CITADEL
StandbyNothing is live right now.
Eval accuracylast measured Sat 21 Nov, 08:25 UTC
%
0.9640482 of 500 eval, 18 wrong
▲ +2.60 pts across 2 evals▲ +2.60 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
Sat 21 Nov, 05:11 UTC
ran for
184:40
config
batch 16, seed 0

run6/11step7274/7274epoch2/2config5/9seed0batch16modelettin-encoder-400m 400Mrev7662476items58183+500

Loss train 0.0129 ▼ −2.9902 lowest 0.0001 at step 6820 run before, run 5
Timebase
Range
Smoothing
Reference
728 readings, 60 min/div60 min/div30 min/div
0123400:0060:00120:00180:00phase starting, 0.0 s to 58.5 sphase training, 58.5 s to 184:40epoch 1, 92:45epoch 2, 184:22trainingepoch 1−2.9902since first3.0031Lowest last train loss across this project's 11 runs, run 10lowest last 0.0002, run 100.01290123400:0060:00120:00180:00phase starting, 0.0 s to 58.5 sphase training, 58.5 s to 184:40epoch 1, 92:45epoch 2, 184:22trainingepoch 13.0031Lowest last train loss across this project's 11 runs, run 10lowest last 0.0002, run 100.0129▼ −2.99020123400:0030:0060:0090:00120:00150:00180:00phase starting, 0.0 s to 58.5 sphase training, 58.5 s to 184:40epoch 1, 92:45epoch 2, 184:22trainingepoch 13.0031Lowest last train loss across this project's 11 runs, run 10lowest last 0.0002, run 100.0129▼ −2.9902

Eval measures

Elapsed

5496.6 s

epoch interval

5559.4 sto eval 15502.8 sto epoch 2

5559.4 s to eval 1, 5502.8 s to epoch 2

7274 of 7274 steps, 2 of 2 epochs, 100%

1.51 s per step, 39.6 steps a minute

Eval accuracy by eval

+0.0260

▲ on the eval before

eval 1 0.9380, eval 2 0.9640

Against the other runs

1 of 11

by last eval accuracy

best so far 0.9640, run 6

▲ +0.0340 against run 5

Set-up

58183 train, 500 eval

synthetic, batch size 16, epochs 2, max length 512, precision bf16-mixed, loss ce, optimizer adamw, scheduler linear, seed 0

Learning rate

0.00e+0

▼ −1.15e-6 since first

On one axis

phase starting, 0.0 s to 58.5 sphase training, 58.5 s to 184:40E1E2traininglowest last 0.0002Train loss0.0129▼ −2.99024.000.003.00310.0129last ▼ −2.9902best 0.9640, run 6Last-value range, 11 runs, 0.8854 to 0.9640Eval accuracy0.9640▲ +0.02601.000.850.93800.9640last ▲ +0.0260Learning rate0.00e+0▼ −1.15e-66.0e-50.0e+01.15e-6last 0.00e+0Power, whole chip, CPU and GPU119 W▲ +4 W125 W110 W114 Wlast 119 WHottest, one board zone95 °C▲ +1 °C96 °C90 °C94 °Clast 95 °C00:0030:0060:0090:00120:00150:00180:00
GPUpower: whole chip, CPU and GPUtemperature: one board zone

GPU 0

Wwhole chip, CPU and GPU

119 W whole chip, CPU and GPU

use89%
memory, whole GPU2.1 GB

peak 42.9 GB, whole GPU

95 °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 0.9100, 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

6 of 11 finished, 0.9640, config 5 of 9 (batch 16), seed 0, Sat 21 Nov, 05:11 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, full pool, synthetic58183 train500 eval
Config
batch size 16, epochs 2, 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
13.0 swent public
58.5 sphase training
5559.4 seval 1
5565.6 sepoch 1
11062.2 sepoch 2
11080.4 seval 2

Readings

729 rows as a table

rows 1 to 100 of 729

StepSecondsTrain lossEval accuracyLearning rate
1058.53.0031-1.15e-6
2073.52.0280-2.29e-6
3088.31.4521-3.44e-6
40103.31.5991-4.59e-6
50118.31.4062-5.73e-6
60133.51.5048-6.88e-6
70148.82.1310-8.03e-6
80164.01.7405-9.17e-6
90179.21.4711-1.03e-5
100194.41.5154-1.15e-5
110209.51.7124-1.26e-5
120224.41.4923-1.38e-5
130239.50.7089-1.49e-5
140254.70.9761-1.61e-5
150269.81.0377-1.72e-5
160284.90.8625-1.83e-5
170299.90.9005-1.95e-5
180315.21.3707-2.06e-5
190330.30.9165-2.18e-5
200345.70.8155-2.29e-5
210360.60.7352-2.41e-5
220375.70.8781-2.52e-5
230390.90.6841-2.64e-5
240405.80.4415-2.75e-5
250420.90.6461-2.87e-5
260435.80.5461-2.98e-5
270450.81.4318-3.10e-5
280466.00.4165-3.21e-5
290481.10.7847-3.33e-5
300496.20.8679-3.44e-5
310511.30.4549-3.56e-5
320526.30.5556-3.67e-5
330541.50.7284-3.78e-5
340556.40.4652-3.90e-5
350571.60.2973-4.01e-5
360586.60.5746-4.13e-5
370601.60.5704-4.24e-5
380616.80.6842-4.36e-5
390631.80.4611-4.47e-5
400646.81.4593-4.59e-5
410661.90.4703-4.70e-5
420676.90.5967-4.82e-5
430691.91.2034-4.93e-5
440707.00.4253-5.00e-5
450722.30.2618-4.99e-5
460737.40.2413-4.98e-5
470752.50.2387-4.98e-5
480767.20.6592-4.97e-5
490782.40.5888-4.96e-5
500797.60.9645-4.95e-5
510813.00.4528-4.95e-5
520828.20.5965-4.94e-5
530843.20.2147-4.93e-5
540858.40.3955-4.92e-5
550873.70.5555-4.92e-5
560888.90.5452-4.91e-5
570904.00.5040-4.90e-5
580919.30.3950-4.89e-5
590934.50.2972-4.89e-5
600949.80.1977-4.88e-5
610964.80.8795-4.87e-5
620979.80.5252-4.87e-5
630994.90.2241-4.86e-5
6401010.20.2392-4.85e-5
6501025.20.4713-4.84e-5
6601040.30.3011-4.84e-5
6701055.40.4879-4.83e-5
6801070.50.1024-4.82e-5
6901085.70.2813-4.81e-5
7001100.70.4635-4.81e-5
7101115.70.6206-4.80e-5
7201131.00.2158-4.79e-5
7301146.30.8613-4.79e-5
7401161.50.5357-4.78e-5
7501176.50.8118-4.77e-5
7601191.70.1752-4.76e-5
7701206.60.2098-4.76e-5
7801221.71.1665-4.75e-5
7901236.90.3343-4.74e-5
8001252.40.3752-4.73e-5
8101267.70.0394-4.73e-5
8201283.00.3968-4.72e-5
8301298.00.4288-4.71e-5
8401313.20.3165-4.70e-5
8501328.20.3906-4.70e-5
8601343.20.5769-4.69e-5
8701358.30.4086-4.68e-5
8801373.30.2096-4.68e-5
8901388.60.4823-4.67e-5
9001404.10.4267-4.66e-5
9101419.00.1757-4.65e-5
9201434.00.3067-4.65e-5
9301449.30.3538-4.64e-5
9401464.30.3301-4.63e-5
9501479.40.7673-4.62e-5
9601494.60.1334-4.62e-5
9701509.70.2267-4.61e-5
9801524.90.0737-4.60e-5
9901540.30.2493-4.59e-5
10001555.71.0383-4.59e-5

GPU readings

729 rows as a table

rows 1 to 100 of 342

StepGPUUse %Power W, whole chip, CPU and GPUMemory GB, whole GPUTemp C, one board zone
100100114.320.693.8
2000120.320.690.8
500100116.623.992.8
7000124.231.593.8
9000118.231.593.8
1100100116.631.594.8
13000120.331.594.8
1500100112.531.594.8
1700100119.431.593.8
19000120.231.593.8
2200100115.731.593.8
24000120.231.594.8
2600100117.531.594.8
28000120.131.594.8
3000100114.531.594.8
3200100122.831.594.8
3500100114.331.594.8
37000123.631.594.8
3900100113.331.594.8
4100100122.631.594.8
4300100114.531.594.8
4500100119.031.593.8
4700100115.431.593.8
5000100118.331.593.8
52000122.731.594.8
5400100112.631.594.8
5600100115.731.594.8
5800100119.731.594.8
60000119.831.593.8
6200100116.431.594.8
64000123.231.594.8
6700100114.131.594.8
6900100120.631.594.8
7100100113.131.594.8
7300100113.131.594.8
7500100117.931.593.8
77000113.131.594.8
7900100113.031.593.8
8200100115.631.594.8
84000124.031.594.8
8600100115.931.593.8
88000120.831.594.8
90000120.335.294.8
9200100116.135.293.8
94000121.935.294.8
9700100114.035.294.8
9900100115.535.293.8
10100100119.138.994.8
103000120.538.994.8
105000119.238.993.8
10700100119.038.993.8
109000114.638.993.8
11100100116.238.993.8
114000120.238.994.8
11600100114.138.993.8
11800100117.038.993.8
120000117.038.993.8
12200100116.638.993.8
124000120.238.994.8
12600100112.538.994.8
12900100117.038.994.8
131000121.738.994.8
13300100115.438.994.8
135000122.338.993.8
13700100115.638.994.8
139000123.538.994.8
141000120.438.993.8
14400100113.838.993.8
14600100122.738.993.8
148000113.438.994.8
15000100121.338.993.8
152000119.538.993.8
15400100113.838.993.8
156000123.638.994.8
158000118.438.993.8
16100100114.738.993.8
16300100119.538.993.8
16500100112.238.994.8
16700100117.838.993.8
169000119.538.993.8
17100100112.738.993.8
17300100120.238.993.8
17600100114.238.993.8
17800100120.038.993.8
18000100112.838.993.8
18200100118.338.993.8
184000119.138.994.8
18600100115.438.994.8
188000122.238.993.8
19100100114.238.993.8
19300100123.238.993.8
195000114.938.993.8
19700100113.138.994.8
19900100118.938.993.8
201000119.538.993.8
20300100113.638.993.8
20600100119.138.993.8
20800100112.638.993.8
21000100116.738.993.8
212000123.938.994.8

Worked out

ReadoutValueWorked out from
Run complete100%7274 steps over 7274
Step time1.51 s10993.9 s over 7257 steps, between readings. Startup and intervals that hold an eval are left out
Steps per minute39.660 s over the step time of 1.51 s
Epoch interval5496.6 sThe one gap between epoch marks
Train loss since first reading−2.99020.0129 at step 7274 minus 3.0031 at step 10
Lowest train loss0.0001Lowest of 728 readings, at step 6820
Eval accuracy, as shown96.4%0.9640 times 100, to one decimal. Changes on that plate are percentage points
Eval accuracy since first eval+0.02600.9640 minus 0.9380
Eval change+0.0260Eval 2 at 0.9640 minus eval 1 at 0.9380
Wrong answers18500 eval items minus 482 correct. 482 over 500 is the only count that gives 0.9640
Rank1 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+0.03400.9640 minus 0.9300 on run 5
Lowest last train loss0.0002Lowest last train loss across this project's 11 runs, run 10
Items per second10.61 step over 1.51 s, times batch size 16
Learning rate since first reading−1.15e-60.00e+0 minus 1.15e-6
Peak memory per GPU, whole GPU42.9 GBHighest sensor reading of any GPU, GPU 0, counting every process on it