Commit Graph
109 Commits
Author SHA1 Message Date
Dr. Kashif Rasul ad482cb526 formatting 2020-01-28 14:01:48 +01:00
Dr. Kashif Rasul 7bede84f45 enc_out needs to be repeated at dim=1 2020-01-28 13:44:17 +01:00
Dr. Kashif Rasul 438785cde1 copy trainer network to device 2020-01-28 12:12:32 +01:00
Dr. Kashif Rasul 23b0d39412 set tgt_mask as a buffer 2020-01-28 12:03:29 +01:00
Dr. Kashif Rasul 8760a9dc08 no cache reset in pytorch 2020-01-28 11:45:32 +01:00
Dr. Kashif Rasul 79902625d7 another typo 2020-01-28 11:31:32 +01:00
Dr. Kashif Rasul f19d402475 fix typo 2020-01-28 11:30:01 +01:00
Dr. Kashif Rasul 9322ce004a fix argumment name 2020-01-28 11:27:58 +01:00
Dr. Kashif Rasul 1ab2babbe3 transfomer takes [T, B, F] tensors 2020-01-28 11:21:06 +01:00
Dr. Kashif Rasul 5141ae69e9 fix loss 2020-01-28 10:50:53 +01:00
Dr. Kashif Rasul 969b8c41b3 fix typos 2020-01-28 10:30:53 +01:00
Dr. Kashif Rasul 1c257bd317 added encoder and decoder input layers 2020-01-28 10:17:00 +01:00
Dr. Kashif Rasul 72710c21d9 api to set the transformer dim and ff_dim 2020-01-28 10:10:47 +01:00
Kashif Rasul 8d92ae4b89 added static real features 2020-01-27 21:19:16 +01:00
Kashif Rasul 064baee01c fix mask
TODO check this
2020-01-27 21:08:41 +01:00
Kashif Rasul ff92cbad50 more typos 2020-01-27 20:28:38 +01:00
Kashif Rasul 9c22720b60 import missing 2020-01-27 20:19:16 +01:00
Kashif Rasul d95bcfe57f fix imports 2020-01-27 20:15:03 +01:00
Kashif Rasul eea4d90aba formatting 2020-01-27 20:05:12 +01:00
Kashif Rasul da584d309c typos 2020-01-27 20:03:10 +01:00
Kashif Rasul b1d1cdbea6 fix api to nn.Transformer 2020-01-27 19:57:54 +01:00
Dr. Kashif Rasul 2dc6f23efe use nn.Transformer 2020-01-27 17:43:51 +01:00
Dr. Kashif Rasul 4a1de2d06f fix typo 2020-01-27 17:00:09 +01:00
Dr. Kashif Rasul 3184f5d588 initial transformer model
todo decoder encoder
2020-01-27 16:24:02 +01:00
Dr. Kashif Rasul d1005e2aa5 renamed flag to dequantize 2020-01-22 11:05:39 +01:00
Dr. Kashif Rasul 04bdd7bbf2 added MAF option to model 2020-01-17 11:42:46 +01:00
Dr. Kashif Rasul 4c5b02864f added quantization option to model 2020-01-17 10:26:42 +01:00
Dr. Kashif Rasul e34561373e another typo 2020-01-16 10:51:16 +01:00
Dr. Kashif Rasul 177834451a another typo 2020-01-16 10:43:15 +01:00
Dr. Kashif Rasul 28f6ed1dbc fix typo 2020-01-16 10:41:38 +01:00
Dr. Kashif Rasul 00d790a66d use pytorch Dataloader and TransformedIterableDataset 2020-01-16 10:32:30 +01:00
Dr. Kashif Rasul 7fd92875fe RNN always gets mean scaled values 2020-01-15 15:49:09 +01:00
Dr. Kashif Rasul 6ee817ab72 dont set scale if scaling is false 2020-01-15 15:24:11 +01:00
Dr. Kashif Rasul b5ef5f90c2 put back scaling to flow 2020-01-15 13:12:23 +01:00
Kashif RasulandGitHub Enterprise 9938121a42 Test conditioning (#1)
* explicitly set the conditioning to flow

* try ELUs

* return first element of tuple

* typo

* revert ELU
2020-01-15 11:39:10 +01:00
Dr. Kashif Rasul 77793fa55f cleanup more 2020-01-14 17:12:34 +01:00
Dr. Kashif Rasul 53bb87e952 formatting 2020-01-14 11:57:46 +01:00
Dr. Kashif Rasul a1fe683254 remove comments and cleanup 2020-01-14 11:55:33 +01:00
Dr. Kashif Rasul 6979042497 cleanup and loss is mean over time 2020-01-14 11:14:09 +01:00
Kashif Rasul 0a8e9fdec8 set prediction network to eval mode 2020-01-13 20:01:51 +01:00
Dr. Kashif Rasul 748e0c4ad5 output_transform is none 2020-01-13 13:42:57 +01:00
Dr. Kashif Rasul 2f019604ff flow's input size is target_dim 2020-01-13 12:06:22 +01:00
Dr. Kashif Rasul 863c675981 rename
TODO remove and refactor arg. names
2020-01-13 12:04:11 +01:00
Dr. Kashif Rasul 127f99299d added missing variables 2020-01-13 11:57:50 +01:00
Dr. Kashif Rasul a8644501dc typo 2020-01-13 11:54:29 +01:00
Dr. Kashif Rasul 056f831285 store cell_type 2020-01-13 11:53:32 +01:00
Dr. Kashif Rasul 2d14e9d842 fix arguments to training network 2020-01-13 11:51:50 +01:00
Dr. Kashif Rasul c3744da5d7 only works for multivariate 2020-01-13 11:50:12 +01:00
Dr. Kashif Rasul 3d5ac3fe3e fix arg. 2020-01-13 11:42:29 +01:00
Dr. Kashif Rasul 77d27e1fbb another one 2020-01-13 11:41:30 +01:00