300LiterPropofol

Thanks! I can see #451 is still open and I guess this patch hasn't been released in the latest version of serverless. I will close this issue once a new...

try to change all .cuda() to .to("cuda:1"), or at the very beginning of code write torch.cuda.set_device(1) and keep all .cuda() unchanged

I am confused too. Did you solve it?

The leap is normal because the current repository only supports batch_size = 1 training, it means it randomly select out one image instead of a mini batch that includes multiple...

> You need to revise `dataloader`(crop/resize/pad image to the same shape), `RPN`, `Trainer` etc. see [faster-rcnn.pytorch](https://github.com/jwyang/faster-rcnn.pytorch) for more detail Hi! I wonder which parts exactly I need to look into...

@whizzzkid Thank you a lot for this schema reference AVRO contribution! I wonder is there a description of how to use your commit? How should I register my producer on...

@whizzzkid Hi! I couldn't figure out how to use your fork. I mean, if I `npm install git+https://github.com/whizzzkid/confluent-schema-registry.git\#feature/schema-references`, your module seems to have a dependency compatibility issue. This command will...

> @300LiterPropofol that will not work because you need to build the module first, This project uses Azure to build the module. Internally we added a manual build step in...

@whizzzkid Thank you very much! I will give a try.

@whizzzkid Sorry to disturb you again. I tried to build the module locally and installed from local. I hit an error that ``` Error: undefined type name: selfdefinednamespace.selfdefinedname at Function.Type.forSchema...