Orion Reblitz-Richardson

Any progress on this? Having the ability to diff ipynb files in GitHub PRs would be huge.

Latest Caffe2 does support python3. We still need to update the documentation and tutorials to support this, though. Thanks for all the feedback and we're working on those doc updates.

Hi @kyw2539 - how did you install caffe2 in this case? Are you using conda or pip? It looks like there's some mismatch between the google base library you used...

Actually looks like you said this was conda above. Are you on a Mac or Linux? If Mac, you might want to try the new `conda build conda` steps at...

@WormCoder you are right - PyTorch and Caffe2 have merged and now use the same build and install system. However, PyTorch by default doesn't use OpenCV or ffmpeg. In order...

Thank you @BrettRyland for the alert. Looks like a new OpenCV dependency on `cv::calcOpticalFlowFarneback` was added with https://github.com/caffe2/caffe2/commit/4ba9624d1d0965ce9421370fba4423093dc4c8d1 5 days ago. Adding @pietern to see if (a) we need to...

Hi @jiminHwa - it looks like the SqueezeNet model in Amazon S3 (where the model downloader gets files from) was corrupt. I've just updated with the one in GitHub. Can...

Hi @CriCL. Welcome! The init_net typically stores all the weights, offsets, etc. used by the network, while the predict_net stores the operators that execute the net. Technically it's more like...

@CriCL if you want to retrain, you'll probably need to create a Python training script and use that to recreate the model. You can then load weights from the `pb`...

There is work being done by ROCm (https://rocm.github.io/index.html) on Caffe2 at https://github.com/ROCmSoftwarePlatform/caffe2 for OpenCL. Feel free to take a look at that as well.