Closed L-M-Sherlock closed 3 weeks ago
I'm going to wait until the burn v0.15.0 upgrade is complete to batch all the work together since the merge action failed due to a burn update.
I would wait for burn v0.16.0. The v0.15.0 slows down the training. I guess the next release won't happen soon. By the way, I'm curious whether burn has supported wasm with multi-threading. If so, we may not need to maintain a fork.
closed by e35f1c5fe531b9db5be3432df11fcf18b63111f2
I guess the next release won't happen soon.
Looks like they fixed it already - maybe they'll make a small patch release.
By the way, I'm curious whether burn has supported wasm with multi-threading. If so, we may not need to maintain a fork.
I believe they're unwilling to use Rust nightly, which is required for wasm with multi-threading:
This feature should be supported with Rust stable, since supporting another Rust toolchain is a lot of work and may slow down the development of other more important issues.
Until that changes, we must do this dance >_>
Sucks too because like, this feature is ~100 lines long... if you ignore CI. Which is a big if.
Fine. I can wait. The patch only improves 6% performance of FSRS. When will Rust support it? 2025?
As far as I can tell, it's not being actively worked on: https://github.com/rust-lang/rust/issues/77839
So no ETA >_>
This, by the way, is why I put so much effort into the CI - I expect this dance to go on for a long time.
After merging #246, the action failed. For details, see: https://github.com/open-spaced-repetition/fsrs-rs/actions/runs/11547276816/job/32136903375