There are lots of deep learning frameworks accessed by researchers and developers. Each framework has its own advantages, one of these features may be speed for some specific models. You can see some papers for more information about different frameworks.
Use better GPUs or FPGAs. Most of frameworks support for distribution computing.
Can we use pvanet with end2end (from faster-rcnn)? If so, how? What is the command like? I notice that it takes pretty long to train my model.
If its not possible, what can I do to speed up the training process?
I want to get good mAP, so I'd have to use the full model, not compressed.
This is my overall loss after 100k iterations:![loss](https://cloud.githubusercontent.com/assets/20983161/23259986/c81c1708-fa0a-11e6-937b-5d057becd768.png)