Block a user
2f333bfc1d
config.h: reduce default CUDA blocksize
07369c3905
free: Add comment
e587261e4d
test/cnn_convolution: update to newer function
329e213e1f
write image: automatically detect padding
06abf0bc6b
train.c: pick architecture based on dataset type
003183d3fd
Add VGG16 architecture
de48f11e78
Fix a mistake i just did
19757b1c0d
Fix some mistakes
8de03863fa
Removal of useless lines
772b3511cf
common/utils: change function return type (
int
-> bool
)
321994df2b
Fix cuda compilation
c0808b9240
Change notation toward a consensus
521a1bb729
Add definition of 'convolution_not_outside'
3dd2e33fa9
Fix issues with the network not converging
f316882eeb
Add 'stride' and 'padding' to the forward
47a475a370
Removed extra memory allocation
49a2299c1c
Revert
fd36226
for cnn/creation.c
3ac318dd2c
Update .gitignore
9474734a8f
Update jpeg test
693dee581e
Move 'min' and 'max' definition to commons/utils
3672d07dff
Update Makefile
7c2c911976
cnn/export.c: Ajout de visual-propagation
af288166d6
Ajout stride dans average_ pooling et max_pooling
6e022fbd44
Ajout de 'stride' dans 'make_convolution'
augustin64
synced and deleted reference 2023-03-31 15:52:30 +02:00
cuda-backpropagation
at augustin64/tipe from mirror
710306a286
Remove duplicated functions
d6d03162b2
Update test/cnn_function.cu
05315a3567
Merge pull request #2 from augustin64/cuda-backpropagation
a3a803ba40
Fix previous compilation errors
7511856621
Translate backward convolution to CUDA
augustin64
synced commits to cuda-backpropagation at augustin64/tipe from mirror
2023-03-30 23:32:30 +02:00
augustin64
synced new reference cuda-backpropagation to augustin64/tipe from mirror
2023-03-30 23:32:30 +02:00
eeff720ae4
Add CUDA blocksize to config.h
2d6b4fe011
Implement copy_3d_array in CUDA