Releases · open-spaced-repetition/fsrs-rs

What's Changed

Feat/weight clipper by @Luc-Mcgrady in #9
Feat/rewrite FSRS item by @L-M-Sherlock in #8
Add a check that the code is formatted by @dae in #10
Run cargo fmt by @dae in #11
Add clippy to CI checks by @dae in #12
simplify by @asukaminato0721 in #15
rewrite test for convertor by @L-M-Sherlock in #14
Add BCELoss by @L-M-Sherlock in #18
Simplify Revlog->FSRSItem conversion by @dae in #13
Sequence padding by @asukaminato0721 in #16
Fix/feature dimension by @L-M-Sherlock in #26
add explanation for the sql query by @asukaminato0721 in #23
Feat/cosine_annealing_lr by @L-M-Sherlock in #25
Update to the latest Burn and pin dependencies by @dae in #33
use path join by @asukaminato0721 in #30
Feat/sort FSRSItem by length to speed up training by @L-M-Sherlock in #32
Feat/clip weights during training by @L-M-Sherlock in #35
Freezes weights by @GraesonB in #29
Expose a public API for training by @dae in #36
Fix/remove revlogs after manual reset & use const by @L-M-Sherlock in #38
add test into ci by @asukaminato0721 in #31
remove reshape() calls by @L-M-Sherlock in #39
Feat/pre-training for initial stability by @L-M-Sherlock in #42
Create LICENSE by @L-M-Sherlock in #44
Feat/insert default point to avoid overfitting with small dataset by @L-M-Sherlock in #46
Error handling and refactoring by @dae in #47
Feat/find optimal retention by @L-M-Sherlock in #48
Use an enum instead of a hashmap for table lookups by @dae in #49
more izip by @asukaminato0721 in #50
Feat/evaluation by @L-M-Sherlock in #52
Update burn and add progress handling/cancellation by @dae in #53
Feat/tune hype parameters by @L-M-Sherlock in #55
Feat/support scheduler by @L-M-Sherlock in #58
Benchmark and API tweaks by @dae in #59
A couple of minor fixes by @dae in #60
Some more minor tweaks to API by @dae in #62
Feat/answer buttons ratio & cost by @L-M-Sherlock in #65
use tryFrom and some path by @asukaminato0721 in #66
add more tests for convertor tests by @L-M-Sherlock in #64
Drop constraint on FloatElem by @dae in #67
add badge, add tests by @asukaminato0721 in #68
Update burn by @dae in #69
Avoid updating memory state when no days elapsed by @dae in #70
Update burn by @dae in #71
Feat/extract optimal retention parameters from revlog by @L-M-Sherlock in #73
Clip user-provided weights by @dae in #74
remove one mut by @asukaminato0721 in #75
remove feature dimension by @asukaminato0721 in #76
polyfill pow by @asukaminato0721 in #77
memory_state_from_sm2() and next_interval() by @dae in #82
Feat/outlier filter by @L-M-Sherlock in #80
Fix/correct lr_scheduler & increase batch_size & lr by @L-M-Sherlock in #83
Use u32 for inputs by @dae in #85
Feat/loss_aversion by @L-M-Sherlock in #86
update burn by @AlexErrant in #90
Run retention calculation in parallel by @dae in #91
reduce some dup code by @asukaminato0721 in #92
Support overriding initial S/D when revlog incomplete by @dae in #93
Laplace smoothing by @asukaminato0721 in #94
Fix/BatchShuffledDataLoader by @L-M-Sherlock in #96
Feat/stratified k fold by @L-M-Sherlock in #95
Update burn for tracel-ai/burn#839 by @dae in #97
Update burn for tracel-ai/burn#843 by @dae in #98
bump toolchain version by @asukaminato0721 in #100
Factor sm2 retention into memory state calculation by @dae in #101
Fix panic when training interrupted by @dae in #102
Correctly track progress with n_splits > 1 by @dae in #104
Feat/update initial values of S0 by @L-M-Sherlock in #105
Fix/Use power forgetting curve in memory_state_from_sm2 by @user1823 in #103
Update burn-rs by @dae in #108
Feat/early stop by @L-M-Sherlock in #109
We can now use burn from crates.io by @dae in #110
Feat/calc SInc based on last d by @L-M-Sherlock in #111
Clippy nursery by @asukaminato0721 in #112
Fix/clamp the upper limit of w16 by @L-M-Sherlock in #115
let-else by @asukaminato0721 in #116
Doc/rewrite README by @L-M-Sherlock in #120
Feat/filter outlier in trainset by @L-M-Sherlock in #119
Feat/speed up finding optimal retention via brent's method by @L-M-Sherlock in #122
Fix progress monitoring sometimes not terminating by @dae in #123
Rework progress code for optimal retention by @dae in #124
Add sanity check for NaN values by @dae in #125
Dependency updates by @dae in #127
Feat/update default weights by @L-M-Sherlock in #128
Increase the sample size.rs by @Expertium in #129
Use .is_finite() instead of .is_normal() by @dae in #131
Fix/pretrain default S0 should be from DEFAULT_WEIGHTS by @L-M-Sherlock in #132
Feat/remove items when retention is 100% & use mode parameters by @L-M-Sherlock in #135
Feat/flat power forgetting curve by @L-M-Sherlock in #134
Feat/add an env var to disable outlier filter by @L-M-Sherlock in #137
Fix/sort init_s0 on generating from r_s0_default by @L-M-Sherlock in #139
Use array ins...