Changes
Huge speed boost on high-performant hardware
Parts of the restoration pipeline are now being run in parallel with buffering in between each step. This enhances utilization / reduces wait times.
Speed on my RTX 3090 on average doubled compared to v0.3.2. On the Laptop with a throttled RTX 3050 I didn't see a noticeable difference (was already fully utilized on v0.3.2)
You can find detailed speed performance numbers in the README.
Full Changelog: v0.3.2...v0.4.0