Open leemgs opened 7 years ago
IMO, if you plan to use DNNs for production systems, in particular mobile, definitely caffe2.
@bgchun , Thank you for your opinion.
To Caffe2-based developers and commercial users
I think that Facebook and lots of contributors have submitted their patches to deploy Caffe2 as a commercial deep learning framework (= production edge deployment) against Caffe. In case of Android mobile device, we could read an official instruction as following:
Anyone who already tried to migrate a base deep learning framework from Caffe to a Caffe2-based project?
\CC: @myungjoo , @jijoongmoon , @wonsubkim
Yes, as we've guessed, we should move on to Caffe2 as soon as possible.
Let's look at how we will be able to implement "Caffe2 Driver" for specific hardware and how we can access Caffe2 from our API sets.
https://github.com/caffe2/caffe2/issues/274#issuecomment-295381560
@Yangqing, @akyrola, Please, have a look at this. Could you give me your experience or opinion about this topic?
Dear All,
There are many existing projects based on the existing version, Caffe. After Facebook open sourced Caffe2 (Apr 18, 2017) as a next version of Caffe, It seems that Caffe2 include lots of new features (https://caffe2.ai/docs/caffe-migration.html#how-does-caffe-compare-to-caffe2) as following:
The official website provides how to convert from Caffe as https://caffe2.ai/docs/caffe-migration.html#caffe-to-caffe2. Nvidia also has already started Caffe2 related instruction as a 3rd party at https://developer.nvidia.com/caffe2.
At this time, does it reasonable time that we try to change a base deep learning framework from Caffe to Caffe2 in case of a commercial goal? Where can we find a detail comparison of two versions that is compared between Caffe and Caffe2?