julienChemillier
5ec7bab488
Add finetuning option
2023-05-25 16:32:37 +02:00
ec138581c2
cnn: Add option to disable random offset
2023-05-25 13:31:55 +02:00
08993ade85
cnn/train: add newline & change initialisation
2023-05-19 23:29:46 +02:00
5738892142
Add detailed train timings
2023-05-19 21:48:08 +02:00
dbd5362d7d
utils:cuda: Select best available GPU
2023-05-16 13:21:26 +02:00
9db15e96ac
Update headers and Makefile to reflect 4910137
2023-05-15 18:25:29 +02:00
329e213e1f
write image: automatically detect padding
2023-05-15 11:34:23 +02:00
06abf0bc6b
train.c: pick architecture based on dataset type
2023-05-15 10:45:14 +02:00
19005366d3
mem management: add faster way to free all memory
2023-05-15 10:44:09 +02:00
julienChemillier
c0808b9240
Change notation toward a consensus
2023-05-13 17:22:47 +02:00
julienChemillier
2790883586
Removal of warnings
2023-05-13 10:05:54 +02:00
fd36226600
Move generic files to src/common
2023-05-12 16:16:34 +02:00
3198e6f2d0
Load image from a separate thread
2023-04-02 17:34:31 +02:00
7511856621
Translate backward convolution to CUDA
...
Not working yet, CUDA kernels in `backpropagation.cu` don't have access to activation functions declared in `function.cu` using `get_activation_function_cuda`.
Temporary workaround: copy `backpropagation.cu` parts that don't work to `function.cu` (all the parts using function pointers in kernels
2023-03-30 18:16:41 +02:00
b893f11da0
Add random offset to mnist images during learning
2023-03-22 13:03:19 +01:00
julienChemillier
1bd92074ab
Back to multiple bias implementation
2023-03-18 13:25:58 +01:00
04087c3de4
Rename mnist folder to dense
2023-03-13 13:55:09 +01:00
81ff4f4d00
Trying to improve train.c readability
2023-03-11 13:45:00 +01:00
9b8aa1caf8
Add loss information
2023-02-24 14:36:48 +01:00
383d3bebfd
Check device memory block size
...
& add pretty-prints
& remove some warnings
2023-02-19 15:01:58 +01:00
db92b367ad
Fix abusive unnecessary use of shared memory
2023-02-19 10:22:22 +01:00
5c7c013d91
Change headers
2023-02-18 13:10:00 +01:00
1121d11577
Really exiting
2023-01-29 09:47:29 +01:00
bc18a7cb34
Update utils.c/u
2023-01-29 09:40:55 +01:00
64e1cf0ad5
Implement custom memory management
2023-01-28 22:04:38 +01:00
fe880f9aae
Add copy_network_parameters
2023-01-28 13:09:52 +01:00
7deef7c5c5
Add simple_one
2023-01-21 18:59:59 +01:00
5f47b93672
Add loss computation
2023-01-20 13:41:38 +01:00
cea4132e2d
Remove white spaces
2023-01-17 15:34:29 +01:00
61468044a9
Update train.c
2023-01-17 12:50:35 +01:00
771d69af20
Fix dataset memory leak
2023-01-14 15:28:02 +01:00
df48b92cf2
Fix various multithreading related issues
2023-01-14 15:02:57 +01:00
03ad4b01e5
Use OpenMP instead of clock time
...
It was previously not accurate when using threads
2023-01-13 15:58:11 +01:00
c913bf3195
Update train.c
2022-12-19 15:49:03 +01:00
cedb240df2
Add recovery option
2022-12-07 13:09:39 +01:00
963a4afcff
Add cnn knuth shuffle
2022-12-07 10:44:28 +01:00
7da3544d8b
Add recognize option
2022-11-25 15:17:47 +01:00
4cafabdbee
Update train.c
2022-11-23 11:37:26 +01:00
11ce67f845
Update multithreading
2022-11-23 10:41:19 +01:00
18900f3fe9
Change images aspect ration
2022-11-19 22:22:24 +01:00
d0b7781695
Add jpeg dataset implementation
2022-11-19 16:09:07 +01:00
5a0f807a00
Update update.c
2022-11-18 14:09:49 +01:00
d03f7493b2
Move update
2022-11-16 10:38:01 +01:00
julienChemillier
50af676027
Learning_rate is a (NON NULL) float
2022-11-15 18:15:18 +01:00
3e7309c6e1
Add current_accuracy estimation
2022-11-15 17:50:33 +01:00
julienChemillier
f5f22c6b3e
Fix issues
2022-11-15 12:58:00 +01:00
julienChemillier
d94f61bb77
First results in backprop ?
2022-11-15 12:50:38 +01:00
julienChemillier
7a0ee3be7b
Implementation of the initialisation
2022-11-04 10:54:32 +01:00
4637d62e73
Update backpropagation.c
2022-11-03 18:13:01 +01:00
cf050de1d6
Move to Makefile
2022-10-24 12:54:51 +02:00
