Releases · chaobrain/brainstate

This release standardizes the features and functionalities of brainstate.

BrainState establishes a State-based compilation and augmentation system for JAX code. This State-based syntax supports most of JAX transformations, including:

jit
vmap
grad
and others

What's Changed

Update documtation by @chaoming0625 in #1
add brainstate.nn.Embedding; support reset_state() by @chaoming0625 in #2
rewrite _make_jaxpr() function to be compatible with jax==0.4.29 by @chaoming0625 in #3
remove math module by @chaoming0625 in #4
Linear interpolation of the delay by @chaoming0625 in #5
Simplify the linear interpolation to accelerate brainstate.Delay by @chaoming0625 in #6
Fix jit error by @chaoming0625 in #7
Updates by @chaoming0625 in #8
Several enhancements by @chaoming0625 in #9
upgrade to support brainunit>=0.0.2 by @chaoming0625 in #10
Enable the environ setting to interact with the user defined behaviors by @chaoming0625 in #11
Update by @chaoming0625 in #12
Checkpointed scan and for_loop by @chaoming0625 in #13
update delays for compatible with brainunit by @chaoming0625 in #14
New version of brainstate by @chaoming0625 in #17
Update code and doc by @chaoming0625 in #18
remove example by @chaoming0625 in #19
update examples by @chaoming0625 in #20
Update docs and examples by @chaoming0625 in #21
Update documentation by @chaoming0625 in #22
change code indent, remove keep_size, add HiddenState by @chaoming0625 in #24
update examples to match new Dynamics design by @chaoming0625 in #25
unify interface of OptaxOptimizer and built-in optimizers by @chaoming0625 in #26
Update doc of concepts and pygraph & APIs related to program augmentation by @chaoming0625 in #27
Standarize program augmentation APIs and their documentation by @chaoming0625 in #28
add docs about tutorival of event_driven_computation by @hjflo in #29
[doc] Add docs about tutorial of checkpointing, random numbers and snn training by @Routhleck in #23
[doc] Fix formula rendering issues in snn training and correct incorrect references to checkpoints by @Routhleck in #30
Event driven by @hjflo in #31
Ann train snn sim by @senngadaisuki in #33
State & compilation docs by @ztqakita in #34
New styles of event-driven computation operators by @chaoming0625 in #35
ann tutorial by @senngadaisuki in #37
Event-driven operator updates by @chaoming0625 in #36
update file names by @chaoming0625 in #39
State compilation docs by @ztqakita in #38
[doc] Add lightweight approach checkpoint method of braintools by @Routhleck in #40
[docs] Update program_compilation-en.ipynb by @ztqakita in #41
Support physical unit-aware gradient computation using brainunit.autograd by @chaoming0625 in #42
add LayerNorm, RMSNorm, GroupNorm by @chaoming0625 in #43
compatible with jax==0.4.36 by @chaoming0625 in #45
[compatible with jax 0.4.36] remove ensure_compile_time_eval context, fix progress bar in loop transformation by @chaoming0625 in #46
unify CSRLinear, CSCLinear, COOLinear using SparseLinear by @chaoming0625 in #48
Snn sim by @senngadaisuki in #49
fix equantion show by @senngadaisuki in #50
Updates by @chaoming0625 in #51
Several updates by @chaoming0625 in #52
Enable to set bf16 precision in brainstate.environ module by @chaoming0625 in #53
Version compatible for custom op registeration by @chaoming0625 in #54
Many updates and compatible with jax>=0.5.0 by @chaoming0625 in #55
Remove brainstate.event module, please refer to brainevent library by @chaoming0625 in #57
Remove docs about event-driven operators by @chaoming0625 in #58
Update progress bar by @chaoming0625 in #59
Standardize state value reading and writing by @chaoming0625 in #60
Refactor Poisson input handling by @chaoming0625 in #62
update pretty representation by @chaoming0625 in #61
Many updates by @chaoming0625 in #63