deep-spin entmax issues

deep-spin / entmax

The entmax mapping and its loss, a family of sparse softmax alternatives.

MIT License

407 stars 43 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Enable returning support size with exact algorithms for alpha=1.5 and alpha=2

#37 bpopeters closed 4 months ago
3
Instability of entmax bisect around 0?

#36 azhx opened 5 months ago
1
Nans gradients for tensors with singleton dimensions for budget_bisect.

#35 ssantos97 opened 7 months ago
5
Update minimum requirements and azure tests

#34 bpopeters closed 8 months ago
0
Implementation of normmax activation and loss and budget (k-subsets) activation.

#33 andre-martins closed 8 months ago
0
Fix bug in entmax bisection

#32 andre-martins closed 9 months ago
1
Add support for automatic mixed precision

#31 bpopeters opened 11 months ago
1
entmax_bisect bugs with fp16/bf16

#30 bpopeters opened 1 year ago
1
alpha-entmax

#29 levtelyatnikov opened 1 year ago
0
Code to reproduce figure 3

#28 erikdao closed 2 years ago
1
Fix setup.py

#27 cifkao closed 1 year ago
2
Release patch 1.0.1 with torch install_requires fix

#26 tgaddair closed 1 year ago
5
Do not import entmax => torch in setup.py

#25 frankier closed 1 year ago
1
`entmax_bisect` is not stable around `alpha=1`

#24 MartinXPN closed 3 years ago
3
A bug when alpha = 1 for entmax_bisect?

#23 mysteriouslfz closed 3 years ago
3
Change root finding to support alpha < 1.

#22 andre-martins closed 3 years ago
0
Index -1 is out of bounds

#21 liehtman opened 3 years ago
1
Alpha value less than one?

#20 kayuksel closed 3 years ago
8
Entmax fails when all inputs are -inf

#19 erickrf opened 3 years ago
1
Fix rare bug in activations.py

#18 antoniogois closed 4 years ago
0
Problem with sparse activations

#17 zylm opened 4 years ago
9
entmax_bisect leads to loss becoming nan

#16 prajjwal1 closed 4 years ago
16
Replicating the behaviour from the paper

#15 prajjwal1 closed 4 years ago
1
entmax implementation for Tesnoflow 2

#14 deepgradient opened 4 years ago
1
Unexpected behaviour of sparsemax gradients for 3d tensors

#13 mtreviso closed 3 years ago
8
make EntmaxBisect forward not take alpha argument

#12 bpopeters closed 4 years ago
0
Add torch>=1.0 requirement to setup.py

#11 mtreviso closed 4 years ago
0
Usage of alpha

#10 alibabadoufu closed 4 years ago
3
Sparse losses return nan when there is -inf in the input

#9 erickrf opened 5 years ago
2
ignore_index in loss functions

#8 erickrf opened 5 years ago
0
EntmaxBisect forward requires unused argument

#7 bpopeters closed 1 year ago
2
Setup, versioning, benchmarks

#6 vene closed 5 years ago
1
Gradient wrt alpha

#5 vene closed 5 years ago
14
[WIP] class methods, combining top k and naive entmax

#4 bpopeters closed 5 years ago
0
Errors when using the loss function

#3 berlino closed 5 years ago
9
Tensorflow implementation of entmax1.5

#2 justheuristic opened 5 years ago
10
Rename tsallis to entmax

#1 bpopeters closed 5 years ago
0