Commit Graph

74 Commits

Author SHA1 Message Date
andrea b2be446446 fixed lr 2021-02-08 09:02:16 +01:00
andrea 80d0693cb1 removed unused lstm_class.py module 2021-02-05 11:55:19 +01:00
andrea ec6886dbab fixed bug in common 'l_test_raw_index_zero_shot' 2021-02-05 10:54:04 +01:00
andrea a6be7857a3 implemented zero-shot experiment code for VanillaFunGen and WordClassGen 2021-02-04 16:50:09 +01:00
andrea 495a0b6af9 implemented zero-shot experiment code for VanillaFunGen and WordClassGen 2021-02-04 13:00:18 +01:00
andrea f3fafd0f00 implemented zero-shot experiment code for VanillaFunGen and WordClassGen 2021-02-04 12:44:36 +01:00
andrea 8968570d82 implemented zero-shot experiment code for VanillaFunGen and WordClassGen 2021-02-04 12:24:57 +01:00
andrea 7affa1fab4 implemented zero-shot experiment code for VanillaFunGen and WordClassGen 2021-02-04 11:43:47 +01:00
andrea ab3bacb29c setting up zero-shot experiments (implemented for Recurrent and Bert but not tested for Bert) 2021-02-02 17:13:31 +01:00
andrea 6361a4eba0 setting up zero-shot experiments (implemented for Recurrent and Bert but not tested) 2021-02-02 16:12:08 +01:00
andrea ee98c5f610 setting up zero-shot experiments (implemented for Recurrent and Bert but not tested) 2021-02-02 16:09:49 +01:00
andrea 5821325c86 setting up zero-shot experiments (done and tested for WordClassGen) 2021-02-02 15:15:23 +01:00
andrea 7f493da0f8 setting up zero-shot experiments (done and tested for MuseGen) 2021-02-02 12:57:27 +01:00
andrea 10bed81916 Set arguments in order to reproduce 'master' performances with Neural setting 2021-02-02 11:23:55 +01:00
andrea 0e01d654cf lr monitor 2021-01-29 16:57:21 +01:00
andrea 2c70f37823 logging dirs 2021-01-29 10:49:47 +01:00
andrea e52b153ad4 fixed view generators' transform method 2021-01-28 18:12:20 +01:00
andrea bb84422d24 early stopping + typos 2021-01-26 18:56:24 +01:00
andrea 79cdaa0beb early stopping + typos 2021-01-26 18:13:34 +01:00
andrea e9a410faa4 early stopping 2021-01-26 18:12:14 +01:00
andrea 5cd36d27fc fixed cuda oom at inference time 2021-01-26 18:04:15 +01:00
andrea 20c76a2103 fixed imports 2021-01-26 15:52:09 +01:00
andrea ce4e32aad2 fixed imports 2021-01-26 15:19:09 +01:00
andrea a5912a22a9 merged with refactor 2021-01-26 15:10:32 +01:00
andrea 79eae9003b refactoring 2021-01-19 10:34:01 +01:00
andrea a5322ba227 refactoring 2021-01-19 10:31:57 +01:00
andrea 515acae15b rsc branch;
load pre-computed VGs' output if already stored in memory
2020-11-19 14:30:10 +01:00
andrea 8af763b130 fixed transform_mean() when dealing with only one VG;
modified default settings (new default: aggregation: mean, all_prob:True)
2020-10-29 12:55:07 +01:00
andrea 20dca61e22 minor fixes 2020-10-27 22:55:25 +01:00
andrea 94bfe6a036 fixed batcher 2020-10-27 15:08:39 +01:00
andrea 5906f85f33 branching to devel 2020-10-27 11:12:26 +01:00
andrea 526cf80b66 moved .sh files 2020-10-22 15:33:54 +02:00
andrea 90f24dab8e First commit 2020-10-22 15:11:49 +02:00
andrea 4a7a594a41 minor changes 2020-07-28 18:00:54 +02:00
andrea ba913f770a fixed train validation splitting 2020-07-27 17:15:06 +02:00
andrea 92f16e60eb baseline multilingual Bert
removed some comments
2020-07-27 12:05:18 +02:00
andrea a594a84dab baseline multilingual Bert 2020-07-27 12:00:33 +02:00
andrea d1fdad5f6e baseline multilingual Bert 2020-07-27 11:56:09 +02:00
Alejandro Moreo Fernandez 22b7ea7e66 huge refactoring, deep learning, and other stuff 2020-01-30 17:08:52 +01:00
Alejandro Moreo Fernandez d249c4801f bugfix in muse extract method 2020-01-20 14:53:14 +01:00
Alejandro Moreo Fernandez 5bb5c913c0 assert added 2020-01-20 12:35:36 +01:00
Alejandro Moreo Fernandez c14e8226b1 refactor: added MUSE to learning/transformers.py 2020-01-20 12:32:47 +01:00
Alejandro Moreo Fernandez 9bf1986402 bugfix 2020-01-17 18:17:27 +01:00
Alejandro Moreo Fernandez 1ba5e60031 refactor 2020-01-17 18:16:29 +01:00
Alejandro Moreo Fernandez 73d1e70ae9 refactor 2020-01-16 17:28:54 +01:00
Alejandro Moreo Fernandez cfd3a609a2 merged? 2020-01-16 14:30:20 +01:00
Alejandro Moreo Fernandez e96968fc45 merging 2020-01-09 10:50:35 +01:00
andrea 53198a7e2c implemented method to compute WCE only for well represented classes;
refactored MLE class in order to support WCE, standard embeddings and combinations;
sketched out NN implementation for WE compositionality;
still TODO SIF embeddings;
2020-01-07 17:05:41 +01:00
andrea 0e66fbf197 implemented method to compute WCE only for well represented classes -
refactored MLE class in order to support WCE, standard embeddings and combinations
2019-12-29 11:54:05 +01:00
andrea 56ee88220b typos 2019-12-17 10:42:29 +01:00