jiayq84

7 ms·

31.
by jiayq84 10y ago
There is no push to migrate from Caffe to Caffe2 for sure. After Facebook "dogfooding" our own implementation, I think it is safe to say that C2 is now much stable. I would encourage you to try migrating it, letting us know if you
32.
by jiayq84 10y ago
Haha yeah, I definitely feel your pain - as a caffe developer it really makes me cry when things get so incompatible. I've made some improvements in caffe2 to make it more modular - checkout http://GitHub.com/caffe2&#x2
33.
by jiayq84 10y ago
PyTorch definitely makes experimentation much better. For example, if you want to train some system that is highly dynamic (reinforcement learning, for example), you might want to use a real scripting language which is Python, and PyTorch m
34.
by jiayq84 10y ago
TensorRT currently is supporting Caffe; it would definitely be interesting to make a direct Caffe2 compatibility support.
35.
by jiayq84 10y ago
I am not aware of that right now. Internally we measure the performance by looking at the theoretical peak possible (like the scaling efficiency when you use distributed training). C2 has been doing pretty well; if you find performance degr
36.
by jiayq84 10y ago
Thanks - Python 3 support is coming soon, as you may expect, there are a few issues regarding string manipulation but we are figuring out the details.
37.
by jiayq84 10y ago
Yeah, Caffe2 has the same root as Caffe, so you'll see the image side really polished - for object recognition, AR, etc. Over the next weeks/months we'll share more examples on other applications such as RNNs.
38.
by jiayq84 10y ago
Ah, it is because we do not want to create too much chaos for people to migrate between codebases. Evan and I had played both scenarios and decided that it is cleaner to put the code in separate codebases. A lot of the runtime code are shar
39.
by jiayq84 10y ago
RE the performance - many frameworks nowadays enjoy the high-performance libraries such as CuDNN for optimized runtime on different platforms. For example, we've been collaborating with NVidia on optimized distributed training on GPUs,
40.
by jiayq84 10y ago
MKL is going to be merged very soon - a full MKL pull request is here if you'd like to track it: https://github.com/caffe2/caffe2/pull/270 And thanks so much for the awesome work from Intel!
41.
by jiayq84 10y ago
So we carefully made the core much small and also made the platform more modular, so that the dependencies can be minimal when you build on Android/iOS. With our build system (buck) we are having very small binary footprints, which he
42.
by jiayq84 10y ago
Yep, if you look under the scripts/ folder we are putting on example scripts that you can use to build on specific platforms. Let me know what you think and feel free to send issues/PRs!
43.
by jiayq84 10y ago
Hey guys - Yangqing here, worked on Caffe and Caffe2 over the years, super excited to have continuous contribution to the OSS community. I am more than happy to answer questions if you are curious.