v0.8.20

Bump version to v0.8.20

v0.8.19

What's Changed

• Remove the unnecessary async functions in loader by @guzongmin in #34

New Contributors

• @guzongmin made their first contribution in #34

Full Changelog: v0.8.18...v0.8.19

v0.8.18

Update dependencies.

v0.8.17

Allow specifying `rescale` layers when loading.

v0.8.16

Fix initial state loading (when loading buffer from CPU, don't use ca…

v0.8.15

Fix no subgroups quant GEMV.

v0.8.14

Limit buffer cache size.

v0.8.12

Update `web-rwkv` to v0.20.1

v0.8.11

Rename state GPU APIs to `write` and `read`.

v0.8.10

Simplify inference API.