search

mlcommons / training

Reference implementations of MLPerf™ training benchmarks
https://mlcommons.org/en/groups/training
Apache License 2.0
1.62k stars 561 forks source link

[MaskRCNN bug] when MaskRCNN saves checkpoint after training, an error is reported #671

Closed Xiao-Yamin closed 3 months ago

Xiao-Yamin commented 1 year ago

Question: When MaskRCNN saves checkpoint after training, this variable arguments["save_checkpoints"] is used, but no value is assigned to it. image

Solution: In training\object_detection\pytorch\tools\train_net.py，add the following code: arguments["save_checkpoints"] = cfg.SAVE_CHECKPOINTS

ShriyaPalsamudram commented 3 months ago

Sorry but the maskrcnn benchmark is dropped from the training benchmarks suite so this issue cannot be addressed at this time.