RoisulIslamRumi
commented
1 year ago Hello @tiamjiakun, were you able to find a solution to this issue? I was fine-tuning the model that was trained on all the datasets and faced similar warnings. In my case, they are as follows:
WARNING [08/25 20:57:38 fvcore.common.checkpoint]: Some model parameters or buffers are not found in the checkpoint: roi_heads.recognizer_head.backbone.conv1.bias roi_heads.recognizer_head.backbone.conv2.bias
WARNING [08/25 20:57:38 fvcore.common.checkpoint]: The checkpoint state_dict contains keys that are not used by the model: roi_heads.recognizer_head.backbone.conv1.norm.{bias, num_batches_tracked, running_mean, running_var, weight} roi_heads.recognizer_head.backbone.conv2.norm.{bias, num_batches_tracked, running_mean, running_var, weight}
Kindly let me know if you have found any workaround for this warning.
When I was using the first fine-tuning model(IC15) you provided, I received the following prompt： WARNING [04/17 15:55:54 fvcore.common.checkpoint]: Some model parameters or buffers are not found in the checkpoint: roi_heads.box_predictor.orientation_pred.{bias, weight} roi_heads.recognizer_head.backbone.conv1.bias roi_heads.recognizer_head.backbone.conv2.bias WARNING [04/17 15:55:54 fvcore.common.checkpoint]: The checkpoint state_dict contains keys that are not used by the model: roi_heads.recognizer_head.backbone.conv1.norm.{bias, num_batches_tracked, running_mean, running_var, weight} roi_heads.recognizer_head.backbone.conv2.norm.{bias, num_batches_tracked, running_mean, running_var, weight}
I found that the code used 'loss_orientation' during the training , but it is not included in equation 4 of the paper. There is an 'roi_heads.box_predictor.orientation_pred' in the code but not in the model file you provided, under what circumstances should the 'roi_heads.box_predictor.orientation_pred' be used?