Jiong Gong

Results 140 comments of Jiong Gong

I got the same problem with batch size 8 but better with batch size 7. Why is there big difference with subtle batch size change?

No. I gave up eventually and turned to Berkeley’s FCN. You mean your change can get the network converge eventually? From: Yubin Wang [mailto:[email protected]] Sent: Wednesday, March 01, 2017 6:12...

The proposed list contains modules that fall into the following categories: 1. Intel was the primary contributor and is willing to own the modules (“CPU frontend”, “oneDNN graph integration” and...

@malfet I applied the changes in CODEOWNERS to the merge_rules.json. Could you please review? For modules in category 2 and 3, there should be Meta owners and/or third-party owners too...

@malfet I removed NNC from the merge_rules.json and reverted changes from CODEOWNERS. Please review. Thanks.

Hi @malfet Does the PR look good to you? Any more comments?

> @yiqianglee, @jgong5, @vladimir-paramuzov, @dmitry-gorokhov, we need your feedback on this proposal. Currently, we are leaning towards making memory descriptor opaque and removing operation descriptors from API. Please let us...

@igorsafo The RFC proposed both short-term and long-term changes. May I know what changes target v3.0? Some comments from PyTorch side: 1. PyTorch has asymmetric quantization as default while this...