summaryrefslogtreecommitdiff
path: root/play.py
diff options
context:
space:
mode:
authorhaoyuren <13851610112@163.com>2026-02-22 12:06:23 -0600
committerhaoyuren <13851610112@163.com>2026-02-22 12:06:23 -0600
commit800e1f1f33d93cb7a1812dff1dc0ef85289ef075 (patch)
tree7a72228f3cb639046269a89931f8c8cebf33ee84 /play.py
parentdda6db0777620f8139bd476e27e6b275c0679358 (diff)
Auto-calibrate collect_batch when not specified
Benchmarks batch sizes [64,128,256,512] and picks smallest within 10% of peak throughput. Smaller batches = more frequent PPO updates = better training quality at similar speed. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Diffstat (limited to 'play.py')
0 files changed, 0 insertions, 0 deletions