CITADEL
StandbyNothing is live right now.
Eval accuracylast measured Fri 20 Nov, 00:08 UTC
%
0.9154119 of 130 eval, 11 wrong
▲ +40.77 pts across 2 evals▲ +40.77 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
Thu 19 Nov, 23:24 UTC
ran for
43:49
config
learning rate 0.000005, batch 8, seed 1 of 3

run2/7step339/339epoch3/3config1/4seed1/3lr0.000005batch8modelOneJev1.6-0.8B 0.8Brevc89933citems1800+130

Loss train 0.0000 ▼ −0.8291 eval 0.7348 lowest 0.0000 at step 300 run before, run 1 ▼ −0.0000 step 339
Timebase
Range
Smoothing
Reference
68 readings, 30 min/div10 min/div5 min/div
00.511.500:0030:00phase starting, 0.0 s to 02:25phase training, 02:25 to 43:49epoch 1, 02:25epoch 2, 18:16epoch 3, 31:17trainingepoch 2epoch 3−0.8291since first0.8291Lowest last train loss across this project's 6 runs, run 2lowest last 0.0000, run 20.000000.511.500:0010:0020:0030:0040:00phase starting, 0.0 s to 02:25phase training, 02:25 to 43:49epoch 1, 02:25epoch 2, 18:16epoch 3, 31:17trainingepoch 2epoch 30.8291Lowest last train loss across this project's 6 runs, run 2lowest last 0.0000, run 2eval 0.73480.0000▼ −0.829100.511.500:0005:0010:0015:0020:0025:0030:0035:0040:00phase starting, 0.0 s to 02:25phase training, 02:25 to 43:49epoch 1, 02:25epoch 2, 18:16epoch 3, 31:17trainingepoch 2epoch 30.8291Lowest last train loss across this project's 6 runs, run 2lowest last 0.0000, run 2eval 0.73480.0000▼ −0.8291

Eval measures

Eval losslower is better
0.7348▼ −0.3278 since first
02 →
ECEcalibration error, lower is better
0.0929▼ −0.1107 since first
01

Elapsed

865.9 s

epoch interval

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

7.46 s per step, 8.0 steps a minute

Eval accuracy by eval

+0.4077

▲ on the eval before

eval 1 0.5077, eval 2 0.9154

Against the other runs

2 of 7

by last eval accuracy

best so far 0.9385, run 7

▲ +0.0154 against run 1

Train minus eval

−0.7348

step 339

train 0.0000, eval 0.7348

Set-up

1800 train, 130 eval

synthetic, learning rate 0.000005, batch size 8, grad accum 2, epochs 3, optimizer adamw, scheduler cosine, seed 1

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 02:25phase training, 02:25 to 43:49E1E2E3traininglowest last 0.0000Train loss0.0000▼ −0.82911.500.000.82910.0000last ▼ −0.8291best 0.9385, run 7Last-value range, 7 runs, 0.3000 to 0.9385Eval accuracy0.9154▲ +0.40771.000.250.50770.9154last ▲ +0.4077Learning rate0.00e+0▼ −2.50e-66.0e-60.0e+02.50e-6last 0.00e+0Power, whole chip, CPU and GPU54 W▲ +2 W55 W52 W53 Wlast 54 WHottest, one board zone69 °C▲ +1 °C69 °C66 °C68 °Clast 69 °C00:0005:0010:0015:0020:0025:0030:0035:0040:00
GPUpower: whole chip, CPU and GPUtemperature: one board zone

GPU 0

Wwhole chip, CPU and GPU

54 W whole chip, CPU and GPU

use0%
memory, whole GPU37.8 GB

peak 42.8 GB, whole GPU

69 °C one board zone

history 0 to 55 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
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

2 of 7 finished, 0.9154, config 1 of 4 (learning rate 0.000005, batch 8), seed 1 of 3, Thu 19 Nov, 23:24 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 1
Hardware
Radeon 8060S, ROCm

Log

Since startMark
0.0 sphase starting
12.4 swent public
74.9 seval 1
145.6 sphase training
145.6 sepoch 1
1096.3 sepoch 2
1877.4 sepoch 3
2629.5 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
074.9-1.06260.50770.2036-
5145.60.8291---2.50e-6
10203.40.7385---5.00e-6
15254.70.8285---5.00e-6
20313.90.8188---4.99e-6
25371.70.3701---4.97e-6
30417.50.4389---4.95e-6
35470.00.2824---4.93e-6
40520.00.2368---4.90e-6
45558.60.4325---4.86e-6
50608.40.4037---4.82e-6
55643.80.4431---4.77e-6
60682.80.3990---4.72e-6
65724.30.0403---4.66e-6
70772.00.2863---4.60e-6
75810.20.3174---4.53e-6
80853.90.0976---4.46e-6
85885.00.0382---4.39e-6
90920.70.4913---4.31e-6
95960.80.0341---4.22e-6
1001000.30.2748---4.13e-6
1051033.50.0084---4.04e-6
1101067.70.0571---3.94e-6
1151096.30.0067---3.85e-6
1201131.50.3512---3.74e-6
1251163.80.1441---3.64e-6
1301199.20.3070---3.53e-6
1351231.40.1132---3.42e-6
1401262.90.0276---3.31e-6
1451297.70.0585---3.19e-6
1501335.80.2336---3.08e-6
1551366.40.0040---2.96e-6
1601409.10.0001---2.85e-6
1651443.30.0056---2.73e-6
1701476.00.0009---2.61e-6
1751506.10.8307---2.49e-6
1801541.30.6909---2.37e-6
1851572.60.0324---2.25e-6
1901607.50.5224---2.13e-6
1951640.00.0139---2.01e-6
2001670.90.1267---1.90e-6
2051710.20.0015---1.78e-6
2101745.50.5644---1.67e-6
2151782.50.0005---1.56e-6
2201813.40.0888---1.45e-6
2251845.50.0003---1.34e-6
2301877.40.0005---1.24e-6
2351908.80.0001---1.13e-6
2401941.00.0001---1.04e-6
2451972.90.0001---9.41e-7
2502005.70.0025---8.50e-7
2552037.50.0004---7.62e-7
2602069.40.0285---6.78e-7
2652109.40.0000---5.99e-7
2702140.70.0001---5.23e-7
2752172.20.4593---4.53e-7
2802203.30.4703---3.86e-7
2852234.40.1381---3.25e-7
2902268.80.0046---2.69e-7
2952300.90.0000---2.17e-7
3002332.00.0000---1.71e-7
3052363.30.4417---1.31e-7
3102393.90.0000---9.52e-8
3152435.00.0003---6.54e-8
3202475.10.0009---4.10e-8
3252507.30.0000---2.23e-8
3302542.00.0000---9.23e-9
3352573.80.0002---1.82e-9
3392629.50.00000.73480.91540.09290.00e+0

GPU readings

69 rows as a table
StepGPUUse %Power W, whole chip, CPU and GPUMemory GB, whole GPUTemp C, one board zone
002752.511.967.8
503353.119.567.8
1002753.733.366.8
1504052.832.467.8
2003854.023.267.8
2502253.428.367.8
3002553.818.267.8
350053.332.967.8
4002953.033.866.8
450052.929.767.8
5002553.626.266.8
550653.233.666.8
6002353.020.666.8
6502454.820.367.8
700952.833.366.8
750152.933.067.8
800153.134.066.8
850253.133.567.8
9005952.834.467.8
950553.334.567.8
1000052.935.466.8
1050053.030.767.8
1100653.334.066.8
1150853.034.066.8
1200053.134.666.8
1250052.934.367.8
13001352.937.166.8
1350053.137.166.8
1400252.837.166.8
14502553.022.567.8
1500053.135.766.8
1550053.035.767.8
1600052.935.766.8
1700053.235.766.8
1750053.236.866.8
18001553.141.966.8
18502353.817.566.8
19001953.035.466.8
1950453.236.866.8
2000053.132.867.8
20502253.323.967.8
2100053.129.667.8
2150353.137.366.8
2200053.137.366.8
2250453.242.866.8
2300853.142.866.8
2350053.142.866.8
2400153.142.866.8
2450152.942.866.8
2500053.042.866.8
2600053.142.866.8
2650053.042.866.8
2700052.642.866.8
2750053.042.866.8
2850053.035.266.8
2950953.036.166.8
3000052.936.166.8
3050353.036.166.8
31001052.936.166.8
3150053.536.166.8
3200653.136.166.8
3250253.131.167.8
3300253.137.866.8
3390054.137.868.8

Worked out

ReadoutValueWorked out from
Run complete100%339 steps over 339
Step time7.46 s2499.0 s over 335 steps, between readings. Startup and intervals that hold an eval are left out
Steps per minute8.060 s over the step time of 7.46 s
Epoch interval865.9 sMean of 2 gaps between epoch marks
Train loss since first reading−0.82910.0000 at step 339 minus 0.8291 at step 5
Lowest train loss0.0000Lowest of 68 readings, at step 300
Eval accuracy, as shown91.5%0.9154 times 100, to one decimal. Changes on that plate are percentage points
Eval accuracy since first eval+0.40770.9154 minus 0.5077
Eval loss since first eval−0.32780.7348 minus 1.0626
ECE since first eval−0.11070.0929 minus 0.2036
Eval change+0.4077Eval 2 at 0.9154 minus eval 1 at 0.5077
Train minus eval loss−0.73480.0000 minus 0.7348, both at step 339
Wrong answers11130 eval items minus 119 correct. 119 over 130 is the only count that gives 0.9154
Rank2 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+0.01540.9154 minus 0.9000 on run 1
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 second1.11 step over 7.46 s, times batch size 8
Train loss against run 1−0.00000.0000 minus 0.0000 on run 1, both at step 339
Learning rate since first reading−2.50e-60.00e+0 minus 2.50e-6
Peak memory per GPU, whole GPU42.8 GBHighest sensor reading of any GPU, GPU 0, counting every process on it