CITADEL
StandbyNothing is live right now.
Eval accuracylast measured Sun 22 Nov, 02:07 UTC
%
0.9385122 of 130 eval, 8 wrong
▲ +43.08 pts across 2 evals▲ +43.08 pts on the eval before
0%25%50%75%100%0%25%50%75%100%0%25%50%75%100%

Claviger: OneJev 0.8B

finishedRadeon 8060S

A small model that reads one Discord message and decides whether an agent should reply, and which one. OneJev 0.8B fine-tuned on invented chat, beside the Laya model.

started
Sun 22 Nov, 01:31 UTC
ran for
35:44
config
learning rate 0.000005, batch 8, seed 2 of 3

run7/7step339/339epoch3/3config1/4seed2/3lr0.000005batch8modelOneJev1.6-0.8B 0.8Brevc89933citems1800+130

Loss train 0.0001 ▼ −0.7239 eval 0.4932 lowest 0.0000 at step 310 run before, run 6 ▼ −0.9521 step 339
Timebase
Range
Smoothing
Reference
68 readings, 10 min/div10 min/div5 min/div
00.511.500:0010:0020:0030:00phase starting, 0.0 s to 87.4 sphase training, 87.4 s to 35:44epoch 1, 87.4 sepoch 2, 13:51epoch 3, 24:54trainingepoch 2epoch 3−0.7239since first0.7239Lowest last train loss across this project's 6 runs, run 2lowest last 0.0000, run 20.000100.511.500:0010:0020:0030:00phase starting, 0.0 s to 87.4 sphase training, 87.4 s to 35:44epoch 1, 87.4 sepoch 2, 13:51epoch 3, 24:54trainingepoch 2epoch 30.7239Lowest last train loss across this project's 6 runs, run 2lowest last 0.0000, run 2eval 0.49320.0001▼ −0.723900.511.500:0005:0010:0015:0020:0025:0030:0035:00phase starting, 0.0 s to 87.4 sphase training, 87.4 s to 35:44epoch 1, 87.4 sepoch 2, 13:51epoch 3, 24:54trainingepoch 2epoch 30.7239Lowest last train loss across this project's 6 runs, run 2lowest last 0.0000, run 2eval 0.49320.0001▼ −0.7239

Eval measures

Eval losslower is better
0.4932▼ −0.5694 since first
02 →
ECEcalibration error, lower is better
0.0611▼ −0.1425 since first
01

Elapsed

703.3 s

epoch interval

339 of 339 steps, 3 of 3 epochs, 100%

6.14 s per step, 9.8 steps a minute

Eval accuracy by eval

+0.4308

▲ on the eval before

eval 1 0.5077, eval 2 0.9385

Against the other runs

1 of 7

by last eval accuracy

best so far 0.9385, run 7

▲ +0.3770 against run 6

Train minus eval

−0.4932

step 339

train 0.0001, eval 0.4932

Set-up

1800 train, 130 eval

synthetic, learning rate 0.000005, batch size 8, grad accum 2, epochs 3, optimizer adamw, scheduler cosine, seed 2

Seed spread, observed

0.0385
0.900.95

3 finished seeds

Learning rate

0.00e+0

▼ −2.50e-6 since first

On one axis

phase starting, 0.0 s to 87.4 sphase training, 87.4 s to 35:44E1E2E3traininglowest last 0.0000Train loss0.0001▼ −0.72391.500.000.72390.0001last ▼ −0.7239best 0.9385, run 7Last-value range, 7 runs, 0.3000 to 0.9385Eval accuracy0.9385▲ +0.43081.000.250.50770.9385last ▲ +0.4308Learning rate0.00e+0▼ −2.50e-66.0e-60.0e+02.50e-6last 0.00e+0Power, whole chip, CPU and GPU118 W▼ −2 W140 W80 W121 Wlast 118 WHottest, one board zone90 °C▼ −8 °C100 °C85 °C98 °Clast 90 °C00:0005:0010:0015:0020:0025:0030:0035:00
GPUpower: whole chip, CPU and GPUtemperature: one board zone

GPU 0

Wwhole chip, CPU and GPU

118 W whole chip, CPU and GPU

use100%
memory, whole GPU34.8 GB

peak 42.3 GB, whole GPU

90 °C one board zone

history 0 to 122 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 ▸
0.950.300.950.30finished, eval accuracy 0.9000, config 1 of 4 (learning rate 0.000005, batch 8), seed 0 of 3, Thu 19 Nov, 22:36 UTCfinished, eval accuracy 0.9154, config 1 of 4 (learning rate 0.000005, batch 8), seed 1 of 3, Thu 19 Nov, 23:24 UTCfailed, eval accuracy 0.5077, config 1 of 4 (learning rate 0.000005, batch 8), seed 2 of 3, Fri 20 Nov, 00:08 UTCfinished, eval accuracy 0.3000, config 2 of 4 (learning rate 0.0001, batch 8), seed 0, Sat 21 Nov, 23:42 UTCfinished, eval accuracy 0.8769, config 3 of 4 (learning rate 0.00002, batch 8), seed 0, Sun 22 Nov, 00:18 UTCfinished, eval accuracy 0.5615, config 4 of 4 (learning rate 0.00005, batch 8), seed 0, Sun 22 Nov, 00:55 UTCfinished, eval accuracy 0.9385, config 1 of 4 (learning rate 0.000005, batch 8), seed 2 of 3, Sun 22 Nov, 01:31 UTC0.950.30finished, eval accuracy 0.9000, config 1 of 4 (learning rate 0.000005, batch 8), seed 0 of 3, Thu 19 Nov, 22:36 UTCfinished, eval accuracy 0.9154, config 1 of 4 (learning rate 0.000005, batch 8), seed 1 of 3, Thu 19 Nov, 23:24 UTCfailed, eval accuracy 0.5077, config 1 of 4 (learning rate 0.000005, batch 8), seed 2 of 3, Fri 20 Nov, 00:08 UTCfinished, eval accuracy 0.3000, config 2 of 4 (learning rate 0.0001, batch 8), seed 0, Sat 21 Nov, 23:42 UTCfinished, eval accuracy 0.8769, config 3 of 4 (learning rate 0.00002, batch 8), seed 0, Sun 22 Nov, 00:18 UTCfinished, eval accuracy 0.5615, config 4 of 4 (learning rate 0.00005, batch 8), seed 0, Sun 22 Nov, 00:55 UTCfinished, eval accuracy 0.9385, config 1 of 4 (learning rate 0.000005, batch 8), seed 2 of 3, Sun 22 Nov, 01:31 UTC

7 of 7 finished, 0.9385, config 1 of 4 (learning rate 0.000005, batch 8), seed 2 of 3, Sun 22 Nov, 01:31 UTC

▸
Runs as a table
StartedStatusEval accuracyStepsConfig
Sun 22 Nov, 01:31 UTC| finished0.9385339 of 339learning rate 0.000005, batch 8, seed 2 of 3
Sun 22 Nov, 00:55 UTC| finished0.5615339 of 339learning rate 0.00005, batch 8, seed 0
Sun 22 Nov, 00:18 UTC| finished0.8769339 of 339learning rate 0.00002, batch 8, seed 0
Sat 21 Nov, 23:42 UTC| finished0.3000339 of 339learning rate 0.0001, batch 8, seed 0
Fri 20 Nov, 00:08 UTC× failed0.50770learning rate 0.000005, batch 8, seed 2 of 3
Thu 19 Nov, 23:24 UTC| finished0.9154339 of 339learning rate 0.000005, batch 8, seed 1 of 3
Thu 19 Nov, 22:36 UTC| finished0.9000339 of 339learning rate 0.000005, batch 8, seed 0 of 3

Plate

Model
OmniJev/OneJev1.6-0.8B, 0.8B parameters revision c89933c
Data
Synthetic Discord triage set, 2,000-row draw, synthetic1800 train130 eval
Config
learning rate 0.000005, batch size 8, grad accum 2, epochs 3, optimizer adamw, scheduler cosine, seed 2
Hardware
Radeon 8060S, ROCm

Log

Since startMark
0.0 sphase starting
4.5 swent public
39.8 seval 1
87.4 sphase training
87.4 sepoch 1
831.3 sepoch 2
1494.1 sepoch 3
2145.0 seval 2

Seeds

SeedStatusEval accuracy
seed 0| finished0.9000
seed 1| finished0.9154
seed 2× failed0.5077
seed 2| finished0.9385

observed spread 0.0385, 3 finished seeds

Readings

69 rows as a table
StepSecondsTrain lossEval lossEval accuracyECELearning rate
039.8-1.06260.50770.2036-
587.40.7239---2.50e-6
10128.20.6374---5.00e-6
15170.00.4155---5.00e-6
20208.90.8375---4.99e-6
25246.20.2964---4.97e-6
30285.40.2117---4.95e-6
35321.80.2893---4.93e-6
40353.90.1518---4.90e-6
45385.80.5385---4.86e-6
50424.70.3496---4.82e-6
55458.50.2775---4.77e-6
60489.70.3376---4.72e-6
65525.70.3290---4.66e-6
70559.80.1074---4.60e-6
75589.20.3840---4.53e-6
80619.90.2155---4.46e-6
85652.60.5599---4.39e-6
90683.70.1048---4.31e-6
95711.50.3470---4.22e-6
100741.60.8191---4.13e-6
105772.30.2620---4.04e-6
110802.40.7802---3.94e-6
115831.30.1266---3.85e-6
120859.50.4028---3.74e-6
125889.30.5512---3.64e-6
130918.90.0023---3.53e-6
135947.10.0554---3.42e-6
140975.30.0401---3.31e-6
1451006.00.0149---3.19e-6
1501034.30.4875---3.08e-6
1551064.30.6180---2.96e-6
1601092.40.0201---2.85e-6
1651121.30.3865---2.73e-6
1701149.70.3920---2.61e-6
1751180.00.0019---2.49e-6
1801208.10.0003---2.37e-6
1851238.50.8803---2.25e-6
1901265.00.0010---2.13e-6
1951294.10.0007---2.01e-6
2001325.20.3768---1.90e-6
2051353.90.0320---1.78e-6
2101383.60.3848---1.67e-6
2151412.50.0009---1.56e-6
2201440.90.0002---1.45e-6
2251468.00.0040---1.34e-6
2301494.10.0001---1.24e-6
2351522.90.0314---1.13e-6
2401551.10.0002---1.04e-6
2451580.10.0003---9.41e-7
2501608.00.0002---8.50e-7
2551636.30.0005---7.62e-7
2601666.50.3729---6.78e-7
2651695.70.0001---5.99e-7
2701725.50.0001---5.23e-7
2751755.10.0003---4.53e-7
2801782.90.2737---3.86e-7
2851811.30.0001---3.25e-7
2901839.70.0003---2.69e-7
2951870.50.0013---2.17e-7
3001898.80.0000---1.71e-7
3051927.40.0000---1.31e-7
3101955.00.0000---9.52e-8
3151983.90.0001---6.54e-8
3202011.90.0001---4.10e-8
3252040.40.0001---2.23e-8
3302069.50.0004---9.23e-9
3352098.20.0000---1.82e-9
3392145.00.00010.49320.93850.06110.00e+0

GPU readings

69 rows as a table
StepGPUUse %Power W, whole chip, CPU and GPUMemory GB, whole GPUTemp C, one board zone
000120.530.397.8
505895.028.797.8
10074105.029.097.8
15068103.826.297.8
200100115.831.997.8
25078103.429.297.8
30066117.533.097.8
35090116.434.097.8
4008497.219.895.8
45065117.432.197.8
500100105.018.897.8
55071106.029.594.8
60068118.834.396.8
650100113.730.197.8
7000121.634.996.8
7506397.026.493.8
800100117.534.793.8
85082115.230.097.8
900100115.533.594.8
9500121.130.295.8
1000100117.434.694.8
10506599.333.595.8
110094116.932.396.8
1150100117.634.694.8
1200100113.831.095.8
12500118.934.394.8
1350100112.137.594.8
14005088.823.994.8
14500117.537.394.8
1500100117.637.393.8
1550100111.635.494.8
16000118.137.695.8
1650100111.137.693.8
1700100121.530.395.8
1750100111.335.393.8
1800100116.941.294.8
190065118.642.393.8
195064117.242.394.8
2000100114.232.596.8
2050100117.034.094.8
21000118.436.494.8
21500117.936.493.8
220083114.738.092.8
225058120.238.094.8
23000118.838.091.8
2400100116.738.094.8
245094116.438.093.8
25000118.338.094.8
2550100117.738.094.8
260068117.540.295.8
2650100112.740.295.8
27000121.833.295.8
2750100111.834.293.8
28000118.434.294.8
29000120.336.593.8
2950100116.836.794.8
300066118.933.893.8
30500117.334.793.8
3100100115.436.294.8
315066119.636.293.8
32000117.236.294.8
325065118.433.994.8
33000117.434.892.8
3390100118.134.889.8

Worked out

ReadoutValueWorked out from
Run complete100%339 steps over 339
Step time6.14 s2058.3 s over 335 steps, between readings. Startup and intervals that hold an eval are left out
Steps per minute9.860 s over the step time of 6.14 s
Epoch interval703.3 sMean of 2 gaps between epoch marks
Train loss since first reading−0.72390.0001 at step 339 minus 0.7239 at step 5
Lowest train loss0.0000Lowest of 68 readings, at step 310
Eval accuracy, as shown93.9%0.9385 times 100, to one decimal. Changes on that plate are percentage points
Eval accuracy since first eval+0.43080.9385 minus 0.5077
Eval loss since first eval−0.56940.4932 minus 1.0626
ECE since first eval−0.14250.0611 minus 0.2036
Eval change+0.4308Eval 2 at 0.9385 minus eval 1 at 0.5077
Train minus eval loss−0.49320.0001 minus 0.4932, both at step 339
Wrong answers8130 eval items minus 122 correct. 122 over 130 is the only count that gives 0.9385
Rank1 of 7By last eval accuracy, across this project's 7 runs with an eval
Best so far0.9385Highest last eval accuracy, run 7
Against the run before+0.37700.9385 minus 0.5615 on run 6
Lowest last train loss0.0000Lowest last train loss across this project's 6 runs, run 2
Seed spread, observed0.0385High 0.9385 minus low 0.9000 across 3 finished seeds
Items per second1.31 step over 6.14 s, times batch size 8
Train loss against run 6−0.95210.0001 minus 0.9522 on run 6, both at step 339
Learning rate since first reading−2.50e-60.00e+0 minus 2.50e-6
Peak memory per GPU, whole GPU42.3 GBHighest sensor reading of any GPU, GPU 0, counting every process on it