CITADEL
StandbyNothing is live right now.
Eval accuracylast measured Tue 10 Nov, 14:30 UTC
%
0.9233
one eval
0%25%50%75%100%0%25%50%75%100%0%25%50%75%100%

Claviger router: OneJev 0.8B (backfilled)

finishedRadeon 8060S

A model that routes a request to the one destination, of many, that should take it. Backfilled: recorded after the fact from the run's own logs and records. Times inside a run are estimated, and there are no GPU sensor readings.

started
Tue 10 Nov, 13:33 UTC
ran for
45:47
config
learning rate 0.00001, batch 1, seed 0 of 3

run1/5step42/42epoch1/1config1/3seed0/3lr0.00001batch1modelOneJev1.6-0.8B 0.8Brevc89933c

Loss train 1.0576 ▲ +0.9700 eval 0.6839 lowest 0.0001 at step 7
Timebase
Range
Smoothing
42 readings, 30 min/div10 min/div5 min/div
012300:0030:00phase starting, 0.0 s to 65.4 sphase training, 65.4 s to 45:47epoch 1, 65.4 straining+0.9700since first0.0876Lowest last train loss across this project's 5 runs, run 4lowest last 0.0014, run 41.0576012300:0010:0020:0030:0040:00phase starting, 0.0 s to 65.4 sphase training, 65.4 s to 45:47epoch 1, 65.4 straining0.0876Lowest last train loss across this project's 5 runs, run 4lowest last 0.0014, run 4eval 0.68391.0576▲ +0.9700012300:0005:0010:0015:0020:0025:0030:0035:0040:0045:00phase starting, 0.0 s to 65.4 sphase training, 65.4 s to 45:47epoch 1, 65.4 straining0.0876Lowest last train loss across this project's 5 runs, run 4lowest last 0.0014, run 4eval 0.68391.0576▲ +0.9700

Eval measures

Eval losslower is better
0.6839
02 →
ECEcalibration error, lower is better
0.0707
01

Elapsed

65.4 s to epoch 1, 2682.5 s to eval 1

42 of 42 steps, 1 of 1 epochs, 100%

65.4 s per step, 0.9 steps a minute

Against the other runs

1 of 3

by last eval accuracy

best so far 0.9233, run 1

Train minus eval

+0.3737

step 42

train 1.0576, eval 0.6839

Set-up

666 train

synthetic, learning rate 0.00001, batch size 1, grad accum 16, epochs 1, optimizer adamw, seed 0

Learning rate

0.00e+0

▼ −1.00e-5 since first

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 of 5 finished, 0.9233, config 1 of 3 (learning rate 0.00001, batch 1), seed 0 of 3, Tue 10 Nov, 13:33 UTC

◂
Runs as a table
StartedStatusEval accuracyStepsConfig
Tue 10 Nov, 17:10 UTC| finished-42 of 42learning rate 0.00001, batch 1, seed 2 of 3
Tue 10 Nov, 16:24 UTC| finished-42 of 42learning rate 0.00001, batch 1, seed 1 of 3
Tue 10 Nov, 15:27 UTC| finished0.084742 of 42learning rate 0.00005, batch 1, seed 0
Tue 10 Nov, 14:30 UTC| finished0.912742 of 42learning rate 0.00002, batch 1, seed 0
Tue 10 Nov, 13:33 UTC| finished0.923342 of 42learning rate 0.00001, batch 1, seed 0 of 3

Plate

Model
OmniJev/OneJev1.6-0.8B, 0.8B parameters revision c89933c
Data
Synthetic routing worlds, synthetic
Config
learning rate 0.00001, batch size 1, grad accum 16, epochs 1, optimizer adamw, seed 0
Hardware
Radeon 8060S, ROCm

Log

Since startMark
0.0 sphase starting
65.4 sphase training
65.4 sepoch 1
2747.9 seval 1
191 h 18 min laterwent public

Seeds

SeedStatusEval accuracy
seed 0| finished0.9233
seed 1| finished-
seed 2| finished-

Readings

42 rows as a table
StepSecondsTrain lossEval lossEval accuracyECELearning rate
165.40.0876---1.00e-5
2130.90.0181---9.99e-6
3196.30.0025---9.94e-6
4261.70.0499---9.87e-6
5327.10.0001---9.77e-6
6392.60.2441---9.64e-6
7458.00.0001---9.48e-6
8523.41.9232---9.30e-6
9588.80.4464---9.09e-6
10654.30.0129---8.86e-6
11719.70.0076---8.60e-6
12785.11.1170---8.33e-6
13850.50.1006---8.03e-6
14916.01.6922---7.72e-6
15981.41.5120---7.39e-6
161046.80.9942---7.05e-6
171112.20.2331---6.69e-6
181177.71.2901---6.32e-6
191243.12.2822---5.95e-6
201308.50.6702---5.57e-6
211374.00.0064---5.19e-6
221439.41.3281---4.81e-6
231504.80.5506---4.43e-6
241570.20.0006---4.05e-6
251635.70.6179---3.68e-6
261701.10.2829---3.31e-6
271766.50.0149---2.95e-6
281831.90.8996---2.61e-6
291897.40.1055---2.28e-6
301962.80.0092---1.97e-6
312028.20.0024---1.67e-6
322093.60.0004---1.40e-6
332159.10.8714---1.14e-6
342224.50.4984---9.10e-7
352289.90.1254---7.02e-7
362355.30.4071---5.19e-7
372420.81.9418---3.62e-7
382486.20.0716---2.33e-7
392551.60.0174---1.32e-7
402617.00.9824---5.86e-8
412682.50.0798---1.47e-8
422747.91.05760.68390.92330.07070.00e+0

Worked out

ReadoutValueWorked out from
Run complete100%42 steps over 42
Step time65.4 s2617.1 s over 40 steps, between readings. Startup and intervals that hold an eval are left out
Steps per minute0.960 s over the step time of 65.43 s
Train loss since first reading+0.97001.0576 at step 42 minus 0.0876 at step 1
Lowest train loss0.0001Lowest of 42 readings, at step 7
Eval accuracy, as shown92.3%0.9233 times 100, to one decimal. Changes on that plate are percentage points
Train minus eval loss+0.37371.0576 minus 0.6839, both at step 42
Rank1 of 3By last eval accuracy, across this project's 3 runs with an eval
Best so far0.9233Highest last eval accuracy, run 1
Lowest last train loss0.0014Lowest last train loss across this project's 5 runs, run 4
Items per second0.01 step over 65.43 s, times batch size 1
Learning rate since first reading−1.00e-50.00e+0 minus 1.00e-5