π0 Labeller

Converting videos...

π0 Labeller

Select a video 0 frames Expert Intervention

γ (Gamma):

λ (Lambda):

Penalty:

Adv Quantile (%):

Cutoff:

⚡ RECAP State

Generalist Not Ready

Finetuned Policy Not Ready

Trains small π0 on binarized advantages

Value Networks

Rollouts

Joint State

Dim 0/0

Action Editor

T=0

Tasks

Train Value Network

Select configuration and train on calculated returns.

Mock (Ultra Fast) 16 dim, 1 depth, 2 heads, 32x32px

Small 64 dim, 2 depth, 4 heads (Extremely Small)

Medium 128 dim, 4 depth, 8 heads

Large 256 dim, 6 depth, 8 heads

Epoch -

Loss -

✓

Training Complete!

Model: -

Epochs: -

Final Loss: -

Finetune Policy Network

Finetune policy using binarized advantages.

Mock (Ultra Fast) 8 dim, 1 depth, 1 heads, 32px

Small 16 dim, 1 depth, 2 heads, 32px

Epoch -

Loss -

✓

Finetuning Complete!

Model: -

Epochs: -

Final Loss: -