Goncalves Henriques, Andre (UG - Computer Science)
|
9257404edd
|
Move the name of the module
|
2024-04-21 15:15:00 +01:00 |
|
sugarme
|
bf24e57df9
|
upgrade libtorch2.1.0
|
2023-10-11 14:53:35 +11:00 |
|
sugarme
|
52d0bc412b
|
update installation script
|
2023-10-11 12:31:45 +11:00 |
|
sugarme
|
1cffab577c
|
fixed Linear.Forward with bias is nil and clean up
|
2023-08-12 22:09:34 +10:00 |
|
sugarme
|
b3d821d34e
|
temp use of ts.Randn() instead of ts.Uniform() as it causes mem leak
|
2023-08-12 20:42:39 +10:00 |
|
sugarme
|
163e625426
|
fixed mem blow-up due to gradient accumulation when loading model pretrained
|
2023-08-12 15:46:51 +10:00 |
|
sugarme
|
ef00723027
|
remove linear.bias init when not required
|
2023-08-09 16:25:09 +10:00 |
|
sugarme
|
01f1e3a7fc
|
wired up ts.Drop() for directly free mem
|
2023-07-07 22:30:08 +10:00 |
|
sugarme
|
c1ee7689ad
|
fixed check null at tensor ops return slice of tensors and clean-up
|
2023-07-07 16:08:15 +10:00 |
|
sugarme
|
f9cb2f5cc6
|
Clean up
|
2023-07-07 13:20:51 +10:00 |
|
sugarme
|
34e87b1302
|
added dtype option to nn package
|
2023-07-07 12:24:56 +10:00 |
|
sugarme
|
f45f0a7ed0
|
changed []ts.Tensor -> []*ts.Tensor
|
2023-07-05 23:56:48 +10:00 |
|
sugarme
|
66401b2730
|
updated KaimingUniform initialization and mnist CNN
|
2022-11-24 13:32:46 +11:00 |
|
sugarme
|
cc5792ecbf
|
added nn.MSELoss()
|
2022-03-16 20:46:04 +11:00 |
|
sugarme
|
b662d09a0e
|
added fields to layernorm config
|
2022-03-16 12:47:48 +11:00 |
|
sugarme
|
664928551b
|
added API Path.Remove()
|
2022-03-12 21:57:23 +11:00 |
|
sugarme
|
1c2e6f2599
|
change package 'tensor' to 'ts'
|
2022-03-12 18:20:20 +11:00 |
|
sugarme
|
5a6fac51f3
|
varstore reworked and update
|
2022-03-12 15:47:15 +11:00 |
|
sugarme
|
d95eaba5b3
|
WIP: varstore rework
|
2022-03-11 13:21:39 +11:00 |
|
sugarme
|
73d6c0ae86
|
added more building block modules
|
2022-02-15 12:40:28 +11:00 |
|
sugarme
|
a4e5f38705
|
fixed nn/rnn NewLSTM() clashed weight names
|
2022-01-21 11:05:11 +11:00 |
|
sugarme
|
653caf4be5
|
added nn.Path.Paths() and nn.VarStore.Summary() methods
|
2021-11-13 22:23:38 +11:00 |
|
sugarme
|
880a1b25df
|
fixed conv3D config
|
2021-08-03 17:56:25 +10:00 |
|
sugarme
|
620fccf452
|
added conv3dconfig
|
2021-08-03 12:36:10 +10:00 |
|
sugarme
|
dd03baa6b0
|
export nn/scheduler DefaultSchedulerOptions()
|
2021-07-25 18:29:11 +10:00 |
|
sugarme
|
bb00d244ed
|
added nn/conv config options and bump version
|
2021-07-25 18:23:00 +10:00 |
|
sugarme
|
d490db3fdf
|
added nn/schedulear NewLRScheduler
|
2021-07-25 17:37:44 +10:00 |
|
sugarme
|
5d8751bfaf
|
updated APIs changes
|
2021-07-23 00:54:41 +10:00 |
|
sugarme
|
731513a986
|
fixed #45 #48 RandInit
|
2021-07-21 00:04:53 +10:00 |
|
sugarme
|
c89e4b3ba1
|
reworked on resnet, densenet, added BCELoss, CrossEntropyLoss, changed DataLoader.Reset()
|
2021-07-14 10:38:11 +10:00 |
|
sugarme
|
3798a51c31
|
fixed incorrect message at nn.Varstore.Load method
|
2021-06-19 18:14:15 +10:00 |
|
sugarme
|
efa71f3d82
|
added warning to varstore.LoadPartial
|
2021-06-19 14:31:28 +10:00 |
|
sugarme
|
22d9c5695a
|
fixed multiple memory leakage at dataloader and vision image and efficientnet
|
2021-06-16 22:44:13 +10:00 |
|
sugarme
|
6f8ec3b69f
|
added AdamW optimizer
|
2021-06-12 18:40:42 +10:00 |
|
sugarme
|
50dd5b181c
|
added OneCycleLR
|
2021-06-12 18:21:43 +10:00 |
|
sugarme
|
2fd99cdffd
|
WIP: OneCycleLR
|
2021-06-12 14:54:05 +10:00 |
|
sugarme
|
5f57a216c7
|
added CosineAnnealingWarmRestarts
|
2021-06-12 10:54:59 +10:00 |
|
sugarme
|
056bac2ff8
|
added CyclicLR
|
2021-06-10 23:21:11 +10:00 |
|
sugarme
|
ccb7beaf51
|
changed to SchedulerOptions and added ReduceLROnPlateau
|
2021-06-10 15:06:00 +10:00 |
|
sugarme
|
aabb8814ac
|
Rewrote scheduler
|
2021-06-09 20:47:10 +10:00 |
|
sugarme
|
3f17f0b4cb
|
WIP: CosineAnnealingLR
|
2021-06-08 22:59:25 +10:00 |
|
sugarme
|
be07ef9288
|
added StepLR and MultiStepLR
|
2021-06-08 11:46:59 +10:00 |
|
sugarme
|
8e76313b30
|
updated LambdaLR
|
2021-06-07 22:34:17 +10:00 |
|
sugarme
|
f26d51d358
|
added LambdaLR
|
2021-06-07 22:26:31 +10:00 |
|
sugarme
|
dbab26c5aa
|
added libtorch APIs get/set optimizer learning rates
|
2021-06-06 18:04:08 +10:00 |
|
sugarme
|
b02185df22
|
added libtorch APIs GetLRs and ParamGroupNum
|
2021-06-04 23:07:59 +10:00 |
|
sugarme
|
720beffa62
|
fixed nn.Seq Forward nil pointer if layers length = 1
|
2021-05-15 17:50:08 +10:00 |
|
sugarme
|
d6fb8d88d8
|
example/jit-train: added
|
2021-01-02 16:49:36 +11:00 |
|
sugarme
|
82113d7225
|
added(nn/jit): ForwadT method
|
2021-01-02 14:24:29 +11:00 |
|
sugarme
|
d0727911c4
|
feat(tensor/jit): added all 'atm_' API; feat(nn/jit): added TrainableCModule struct
|
2021-01-02 14:18:49 +11:00 |
|