Joshua Z. Zhang
Joshua Z. Zhang
@maram992 just specify use `--cpu`
There's no features for each box technically, the features are shared, for example, conv4_3_output
You are correct, fixed now.
No-resizing is potentially good for performance, but training is more complicated since you can only do batch-1 training, and there's not a easy solution for mini-batching like fast-rcnn.
I'm verifying this.
seems like it was related to some parameter change including lr_mult and learning rate. I did modified them for sake of a lot of GPUs, and it turns out to...
Using multi devices will reduce the effective batch-size, 'valid' normalization in softmaxOutput and makeloss will be affected.
can you show the full stack trace? Might be related to CPU because no body even my self have trained on CPU before.
@arisha07 You might be using a very old mxnet version
please check mxnet version, make sure it's 1.1 or 1.2