|
|
5860faf85e
|
Use 3d memory layout for convolution
|
2024-03-20 19:15:27 +01:00 |
|
|
|
ef63cbd9f1
|
Implement avg pooling
|
2024-03-19 22:33:43 +01:00 |
|
|
|
a0fc1b00ae
|
Implement max pooling layer
|
2024-03-19 22:04:58 +01:00 |
|
|
|
364715ff70
|
Refactor kernels
|
2024-03-19 21:37:25 +01:00 |
|
|
|
b6c4b7d2ae
|
Refactor layers
|
2024-03-19 21:35:05 +01:00 |
|
|
|
8d14b74f66
|
Implement Add layer
|
2024-03-18 20:37:13 +01:00 |
|
|
|
6cf604423a
|
Combine padding and conv kernel
|
2024-03-18 19:53:40 +01:00 |
|
|
|
e6d3757312
|
Change unsigned int to int
|
2024-03-18 19:40:00 +01:00 |
|
|
|
aac0c3a826
|
Implement concat layer
|
2024-03-17 21:38:29 +01:00 |
|
|
|
cbdb4e7707
|
Test softmax
|
2024-03-17 19:08:16 +01:00 |
|
|
|
42d646750b
|
Abstract activation and implement softmax
|
2024-03-17 18:37:15 +01:00 |
|
|
|
0c22fac64e
|
Add toplevel CUDANet namespace
|
2024-03-17 16:08:53 +01:00 |
|
|
|
dc86cddeb7
|
Use tiling shmem for mat vec mul kernel
|
2024-03-15 23:33:09 +01:00 |
|
|
|
77004c16be
|
Use shared memory for mat vec mul kernel
|
2024-03-13 22:13:11 +01:00 |
|
|
|
708164e4d0
|
Implement simple input layer
|
2024-03-12 21:16:46 +01:00 |
|
|
|
9d91896f13
|
Change forward function to return output pointer
|
2024-03-12 20:50:49 +01:00 |
|
|
|
a3973f0b21
|
Add activation to conv2d
|
2024-03-11 21:05:38 +01:00 |
|
|
|
d2ab78fbc7
|
Add Kernels namespace
|
2024-03-11 21:04:23 +01:00 |
|
|
|
e0178e2d5c
|
Cleanup and refactor
|
2024-03-11 20:39:44 +01:00 |
|
|
|
f3112311da
|
Make conv2d work again
|
2024-03-10 19:13:22 +01:00 |
|
|
|
d177a67cd6
|
Add bias to conv2d
|
2024-03-09 23:03:23 +01:00 |
|
|
|
4f3c4f1afb
|
Fix conv2d kernel dims
|
2024-03-09 22:55:37 +01:00 |
|
|
|
fceef07a9b
|
Remove linear activation kernel
|
2024-03-09 22:54:23 +01:00 |
|
|
|
a3d85a10fc
|
Working conv2d forward
|
2024-03-09 21:08:16 +01:00 |
|
|
|
e51aabc2f2
|
Initial cuda conv kernel implementation
|
2024-03-08 23:35:54 +01:00 |
|
|
|
4b6fcbc191
|
Implement simple test for host conv2d
|
2024-03-08 23:12:04 +01:00 |
|
|
|
7e75943a6b
|
Add stride to index calculation
|
2024-03-07 21:32:39 +01:00 |
|
|
|
fc2c1616b4
|
Initial cpu conv implementation
|
2024-03-07 21:24:59 +01:00 |
|
|
|
07f231a30b
|
Switch padding kernel to row major
|
2024-03-05 21:04:11 +01:00 |
|
|
|
f4257afd5a
|
Remove cublas dependency
|
2024-03-05 18:41:35 +01:00 |
|
|
|
98ad84c659
|
Add matrix math kernels
|
2024-03-05 17:38:46 +01:00 |
|
|
|
cfc5c46d5e
|
Initialize conv2d layer
|
2024-03-04 22:16:03 +01:00 |
|
|
|
f37320594a
|
Add activations enum
|
2024-03-03 15:24:54 +01:00 |
|
|
|
7e4460cc5e
|
Implement padding kernel
|
2024-03-03 15:24:39 +01:00 |
|
|
|
019ccc33d9
|
Start implementing padding kernel
|
2024-02-29 22:21:48 +01:00 |
|
|
|
045359cca2
|
Remove not needed code
|
2024-02-29 22:21:32 +01:00 |
|
|
|
9747abe53e
|
Fix dense layer forward prop
|
2024-02-27 21:47:46 +01:00 |
|
|
|
b1eb8b5806
|
Add activations test
|
2024-02-27 20:19:17 +01:00 |
|
|
|
48ba09b28d
|
Format source code using clang-format
|
2024-02-27 18:52:12 +01:00 |
|
|
|
5e1e0ed1d1
|
Initial activations implementation
|
2024-02-27 00:24:57 +01:00 |
|
|
|
6e99525ad0
|
Rename hheader files to .cuh
|
2024-02-26 19:53:46 +01:00 |
|
|
|
035f3b053b
|
Rename files to .cu and fix IDX2C usage
|
2024-02-21 20:03:04 +01:00 |
|
|
|
02fc9e4e8b
|
Use IDX2C macro properly
|
2024-02-19 22:26:54 +01:00 |
|
|
|
4b6fff9bfd
|
Improve test dense
|
2024-02-18 13:12:49 +01:00 |
|
|
|
ee1a8cc6e6
|
Set up basic tests with gtest
|
2024-02-17 23:07:26 +01:00 |
|
|
|
f541e2f7f8
|
Set up cmake to compile library
|
2024-02-17 23:07:09 +01:00 |
|
|
|
22e33a395b
|
Initialize and use cuBLAS properly
|
2024-02-10 17:17:07 +01:00 |
|
|
|
3e6b7dc7e6
|
Implement to_cuda function
|
2024-02-09 19:18:30 +01:00 |
|
|
|
6645cd31ba
|
Fix issues with cuda memory allocation
|
2024-02-09 18:18:02 +01:00 |
|
|
|
950021389c
|
Start dense layer implementation
|
2024-02-08 19:19:51 +01:00 |
|
