← Back to Physical AI
·
run-a8df21e2c
·
OpenPI π0.5
·
powered by Positronic Robotics
π
warehouse-pick-v3
Active
Overview
Metrics
Logs
Checkpoints
Config
Training progress
14,720 / 30,000 steps
Elapsed · ETA
14m 32s · ETA 15m 12s
Loss (smoothed)
0.0142
▼ from 0.892 at step 0
Throughput
12.4 it/s
stable across last 200 steps
GPU utilization
94%
avg · 1× H100 NVLink
VRAM
62.1 / 80 GB
78% headroom used
Live logs
tailing · positronic.vendors.openpi.train
14:32:08 INFO step 14,718 · loss 0.0148 · grad_norm 0.412 · lr 3.0e-5
14:32:09 INFO step 14,719 · loss 0.0145 · grad_norm 0.398 · lr 3.0e-5
14:32:10 INFO step 14,720 · loss 0.0142 · grad_norm 0.405 · lr 3.0e-5
14:32:10 CKPT saved s3://positronic-checkpoints/warehouse-pick-v3/step-14500.safetensors
14:32:11 INFO eval/success_rate 0.71 on holdout (32 episodes)
14:32:11 INFO data_loader queue 8/16 · gpu_util 94% · vram 62.1G
14:32:12 INFO step 14,721 · loss 0.0140 · grad_norm 0.391 · lr 3.0e-5
14:32:13 INFO step 14,722 · loss 0.0143 · grad_norm 0.402 · lr 3.0e-5
14:32:14 INFO step 14,723 · loss 0.0139 · grad_norm 0.388 · lr 3.0e-5
14:32:09 INFO step 14,719 · loss 0.0145 · grad_norm 0.398 · lr 3.0e-5
14:32:10 INFO step 14,720 · loss 0.0142 · grad_norm 0.405 · lr 3.0e-5
14:32:10 CKPT saved s3://positronic-checkpoints/warehouse-pick-v3/step-14500.safetensors
14:32:11 INFO eval/success_rate 0.71 on holdout (32 episodes)
14:32:11 INFO data_loader queue 8/16 · gpu_util 94% · vram 62.1G
14:32:12 INFO step 14,721 · loss 0.0140 · grad_norm 0.391 · lr 3.0e-5
14:32:13 INFO step 14,722 · loss 0.0143 · grad_norm 0.402 · lr 3.0e-5
14:32:14 INFO step 14,723 · loss 0.0139 · grad_norm 0.388 · lr 3.0e-5
Cost so far
$1.21
$5.04 / GPU-hour · 14m 32s elapsed
Est. final
~$4.20
Project budget
$481 / $2,000
Best checkpoint
step-14500
eval/success_rate 0.71 · loss 0.0148
Sample evaluation episode: phail.ai/episode/8872
Deploy checkpoint →
