, 0.0 s to 87.7 sphase training, 87.7 s to 101:10epoch 1, 34:28epoch 2, 67:08epoch 3, 100:50trainingepoch 1epoch 21.3132Lowest last train loss across this project's 11 runs, run 10lowest last 0.0002, run 100.0993▼ −1.2139

Eval measures

Elapsed

1990.9 s

epoch interval

2046.6 sto eval 11981.8 sto epoch 22021.8 sto epoch 3

2046.6 s to eval 1, 1981.8 s to epoch 2, 2021.8 s to epoch 3

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

5.35 s per step, 11.2 steps a minute

Eval accuracy by eval

+0.0180

▲ on the eval before

eval 1 0.8360, eval 3 0.9240

Against the other runs

3 of 11

by last eval accuracy

best so far 0.9640, run 6

▲ +0.0120 against run 2

Set-up

6000 train, 500 eval

synthetic, learning rate 0.00005, batch size 16, epochs 3, max length 512, precision bf16-mixed, loss ce, optimizer adamw, scheduler linear, seed 0

Learning rate

0.00e+0

▼ −7.46e-6 since first

On one axis

phase starting, 0.0 s to 87.7 sphase training, 87.7 s to 101:10E1E2E3traininglowest last 0.0002Train loss0.0993▼ −1.21393.000.001.31320.0993last ▼ −1.2139best 0.9640, run 6Last-value range, 11 runs, 0.8854 to 0.9640Eval accuracy0.9240▲ +0.08801.000.800.83600.9240last ▲ +0.0880Learning rate0.00e+0▼ −7.46e-66.0e-50.0e+07.46e-6last 0.00e+0Power, whole chip, CPU and GPU53 W▲ +0 W56 W50 W53 Wlast 53 WHottest, one board zone68 °C▲ +1 °C75 °C60 °C67 °Clast 68 °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

use54%
memory, whole GPU2.6 GB

peak 29.2 GB, whole GPU

68 °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 0.9100, config 1 of 9 (learning rate 0.0001, batch 16), seed 0, Wed 18 Nov, 18:39 UTCfinished, eval accuracy 0.9120, config 2 of 9 (learning rate 0.00002, batch 16), seed 0, Wed 18 Nov, 20:29 UTCfinished, eval accuracy 0.9240, config 3 of 9 (learning rate 0.00005, batch 16), seed 0, Wed 18 Nov, 22:27 UTCfinished, eval accuracy 0.9200, config 4 of 9 (batch 16), seed 1 of 2, Thu 19 Nov, 04:02 UTCfinished, eval accuracy 0.9300, config 4 of 9 (batch 16), seed 2 of 2, Thu 19 Nov, 05:50 UTCfinished, eval accuracy 0.9640, config 5 of 9 (batch 16), seed 0, Sat 21 Nov, 05:11 UTCfinished, eval accuracy 0.8854, config 6 of 9 (learning rate 0.0001, batch 16), seed 0, Sat 21 Nov, 14:33 UTCfinished, eval accuracy 0.9034, config 7 of 9 (learning rate 0.00002, batch 16), seed 0, Sat 21 Nov, 15:12 UTCfinished, eval accuracy 0.8854, config 8 of 9 (learning rate 0.00005, batch 16), seed 0, Sat 21 Nov, 15:51 UTCfinished, eval accuracy 0.8899, config 9 of 9 (batch 16), seed 1 of 2, Sat 21 Nov, 20:19 UTCfinished, eval accuracy 0.8989, config 9 of 9 (batch 16), seed 2 of 2, Sat 21 Nov, 20:58 UTC1.000.85finished, eval accuracy 0.9100, config 1 of 9 (learning rate 0.0001, batch 16), seed 0, Wed 18 Nov, 18:39 UTCfinished, eval accuracy 0.9120, config 2 of 9 (learning rate 0.00002, batch 16), seed 0, Wed 18 Nov, 20:29 UTCfinished, eval accuracy 0.9240, config 3 of 9 (learning rate 0.00005, batch 16), seed 0, Wed 18 Nov, 22:27 UTCfinished, eval accuracy 0.9200, config 4 of 9 (batch 16), seed 1 of 2, Thu 19 Nov, 04:02 UTCfinished, eval accuracy 0.9300, config 4 of 9 (batch 16), seed 2 of 2, Thu 19 Nov, 05:50 UTCfinished, eval accuracy 0.9640, config 5 of 9 (batch 16), seed 0, Sat 21 Nov, 05:11 UTCfinished, eval accuracy 0.8854, config 6 of 9 (learning rate 0.0001, batch 16), seed 0, Sat 21 Nov, 14:33 UTCfinished, eval accuracy 0.9034, config 7 of 9 (learning rate 0.00002, batch 16), seed 0, Sat 21 Nov, 15:12 UTCfinished, eval accuracy 0.8854, config 8 of 9 (learning rate 0.00005, batch 16), seed 0, Sat 21 Nov, 15:51 UTCfinished, eval accuracy 0.8899, config 9 of 9 (batch 16), seed 1 of 2, Sat 21 Nov, 20:19 UTCfinished, eval accuracy 0.8989, config 9 of 9 (batch 16), seed 2 of 2, Sat 21 Nov, 20:58 UTC

3 of 11 finished, 0.9240, config 3 of 9 (learning rate 0.00005, batch 16), seed 0, Wed 18 Nov, 22:27 UTC

Runs as a table
StartedStatusEval accuracyStepsConfig
Sat 21 Nov, 20:58 UTC| finished0.89891032 of 1032batch 16, seed 2 of 2
Sat 21 Nov, 20:19 UTC| finished0.88991032 of 1032batch 16, seed 1 of 2
Sat 21 Nov, 15:51 UTC| finished0.88541032 of 1032learning rate 0.00005, batch 16, seed 0
Sat 21 Nov, 15:12 UTC| finished0.90341032 of 1032learning rate 0.00002, batch 16, seed 0
Sat 21 Nov, 14:33 UTC| finished0.88541032 of 1032learning rate 0.0001, batch 16, seed 0
Sat 21 Nov, 05:11 UTC| finished0.96407274 of 7274batch 16, seed 0
Thu 19 Nov, 05:50 UTC| finished0.93001125 of 1125batch 16, seed 2 of 2
Thu 19 Nov, 04:02 UTC| finished0.92001125 of 1125batch 16, seed 1 of 2
Wed 18 Nov, 22:27 UTC| finished0.92401125 of 1125learning rate 0.00005, batch 16, seed 0
Wed 18 Nov, 20:29 UTC| finished0.91201125 of 1125learning rate 0.00002, batch 16, seed 0
Wed 18 Nov, 18:39 UTC| finished0.91001125 of 1125learning rate 0.0001, batch 16, seed 0

Plate

Model
jhu-clsp/ettin-encoder-400m, 400M parameters revision 7662476
Data
Familiar decisions, shared draw, synthetic6000 train500 eval
Config
learning rate 0.00005, 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
13.8 swent public
87.7 sphase training
2046.6 seval 1
2068.5 sepoch 1
4028.4 sepoch 2
4049.0 seval 2
6050.3 sepoch 3
6070.7 seval 3

Readings

114 rows as a table

rows 1 to 100 of 114

StepSecondsTrain lossEval accuracyLearning rate
1087.71.3132-7.46e-6
20131.41.5311-1.49e-5
30182.10.9348-2.24e-5
40234.71.2720-2.99e-5
50306.01.9182-3.73e-5
60351.31.7628-4.48e-5
70386.61.3766-4.99e-5
80441.41.1811-4.94e-5
90494.61.1278-4.89e-5
100539.21.1402-4.84e-5
110583.61.0269-4.80e-5
120636.90.5145-4.75e-5
130707.41.0485-4.70e-5
140770.30.9026-4.66e-5
150823.00.5987-4.61e-5
160876.80.8969-4.56e-5
170913.40.7220-4.51e-5
180958.40.6424-4.47e-5
1901021.50.9954-4.42e-5
2001067.90.5226-4.37e-5
2101139.00.2864-4.32e-5
2201194.31.0001-4.28e-5
2301256.60.6976-4.23e-5
2401292.30.2254-4.18e-5
2501336.60.8012-4.14e-5
2601390.30.5900-4.09e-5
2701443.30.6760-4.04e-5
2801488.31.0987-3.99e-5
2901551.20.9204-3.95e-5
3001587.10.6478-3.90e-5
3101650.40.4726-3.85e-5
3201704.20.2488-3.80e-5
3301775.50.1928-3.76e-5
3401856.10.6937-3.71e-5
3501909.30.4104-3.66e-5
3601964.60.9733-3.62e-5
3702008.60.3698-3.57e-5
3752046.6-0.8360-
3802068.50.4294-3.52e-5
3902120.90.1496-3.47e-5
4002156.00.6744-3.43e-5
4102191.10.4599-3.38e-5
4202252.90.4126-3.33e-5
4302304.80.4503-3.28e-5
4402356.80.3364-3.24e-5
4502400.30.6504-3.19e-5
4602471.10.3143-3.14e-5
4702532.40.2253-3.10e-5
4802584.50.3484-3.05e-5
4902646.30.3498-3.00e-5
5002708.20.3761-2.95e-5
5102753.60.2889-2.91e-5
5202806.50.5890-2.86e-5
5302866.80.4606-2.81e-5
5402919.50.2866-2.76e-5
5502972.61.1023-2.72e-5
5603009.10.1834-2.67e-5
5703062.80.1177-2.62e-5
5803115.40.3058-2.58e-5
5903168.30.2342-2.53e-5
6003228.00.6162-2.48e-5
6103297.70.2106-2.43e-5
6203349.40.3089-2.39e-5
6303375.80.2121-2.34e-5
6403437.20.4055-2.29e-5
6503489.20.5423-2.24e-5
6603567.00.1405-2.20e-5
6703618.60.1373-2.15e-5
6803653.20.3315-2.10e-5
6903716.10.1569-2.06e-5
7003777.10.3034-2.01e-5
7103820.90.2696-1.96e-5
7203890.30.5652-1.91e-5
7303924.70.1638-1.87e-5
7403984.20.4842-1.82e-5
7504049.00.46440.90601.77e-5
7604128.40.0983-1.72e-5
7704181.40.2337-1.68e-5
7804216.50.3859-1.63e-5
7904278.10.0952-1.58e-5
8004323.00.1848-1.54e-5
8104384.90.3320-1.49e-5
8204463.90.1315-1.44e-5
8304535.00.1418-1.39e-5
8404596.80.0627-1.35e-5
8504657.50.0665-1.30e-5
8604719.90.1321-1.25e-5
8704798.70.2060-1.21e-5
8804834.60.1383-1.16e-5
8904879.10.2252-1.11e-5
9004940.90.2380-1.06e-5
9104985.50.0895-1.02e-5
9205046.90.1250-9.69e-6
9305098.90.1360-9.22e-6
9405152.40.5020-8.74e-6
9505206.70.2249-8.27e-6
9605250.40.0525-7.80e-6
9705312.40.0309-7.33e-6
9805375.50.1234-6.85e-6
9905410.80.4274-6.38e-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
100052.921.666.8
200250.925.369.8
3001051.529.270.8
400053.129.269.8
500055.129.271.8
600053.329.267.8
700054.629.270.8
800053.029.268.8
900252.029.271.8
1000053.829.271.8
1100054.629.270.8
1200055.129.270.8
1300055.129.269.8
1400051.529.270.8
1500054.429.270.8
1600053.129.271.8
1700053.129.268.8
1800054.429.270.8
1900053.129.266.8
2000052.029.269.8
2100053.429.268.8
2200053.329.268.8
2300052.729.264.8
2400053.429.270.8
2500053.129.267.8
2600653.029.268.8
2700053.929.271.8
2800053.629.268.8
29001952.529.269.8
3000052.929.269.8
3100053.529.265.8
3200054.029.268.8
3300053.229.269.8
3400053.229.267.8
3500052.929.271.8
3600051.829.270.8
3700053.029.268.8
3750053.129.266.8
3800053.129.268.8
3900053.429.268.8
4000052.929.268.8
4100053.529.269.8
4200052.129.270.8
4300053.029.268.8
44001652.929.266.8
4500053.229.268.8
4600053.129.268.8
4700053.029.269.8
4800054.529.268.8
4900052.829.268.8
5000053.429.265.8
5100051.329.269.8
5200054.429.268.8
5300053.129.266.8
5400053.029.270.8
55001252.329.269.8
5600052.929.268.8
5700053.529.267.8
5800951.629.267.8
5900054.529.268.8
6000052.929.267.8
6100052.229.269.8
6200052.029.264.8
6300053.029.267.8
6400852.129.268.8
6500652.529.267.8
6600053.329.268.8
6700353.029.270.8
6800053.129.268.8
6900053.229.270.8
7000253.129.267.8
7100051.329.268.8
7200053.229.274.8
7300053.729.269.8
7400053.229.268.8
7500053.329.270.8
7600052.829.269.8
7700052.129.268.8
7800051.429.270.8
7900053.929.269.8
8000054.729.269.8
8100053.329.268.8
8200053.329.269.8
8300053.629.268.8
8400054.129.268.8
8500054.229.271.8
8600054.229.267.8
8700053.429.272.8
8800055.529.268.8
8900053.329.269.8
9000052.829.267.8
9100053.329.271.8
9200053.329.268.8
9300053.029.268.8
9400052.929.269.8
9500052.529.267.8
9600054.329.269.8
9700053.029.267.8
9800052.129.267.8
9900055.329.271.8

Worked out

ReadoutValueWorked out from
Run complete100%1125 steps over 1125
Step time5.35 s5858.6 s over 1095 steps, between readings. Startup and intervals that hold an eval are left out
Steps per minute11.260 s over the step time of 5.35 s
Epoch interval1990.9 sMean of 2 gaps between epoch marks
Train loss since first reading−1.21390.0993 at step 1125 minus 1.3132 at step 10
Lowest train loss0.0240Lowest of 113 readings, at step 1110
Eval accuracy, as shown92.4%0.9240 times 100, to one decimal. Changes on that plate are percentage points
Eval accuracy since first eval+0.08800.9240 minus 0.8360
Eval change+0.0180Eval 3 at 0.9240 minus eval 2 at 0.9060
Wrong answers38500 eval items minus 462 correct. 462 over 500 is the only count that gives 0.9240
Rank3 of 11By last eval accuracy, across this project's 11 runs with an eval
Best so far0.9640Highest last eval accuracy, run 6
Against the run before+0.01200.9240 minus 0.9120 on run 2
Lowest last train loss0.0002Lowest last train loss across this project's 11 runs, run 10
Items per second3.01 step over 5.35 s, times batch size 16
Train loss against run 2−0.04620.0993 minus 0.1455 on run 2, both at step 1125
Learning rate since first reading−7.46e-60.00e+0 minus 7.46e-6
Peak memory per GPU, whole GPU29.2 GBHighest sensor reading of any GPU, GPU 0, counting every process on it