CITADEL
StandbyNothing is live right now.
Eval accuracylast measured Wed 18 Nov, 20:29 UTC
%
0.9100455 of 500 eval, 45 wrong
▲ +7.40 pts across 3 evals▲ +1.60 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, 18:39 UTC
ran for
98:09
config
learning rate 0.0001, batch 16, seed 0

run1/11step1125/1125epoch3/3config1/9seed0lr0.0001batch16modelettin-encoder-400m 400Mrev7662476items6000+500

Loss train 0.1872 ▼ −1.0176 lowest 0.0420 at step 1000
Timebase
Range
Smoothing
113 readings, 30 min/div30 min/div10 min/div
00.511.5200:0030:0060:0090:00phase starting, 0.0 s to 90.1 sphase training, 90.1 s to 98:09epoch 1, 34:17epoch 2, 65:43epoch 3, 97:48trainingepoch 1epoch 2−1.0176since first1.2048Lowest last train loss across this project's 11 runs, run 10lowest last 0.0002, run 100.187200.511.5200:0030:0060:0090:00phase starting, 0.0 s to 90.1 sphase training, 90.1 s to 98:09epoch 1, 34:17epoch 2, 65:43epoch 3, 97:48trainingepoch 1epoch 21.2048Lowest last train loss across this project's 11 runs, run 10lowest last 0.0002, run 100.1872▼ −1.017600.511.5200:0010:0020:0030:0040:0050:0060:0070:0080:0090:00phase starting, 0.0 s to 90.1 sphase training, 90.1 s to 98:09epoch 1, 34:17epoch 2, 65:43epoch 3, 97:48trainingepoch 1epoch 21.2048Lowest last train loss across this project's 11 runs, run 10lowest last 0.0002, run 100.1872▼ −1.0176

Eval measures

Elapsed

1905.5 s

epoch interval

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

5.16 s per step, 11.6 steps a minute

Eval accuracy by eval

+0.0160

▲ on the eval before

eval 1 0.8360, eval 3 0.9100

Against the other runs

6 of 11

by last eval accuracy

best so far 0.9640, run 6

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 90.1 sphase training, 90.1 s to 98:09E1E2E3traininglowest last 0.0002Train loss0.1872▼ −1.01762.000.001.20480.1872last ▼ −1.0176best 0.9640, run 6Last-value range, 11 runs, 0.8854 to 0.9640Eval accuracy0.9100▲ +0.07401.000.800.83600.9100last ▲ +0.0740Learning rate0.00e+0▼ −1.49e-51.0e-40.0e+01.49e-5last 0.00e+0Power, whole chip, CPU and GPU53 W▼ −0 W58 W50 W53 Wlast 53 WHottest, one board zone68 °C0 °C75 °C60 °C68 °Clast 68 °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

53 W whole chip, CPU and GPU

use47%
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 ◂ older
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

1 of 11 finished, 0.9100, config 1 of 9 (learning rate 0.0001, batch 16), seed 0, Wed 18 Nov, 18:39 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 0.0001, 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
3.8 swent public
90.1 sphase training
2026.2 seval 1
2057.7 sepoch 1
3943.4 sepoch 2
3963.9 seval 2
5868.6 sepoch 3
5889.0 seval 3

Readings

114 rows as a table

rows 1 to 100 of 114

StepSecondsTrain lossEval accuracyLearning rate
1090.11.2048-1.49e-5
20140.81.3741-2.99e-5
30229.61.0748-4.48e-5
40283.61.2025-5.97e-5
50337.11.8886-7.46e-5
60391.91.7102-8.96e-5
70409.61.0523-9.97e-5
80444.91.1303-9.88e-5
90534.41.0810-9.78e-5
100615.70.9012-9.69e-5
110650.90.8832-9.59e-5
120677.90.5054-9.50e-5
130742.51.2702-9.40e-5
140798.40.7418-9.31e-5
150833.70.5214-9.22e-5
160887.11.2078-9.12e-5
170941.70.7921-9.03e-5
1801014.50.5986-8.93e-5
1901051.21.0857-8.84e-5
2001115.00.6489-8.74e-5
2101188.60.3932-8.65e-5
2201232.90.9364-8.55e-5
2301294.70.7927-8.46e-5
2401340.60.2933-8.36e-5
2501367.01.0228-8.27e-5
2601412.50.6997-8.18e-5
2701457.00.6417-8.08e-5
2801503.41.1161-7.99e-5
2901547.50.7507-7.89e-5
3001619.70.6632-7.80e-5
3101664.80.2650-7.70e-5
3201726.90.2760-7.61e-5
3301762.80.2090-7.51e-5
3401834.70.8542-7.42e-5
3501879.10.4456-7.33e-5
3601933.60.8630-7.23e-5
3701997.00.3845-7.14e-5
3752026.2-0.8360-
3802057.70.4727-7.04e-5
3902101.80.2490-6.95e-5
4002155.70.7080-6.85e-5
4102202.00.4889-6.76e-5
4202264.70.3867-6.66e-5
4302310.00.6326-6.57e-5
4402345.00.2725-6.47e-5
4502389.10.6345-6.38e-5
4602442.70.4006-6.29e-5
4702486.90.1079-6.19e-5
4802558.90.4096-6.10e-5
4902594.40.3996-6.00e-5
5002649.50.4160-5.91e-5
5102711.50.3374-5.81e-5
5202782.10.5503-5.72e-5
5302800.30.4686-5.62e-5
5402881.60.2057-5.53e-5
5502945.41.2508-5.43e-5
5602981.00.1354-5.34e-5
5703007.50.1372-5.25e-5
5803060.40.3582-5.15e-5
5903114.40.3132-5.06e-5
6003167.20.5906-4.96e-5
6103195.30.2227-4.87e-5
6203239.60.4643-4.77e-5
6303292.60.2917-4.68e-5
6403337.10.5756-4.58e-5
6503392.10.6058-4.49e-5
6603446.00.1007-4.40e-5
6703472.40.2054-4.30e-5
6803534.60.4113-4.21e-5
6903579.60.2189-4.11e-5
7003641.10.3201-4.02e-5
7103693.80.4657-3.92e-5
7203764.90.8539-3.83e-5
7303826.70.2866-3.73e-5
7403879.50.4135-3.64e-5
7503963.90.44500.89403.54e-5
7604018.70.1089-3.45e-5
7704074.10.1920-3.36e-5
7804100.30.2893-3.26e-5
7904154.30.2439-3.17e-5
8004215.30.2907-3.07e-5
8104286.10.2982-2.98e-5
8204330.90.1230-2.88e-5
8304394.60.1200-2.79e-5
8404429.70.1385-2.69e-5
8504482.00.0632-2.60e-5
8604544.80.1029-2.50e-5
8704616.70.1442-2.41e-5
8804688.10.1044-2.32e-5
8904732.10.2552-2.22e-5
9004769.60.3027-2.13e-5
9104787.30.3102-2.03e-5
9204822.90.1051-1.94e-5
9304875.70.1693-1.84e-5
9404911.20.2908-1.75e-5
9504964.40.1702-1.65e-5
9605008.30.0784-1.56e-5
9705072.20.0651-1.47e-5
9805108.50.1039-1.37e-5
9905154.80.4125-1.28e-5

GPU readings

114 rows as a table

rows 1 to 100 of 111

StepGPUUse %Power W, whole chip, CPU and GPUMemory GB, whole GPUTemp C, one board zone
100053.121.667.8
200053.025.369.8
300052.225.370.8
400054.729.271.8
500051.729.269.8
700053.329.266.8
800051.729.271.8
900053.429.272.8
1000053.029.267.8
1200054.429.273.8
1300053.129.272.8
1400056.329.274.8
1500053.729.272.8
1600055.229.271.8
1700052.929.271.8
1800053.029.268.8
1900053.829.271.8
2000053.429.268.8
2100056.329.270.8
2200053.229.270.8
2300053.229.269.8
2400056.329.270.8
2500251.529.265.8
2600054.729.270.8
2700052.329.269.8
2800053.829.270.8
2900051.729.272.8
3000053.229.267.8
3100053.329.271.8
3200054.829.270.8
3300054.729.270.8
3400053.329.269.8
3500053.929.271.8
3600055.029.272.8
3700353.029.264.8
3750052.629.268.8
3800053.029.270.8
3900053.629.269.8
4000053.329.271.8
4100054.029.271.8
4200052.929.270.8
43002053.129.267.8
4400051.929.269.8
4500052.329.269.8
4600054.129.271.8
4700053.729.270.8
4800053.029.267.8
4900052.829.271.8
5000053.429.270.8
5100053.329.270.8
5300051.629.270.8
5400052.929.269.8
5500053.129.266.8
5600053.829.267.8
5700052.229.271.8
5800053.029.271.8
5900053.229.268.8
6000053.929.272.8
6100055.629.270.8
6200252.129.269.8
6300052.829.267.8
6400051.229.273.8
6500053.829.270.8
6600052.629.269.8
6700053.429.269.8
6800053.129.270.8
6900054.829.271.8
7000055.029.271.8
7100053.529.272.8
7200056.029.270.8
7300052.829.270.8
7400051.229.271.8
7500055.329.269.8
7600053.129.270.8
7700053.429.265.8
7800053.729.267.8
7900052.929.267.8
8000251.929.271.8
8100053.129.270.8
8200052.429.268.8
8300054.629.273.8
8400052.929.270.8
8500053.429.271.8
8600054.829.271.8
8700055.529.270.8
8800053.129.271.8
8900052.729.268.8
9000053.629.265.8
9100053.729.272.8
9200353.229.267.8
9300053.129.268.8
9400054.729.270.8
9500053.629.269.8
9600054.129.270.8
9700052.229.267.8
9800053.829.270.8
9900053.229.274.8
10000054.629.270.8
10100051.229.271.8
10200053.129.269.8

Worked out

ReadoutValueWorked out from
Run complete100%1125 steps over 1125
Step time5.16 s5647.6 s over 1095 steps, between readings. Startup and intervals that hold an eval are left out
Steps per minute11.660 s over the step time of 5.16 s
Epoch interval1905.5 sMean of 2 gaps between epoch marks
Train loss since first reading−1.01760.1872 at step 1125 minus 1.2048 at step 10
Lowest train loss0.0420Lowest of 113 readings, at step 1000
Eval accuracy, as shown91.0%0.9100 times 100, to one decimal. Changes on that plate are percentage points
Eval accuracy since first eval+0.07400.9100 minus 0.8360
Eval change+0.0160Eval 3 at 0.9100 minus eval 2 at 0.8940
Wrong answers45500 eval items minus 455 correct. 455 over 500 is the only count that gives 0.9100
Rank6 of 11By last eval accuracy, across this project's 11 runs with an eval
Best so far0.9640Highest last eval accuracy, run 6
Lowest last train loss0.0002Lowest last train loss across this project's 11 runs, run 10
Items per second3.11 step over 5.16 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