MogaNet
MogaNet copied to clipboard
Segmentation Config Issues: Log Mismatch and Parameter Name Inconsistency
Hi, thank you for your great work. I noticed two issues while reproducing the results:
The provided config file for semantic segmentation does not fully match the training log. For example, the log indicates frozen_stages=1 and norm_eval=False, but these options are missing in the config.
There is a mismatch between the config parameter names and those expected by the model definition. Specifically, the model requires parameters named stem_norm_type and conv_norm_type, but the config uses stem_norm_cfg and conv_norm_cfg instead.
Could you please clarify these inconsistencies? Thank you!