Jiewei Qian
Jiewei Qian
pr welcome
FYI, I observe `torch.compile + channel_last` provides 2x speedup (no tiling, no face enhancing, fp16) on NVIDIA A4000. ``` self.model = self.model.to(memory_format=torch.channels_last) self.model = torch.compile(self.model) ``` Might be worth a...
> Thanks for your comments! Which model are you using? On my side, using channel_last seems to reduce performance by half. "Officlal" Real-ESRGAN x4 I suspect channel_last / channel_first gain...
I don't have a MPS capable device, so I don't have further comments here. This thread suggests performance gain from changing layout varies by device and network architecture: https://github.com/pytorch/pytorch/issues/92542 My...
Same problem. Ubuntu 14.04, x86_64
V1 `playUrl` seems to be still working. Assuming response data `quality` field is correct
v2 provides, qn=10000 （原画？） * (proto 0?) flv: hevc/h264; H264 URL appears to agree with v1 API * (proto 1?) m3u8: hevc/h264; Much slower TTFB Open Q: * Unclear if...
Warn: new HLS stream m4s chunk pacing is broken in high latency situations. Guaranteed jank.
Fecha does not support parsing / formatting of `week`. I guess element need to hack it somehow? Modify week's parser might solve the problem. I might consider removing week in...
Interesting. Can you make a PR for it? I recommend integrate week format into our custom fecha (`/src/utils/date.js`)? So date-picker can treat week number as normal dates. (Thus removing custom...