patrick-kidger/equinox
Elegant easy-to-use neural networks + scientific computing in JAX. https://docs.kidger.site/equinox/
PythonApache-2.0
Issues
- 2
Right way to count cache size
#743 opened by AakashKumarNain - 2
Test fail with latest jaxtyping (0.2.29)
#740 opened by GaetanLepage - 5
Error with data sharding
#738 opened by andremfreitas - 2
docs request for filter_jacs/hessian
#741 opened by aidancrilly - 4
A simple DeepMLP is taking 11GB/16GB in GPU RAM
#737 opened by Qazalbash - 6
Flash Attention
#732 opened by neel04 - 4
Conv Transpose doesn't exactly transpose
#728 opened by lockwo - 11
Unroll in eqx's internal `scan`
#726 opened by neel04 - 2
Kolmogorov Arnold Networks with equinox [kanx]
#730 opened by stergiosba - 3
Custom JVP/VJP definition within Module?
#725 opened by leonard-gleyzer - 1
Static/Dynamic issues with filter_eval_shape
#729 opened by quattro - 7
- 34
Possible memory leak?
#673 opened by AakashKumarNain - 3
Comparisons to NNX
#718 opened by jlperla - 5
Strange error with multiple tuples as attributes and `equinox.tree_at`?
#715 opened by leonard-gleyzer - 2
How to serialize a model + state pair?
#714 opened by Artur-Galstyan - 7
- 2
merging multiple eqx.Module classes
#713 opened by ToshiyukiBandai - 1
Getting a type error when initializing Linear module (unexpected keyword argument 'key')
#712 opened by AbhinavRao23 - 5
`eqx.tree_at` fails with `TypeError: Operation undefined, {x} is not a leaf of the pytree` when custom pytree attributes are modified within `__init__`
#711 opened by nicolasJouvin - 5
- 9
Regarding trainable parameters in equinox
#710 opened by Basant1861 - 1
How to print HLO for jit accelated function?
#700 opened by nanoric - 5
lax.scan for equinox Modules ?
#709 opened by mchagneux - 3
Incorrect error handling in error_if
#696 opened by danielward27 - 6
- 4
- 1
Graph Neural Networks
#702 opened by stergiosba - 0
Broken link in documentation
#698 opened by lockwo - 19
Sharding - shard `eqx.Module` as well as inputs?
#688 opened by homerjed - 3
Add a [source] to docs [feature request]
#683 opened by lockwo - 2
- 12
- 4
Optimally training ensembles.
#693 opened by BabaYara - 1
Type annotations for "struct of arrays" ?
#692 opened by nboyd - 6
- 2
Why is this function recompiling?
#686 opened by jdpeiffer6 - 9
[Enhancement] Adding `dtype` to nn layers
#679 opened by AakashKumarNain - 3
- 9
What is the recommended pattern for optimization loops in transformed parameter spaces?
#676 opened by mjo22 - 2
Feature request for custom behavior in apply_updates
#678 opened by mjo22 - 4
Question: Support for Transformer Engine
#669 opened by ConnorBaker - 0
- 1
Masking in Embedding layer
#668 opened by AakashKumarNain - 5
Recursively setting weights of submodules
#657 opened by AakashKumarNain - 5
BatchNorm training instability
#659 opened by andrewdipper - 1
eqx.nn.Conv2d slower than torch.nn.Conv2d
#667 opened by lowlorenz - 2
Speeding up Newton Update
#666 opened by BabaYara - 11
- 1