| commit | 630d3a5984a7ad3fc2e4f92c1ccb8594a9b00209 | [log] [tgz] |
|---|---|---|
| author | Dmytro Dzhulgakov <dzhulgakov@fb.com> | Wed Jan 25 11:06:25 2017 -0800 |
| committer | Facebook Github Bot <facebook-github-bot@users.noreply.github.com> | Wed Jan 25 11:14:51 2017 -0800 |
| tree | 6bedaebec17737f2e7cdd0ec3dba129b29d8e0d9 | |
| parent | 65f7c915fd471cf7845f6a9fe2d8172abb3c060c [diff] |
Fix blob serialization in KVStore ops Summary: Fixes the problem surfaced by D4446583. Our serialization interface is designed for chunking but recepients in distributed training didn't expect that. For now I just fixed the naming of the tensor and since our blobs are small it should work. I believe it's still wrong however for big tensors as we just concatenate the serialized proto strings of chunks here: https://fburl.com/6wayxglz and here: https://fburl.com/7k4nhjja . Deserialization path though just tries to deserialize it as a single proto. I'll make Blob::Serialize(name) version use non-chunking version in a separate diff. Just sending it to unblock for now. Side note - oujin - why do we have two versions of operator setting the blob? :) Is one of them added by Pieter? Maybe we should unify them a bit. Reviewed By: kennyhorror Differential Revision: D4460974 fbshipit-source-id: 485b4de7c8af8cd9eac44c06a1246deaf0b4d502
Caffe2 is a deep learning framework made with expression, speed, and modularity in mind. It is an experimental refactoring of Caffe, and allows a more flexible way to organize computation.
Caffe2 is released under the BSD 2-Clause license.
git clone --recursive https://github.com/bwasti/caffe2.git cd caffe2
brew install automake protobuf mkdir build && cd build cmake .. make
sudo apt-get install libprotobuf-dev protobuf-compiler libatlas-base-dev libgoogle-glog-dev libgtest-dev liblmdb-dev libleveldb-dev libsnappy-dev python-dev python-pip libiomp-dev libopencv-dev libpthread-stubs0-dev cmake
sudo pip install numpy
wget http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1404/x86_64/cuda-repo-ubuntu1404_8.0.44-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu1404_8.0.44-1_amd64.deb
sudo apt-get update
sudo apt-get install cuda
sudo apt-get install git
CUDNN_URL="http://developer.download.nvidia.com/compute/redist/cudnn/v5.1/cudnn-8.0-linux-x64-v5.1.tgz" &&
curl -fsSL ${CUDNN_URL} -O &&
sudo tar -xzf cudnn-8.0-linux-x64-v5.1.tgz -C /usr/local &&
rm cudnn-8.0-linux-x64-v5.1.tgz &&
sudo ldconfig
mkdir build && cd build
cmake ..
make
To run the tutorials you'll need ipython-notebooks and matplotlib, which can be installed on OS X with:
brew install matplotlib --with-python3 pip install ipython notebook
Ubuntu 14.04 (GCC)
OS X (Clang)
Options (both Clang and GCC)
BLAS
Other