Skip to main content
shion_honda
Projects
reviewer-2-bot-dpo-tiny-llama
Log in
Sign up
Overview
Workspace
Runs
Automat.
Sweeps
Reports
Artifacts
Shion_honda's workspace
Personal workspace
Automated workspace
Changes are only visible to you.
Runs
9
Name
1 visualized
5e-4 long
5e-4 long
7e-4 + long
7e-4 + long
7e-4 + warmup
7e-4 + warmup
1e-3 + warmup
1e-3 + warmup
1e-3 + warmup
1e-3 + warmup
5e-4 + warmup
5e-4 + warmup
2e-4
2e-4
1e-3
1e-3
5e-4
5e-4
1-9
of 9
Settings
Add panels
eval/logits
2
eval/logps
2
eval
2
eval/steps_per_second
eval/steps_per_second
10
20
30
40
50
Step
0.486
0.4865
0.487
0.4875
0.488
0.4885
0.489
5e-4
eval/samples_per_second
eval/samples_per_second
10
20
30
40
50
Step
3.74
3.745
3.75
3.755
3.76
5e-4
eval/rewards
4
1-4 of 4
eval/rewards/accuracies
eval/rewards/accuracies
10
20
30
40
50
Step
0
0.005
0.01
0.015
5e-4
eval/rewards/margins
eval/rewards/margins
10
20
30
40
50
Step
0.017
0.0175
0.018
0.0185
5e-4
eval/rewards/rejected
eval/rewards/rejected
10
20
30
40
50
Step
-0.14
-0.12
-0.1
5e-4
train
4
1-4 of 4
Add section