Hi,
Thank you for the excellent repo. I used multi gpus to train the model from scratch and when I check the performance of the network on validation set, it is little lower compared to the pre-trained checkpoint provided.
Any additional changes/suggestions for network training would be really helpful.
Regards
Niranjan.
Hi, Thank you for the excellent repo. I used multi gpus to train the model from scratch and when I check the performance of the network on validation set, it is little lower compared to the pre-trained checkpoint provided. Any additional changes/suggestions for network training would be really helpful. Regards Niranjan.