ddpt-policy-multiwoz21 / train_INFO.log
ChrisGeishauser's picture
Upload 3 files
2285042
raw history blame
No virus
10.2 kB
Visible device: cuda
Seed used: 1
Batch size: 64
Epochs: 40
Learning rate: 1e-05
Entropy weight: 0.01
Regularization weight: 0.0
Only use multiwoz like domains: False
We use: 100.0% of the data
Dialogue order used: 0
Vectorizer: Data set used is multiwoz21
We filter state by active domains: True
Vectorizer: Data set used is multiwoz21
Embedding semantic descriptions: True
Embedded descriptions successfully. Size: torch.Size([338, 768])
Data set used for descriptions: multiwoz21
We use Roberta to embed actions.
Didnt load a model
Start training
Epoch: 0
Average actions: 1.957058072090149
Average target actions: 2.669339895248413
Precision: 0.13822525597269625
Recall: 0.10146667362597213
F1: 0.11702736056346508
<<dialog policy>> epoch 0: saved network to mdl
Best Precision: 0.13822525597269625
Best Recall: 0.10146667362597213
Best F1: 0.11702736056346508
Epoch: 1
Precision: 0.13822525597269625
Recall: 0.10146667362597213
F1: 0.11702736056346508
Best Precision: 0.13822525597269625
Best Recall: 0.10146667362597213
Best F1: 0.11702736056346508
Epoch: 2
Average actions: 2.0794308185577393
Average target actions: 2.6675729751586914
Precision: 0.22303363258743134
Recall: 0.1737564591053813
F1: 0.19533519143318176
<<dialog policy>> epoch 2: saved network to mdl
Best Precision: 0.22303363258743134
Best Recall: 0.1737564591053813
Best F1: 0.19533519143318176
Epoch: 3
Precision: 0.22303363258743134
Recall: 0.1737564591053813
F1: 0.19533519143318176
Best Precision: 0.22303363258743134
Best Recall: 0.1737564591053813
Best F1: 0.19533519143318176
Epoch: 4
Average actions: 2.0110926628112793
Average target actions: 2.665806293487549
Precision: 0.26409084614319345
Recall: 0.19907093272091445
F1: 0.22701705306389688
<<dialog policy>> epoch 4: saved network to mdl
Best Precision: 0.26409084614319345
Best Recall: 0.19907093272091445
Best F1: 0.22701705306389688
Epoch: 5
Precision: 0.26409084614319345
Recall: 0.19907093272091445
F1: 0.22701705306389688
Best Precision: 0.26409084614319345
Best Recall: 0.19907093272091445
Best F1: 0.22701705306389688
Epoch: 6
Average actions: 1.9673057794570923
Average target actions: 2.667219877243042
Precision: 0.2910210146465719
Recall: 0.21467717521791324
F1: 0.2470863871200288
<<dialog policy>> epoch 6: saved network to mdl
Best Precision: 0.2910210146465719
Best Recall: 0.21467717521791324
Best F1: 0.2470863871200288
Epoch: 7
Precision: 0.2910210146465719
Recall: 0.21467717521791324
F1: 0.2470863871200288
Best Precision: 0.2910210146465719
Best Recall: 0.21467717521791324
Best F1: 0.2470863871200288
Epoch: 8
Average actions: 1.8258512020111084
Average target actions: 2.667926549911499
Precision: 0.30450038138825325
Recall: 0.20836160551176994
F1: 0.24742012457776819
<<dialog policy>> epoch 8: saved network to mdl
Best Precision: 0.30450038138825325
Best Recall: 0.21467717521791324
Best F1: 0.24742012457776819
Epoch: 9
Precision: 0.30450038138825325
Recall: 0.20836160551176994
F1: 0.24742012457776819
Best Precision: 0.30450038138825325
Best Recall: 0.21467717521791324
Best F1: 0.24742012457776819
Epoch: 10
Average actions: 1.7796674966812134
Average target actions: 2.66333270072937
Precision: 0.3297132588483475
Recall: 0.2202620178506185
F1: 0.2640966268227048
<<dialog policy>> epoch 10: saved network to mdl
Best Precision: 0.3297132588483475
Best Recall: 0.2202620178506185
Best F1: 0.2640966268227048
Epoch: 11
Precision: 0.3297132588483475
Recall: 0.2202620178506185
F1: 0.2640966268227048
Best Precision: 0.3297132588483475
Best Recall: 0.2202620178506185
Best F1: 0.2640966268227048
Epoch: 12
Average actions: 1.8398014307022095
Average target actions: 2.67004656791687
Precision: 0.34064769975786924
Recall: 0.23498094890129964
F1: 0.27811583011583013
<<dialog policy>> epoch 12: saved network to mdl
Best Precision: 0.34064769975786924
Best Recall: 0.23498094890129964
Best F1: 0.27811583011583013
Epoch: 13
Precision: 0.34064769975786924
Recall: 0.23498094890129964
F1: 0.27811583011583013
Best Precision: 0.34064769975786924
Best Recall: 0.23498094890129964
Best F1: 0.27811583011583013
Epoch: 14
Average actions: 1.7070426940917969
Average target actions: 2.667219877243042
Precision: 0.35462034091835903
Recall: 0.22694295109348087
F1: 0.2767663908338638
Best Precision: 0.35462034091835903
Best Recall: 0.23498094890129964
Best F1: 0.27811583011583013
Epoch: 15
Precision: 0.35462034091835903
Recall: 0.22694295109348087
F1: 0.2767663908338638
Best Precision: 0.35462034091835903
Best Recall: 0.23498094890129964
Best F1: 0.27811583011583013
Epoch: 16
Average actions: 1.6812468767166138
Average target actions: 2.6643927097320557
Precision: 0.34859650575474044
Recall: 0.21974006994101988
F1: 0.2695607632219234
Best Precision: 0.35462034091835903
Best Recall: 0.23498094890129964
Best F1: 0.27811583011583013
Epoch: 17
Precision: 0.34859650575474044
Recall: 0.21974006994101988
F1: 0.2695607632219234
Best Precision: 0.35462034091835903
Best Recall: 0.23498094890129964
Best F1: 0.27811583011583013
Epoch: 18
Average actions: 1.675270438194275
Average target actions: 2.6640396118164062
Precision: 0.35976419794088343
Recall: 0.22616002922908293
F1: 0.27772970547703746
Best Precision: 0.35976419794088343
Best Recall: 0.23498094890129964
Best F1: 0.27811583011583013
Epoch: 19
Precision: 0.35976419794088343
Recall: 0.22616002922908293
F1: 0.27772970547703746
Best Precision: 0.35976419794088343
Best Recall: 0.23498094890129964
Best F1: 0.27811583011583013
Epoch: 20
Average actions: 1.5666790008544922
Average target actions: 2.6647462844848633
Precision: 0.3769442716203004
Recall: 0.2213581084607756
F1: 0.27892140743176586
<<dialog policy>> epoch 20: saved network to mdl
Best Precision: 0.3769442716203004
Best Recall: 0.23498094890129964
Best F1: 0.27892140743176586
Epoch: 21
Precision: 0.3769442716203004
Recall: 0.2213581084607756
F1: 0.27892140743176586
Best Precision: 0.3769442716203004
Best Recall: 0.23498094890129964
Best F1: 0.27892140743176586
Epoch: 22
Average actions: 1.6693706512451172
Average target actions: 2.6661596298217773
Precision: 0.3716379382130069
Recall: 0.23294535205386502
F1: 0.2863834702258727
<<dialog policy>> epoch 22: saved network to mdl
Best Precision: 0.3769442716203004
Best Recall: 0.23498094890129964
Best F1: 0.2863834702258727
Epoch: 23
Precision: 0.3716379382130069
Recall: 0.23294535205386502
F1: 0.2863834702258727
Best Precision: 0.3769442716203004
Best Recall: 0.23498094890129964
Best F1: 0.2863834702258727
Epoch: 24
Average actions: 1.6701388359069824
Average target actions: 2.6643927097320557
Precision: 0.3714618714618715
Recall: 0.23289315726290516
F1: 0.2862917455327067
Best Precision: 0.3769442716203004
Best Recall: 0.23498094890129964
Best F1: 0.2863834702258727
Epoch: 25
Precision: 0.3714618714618715
Recall: 0.23289315726290516
F1: 0.2862917455327067
Best Precision: 0.3769442716203004
Best Recall: 0.23498094890129964
Best F1: 0.2863834702258727
Epoch: 26
Average actions: 1.6909722089767456
Average target actions: 2.665099620819092
Precision: 0.3781160016454134
Recall: 0.2398872592515267
F1: 0.2935428242958421
<<dialog policy>> epoch 26: saved network to mdl
Best Precision: 0.3781160016454134
Best Recall: 0.2398872592515267
Best F1: 0.2935428242958421
Epoch: 27
Precision: 0.3781160016454134
Recall: 0.2398872592515267
F1: 0.2935428242958421
Best Precision: 0.3781160016454134
Best Recall: 0.2398872592515267
Best F1: 0.2935428242958421
Epoch: 28
Average actions: 1.8047566413879395
Average target actions: 2.6643927097320557
Precision: 0.3654779326811985
Recall: 0.24766428310454616
F1: 0.29525231783958683
<<dialog policy>> epoch 28: saved network to mdl
Best Precision: 0.3781160016454134
Best Recall: 0.24766428310454616
Best F1: 0.29525231783958683
Epoch: 29
Precision: 0.3654779326811985
Recall: 0.24766428310454616
F1: 0.29525231783958683
Best Precision: 0.3781160016454134
Best Recall: 0.24766428310454616
Best F1: 0.29525231783958683
Epoch: 30
Average actions: 1.680601716041565
Average target actions: 2.6640396118164062
Precision: 0.37665562913907286
Recall: 0.23748629886737305
F1: 0.2913025384935497
Best Precision: 0.3781160016454134
Best Recall: 0.24766428310454616
Best F1: 0.29525231783958683
Epoch: 31
Precision: 0.37665562913907286
Recall: 0.23748629886737305
F1: 0.2913025384935497
Best Precision: 0.3781160016454134
Best Recall: 0.24766428310454616
Best F1: 0.29525231783958683
Epoch: 32
Average actions: 1.7778853178024292
Average target actions: 2.667219877243042
Precision: 0.3660120491354354
Recall: 0.2441672321102354
F1: 0.2929242329367564
Best Precision: 0.3781160016454134
Best Recall: 0.24766428310454616
Best F1: 0.29525231783958683
Epoch: 33
Precision: 0.3660120491354354
Recall: 0.2441672321102354
F1: 0.2929242329367564
Best Precision: 0.3781160016454134
Best Recall: 0.24766428310454616
Best F1: 0.29525231783958683
Epoch: 34
Average actions: 1.726846694946289
Average target actions: 2.66333270072937
Precision: 0.3723121526938874
Recall: 0.24129651860744297
F1: 0.29281732961743095
Best Precision: 0.3781160016454134
Best Recall: 0.24766428310454616
Best F1: 0.29525231783958683
Epoch: 35
Precision: 0.3723121526938874
Recall: 0.24129651860744297
F1: 0.29281732961743095
Best Precision: 0.3781160016454134
Best Recall: 0.24766428310454616
Best F1: 0.29525231783958683
Epoch: 36
Average actions: 1.8067078590393066
Average target actions: 2.6675729751586914
Precision: 0.37099753694581283
Recall: 0.2515788924265358
F1: 0.29983515287238344
<<dialog policy>> epoch 36: saved network to mdl
Best Precision: 0.3781160016454134
Best Recall: 0.2515788924265358
Best F1: 0.29983515287238344
Epoch: 37
Precision: 0.37099753694581283
Recall: 0.2515788924265358
F1: 0.29983515287238344
Best Precision: 0.3781160016454134
Best Recall: 0.2515788924265358
Best F1: 0.29983515287238344
Epoch: 38
Average actions: 1.7964909076690674
Average target actions: 2.6647462844848633
Precision: 0.36536823356307596
Recall: 0.2462550237486299
F1: 0.2942130207034173
Best Precision: 0.3781160016454134
Best Recall: 0.2515788924265358
Best F1: 0.29983515287238344
Epoch: 39
Precision: 0.36536823356307596
Recall: 0.2462550237486299
F1: 0.2942130207034173
Best Precision: 0.3781160016454134
Best Recall: 0.2515788924265358
Best F1: 0.29983515287238344