Results 123 comments of Xin Yao

The precision interval for fp16 between 2048 and 4096 is 2. https://en.wikipedia.org/wiki/Half-precision_floating-point_format. Considering elements in `feat_fp16` are between 0 and 1, they will be ignored due to the round-off error....

@eric-wieser, can you review it please?

Since I'm not a native speaker, I expect your help in improving the documentation. @eric-wieser

Any further advice? @eric-wieser

Anything I can do? @charris

So I need to checkout to a new branch (e.g. `enh-multi-dim-weights`) in my repo, and create another PR now? @mattip

> When I upgrade macOS Big Sur today, I have same issues. I have try to upgrade ss-local, also using v2ray-plugin master branch instead of old release version, Now the...

Still tracking it internally.