diff --git a/results_tencent/hubert_base/layer3/run4/model_best.pt b/results_tencent/hubert_base/layer3/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b3f0be55f03088717a1cbc80a50fa9f02e4d2157 --- /dev/null +++ b/results_tencent/hubert_base/layer3/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:631e2d40a5cd386ac92c257983f300df8ed14132d20eb6194a657d9320710394 +size 2749638 diff --git a/results_tencent/hubert_base/layer3/run4/train.log b/results_tencent/hubert_base/layer3/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3b6f0a5c053a7a92293dd6b6cddd1481ced7e573 --- /dev/null +++ b/results_tencent/hubert_base/layer3/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.585044427037239 +INFO:root: +[Valid][0][UTT][ MSE = 0.3097 | LCC = 0.9209 | SRCC = 0.9227 ] +INFO:root:Average loss=0.18621259391307832 +INFO:root: +[Valid][1][UTT][ MSE = 0.2580 | LCC = 0.9450 | SRCC = 0.9464 ] +INFO:root:Average loss=0.15076538735628128 +INFO:root: +[Valid][2][UTT][ MSE = 0.1899 | LCC = 0.9496 | SRCC = 0.9501 ] +INFO:root:Average loss=0.12869137960672378 +INFO:root: +[Valid][3][UTT][ MSE = 0.1957 | LCC = 0.9526 | SRCC = 0.9530 ] +INFO:root:Average loss=0.11905399978160858 +INFO:root: +[Valid][4][UTT][ MSE = 0.2246 | LCC = 0.9558 | SRCC = 0.9565 ] +INFO:root:Average loss=0.1064041962325573 +INFO:root: +[Valid][5][UTT][ MSE = 0.1910 | LCC = 0.9558 | SRCC = 0.9565 ] +INFO:root:Average loss=0.10023876416683197 +INFO:root: +[Valid][6][UTT][ MSE = 0.1964 | LCC = 0.9565 | SRCC = 0.9569 ] +INFO:root:Average loss=0.09155132782459259 +INFO:root: +[Valid][7][UTT][ MSE = 0.2142 | LCC = 0.9591 | SRCC = 0.9595 ] +INFO:root:Average loss=0.08545256182551383 +INFO:root: +[Valid][8][UTT][ MSE = 0.1806 | LCC = 0.9592 | SRCC = 0.9593 ] +INFO:root:Average loss=0.08068941643834114 +INFO:root: +[Valid][9][UTT][ MSE = 0.1953 | LCC = 0.9610 | SRCC = 0.9615 ] +INFO:root:Average loss=0.08015629923343659 +INFO:root: +[Valid][10][UTT][ MSE = 0.1945 | LCC = 0.9603 | SRCC = 0.9610 ] +INFO:root:Average loss=0.07107420200109482 +INFO:root: +[Valid][11][UTT][ MSE = 0.1723 | LCC = 0.9613 | SRCC = 0.9613 ] +INFO:root:Average loss=0.0715145302414894 +INFO:root: +[Valid][12][UTT][ MSE = 0.1820 | LCC = 0.9609 | SRCC = 0.9610 ] +INFO:root:Average loss=0.06508512896299362 +INFO:root: +[Valid][13][UTT][ MSE = 0.1582 | LCC = 0.9612 | SRCC = 0.9619 ] +INFO:root:Average loss=0.06264670187234879 +INFO:root: +[Valid][14][UTT][ MSE = 0.1757 | LCC = 0.9610 | SRCC = 0.9613 ] +INFO:root:Average loss=0.05488259424269199 +INFO:root: +[Valid][15][UTT][ MSE = 0.1552 | LCC = 0.9616 | SRCC = 0.9616 ] +INFO:root:Average loss=0.05504263737797737 +INFO:root: +[Valid][16][UTT][ MSE = 0.1690 | LCC = 0.9613 | SRCC = 0.9617 ] +INFO:root:Average loss=0.054654211789369585 +INFO:root: +[Valid][17][UTT][ MSE = 0.1705 | LCC = 0.9627 | SRCC = 0.9627 ] +INFO:root:Average loss=0.050548502132296565 +INFO:root: +[Valid][18][UTT][ MSE = 0.1641 | LCC = 0.9611 | SRCC = 0.9610 ] +INFO:root:Average loss=0.052234662771224974 +INFO:root: +[Valid][19][UTT][ MSE = 0.1602 | LCC = 0.9628 | SRCC = 0.9629 ] +INFO:root:Average loss=0.047292057484388354 +INFO:root: +[Valid][20][UTT][ MSE = 0.1571 | LCC = 0.9621 | SRCC = 0.9625 ] +INFO:root:Average loss=0.042859895259141925 +INFO:root: +[Valid][21][UTT][ MSE = 0.1713 | LCC = 0.9601 | SRCC = 0.9602 ] +INFO:root:Average loss=0.04676552534103393 +INFO:root: +[Valid][22][UTT][ MSE = 0.1733 | LCC = 0.9626 | SRCC = 0.9625 ] +INFO:root:Average loss=0.04063004079461098 +INFO:root: +[Valid][23][UTT][ MSE = 0.1769 | LCC = 0.9635 | SRCC = 0.9635 ] +INFO:root:Average loss=0.04098973496258259 +INFO:root: +[Valid][24][UTT][ MSE = 0.1704 | LCC = 0.9624 | SRCC = 0.9622 ] +INFO:root:Average loss=0.038265125632286075 +INFO:root: +[Valid][25][UTT][ MSE = 0.1553 | LCC = 0.9628 | SRCC = 0.9626 ] +INFO:root:Average loss=0.03658791047334671 +INFO:root: +[Valid][26][UTT][ MSE = 0.1662 | LCC = 0.9628 | SRCC = 0.9629 ] +INFO:root:Average loss=0.035070098757743834 +INFO:root: +[Valid][27][UTT][ MSE = 0.1590 | LCC = 0.9631 | SRCC = 0.9630 ] +INFO:root:Average loss=0.03913343170285225 +INFO:root: +[Valid][28][UTT][ MSE = 0.1528 | LCC = 0.9640 | SRCC = 0.9638 ] +INFO:root:Average loss=0.03600745689868927 +INFO:root: +[Valid][29][UTT][ MSE = 0.1657 | LCC = 0.9620 | SRCC = 0.9622 ] +INFO:root:Average loss=0.032543361201882365 +INFO:root: +[Valid][30][UTT][ MSE = 0.1691 | LCC = 0.9631 | SRCC = 0.9630 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1539 | LCC = 0.9657 | SRCC = 0.9658 ] diff --git a/results_tencent/hubert_base/layer4/run0/config.gin b/results_tencent/hubert_base/layer4/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer4/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer4/run0/model.pt b/results_tencent/hubert_base/layer4/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..49f66dc5f6fee64a81340777a336b7d0ca73203a --- /dev/null +++ b/results_tencent/hubert_base/layer4/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b5072cbdf0d4aa98d18c580e97f75346cea1701ce367ccdfd9fd3742bddd0cc8 +size 2747979 diff --git a/results_tencent/hubert_base/layer4/run0/model_best.pt b/results_tencent/hubert_base/layer4/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7292e1cca7d6f92770902bd7f5c01eae6d41a9ae --- /dev/null +++ b/results_tencent/hubert_base/layer4/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6d29521f53773706fa046312d38bdbc0dd52dc65e0b759397171c6cef8db264f +size 2749638 diff --git a/results_tencent/hubert_base/layer4/run0/train.log b/results_tencent/hubert_base/layer4/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ce788f380ba024708872db346346b4b169283fcb --- /dev/null +++ b/results_tencent/hubert_base/layer4/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4132269891500473 +INFO:root: +[Valid][0][UTT][ MSE = 0.3494 | LCC = 0.9285 | SRCC = 0.9277 ] +INFO:root:Average loss=0.1983527769446373 +INFO:root: +[Valid][1][UTT][ MSE = 0.2592 | LCC = 0.9408 | SRCC = 0.9408 ] +INFO:root:Average loss=0.15936129015684128 +INFO:root: +[Valid][2][UTT][ MSE = 0.2257 | LCC = 0.9478 | SRCC = 0.9487 ] +INFO:root:Average loss=0.13534067189693452 +INFO:root: +[Valid][3][UTT][ MSE = 0.1982 | LCC = 0.9505 | SRCC = 0.9509 ] +INFO:root:Average loss=0.11947703397274018 +INFO:root: +[Valid][4][UTT][ MSE = 0.2073 | LCC = 0.9528 | SRCC = 0.9535 ] +INFO:root:Average loss=0.11286948585510254 +INFO:root: +[Valid][5][UTT][ MSE = 0.1985 | LCC = 0.9563 | SRCC = 0.9572 ] +INFO:root:Average loss=0.10170915728807449 +INFO:root: +[Valid][6][UTT][ MSE = 0.1775 | LCC = 0.9574 | SRCC = 0.9576 ] +INFO:root:Average loss=0.09279373687505722 +INFO:root: +[Valid][7][UTT][ MSE = 0.1747 | LCC = 0.9582 | SRCC = 0.9583 ] +INFO:root:Average loss=0.08608387839794159 +INFO:root: +[Valid][8][UTT][ MSE = 0.1873 | LCC = 0.9584 | SRCC = 0.9584 ] +INFO:root:Average loss=0.08231061291694641 +INFO:root: +[Valid][9][UTT][ MSE = 0.1632 | LCC = 0.9590 | SRCC = 0.9597 ] +INFO:root:Average loss=0.07583020102977753 +INFO:root: +[Valid][10][UTT][ MSE = 0.1950 | LCC = 0.9600 | SRCC = 0.9600 ] +INFO:root:Average loss=0.07406844142079354 +INFO:root: +[Valid][11][UTT][ MSE = 0.1876 | LCC = 0.9601 | SRCC = 0.9600 ] +INFO:root:Average loss=0.06619387859106064 +INFO:root: +[Valid][12][UTT][ MSE = 0.1638 | LCC = 0.9601 | SRCC = 0.9601 ] +INFO:root:Average loss=0.06396218147873879 +INFO:root: +[Valid][13][UTT][ MSE = 0.1850 | LCC = 0.9614 | SRCC = 0.9612 ] +INFO:root:Average loss=0.06071103432774544 +INFO:root: +[Valid][14][UTT][ MSE = 0.1952 | LCC = 0.9615 | SRCC = 0.9610 ] +INFO:root:Average loss=0.05676183819770813 +INFO:root: +[Valid][15][UTT][ MSE = 0.1771 | LCC = 0.9616 | SRCC = 0.9614 ] +INFO:root:Average loss=0.05342316171526909 +INFO:root: +[Valid][16][UTT][ MSE = 0.1842 | LCC = 0.9589 | SRCC = 0.9581 ] +INFO:root:Average loss=0.05234560041129589 +INFO:root: +[Valid][17][UTT][ MSE = 0.1949 | LCC = 0.9610 | SRCC = 0.9602 ] +INFO:root:Average loss=0.050308501183986665 +INFO:root: +[Valid][18][UTT][ MSE = 0.1733 | LCC = 0.9605 | SRCC = 0.9600 ] +INFO:root:Average loss=0.04784963104128838 +INFO:root: +[Valid][19][UTT][ MSE = 0.1871 | LCC = 0.9606 | SRCC = 0.9605 ] +INFO:root:Average loss=0.04462337051331997 +INFO:root: +[Valid][20][UTT][ MSE = 0.1605 | LCC = 0.9618 | SRCC = 0.9615 ] +INFO:root:Average loss=0.04357859367132187 +INFO:root: +[Valid][21][UTT][ MSE = 0.1915 | LCC = 0.9620 | SRCC = 0.9618 ] +INFO:root:Average loss=0.040906944751739505 +INFO:root: +[Valid][22][UTT][ MSE = 0.1737 | LCC = 0.9614 | SRCC = 0.9612 ] +INFO:root:Average loss=0.04050542619824409 +INFO:root: +[Valid][23][UTT][ MSE = 0.1823 | LCC = 0.9619 | SRCC = 0.9612 ] +INFO:root:Average loss=0.03871918366849422 +INFO:root: +[Valid][24][UTT][ MSE = 0.1588 | LCC = 0.9622 | SRCC = 0.9617 ] +INFO:root:Average loss=0.036375430226325986 +INFO:root: +[Valid][25][UTT][ MSE = 0.1763 | LCC = 0.9623 | SRCC = 0.9617 ] +INFO:root:Average loss=0.03660297678411007 +INFO:root: +[Valid][26][UTT][ MSE = 0.1737 | LCC = 0.9621 | SRCC = 0.9612 ] +INFO:root:Average loss=0.03497839273512363 +INFO:root: +[Valid][27][UTT][ MSE = 0.1917 | LCC = 0.9624 | SRCC = 0.9626 ] +INFO:root:Average loss=0.03361639212071896 +INFO:root: +[Valid][28][UTT][ MSE = 0.1636 | LCC = 0.9621 | SRCC = 0.9611 ] +INFO:root:Average loss=0.03166963692009449 +INFO:root: +[Valid][29][UTT][ MSE = 0.1532 | LCC = 0.9626 | SRCC = 0.9624 ] +INFO:root:Average loss=0.03618527166545391 +INFO:root: +[Valid][30][UTT][ MSE = 0.1715 | LCC = 0.9631 | SRCC = 0.9624 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1782 | LCC = 0.9638 | SRCC = 0.9645 ] diff --git a/results_tencent/hubert_base/layer4/run1/config.gin b/results_tencent/hubert_base/layer4/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer4/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer4/run1/model.pt b/results_tencent/hubert_base/layer4/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f56647ff44a7273850779d591435bd7c3ea15258 --- /dev/null +++ b/results_tencent/hubert_base/layer4/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7774f084d32dabf7c973b897e0a57f680264ad1a7e85eb0133d71abf0cc67af3 +size 2747979 diff --git a/results_tencent/hubert_base/layer4/run1/model_best.pt b/results_tencent/hubert_base/layer4/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..afeb5a37c9618013d5910409bd3d547ae68c52e3 --- /dev/null +++ b/results_tencent/hubert_base/layer4/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cdf0a10b1af29b0ba4bc168d4fceb76de6c3f447d147387ef8f31b11da191d69 +size 2749638 diff --git a/results_tencent/hubert_base/layer4/run1/train.log b/results_tencent/hubert_base/layer4/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b7c3031a8e80742d555759d27d3421bbecbf085f --- /dev/null +++ b/results_tencent/hubert_base/layer4/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46159341096878054 +INFO:root: +[Valid][0][UTT][ MSE = 0.2647 | LCC = 0.9246 | SRCC = 0.9232 ] +INFO:root:Average loss=0.17943217718601226 +INFO:root: +[Valid][1][UTT][ MSE = 0.2516 | LCC = 0.9424 | SRCC = 0.9412 ] +INFO:root:Average loss=0.14303919917345048 +INFO:root: +[Valid][2][UTT][ MSE = 0.2039 | LCC = 0.9527 | SRCC = 0.9529 ] +INFO:root:Average loss=0.1231202420592308 +INFO:root: +[Valid][3][UTT][ MSE = 0.2090 | LCC = 0.9524 | SRCC = 0.9523 ] +INFO:root:Average loss=0.11004208463430405 +INFO:root: +[Valid][4][UTT][ MSE = 0.1795 | LCC = 0.9580 | SRCC = 0.9575 ] +INFO:root:Average loss=0.10258146399259567 +INFO:root: +[Valid][5][UTT][ MSE = 0.1911 | LCC = 0.9556 | SRCC = 0.9560 ] +INFO:root:Average loss=0.09297062692046165 +INFO:root: +[Valid][6][UTT][ MSE = 0.1581 | LCC = 0.9597 | SRCC = 0.9599 ] +INFO:root:Average loss=0.08695806893706322 +INFO:root: +[Valid][7][UTT][ MSE = 0.1604 | LCC = 0.9600 | SRCC = 0.9601 ] +INFO:root:Average loss=0.07936018177866935 +INFO:root: +[Valid][8][UTT][ MSE = 0.1651 | LCC = 0.9610 | SRCC = 0.9611 ] +INFO:root:Average loss=0.07519185608625412 +INFO:root: +[Valid][9][UTT][ MSE = 0.1544 | LCC = 0.9620 | SRCC = 0.9618 ] +INFO:root:Average loss=0.07150062566995621 +INFO:root: +[Valid][10][UTT][ MSE = 0.1695 | LCC = 0.9620 | SRCC = 0.9619 ] +INFO:root:Average loss=0.0688721090555191 +INFO:root: +[Valid][11][UTT][ MSE = 0.1699 | LCC = 0.9620 | SRCC = 0.9619 ] +INFO:root:Average loss=0.062040556102991105 +INFO:root: +[Valid][12][UTT][ MSE = 0.1548 | LCC = 0.9621 | SRCC = 0.9617 ] +INFO:root:Average loss=0.06254649198055268 +INFO:root: +[Valid][13][UTT][ MSE = 0.1629 | LCC = 0.9624 | SRCC = 0.9619 ] +INFO:root:Average loss=0.05622623857855797 +INFO:root: +[Valid][14][UTT][ MSE = 0.1584 | LCC = 0.9627 | SRCC = 0.9622 ] +INFO:root:Average loss=0.05377035996317864 +INFO:root: +[Valid][15][UTT][ MSE = 0.1703 | LCC = 0.9625 | SRCC = 0.9621 ] +INFO:root:Average loss=0.05389455099403858 +INFO:root: +[Valid][16][UTT][ MSE = 0.1509 | LCC = 0.9626 | SRCC = 0.9623 ] +INFO:root:Average loss=0.051428915783762935 +INFO:root: +[Valid][17][UTT][ MSE = 0.1588 | LCC = 0.9629 | SRCC = 0.9629 ] +INFO:root:Average loss=0.04775519397854805 +INFO:root: +[Valid][18][UTT][ MSE = 0.1972 | LCC = 0.9626 | SRCC = 0.9625 ] +INFO:root:Average loss=0.04895164597034454 +INFO:root: +[Valid][19][UTT][ MSE = 0.1733 | LCC = 0.9615 | SRCC = 0.9608 ] +INFO:root:Average loss=0.04453029118478298 +INFO:root: +[Valid][20][UTT][ MSE = 0.1857 | LCC = 0.9628 | SRCC = 0.9622 ] +INFO:root:Average loss=0.04112897390127182 +INFO:root: +[Valid][21][UTT][ MSE = 0.1653 | LCC = 0.9629 | SRCC = 0.9629 ] +INFO:root:Average loss=0.04150425037741661 +INFO:root: +[Valid][22][UTT][ MSE = 0.1566 | LCC = 0.9629 | SRCC = 0.9627 ] +INFO:root:Average loss=0.0384238301217556 +INFO:root: +[Valid][23][UTT][ MSE = 0.1699 | LCC = 0.9625 | SRCC = 0.9622 ] +INFO:root:Average loss=0.03549708844721317 +INFO:root: +[Valid][24][UTT][ MSE = 0.1550 | LCC = 0.9628 | SRCC = 0.9627 ] +INFO:root:Average loss=0.03578303168714046 +INFO:root: +[Valid][25][UTT][ MSE = 0.1473 | LCC = 0.9626 | SRCC = 0.9621 ] +INFO:root:Average loss=0.03402412061393261 +INFO:root: +[Valid][26][UTT][ MSE = 0.1626 | LCC = 0.9614 | SRCC = 0.9611 ] +INFO:root:Average loss=0.03447216331958771 +INFO:root: +[Valid][27][UTT][ MSE = 0.1545 | LCC = 0.9626 | SRCC = 0.9624 ] +INFO:root:Average loss=0.033135580465197566 +INFO:root: +[Valid][28][UTT][ MSE = 0.1596 | LCC = 0.9632 | SRCC = 0.9625 ] +INFO:root:Average loss=0.03470871095359326 +INFO:root: +[Valid][29][UTT][ MSE = 0.1561 | LCC = 0.9624 | SRCC = 0.9621 ] +INFO:root:Average loss=0.030326837614178658 +INFO:root: +[Valid][30][UTT][ MSE = 0.1778 | LCC = 0.9636 | SRCC = 0.9631 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1818 | LCC = 0.9636 | SRCC = 0.9646 ] diff --git a/results_tencent/hubert_base/layer4/run2/config.gin b/results_tencent/hubert_base/layer4/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer4/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer4/run2/model.pt b/results_tencent/hubert_base/layer4/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a6bdbcd887a8e83ce5753d4cec4bff2d0e09b907 --- /dev/null +++ b/results_tencent/hubert_base/layer4/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:31920c8a8d46c20a73d25446748cc3081817656848e8e930352e07db191f910f +size 2747979 diff --git a/results_tencent/hubert_base/layer4/run2/model_best.pt b/results_tencent/hubert_base/layer4/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..417ec33e6f6bc058b161c8f0bcdf7663cfdfbe37 --- /dev/null +++ b/results_tencent/hubert_base/layer4/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4ba0037fc4ec77b9c450bf9ec8156a248f586e572e0b8aa339723763b8c55cff +size 2749638 diff --git a/results_tencent/hubert_base/layer4/run2/train.log b/results_tencent/hubert_base/layer4/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6cdea835d47cc3b13c6a5d0f9ef39fd6a80475c1 --- /dev/null +++ b/results_tencent/hubert_base/layer4/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4995166178941727 +INFO:root: +[Valid][0][UTT][ MSE = 0.2875 | LCC = 0.9220 | SRCC = 0.9156 ] +INFO:root:Average loss=0.19209776985645294 +INFO:root: +[Valid][1][UTT][ MSE = 0.2143 | LCC = 0.9409 | SRCC = 0.9387 ] +INFO:root:Average loss=0.1524425759911537 +INFO:root: +[Valid][2][UTT][ MSE = 0.2173 | LCC = 0.9470 | SRCC = 0.9460 ] +INFO:root:Average loss=0.13451571381092073 +INFO:root: +[Valid][3][UTT][ MSE = 0.1917 | LCC = 0.9519 | SRCC = 0.9509 ] +INFO:root:Average loss=0.1196074041724205 +INFO:root: +[Valid][4][UTT][ MSE = 0.2000 | LCC = 0.9542 | SRCC = 0.9536 ] +INFO:root:Average loss=0.11005720752477646 +INFO:root: +[Valid][5][UTT][ MSE = 0.1752 | LCC = 0.9562 | SRCC = 0.9562 ] +INFO:root:Average loss=0.09841415452957153 +INFO:root: +[Valid][6][UTT][ MSE = 0.1990 | LCC = 0.9575 | SRCC = 0.9573 ] +INFO:root:Average loss=0.09684673476219177 +INFO:root: +[Valid][7][UTT][ MSE = 0.1586 | LCC = 0.9583 | SRCC = 0.9583 ] +INFO:root:Average loss=0.08796339356899262 +INFO:root: +[Valid][8][UTT][ MSE = 0.1773 | LCC = 0.9587 | SRCC = 0.9585 ] +INFO:root:Average loss=0.08198543867468834 +INFO:root: +[Valid][9][UTT][ MSE = 0.1740 | LCC = 0.9586 | SRCC = 0.9584 ] +INFO:root:Average loss=0.07829059413075447 +INFO:root: +[Valid][10][UTT][ MSE = 0.1532 | LCC = 0.9606 | SRCC = 0.9605 ] +INFO:root:Average loss=0.07272988843917846 +INFO:root: +[Valid][11][UTT][ MSE = 0.1798 | LCC = 0.9604 | SRCC = 0.9603 ] +INFO:root:Average loss=0.07076172545552253 +INFO:root: +[Valid][12][UTT][ MSE = 0.1845 | LCC = 0.9605 | SRCC = 0.9601 ] +INFO:root:Average loss=0.06509678456187248 +INFO:root: +[Valid][13][UTT][ MSE = 0.1585 | LCC = 0.9612 | SRCC = 0.9609 ] +INFO:root:Average loss=0.0646875624358654 +INFO:root: +[Valid][14][UTT][ MSE = 0.1955 | LCC = 0.9604 | SRCC = 0.9599 ] +INFO:root:Average loss=0.059271256491541864 +INFO:root: +[Valid][15][UTT][ MSE = 0.1885 | LCC = 0.9617 | SRCC = 0.9613 ] +INFO:root:Average loss=0.05441617016494274 +INFO:root: +[Valid][16][UTT][ MSE = 0.1361 | LCC = 0.9611 | SRCC = 0.9606 ] +INFO:root:Average loss=0.05317151245474815 +INFO:root: +[Valid][17][UTT][ MSE = 0.1677 | LCC = 0.9618 | SRCC = 0.9616 ] +INFO:root:Average loss=0.05185954561829567 +INFO:root: +[Valid][18][UTT][ MSE = 0.1465 | LCC = 0.9613 | SRCC = 0.9611 ] +INFO:root:Average loss=0.05082720635831356 +INFO:root: +[Valid][19][UTT][ MSE = 0.1561 | LCC = 0.9617 | SRCC = 0.9611 ] +INFO:root:Average loss=0.04718521195650101 +INFO:root: +[Valid][20][UTT][ MSE = 0.1727 | LCC = 0.9616 | SRCC = 0.9613 ] +INFO:root:Average loss=0.04441496774554252 +INFO:root: +[Valid][21][UTT][ MSE = 0.1624 | LCC = 0.9611 | SRCC = 0.9609 ] +INFO:root:Average loss=0.04399840787053108 +INFO:root: +[Valid][22][UTT][ MSE = 0.1665 | LCC = 0.9621 | SRCC = 0.9617 ] +INFO:root:Average loss=0.04263019375503063 +INFO:root: +[Valid][23][UTT][ MSE = 0.1582 | LCC = 0.9622 | SRCC = 0.9616 ] +INFO:root:Average loss=0.04249238841235638 +INFO:root: +[Valid][24][UTT][ MSE = 0.1437 | LCC = 0.9624 | SRCC = 0.9617 ] +INFO:root:Average loss=0.039484340891242024 +INFO:root: +[Valid][25][UTT][ MSE = 0.1466 | LCC = 0.9624 | SRCC = 0.9618 ] +INFO:root:Average loss=0.0373617038577795 +INFO:root: +[Valid][26][UTT][ MSE = 0.1524 | LCC = 0.9629 | SRCC = 0.9623 ] +INFO:root:Average loss=0.0370854345113039 +INFO:root: +[Valid][27][UTT][ MSE = 0.1691 | LCC = 0.9615 | SRCC = 0.9606 ] +INFO:root:Average loss=0.0355366690158844 +INFO:root: +[Valid][28][UTT][ MSE = 0.1702 | LCC = 0.9618 | SRCC = 0.9609 ] +INFO:root:Average loss=0.03367051811516285 +INFO:root: +[Valid][29][UTT][ MSE = 0.1715 | LCC = 0.9621 | SRCC = 0.9617 ] +INFO:root:Average loss=0.03245631115138531 +INFO:root: +[Valid][30][UTT][ MSE = 0.1599 | LCC = 0.9619 | SRCC = 0.9612 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1507 | LCC = 0.9649 | SRCC = 0.9658 ] diff --git a/results_tencent/hubert_base/layer4/run3/config.gin b/results_tencent/hubert_base/layer4/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer4/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer4/run3/model.pt b/results_tencent/hubert_base/layer4/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f67308dc913f2a08f9d070ceacc04b4f377d1b44 --- /dev/null +++ b/results_tencent/hubert_base/layer4/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df26764fc7a1bf316af3561f7506ad848a9918d870551939385c58d2eed73420 +size 2747979 diff --git a/results_tencent/hubert_base/layer4/run3/model_best.pt b/results_tencent/hubert_base/layer4/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..275aad12b7f4e7dffa3a00983200e533e1e771c8 --- /dev/null +++ b/results_tencent/hubert_base/layer4/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a704fcf06b002778ad75ab700f7b64e8f3f269e88563cb463f1d2c4c4fbfe336 +size 2749638 diff --git a/results_tencent/hubert_base/layer4/run3/train.log b/results_tencent/hubert_base/layer4/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ac009428c602c23a61547dca8b60e4b357c7e104 --- /dev/null +++ b/results_tencent/hubert_base/layer4/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44030873358249667 +INFO:root: +[Valid][0][UTT][ MSE = 0.3020 | LCC = 0.9279 | SRCC = 0.9280 ] +INFO:root:Average loss=0.18206725883483887 +INFO:root: +[Valid][1][UTT][ MSE = 0.2405 | LCC = 0.9444 | SRCC = 0.9435 ] +INFO:root:Average loss=0.1476012277007103 +INFO:root: +[Valid][2][UTT][ MSE = 0.2160 | LCC = 0.9506 | SRCC = 0.9513 ] +INFO:root:Average loss=0.129012701690197 +INFO:root: +[Valid][3][UTT][ MSE = 0.2141 | LCC = 0.9525 | SRCC = 0.9529 ] +INFO:root:Average loss=0.11525272637605667 +INFO:root: +[Valid][4][UTT][ MSE = 0.2194 | LCC = 0.9549 | SRCC = 0.9548 ] +INFO:root:Average loss=0.10601677018404007 +INFO:root: +[Valid][5][UTT][ MSE = 0.2011 | LCC = 0.9578 | SRCC = 0.9583 ] +INFO:root:Average loss=0.09903485164046287 +INFO:root: +[Valid][6][UTT][ MSE = 0.1959 | LCC = 0.9590 | SRCC = 0.9585 ] +INFO:root:Average loss=0.09287466624379158 +INFO:root: +[Valid][7][UTT][ MSE = 0.1695 | LCC = 0.9591 | SRCC = 0.9597 ] +INFO:root:Average loss=0.08375035202503205 +INFO:root: +[Valid][8][UTT][ MSE = 0.1564 | LCC = 0.9602 | SRCC = 0.9604 ] +INFO:root:Average loss=0.07879621079564095 +INFO:root: +[Valid][9][UTT][ MSE = 0.1944 | LCC = 0.9609 | SRCC = 0.9609 ] +INFO:root:Average loss=0.07444575351476669 +INFO:root: +[Valid][10][UTT][ MSE = 0.1809 | LCC = 0.9597 | SRCC = 0.9598 ] +INFO:root:Average loss=0.07514490965008735 +INFO:root: +[Valid][11][UTT][ MSE = 0.1948 | LCC = 0.9613 | SRCC = 0.9612 ] +INFO:root:Average loss=0.0638214111328125 +INFO:root: +[Valid][12][UTT][ MSE = 0.2024 | LCC = 0.9618 | SRCC = 0.9617 ] +INFO:root:Average loss=0.06424891629815102 +INFO:root: +[Valid][13][UTT][ MSE = 0.1626 | LCC = 0.9602 | SRCC = 0.9607 ] +INFO:root:Average loss=0.060301953971385956 +INFO:root: +[Valid][14][UTT][ MSE = 0.1562 | LCC = 0.9604 | SRCC = 0.9609 ] +INFO:root:Average loss=0.05698407934606075 +INFO:root: +[Valid][15][UTT][ MSE = 0.1949 | LCC = 0.9627 | SRCC = 0.9621 ] +INFO:root:Average loss=0.053353609368205074 +INFO:root: +[Valid][16][UTT][ MSE = 0.1880 | LCC = 0.9602 | SRCC = 0.9600 ] +INFO:root:Average loss=0.05258745715022087 +INFO:root: +[Valid][17][UTT][ MSE = 0.1667 | LCC = 0.9629 | SRCC = 0.9629 ] +INFO:root:Average loss=0.05299400274455547 +INFO:root: +[Valid][18][UTT][ MSE = 0.2154 | LCC = 0.9620 | SRCC = 0.9624 ] +INFO:root:Average loss=0.04674607641994953 +INFO:root: +[Valid][19][UTT][ MSE = 0.1735 | LCC = 0.9634 | SRCC = 0.9628 ] +INFO:root:Average loss=0.04557451544702053 +INFO:root: +[Valid][20][UTT][ MSE = 0.1831 | LCC = 0.9633 | SRCC = 0.9628 ] +INFO:root:Average loss=0.04382823812961578 +INFO:root: +[Valid][21][UTT][ MSE = 0.1703 | LCC = 0.9635 | SRCC = 0.9627 ] +INFO:root:Average loss=0.040797299072146416 +INFO:root: +[Valid][22][UTT][ MSE = 0.1625 | LCC = 0.9628 | SRCC = 0.9627 ] +INFO:root:Average loss=0.040290260583162305 +INFO:root: +[Valid][23][UTT][ MSE = 0.1685 | LCC = 0.9618 | SRCC = 0.9614 ] +INFO:root:Average loss=0.040101629719138145 +INFO:root: +[Valid][24][UTT][ MSE = 0.1883 | LCC = 0.9627 | SRCC = 0.9619 ] +INFO:root:Average loss=0.037949363678693775 +INFO:root: +[Valid][25][UTT][ MSE = 0.1712 | LCC = 0.9625 | SRCC = 0.9615 ] +INFO:root:Average loss=0.034947402223944664 +INFO:root: +[Valid][26][UTT][ MSE = 0.1597 | LCC = 0.9618 | SRCC = 0.9616 ] +INFO:root:Average loss=0.033867306530475615 +INFO:root: +[Valid][27][UTT][ MSE = 0.1929 | LCC = 0.9627 | SRCC = 0.9617 ] +INFO:root:Average loss=0.03201555570960045 +INFO:root: +[Valid][28][UTT][ MSE = 0.1508 | LCC = 0.9622 | SRCC = 0.9619 ] +INFO:root:Average loss=0.03369861473143101 +INFO:root: +[Valid][29][UTT][ MSE = 0.1673 | LCC = 0.9634 | SRCC = 0.9624 ] +INFO:root:Average loss=0.032580247417092326 +INFO:root: +[Valid][30][UTT][ MSE = 0.1701 | LCC = 0.9631 | SRCC = 0.9623 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1754 | LCC = 0.9639 | SRCC = 0.9642 ] diff --git a/results_tencent/hubert_base/layer4/run4/config.gin b/results_tencent/hubert_base/layer4/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer4/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer4/run4/model.pt b/results_tencent/hubert_base/layer4/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..df160b82e543240c64d0145be162179cad8e5e40 --- /dev/null +++ b/results_tencent/hubert_base/layer4/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d750b1338960daf96e5249d2a182270eafbbe2cf2ab6874d63f430eefd1ebc9e +size 2747979 diff --git a/results_tencent/hubert_base/layer4/run4/model_best.pt b/results_tencent/hubert_base/layer4/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..3af1cead638abaa5623a157179b8f2b8127cca0a --- /dev/null +++ b/results_tencent/hubert_base/layer4/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:68a4b8768aa48a21cceb33d42a522e5179cfbdab1a5d75e2ae234858c6a7a0fb +size 2749638 diff --git a/results_tencent/hubert_base/layer4/run4/train.log b/results_tencent/hubert_base/layer4/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..220807fb1250af8e2778490cc1b8dd5dfedc874e --- /dev/null +++ b/results_tencent/hubert_base/layer4/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4231206508874893 +INFO:root: +[Valid][0][UTT][ MSE = 0.2700 | LCC = 0.9236 | SRCC = 0.9223 ] +INFO:root:Average loss=0.1873604558110237 +INFO:root: +[Valid][1][UTT][ MSE = 0.2261 | LCC = 0.9445 | SRCC = 0.9434 ] +INFO:root:Average loss=0.14630690729618073 +INFO:root: +[Valid][2][UTT][ MSE = 0.2074 | LCC = 0.9504 | SRCC = 0.9494 ] +INFO:root:Average loss=0.12512446576356887 +INFO:root: +[Valid][3][UTT][ MSE = 0.2110 | LCC = 0.9547 | SRCC = 0.9545 ] +INFO:root:Average loss=0.11086751115322113 +INFO:root: +[Valid][4][UTT][ MSE = 0.2034 | LCC = 0.9578 | SRCC = 0.9579 ] +INFO:root:Average loss=0.10226034927368165 +INFO:root: +[Valid][5][UTT][ MSE = 0.1989 | LCC = 0.9593 | SRCC = 0.9594 ] +INFO:root:Average loss=0.09708944684267044 +INFO:root: +[Valid][6][UTT][ MSE = 0.1803 | LCC = 0.9599 | SRCC = 0.9600 ] +INFO:root:Average loss=0.08576299405097962 +INFO:root: +[Valid][7][UTT][ MSE = 0.1588 | LCC = 0.9612 | SRCC = 0.9610 ] +INFO:root:Average loss=0.08215700510144233 +INFO:root: +[Valid][8][UTT][ MSE = 0.1715 | LCC = 0.9613 | SRCC = 0.9613 ] +INFO:root:Average loss=0.07669090881943702 +INFO:root: +[Valid][9][UTT][ MSE = 0.1798 | LCC = 0.9613 | SRCC = 0.9612 ] +INFO:root:Average loss=0.07130826711654663 +INFO:root: +[Valid][10][UTT][ MSE = 0.1576 | LCC = 0.9609 | SRCC = 0.9606 ] +INFO:root:Average loss=0.07080629828572273 +INFO:root: +[Valid][11][UTT][ MSE = 0.1713 | LCC = 0.9614 | SRCC = 0.9609 ] +INFO:root:Average loss=0.06348666182160377 +INFO:root: +[Valid][12][UTT][ MSE = 0.1769 | LCC = 0.9615 | SRCC = 0.9611 ] +INFO:root:Average loss=0.06148102152347565 +INFO:root: +[Valid][13][UTT][ MSE = 0.1813 | LCC = 0.9624 | SRCC = 0.9615 ] +INFO:root:Average loss=0.056300003230571743 +INFO:root: +[Valid][14][UTT][ MSE = 0.1847 | LCC = 0.9626 | SRCC = 0.9621 ] +INFO:root:Average loss=0.0544388337135315 +INFO:root: +[Valid][15][UTT][ MSE = 0.1692 | LCC = 0.9618 | SRCC = 0.9610 ] +INFO:root:Average loss=0.050749091237783435 +INFO:root: +[Valid][16][UTT][ MSE = 0.1722 | LCC = 0.9609 | SRCC = 0.9602 ] +INFO:root:Average loss=0.049141381353139876 +INFO:root: +[Valid][17][UTT][ MSE = 0.1528 | LCC = 0.9624 | SRCC = 0.9624 ] +INFO:root:Average loss=0.05031431767344475 +INFO:root: +[Valid][18][UTT][ MSE = 0.1539 | LCC = 0.9621 | SRCC = 0.9618 ] +INFO:root:Average loss=0.0469830205142498 +INFO:root: +[Valid][19][UTT][ MSE = 0.1581 | LCC = 0.9623 | SRCC = 0.9618 ] +INFO:root:Average loss=0.04456768381595612 +INFO:root: +[Valid][20][UTT][ MSE = 0.1813 | LCC = 0.9591 | SRCC = 0.9601 ] +INFO:root:Average loss=0.04225409373641014 +INFO:root: +[Valid][21][UTT][ MSE = 0.1595 | LCC = 0.9625 | SRCC = 0.9620 ] +INFO:root:Average loss=0.041059058561921116 +INFO:root: +[Valid][22][UTT][ MSE = 0.1494 | LCC = 0.9626 | SRCC = 0.9623 ] +INFO:root:Average loss=0.04062368324398995 +INFO:root: +[Valid][23][UTT][ MSE = 0.1584 | LCC = 0.9634 | SRCC = 0.9626 ] +INFO:root:Average loss=0.037790130510926244 +INFO:root: +[Valid][24][UTT][ MSE = 0.1613 | LCC = 0.9625 | SRCC = 0.9620 ] +INFO:root:Average loss=0.036218619644641875 +INFO:root: +[Valid][25][UTT][ MSE = 0.1662 | LCC = 0.9617 | SRCC = 0.9613 ] +INFO:root:Average loss=0.03467911140620709 +INFO:root: +[Valid][26][UTT][ MSE = 0.1522 | LCC = 0.9625 | SRCC = 0.9617 ] +INFO:root:Average loss=0.03379358433187008 +INFO:root: +[Valid][27][UTT][ MSE = 0.1568 | LCC = 0.9633 | SRCC = 0.9626 ] +INFO:root:Average loss=0.03456975008547306 +INFO:root: +[Valid][28][UTT][ MSE = 0.1692 | LCC = 0.9626 | SRCC = 0.9619 ] +INFO:root:Average loss=0.030540536373853684 +INFO:root: +[Valid][29][UTT][ MSE = 0.1527 | LCC = 0.9631 | SRCC = 0.9629 ] +INFO:root:Average loss=0.03388060097396374 +INFO:root: +[Valid][30][UTT][ MSE = 0.1577 | LCC = 0.9622 | SRCC = 0.9617 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1579 | LCC = 0.9651 | SRCC = 0.9653 ] diff --git a/results_tencent/hubert_base/layer5/run0/config.gin b/results_tencent/hubert_base/layer5/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer5/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer5/run0/model.pt b/results_tencent/hubert_base/layer5/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..38e7a9b6c8fb96222d6f29180811060edb4e044b --- /dev/null +++ b/results_tencent/hubert_base/layer5/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:db67133686c06c5fbad2a8eee70b613c78efff291e00cfca2f58bf9d8a25e298 +size 2747979 diff --git a/results_tencent/hubert_base/layer5/run0/model_best.pt b/results_tencent/hubert_base/layer5/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2127b4363cf6edbd49abbc9fa49ca526dbcbc894 --- /dev/null +++ b/results_tencent/hubert_base/layer5/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d032228cee1b180d603c6623adb2dfbd756a037b7b54881a86cb500920628a5c +size 2749702 diff --git a/results_tencent/hubert_base/layer5/run0/train.log b/results_tencent/hubert_base/layer5/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e394d1ebd949c831cbde3a4c21dd6cf879332048 --- /dev/null +++ b/results_tencent/hubert_base/layer5/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.509871609210968 +INFO:root: +[Valid][0][UTT][ MSE = 0.2946 | LCC = 0.9260 | SRCC = 0.9235 ] +INFO:root:Average loss=0.17839222133159638 +INFO:root: +[Valid][1][UTT][ MSE = 0.2250 | LCC = 0.9480 | SRCC = 0.9470 ] +INFO:root:Average loss=0.13951319366693496 +INFO:root: +[Valid][2][UTT][ MSE = 0.1932 | LCC = 0.9529 | SRCC = 0.9524 ] +INFO:root:Average loss=0.12330902749300003 +INFO:root: +[Valid][3][UTT][ MSE = 0.2062 | LCC = 0.9554 | SRCC = 0.9547 ] +INFO:root:Average loss=0.10927059602737427 +INFO:root: +[Valid][4][UTT][ MSE = 0.1682 | LCC = 0.9570 | SRCC = 0.9566 ] +INFO:root:Average loss=0.10175281023979187 +INFO:root: +[Valid][5][UTT][ MSE = 0.1955 | LCC = 0.9590 | SRCC = 0.9588 ] +INFO:root:Average loss=0.09398409953713417 +INFO:root: +[Valid][6][UTT][ MSE = 0.1893 | LCC = 0.9590 | SRCC = 0.9589 ] +INFO:root:Average loss=0.08592596876621246 +INFO:root: +[Valid][7][UTT][ MSE = 0.1933 | LCC = 0.9593 | SRCC = 0.9586 ] +INFO:root:Average loss=0.08179879042506218 +INFO:root: +[Valid][8][UTT][ MSE = 0.1892 | LCC = 0.9604 | SRCC = 0.9600 ] +INFO:root:Average loss=0.07720961457490921 +INFO:root: +[Valid][9][UTT][ MSE = 0.1961 | LCC = 0.9604 | SRCC = 0.9601 ] +INFO:root:Average loss=0.07389661473035812 +INFO:root: +[Valid][10][UTT][ MSE = 0.1846 | LCC = 0.9605 | SRCC = 0.9602 ] +INFO:root:Average loss=0.06821288815140725 +INFO:root: +[Valid][11][UTT][ MSE = 0.1817 | LCC = 0.9611 | SRCC = 0.9604 ] +INFO:root:Average loss=0.06425552794337272 +INFO:root: +[Valid][12][UTT][ MSE = 0.1631 | LCC = 0.9618 | SRCC = 0.9616 ] +INFO:root:Average loss=0.06127037313580513 +INFO:root: +[Valid][13][UTT][ MSE = 0.1900 | LCC = 0.9610 | SRCC = 0.9603 ] +INFO:root:Average loss=0.05727958863973617 +INFO:root: +[Valid][14][UTT][ MSE = 0.1928 | LCC = 0.9616 | SRCC = 0.9611 ] +INFO:root:Average loss=0.05375721390545368 +INFO:root: +[Valid][15][UTT][ MSE = 0.1795 | LCC = 0.9611 | SRCC = 0.9603 ] +INFO:root:Average loss=0.05360283032059669 +INFO:root: +[Valid][16][UTT][ MSE = 0.1750 | LCC = 0.9614 | SRCC = 0.9606 ] +INFO:root:Average loss=0.05083597853779793 +INFO:root: +[Valid][17][UTT][ MSE = 0.1804 | LCC = 0.9609 | SRCC = 0.9600 ] +INFO:root:Average loss=0.04944618593156338 +INFO:root: +[Valid][18][UTT][ MSE = 0.1557 | LCC = 0.9607 | SRCC = 0.9601 ] +INFO:root:Average loss=0.04685955688357353 +INFO:root: +[Valid][19][UTT][ MSE = 0.1642 | LCC = 0.9611 | SRCC = 0.9601 ] +INFO:root:Average loss=0.04339661008119583 +INFO:root: +[Valid][20][UTT][ MSE = 0.1665 | LCC = 0.9614 | SRCC = 0.9609 ] +INFO:root:Average loss=0.0437190684825182 +INFO:root: +[Valid][21][UTT][ MSE = 0.2013 | LCC = 0.9606 | SRCC = 0.9602 ] +INFO:root:Average loss=0.041316692784428594 +INFO:root: +[Valid][22][UTT][ MSE = 0.1764 | LCC = 0.9611 | SRCC = 0.9605 ] +INFO:root:Average loss=0.03744974264502526 +INFO:root: +[Valid][23][UTT][ MSE = 0.1652 | LCC = 0.9615 | SRCC = 0.9605 ] +INFO:root:Average loss=0.037872024461627005 +INFO:root: +[Valid][24][UTT][ MSE = 0.1745 | LCC = 0.9612 | SRCC = 0.9604 ] +INFO:root:Average loss=0.03666692753136158 +INFO:root: +[Valid][25][UTT][ MSE = 0.2224 | LCC = 0.9584 | SRCC = 0.9600 ] +INFO:root:Average loss=0.033550654053688046 +INFO:root: +[Valid][26][UTT][ MSE = 0.1842 | LCC = 0.9610 | SRCC = 0.9608 ] +INFO:root:Average loss=0.032786446258425715 +INFO:root: +[Valid][27][UTT][ MSE = 0.1696 | LCC = 0.9613 | SRCC = 0.9605 ] +INFO:root:Average loss=0.0310258277207613 +INFO:root: +[Valid][28][UTT][ MSE = 0.1692 | LCC = 0.9608 | SRCC = 0.9595 ] +INFO:root:Average loss=0.03211408606171608 +INFO:root: +[Valid][29][UTT][ MSE = 0.1635 | LCC = 0.9609 | SRCC = 0.9595 ] +INFO:root:Average loss=0.029530202552676202 +INFO:root: +[Valid][30][UTT][ MSE = 0.1799 | LCC = 0.9614 | SRCC = 0.9602 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1713 | LCC = 0.9609 | SRCC = 0.9614 ] diff --git a/results_tencent/hubert_base/layer5/run1/config.gin b/results_tencent/hubert_base/layer5/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer5/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer5/run1/model.pt b/results_tencent/hubert_base/layer5/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a553182931d50692dd3899f61cfb2d8790c4d366 --- /dev/null +++ b/results_tencent/hubert_base/layer5/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4d0ef3afe630a483b3f2b6e0fa71ae16d937f71c589ccfa4dcdf8e1697450e4b +size 2747979 diff --git a/results_tencent/hubert_base/layer5/run1/model_best.pt b/results_tencent/hubert_base/layer5/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..513706ff738774c7a8c7dfdbaf4c6ef5681ed110 --- /dev/null +++ b/results_tencent/hubert_base/layer5/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be3c00bf8079c4660e5f84a864c8a7a6d248c5e711d9031371756f0848572904 +size 2749702 diff --git a/results_tencent/hubert_base/layer5/run1/train.log b/results_tencent/hubert_base/layer5/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..55de4ac188f2e9c2c6784a02bff78d0074b113af --- /dev/null +++ b/results_tencent/hubert_base/layer5/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4485912972688675 +INFO:root: +[Valid][0][UTT][ MSE = 0.2821 | LCC = 0.9280 | SRCC = 0.9262 ] +INFO:root:Average loss=0.18385643416643144 +INFO:root: +[Valid][1][UTT][ MSE = 0.2263 | LCC = 0.9450 | SRCC = 0.9430 ] +INFO:root:Average loss=0.14546715712547303 +INFO:root: +[Valid][2][UTT][ MSE = 0.2162 | LCC = 0.9490 | SRCC = 0.9477 ] +INFO:root:Average loss=0.128211097240448 +INFO:root: +[Valid][3][UTT][ MSE = 0.2271 | LCC = 0.9513 | SRCC = 0.9498 ] +INFO:root:Average loss=0.11424678936600685 +INFO:root: +[Valid][4][UTT][ MSE = 0.1871 | LCC = 0.9559 | SRCC = 0.9553 ] +INFO:root:Average loss=0.1077485965192318 +INFO:root: +[Valid][5][UTT][ MSE = 0.1991 | LCC = 0.9563 | SRCC = 0.9556 ] +INFO:root:Average loss=0.0988824163377285 +INFO:root: +[Valid][6][UTT][ MSE = 0.2076 | LCC = 0.9575 | SRCC = 0.9569 ] +INFO:root:Average loss=0.08992268896102905 +INFO:root: +[Valid][7][UTT][ MSE = 0.2025 | LCC = 0.9572 | SRCC = 0.9563 ] +INFO:root:Average loss=0.08608752313256264 +INFO:root: +[Valid][8][UTT][ MSE = 0.2205 | LCC = 0.9596 | SRCC = 0.9593 ] +INFO:root:Average loss=0.07951096522808075 +INFO:root: +[Valid][9][UTT][ MSE = 0.1917 | LCC = 0.9578 | SRCC = 0.9571 ] +INFO:root:Average loss=0.07338705691695213 +INFO:root: +[Valid][10][UTT][ MSE = 0.1596 | LCC = 0.9595 | SRCC = 0.9595 ] +INFO:root:Average loss=0.06975187465548516 +INFO:root: +[Valid][11][UTT][ MSE = 0.1824 | LCC = 0.9601 | SRCC = 0.9593 ] +INFO:root:Average loss=0.06577360579371452 +INFO:root: +[Valid][12][UTT][ MSE = 0.1613 | LCC = 0.9596 | SRCC = 0.9597 ] +INFO:root:Average loss=0.06473290187120438 +INFO:root: +[Valid][13][UTT][ MSE = 0.2088 | LCC = 0.9608 | SRCC = 0.9605 ] +INFO:root:Average loss=0.06011417979001999 +INFO:root: +[Valid][14][UTT][ MSE = 0.1673 | LCC = 0.9607 | SRCC = 0.9606 ] +INFO:root:Average loss=0.05521336568892002 +INFO:root: +[Valid][15][UTT][ MSE = 0.1820 | LCC = 0.9588 | SRCC = 0.9580 ] +INFO:root:Average loss=0.05224400761723518 +INFO:root: +[Valid][16][UTT][ MSE = 0.1741 | LCC = 0.9615 | SRCC = 0.9608 ] +INFO:root:Average loss=0.04970773628354073 +INFO:root: +[Valid][17][UTT][ MSE = 0.1725 | LCC = 0.9616 | SRCC = 0.9610 ] +INFO:root:Average loss=0.04900969077646732 +INFO:root: +[Valid][18][UTT][ MSE = 0.1653 | LCC = 0.9610 | SRCC = 0.9608 ] +INFO:root:Average loss=0.04554431425035 +INFO:root: +[Valid][19][UTT][ MSE = 0.1803 | LCC = 0.9599 | SRCC = 0.9600 ] +INFO:root:Average loss=0.04634648248553276 +INFO:root: +[Valid][20][UTT][ MSE = 0.1741 | LCC = 0.9608 | SRCC = 0.9605 ] +INFO:root:Average loss=0.04277595445513725 +INFO:root: +[Valid][21][UTT][ MSE = 0.1693 | LCC = 0.9600 | SRCC = 0.9590 ] +INFO:root:Average loss=0.042401527494192126 +INFO:root: +[Valid][22][UTT][ MSE = 0.1857 | LCC = 0.9611 | SRCC = 0.9600 ] +INFO:root:Average loss=0.040588512673974037 +INFO:root: +[Valid][23][UTT][ MSE = 0.1740 | LCC = 0.9608 | SRCC = 0.9599 ] +INFO:root:Average loss=0.039168834745883944 +INFO:root: +[Valid][24][UTT][ MSE = 0.1887 | LCC = 0.9591 | SRCC = 0.9581 ] +INFO:root:Average loss=0.03717134357988834 +INFO:root: +[Valid][25][UTT][ MSE = 0.1711 | LCC = 0.9604 | SRCC = 0.9596 ] +INFO:root:Average loss=0.036077167302370074 +INFO:root: +[Valid][26][UTT][ MSE = 0.1827 | LCC = 0.9618 | SRCC = 0.9610 ] +INFO:root:Average loss=0.033301551312208175 +INFO:root: +[Valid][27][UTT][ MSE = 0.1701 | LCC = 0.9616 | SRCC = 0.9609 ] +INFO:root:Average loss=0.03407234816253185 +INFO:root: +[Valid][28][UTT][ MSE = 0.1701 | LCC = 0.9614 | SRCC = 0.9607 ] +INFO:root:Average loss=0.03406850910186768 +INFO:root: +[Valid][29][UTT][ MSE = 0.1913 | LCC = 0.9599 | SRCC = 0.9591 ] +INFO:root:Average loss=0.030076176837086677 +INFO:root: +[Valid][30][UTT][ MSE = 0.1648 | LCC = 0.9615 | SRCC = 0.9606 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1903 | LCC = 0.9602 | SRCC = 0.9608 ] diff --git a/results_tencent/hubert_base/layer5/run2/config.gin b/results_tencent/hubert_base/layer5/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer5/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer5/run2/model.pt b/results_tencent/hubert_base/layer5/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a527d82bcac14a5fdcfe5267428ae902fb7b4199 --- /dev/null +++ b/results_tencent/hubert_base/layer5/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1262d494f321c5b50f690458620ce480e9832e29253d036b8672d09fc3ce9461 +size 2747979 diff --git a/results_tencent/hubert_base/layer5/run2/model_best.pt b/results_tencent/hubert_base/layer5/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a0d30eb84c1f92e93997fcc2e340fb886862c350 --- /dev/null +++ b/results_tencent/hubert_base/layer5/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6c3eee747a8c205b163eb4452bbd2105d597ee426445187e7a8551ff131ee3eb +size 2749702 diff --git a/results_tencent/hubert_base/layer5/run2/train.log b/results_tencent/hubert_base/layer5/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..647d027274966fe355aecce3149575289e57c1e2 --- /dev/null +++ b/results_tencent/hubert_base/layer5/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5340962796211243 +INFO:root: +[Valid][0][UTT][ MSE = 0.3195 | LCC = 0.9308 | SRCC = 0.9277 ] +INFO:root:Average loss=0.18154503667354585 +INFO:root: +[Valid][1][UTT][ MSE = 0.2656 | LCC = 0.9422 | SRCC = 0.9392 ] +INFO:root:Average loss=0.1449539442062378 +INFO:root: +[Valid][2][UTT][ MSE = 0.2026 | LCC = 0.9515 | SRCC = 0.9513 ] +INFO:root:Average loss=0.12917825067043304 +INFO:root: +[Valid][3][UTT][ MSE = 0.2204 | LCC = 0.9543 | SRCC = 0.9538 ] +INFO:root:Average loss=0.11683966714143754 +INFO:root: +[Valid][4][UTT][ MSE = 0.2065 | LCC = 0.9574 | SRCC = 0.9565 ] +INFO:root:Average loss=0.1063873569369316 +INFO:root: +[Valid][5][UTT][ MSE = 0.2155 | LCC = 0.9578 | SRCC = 0.9578 ] +INFO:root:Average loss=0.09734632813930512 +INFO:root: +[Valid][6][UTT][ MSE = 0.2446 | LCC = 0.9568 | SRCC = 0.9578 ] +INFO:root:Average loss=0.08912296891212464 +INFO:root: +[Valid][7][UTT][ MSE = 0.2054 | LCC = 0.9591 | SRCC = 0.9592 ] +INFO:root:Average loss=0.08710124406218529 +INFO:root: +[Valid][8][UTT][ MSE = 0.2206 | LCC = 0.9593 | SRCC = 0.9596 ] +INFO:root:Average loss=0.07703453764319419 +INFO:root: +[Valid][9][UTT][ MSE = 0.1998 | LCC = 0.9600 | SRCC = 0.9595 ] +INFO:root:Average loss=0.07147296983003616 +INFO:root: +[Valid][10][UTT][ MSE = 0.2122 | LCC = 0.9605 | SRCC = 0.9608 ] +INFO:root:Average loss=0.06867362049221992 +INFO:root: +[Valid][11][UTT][ MSE = 0.1918 | LCC = 0.9606 | SRCC = 0.9603 ] +INFO:root:Average loss=0.0661231769323349 +INFO:root: +[Valid][12][UTT][ MSE = 0.2198 | LCC = 0.9596 | SRCC = 0.9602 ] +INFO:root:Average loss=0.06014017602801323 +INFO:root: +[Valid][13][UTT][ MSE = 0.2083 | LCC = 0.9608 | SRCC = 0.9616 ] +INFO:root:Average loss=0.059064438670873645 +INFO:root: +[Valid][14][UTT][ MSE = 0.1789 | LCC = 0.9606 | SRCC = 0.9597 ] +INFO:root:Average loss=0.055299815952777866 +INFO:root: +[Valid][15][UTT][ MSE = 0.1894 | LCC = 0.9615 | SRCC = 0.9613 ] +INFO:root:Average loss=0.054188394337892534 +INFO:root: +[Valid][16][UTT][ MSE = 0.1957 | LCC = 0.9619 | SRCC = 0.9617 ] +INFO:root:Average loss=0.0503035536557436 +INFO:root: +[Valid][17][UTT][ MSE = 0.1847 | LCC = 0.9625 | SRCC = 0.9621 ] +INFO:root:Average loss=0.04701666863262653 +INFO:root: +[Valid][18][UTT][ MSE = 0.1842 | LCC = 0.9621 | SRCC = 0.9612 ] +INFO:root:Average loss=0.04411008244752884 +INFO:root: +[Valid][19][UTT][ MSE = 0.1756 | LCC = 0.9607 | SRCC = 0.9602 ] +INFO:root:Average loss=0.04376040953397751 +INFO:root: +[Valid][20][UTT][ MSE = 0.2026 | LCC = 0.9607 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04142147873342037 +INFO:root: +[Valid][21][UTT][ MSE = 0.1752 | LCC = 0.9610 | SRCC = 0.9603 ] +INFO:root:Average loss=0.03836024996638298 +INFO:root: +[Valid][22][UTT][ MSE = 0.1602 | LCC = 0.9611 | SRCC = 0.9602 ] +INFO:root:Average loss=0.03753511606156826 +INFO:root: +[Valid][23][UTT][ MSE = 0.1799 | LCC = 0.9627 | SRCC = 0.9619 ] +INFO:root:Average loss=0.036446436583995816 +INFO:root: +[Valid][24][UTT][ MSE = 0.1678 | LCC = 0.9616 | SRCC = 0.9606 ] +INFO:root:Average loss=0.0363024819791317 +INFO:root: +[Valid][25][UTT][ MSE = 0.1751 | LCC = 0.9617 | SRCC = 0.9606 ] +INFO:root:Average loss=0.032234612718224524 +INFO:root: +[Valid][26][UTT][ MSE = 0.1759 | LCC = 0.9622 | SRCC = 0.9611 ] +INFO:root:Average loss=0.03274307750165462 +INFO:root: +[Valid][27][UTT][ MSE = 0.1948 | LCC = 0.9622 | SRCC = 0.9615 ] +INFO:root:Average loss=0.03142600749433041 +INFO:root: +[Valid][28][UTT][ MSE = 0.1797 | LCC = 0.9614 | SRCC = 0.9604 ] +INFO:root:Average loss=0.029825103282928467 +INFO:root: +[Valid][29][UTT][ MSE = 0.1974 | LCC = 0.9622 | SRCC = 0.9615 ] +INFO:root:Average loss=0.030201896876096725 +INFO:root: +[Valid][30][UTT][ MSE = 0.1854 | LCC = 0.9620 | SRCC = 0.9610 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1919 | LCC = 0.9603 | SRCC = 0.9608 ] diff --git a/results_tencent/hubert_base/layer5/run3/config.gin b/results_tencent/hubert_base/layer5/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer5/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer5/run3/model.pt b/results_tencent/hubert_base/layer5/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..32929f70da5f62d51d83996ea2e574dfeb79cd9d --- /dev/null +++ b/results_tencent/hubert_base/layer5/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4e8c24368de5f1de0d45bd9a53f22e11783c99aa25c64a0a5b35c3e6e597a75f +size 2747979 diff --git a/results_tencent/hubert_base/layer5/run3/model_best.pt b/results_tencent/hubert_base/layer5/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..034c91054ec45256abf1a33e43ddd0a32d08972f --- /dev/null +++ b/results_tencent/hubert_base/layer5/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c6be1a0fa84565f105d7685f765cb57cf73260ac0fe7411d5e2a03ca0c8bf180 +size 2749702 diff --git a/results_tencent/hubert_base/layer5/run3/train.log b/results_tencent/hubert_base/layer5/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f9dcff3b71c6a44209b6afed489aa840e32a495c --- /dev/null +++ b/results_tencent/hubert_base/layer5/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4258647834062576 +INFO:root: +[Valid][0][UTT][ MSE = 0.2839 | LCC = 0.9219 | SRCC = 0.9187 ] +INFO:root:Average loss=0.19219473093748093 +INFO:root: +[Valid][1][UTT][ MSE = 0.2342 | LCC = 0.9411 | SRCC = 0.9403 ] +INFO:root:Average loss=0.15391053366661073 +INFO:root: +[Valid][2][UTT][ MSE = 0.2250 | LCC = 0.9477 | SRCC = 0.9482 ] +INFO:root:Average loss=0.13024138921499254 +INFO:root: +[Valid][3][UTT][ MSE = 0.2151 | LCC = 0.9524 | SRCC = 0.9524 ] +INFO:root:Average loss=0.11660204619169236 +INFO:root: +[Valid][4][UTT][ MSE = 0.2066 | LCC = 0.9552 | SRCC = 0.9548 ] +INFO:root:Average loss=0.10466550648212433 +INFO:root: +[Valid][5][UTT][ MSE = 0.1827 | LCC = 0.9533 | SRCC = 0.9534 ] +INFO:root:Average loss=0.09522219270467758 +INFO:root: +[Valid][6][UTT][ MSE = 0.1834 | LCC = 0.9553 | SRCC = 0.9545 ] +INFO:root:Average loss=0.09436587470769882 +INFO:root: +[Valid][7][UTT][ MSE = 0.1765 | LCC = 0.9509 | SRCC = 0.9521 ] +INFO:root:Average loss=0.08626363936066628 +INFO:root: +[Valid][8][UTT][ MSE = 0.1691 | LCC = 0.9563 | SRCC = 0.9558 ] +INFO:root:Average loss=0.08010093629360199 +INFO:root: +[Valid][9][UTT][ MSE = 0.1804 | LCC = 0.9543 | SRCC = 0.9555 ] +INFO:root:Average loss=0.07669223803281784 +INFO:root: +[Valid][10][UTT][ MSE = 0.1508 | LCC = 0.9567 | SRCC = 0.9578 ] +INFO:root:Average loss=0.0685383132994175 +INFO:root: +[Valid][11][UTT][ MSE = 0.1614 | LCC = 0.9586 | SRCC = 0.9588 ] +INFO:root:Average loss=0.06738250184059143 +INFO:root: +[Valid][12][UTT][ MSE = 0.1597 | LCC = 0.9585 | SRCC = 0.9589 ] +INFO:root:Average loss=0.06330720314383506 +INFO:root: +[Valid][13][UTT][ MSE = 0.1759 | LCC = 0.9575 | SRCC = 0.9577 ] +INFO:root:Average loss=0.0582011576294899 +INFO:root: +[Valid][14][UTT][ MSE = 0.1573 | LCC = 0.9570 | SRCC = 0.9578 ] +INFO:root:Average loss=0.05550464406609535 +INFO:root: +[Valid][15][UTT][ MSE = 0.1774 | LCC = 0.9600 | SRCC = 0.9598 ] +INFO:root:Average loss=0.0553211199939251 +INFO:root: +[Valid][16][UTT][ MSE = 0.1994 | LCC = 0.9580 | SRCC = 0.9577 ] +INFO:root:Average loss=0.05223307606577873 +INFO:root: +[Valid][17][UTT][ MSE = 0.1679 | LCC = 0.9572 | SRCC = 0.9576 ] +INFO:root:Average loss=0.048624135598540306 +INFO:root: +[Valid][18][UTT][ MSE = 0.1634 | LCC = 0.9603 | SRCC = 0.9598 ] +INFO:root:Average loss=0.0455592223405838 +INFO:root: +[Valid][19][UTT][ MSE = 0.1716 | LCC = 0.9608 | SRCC = 0.9604 ] +INFO:root:Average loss=0.04710148233175278 +INFO:root: +[Valid][20][UTT][ MSE = 0.1615 | LCC = 0.9578 | SRCC = 0.9592 ] +INFO:root:Average loss=0.04371048204600811 +INFO:root: +[Valid][21][UTT][ MSE = 0.1577 | LCC = 0.9589 | SRCC = 0.9586 ] +INFO:root:Average loss=0.040241401448845866 +INFO:root: +[Valid][22][UTT][ MSE = 0.1686 | LCC = 0.9596 | SRCC = 0.9587 ] +INFO:root:Average loss=0.03926335689425468 +INFO:root: +[Valid][23][UTT][ MSE = 0.1599 | LCC = 0.9571 | SRCC = 0.9572 ] +INFO:root:Average loss=0.03711808468401432 +INFO:root: +[Valid][24][UTT][ MSE = 0.1708 | LCC = 0.9601 | SRCC = 0.9597 ] +INFO:root:Average loss=0.03830269975960255 +INFO:root: +[Valid][25][UTT][ MSE = 0.1716 | LCC = 0.9596 | SRCC = 0.9585 ] +INFO:root:Average loss=0.03531160207092762 +INFO:root: +[Valid][26][UTT][ MSE = 0.1621 | LCC = 0.9597 | SRCC = 0.9594 ] +INFO:root:Average loss=0.03396486204862595 +INFO:root: +[Valid][27][UTT][ MSE = 0.1631 | LCC = 0.9576 | SRCC = 0.9579 ] +INFO:root:Average loss=0.03429593308269978 +INFO:root: +[Valid][28][UTT][ MSE = 0.1773 | LCC = 0.9593 | SRCC = 0.9593 ] +INFO:root:Average loss=0.0324753201007843 +INFO:root: +[Valid][29][UTT][ MSE = 0.1659 | LCC = 0.9572 | SRCC = 0.9580 ] +INFO:root:Average loss=0.033033864825963974 +INFO:root: +[Valid][30][UTT][ MSE = 0.1658 | LCC = 0.9604 | SRCC = 0.9597 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1802 | LCC = 0.9600 | SRCC = 0.9604 ] diff --git a/results_tencent/hubert_base/layer5/run4/config.gin b/results_tencent/hubert_base/layer5/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer5/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer5/run4/model.pt b/results_tencent/hubert_base/layer5/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..de973fffbaa68184aedf2e65d8ef61cfbdb1a7cc --- /dev/null +++ b/results_tencent/hubert_base/layer5/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7bb0e742fd88269c090319bfacf25dcc2a6db8c6cfc9e67fac1d9d067313822e +size 2747979 diff --git a/results_tencent/hubert_base/layer5/run4/model_best.pt b/results_tencent/hubert_base/layer5/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..60e302149081467741886ff2b6fc5d048066b74d --- /dev/null +++ b/results_tencent/hubert_base/layer5/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:506a1b4682972103e9ae09ef95a3994c4965217d709f649a10c0f0ea92b779b9 +size 2749702 diff --git a/results_tencent/hubert_base/layer5/run4/train.log b/results_tencent/hubert_base/layer5/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0f97a2d71d9c659b05a53787bc106876925a46d4 --- /dev/null +++ b/results_tencent/hubert_base/layer5/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5370765228271485 +INFO:root: +[Valid][0][UTT][ MSE = 0.3321 | LCC = 0.9151 | SRCC = 0.9118 ] +INFO:root:Average loss=0.1923546217083931 +INFO:root: +[Valid][1][UTT][ MSE = 0.2317 | LCC = 0.9417 | SRCC = 0.9393 ] +INFO:root:Average loss=0.15295204663276674 +INFO:root: +[Valid][2][UTT][ MSE = 0.2365 | LCC = 0.9484 | SRCC = 0.9464 ] +INFO:root:Average loss=0.13356065195798875 +INFO:root: +[Valid][3][UTT][ MSE = 0.2033 | LCC = 0.9523 | SRCC = 0.9507 ] +INFO:root:Average loss=0.11815218222141266 +INFO:root: +[Valid][4][UTT][ MSE = 0.2046 | LCC = 0.9539 | SRCC = 0.9522 ] +INFO:root:Average loss=0.10718410271406174 +INFO:root: +[Valid][5][UTT][ MSE = 0.1783 | LCC = 0.9555 | SRCC = 0.9547 ] +INFO:root:Average loss=0.09637917339801788 +INFO:root: +[Valid][6][UTT][ MSE = 0.1929 | LCC = 0.9557 | SRCC = 0.9544 ] +INFO:root:Average loss=0.0902817822098732 +INFO:root: +[Valid][7][UTT][ MSE = 0.1979 | LCC = 0.9588 | SRCC = 0.9575 ] +INFO:root:Average loss=0.08579420790076256 +INFO:root: +[Valid][8][UTT][ MSE = 0.2054 | LCC = 0.9580 | SRCC = 0.9565 ] +INFO:root:Average loss=0.07984266909956932 +INFO:root: +[Valid][9][UTT][ MSE = 0.1855 | LCC = 0.9576 | SRCC = 0.9561 ] +INFO:root:Average loss=0.07675537443161011 +INFO:root: +[Valid][10][UTT][ MSE = 0.2074 | LCC = 0.9584 | SRCC = 0.9567 ] +INFO:root:Average loss=0.07353569087386132 +INFO:root: +[Valid][11][UTT][ MSE = 0.1952 | LCC = 0.9598 | SRCC = 0.9591 ] +INFO:root:Average loss=0.06693949377536773 +INFO:root: +[Valid][12][UTT][ MSE = 0.1671 | LCC = 0.9593 | SRCC = 0.9584 ] +INFO:root:Average loss=0.065351326495409 +INFO:root: +[Valid][13][UTT][ MSE = 0.1959 | LCC = 0.9600 | SRCC = 0.9589 ] +INFO:root:Average loss=0.0597770706564188 +INFO:root: +[Valid][14][UTT][ MSE = 0.1958 | LCC = 0.9599 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05761149644851685 +INFO:root: +[Valid][15][UTT][ MSE = 0.1928 | LCC = 0.9602 | SRCC = 0.9595 ] +INFO:root:Average loss=0.05435423913598061 +INFO:root: +[Valid][16][UTT][ MSE = 0.2040 | LCC = 0.9602 | SRCC = 0.9596 ] +INFO:root:Average loss=0.05272797477245331 +INFO:root: +[Valid][17][UTT][ MSE = 0.1697 | LCC = 0.9606 | SRCC = 0.9594 ] +INFO:root:Average loss=0.04896588778495788 +INFO:root: +[Valid][18][UTT][ MSE = 0.1696 | LCC = 0.9613 | SRCC = 0.9601 ] +INFO:root:Average loss=0.0479572434425354 +INFO:root: +[Valid][19][UTT][ MSE = 0.1828 | LCC = 0.9610 | SRCC = 0.9596 ] +INFO:root:Average loss=0.043995730489492416 +INFO:root: +[Valid][20][UTT][ MSE = 0.1558 | LCC = 0.9603 | SRCC = 0.9589 ] +INFO:root:Average loss=0.043300458177924155 +INFO:root: +[Valid][21][UTT][ MSE = 0.1677 | LCC = 0.9610 | SRCC = 0.9598 ] +INFO:root:Average loss=0.04187612639367581 +INFO:root: +[Valid][22][UTT][ MSE = 0.2056 | LCC = 0.9598 | SRCC = 0.9594 ] +INFO:root:Average loss=0.04159242367744446 +INFO:root: +[Valid][23][UTT][ MSE = 0.1922 | LCC = 0.9604 | SRCC = 0.9593 ] +INFO:root:Average loss=0.03936800995469093 +INFO:root: +[Valid][24][UTT][ MSE = 0.1853 | LCC = 0.9607 | SRCC = 0.9596 ] +INFO:root:Average loss=0.037449844270944596 +INFO:root: +[Valid][25][UTT][ MSE = 0.1692 | LCC = 0.9607 | SRCC = 0.9596 ] +INFO:root:Average loss=0.033755977541208265 +INFO:root: +[Valid][26][UTT][ MSE = 0.1770 | LCC = 0.9603 | SRCC = 0.9592 ] +INFO:root:Average loss=0.034037769943475724 +INFO:root: +[Valid][27][UTT][ MSE = 0.1802 | LCC = 0.9605 | SRCC = 0.9588 ] +INFO:root:Average loss=0.037080147787928584 +INFO:root: +[Valid][28][UTT][ MSE = 0.2004 | LCC = 0.9605 | SRCC = 0.9600 ] +INFO:root:Average loss=0.03280720244348049 +INFO:root: +[Valid][29][UTT][ MSE = 0.1724 | LCC = 0.9601 | SRCC = 0.9588 ] +INFO:root:Average loss=0.03155486150085926 +INFO:root: +[Valid][30][UTT][ MSE = 0.1787 | LCC = 0.9607 | SRCC = 0.9593 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1789 | LCC = 0.9605 | SRCC = 0.9610 ] diff --git a/results_tencent/hubert_base/layer6/run0/config.gin b/results_tencent/hubert_base/layer6/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer6/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer6/run0/model.pt b/results_tencent/hubert_base/layer6/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..eb412715530b6b6e5cdd3f8f40d821bbaeed6d4f --- /dev/null +++ b/results_tencent/hubert_base/layer6/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b4793deea6eb25fd1d87c5288f93e8a290cfaa9e6126471dd2dabad6222083e5 +size 2747979 diff --git a/results_tencent/hubert_base/layer6/run0/model_best.pt b/results_tencent/hubert_base/layer6/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c0dc0d3d47e76027043cecf85ac6ca1451fbfd47 --- /dev/null +++ b/results_tencent/hubert_base/layer6/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ae7a0dffc7eea72b8420d150f6159f2366670edd5abacbb0006427b6e11d4bd8 +size 2749702 diff --git a/results_tencent/hubert_base/layer6/run0/train.log b/results_tencent/hubert_base/layer6/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..cb46ed19eb42bd9fb36d8abefa2ae5d7eddb8a17 --- /dev/null +++ b/results_tencent/hubert_base/layer6/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4860212173461914 +INFO:root: +[Valid][0][UTT][ MSE = 0.3023 | LCC = 0.9217 | SRCC = 0.9193 ] +INFO:root:Average loss=0.18159495234489442 +INFO:root: +[Valid][1][UTT][ MSE = 0.2373 | LCC = 0.9430 | SRCC = 0.9409 ] +INFO:root:Average loss=0.14760519045591355 +INFO:root: +[Valid][2][UTT][ MSE = 0.2639 | LCC = 0.9478 | SRCC = 0.9475 ] +INFO:root:Average loss=0.12826874774694444 +INFO:root: +[Valid][3][UTT][ MSE = 0.2005 | LCC = 0.9517 | SRCC = 0.9507 ] +INFO:root:Average loss=0.11358935910463333 +INFO:root: +[Valid][4][UTT][ MSE = 0.2137 | LCC = 0.9538 | SRCC = 0.9531 ] +INFO:root:Average loss=0.10559042143821716 +INFO:root: +[Valid][5][UTT][ MSE = 0.1757 | LCC = 0.9527 | SRCC = 0.9525 ] +INFO:root:Average loss=0.09736696517467498 +INFO:root: +[Valid][6][UTT][ MSE = 0.1991 | LCC = 0.9544 | SRCC = 0.9534 ] +INFO:root:Average loss=0.0931207927763462 +INFO:root: +[Valid][7][UTT][ MSE = 0.1838 | LCC = 0.9547 | SRCC = 0.9536 ] +INFO:root:Average loss=0.08458654436469078 +INFO:root: +[Valid][8][UTT][ MSE = 0.1941 | LCC = 0.9566 | SRCC = 0.9559 ] +INFO:root:Average loss=0.07769061848521233 +INFO:root: +[Valid][9][UTT][ MSE = 0.1768 | LCC = 0.9556 | SRCC = 0.9548 ] +INFO:root:Average loss=0.0725905313193798 +INFO:root: +[Valid][10][UTT][ MSE = 0.1730 | LCC = 0.9578 | SRCC = 0.9568 ] +INFO:root:Average loss=0.06779861024022102 +INFO:root: +[Valid][11][UTT][ MSE = 0.1601 | LCC = 0.9576 | SRCC = 0.9572 ] +INFO:root:Average loss=0.0643130929172039 +INFO:root: +[Valid][12][UTT][ MSE = 0.1902 | LCC = 0.9578 | SRCC = 0.9565 ] +INFO:root:Average loss=0.0600448796749115 +INFO:root: +[Valid][13][UTT][ MSE = 0.1914 | LCC = 0.9577 | SRCC = 0.9577 ] +INFO:root:Average loss=0.05863547882437706 +INFO:root: +[Valid][14][UTT][ MSE = 0.1744 | LCC = 0.9577 | SRCC = 0.9568 ] +INFO:root:Average loss=0.053800854325294495 +INFO:root: +[Valid][15][UTT][ MSE = 0.1743 | LCC = 0.9584 | SRCC = 0.9572 ] +INFO:root:Average loss=0.050546395227313044 +INFO:root: +[Valid][16][UTT][ MSE = 0.1998 | LCC = 0.9584 | SRCC = 0.9578 ] +INFO:root:Average loss=0.048676232919096944 +INFO:root: +[Valid][17][UTT][ MSE = 0.1766 | LCC = 0.9576 | SRCC = 0.9563 ] +INFO:root:Average loss=0.04625153458118439 +INFO:root: +[Valid][18][UTT][ MSE = 0.1800 | LCC = 0.9594 | SRCC = 0.9587 ] +INFO:root:Average loss=0.043752063050866125 +INFO:root: +[Valid][19][UTT][ MSE = 0.2023 | LCC = 0.9567 | SRCC = 0.9572 ] +INFO:root:Average loss=0.041726768523454666 +INFO:root: +[Valid][20][UTT][ MSE = 0.1828 | LCC = 0.9593 | SRCC = 0.9580 ] +INFO:root:Average loss=0.03992378796637058 +INFO:root: +[Valid][21][UTT][ MSE = 0.1809 | LCC = 0.9585 | SRCC = 0.9575 ] +INFO:root:Average loss=0.03818070361018181 +INFO:root: +[Valid][22][UTT][ MSE = 0.1818 | LCC = 0.9582 | SRCC = 0.9572 ] +INFO:root:Average loss=0.03598105804622173 +INFO:root: +[Valid][23][UTT][ MSE = 0.1943 | LCC = 0.9586 | SRCC = 0.9580 ] +INFO:root:Average loss=0.034207244515419004 +INFO:root: +[Valid][24][UTT][ MSE = 0.1759 | LCC = 0.9583 | SRCC = 0.9574 ] +INFO:root:Average loss=0.03371845646202564 +INFO:root: +[Valid][25][UTT][ MSE = 0.1783 | LCC = 0.9597 | SRCC = 0.9588 ] +INFO:root:Average loss=0.032926248893141746 +INFO:root: +[Valid][26][UTT][ MSE = 0.1741 | LCC = 0.9602 | SRCC = 0.9593 ] +INFO:root:Average loss=0.029713588669896127 +INFO:root: +[Valid][27][UTT][ MSE = 0.1704 | LCC = 0.9598 | SRCC = 0.9590 ] +INFO:root:Average loss=0.030475397616624832 +INFO:root: +[Valid][28][UTT][ MSE = 0.1756 | LCC = 0.9604 | SRCC = 0.9594 ] +INFO:root:Average loss=0.029691836163401604 +INFO:root: +[Valid][29][UTT][ MSE = 0.1922 | LCC = 0.9595 | SRCC = 0.9588 ] +INFO:root:Average loss=0.028685176894068718 +INFO:root: +[Valid][30][UTT][ MSE = 0.1894 | LCC = 0.9591 | SRCC = 0.9589 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1852 | LCC = 0.9585 | SRCC = 0.9585 ] diff --git a/results_tencent/hubert_base/layer6/run1/config.gin b/results_tencent/hubert_base/layer6/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer6/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer6/run1/model.pt b/results_tencent/hubert_base/layer6/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..215e324ccafddaf9f6e2b49c79c19fedaf3490f7 --- /dev/null +++ b/results_tencent/hubert_base/layer6/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:89403567a273d9a5fd11ecaf843e8f06a3bae3f364a7fb39a3ee5cecf5cc17bf +size 2747979 diff --git a/results_tencent/hubert_base/layer6/run1/model_best.pt b/results_tencent/hubert_base/layer6/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..66a4de87c2f2b2fb33e239fd895ef3d212354f88 --- /dev/null +++ b/results_tencent/hubert_base/layer6/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e11b242c3e5ca0356be5c3eb78477d8ae2b19b7007f48c7e731ca30ae7169215 +size 2749702 diff --git a/results_tencent/hubert_base/layer6/run1/train.log b/results_tencent/hubert_base/layer6/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6288a1f4ce676ad22b7875bf93805fb05602f1d3 --- /dev/null +++ b/results_tencent/hubert_base/layer6/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.45139190936088563 +INFO:root: +[Valid][0][UTT][ MSE = 0.3029 | LCC = 0.9204 | SRCC = 0.9195 ] +INFO:root:Average loss=0.18781933081150054 +INFO:root: +[Valid][1][UTT][ MSE = 0.2384 | LCC = 0.9379 | SRCC = 0.9376 ] +INFO:root:Average loss=0.15095506459474564 +INFO:root: +[Valid][2][UTT][ MSE = 0.2074 | LCC = 0.9429 | SRCC = 0.9440 ] +INFO:root:Average loss=0.13250942754745484 +INFO:root: +[Valid][3][UTT][ MSE = 0.1992 | LCC = 0.9477 | SRCC = 0.9472 ] +INFO:root:Average loss=0.12005011171102524 +INFO:root: +[Valid][4][UTT][ MSE = 0.1759 | LCC = 0.9512 | SRCC = 0.9516 ] +INFO:root:Average loss=0.10952923047542572 +INFO:root: +[Valid][5][UTT][ MSE = 0.1848 | LCC = 0.9537 | SRCC = 0.9534 ] +INFO:root:Average loss=0.09852313885092735 +INFO:root: +[Valid][6][UTT][ MSE = 0.1853 | LCC = 0.9516 | SRCC = 0.9516 ] +INFO:root:Average loss=0.09237195628881455 +INFO:root: +[Valid][7][UTT][ MSE = 0.1794 | LCC = 0.9560 | SRCC = 0.9552 ] +INFO:root:Average loss=0.08177016216516494 +INFO:root: +[Valid][8][UTT][ MSE = 0.1775 | LCC = 0.9551 | SRCC = 0.9541 ] +INFO:root:Average loss=0.08053918594121932 +INFO:root: +[Valid][9][UTT][ MSE = 0.1877 | LCC = 0.9563 | SRCC = 0.9553 ] +INFO:root:Average loss=0.07255291226506233 +INFO:root: +[Valid][10][UTT][ MSE = 0.1627 | LCC = 0.9562 | SRCC = 0.9555 ] +INFO:root:Average loss=0.07130857136845589 +INFO:root: +[Valid][11][UTT][ MSE = 0.1682 | LCC = 0.9570 | SRCC = 0.9559 ] +INFO:root:Average loss=0.06450583267211914 +INFO:root: +[Valid][12][UTT][ MSE = 0.1548 | LCC = 0.9563 | SRCC = 0.9561 ] +INFO:root:Average loss=0.06175072214007377 +INFO:root: +[Valid][13][UTT][ MSE = 0.1790 | LCC = 0.9573 | SRCC = 0.9564 ] +INFO:root:Average loss=0.05693508513271808 +INFO:root: +[Valid][14][UTT][ MSE = 0.1912 | LCC = 0.9566 | SRCC = 0.9553 ] +INFO:root:Average loss=0.05695650634169579 +INFO:root: +[Valid][15][UTT][ MSE = 0.1694 | LCC = 0.9561 | SRCC = 0.9549 ] +INFO:root:Average loss=0.04971479913592339 +INFO:root: +[Valid][16][UTT][ MSE = 0.1613 | LCC = 0.9556 | SRCC = 0.9561 ] +INFO:root:Average loss=0.04990753784775734 +INFO:root: +[Valid][17][UTT][ MSE = 0.1783 | LCC = 0.9574 | SRCC = 0.9565 ] +INFO:root:Average loss=0.04656219618022442 +INFO:root: +[Valid][18][UTT][ MSE = 0.1590 | LCC = 0.9578 | SRCC = 0.9573 ] +INFO:root:Average loss=0.04499005725979805 +INFO:root: +[Valid][19][UTT][ MSE = 0.1799 | LCC = 0.9580 | SRCC = 0.9572 ] +INFO:root:Average loss=0.04280725106596947 +INFO:root: +[Valid][20][UTT][ MSE = 0.1967 | LCC = 0.9575 | SRCC = 0.9566 ] +INFO:root:Average loss=0.04136967197060585 +INFO:root: +[Valid][21][UTT][ MSE = 0.1813 | LCC = 0.9574 | SRCC = 0.9567 ] +INFO:root:Average loss=0.04063347715139389 +INFO:root: +[Valid][22][UTT][ MSE = 0.1614 | LCC = 0.9580 | SRCC = 0.9573 ] +INFO:root:Average loss=0.037435056164860724 +INFO:root: +[Valid][23][UTT][ MSE = 0.1836 | LCC = 0.9580 | SRCC = 0.9569 ] +INFO:root:Average loss=0.036187818929553035 +INFO:root: +[Valid][24][UTT][ MSE = 0.1881 | LCC = 0.9575 | SRCC = 0.9562 ] +INFO:root:Average loss=0.03592694228887558 +INFO:root: +[Valid][25][UTT][ MSE = 0.1838 | LCC = 0.9577 | SRCC = 0.9568 ] +INFO:root:Average loss=0.034170073181390766 +INFO:root: +[Valid][26][UTT][ MSE = 0.1940 | LCC = 0.9579 | SRCC = 0.9568 ] +INFO:root:Average loss=0.03472886946797371 +INFO:root: +[Valid][27][UTT][ MSE = 0.1771 | LCC = 0.9587 | SRCC = 0.9575 ] +INFO:root:Average loss=0.033754281431436536 +INFO:root: +[Valid][28][UTT][ MSE = 0.1748 | LCC = 0.9590 | SRCC = 0.9577 ] +INFO:root:Average loss=0.030566699728369714 +INFO:root: +[Valid][29][UTT][ MSE = 0.1775 | LCC = 0.9573 | SRCC = 0.9564 ] +INFO:root:Average loss=0.02976990607380867 +INFO:root: +[Valid][30][UTT][ MSE = 0.1704 | LCC = 0.9587 | SRCC = 0.9578 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1787 | LCC = 0.9582 | SRCC = 0.9580 ] diff --git a/results_tencent/hubert_base/layer6/run2/config.gin b/results_tencent/hubert_base/layer6/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer6/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer6/run2/model.pt b/results_tencent/hubert_base/layer6/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2573d2552ce595f757521289cd8802b102ec2060 --- /dev/null +++ b/results_tencent/hubert_base/layer6/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fcded068edaaadb881e548ee867c348df75b611d3f03d85e2b1d923eaa258e1d +size 2747979 diff --git a/results_tencent/hubert_base/layer6/run2/model_best.pt b/results_tencent/hubert_base/layer6/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..87061277c73af52d01e0857748ddcdda09ec89af --- /dev/null +++ b/results_tencent/hubert_base/layer6/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:85b7ff94826982bc00bf109ac537d2b8d8b7c40875ba389a4471ce4b764050b1 +size 2749702 diff --git a/results_tencent/hubert_base/layer6/run2/train.log b/results_tencent/hubert_base/layer6/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..fe276f3a5ba177c22fb5e588a56523f10dbc9b2c --- /dev/null +++ b/results_tencent/hubert_base/layer6/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4315011500120163 +INFO:root: +[Valid][0][UTT][ MSE = 0.2612 | LCC = 0.9236 | SRCC = 0.9226 ] +INFO:root:Average loss=0.18196563023328782 +INFO:root: +[Valid][1][UTT][ MSE = 0.2135 | LCC = 0.9451 | SRCC = 0.9438 ] +INFO:root:Average loss=0.15135452783107758 +INFO:root: +[Valid][2][UTT][ MSE = 0.2196 | LCC = 0.9496 | SRCC = 0.9488 ] +INFO:root:Average loss=0.12991687697172166 +INFO:root: +[Valid][3][UTT][ MSE = 0.1907 | LCC = 0.9503 | SRCC = 0.9503 ] +INFO:root:Average loss=0.1201533374786377 +INFO:root: +[Valid][4][UTT][ MSE = 0.2107 | LCC = 0.9509 | SRCC = 0.9491 ] +INFO:root:Average loss=0.10485779693722724 +INFO:root: +[Valid][5][UTT][ MSE = 0.1942 | LCC = 0.9548 | SRCC = 0.9540 ] +INFO:root:Average loss=0.09728343737125397 +INFO:root: +[Valid][6][UTT][ MSE = 0.1960 | LCC = 0.9545 | SRCC = 0.9530 ] +INFO:root:Average loss=0.08978805124759674 +INFO:root: +[Valid][7][UTT][ MSE = 0.1824 | LCC = 0.9555 | SRCC = 0.9545 ] +INFO:root:Average loss=0.08092806208133697 +INFO:root: +[Valid][8][UTT][ MSE = 0.1857 | LCC = 0.9555 | SRCC = 0.9544 ] +INFO:root:Average loss=0.07827318277955055 +INFO:root: +[Valid][9][UTT][ MSE = 0.1790 | LCC = 0.9559 | SRCC = 0.9549 ] +INFO:root:Average loss=0.07251308035850525 +INFO:root: +[Valid][10][UTT][ MSE = 0.1865 | LCC = 0.9524 | SRCC = 0.9522 ] +INFO:root:Average loss=0.06664940357208252 +INFO:root: +[Valid][11][UTT][ MSE = 0.1821 | LCC = 0.9523 | SRCC = 0.9523 ] +INFO:root:Average loss=0.0678919721841812 +INFO:root: +[Valid][12][UTT][ MSE = 0.1800 | LCC = 0.9576 | SRCC = 0.9565 ] +INFO:root:Average loss=0.06060925522446632 +INFO:root: +[Valid][13][UTT][ MSE = 0.1852 | LCC = 0.9575 | SRCC = 0.9565 ] +INFO:root:Average loss=0.056369058161973955 +INFO:root: +[Valid][14][UTT][ MSE = 0.1677 | LCC = 0.9580 | SRCC = 0.9570 ] +INFO:root:Average loss=0.053753947377204896 +INFO:root: +[Valid][15][UTT][ MSE = 0.1793 | LCC = 0.9581 | SRCC = 0.9566 ] +INFO:root:Average loss=0.0497821713835001 +INFO:root: +[Valid][16][UTT][ MSE = 0.1692 | LCC = 0.9551 | SRCC = 0.9542 ] +INFO:root:Average loss=0.049183897912502286 +INFO:root: +[Valid][17][UTT][ MSE = 0.1697 | LCC = 0.9587 | SRCC = 0.9576 ] +INFO:root:Average loss=0.04506975698471069 +INFO:root: +[Valid][18][UTT][ MSE = 0.1674 | LCC = 0.9581 | SRCC = 0.9569 ] +INFO:root:Average loss=0.04591934308409691 +INFO:root: +[Valid][19][UTT][ MSE = 0.1699 | LCC = 0.9572 | SRCC = 0.9566 ] +INFO:root:Average loss=0.04380649127066136 +INFO:root: +[Valid][20][UTT][ MSE = 0.1626 | LCC = 0.9579 | SRCC = 0.9570 ] +INFO:root:Average loss=0.03941837926208973 +INFO:root: +[Valid][21][UTT][ MSE = 0.1702 | LCC = 0.9588 | SRCC = 0.9577 ] +INFO:root:Average loss=0.03901199531555176 +INFO:root: +[Valid][22][UTT][ MSE = 0.1558 | LCC = 0.9571 | SRCC = 0.9567 ] +INFO:root:Average loss=0.037800316989421844 +INFO:root: +[Valid][23][UTT][ MSE = 0.1791 | LCC = 0.9578 | SRCC = 0.9566 ] +INFO:root:Average loss=0.03506005170941353 +INFO:root: +[Valid][24][UTT][ MSE = 0.1585 | LCC = 0.9568 | SRCC = 0.9570 ] +INFO:root:Average loss=0.035142500698566434 +INFO:root: +[Valid][25][UTT][ MSE = 0.1614 | LCC = 0.9591 | SRCC = 0.9576 ] +INFO:root:Average loss=0.03380853132903576 +INFO:root: +[Valid][26][UTT][ MSE = 0.1732 | LCC = 0.9585 | SRCC = 0.9573 ] +INFO:root:Average loss=0.03261709921061993 +INFO:root: +[Valid][27][UTT][ MSE = 0.1776 | LCC = 0.9570 | SRCC = 0.9571 ] +INFO:root:Average loss=0.03214026808738708 +INFO:root: +[Valid][28][UTT][ MSE = 0.1641 | LCC = 0.9582 | SRCC = 0.9571 ] +INFO:root:Average loss=0.02954187139868736 +INFO:root: +[Valid][29][UTT][ MSE = 0.1733 | LCC = 0.9589 | SRCC = 0.9577 ] +INFO:root:Average loss=0.029788785591721535 +INFO:root: +[Valid][30][UTT][ MSE = 0.1599 | LCC = 0.9581 | SRCC = 0.9573 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1675 | LCC = 0.9569 | SRCC = 0.9570 ] diff --git a/results_tencent/hubert_base/layer6/run3/config.gin b/results_tencent/hubert_base/layer6/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer6/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer6/run3/model.pt b/results_tencent/hubert_base/layer6/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a1456edae39161ab56124a3374ef4be7100a077f --- /dev/null +++ b/results_tencent/hubert_base/layer6/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b9155d11a3edc8a1ff708211a3ea8d1f4b80a1b2aeea3d97e43fa632b3f4640e +size 2747979 diff --git a/results_tencent/hubert_base/layer6/run3/model_best.pt b/results_tencent/hubert_base/layer6/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..402c2fbcb6d7784f2f84c8e7cb362c91206e8142 --- /dev/null +++ b/results_tencent/hubert_base/layer6/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:919e3a618d52f78f530128edf69c98679dc526d67852ef6b14d3ca8115159e23 +size 2749638 diff --git a/results_tencent/hubert_base/layer6/run3/train.log b/results_tencent/hubert_base/layer6/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..bac45d5cd8d018a9b79f138d785dc6bd8042ed5d --- /dev/null +++ b/results_tencent/hubert_base/layer6/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5143626892566681 +INFO:root: +[Valid][0][UTT][ MSE = 0.2925 | LCC = 0.9217 | SRCC = 0.9196 ] +INFO:root:Average loss=0.19196676856279374 +INFO:root: +[Valid][1][UTT][ MSE = 0.2106 | LCC = 0.9400 | SRCC = 0.9390 ] +INFO:root:Average loss=0.15613075160980225 +INFO:root: +[Valid][2][UTT][ MSE = 0.2254 | LCC = 0.9472 | SRCC = 0.9458 ] +INFO:root:Average loss=0.13340931111574172 +INFO:root: +[Valid][3][UTT][ MSE = 0.2005 | LCC = 0.9491 | SRCC = 0.9488 ] +INFO:root:Average loss=0.12498688286542893 +INFO:root: +[Valid][4][UTT][ MSE = 0.2123 | LCC = 0.9497 | SRCC = 0.9495 ] +INFO:root:Average loss=0.1095610525906086 +INFO:root: +[Valid][5][UTT][ MSE = 0.2022 | LCC = 0.9528 | SRCC = 0.9523 ] +INFO:root:Average loss=0.10103139209747314 +INFO:root: +[Valid][6][UTT][ MSE = 0.1946 | LCC = 0.9525 | SRCC = 0.9520 ] +INFO:root:Average loss=0.09304774567484855 +INFO:root: +[Valid][7][UTT][ MSE = 0.1875 | LCC = 0.9540 | SRCC = 0.9536 ] +INFO:root:Average loss=0.08797001099586486 +INFO:root: +[Valid][8][UTT][ MSE = 0.2023 | LCC = 0.9551 | SRCC = 0.9552 ] +INFO:root:Average loss=0.07864772894978524 +INFO:root: +[Valid][9][UTT][ MSE = 0.2022 | LCC = 0.9553 | SRCC = 0.9552 ] +INFO:root:Average loss=0.0738165755867958 +INFO:root: +[Valid][10][UTT][ MSE = 0.1951 | LCC = 0.9543 | SRCC = 0.9537 ] +INFO:root:Average loss=0.07231524133682252 +INFO:root: +[Valid][11][UTT][ MSE = 0.2031 | LCC = 0.9563 | SRCC = 0.9557 ] +INFO:root:Average loss=0.06809905111789703 +INFO:root: +[Valid][12][UTT][ MSE = 0.1723 | LCC = 0.9546 | SRCC = 0.9546 ] +INFO:root:Average loss=0.06349330937862396 +INFO:root: +[Valid][13][UTT][ MSE = 0.1891 | LCC = 0.9556 | SRCC = 0.9556 ] +INFO:root:Average loss=0.059808895021677015 +INFO:root: +[Valid][14][UTT][ MSE = 0.1918 | LCC = 0.9574 | SRCC = 0.9566 ] +INFO:root:Average loss=0.05775750303268433 +INFO:root: +[Valid][15][UTT][ MSE = 0.1883 | LCC = 0.9566 | SRCC = 0.9558 ] +INFO:root:Average loss=0.054320946246385574 +INFO:root: +[Valid][16][UTT][ MSE = 0.2025 | LCC = 0.9559 | SRCC = 0.9550 ] +INFO:root:Average loss=0.05170893862843513 +INFO:root: +[Valid][17][UTT][ MSE = 0.2164 | LCC = 0.9542 | SRCC = 0.9541 ] +INFO:root:Average loss=0.049963758125901224 +INFO:root: +[Valid][18][UTT][ MSE = 0.1866 | LCC = 0.9540 | SRCC = 0.9534 ] +INFO:root:Average loss=0.04549326673150063 +INFO:root: +[Valid][19][UTT][ MSE = 0.1785 | LCC = 0.9562 | SRCC = 0.9559 ] +INFO:root:Average loss=0.04498056173324585 +INFO:root: +[Valid][20][UTT][ MSE = 0.1955 | LCC = 0.9553 | SRCC = 0.9545 ] +INFO:root:Average loss=0.04472668056190014 +INFO:root: +[Valid][21][UTT][ MSE = 0.1910 | LCC = 0.9543 | SRCC = 0.9541 ] +INFO:root:Average loss=0.040495546087622646 +INFO:root: +[Valid][22][UTT][ MSE = 0.1745 | LCC = 0.9558 | SRCC = 0.9554 ] +INFO:root:Average loss=0.03945193675160408 +INFO:root: +[Valid][23][UTT][ MSE = 0.1906 | LCC = 0.9570 | SRCC = 0.9561 ] +INFO:root:Average loss=0.03648126123845577 +INFO:root: +[Valid][24][UTT][ MSE = 0.1887 | LCC = 0.9568 | SRCC = 0.9563 ] +INFO:root:Average loss=0.03649713271856308 +INFO:root: +[Valid][25][UTT][ MSE = 0.1825 | LCC = 0.9555 | SRCC = 0.9547 ] +INFO:root:Average loss=0.03557686381042004 +INFO:root: +[Valid][26][UTT][ MSE = 0.1901 | LCC = 0.9559 | SRCC = 0.9549 ] +INFO:root:Average loss=0.033247600451111796 +INFO:root: +[Valid][27][UTT][ MSE = 0.1892 | LCC = 0.9559 | SRCC = 0.9552 ] +INFO:root:Average loss=0.03273570615053177 +INFO:root: +[Valid][28][UTT][ MSE = 0.1909 | LCC = 0.9568 | SRCC = 0.9562 ] +INFO:root:Average loss=0.031908841915428636 +INFO:root: +[Valid][29][UTT][ MSE = 0.1675 | LCC = 0.9549 | SRCC = 0.9551 ] +INFO:root:Average loss=0.03125849361717701 +INFO:root: +[Valid][30][UTT][ MSE = 0.2131 | LCC = 0.9568 | SRCC = 0.9565 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1975 | LCC = 0.9574 | SRCC = 0.9584 ] diff --git a/results_tencent/hubert_base/layer6/run4/config.gin b/results_tencent/hubert_base/layer6/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer6/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer6/run4/model.pt b/results_tencent/hubert_base/layer6/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..24eaddf9876ba68714d89b44ec1323adc3889573 --- /dev/null +++ b/results_tencent/hubert_base/layer6/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8a3d4f2007f727f1dffa27aeff0680522527a221f90707b168bb3ec4ca66f78c +size 2747979 diff --git a/results_tencent/hubert_base/layer6/run4/model_best.pt b/results_tencent/hubert_base/layer6/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..02499e91ce032f28f4e6e2f3d2968d0a29206741 --- /dev/null +++ b/results_tencent/hubert_base/layer6/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b29423fab674938438934e18cd4d364fd660c02f8a9250de09fb383f19a3f725 +size 2749638 diff --git a/results_tencent/hubert_base/layer6/run4/train.log b/results_tencent/hubert_base/layer6/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..76467ecbb46e86d56a87206abf994d9b5fe70d72 --- /dev/null +++ b/results_tencent/hubert_base/layer6/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5520982341766357 +INFO:root: +[Valid][0][UTT][ MSE = 0.2943 | LCC = 0.9149 | SRCC = 0.9127 ] +INFO:root:Average loss=0.20228598046302795 +INFO:root: +[Valid][1][UTT][ MSE = 0.2331 | LCC = 0.9354 | SRCC = 0.9332 ] +INFO:root:Average loss=0.15975645339488984 +INFO:root: +[Valid][2][UTT][ MSE = 0.2272 | LCC = 0.9441 | SRCC = 0.9430 ] +INFO:root:Average loss=0.13706517285108566 +INFO:root: +[Valid][3][UTT][ MSE = 0.1999 | LCC = 0.9453 | SRCC = 0.9434 ] +INFO:root:Average loss=0.12296825575828552 +INFO:root: +[Valid][4][UTT][ MSE = 0.2055 | LCC = 0.9469 | SRCC = 0.9453 ] +INFO:root:Average loss=0.11711147662997246 +INFO:root: +[Valid][5][UTT][ MSE = 0.1975 | LCC = 0.9514 | SRCC = 0.9503 ] +INFO:root:Average loss=0.1016559660434723 +INFO:root: +[Valid][6][UTT][ MSE = 0.1999 | LCC = 0.9504 | SRCC = 0.9492 ] +INFO:root:Average loss=0.09599425464868545 +INFO:root: +[Valid][7][UTT][ MSE = 0.2182 | LCC = 0.9528 | SRCC = 0.9514 ] +INFO:root:Average loss=0.0900737980902195 +INFO:root: +[Valid][8][UTT][ MSE = 0.1703 | LCC = 0.9536 | SRCC = 0.9530 ] +INFO:root:Average loss=0.07985910657048226 +INFO:root: +[Valid][9][UTT][ MSE = 0.2189 | LCC = 0.9540 | SRCC = 0.9531 ] +INFO:root:Average loss=0.07896288481354713 +INFO:root: +[Valid][10][UTT][ MSE = 0.1841 | LCC = 0.9551 | SRCC = 0.9537 ] +INFO:root:Average loss=0.07397510513663291 +INFO:root: +[Valid][11][UTT][ MSE = 0.1753 | LCC = 0.9547 | SRCC = 0.9540 ] +INFO:root:Average loss=0.06598313796520233 +INFO:root: +[Valid][12][UTT][ MSE = 0.1804 | LCC = 0.9547 | SRCC = 0.9542 ] +INFO:root:Average loss=0.06503885313868522 +INFO:root: +[Valid][13][UTT][ MSE = 0.1829 | LCC = 0.9557 | SRCC = 0.9545 ] +INFO:root:Average loss=0.05884318825602532 +INFO:root: +[Valid][14][UTT][ MSE = 0.1819 | LCC = 0.9534 | SRCC = 0.9528 ] +INFO:root:Average loss=0.05667384508252144 +INFO:root: +[Valid][15][UTT][ MSE = 0.2017 | LCC = 0.9562 | SRCC = 0.9550 ] +INFO:root:Average loss=0.053415642112493514 +INFO:root: +[Valid][16][UTT][ MSE = 0.1868 | LCC = 0.9534 | SRCC = 0.9523 ] +INFO:root:Average loss=0.051416541159152986 +INFO:root: +[Valid][17][UTT][ MSE = 0.1890 | LCC = 0.9562 | SRCC = 0.9550 ] +INFO:root:Average loss=0.04558428977429867 +INFO:root: +[Valid][18][UTT][ MSE = 0.1802 | LCC = 0.9549 | SRCC = 0.9536 ] +INFO:root:Average loss=0.045692944407463076 +INFO:root: +[Valid][19][UTT][ MSE = 0.1663 | LCC = 0.9533 | SRCC = 0.9528 ] +INFO:root:Average loss=0.04321558313071728 +INFO:root: +[Valid][20][UTT][ MSE = 0.1903 | LCC = 0.9560 | SRCC = 0.9549 ] +INFO:root:Average loss=0.04263185209035873 +INFO:root: +[Valid][21][UTT][ MSE = 0.1781 | LCC = 0.9554 | SRCC = 0.9544 ] +INFO:root:Average loss=0.03884760373830795 +INFO:root: +[Valid][22][UTT][ MSE = 0.1947 | LCC = 0.9534 | SRCC = 0.9519 ] +INFO:root:Average loss=0.04034323942661285 +INFO:root: +[Valid][23][UTT][ MSE = 0.1859 | LCC = 0.9542 | SRCC = 0.9530 ] +INFO:root:Average loss=0.037458357259631155 +INFO:root: +[Valid][24][UTT][ MSE = 0.1937 | LCC = 0.9551 | SRCC = 0.9538 ] +INFO:root:Average loss=0.036529575929045675 +INFO:root: +[Valid][25][UTT][ MSE = 0.2035 | LCC = 0.9551 | SRCC = 0.9541 ] +INFO:root:Average loss=0.035926210626959804 +INFO:root: +[Valid][26][UTT][ MSE = 0.1719 | LCC = 0.9568 | SRCC = 0.9559 ] +INFO:root:Average loss=0.03399308417737484 +INFO:root: +[Valid][27][UTT][ MSE = 0.1765 | LCC = 0.9564 | SRCC = 0.9556 ] +INFO:root:Average loss=0.03124421052634716 +INFO:root: +[Valid][28][UTT][ MSE = 0.1772 | LCC = 0.9565 | SRCC = 0.9556 ] +INFO:root:Average loss=0.0304555334597826 +INFO:root: +[Valid][29][UTT][ MSE = 0.1919 | LCC = 0.9543 | SRCC = 0.9531 ] +INFO:root:Average loss=0.029286164954304694 +INFO:root: +[Valid][30][UTT][ MSE = 0.1879 | LCC = 0.9558 | SRCC = 0.9543 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1687 | LCC = 0.9592 | SRCC = 0.9587 ] diff --git a/results_tencent/hubert_base/layer7/run0/config.gin b/results_tencent/hubert_base/layer7/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer7/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer7/run0/model.pt b/results_tencent/hubert_base/layer7/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f3a3fd9f5b95a6a4bca582b9ba2b0d5910600904 --- /dev/null +++ b/results_tencent/hubert_base/layer7/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f30e2be1e268bdb9811c47880de17f2a32cb3e90145e67ff6a75d80068f5661c +size 2747979 diff --git a/results_tencent/hubert_base/layer7/run0/model_best.pt b/results_tencent/hubert_base/layer7/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f5a77719d7309e14aba94fa49fb5817c621cd598 --- /dev/null +++ b/results_tencent/hubert_base/layer7/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:66ba4bc8246b85f6f7768cbd81a140dfd25c4410a7ea7d74ecb240659aca120a +size 2749702 diff --git a/results_tencent/hubert_base/layer7/run0/train.log b/results_tencent/hubert_base/layer7/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3843e20f25e577032c28ffe840d01febacabbb9b --- /dev/null +++ b/results_tencent/hubert_base/layer7/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.43120053255558016 +INFO:root: +[Valid][0][UTT][ MSE = 0.3199 | LCC = 0.9089 | SRCC = 0.9037 ] +INFO:root:Average loss=0.20694168853759765 +INFO:root: +[Valid][1][UTT][ MSE = 0.3368 | LCC = 0.9238 | SRCC = 0.9218 ] +INFO:root:Average loss=0.16600096023082733 +INFO:root: +[Valid][2][UTT][ MSE = 0.2285 | LCC = 0.9424 | SRCC = 0.9411 ] +INFO:root:Average loss=0.14629348331689834 +INFO:root: +[Valid][3][UTT][ MSE = 0.1957 | LCC = 0.9472 | SRCC = 0.9471 ] +INFO:root:Average loss=0.12836238610744477 +INFO:root: +[Valid][4][UTT][ MSE = 0.2148 | LCC = 0.9501 | SRCC = 0.9497 ] +INFO:root:Average loss=0.11403924262523651 +INFO:root: +[Valid][5][UTT][ MSE = 0.1935 | LCC = 0.9514 | SRCC = 0.9518 ] +INFO:root:Average loss=0.10376023507118225 +INFO:root: +[Valid][6][UTT][ MSE = 0.2228 | LCC = 0.9520 | SRCC = 0.9513 ] +INFO:root:Average loss=0.09594639700651168 +INFO:root: +[Valid][7][UTT][ MSE = 0.2175 | LCC = 0.9529 | SRCC = 0.9531 ] +INFO:root:Average loss=0.08920199757814408 +INFO:root: +[Valid][8][UTT][ MSE = 0.2080 | LCC = 0.9539 | SRCC = 0.9534 ] +INFO:root:Average loss=0.08799193593859672 +INFO:root: +[Valid][9][UTT][ MSE = 0.2192 | LCC = 0.9529 | SRCC = 0.9530 ] +INFO:root:Average loss=0.07680274310708046 +INFO:root: +[Valid][10][UTT][ MSE = 0.2341 | LCC = 0.9530 | SRCC = 0.9522 ] +INFO:root:Average loss=0.07049426981806756 +INFO:root: +[Valid][11][UTT][ MSE = 0.1982 | LCC = 0.9537 | SRCC = 0.9531 ] +INFO:root:Average loss=0.06457051950693131 +INFO:root: +[Valid][12][UTT][ MSE = 0.1969 | LCC = 0.9525 | SRCC = 0.9523 ] +INFO:root:Average loss=0.06419781291484833 +INFO:root: +[Valid][13][UTT][ MSE = 0.2006 | LCC = 0.9538 | SRCC = 0.9530 ] +INFO:root:Average loss=0.0590320862531662 +INFO:root: +[Valid][14][UTT][ MSE = 0.1855 | LCC = 0.9551 | SRCC = 0.9549 ] +INFO:root:Average loss=0.055687815994024274 +INFO:root: +[Valid][15][UTT][ MSE = 0.2147 | LCC = 0.9514 | SRCC = 0.9522 ] +INFO:root:Average loss=0.05222860746085644 +INFO:root: +[Valid][16][UTT][ MSE = 0.1923 | LCC = 0.9541 | SRCC = 0.9541 ] +INFO:root:Average loss=0.04805498649179935 +INFO:root: +[Valid][17][UTT][ MSE = 0.2280 | LCC = 0.9535 | SRCC = 0.9533 ] +INFO:root:Average loss=0.04914399562776089 +INFO:root: +[Valid][18][UTT][ MSE = 0.2150 | LCC = 0.9545 | SRCC = 0.9543 ] +INFO:root:Average loss=0.04460493475198746 +INFO:root: +[Valid][19][UTT][ MSE = 0.2152 | LCC = 0.9545 | SRCC = 0.9541 ] +INFO:root:Average loss=0.0418499405682087 +INFO:root: +[Valid][20][UTT][ MSE = 0.1767 | LCC = 0.9558 | SRCC = 0.9552 ] +INFO:root:Average loss=0.03925532965362072 +INFO:root: +[Valid][21][UTT][ MSE = 0.1791 | LCC = 0.9557 | SRCC = 0.9554 ] +INFO:root:Average loss=0.04052885316312313 +INFO:root: +[Valid][22][UTT][ MSE = 0.1766 | LCC = 0.9550 | SRCC = 0.9544 ] +INFO:root:Average loss=0.037357772558927534 +INFO:root: +[Valid][23][UTT][ MSE = 0.2097 | LCC = 0.9550 | SRCC = 0.9550 ] +INFO:root:Average loss=0.03640632905066013 +INFO:root: +[Valid][24][UTT][ MSE = 0.2012 | LCC = 0.9532 | SRCC = 0.9527 ] +INFO:root:Average loss=0.03609071524441242 +INFO:root: +[Valid][25][UTT][ MSE = 0.1948 | LCC = 0.9556 | SRCC = 0.9556 ] +INFO:root:Average loss=0.03330813436210155 +INFO:root: +[Valid][26][UTT][ MSE = 0.2004 | LCC = 0.9545 | SRCC = 0.9544 ] +INFO:root:Average loss=0.03234934736788273 +INFO:root: +[Valid][27][UTT][ MSE = 0.1914 | LCC = 0.9564 | SRCC = 0.9561 ] +INFO:root:Average loss=0.03148770202696324 +INFO:root: +[Valid][28][UTT][ MSE = 0.1863 | LCC = 0.9546 | SRCC = 0.9546 ] +INFO:root:Average loss=0.028437227055430413 +INFO:root: +[Valid][29][UTT][ MSE = 0.1803 | LCC = 0.9560 | SRCC = 0.9555 ] +INFO:root:Average loss=0.02833978758752346 +INFO:root: +[Valid][30][UTT][ MSE = 0.1781 | LCC = 0.9564 | SRCC = 0.9559 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1979 | LCC = 0.9546 | SRCC = 0.9549 ] diff --git a/results_tencent/hubert_base/layer7/run1/config.gin b/results_tencent/hubert_base/layer7/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer7/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer7/run1/model.pt b/results_tencent/hubert_base/layer7/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bc057512992800c5e0c216032eb4cfcf255e6855 --- /dev/null +++ b/results_tencent/hubert_base/layer7/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4bceae328533961db5dfb76f74332ec2eb85e391fc0987c6a4bc73082e012bfc +size 2747979 diff --git a/results_tencent/hubert_base/layer7/run1/model_best.pt b/results_tencent/hubert_base/layer7/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ccf5bf7fe582a55c0f7fe76270837adda0a43ef2 --- /dev/null +++ b/results_tencent/hubert_base/layer7/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ce472f9992140ac430a3a2ec1ec337e934024ee57d8e1c8e8f351c546bacde68 +size 2749702 diff --git a/results_tencent/hubert_base/layer7/run1/train.log b/results_tencent/hubert_base/layer7/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0a8b4be7b07ace4226d74b82b90add49cd13bd97 --- /dev/null +++ b/results_tencent/hubert_base/layer7/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42486579084396364 +INFO:root: +[Valid][0][UTT][ MSE = 0.3343 | LCC = 0.9192 | SRCC = 0.9178 ] +INFO:root:Average loss=0.19630204206705093 +INFO:root: +[Valid][1][UTT][ MSE = 0.2947 | LCC = 0.9364 | SRCC = 0.9384 ] +INFO:root:Average loss=0.1647964768409729 +INFO:root: +[Valid][2][UTT][ MSE = 0.2392 | LCC = 0.9436 | SRCC = 0.9434 ] +INFO:root:Average loss=0.14248754012584686 +INFO:root: +[Valid][3][UTT][ MSE = 0.2362 | LCC = 0.9471 | SRCC = 0.9473 ] +INFO:root:Average loss=0.12802432626485824 +INFO:root: +[Valid][4][UTT][ MSE = 0.2473 | LCC = 0.9512 | SRCC = 0.9509 ] +INFO:root:Average loss=0.11395532363653182 +INFO:root: +[Valid][5][UTT][ MSE = 0.2591 | LCC = 0.9521 | SRCC = 0.9518 ] +INFO:root:Average loss=0.10804945474863052 +INFO:root: +[Valid][6][UTT][ MSE = 0.2330 | LCC = 0.9530 | SRCC = 0.9528 ] +INFO:root:Average loss=0.09726630833745002 +INFO:root: +[Valid][7][UTT][ MSE = 0.2351 | LCC = 0.9494 | SRCC = 0.9493 ] +INFO:root:Average loss=0.08794758522510529 +INFO:root: +[Valid][8][UTT][ MSE = 0.2136 | LCC = 0.9539 | SRCC = 0.9533 ] +INFO:root:Average loss=0.08019592061638832 +INFO:root: +[Valid][9][UTT][ MSE = 0.2328 | LCC = 0.9538 | SRCC = 0.9530 ] +INFO:root:Average loss=0.0787965886592865 +INFO:root: +[Valid][10][UTT][ MSE = 0.2116 | LCC = 0.9531 | SRCC = 0.9525 ] +INFO:root:Average loss=0.0710307814180851 +INFO:root: +[Valid][11][UTT][ MSE = 0.2202 | LCC = 0.9558 | SRCC = 0.9548 ] +INFO:root:Average loss=0.06724451208114623 +INFO:root: +[Valid][12][UTT][ MSE = 0.2110 | LCC = 0.9555 | SRCC = 0.9551 ] +INFO:root:Average loss=0.06280812719464302 +INFO:root: +[Valid][13][UTT][ MSE = 0.2036 | LCC = 0.9563 | SRCC = 0.9556 ] +INFO:root:Average loss=0.05889228507876396 +INFO:root: +[Valid][14][UTT][ MSE = 0.1937 | LCC = 0.9548 | SRCC = 0.9544 ] +INFO:root:Average loss=0.05364346653223038 +INFO:root: +[Valid][15][UTT][ MSE = 0.2197 | LCC = 0.9565 | SRCC = 0.9553 ] +INFO:root:Average loss=0.05280550816655159 +INFO:root: +[Valid][16][UTT][ MSE = 0.1906 | LCC = 0.9555 | SRCC = 0.9544 ] +INFO:root:Average loss=0.04919363752007484 +INFO:root: +[Valid][17][UTT][ MSE = 0.2163 | LCC = 0.9550 | SRCC = 0.9539 ] +INFO:root:Average loss=0.04733271205425262 +INFO:root: +[Valid][18][UTT][ MSE = 0.2216 | LCC = 0.9570 | SRCC = 0.9561 ] +INFO:root:Average loss=0.04455510634183884 +INFO:root: +[Valid][19][UTT][ MSE = 0.1933 | LCC = 0.9549 | SRCC = 0.9550 ] +INFO:root:Average loss=0.043284965977072716 +INFO:root: +[Valid][20][UTT][ MSE = 0.2119 | LCC = 0.9571 | SRCC = 0.9566 ] +INFO:root:Average loss=0.04111791896820068 +INFO:root: +[Valid][21][UTT][ MSE = 0.2123 | LCC = 0.9565 | SRCC = 0.9562 ] +INFO:root:Average loss=0.04241078513860703 +INFO:root: +[Valid][22][UTT][ MSE = 0.2186 | LCC = 0.9557 | SRCC = 0.9549 ] +INFO:root:Average loss=0.03671489106118679 +INFO:root: +[Valid][23][UTT][ MSE = 0.2309 | LCC = 0.9569 | SRCC = 0.9565 ] +INFO:root:Average loss=0.03872574009001255 +INFO:root: +[Valid][24][UTT][ MSE = 0.2069 | LCC = 0.9579 | SRCC = 0.9568 ] +INFO:root:Average loss=0.034266401648521426 +INFO:root: +[Valid][25][UTT][ MSE = 0.2272 | LCC = 0.9568 | SRCC = 0.9558 ] +INFO:root:Average loss=0.03242908878624439 +INFO:root: +[Valid][26][UTT][ MSE = 0.2097 | LCC = 0.9575 | SRCC = 0.9565 ] +INFO:root:Average loss=0.03079522776603699 +INFO:root: +[Valid][27][UTT][ MSE = 0.2151 | LCC = 0.9559 | SRCC = 0.9548 ] +INFO:root:Average loss=0.0322249783873558 +INFO:root: +[Valid][28][UTT][ MSE = 0.2083 | LCC = 0.9577 | SRCC = 0.9564 ] +INFO:root:Average loss=0.03044900159537792 +INFO:root: +[Valid][29][UTT][ MSE = 0.2036 | LCC = 0.9568 | SRCC = 0.9557 ] +INFO:root:Average loss=0.029491838589310645 +INFO:root: +[Valid][30][UTT][ MSE = 0.2039 | LCC = 0.9558 | SRCC = 0.9546 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2180 | LCC = 0.9551 | SRCC = 0.9552 ] diff --git a/results_tencent/hubert_base/layer7/run2/config.gin b/results_tencent/hubert_base/layer7/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer7/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer7/run2/model.pt b/results_tencent/hubert_base/layer7/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1db89770d77ef5814d2dc24b364939857e1e837d --- /dev/null +++ b/results_tencent/hubert_base/layer7/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7d0557411e912786b24748d2ed7637fbef098f8b32398988765875fea6dc075b +size 2747979 diff --git a/results_tencent/hubert_base/layer7/run2/model_best.pt b/results_tencent/hubert_base/layer7/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ecfee88b85975f46061c09c133cbe31435b7d73b --- /dev/null +++ b/results_tencent/hubert_base/layer7/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5a236866ca3b658e258c915da7c0e388c8161a2b58ceb4c2aea9b6456a550bab +size 2749638 diff --git a/results_tencent/hubert_base/layer7/run2/train.log b/results_tencent/hubert_base/layer7/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..887f90ecfc3f78cac13909e592596bbd1f04d9f7 --- /dev/null +++ b/results_tencent/hubert_base/layer7/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4415961595773697 +INFO:root: +[Valid][0][UTT][ MSE = 0.2905 | LCC = 0.9189 | SRCC = 0.9186 ] +INFO:root:Average loss=0.1955843676328659 +INFO:root: +[Valid][1][UTT][ MSE = 0.2316 | LCC = 0.9369 | SRCC = 0.9374 ] +INFO:root:Average loss=0.1597661380171776 +INFO:root: +[Valid][2][UTT][ MSE = 0.2373 | LCC = 0.9440 | SRCC = 0.9434 ] +INFO:root:Average loss=0.1371840539574623 +INFO:root: +[Valid][3][UTT][ MSE = 0.2258 | LCC = 0.9476 | SRCC = 0.9462 ] +INFO:root:Average loss=0.12583478808403015 +INFO:root: +[Valid][4][UTT][ MSE = 0.2029 | LCC = 0.9495 | SRCC = 0.9492 ] +INFO:root:Average loss=0.11565796700119972 +INFO:root: +[Valid][5][UTT][ MSE = 0.2063 | LCC = 0.9496 | SRCC = 0.9497 ] +INFO:root:Average loss=0.10485636216402054 +INFO:root: +[Valid][6][UTT][ MSE = 0.2101 | LCC = 0.9522 | SRCC = 0.9522 ] +INFO:root:Average loss=0.09825711476802826 +INFO:root: +[Valid][7][UTT][ MSE = 0.1798 | LCC = 0.9516 | SRCC = 0.9520 ] +INFO:root:Average loss=0.08968968188762665 +INFO:root: +[Valid][8][UTT][ MSE = 0.2045 | LCC = 0.9539 | SRCC = 0.9533 ] +INFO:root:Average loss=0.08089641299843787 +INFO:root: +[Valid][9][UTT][ MSE = 0.1746 | LCC = 0.9538 | SRCC = 0.9541 ] +INFO:root:Average loss=0.07442639634013176 +INFO:root: +[Valid][10][UTT][ MSE = 0.1743 | LCC = 0.9545 | SRCC = 0.9541 ] +INFO:root:Average loss=0.0707465327680111 +INFO:root: +[Valid][11][UTT][ MSE = 0.1694 | LCC = 0.9553 | SRCC = 0.9549 ] +INFO:root:Average loss=0.06794376775622368 +INFO:root: +[Valid][12][UTT][ MSE = 0.1915 | LCC = 0.9550 | SRCC = 0.9538 ] +INFO:root:Average loss=0.0634960914850235 +INFO:root: +[Valid][13][UTT][ MSE = 0.1765 | LCC = 0.9558 | SRCC = 0.9552 ] +INFO:root:Average loss=0.05859781953692436 +INFO:root: +[Valid][14][UTT][ MSE = 0.1705 | LCC = 0.9556 | SRCC = 0.9551 ] +INFO:root:Average loss=0.05590595382452011 +INFO:root: +[Valid][15][UTT][ MSE = 0.1703 | LCC = 0.9562 | SRCC = 0.9551 ] +INFO:root:Average loss=0.05255322228372097 +INFO:root: +[Valid][16][UTT][ MSE = 0.1637 | LCC = 0.9556 | SRCC = 0.9550 ] +INFO:root:Average loss=0.048931100338697436 +INFO:root: +[Valid][17][UTT][ MSE = 0.1603 | LCC = 0.9567 | SRCC = 0.9561 ] +INFO:root:Average loss=0.047690877318382265 +INFO:root: +[Valid][18][UTT][ MSE = 0.1510 | LCC = 0.9552 | SRCC = 0.9553 ] +INFO:root:Average loss=0.042998849719762804 +INFO:root: +[Valid][19][UTT][ MSE = 0.1726 | LCC = 0.9554 | SRCC = 0.9548 ] +INFO:root:Average loss=0.04215073348581791 +INFO:root: +[Valid][20][UTT][ MSE = 0.1700 | LCC = 0.9551 | SRCC = 0.9541 ] +INFO:root:Average loss=0.04348738625645637 +INFO:root: +[Valid][21][UTT][ MSE = 0.1724 | LCC = 0.9553 | SRCC = 0.9547 ] +INFO:root:Average loss=0.039724914968013766 +INFO:root: +[Valid][22][UTT][ MSE = 0.1653 | LCC = 0.9568 | SRCC = 0.9558 ] +INFO:root:Average loss=0.03842639607191086 +INFO:root: +[Valid][23][UTT][ MSE = 0.1725 | LCC = 0.9546 | SRCC = 0.9542 ] +INFO:root:Average loss=0.03502328358590603 +INFO:root: +[Valid][24][UTT][ MSE = 0.1709 | LCC = 0.9557 | SRCC = 0.9547 ] +INFO:root:Average loss=0.0357869471013546 +INFO:root: +[Valid][25][UTT][ MSE = 0.1723 | LCC = 0.9560 | SRCC = 0.9552 ] +INFO:root:Average loss=0.03330626989901066 +INFO:root: +[Valid][26][UTT][ MSE = 0.1849 | LCC = 0.9535 | SRCC = 0.9535 ] +INFO:root:Average loss=0.032319654032588005 +INFO:root: +[Valid][27][UTT][ MSE = 0.1811 | LCC = 0.9547 | SRCC = 0.9544 ] +INFO:root:Average loss=0.03316288481652736 +INFO:root: +[Valid][28][UTT][ MSE = 0.1766 | LCC = 0.9564 | SRCC = 0.9554 ] +INFO:root:Average loss=0.030585488311946393 +INFO:root: +[Valid][29][UTT][ MSE = 0.1759 | LCC = 0.9561 | SRCC = 0.9547 ] +INFO:root:Average loss=0.028403691306710242 +INFO:root: +[Valid][30][UTT][ MSE = 0.1718 | LCC = 0.9566 | SRCC = 0.9555 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1682 | LCC = 0.9578 | SRCC = 0.9585 ] diff --git a/results_tencent/hubert_base/layer7/run3/config.gin b/results_tencent/hubert_base/layer7/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer7/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer7/run3/model.pt b/results_tencent/hubert_base/layer7/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8314b1c1dc137a0c97eb7822621f4ee270fde9ce --- /dev/null +++ b/results_tencent/hubert_base/layer7/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f9ed12421505d0ee2d5bcdc83029d58f881cb1416ff39304f2387c8ce6c5b5a3 +size 2747979 diff --git a/results_tencent/hubert_base/layer7/run3/model_best.pt b/results_tencent/hubert_base/layer7/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f879d6063634271b8bc6b624596c415f368640e1 --- /dev/null +++ b/results_tencent/hubert_base/layer7/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6dc2b5bae9f504084fdfebd74d60b2ee5bd86d453ae2450b974c873e4d11397 +size 2749702 diff --git a/results_tencent/hubert_base/layer7/run3/train.log b/results_tencent/hubert_base/layer7/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e8baf815fcbe79494e4f72b3039c568de3590731 --- /dev/null +++ b/results_tencent/hubert_base/layer7/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4391294817924499 +INFO:root: +[Valid][0][UTT][ MSE = 0.2934 | LCC = 0.9157 | SRCC = 0.9196 ] +INFO:root:Average loss=0.19564815402030944 +INFO:root: +[Valid][1][UTT][ MSE = 0.2164 | LCC = 0.9418 | SRCC = 0.9411 ] +INFO:root:Average loss=0.15681724441051484 +INFO:root: +[Valid][2][UTT][ MSE = 0.1900 | LCC = 0.9444 | SRCC = 0.9461 ] +INFO:root:Average loss=0.13795338195562362 +INFO:root: +[Valid][3][UTT][ MSE = 0.1903 | LCC = 0.9490 | SRCC = 0.9506 ] +INFO:root:Average loss=0.12565167510509492 +INFO:root: +[Valid][4][UTT][ MSE = 0.2136 | LCC = 0.9508 | SRCC = 0.9501 ] +INFO:root:Average loss=0.11552419537305832 +INFO:root: +[Valid][5][UTT][ MSE = 0.2176 | LCC = 0.9526 | SRCC = 0.9527 ] +INFO:root:Average loss=0.10422927802801132 +INFO:root: +[Valid][6][UTT][ MSE = 0.1873 | LCC = 0.9536 | SRCC = 0.9544 ] +INFO:root:Average loss=0.09619856017827988 +INFO:root: +[Valid][7][UTT][ MSE = 0.2167 | LCC = 0.9526 | SRCC = 0.9519 ] +INFO:root:Average loss=0.08652054041624069 +INFO:root: +[Valid][8][UTT][ MSE = 0.1764 | LCC = 0.9560 | SRCC = 0.9555 ] +INFO:root:Average loss=0.08260237237811088 +INFO:root: +[Valid][9][UTT][ MSE = 0.1953 | LCC = 0.9556 | SRCC = 0.9557 ] +INFO:root:Average loss=0.0754458548128605 +INFO:root: +[Valid][10][UTT][ MSE = 0.1869 | LCC = 0.9542 | SRCC = 0.9547 ] +INFO:root:Average loss=0.07118793252110481 +INFO:root: +[Valid][11][UTT][ MSE = 0.1930 | LCC = 0.9565 | SRCC = 0.9562 ] +INFO:root:Average loss=0.06716463878750802 +INFO:root: +[Valid][12][UTT][ MSE = 0.2093 | LCC = 0.9516 | SRCC = 0.9533 ] +INFO:root:Average loss=0.06443686655163765 +INFO:root: +[Valid][13][UTT][ MSE = 0.1784 | LCC = 0.9528 | SRCC = 0.9551 ] +INFO:root:Average loss=0.0572564001083374 +INFO:root: +[Valid][14][UTT][ MSE = 0.1950 | LCC = 0.9564 | SRCC = 0.9566 ] +INFO:root:Average loss=0.052625287055969236 +INFO:root: +[Valid][15][UTT][ MSE = 0.1782 | LCC = 0.9546 | SRCC = 0.9543 ] +INFO:root:Average loss=0.04953575287759304 +INFO:root: +[Valid][16][UTT][ MSE = 0.2045 | LCC = 0.9556 | SRCC = 0.9549 ] +INFO:root:Average loss=0.047424980133771895 +INFO:root: +[Valid][17][UTT][ MSE = 0.1970 | LCC = 0.9537 | SRCC = 0.9547 ] +INFO:root:Average loss=0.04602526284754276 +INFO:root: +[Valid][18][UTT][ MSE = 0.1978 | LCC = 0.9527 | SRCC = 0.9544 ] +INFO:root:Average loss=0.043983100980520246 +INFO:root: +[Valid][19][UTT][ MSE = 0.1922 | LCC = 0.9566 | SRCC = 0.9562 ] +INFO:root:Average loss=0.0415740285217762 +INFO:root: +[Valid][20][UTT][ MSE = 0.1714 | LCC = 0.9564 | SRCC = 0.9563 ] +INFO:root:Average loss=0.041035820916295054 +INFO:root: +[Valid][21][UTT][ MSE = 0.1803 | LCC = 0.9555 | SRCC = 0.9557 ] +INFO:root:Average loss=0.03821026559174061 +INFO:root: +[Valid][22][UTT][ MSE = 0.1778 | LCC = 0.9551 | SRCC = 0.9546 ] +INFO:root:Average loss=0.036418177738785744 +INFO:root: +[Valid][23][UTT][ MSE = 0.1757 | LCC = 0.9565 | SRCC = 0.9561 ] +INFO:root:Average loss=0.03385404494404793 +INFO:root: +[Valid][24][UTT][ MSE = 0.1819 | LCC = 0.9558 | SRCC = 0.9557 ] +INFO:root:Average loss=0.03452560409903526 +INFO:root: +[Valid][25][UTT][ MSE = 0.1677 | LCC = 0.9568 | SRCC = 0.9561 ] +INFO:root:Average loss=0.03306816984713078 +INFO:root: +[Valid][26][UTT][ MSE = 0.1823 | LCC = 0.9549 | SRCC = 0.9558 ] +INFO:root:Average loss=0.031239396899938585 +INFO:root: +[Valid][27][UTT][ MSE = 0.1895 | LCC = 0.9565 | SRCC = 0.9559 ] +INFO:root:Average loss=0.03094064651429653 +INFO:root: +[Valid][28][UTT][ MSE = 0.1713 | LCC = 0.9553 | SRCC = 0.9561 ] +INFO:root:Average loss=0.03042952747642994 +INFO:root: +[Valid][29][UTT][ MSE = 0.2128 | LCC = 0.9565 | SRCC = 0.9559 ] +INFO:root:Average loss=0.028207323640584946 +INFO:root: +[Valid][30][UTT][ MSE = 0.1917 | LCC = 0.9557 | SRCC = 0.9556 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1763 | LCC = 0.9534 | SRCC = 0.9536 ] diff --git a/results_tencent/hubert_base/layer7/run4/config.gin b/results_tencent/hubert_base/layer7/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer7/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer7/run4/model.pt b/results_tencent/hubert_base/layer7/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5792f2bcebec6d34721e2c43e69f830828c6e6ca --- /dev/null +++ b/results_tencent/hubert_base/layer7/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:86966df2ee2c29d449025bed8aead143195d98b593192c0351af7827ff9068f2 +size 2747979 diff --git a/results_tencent/hubert_base/layer7/run4/model_best.pt b/results_tencent/hubert_base/layer7/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..cf57bf97ef7a96e8b0afd106e06362d2052093ae --- /dev/null +++ b/results_tencent/hubert_base/layer7/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d7a628de70b28960b62ab70074926a371ba762190be079c58b648c0828a83ec +size 2749702 diff --git a/results_tencent/hubert_base/layer7/run4/train.log b/results_tencent/hubert_base/layer7/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..78c460bded4268fa5323c6635ae691f70c3453b1 --- /dev/null +++ b/results_tencent/hubert_base/layer7/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46731732189655306 +INFO:root: +[Valid][0][UTT][ MSE = 0.2826 | LCC = 0.9168 | SRCC = 0.9150 ] +INFO:root:Average loss=0.2047037230730057 +INFO:root: +[Valid][1][UTT][ MSE = 0.2316 | LCC = 0.9325 | SRCC = 0.9307 ] +INFO:root:Average loss=0.1652503855228424 +INFO:root: +[Valid][2][UTT][ MSE = 0.2156 | LCC = 0.9422 | SRCC = 0.9408 ] +INFO:root:Average loss=0.14013947534561158 +INFO:root: +[Valid][3][UTT][ MSE = 0.2084 | LCC = 0.9472 | SRCC = 0.9471 ] +INFO:root:Average loss=0.1303021105527878 +INFO:root: +[Valid][4][UTT][ MSE = 0.1934 | LCC = 0.9489 | SRCC = 0.9483 ] +INFO:root:Average loss=0.11675053012371063 +INFO:root: +[Valid][5][UTT][ MSE = 0.2256 | LCC = 0.9525 | SRCC = 0.9521 ] +INFO:root:Average loss=0.10543711680173874 +INFO:root: +[Valid][6][UTT][ MSE = 0.2017 | LCC = 0.9506 | SRCC = 0.9495 ] +INFO:root:Average loss=0.09709864541888237 +INFO:root: +[Valid][7][UTT][ MSE = 0.1995 | LCC = 0.9518 | SRCC = 0.9511 ] +INFO:root:Average loss=0.08984467133879662 +INFO:root: +[Valid][8][UTT][ MSE = 0.2105 | LCC = 0.9545 | SRCC = 0.9536 ] +INFO:root:Average loss=0.08196622487902641 +INFO:root: +[Valid][9][UTT][ MSE = 0.1943 | LCC = 0.9524 | SRCC = 0.9516 ] +INFO:root:Average loss=0.07746868342161178 +INFO:root: +[Valid][10][UTT][ MSE = 0.1879 | LCC = 0.9536 | SRCC = 0.9530 ] +INFO:root:Average loss=0.07456634026765824 +INFO:root: +[Valid][11][UTT][ MSE = 0.2105 | LCC = 0.9552 | SRCC = 0.9547 ] +INFO:root:Average loss=0.06640736156702041 +INFO:root: +[Valid][12][UTT][ MSE = 0.2020 | LCC = 0.9530 | SRCC = 0.9525 ] +INFO:root:Average loss=0.06519628220796585 +INFO:root: +[Valid][13][UTT][ MSE = 0.2538 | LCC = 0.9542 | SRCC = 0.9545 ] +INFO:root:Average loss=0.058726285964250564 +INFO:root: +[Valid][14][UTT][ MSE = 0.1871 | LCC = 0.9546 | SRCC = 0.9538 ] +INFO:root:Average loss=0.05649688503146172 +INFO:root: +[Valid][15][UTT][ MSE = 0.1983 | LCC = 0.9550 | SRCC = 0.9543 ] +INFO:root:Average loss=0.05476478682458401 +INFO:root: +[Valid][16][UTT][ MSE = 0.1906 | LCC = 0.9545 | SRCC = 0.9533 ] +INFO:root:Average loss=0.05061878323554993 +INFO:root: +[Valid][17][UTT][ MSE = 0.1908 | LCC = 0.9548 | SRCC = 0.9538 ] +INFO:root:Average loss=0.046200878024101256 +INFO:root: +[Valid][18][UTT][ MSE = 0.1770 | LCC = 0.9551 | SRCC = 0.9547 ] +INFO:root:Average loss=0.04805988961458206 +INFO:root: +[Valid][19][UTT][ MSE = 0.2173 | LCC = 0.9555 | SRCC = 0.9551 ] +INFO:root:Average loss=0.04469612504541874 +INFO:root: +[Valid][20][UTT][ MSE = 0.1850 | LCC = 0.9541 | SRCC = 0.9533 ] +INFO:root:Average loss=0.04281171567738056 +INFO:root: +[Valid][21][UTT][ MSE = 0.1831 | LCC = 0.9566 | SRCC = 0.9556 ] +INFO:root:Average loss=0.040705157414078715 +INFO:root: +[Valid][22][UTT][ MSE = 0.1731 | LCC = 0.9554 | SRCC = 0.9547 ] +INFO:root:Average loss=0.03912829069793224 +INFO:root: +[Valid][23][UTT][ MSE = 0.2096 | LCC = 0.9545 | SRCC = 0.9538 ] +INFO:root:Average loss=0.0372146637737751 +INFO:root: +[Valid][24][UTT][ MSE = 0.1729 | LCC = 0.9551 | SRCC = 0.9542 ] +INFO:root:Average loss=0.03580027441680431 +INFO:root: +[Valid][25][UTT][ MSE = 0.1825 | LCC = 0.9559 | SRCC = 0.9550 ] +INFO:root:Average loss=0.03427510191500187 +INFO:root: +[Valid][26][UTT][ MSE = 0.1767 | LCC = 0.9560 | SRCC = 0.9551 ] +INFO:root:Average loss=0.033660008400678634 +INFO:root: +[Valid][27][UTT][ MSE = 0.1772 | LCC = 0.9562 | SRCC = 0.9549 ] +INFO:root:Average loss=0.0322692047059536 +INFO:root: +[Valid][28][UTT][ MSE = 0.1744 | LCC = 0.9558 | SRCC = 0.9549 ] +INFO:root:Average loss=0.03159766614437103 +INFO:root: +[Valid][29][UTT][ MSE = 0.1743 | LCC = 0.9520 | SRCC = 0.9522 ] +INFO:root:Average loss=0.03212705464661121 +INFO:root: +[Valid][30][UTT][ MSE = 0.1900 | LCC = 0.9560 | SRCC = 0.9548 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1943 | LCC = 0.9551 | SRCC = 0.9557 ] diff --git a/results_tencent/hubert_base/layer8/run0/config.gin b/results_tencent/hubert_base/layer8/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer8/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer8/run0/model.pt b/results_tencent/hubert_base/layer8/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..28601b01a3be258b2e8fa321dee6bcd07eb70adf --- /dev/null +++ b/results_tencent/hubert_base/layer8/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:10b922c658c87239d418cc8561f862ca328cba864039b6cad25a38e585e734f8 +size 2747979 diff --git a/results_tencent/hubert_base/layer8/run0/model_best.pt b/results_tencent/hubert_base/layer8/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7e00673351563837f8206f535c236a4d5c61adfc --- /dev/null +++ b/results_tencent/hubert_base/layer8/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9913eb02e82090d01751b940971bb2dbc29936520875dff570515c7240cb3b04 +size 2749702 diff --git a/results_tencent/hubert_base/layer8/run0/train.log b/results_tencent/hubert_base/layer8/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..09ee339279e3fc662a34b4f6e413900330988747 --- /dev/null +++ b/results_tencent/hubert_base/layer8/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4663961594104767 +INFO:root: +[Valid][0][UTT][ MSE = 0.2961 | LCC = 0.9265 | SRCC = 0.9266 ] +INFO:root:Average loss=0.19447220706939697 +INFO:root: +[Valid][1][UTT][ MSE = 0.2411 | LCC = 0.9401 | SRCC = 0.9406 ] +INFO:root:Average loss=0.1578968061208725 +INFO:root: +[Valid][2][UTT][ MSE = 0.2016 | LCC = 0.9460 | SRCC = 0.9463 ] +INFO:root:Average loss=0.14024895823001862 +INFO:root: +[Valid][3][UTT][ MSE = 0.2104 | LCC = 0.9495 | SRCC = 0.9502 ] +INFO:root:Average loss=0.1272516987323761 +INFO:root: +[Valid][4][UTT][ MSE = 0.2048 | LCC = 0.9517 | SRCC = 0.9519 ] +INFO:root:Average loss=0.10956569343805313 +INFO:root: +[Valid][5][UTT][ MSE = 0.2207 | LCC = 0.9537 | SRCC = 0.9540 ] +INFO:root:Average loss=0.10328154814243316 +INFO:root: +[Valid][6][UTT][ MSE = 0.1969 | LCC = 0.9547 | SRCC = 0.9549 ] +INFO:root:Average loss=0.09882571813464165 +INFO:root: +[Valid][7][UTT][ MSE = 0.1774 | LCC = 0.9545 | SRCC = 0.9551 ] +INFO:root:Average loss=0.088814786195755 +INFO:root: +[Valid][8][UTT][ MSE = 0.2097 | LCC = 0.9542 | SRCC = 0.9543 ] +INFO:root:Average loss=0.08160611346364022 +INFO:root: +[Valid][9][UTT][ MSE = 0.1923 | LCC = 0.9550 | SRCC = 0.9549 ] +INFO:root:Average loss=0.0755985603928566 +INFO:root: +[Valid][10][UTT][ MSE = 0.1852 | LCC = 0.9559 | SRCC = 0.9555 ] +INFO:root:Average loss=0.07043744254112244 +INFO:root: +[Valid][11][UTT][ MSE = 0.1993 | LCC = 0.9559 | SRCC = 0.9553 ] +INFO:root:Average loss=0.06776206141710281 +INFO:root: +[Valid][12][UTT][ MSE = 0.1924 | LCC = 0.9553 | SRCC = 0.9548 ] +INFO:root:Average loss=0.06424590033292771 +INFO:root: +[Valid][13][UTT][ MSE = 0.2072 | LCC = 0.9549 | SRCC = 0.9546 ] +INFO:root:Average loss=0.05955948308110237 +INFO:root: +[Valid][14][UTT][ MSE = 0.1906 | LCC = 0.9564 | SRCC = 0.9562 ] +INFO:root:Average loss=0.05592438313364983 +INFO:root: +[Valid][15][UTT][ MSE = 0.2189 | LCC = 0.9569 | SRCC = 0.9569 ] +INFO:root:Average loss=0.05342322885990143 +INFO:root: +[Valid][16][UTT][ MSE = 0.1812 | LCC = 0.9568 | SRCC = 0.9564 ] +INFO:root:Average loss=0.04808048352599144 +INFO:root: +[Valid][17][UTT][ MSE = 0.1845 | LCC = 0.9565 | SRCC = 0.9561 ] +INFO:root:Average loss=0.04539513470232487 +INFO:root: +[Valid][18][UTT][ MSE = 0.1731 | LCC = 0.9558 | SRCC = 0.9557 ] +INFO:root:Average loss=0.043561947286129 +INFO:root: +[Valid][19][UTT][ MSE = 0.1867 | LCC = 0.9560 | SRCC = 0.9557 ] +INFO:root:Average loss=0.04194634613394737 +INFO:root: +[Valid][20][UTT][ MSE = 0.1730 | LCC = 0.9554 | SRCC = 0.9552 ] +INFO:root:Average loss=0.041328808963298796 +INFO:root: +[Valid][21][UTT][ MSE = 0.1795 | LCC = 0.9560 | SRCC = 0.9555 ] +INFO:root:Average loss=0.03891950660943985 +INFO:root: +[Valid][22][UTT][ MSE = 0.1774 | LCC = 0.9563 | SRCC = 0.9558 ] +INFO:root:Average loss=0.03971857899427414 +INFO:root: +[Valid][23][UTT][ MSE = 0.1769 | LCC = 0.9561 | SRCC = 0.9559 ] +INFO:root:Average loss=0.034392075568437576 +INFO:root: +[Valid][24][UTT][ MSE = 0.1754 | LCC = 0.9572 | SRCC = 0.9567 ] +INFO:root:Average loss=0.03655870643258095 +INFO:root: +[Valid][25][UTT][ MSE = 0.1818 | LCC = 0.9567 | SRCC = 0.9562 ] +INFO:root:Average loss=0.033906725868582725 +INFO:root: +[Valid][26][UTT][ MSE = 0.1738 | LCC = 0.9568 | SRCC = 0.9562 ] +INFO:root:Average loss=0.03180735279619694 +INFO:root: +[Valid][27][UTT][ MSE = 0.1919 | LCC = 0.9563 | SRCC = 0.9557 ] +INFO:root:Average loss=0.029887414023280143 +INFO:root: +[Valid][28][UTT][ MSE = 0.1727 | LCC = 0.9567 | SRCC = 0.9563 ] +INFO:root:Average loss=0.03003402154147625 +INFO:root: +[Valid][29][UTT][ MSE = 0.1858 | LCC = 0.9573 | SRCC = 0.9567 ] +INFO:root:Average loss=0.02864827439188957 +INFO:root: +[Valid][30][UTT][ MSE = 0.1911 | LCC = 0.9578 | SRCC = 0.9570 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2093 | LCC = 0.9550 | SRCC = 0.9550 ] diff --git a/results_tencent/hubert_base/layer8/run1/config.gin b/results_tencent/hubert_base/layer8/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer8/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer8/run1/model.pt b/results_tencent/hubert_base/layer8/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4b892a2f63e0d48720e559007c52f15378cc130e --- /dev/null +++ b/results_tencent/hubert_base/layer8/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f22ce421696a41b74568a7e52e7b600d6e78ca0a0a1595db3e6c2298ef5e5f69 +size 2747979 diff --git a/results_tencent/hubert_base/layer8/run1/model_best.pt b/results_tencent/hubert_base/layer8/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..80397ac7b254cfa91b8f77a51976884858a9b3ce --- /dev/null +++ b/results_tencent/hubert_base/layer8/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:31804e406a2bb77caffeb84f391b617388ae5d0ceaf66bae9a6a9bb28853d213 +size 2749702 diff --git a/results_tencent/hubert_base/layer8/run1/train.log b/results_tencent/hubert_base/layer8/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e8714df66ed419f075df8e73c8a45497c7567ea2 --- /dev/null +++ b/results_tencent/hubert_base/layer8/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.518630673289299 +INFO:root: +[Valid][0][UTT][ MSE = 0.3022 | LCC = 0.9196 | SRCC = 0.9169 ] +INFO:root:Average loss=0.20503818786144257 +INFO:root: +[Valid][1][UTT][ MSE = 0.2486 | LCC = 0.9398 | SRCC = 0.9387 ] +INFO:root:Average loss=0.16033215987682342 +INFO:root: +[Valid][2][UTT][ MSE = 0.2023 | LCC = 0.9439 | SRCC = 0.9447 ] +INFO:root:Average loss=0.13764582949876786 +INFO:root: +[Valid][3][UTT][ MSE = 0.2162 | LCC = 0.9487 | SRCC = 0.9481 ] +INFO:root:Average loss=0.12447013837099076 +INFO:root: +[Valid][4][UTT][ MSE = 0.2087 | LCC = 0.9512 | SRCC = 0.9507 ] +INFO:root:Average loss=0.11401066589355469 +INFO:root: +[Valid][5][UTT][ MSE = 0.2072 | LCC = 0.9524 | SRCC = 0.9524 ] +INFO:root:Average loss=0.10385237804055214 +INFO:root: +[Valid][6][UTT][ MSE = 0.2092 | LCC = 0.9527 | SRCC = 0.9524 ] +INFO:root:Average loss=0.09706803920865059 +INFO:root: +[Valid][7][UTT][ MSE = 0.1822 | LCC = 0.9535 | SRCC = 0.9534 ] +INFO:root:Average loss=0.09047950148582458 +INFO:root: +[Valid][8][UTT][ MSE = 0.1836 | LCC = 0.9546 | SRCC = 0.9542 ] +INFO:root:Average loss=0.08417674934864045 +INFO:root: +[Valid][9][UTT][ MSE = 0.1995 | LCC = 0.9551 | SRCC = 0.9543 ] +INFO:root:Average loss=0.0768951720893383 +INFO:root: +[Valid][10][UTT][ MSE = 0.1921 | LCC = 0.9532 | SRCC = 0.9525 ] +INFO:root:Average loss=0.07196023371815681 +INFO:root: +[Valid][11][UTT][ MSE = 0.1944 | LCC = 0.9549 | SRCC = 0.9545 ] +INFO:root:Average loss=0.0686198719739914 +INFO:root: +[Valid][12][UTT][ MSE = 0.2099 | LCC = 0.9548 | SRCC = 0.9544 ] +INFO:root:Average loss=0.06329396867752075 +INFO:root: +[Valid][13][UTT][ MSE = 0.1866 | LCC = 0.9540 | SRCC = 0.9531 ] +INFO:root:Average loss=0.060069068223237994 +INFO:root: +[Valid][14][UTT][ MSE = 0.1749 | LCC = 0.9545 | SRCC = 0.9538 ] +INFO:root:Average loss=0.05463128207623959 +INFO:root: +[Valid][15][UTT][ MSE = 0.1940 | LCC = 0.9550 | SRCC = 0.9536 ] +INFO:root:Average loss=0.05361278021335602 +INFO:root: +[Valid][16][UTT][ MSE = 0.2190 | LCC = 0.9547 | SRCC = 0.9540 ] +INFO:root:Average loss=0.049242006346583364 +INFO:root: +[Valid][17][UTT][ MSE = 0.2170 | LCC = 0.9547 | SRCC = 0.9541 ] +INFO:root:Average loss=0.04583188383281231 +INFO:root: +[Valid][18][UTT][ MSE = 0.2262 | LCC = 0.9548 | SRCC = 0.9543 ] +INFO:root:Average loss=0.044320904284715655 +INFO:root: +[Valid][19][UTT][ MSE = 0.1839 | LCC = 0.9552 | SRCC = 0.9538 ] +INFO:root:Average loss=0.043092190742492674 +INFO:root: +[Valid][20][UTT][ MSE = 0.2023 | LCC = 0.9543 | SRCC = 0.9529 ] +INFO:root:Average loss=0.04400700005888939 +INFO:root: +[Valid][21][UTT][ MSE = 0.1857 | LCC = 0.9562 | SRCC = 0.9548 ] +INFO:root:Average loss=0.039741125538945196 +INFO:root: +[Valid][22][UTT][ MSE = 0.1669 | LCC = 0.9529 | SRCC = 0.9524 ] +INFO:root:Average loss=0.03688927900791168 +INFO:root: +[Valid][23][UTT][ MSE = 0.1800 | LCC = 0.9560 | SRCC = 0.9545 ] +INFO:root:Average loss=0.03670138043165207 +INFO:root: +[Valid][24][UTT][ MSE = 0.1812 | LCC = 0.9560 | SRCC = 0.9544 ] +INFO:root:Average loss=0.033954132318496706 +INFO:root: +[Valid][25][UTT][ MSE = 0.1873 | LCC = 0.9560 | SRCC = 0.9546 ] +INFO:root:Average loss=0.03389924629032612 +INFO:root: +[Valid][26][UTT][ MSE = 0.1820 | LCC = 0.9562 | SRCC = 0.9551 ] +INFO:root:Average loss=0.030851650178432465 +INFO:root: +[Valid][27][UTT][ MSE = 0.1960 | LCC = 0.9567 | SRCC = 0.9552 ] +INFO:root:Average loss=0.031804328069090844 +INFO:root: +[Valid][28][UTT][ MSE = 0.1892 | LCC = 0.9560 | SRCC = 0.9543 ] +INFO:root:Average loss=0.031187222793698312 +INFO:root: +[Valid][29][UTT][ MSE = 0.1941 | LCC = 0.9550 | SRCC = 0.9534 ] +INFO:root:Average loss=0.03039980163425207 +INFO:root: +[Valid][30][UTT][ MSE = 0.1769 | LCC = 0.9552 | SRCC = 0.9534 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2065 | LCC = 0.9560 | SRCC = 0.9559 ] diff --git a/results_tencent/hubert_base/layer8/run2/config.gin b/results_tencent/hubert_base/layer8/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer8/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer8/run2/model.pt b/results_tencent/hubert_base/layer8/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c1e06b1d4f6e51a08950114f7988c166bcea3eff --- /dev/null +++ b/results_tencent/hubert_base/layer8/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c822a794791fa7e009698d7909d99556ec73a0ab2ac026edd5b90c16162dabfc +size 2747979 diff --git a/results_tencent/hubert_base/layer8/run2/model_best.pt b/results_tencent/hubert_base/layer8/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e651afbdf3eacfd473e532512130cd036c88f346 --- /dev/null +++ b/results_tencent/hubert_base/layer8/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e288e83ad00cb9a69212b6b2ce18fa7cf3a5c7f6f2dcaf4f1e58391babba169 +size 2749702 diff --git a/results_tencent/hubert_base/layer8/run2/train.log b/results_tencent/hubert_base/layer8/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2a8909fd8a2a5c8f104fe94061271add72838405 --- /dev/null +++ b/results_tencent/hubert_base/layer8/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4679989229440689 +INFO:root: +[Valid][0][UTT][ MSE = 0.2959 | LCC = 0.9233 | SRCC = 0.9229 ] +INFO:root:Average loss=0.192352716088295 +INFO:root: +[Valid][1][UTT][ MSE = 0.2572 | LCC = 0.9389 | SRCC = 0.9382 ] +INFO:root:Average loss=0.15668708783388138 +INFO:root: +[Valid][2][UTT][ MSE = 0.2320 | LCC = 0.9443 | SRCC = 0.9439 ] +INFO:root:Average loss=0.13768919610977173 +INFO:root: +[Valid][3][UTT][ MSE = 0.2342 | LCC = 0.9499 | SRCC = 0.9502 ] +INFO:root:Average loss=0.12218249440193177 +INFO:root: +[Valid][4][UTT][ MSE = 0.2348 | LCC = 0.9497 | SRCC = 0.9501 ] +INFO:root:Average loss=0.10773323714733124 +INFO:root: +[Valid][5][UTT][ MSE = 0.2097 | LCC = 0.9523 | SRCC = 0.9524 ] +INFO:root:Average loss=0.1072228812277317 +INFO:root: +[Valid][6][UTT][ MSE = 0.2185 | LCC = 0.9536 | SRCC = 0.9538 ] +INFO:root:Average loss=0.09602200922369956 +INFO:root: +[Valid][7][UTT][ MSE = 0.2079 | LCC = 0.9538 | SRCC = 0.9534 ] +INFO:root:Average loss=0.09024425655603409 +INFO:root: +[Valid][8][UTT][ MSE = 0.1887 | LCC = 0.9545 | SRCC = 0.9541 ] +INFO:root:Average loss=0.0818274610042572 +INFO:root: +[Valid][9][UTT][ MSE = 0.2211 | LCC = 0.9546 | SRCC = 0.9541 ] +INFO:root:Average loss=0.07729345923662186 +INFO:root: +[Valid][10][UTT][ MSE = 0.2268 | LCC = 0.9549 | SRCC = 0.9550 ] +INFO:root:Average loss=0.07546352049708366 +INFO:root: +[Valid][11][UTT][ MSE = 0.1915 | LCC = 0.9560 | SRCC = 0.9552 ] +INFO:root:Average loss=0.06733068203926086 +INFO:root: +[Valid][12][UTT][ MSE = 0.2008 | LCC = 0.9529 | SRCC = 0.9521 ] +INFO:root:Average loss=0.06235240095853806 +INFO:root: +[Valid][13][UTT][ MSE = 0.1952 | LCC = 0.9552 | SRCC = 0.9540 ] +INFO:root:Average loss=0.06097330138087273 +INFO:root: +[Valid][14][UTT][ MSE = 0.2295 | LCC = 0.9517 | SRCC = 0.9510 ] +INFO:root:Average loss=0.05518241503834725 +INFO:root: +[Valid][15][UTT][ MSE = 0.1834 | LCC = 0.9552 | SRCC = 0.9542 ] +INFO:root:Average loss=0.05304531915485859 +INFO:root: +[Valid][16][UTT][ MSE = 0.1884 | LCC = 0.9535 | SRCC = 0.9521 ] +INFO:root:Average loss=0.04763919745385647 +INFO:root: +[Valid][17][UTT][ MSE = 0.2049 | LCC = 0.9547 | SRCC = 0.9533 ] +INFO:root:Average loss=0.044720103263854984 +INFO:root: +[Valid][18][UTT][ MSE = 0.1879 | LCC = 0.9553 | SRCC = 0.9541 ] +INFO:root:Average loss=0.04253147074580193 +INFO:root: +[Valid][19][UTT][ MSE = 0.1806 | LCC = 0.9548 | SRCC = 0.9540 ] +INFO:root:Average loss=0.03992520460486412 +INFO:root: +[Valid][20][UTT][ MSE = 0.2002 | LCC = 0.9549 | SRCC = 0.9538 ] +INFO:root:Average loss=0.0396186575293541 +INFO:root: +[Valid][21][UTT][ MSE = 0.2015 | LCC = 0.9553 | SRCC = 0.9544 ] +INFO:root:Average loss=0.03880732327699661 +INFO:root: +[Valid][22][UTT][ MSE = 0.2031 | LCC = 0.9564 | SRCC = 0.9559 ] +INFO:root:Average loss=0.038054704263806347 +INFO:root: +[Valid][23][UTT][ MSE = 0.2132 | LCC = 0.9555 | SRCC = 0.9550 ] +INFO:root:Average loss=0.036402039140462875 +INFO:root: +[Valid][24][UTT][ MSE = 0.1866 | LCC = 0.9561 | SRCC = 0.9550 ] +INFO:root:Average loss=0.032984008938074114 +INFO:root: +[Valid][25][UTT][ MSE = 0.1907 | LCC = 0.9574 | SRCC = 0.9562 ] +INFO:root:Average loss=0.03259932093322277 +INFO:root: +[Valid][26][UTT][ MSE = 0.1985 | LCC = 0.9569 | SRCC = 0.9554 ] +INFO:root:Average loss=0.0329017114341259 +INFO:root: +[Valid][27][UTT][ MSE = 0.1821 | LCC = 0.9566 | SRCC = 0.9554 ] +INFO:root:Average loss=0.030694642826914788 +INFO:root: +[Valid][28][UTT][ MSE = 0.2015 | LCC = 0.9555 | SRCC = 0.9543 ] +INFO:root:Average loss=0.028482242733240126 +INFO:root: +[Valid][29][UTT][ MSE = 0.1846 | LCC = 0.9570 | SRCC = 0.9558 ] +INFO:root:Average loss=0.02797266606986523 +INFO:root: +[Valid][30][UTT][ MSE = 0.2019 | LCC = 0.9561 | SRCC = 0.9547 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1970 | LCC = 0.9570 | SRCC = 0.9567 ] diff --git a/results_tencent/hubert_base/layer8/run3/config.gin b/results_tencent/hubert_base/layer8/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer8/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer8/run3/model.pt b/results_tencent/hubert_base/layer8/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9ef4e0a6ce13e2528ce16e7d013344b6eda1f8bf --- /dev/null +++ b/results_tencent/hubert_base/layer8/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5ac24cf237544859b5f9c22ae7ced85810087d2272000e4d91a84baeba81c999 +size 2747979 diff --git a/results_tencent/hubert_base/layer8/run3/model_best.pt b/results_tencent/hubert_base/layer8/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4507fb357e60c40cd7f1024ef7330028b41946e0 --- /dev/null +++ b/results_tencent/hubert_base/layer8/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:632451ec95f0020aebcd95c576e259b9956a08c8afe67bef3dac30fbb0823199 +size 2749702 diff --git a/results_tencent/hubert_base/layer8/run3/train.log b/results_tencent/hubert_base/layer8/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a92394fa1262bada45457e7622d737d630fc0210 --- /dev/null +++ b/results_tencent/hubert_base/layer8/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.45217075157165526 +INFO:root: +[Valid][0][UTT][ MSE = 0.2706 | LCC = 0.9257 | SRCC = 0.9246 ] +INFO:root:Average loss=0.1912469607591629 +INFO:root: +[Valid][1][UTT][ MSE = 0.2265 | LCC = 0.9421 | SRCC = 0.9422 ] +INFO:root:Average loss=0.1562147473692894 +INFO:root: +[Valid][2][UTT][ MSE = 0.2392 | LCC = 0.9479 | SRCC = 0.9482 ] +INFO:root:Average loss=0.1348691178560257 +INFO:root: +[Valid][3][UTT][ MSE = 0.1916 | LCC = 0.9507 | SRCC = 0.9509 ] +INFO:root:Average loss=0.12074754685163498 +INFO:root: +[Valid][4][UTT][ MSE = 0.2071 | LCC = 0.9524 | SRCC = 0.9519 ] +INFO:root:Average loss=0.10921972411870956 +INFO:root: +[Valid][5][UTT][ MSE = 0.2318 | LCC = 0.9513 | SRCC = 0.9517 ] +INFO:root:Average loss=0.10088136139512062 +INFO:root: +[Valid][6][UTT][ MSE = 0.1995 | LCC = 0.9519 | SRCC = 0.9529 ] +INFO:root:Average loss=0.09485620650649071 +INFO:root: +[Valid][7][UTT][ MSE = 0.2083 | LCC = 0.9515 | SRCC = 0.9522 ] +INFO:root:Average loss=0.08661850732564925 +INFO:root: +[Valid][8][UTT][ MSE = 0.2053 | LCC = 0.9544 | SRCC = 0.9546 ] +INFO:root:Average loss=0.08206195911765099 +INFO:root: +[Valid][9][UTT][ MSE = 0.2351 | LCC = 0.9530 | SRCC = 0.9530 ] +INFO:root:Average loss=0.07639610838890076 +INFO:root: +[Valid][10][UTT][ MSE = 0.2194 | LCC = 0.9510 | SRCC = 0.9505 ] +INFO:root:Average loss=0.0709100875556469 +INFO:root: +[Valid][11][UTT][ MSE = 0.1907 | LCC = 0.9520 | SRCC = 0.9515 ] +INFO:root:Average loss=0.06565909433364868 +INFO:root: +[Valid][12][UTT][ MSE = 0.1769 | LCC = 0.9560 | SRCC = 0.9551 ] +INFO:root:Average loss=0.0617673990726471 +INFO:root: +[Valid][13][UTT][ MSE = 0.1932 | LCC = 0.9546 | SRCC = 0.9538 ] +INFO:root:Average loss=0.060715622335672376 +INFO:root: +[Valid][14][UTT][ MSE = 0.1866 | LCC = 0.9503 | SRCC = 0.9523 ] +INFO:root:Average loss=0.053723014518618585 +INFO:root: +[Valid][15][UTT][ MSE = 0.1909 | LCC = 0.9526 | SRCC = 0.9522 ] +INFO:root:Average loss=0.05216685411334038 +INFO:root: +[Valid][16][UTT][ MSE = 0.1818 | LCC = 0.9560 | SRCC = 0.9555 ] +INFO:root:Average loss=0.05081378179788589 +INFO:root: +[Valid][17][UTT][ MSE = 0.1996 | LCC = 0.9562 | SRCC = 0.9550 ] +INFO:root:Average loss=0.045693426191806795 +INFO:root: +[Valid][18][UTT][ MSE = 0.2361 | LCC = 0.9537 | SRCC = 0.9533 ] +INFO:root:Average loss=0.047957686930894854 +INFO:root: +[Valid][19][UTT][ MSE = 0.1949 | LCC = 0.9528 | SRCC = 0.9524 ] +INFO:root:Average loss=0.04373168759047985 +INFO:root: +[Valid][20][UTT][ MSE = 0.1752 | LCC = 0.9539 | SRCC = 0.9532 ] +INFO:root:Average loss=0.04102198785543442 +INFO:root: +[Valid][21][UTT][ MSE = 0.1901 | LCC = 0.9554 | SRCC = 0.9541 ] +INFO:root:Average loss=0.04173639951646328 +INFO:root: +[Valid][22][UTT][ MSE = 0.1919 | LCC = 0.9539 | SRCC = 0.9537 ] +INFO:root:Average loss=0.03889865100383758 +INFO:root: +[Valid][23][UTT][ MSE = 0.1831 | LCC = 0.9528 | SRCC = 0.9534 ] +INFO:root:Average loss=0.03601724748313427 +INFO:root: +[Valid][24][UTT][ MSE = 0.1749 | LCC = 0.9558 | SRCC = 0.9554 ] +INFO:root:Average loss=0.03509097960591316 +INFO:root: +[Valid][25][UTT][ MSE = 0.1931 | LCC = 0.9562 | SRCC = 0.9556 ] +INFO:root:Average loss=0.03723022292554379 +INFO:root: +[Valid][26][UTT][ MSE = 0.2143 | LCC = 0.9516 | SRCC = 0.9507 ] +INFO:root:Average loss=0.0352638473212719 +INFO:root: +[Valid][27][UTT][ MSE = 0.1907 | LCC = 0.9554 | SRCC = 0.9550 ] +INFO:root:Average loss=0.03318488036096096 +INFO:root: +[Valid][28][UTT][ MSE = 0.1994 | LCC = 0.9560 | SRCC = 0.9548 ] +INFO:root:Average loss=0.03312614426016808 +INFO:root: +[Valid][29][UTT][ MSE = 0.1940 | LCC = 0.9576 | SRCC = 0.9567 ] +INFO:root:Average loss=0.02972747804224491 +INFO:root: +[Valid][30][UTT][ MSE = 0.2027 | LCC = 0.9552 | SRCC = 0.9552 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2041 | LCC = 0.9544 | SRCC = 0.9544 ] diff --git a/results_tencent/hubert_base/layer8/run4/config.gin b/results_tencent/hubert_base/layer8/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer8/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer8/run4/model.pt b/results_tencent/hubert_base/layer8/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..95e206e96ed1f4414fdd9598a8cb19d6232fbba9 --- /dev/null +++ b/results_tencent/hubert_base/layer8/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b067fd6ad5e6f3db897340650d98230000bc97eb77531546f2a5fc5338cc3be4 +size 2747979 diff --git a/results_tencent/hubert_base/layer8/run4/model_best.pt b/results_tencent/hubert_base/layer8/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6353553d85c04744e21a67602c8bd00034b7e56e --- /dev/null +++ b/results_tencent/hubert_base/layer8/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:12f84a88f997d87ca78126a99868bc909d6723a6d4b61687d664a4f215ec5493 +size 2749702 diff --git a/results_tencent/hubert_base/layer8/run4/train.log b/results_tencent/hubert_base/layer8/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e6336a470213a425ffb6bafdbcd4964745b515f7 --- /dev/null +++ b/results_tencent/hubert_base/layer8/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.49778365933895113 +INFO:root: +[Valid][0][UTT][ MSE = 0.2803 | LCC = 0.9231 | SRCC = 0.9247 ] +INFO:root:Average loss=0.1876319111585617 +INFO:root: +[Valid][1][UTT][ MSE = 0.2210 | LCC = 0.9424 | SRCC = 0.9417 ] +INFO:root:Average loss=0.15403024053573608 +INFO:root: +[Valid][2][UTT][ MSE = 0.2261 | LCC = 0.9473 | SRCC = 0.9462 ] +INFO:root:Average loss=0.13674504452943803 +INFO:root: +[Valid][3][UTT][ MSE = 0.2307 | LCC = 0.9516 | SRCC = 0.9510 ] +INFO:root:Average loss=0.12052827125787735 +INFO:root: +[Valid][4][UTT][ MSE = 0.2044 | LCC = 0.9527 | SRCC = 0.9522 ] +INFO:root:Average loss=0.10921627888083459 +INFO:root: +[Valid][5][UTT][ MSE = 0.2110 | LCC = 0.9532 | SRCC = 0.9531 ] +INFO:root:Average loss=0.10356076326966286 +INFO:root: +[Valid][6][UTT][ MSE = 0.1943 | LCC = 0.9528 | SRCC = 0.9518 ] +INFO:root:Average loss=0.09606406143307686 +INFO:root: +[Valid][7][UTT][ MSE = 0.2390 | LCC = 0.9527 | SRCC = 0.9520 ] +INFO:root:Average loss=0.08961870783567429 +INFO:root: +[Valid][8][UTT][ MSE = 0.1934 | LCC = 0.9514 | SRCC = 0.9508 ] +INFO:root:Average loss=0.08016585499048233 +INFO:root: +[Valid][9][UTT][ MSE = 0.1891 | LCC = 0.9556 | SRCC = 0.9545 ] +INFO:root:Average loss=0.07568605738878251 +INFO:root: +[Valid][10][UTT][ MSE = 0.2042 | LCC = 0.9561 | SRCC = 0.9547 ] +INFO:root:Average loss=0.06874957889318466 +INFO:root: +[Valid][11][UTT][ MSE = 0.1785 | LCC = 0.9562 | SRCC = 0.9556 ] +INFO:root:Average loss=0.06619785994291305 +INFO:root: +[Valid][12][UTT][ MSE = 0.2004 | LCC = 0.9570 | SRCC = 0.9557 ] +INFO:root:Average loss=0.06064632150530815 +INFO:root: +[Valid][13][UTT][ MSE = 0.1880 | LCC = 0.9559 | SRCC = 0.9550 ] +INFO:root:Average loss=0.05729746955633164 +INFO:root: +[Valid][14][UTT][ MSE = 0.2091 | LCC = 0.9556 | SRCC = 0.9546 ] +INFO:root:Average loss=0.05386622993648052 +INFO:root: +[Valid][15][UTT][ MSE = 0.2043 | LCC = 0.9545 | SRCC = 0.9535 ] +INFO:root:Average loss=0.052449051231145856 +INFO:root: +[Valid][16][UTT][ MSE = 0.1843 | LCC = 0.9534 | SRCC = 0.9530 ] +INFO:root:Average loss=0.04850220641493797 +INFO:root: +[Valid][17][UTT][ MSE = 0.2193 | LCC = 0.9547 | SRCC = 0.9544 ] +INFO:root:Average loss=0.048823101952672004 +INFO:root: +[Valid][18][UTT][ MSE = 0.1721 | LCC = 0.9552 | SRCC = 0.9541 ] +INFO:root:Average loss=0.04496191173791885 +INFO:root: +[Valid][19][UTT][ MSE = 0.1767 | LCC = 0.9548 | SRCC = 0.9536 ] +INFO:root:Average loss=0.0411451373398304 +INFO:root: +[Valid][20][UTT][ MSE = 0.1661 | LCC = 0.9559 | SRCC = 0.9549 ] +INFO:root:Average loss=0.03885528598725796 +INFO:root: +[Valid][21][UTT][ MSE = 0.2023 | LCC = 0.9557 | SRCC = 0.9550 ] +INFO:root:Average loss=0.037916133925318715 +INFO:root: +[Valid][22][UTT][ MSE = 0.1865 | LCC = 0.9550 | SRCC = 0.9536 ] +INFO:root:Average loss=0.03923205280303955 +INFO:root: +[Valid][23][UTT][ MSE = 0.1841 | LCC = 0.9545 | SRCC = 0.9540 ] +INFO:root:Average loss=0.03508873711526394 +INFO:root: +[Valid][24][UTT][ MSE = 0.1685 | LCC = 0.9541 | SRCC = 0.9530 ] +INFO:root:Average loss=0.033559143096208575 +INFO:root: +[Valid][25][UTT][ MSE = 0.1932 | LCC = 0.9555 | SRCC = 0.9546 ] +INFO:root:Average loss=0.03505440843105316 +INFO:root: +[Valid][26][UTT][ MSE = 0.2017 | LCC = 0.9552 | SRCC = 0.9542 ] +INFO:root:Average loss=0.031006655409932138 +INFO:root: +[Valid][27][UTT][ MSE = 0.1992 | LCC = 0.9539 | SRCC = 0.9531 ] +INFO:root:Average loss=0.03119840358197689 +INFO:root: +[Valid][28][UTT][ MSE = 0.1946 | LCC = 0.9555 | SRCC = 0.9548 ] +INFO:root:Average loss=0.03141229137778282 +INFO:root: +[Valid][29][UTT][ MSE = 0.1733 | LCC = 0.9555 | SRCC = 0.9541 ] +INFO:root:Average loss=0.028157613769173622 +INFO:root: +[Valid][30][UTT][ MSE = 0.1705 | LCC = 0.9546 | SRCC = 0.9541 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2087 | LCC = 0.9559 | SRCC = 0.9559 ] diff --git a/results_tencent/hubert_base/layer9/run0/config.gin b/results_tencent/hubert_base/layer9/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer9/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer9/run0/model.pt b/results_tencent/hubert_base/layer9/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..05c4e136389ecbb91e5ea6ccac78a1bcbe01d085 --- /dev/null +++ b/results_tencent/hubert_base/layer9/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:434d137a9baaa9d7699aed94fe73b317be7ccfe5ec7145e4336342e0fe7890f1 +size 2747979 diff --git a/results_tencent/hubert_base/layer9/run0/model_best.pt b/results_tencent/hubert_base/layer9/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..8fbd7f7adc07a90885ea545c9cf10f2ae17bb65c --- /dev/null +++ b/results_tencent/hubert_base/layer9/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b0ca21c7b82d9cc86e4d434dcc974ec9ea2e8350223a5eb8a2cd190ad793ade +size 2749638 diff --git a/results_tencent/hubert_base/layer9/run0/train.log b/results_tencent/hubert_base/layer9/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..662fc39436e215d6f5808dbafcb051faba4d3ced --- /dev/null +++ b/results_tencent/hubert_base/layer9/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4611553112268448 +INFO:root: +[Valid][0][UTT][ MSE = 0.3348 | LCC = 0.9238 | SRCC = 0.9214 ] +INFO:root:Average loss=0.19505059134960173 +INFO:root: +[Valid][1][UTT][ MSE = 0.2904 | LCC = 0.9376 | SRCC = 0.9356 ] +INFO:root:Average loss=0.1544279429912567 +INFO:root: +[Valid][2][UTT][ MSE = 0.2566 | LCC = 0.9452 | SRCC = 0.9424 ] +INFO:root:Average loss=0.13919421201944351 +INFO:root: +[Valid][3][UTT][ MSE = 0.2164 | LCC = 0.9477 | SRCC = 0.9459 ] +INFO:root:Average loss=0.12286033433675767 +INFO:root: +[Valid][4][UTT][ MSE = 0.2077 | LCC = 0.9499 | SRCC = 0.9485 ] +INFO:root:Average loss=0.10946286499500274 +INFO:root: +[Valid][5][UTT][ MSE = 0.2285 | LCC = 0.9517 | SRCC = 0.9509 ] +INFO:root:Average loss=0.10210601067543029 +INFO:root: +[Valid][6][UTT][ MSE = 0.1896 | LCC = 0.9516 | SRCC = 0.9508 ] +INFO:root:Average loss=0.09265979513525963 +INFO:root: +[Valid][7][UTT][ MSE = 0.2229 | LCC = 0.9512 | SRCC = 0.9495 ] +INFO:root:Average loss=0.08675390714406968 +INFO:root: +[Valid][8][UTT][ MSE = 0.2056 | LCC = 0.9521 | SRCC = 0.9517 ] +INFO:root:Average loss=0.08708859130740165 +INFO:root: +[Valid][9][UTT][ MSE = 0.1822 | LCC = 0.9542 | SRCC = 0.9529 ] +INFO:root:Average loss=0.07884150579571723 +INFO:root: +[Valid][10][UTT][ MSE = 0.1766 | LCC = 0.9498 | SRCC = 0.9488 ] +INFO:root:Average loss=0.0719936857521534 +INFO:root: +[Valid][11][UTT][ MSE = 0.2051 | LCC = 0.9528 | SRCC = 0.9523 ] +INFO:root:Average loss=0.06825064232945442 +INFO:root: +[Valid][12][UTT][ MSE = 0.1841 | LCC = 0.9535 | SRCC = 0.9520 ] +INFO:root:Average loss=0.06143242532014847 +INFO:root: +[Valid][13][UTT][ MSE = 0.2057 | LCC = 0.9533 | SRCC = 0.9525 ] +INFO:root:Average loss=0.05889843487739563 +INFO:root: +[Valid][14][UTT][ MSE = 0.1904 | LCC = 0.9535 | SRCC = 0.9523 ] +INFO:root:Average loss=0.05652226972579956 +INFO:root: +[Valid][15][UTT][ MSE = 0.1942 | LCC = 0.9534 | SRCC = 0.9526 ] +INFO:root:Average loss=0.05658841474354267 +INFO:root: +[Valid][16][UTT][ MSE = 0.2106 | LCC = 0.9541 | SRCC = 0.9534 ] +INFO:root:Average loss=0.05083543930947781 +INFO:root: +[Valid][17][UTT][ MSE = 0.2029 | LCC = 0.9537 | SRCC = 0.9534 ] +INFO:root:Average loss=0.04996746996045113 +INFO:root: +[Valid][18][UTT][ MSE = 0.1686 | LCC = 0.9543 | SRCC = 0.9532 ] +INFO:root:Average loss=0.04749153618514538 +INFO:root: +[Valid][19][UTT][ MSE = 0.1781 | LCC = 0.9532 | SRCC = 0.9524 ] +INFO:root:Average loss=0.04441627797484398 +INFO:root: +[Valid][20][UTT][ MSE = 0.1938 | LCC = 0.9542 | SRCC = 0.9533 ] +INFO:root:Average loss=0.04110548974573612 +INFO:root: +[Valid][21][UTT][ MSE = 0.1712 | LCC = 0.9528 | SRCC = 0.9519 ] +INFO:root:Average loss=0.04161159273982048 +INFO:root: +[Valid][22][UTT][ MSE = 0.1746 | LCC = 0.9543 | SRCC = 0.9534 ] +INFO:root:Average loss=0.039965137347579005 +INFO:root: +[Valid][23][UTT][ MSE = 0.1697 | LCC = 0.9544 | SRCC = 0.9534 ] +INFO:root:Average loss=0.03694696481525898 +INFO:root: +[Valid][24][UTT][ MSE = 0.1802 | LCC = 0.9552 | SRCC = 0.9542 ] +INFO:root:Average loss=0.036090168595314026 +INFO:root: +[Valid][25][UTT][ MSE = 0.1749 | LCC = 0.9537 | SRCC = 0.9526 ] +INFO:root:Average loss=0.036338919654488565 +INFO:root: +[Valid][26][UTT][ MSE = 0.1856 | LCC = 0.9540 | SRCC = 0.9530 ] +INFO:root:Average loss=0.03359812919795513 +INFO:root: +[Valid][27][UTT][ MSE = 0.1821 | LCC = 0.9543 | SRCC = 0.9533 ] +INFO:root:Average loss=0.03167370665073395 +INFO:root: +[Valid][28][UTT][ MSE = 0.1848 | LCC = 0.9543 | SRCC = 0.9532 ] +INFO:root:Average loss=0.03095984573662281 +INFO:root: +[Valid][29][UTT][ MSE = 0.1709 | LCC = 0.9551 | SRCC = 0.9544 ] +INFO:root:Average loss=0.03140510012209415 +INFO:root: +[Valid][30][UTT][ MSE = 0.1698 | LCC = 0.9554 | SRCC = 0.9545 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1768 | LCC = 0.9560 | SRCC = 0.9554 ] diff --git a/results_tencent/hubert_base/layer9/run1/config.gin b/results_tencent/hubert_base/layer9/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer9/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer9/run1/model.pt b/results_tencent/hubert_base/layer9/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6cee2508413891202d08562c28678776d3642d70 --- /dev/null +++ b/results_tencent/hubert_base/layer9/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f3159dfb81cc153e4329475a71038d662763e23c17fe68219bddf950c7a1dfa8 +size 2747979 diff --git a/results_tencent/hubert_base/layer9/run1/model_best.pt b/results_tencent/hubert_base/layer9/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..559d8c684c123f5de6b4ed8418c00fdd1b0a3c75 --- /dev/null +++ b/results_tencent/hubert_base/layer9/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dc8c545db77cc26b42f78d6bcc512d9ead1ec780a2c12e15936d0423d1185266 +size 2749638 diff --git a/results_tencent/hubert_base/layer9/run1/train.log b/results_tencent/hubert_base/layer9/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..94f637fb4e39a118bbab08b93a8e7f2e6cba14e1 --- /dev/null +++ b/results_tencent/hubert_base/layer9/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4794554990530014 +INFO:root: +[Valid][0][UTT][ MSE = 0.3339 | LCC = 0.9231 | SRCC = 0.9196 ] +INFO:root:Average loss=0.19381355267763137 +INFO:root: +[Valid][1][UTT][ MSE = 0.2409 | LCC = 0.9381 | SRCC = 0.9363 ] +INFO:root:Average loss=0.1604388964176178 +INFO:root: +[Valid][2][UTT][ MSE = 0.2546 | LCC = 0.9435 | SRCC = 0.9421 ] +INFO:root:Average loss=0.1389154340624809 +INFO:root: +[Valid][3][UTT][ MSE = 0.2766 | LCC = 0.9465 | SRCC = 0.9459 ] +INFO:root:Average loss=0.12320762157440185 +INFO:root: +[Valid][4][UTT][ MSE = 0.2386 | LCC = 0.9499 | SRCC = 0.9491 ] +INFO:root:Average loss=0.11082123011350632 +INFO:root: +[Valid][5][UTT][ MSE = 0.2323 | LCC = 0.9498 | SRCC = 0.9491 ] +INFO:root:Average loss=0.10533437290787696 +INFO:root: +[Valid][6][UTT][ MSE = 0.2532 | LCC = 0.9514 | SRCC = 0.9510 ] +INFO:root:Average loss=0.09626249805092811 +INFO:root: +[Valid][7][UTT][ MSE = 0.2496 | LCC = 0.9512 | SRCC = 0.9517 ] +INFO:root:Average loss=0.08956621789932251 +INFO:root: +[Valid][8][UTT][ MSE = 0.1912 | LCC = 0.9490 | SRCC = 0.9480 ] +INFO:root:Average loss=0.08556078800559044 +INFO:root: +[Valid][9][UTT][ MSE = 0.2229 | LCC = 0.9529 | SRCC = 0.9518 ] +INFO:root:Average loss=0.08031307113170624 +INFO:root: +[Valid][10][UTT][ MSE = 0.2242 | LCC = 0.9528 | SRCC = 0.9524 ] +INFO:root:Average loss=0.07638917469978332 +INFO:root: +[Valid][11][UTT][ MSE = 0.2158 | LCC = 0.9538 | SRCC = 0.9531 ] +INFO:root:Average loss=0.07028329223394394 +INFO:root: +[Valid][12][UTT][ MSE = 0.2289 | LCC = 0.9532 | SRCC = 0.9524 ] +INFO:root:Average loss=0.06532420751452446 +INFO:root: +[Valid][13][UTT][ MSE = 0.2152 | LCC = 0.9523 | SRCC = 0.9508 ] +INFO:root:Average loss=0.06229425543546677 +INFO:root: +[Valid][14][UTT][ MSE = 0.2130 | LCC = 0.9538 | SRCC = 0.9533 ] +INFO:root:Average loss=0.05829696491360664 +INFO:root: +[Valid][15][UTT][ MSE = 0.2160 | LCC = 0.9537 | SRCC = 0.9527 ] +INFO:root:Average loss=0.05399110153317452 +INFO:root: +[Valid][16][UTT][ MSE = 0.1958 | LCC = 0.9509 | SRCC = 0.9496 ] +INFO:root:Average loss=0.05328322806954384 +INFO:root: +[Valid][17][UTT][ MSE = 0.2116 | LCC = 0.9526 | SRCC = 0.9521 ] +INFO:root:Average loss=0.04820345701277256 +INFO:root: +[Valid][18][UTT][ MSE = 0.2023 | LCC = 0.9535 | SRCC = 0.9524 ] +INFO:root:Average loss=0.048683357164263724 +INFO:root: +[Valid][19][UTT][ MSE = 0.2322 | LCC = 0.9529 | SRCC = 0.9529 ] +INFO:root:Average loss=0.047226441830396655 +INFO:root: +[Valid][20][UTT][ MSE = 0.1825 | LCC = 0.9517 | SRCC = 0.9511 ] +INFO:root:Average loss=0.045452730923891065 +INFO:root: +[Valid][21][UTT][ MSE = 0.2102 | LCC = 0.9538 | SRCC = 0.9538 ] +INFO:root:Average loss=0.04333837373554707 +INFO:root: +[Valid][22][UTT][ MSE = 0.2328 | LCC = 0.9494 | SRCC = 0.9487 ] +INFO:root:Average loss=0.0412999782115221 +INFO:root: +[Valid][23][UTT][ MSE = 0.2004 | LCC = 0.9539 | SRCC = 0.9534 ] +INFO:root:Average loss=0.03957126955688 +INFO:root: +[Valid][24][UTT][ MSE = 0.2058 | LCC = 0.9533 | SRCC = 0.9522 ] +INFO:root:Average loss=0.035049849212169644 +INFO:root: +[Valid][25][UTT][ MSE = 0.2124 | LCC = 0.9523 | SRCC = 0.9518 ] +INFO:root:Average loss=0.037009487852454184 +INFO:root: +[Valid][26][UTT][ MSE = 0.2058 | LCC = 0.9529 | SRCC = 0.9523 ] +INFO:root:Average loss=0.033846070647239684 +INFO:root: +[Valid][27][UTT][ MSE = 0.1918 | LCC = 0.9526 | SRCC = 0.9517 ] +INFO:root:Average loss=0.033949534803628924 +INFO:root: +[Valid][28][UTT][ MSE = 0.1955 | LCC = 0.9526 | SRCC = 0.9517 ] +INFO:root:Average loss=0.033655183762311934 +INFO:root: +[Valid][29][UTT][ MSE = 0.2264 | LCC = 0.9525 | SRCC = 0.9524 ] +INFO:root:Average loss=0.03166666159033776 +INFO:root: +[Valid][30][UTT][ MSE = 0.2084 | LCC = 0.9514 | SRCC = 0.9505 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2072 | LCC = 0.9548 | SRCC = 0.9552 ] diff --git a/results_tencent/hubert_base/layer9/run2/config.gin b/results_tencent/hubert_base/layer9/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer9/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer9/run2/model.pt b/results_tencent/hubert_base/layer9/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c01e79fe5f94dd9c048f0c62004c461d24548e25 --- /dev/null +++ b/results_tencent/hubert_base/layer9/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b9dc0d50bb5f9757bcb34d4d0f967ab6edb2cacb9a59533cc8190ab431a4bea7 +size 2747979 diff --git a/results_tencent/hubert_base/layer9/run2/model_best.pt b/results_tencent/hubert_base/layer9/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ae0220bc3cba3bdd1db91cd384a2fe8c2d6b847f --- /dev/null +++ b/results_tencent/hubert_base/layer9/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f8f8b7288c41670958988b08317bbb6b835b2b3ba09991d17d122e345bc1239f +size 2749638 diff --git a/results_tencent/hubert_base/layer9/run2/train.log b/results_tencent/hubert_base/layer9/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..61d93cd194ca89b6b94b14ba27bc3cd24b6c6d5c --- /dev/null +++ b/results_tencent/hubert_base/layer9/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.516580154299736 +INFO:root: +[Valid][0][UTT][ MSE = 0.3059 | LCC = 0.9163 | SRCC = 0.9136 ] +INFO:root:Average loss=0.20150395971536636 +INFO:root: +[Valid][1][UTT][ MSE = 0.2573 | LCC = 0.9397 | SRCC = 0.9392 ] +INFO:root:Average loss=0.16062315887212752 +INFO:root: +[Valid][2][UTT][ MSE = 0.2384 | LCC = 0.9427 | SRCC = 0.9441 ] +INFO:root:Average loss=0.13932085555791854 +INFO:root: +[Valid][3][UTT][ MSE = 0.2289 | LCC = 0.9491 | SRCC = 0.9489 ] +INFO:root:Average loss=0.11960892224311828 +INFO:root: +[Valid][4][UTT][ MSE = 0.2221 | LCC = 0.9514 | SRCC = 0.9502 ] +INFO:root:Average loss=0.11878076639771462 +INFO:root: +[Valid][5][UTT][ MSE = 0.2088 | LCC = 0.9517 | SRCC = 0.9506 ] +INFO:root:Average loss=0.1071685455441475 +INFO:root: +[Valid][6][UTT][ MSE = 0.2200 | LCC = 0.9518 | SRCC = 0.9517 ] +INFO:root:Average loss=0.09824139600992203 +INFO:root: +[Valid][7][UTT][ MSE = 0.2305 | LCC = 0.9512 | SRCC = 0.9514 ] +INFO:root:Average loss=0.09730610716342926 +INFO:root: +[Valid][8][UTT][ MSE = 0.2629 | LCC = 0.9529 | SRCC = 0.9528 ] +INFO:root:Average loss=0.0847796440422535 +INFO:root: +[Valid][9][UTT][ MSE = 0.1985 | LCC = 0.9540 | SRCC = 0.9534 ] +INFO:root:Average loss=0.08380139741301537 +INFO:root: +[Valid][10][UTT][ MSE = 0.1997 | LCC = 0.9539 | SRCC = 0.9529 ] +INFO:root:Average loss=0.07575084656476974 +INFO:root: +[Valid][11][UTT][ MSE = 0.1930 | LCC = 0.9545 | SRCC = 0.9535 ] +INFO:root:Average loss=0.07232837578654289 +INFO:root: +[Valid][12][UTT][ MSE = 0.2107 | LCC = 0.9524 | SRCC = 0.9520 ] +INFO:root:Average loss=0.06712987869977952 +INFO:root: +[Valid][13][UTT][ MSE = 0.2080 | LCC = 0.9535 | SRCC = 0.9527 ] +INFO:root:Average loss=0.06106889033317566 +INFO:root: +[Valid][14][UTT][ MSE = 0.2404 | LCC = 0.9516 | SRCC = 0.9504 ] +INFO:root:Average loss=0.061450849413871765 +INFO:root: +[Valid][15][UTT][ MSE = 0.1973 | LCC = 0.9528 | SRCC = 0.9514 ] +INFO:root:Average loss=0.05807031652331352 +INFO:root: +[Valid][16][UTT][ MSE = 0.1751 | LCC = 0.9552 | SRCC = 0.9542 ] +INFO:root:Average loss=0.05490294256806374 +INFO:root: +[Valid][17][UTT][ MSE = 0.2304 | LCC = 0.9530 | SRCC = 0.9516 ] +INFO:root:Average loss=0.04886404238641262 +INFO:root: +[Valid][18][UTT][ MSE = 0.2134 | LCC = 0.9507 | SRCC = 0.9502 ] +INFO:root:Average loss=0.04886039619147778 +INFO:root: +[Valid][19][UTT][ MSE = 0.1982 | LCC = 0.9543 | SRCC = 0.9530 ] +INFO:root:Average loss=0.045222493052482604 +INFO:root: +[Valid][20][UTT][ MSE = 0.1996 | LCC = 0.9530 | SRCC = 0.9520 ] +INFO:root:Average loss=0.044287016347050664 +INFO:root: +[Valid][21][UTT][ MSE = 0.2042 | LCC = 0.9530 | SRCC = 0.9516 ] +INFO:root:Average loss=0.041404220759868625 +INFO:root: +[Valid][22][UTT][ MSE = 0.1941 | LCC = 0.9533 | SRCC = 0.9516 ] +INFO:root:Average loss=0.039016685649752615 +INFO:root: +[Valid][23][UTT][ MSE = 0.2132 | LCC = 0.9527 | SRCC = 0.9510 ] +INFO:root:Average loss=0.04030372560024262 +INFO:root: +[Valid][24][UTT][ MSE = 0.2457 | LCC = 0.9514 | SRCC = 0.9500 ] +INFO:root:Average loss=0.03900586904585362 +INFO:root: +[Valid][25][UTT][ MSE = 0.2039 | LCC = 0.9531 | SRCC = 0.9522 ] +INFO:root:Average loss=0.0371421482861042 +INFO:root: +[Valid][26][UTT][ MSE = 0.2042 | LCC = 0.9530 | SRCC = 0.9517 ] +INFO:root:Average loss=0.03468459096550942 +INFO:root: +[Valid][27][UTT][ MSE = 0.1949 | LCC = 0.9546 | SRCC = 0.9534 ] +INFO:root:Average loss=0.035614626318216325 +INFO:root: +[Valid][28][UTT][ MSE = 0.1869 | LCC = 0.9533 | SRCC = 0.9519 ] +INFO:root:Average loss=0.034444830611348155 +INFO:root: +[Valid][29][UTT][ MSE = 0.2027 | LCC = 0.9547 | SRCC = 0.9537 ] +INFO:root:Average loss=0.03169719558954239 +INFO:root: +[Valid][30][UTT][ MSE = 0.2237 | LCC = 0.9524 | SRCC = 0.9509 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1766 | LCC = 0.9567 | SRCC = 0.9568 ] diff --git a/results_tencent/hubert_base/layer9/run3/config.gin b/results_tencent/hubert_base/layer9/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer9/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer9/run3/model.pt b/results_tencent/hubert_base/layer9/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..dbb039ec651705ed8b2242ca4036fd66209a151a --- /dev/null +++ b/results_tencent/hubert_base/layer9/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f14a2e6bfc1edb6fbd5a3e0ecba60b20246b89814d38efc87e44a3dfa268e218 +size 2747979 diff --git a/results_tencent/hubert_base/layer9/run3/model_best.pt b/results_tencent/hubert_base/layer9/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..16252f8635036f1c3af34e7517e91d6b2f4642a1 --- /dev/null +++ b/results_tencent/hubert_base/layer9/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b5dd65c77d698ec94d2e386fa8c14c135fb44811fd099a1281470553b82aed84 +size 2749638 diff --git a/results_tencent/hubert_base/layer9/run3/train.log b/results_tencent/hubert_base/layer9/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..031b054c70892858deebb7e784faa8b3d4858336 --- /dev/null +++ b/results_tencent/hubert_base/layer9/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44489925479888914 +INFO:root: +[Valid][0][UTT][ MSE = 0.2831 | LCC = 0.9205 | SRCC = 0.9175 ] +INFO:root:Average loss=0.19992084527015686 +INFO:root: +[Valid][1][UTT][ MSE = 0.2241 | LCC = 0.9359 | SRCC = 0.9343 ] +INFO:root:Average loss=0.16775040507316588 +INFO:root: +[Valid][2][UTT][ MSE = 0.2586 | LCC = 0.9439 | SRCC = 0.9424 ] +INFO:root:Average loss=0.14465486055612564 +INFO:root: +[Valid][3][UTT][ MSE = 0.2013 | LCC = 0.9465 | SRCC = 0.9447 ] +INFO:root:Average loss=0.12591670167446137 +INFO:root: +[Valid][4][UTT][ MSE = 0.2220 | LCC = 0.9501 | SRCC = 0.9491 ] +INFO:root:Average loss=0.11691068148612976 +INFO:root: +[Valid][5][UTT][ MSE = 0.2018 | LCC = 0.9513 | SRCC = 0.9501 ] +INFO:root:Average loss=0.1081519563794136 +INFO:root: +[Valid][6][UTT][ MSE = 0.2221 | LCC = 0.9519 | SRCC = 0.9517 ] +INFO:root:Average loss=0.10522328591346741 +INFO:root: +[Valid][7][UTT][ MSE = 0.2228 | LCC = 0.9491 | SRCC = 0.9501 ] +INFO:root:Average loss=0.09493446242809296 +INFO:root: +[Valid][8][UTT][ MSE = 0.1810 | LCC = 0.9525 | SRCC = 0.9520 ] +INFO:root:Average loss=0.08875785946846008 +INFO:root: +[Valid][9][UTT][ MSE = 0.1946 | LCC = 0.9535 | SRCC = 0.9526 ] +INFO:root:Average loss=0.08144683456420898 +INFO:root: +[Valid][10][UTT][ MSE = 0.1864 | LCC = 0.9531 | SRCC = 0.9527 ] +INFO:root:Average loss=0.07898153153061867 +INFO:root: +[Valid][11][UTT][ MSE = 0.1997 | LCC = 0.9544 | SRCC = 0.9534 ] +INFO:root:Average loss=0.07077440786361694 +INFO:root: +[Valid][12][UTT][ MSE = 0.1973 | LCC = 0.9532 | SRCC = 0.9520 ] +INFO:root:Average loss=0.07025890320539474 +INFO:root: +[Valid][13][UTT][ MSE = 0.2178 | LCC = 0.9537 | SRCC = 0.9524 ] +INFO:root:Average loss=0.06654996809363366 +INFO:root: +[Valid][14][UTT][ MSE = 0.2051 | LCC = 0.9529 | SRCC = 0.9514 ] +INFO:root:Average loss=0.060564726889133454 +INFO:root: +[Valid][15][UTT][ MSE = 0.1967 | LCC = 0.9528 | SRCC = 0.9514 ] +INFO:root:Average loss=0.05786897917091847 +INFO:root: +[Valid][16][UTT][ MSE = 0.1963 | LCC = 0.9519 | SRCC = 0.9510 ] +INFO:root:Average loss=0.05494560393691063 +INFO:root: +[Valid][17][UTT][ MSE = 0.1778 | LCC = 0.9544 | SRCC = 0.9531 ] +INFO:root:Average loss=0.05354078438878059 +INFO:root: +[Valid][18][UTT][ MSE = 0.1798 | LCC = 0.9542 | SRCC = 0.9530 ] +INFO:root:Average loss=0.04955881415307522 +INFO:root: +[Valid][19][UTT][ MSE = 0.1882 | LCC = 0.9542 | SRCC = 0.9529 ] +INFO:root:Average loss=0.04627042317390442 +INFO:root: +[Valid][20][UTT][ MSE = 0.1786 | LCC = 0.9531 | SRCC = 0.9520 ] +INFO:root:Average loss=0.044572174459695815 +INFO:root: +[Valid][21][UTT][ MSE = 0.1905 | LCC = 0.9537 | SRCC = 0.9522 ] +INFO:root:Average loss=0.04349770082533359 +INFO:root: +[Valid][22][UTT][ MSE = 0.1942 | LCC = 0.9542 | SRCC = 0.9533 ] +INFO:root:Average loss=0.04032326985895634 +INFO:root: +[Valid][23][UTT][ MSE = 0.1997 | LCC = 0.9529 | SRCC = 0.9523 ] +INFO:root:Average loss=0.03920040807127952 +INFO:root: +[Valid][24][UTT][ MSE = 0.1837 | LCC = 0.9538 | SRCC = 0.9527 ] +INFO:root:Average loss=0.04017826262116432 +INFO:root: +[Valid][25][UTT][ MSE = 0.1884 | LCC = 0.9553 | SRCC = 0.9542 ] +INFO:root:Average loss=0.03702337738871574 +INFO:root: +[Valid][26][UTT][ MSE = 0.1794 | LCC = 0.9543 | SRCC = 0.9533 ] +INFO:root:Average loss=0.03617569503188133 +INFO:root: +[Valid][27][UTT][ MSE = 0.1904 | LCC = 0.9542 | SRCC = 0.9531 ] +INFO:root:Average loss=0.036098716512322424 +INFO:root: +[Valid][28][UTT][ MSE = 0.1861 | LCC = 0.9536 | SRCC = 0.9527 ] +INFO:root:Average loss=0.03335657443106174 +INFO:root: +[Valid][29][UTT][ MSE = 0.2034 | LCC = 0.9523 | SRCC = 0.9514 ] +INFO:root:Average loss=0.03307583332061768 +INFO:root: +[Valid][30][UTT][ MSE = 0.1810 | LCC = 0.9537 | SRCC = 0.9524 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1918 | LCC = 0.9568 | SRCC = 0.9577 ] diff --git a/results_tencent/hubert_base/layer9/run4/config.gin b/results_tencent/hubert_base/layer9/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/hubert_base/layer9/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/hubert_base/layer9/run4/model.pt b/results_tencent/hubert_base/layer9/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..02bd612d7bc5100ad70c6713768509415d1e92d5 --- /dev/null +++ b/results_tencent/hubert_base/layer9/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9de4b30df9fa10f50161aa2239de32fe5c47246b8d748f371303626097c1c0ca +size 2747979 diff --git a/results_tencent/hubert_base/layer9/run4/model_best.pt b/results_tencent/hubert_base/layer9/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..03c2143186c61ba247ab6dcc437569858a97c803 --- /dev/null +++ b/results_tencent/hubert_base/layer9/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39fa1a47fea32ed66d2c9daee347f69b9120ac468aa404b7610677fbe0eaa25f +size 2749638 diff --git a/results_tencent/hubert_base/layer9/run4/train.log b/results_tencent/hubert_base/layer9/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..431a235249780dbbe9edc7a520987ce828617195 --- /dev/null +++ b/results_tencent/hubert_base/layer9/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4514622130393982 +INFO:root: +[Valid][0][UTT][ MSE = 0.2982 | LCC = 0.9198 | SRCC = 0.9177 ] +INFO:root:Average loss=0.19023162853717804 +INFO:root: +[Valid][1][UTT][ MSE = 0.3094 | LCC = 0.9384 | SRCC = 0.9366 ] +INFO:root:Average loss=0.16245165157318114 +INFO:root: +[Valid][2][UTT][ MSE = 0.2578 | LCC = 0.9447 | SRCC = 0.9432 ] +INFO:root:Average loss=0.1411689397096634 +INFO:root: +[Valid][3][UTT][ MSE = 0.2277 | LCC = 0.9471 | SRCC = 0.9465 ] +INFO:root:Average loss=0.12703043895959854 +INFO:root: +[Valid][4][UTT][ MSE = 0.2468 | LCC = 0.9502 | SRCC = 0.9496 ] +INFO:root:Average loss=0.11935830095410346 +INFO:root: +[Valid][5][UTT][ MSE = 0.2159 | LCC = 0.9492 | SRCC = 0.9487 ] +INFO:root:Average loss=0.10688745027780533 +INFO:root: +[Valid][6][UTT][ MSE = 0.2559 | LCC = 0.9488 | SRCC = 0.9483 ] +INFO:root:Average loss=0.10164025604724884 +INFO:root: +[Valid][7][UTT][ MSE = 0.2171 | LCC = 0.9517 | SRCC = 0.9510 ] +INFO:root:Average loss=0.09206317946314811 +INFO:root: +[Valid][8][UTT][ MSE = 0.2112 | LCC = 0.9529 | SRCC = 0.9520 ] +INFO:root:Average loss=0.08785891342163087 +INFO:root: +[Valid][9][UTT][ MSE = 0.2527 | LCC = 0.9522 | SRCC = 0.9523 ] +INFO:root:Average loss=0.08156153374910355 +INFO:root: +[Valid][10][UTT][ MSE = 0.2440 | LCC = 0.9534 | SRCC = 0.9527 ] +INFO:root:Average loss=0.07963981875777244 +INFO:root: +[Valid][11][UTT][ MSE = 0.1847 | LCC = 0.9531 | SRCC = 0.9525 ] +INFO:root:Average loss=0.07273643165826797 +INFO:root: +[Valid][12][UTT][ MSE = 0.1928 | LCC = 0.9538 | SRCC = 0.9533 ] +INFO:root:Average loss=0.0679737445116043 +INFO:root: +[Valid][13][UTT][ MSE = 0.2064 | LCC = 0.9530 | SRCC = 0.9518 ] +INFO:root:Average loss=0.06422815892100334 +INFO:root: +[Valid][14][UTT][ MSE = 0.1896 | LCC = 0.9545 | SRCC = 0.9532 ] +INFO:root:Average loss=0.06099076396226883 +INFO:root: +[Valid][15][UTT][ MSE = 0.2162 | LCC = 0.9539 | SRCC = 0.9529 ] +INFO:root:Average loss=0.059164783149957656 +INFO:root: +[Valid][16][UTT][ MSE = 0.2226 | LCC = 0.9546 | SRCC = 0.9536 ] +INFO:root:Average loss=0.0537550276517868 +INFO:root: +[Valid][17][UTT][ MSE = 0.2011 | LCC = 0.9545 | SRCC = 0.9534 ] +INFO:root:Average loss=0.054237641513347624 +INFO:root: +[Valid][18][UTT][ MSE = 0.2296 | LCC = 0.9538 | SRCC = 0.9534 ] +INFO:root:Average loss=0.05039509908854961 +INFO:root: +[Valid][19][UTT][ MSE = 0.2009 | LCC = 0.9543 | SRCC = 0.9535 ] +INFO:root:Average loss=0.046657975643873216 +INFO:root: +[Valid][20][UTT][ MSE = 0.1791 | LCC = 0.9551 | SRCC = 0.9542 ] +INFO:root:Average loss=0.045116498053073884 +INFO:root: +[Valid][21][UTT][ MSE = 0.2003 | LCC = 0.9546 | SRCC = 0.9540 ] +INFO:root:Average loss=0.04341853940486908 +INFO:root: +[Valid][22][UTT][ MSE = 0.2065 | LCC = 0.9548 | SRCC = 0.9539 ] +INFO:root:Average loss=0.03972654163837433 +INFO:root: +[Valid][23][UTT][ MSE = 0.2000 | LCC = 0.9548 | SRCC = 0.9540 ] +INFO:root:Average loss=0.04157589550316334 +INFO:root: +[Valid][24][UTT][ MSE = 0.2207 | LCC = 0.9548 | SRCC = 0.9543 ] +INFO:root:Average loss=0.03843232263624668 +INFO:root: +[Valid][25][UTT][ MSE = 0.2036 | LCC = 0.9539 | SRCC = 0.9530 ] +INFO:root:Average loss=0.03793144614994526 +INFO:root: +[Valid][26][UTT][ MSE = 0.1868 | LCC = 0.9539 | SRCC = 0.9526 ] +INFO:root:Average loss=0.03623988161981106 +INFO:root: +[Valid][27][UTT][ MSE = 0.1914 | LCC = 0.9549 | SRCC = 0.9539 ] +INFO:root:Average loss=0.03451882675290108 +INFO:root: +[Valid][28][UTT][ MSE = 0.1895 | LCC = 0.9543 | SRCC = 0.9531 ] +INFO:root:Average loss=0.03437819467484951 +INFO:root: +[Valid][29][UTT][ MSE = 0.2008 | LCC = 0.9558 | SRCC = 0.9547 ] +INFO:root:Average loss=0.03208430276811123 +INFO:root: +[Valid][30][UTT][ MSE = 0.2046 | LCC = 0.9549 | SRCC = 0.9540 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2155 | LCC = 0.9558 | SRCC = 0.9563 ] diff --git a/results_tencent/w2v2_base/layer0/run0/config.gin b/results_tencent/w2v2_base/layer0/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer0/run0/model.pt b/results_tencent/w2v2_base/layer0/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ce1a77e5cc4d48cc477139e5a90f46f9733dc28a --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e5e0c8f0e426b4fe52d2c96e2c1683ea02774836c9ca9b75d742dd14edb483c +size 2747979 diff --git a/results_tencent/w2v2_base/layer0/run0/model_best.pt b/results_tencent/w2v2_base/layer0/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..bec343a88e3390ad9f41b07b46abb68bf665c800 --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:96672f730296fc002cebd91d1814834bcdfe5651bff0b6e8e846dc1e25ac5bec +size 2749638 diff --git a/results_tencent/w2v2_base/layer0/run0/train.log b/results_tencent/w2v2_base/layer0/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c0b98624640d09232c0b1c2caef11c3177ad40e6 --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5195842726230622 +INFO:root: +[Valid][0][UTT][ MSE = 0.3141 | LCC = 0.9084 | SRCC = 0.9090 ] +INFO:root:Average loss=0.22970261216163634 +INFO:root: +[Valid][1][UTT][ MSE = 0.2463 | LCC = 0.9275 | SRCC = 0.9305 ] +INFO:root:Average loss=0.19075120836496354 +INFO:root: +[Valid][2][UTT][ MSE = 0.2500 | LCC = 0.9341 | SRCC = 0.9360 ] +INFO:root:Average loss=0.1672649342417717 +INFO:root: +[Valid][3][UTT][ MSE = 0.2357 | LCC = 0.9376 | SRCC = 0.9389 ] +INFO:root:Average loss=0.14789790195226668 +INFO:root: +[Valid][4][UTT][ MSE = 0.2307 | LCC = 0.9415 | SRCC = 0.9448 ] +INFO:root:Average loss=0.13114942133426666 +INFO:root: +[Valid][5][UTT][ MSE = 0.2155 | LCC = 0.9456 | SRCC = 0.9481 ] +INFO:root:Average loss=0.12142134097218514 +INFO:root: +[Valid][6][UTT][ MSE = 0.2271 | LCC = 0.9453 | SRCC = 0.9469 ] +INFO:root:Average loss=0.11490187740325927 +INFO:root: +[Valid][7][UTT][ MSE = 0.2145 | LCC = 0.9475 | SRCC = 0.9485 ] +INFO:root:Average loss=0.10652791911363602 +INFO:root: +[Valid][8][UTT][ MSE = 0.2583 | LCC = 0.9480 | SRCC = 0.9499 ] +INFO:root:Average loss=0.09892977917194366 +INFO:root: +[Valid][9][UTT][ MSE = 0.2278 | LCC = 0.9466 | SRCC = 0.9508 ] +INFO:root:Average loss=0.09758923724293708 +INFO:root: +[Valid][10][UTT][ MSE = 0.2160 | LCC = 0.9502 | SRCC = 0.9506 ] +INFO:root:Average loss=0.08828170296549798 +INFO:root: +[Valid][11][UTT][ MSE = 0.2287 | LCC = 0.9507 | SRCC = 0.9519 ] +INFO:root:Average loss=0.08549676489830017 +INFO:root: +[Valid][12][UTT][ MSE = 0.1768 | LCC = 0.9513 | SRCC = 0.9523 ] +INFO:root:Average loss=0.08017281928658486 +INFO:root: +[Valid][13][UTT][ MSE = 0.1896 | LCC = 0.9507 | SRCC = 0.9523 ] +INFO:root:Average loss=0.07650479346513749 +INFO:root: +[Valid][14][UTT][ MSE = 0.2271 | LCC = 0.9518 | SRCC = 0.9524 ] +INFO:root:Average loss=0.07320074528455735 +INFO:root: +[Valid][15][UTT][ MSE = 0.1776 | LCC = 0.9520 | SRCC = 0.9522 ] +INFO:root:Average loss=0.07006070828437805 +INFO:root: +[Valid][16][UTT][ MSE = 0.1935 | LCC = 0.9516 | SRCC = 0.9527 ] +INFO:root:Average loss=0.0666362909078598 +INFO:root: +[Valid][17][UTT][ MSE = 0.1909 | LCC = 0.9527 | SRCC = 0.9532 ] +INFO:root:Average loss=0.06495228761434554 +INFO:root: +[Valid][18][UTT][ MSE = 0.2069 | LCC = 0.9537 | SRCC = 0.9547 ] +INFO:root:Average loss=0.06525403305888176 +INFO:root: +[Valid][19][UTT][ MSE = 0.2375 | LCC = 0.9519 | SRCC = 0.9519 ] +INFO:root:Average loss=0.06052417895197868 +INFO:root: +[Valid][20][UTT][ MSE = 0.1970 | LCC = 0.9533 | SRCC = 0.9537 ] +INFO:root:Average loss=0.05679317280650139 +INFO:root: +[Valid][21][UTT][ MSE = 0.2040 | LCC = 0.9518 | SRCC = 0.9518 ] +INFO:root:Average loss=0.05351372671127319 +INFO:root: +[Valid][22][UTT][ MSE = 0.1939 | LCC = 0.9531 | SRCC = 0.9533 ] +INFO:root:Average loss=0.0529806854724884 +INFO:root: +[Valid][23][UTT][ MSE = 0.1887 | LCC = 0.9548 | SRCC = 0.9549 ] +INFO:root:Average loss=0.049514997616410254 +INFO:root: +[Valid][24][UTT][ MSE = 0.1972 | LCC = 0.9520 | SRCC = 0.9522 ] +INFO:root:Average loss=0.04991406881809234 +INFO:root: +[Valid][25][UTT][ MSE = 0.1840 | LCC = 0.9545 | SRCC = 0.9552 ] +INFO:root:Average loss=0.04938196285068989 +INFO:root: +[Valid][26][UTT][ MSE = 0.1825 | LCC = 0.9552 | SRCC = 0.9560 ] +INFO:root:Average loss=0.04630316327512264 +INFO:root: +[Valid][27][UTT][ MSE = 0.2033 | LCC = 0.9545 | SRCC = 0.9556 ] +INFO:root:Average loss=0.04714117431640625 +INFO:root: +[Valid][28][UTT][ MSE = 0.1892 | LCC = 0.9563 | SRCC = 0.9567 ] +INFO:root:Average loss=0.04429442910850048 +INFO:root: +[Valid][29][UTT][ MSE = 0.1845 | LCC = 0.9567 | SRCC = 0.9577 ] +INFO:root:Average loss=0.042353986442089084 +INFO:root: +[Valid][30][UTT][ MSE = 0.1900 | LCC = 0.9564 | SRCC = 0.9569 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1816 | LCC = 0.9574 | SRCC = 0.9580 ] diff --git a/results_tencent/w2v2_base/layer0/run1/config.gin b/results_tencent/w2v2_base/layer0/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer0/run1/model.pt b/results_tencent/w2v2_base/layer0/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ffa249dc70afe851e866b41f77df79690013a27d --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7d1104618f60d7305338e6750b5fc26fd78d512b1b7dfb70d23df761c1b8c093 +size 2747979 diff --git a/results_tencent/w2v2_base/layer0/run1/model_best.pt b/results_tencent/w2v2_base/layer0/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7bbb5f6cade68d0ea8cdd523b3415e49498db3b1 --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:84b1fe7a81ae2b0b551c2d33fd0cc0cfd494af2f10e221bad0c3a1d5c245b84a +size 2749638 diff --git a/results_tencent/w2v2_base/layer0/run1/train.log b/results_tencent/w2v2_base/layer0/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ce78c9b8b819a8f027a5bcdf6f2e87805b56eb7d --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5789492708444596 +INFO:root: +[Valid][0][UTT][ MSE = 0.3455 | LCC = 0.9010 | SRCC = 0.9025 ] +INFO:root:Average loss=0.2288916565179825 +INFO:root: +[Valid][1][UTT][ MSE = 0.3021 | LCC = 0.9230 | SRCC = 0.9265 ] +INFO:root:Average loss=0.1824639158844948 +INFO:root: +[Valid][2][UTT][ MSE = 0.2537 | LCC = 0.9340 | SRCC = 0.9358 ] +INFO:root:Average loss=0.15899476563930512 +INFO:root: +[Valid][3][UTT][ MSE = 0.2625 | LCC = 0.9381 | SRCC = 0.9395 ] +INFO:root:Average loss=0.14147335380315781 +INFO:root: +[Valid][4][UTT][ MSE = 0.2488 | LCC = 0.9405 | SRCC = 0.9437 ] +INFO:root:Average loss=0.13478086483478546 +INFO:root: +[Valid][5][UTT][ MSE = 0.2620 | LCC = 0.9435 | SRCC = 0.9464 ] +INFO:root:Average loss=0.12303827875852585 +INFO:root: +[Valid][6][UTT][ MSE = 0.2247 | LCC = 0.9449 | SRCC = 0.9462 ] +INFO:root:Average loss=0.11463609707355499 +INFO:root: +[Valid][7][UTT][ MSE = 0.2251 | LCC = 0.9482 | SRCC = 0.9497 ] +INFO:root:Average loss=0.10588287150859833 +INFO:root: +[Valid][8][UTT][ MSE = 0.2462 | LCC = 0.9480 | SRCC = 0.9503 ] +INFO:root:Average loss=0.09960187911987305 +INFO:root: +[Valid][9][UTT][ MSE = 0.2046 | LCC = 0.9489 | SRCC = 0.9510 ] +INFO:root:Average loss=0.09456500878930092 +INFO:root: +[Valid][10][UTT][ MSE = 0.2211 | LCC = 0.9495 | SRCC = 0.9509 ] +INFO:root:Average loss=0.08579783502221107 +INFO:root: +[Valid][11][UTT][ MSE = 0.1959 | LCC = 0.9505 | SRCC = 0.9524 ] +INFO:root:Average loss=0.08736501124501228 +INFO:root: +[Valid][12][UTT][ MSE = 0.2190 | LCC = 0.9513 | SRCC = 0.9531 ] +INFO:root:Average loss=0.08066918399930001 +INFO:root: +[Valid][13][UTT][ MSE = 0.2340 | LCC = 0.9507 | SRCC = 0.9528 ] +INFO:root:Average loss=0.07461446246504784 +INFO:root: +[Valid][14][UTT][ MSE = 0.2097 | LCC = 0.9526 | SRCC = 0.9536 ] +INFO:root:Average loss=0.07494621795415879 +INFO:root: +[Valid][15][UTT][ MSE = 0.1978 | LCC = 0.9537 | SRCC = 0.9546 ] +INFO:root:Average loss=0.07213274383544922 +INFO:root: +[Valid][16][UTT][ MSE = 0.1970 | LCC = 0.9527 | SRCC = 0.9537 ] +INFO:root:Average loss=0.06770552942156792 +INFO:root: +[Valid][17][UTT][ MSE = 0.2209 | LCC = 0.9534 | SRCC = 0.9550 ] +INFO:root:Average loss=0.06781864801049232 +INFO:root: +[Valid][18][UTT][ MSE = 0.1938 | LCC = 0.9541 | SRCC = 0.9556 ] +INFO:root:Average loss=0.06463908773660659 +INFO:root: +[Valid][19][UTT][ MSE = 0.2121 | LCC = 0.9553 | SRCC = 0.9566 ] +INFO:root:Average loss=0.059036570757627485 +INFO:root: +[Valid][20][UTT][ MSE = 0.2055 | LCC = 0.9554 | SRCC = 0.9565 ] +INFO:root:Average loss=0.06033591264486313 +INFO:root: +[Valid][21][UTT][ MSE = 0.2032 | LCC = 0.9550 | SRCC = 0.9561 ] +INFO:root:Average loss=0.055500189900398254 +INFO:root: +[Valid][22][UTT][ MSE = 0.2052 | LCC = 0.9552 | SRCC = 0.9562 ] +INFO:root:Average loss=0.052678505077958106 +INFO:root: +[Valid][23][UTT][ MSE = 0.1945 | LCC = 0.9562 | SRCC = 0.9570 ] +INFO:root:Average loss=0.051283110663294794 +INFO:root: +[Valid][24][UTT][ MSE = 0.1883 | LCC = 0.9547 | SRCC = 0.9558 ] +INFO:root:Average loss=0.05133038827776909 +INFO:root: +[Valid][25][UTT][ MSE = 0.1963 | LCC = 0.9557 | SRCC = 0.9567 ] +INFO:root:Average loss=0.053434564858675 +INFO:root: +[Valid][26][UTT][ MSE = 0.1968 | LCC = 0.9553 | SRCC = 0.9563 ] +INFO:root:Average loss=0.048416729986667635 +INFO:root: +[Valid][27][UTT][ MSE = 0.2110 | LCC = 0.9559 | SRCC = 0.9567 ] +INFO:root:Average loss=0.045006341621279713 +INFO:root: +[Valid][28][UTT][ MSE = 0.1969 | LCC = 0.9564 | SRCC = 0.9573 ] +INFO:root:Average loss=0.04412023569643497 +INFO:root: +[Valid][29][UTT][ MSE = 0.1891 | LCC = 0.9560 | SRCC = 0.9572 ] +INFO:root:Average loss=0.04120352537930012 +INFO:root: +[Valid][30][UTT][ MSE = 0.1900 | LCC = 0.9572 | SRCC = 0.9580 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1929 | LCC = 0.9581 | SRCC = 0.9597 ] diff --git a/results_tencent/w2v2_base/layer0/run2/config.gin b/results_tencent/w2v2_base/layer0/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer0/run2/model.pt b/results_tencent/w2v2_base/layer0/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..62bb3be830d61f82feb18536bf9991684f5364d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e6d014458e3c47050158461caf1e23b9853d73fcf26e9927a8bc0a058447b467 +size 2747979 diff --git a/results_tencent/w2v2_base/layer0/run2/model_best.pt b/results_tencent/w2v2_base/layer0/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e706f14d33f4666060c9c2bf234df1c30b6e6ca1 --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:77b18c8c65ab6b70fdc353c29e0dc9f451678332ea4b1e71757606569ec7301c +size 2749702 diff --git a/results_tencent/w2v2_base/layer0/run2/train.log b/results_tencent/w2v2_base/layer0/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6b1b0a2af6e02b52d2b36db82883ffeb4e133abb --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6351018522977829 +INFO:root: +[Valid][0][UTT][ MSE = 0.3683 | LCC = 0.9012 | SRCC = 0.9007 ] +INFO:root:Average loss=0.22912476527690886 +INFO:root: +[Valid][1][UTT][ MSE = 0.2842 | LCC = 0.9278 | SRCC = 0.9294 ] +INFO:root:Average loss=0.18553024297952653 +INFO:root: +[Valid][2][UTT][ MSE = 0.2618 | LCC = 0.9352 | SRCC = 0.9366 ] +INFO:root:Average loss=0.15586465471982955 +INFO:root: +[Valid][3][UTT][ MSE = 0.2483 | LCC = 0.9365 | SRCC = 0.9387 ] +INFO:root:Average loss=0.14755400055646897 +INFO:root: +[Valid][4][UTT][ MSE = 0.2426 | LCC = 0.9438 | SRCC = 0.9459 ] +INFO:root:Average loss=0.12881267070770264 +INFO:root: +[Valid][5][UTT][ MSE = 0.2517 | LCC = 0.9451 | SRCC = 0.9467 ] +INFO:root:Average loss=0.1233814281821251 +INFO:root: +[Valid][6][UTT][ MSE = 0.2075 | LCC = 0.9479 | SRCC = 0.9499 ] +INFO:root:Average loss=0.11569036644697189 +INFO:root: +[Valid][7][UTT][ MSE = 0.2496 | LCC = 0.9479 | SRCC = 0.9501 ] +INFO:root:Average loss=0.11082203799486161 +INFO:root: +[Valid][8][UTT][ MSE = 0.1914 | LCC = 0.9461 | SRCC = 0.9482 ] +INFO:root:Average loss=0.10248832476139069 +INFO:root: +[Valid][9][UTT][ MSE = 0.2266 | LCC = 0.9370 | SRCC = 0.9384 ] +INFO:root:Average loss=0.10206957650184631 +INFO:root: +[Valid][10][UTT][ MSE = 0.1928 | LCC = 0.9489 | SRCC = 0.9508 ] +INFO:root:Average loss=0.09402166324853897 +INFO:root: +[Valid][11][UTT][ MSE = 0.1916 | LCC = 0.9498 | SRCC = 0.9507 ] +INFO:root:Average loss=0.08773897880315781 +INFO:root: +[Valid][12][UTT][ MSE = 0.1955 | LCC = 0.9522 | SRCC = 0.9531 ] +INFO:root:Average loss=0.08329442968964576 +INFO:root: +[Valid][13][UTT][ MSE = 0.2213 | LCC = 0.9522 | SRCC = 0.9531 ] +INFO:root:Average loss=0.0794297407567501 +INFO:root: +[Valid][14][UTT][ MSE = 0.2013 | LCC = 0.9495 | SRCC = 0.9504 ] +INFO:root:Average loss=0.07486733770370484 +INFO:root: +[Valid][15][UTT][ MSE = 0.1891 | LCC = 0.9513 | SRCC = 0.9518 ] +INFO:root:Average loss=0.07635999566316605 +INFO:root: +[Valid][16][UTT][ MSE = 0.1997 | LCC = 0.9540 | SRCC = 0.9553 ] +INFO:root:Average loss=0.07089489984512329 +INFO:root: +[Valid][17][UTT][ MSE = 0.1931 | LCC = 0.9505 | SRCC = 0.9514 ] +INFO:root:Average loss=0.06867501300573349 +INFO:root: +[Valid][18][UTT][ MSE = 0.1996 | LCC = 0.9546 | SRCC = 0.9558 ] +INFO:root:Average loss=0.0652609361410141 +INFO:root: +[Valid][19][UTT][ MSE = 0.1978 | LCC = 0.9545 | SRCC = 0.9560 ] +INFO:root:Average loss=0.06029323041439057 +INFO:root: +[Valid][20][UTT][ MSE = 0.1916 | LCC = 0.9528 | SRCC = 0.9534 ] +INFO:root:Average loss=0.058304168552160265 +INFO:root: +[Valid][21][UTT][ MSE = 0.1941 | LCC = 0.9513 | SRCC = 0.9523 ] +INFO:root:Average loss=0.05614898610115051 +INFO:root: +[Valid][22][UTT][ MSE = 0.1923 | LCC = 0.9519 | SRCC = 0.9529 ] +INFO:root:Average loss=0.05611237877607345 +INFO:root: +[Valid][23][UTT][ MSE = 0.1927 | LCC = 0.9530 | SRCC = 0.9551 ] +INFO:root:Average loss=0.05392854595184326 +INFO:root: +[Valid][24][UTT][ MSE = 0.1913 | LCC = 0.9506 | SRCC = 0.9517 ] +INFO:root:Average loss=0.053025365024805066 +INFO:root: +[Valid][25][UTT][ MSE = 0.2129 | LCC = 0.9531 | SRCC = 0.9540 ] +INFO:root:Average loss=0.05065995900332928 +INFO:root: +[Valid][26][UTT][ MSE = 0.1927 | LCC = 0.9544 | SRCC = 0.9556 ] +INFO:root:Average loss=0.049712058633565905 +INFO:root: +[Valid][27][UTT][ MSE = 0.1748 | LCC = 0.9568 | SRCC = 0.9581 ] +INFO:root:Average loss=0.04651093254983425 +INFO:root: +[Valid][28][UTT][ MSE = 0.1755 | LCC = 0.9551 | SRCC = 0.9561 ] +INFO:root:Average loss=0.0453324781358242 +INFO:root: +[Valid][29][UTT][ MSE = 0.2019 | LCC = 0.9556 | SRCC = 0.9572 ] +INFO:root:Average loss=0.04560751175880432 +INFO:root: +[Valid][30][UTT][ MSE = 0.1825 | LCC = 0.9571 | SRCC = 0.9578 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1916 | LCC = 0.9570 | SRCC = 0.9579 ] diff --git a/results_tencent/w2v2_base/layer0/run3/config.gin b/results_tencent/w2v2_base/layer0/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer0/run3/model.pt b/results_tencent/w2v2_base/layer0/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2807a3703183a8244a3ae12af0e63090a7ed5312 --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2cadd52ab4ff564517a406a20636bcf5e04a91e5fc3c6f1391898bae4fa8373c +size 2747979 diff --git a/results_tencent/w2v2_base/layer0/run3/model_best.pt b/results_tencent/w2v2_base/layer0/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b521cb4cbc89e2d1fcf22c76ee0f825657cab08f --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6702be10dc2649ad9bd8e7b8d56e0ec31c9a852302ee93aa5151ff85b19c4aac +size 2749638 diff --git a/results_tencent/w2v2_base/layer0/run3/train.log b/results_tencent/w2v2_base/layer0/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ac0a620ff00cfc3805f2588f32a2be6b0adb8259 --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5296984796524048 +INFO:root: +[Valid][0][UTT][ MSE = 0.3072 | LCC = 0.9081 | SRCC = 0.9084 ] +INFO:root:Average loss=0.2306120902299881 +INFO:root: +[Valid][1][UTT][ MSE = 0.3022 | LCC = 0.9256 | SRCC = 0.9279 ] +INFO:root:Average loss=0.18412342357635497 +INFO:root: +[Valid][2][UTT][ MSE = 0.2430 | LCC = 0.9346 | SRCC = 0.9372 ] +INFO:root:Average loss=0.16500952047109604 +INFO:root: +[Valid][3][UTT][ MSE = 0.2712 | LCC = 0.9390 | SRCC = 0.9418 ] +INFO:root:Average loss=0.1472648048400879 +INFO:root: +[Valid][4][UTT][ MSE = 0.2273 | LCC = 0.9411 | SRCC = 0.9436 ] +INFO:root:Average loss=0.13274038177728653 +INFO:root: +[Valid][5][UTT][ MSE = 0.2381 | LCC = 0.9444 | SRCC = 0.9470 ] +INFO:root:Average loss=0.1219447524547577 +INFO:root: +[Valid][6][UTT][ MSE = 0.2190 | LCC = 0.9437 | SRCC = 0.9488 ] +INFO:root:Average loss=0.11503602278232575 +INFO:root: +[Valid][7][UTT][ MSE = 0.2308 | LCC = 0.9475 | SRCC = 0.9495 ] +INFO:root:Average loss=0.10294812867045403 +INFO:root: +[Valid][8][UTT][ MSE = 0.2299 | LCC = 0.9510 | SRCC = 0.9530 ] +INFO:root:Average loss=0.09852634960412979 +INFO:root: +[Valid][9][UTT][ MSE = 0.1986 | LCC = 0.9512 | SRCC = 0.9527 ] +INFO:root:Average loss=0.09113717365264892 +INFO:root: +[Valid][10][UTT][ MSE = 0.2212 | LCC = 0.9526 | SRCC = 0.9540 ] +INFO:root:Average loss=0.08745007339119912 +INFO:root: +[Valid][11][UTT][ MSE = 0.1976 | LCC = 0.9524 | SRCC = 0.9540 ] +INFO:root:Average loss=0.08790134426951408 +INFO:root: +[Valid][12][UTT][ MSE = 0.1855 | LCC = 0.9518 | SRCC = 0.9542 ] +INFO:root:Average loss=0.07957368794083595 +INFO:root: +[Valid][13][UTT][ MSE = 0.2171 | LCC = 0.9539 | SRCC = 0.9550 ] +INFO:root:Average loss=0.07586952623724938 +INFO:root: +[Valid][14][UTT][ MSE = 0.2069 | LCC = 0.9550 | SRCC = 0.9563 ] +INFO:root:Average loss=0.07165376955270768 +INFO:root: +[Valid][15][UTT][ MSE = 0.2170 | LCC = 0.9531 | SRCC = 0.9542 ] +INFO:root:Average loss=0.0700716701745987 +INFO:root: +[Valid][16][UTT][ MSE = 0.1948 | LCC = 0.9552 | SRCC = 0.9562 ] +INFO:root:Average loss=0.06709293469786644 +INFO:root: +[Valid][17][UTT][ MSE = 0.2001 | LCC = 0.9526 | SRCC = 0.9549 ] +INFO:root:Average loss=0.06538243478536605 +INFO:root: +[Valid][18][UTT][ MSE = 0.1968 | LCC = 0.9553 | SRCC = 0.9560 ] +INFO:root:Average loss=0.0598293809145689 +INFO:root: +[Valid][19][UTT][ MSE = 0.1949 | LCC = 0.9563 | SRCC = 0.9571 ] +INFO:root:Average loss=0.062248577058315276 +INFO:root: +[Valid][20][UTT][ MSE = 0.2018 | LCC = 0.9563 | SRCC = 0.9573 ] +INFO:root:Average loss=0.054849408954381944 +INFO:root: +[Valid][21][UTT][ MSE = 0.1890 | LCC = 0.9559 | SRCC = 0.9570 ] +INFO:root:Average loss=0.05636664815247059 +INFO:root: +[Valid][22][UTT][ MSE = 0.1754 | LCC = 0.9562 | SRCC = 0.9570 ] +INFO:root:Average loss=0.05527374270558357 +INFO:root: +[Valid][23][UTT][ MSE = 0.2200 | LCC = 0.9538 | SRCC = 0.9554 ] +INFO:root:Average loss=0.051405704110860824 +INFO:root: +[Valid][24][UTT][ MSE = 0.1664 | LCC = 0.9574 | SRCC = 0.9585 ] +INFO:root:Average loss=0.04921212635934353 +INFO:root: +[Valid][25][UTT][ MSE = 0.2066 | LCC = 0.9562 | SRCC = 0.9569 ] +INFO:root:Average loss=0.04766547568142414 +INFO:root: +[Valid][26][UTT][ MSE = 0.1824 | LCC = 0.9566 | SRCC = 0.9576 ] +INFO:root:Average loss=0.04608997729420662 +INFO:root: +[Valid][27][UTT][ MSE = 0.1863 | LCC = 0.9561 | SRCC = 0.9568 ] +INFO:root:Average loss=0.045872708663344385 +INFO:root: +[Valid][28][UTT][ MSE = 0.2195 | LCC = 0.9534 | SRCC = 0.9555 ] +INFO:root:Average loss=0.04459651826322079 +INFO:root: +[Valid][29][UTT][ MSE = 0.2021 | LCC = 0.9577 | SRCC = 0.9580 ] +INFO:root:Average loss=0.04414351361989975 +INFO:root: +[Valid][30][UTT][ MSE = 0.1849 | LCC = 0.9568 | SRCC = 0.9576 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2013 | LCC = 0.9580 | SRCC = 0.9588 ] diff --git a/results_tencent/w2v2_base/layer0/run4/config.gin b/results_tencent/w2v2_base/layer0/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer0/run4/model.pt b/results_tencent/w2v2_base/layer0/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..88e7d148d4bad23ce1617182b64f96427db7f88b --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:57fabbffbc9ca30c3f5a0e9e342e7c7237df608e24cd5e0b6e72f83d361e72f7 +size 2747979 diff --git a/results_tencent/w2v2_base/layer0/run4/model_best.pt b/results_tencent/w2v2_base/layer0/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d9d8f7c76945025b761276ba287171f44a7f6f65 --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:41071eab58a7810c9ca8e970105e5cd1c624ac50745869701585eaad2d0c69cd +size 2749638 diff --git a/results_tencent/w2v2_base/layer0/run4/train.log b/results_tencent/w2v2_base/layer0/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c61f5b2fb207f1a013e9505be6f3198c1ed2f0dc --- /dev/null +++ b/results_tencent/w2v2_base/layer0/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5260556327104569 +INFO:root: +[Valid][0][UTT][ MSE = 0.4046 | LCC = 0.8986 | SRCC = 0.9023 ] +INFO:root:Average loss=0.2372559632062912 +INFO:root: +[Valid][1][UTT][ MSE = 0.2997 | LCC = 0.9245 | SRCC = 0.9264 ] +INFO:root:Average loss=0.18590517973899842 +INFO:root: +[Valid][2][UTT][ MSE = 0.2772 | LCC = 0.9344 | SRCC = 0.9368 ] +INFO:root:Average loss=0.1627311697602272 +INFO:root: +[Valid][3][UTT][ MSE = 0.2381 | LCC = 0.9380 | SRCC = 0.9397 ] +INFO:root:Average loss=0.13966861814260484 +INFO:root: +[Valid][4][UTT][ MSE = 0.2617 | LCC = 0.9387 | SRCC = 0.9394 ] +INFO:root:Average loss=0.12859041285514833 +INFO:root: +[Valid][5][UTT][ MSE = 0.2272 | LCC = 0.9428 | SRCC = 0.9453 ] +INFO:root:Average loss=0.11799280506372452 +INFO:root: +[Valid][6][UTT][ MSE = 0.2493 | LCC = 0.9458 | SRCC = 0.9476 ] +INFO:root:Average loss=0.11295719945430756 +INFO:root: +[Valid][7][UTT][ MSE = 0.2347 | LCC = 0.9477 | SRCC = 0.9491 ] +INFO:root:Average loss=0.10501770967245103 +INFO:root: +[Valid][8][UTT][ MSE = 0.2044 | LCC = 0.9483 | SRCC = 0.9495 ] +INFO:root:Average loss=0.09967910942435265 +INFO:root: +[Valid][9][UTT][ MSE = 0.2055 | LCC = 0.9495 | SRCC = 0.9516 ] +INFO:root:Average loss=0.0928885864019394 +INFO:root: +[Valid][10][UTT][ MSE = 0.1937 | LCC = 0.9507 | SRCC = 0.9525 ] +INFO:root:Average loss=0.08677258577942848 +INFO:root: +[Valid][11][UTT][ MSE = 0.2135 | LCC = 0.9495 | SRCC = 0.9516 ] +INFO:root:Average loss=0.0872132907807827 +INFO:root: +[Valid][12][UTT][ MSE = 0.2092 | LCC = 0.9517 | SRCC = 0.9528 ] +INFO:root:Average loss=0.08048540252447128 +INFO:root: +[Valid][13][UTT][ MSE = 0.1865 | LCC = 0.9495 | SRCC = 0.9522 ] +INFO:root:Average loss=0.07958024787902832 +INFO:root: +[Valid][14][UTT][ MSE = 0.2100 | LCC = 0.9518 | SRCC = 0.9533 ] +INFO:root:Average loss=0.0708637655377388 +INFO:root: +[Valid][15][UTT][ MSE = 0.1942 | LCC = 0.9482 | SRCC = 0.9500 ] +INFO:root:Average loss=0.06927702268958091 +INFO:root: +[Valid][16][UTT][ MSE = 0.1686 | LCC = 0.9518 | SRCC = 0.9542 ] +INFO:root:Average loss=0.06990379083156585 +INFO:root: +[Valid][17][UTT][ MSE = 0.2089 | LCC = 0.9517 | SRCC = 0.9528 ] +INFO:root:Average loss=0.065038071423769 +INFO:root: +[Valid][18][UTT][ MSE = 0.2101 | LCC = 0.9524 | SRCC = 0.9528 ] +INFO:root:Average loss=0.06325235694646836 +INFO:root: +[Valid][19][UTT][ MSE = 0.1943 | LCC = 0.9537 | SRCC = 0.9549 ] +INFO:root:Average loss=0.05979622822999954 +INFO:root: +[Valid][20][UTT][ MSE = 0.2052 | LCC = 0.9524 | SRCC = 0.9529 ] +INFO:root:Average loss=0.05724852389097214 +INFO:root: +[Valid][21][UTT][ MSE = 0.1952 | LCC = 0.9535 | SRCC = 0.9545 ] +INFO:root:Average loss=0.05509547761082649 +INFO:root: +[Valid][22][UTT][ MSE = 0.1933 | LCC = 0.9543 | SRCC = 0.9555 ] +INFO:root:Average loss=0.05674790167808533 +INFO:root: +[Valid][23][UTT][ MSE = 0.1820 | LCC = 0.9526 | SRCC = 0.9543 ] +INFO:root:Average loss=0.05323053961992264 +INFO:root: +[Valid][24][UTT][ MSE = 0.1878 | LCC = 0.9535 | SRCC = 0.9547 ] +INFO:root:Average loss=0.05126519444584846 +INFO:root: +[Valid][25][UTT][ MSE = 0.1858 | LCC = 0.9498 | SRCC = 0.9513 ] +INFO:root:Average loss=0.04834258399903774 +INFO:root: +[Valid][26][UTT][ MSE = 0.2178 | LCC = 0.9527 | SRCC = 0.9535 ] +INFO:root:Average loss=0.04830804274976253 +INFO:root: +[Valid][27][UTT][ MSE = 0.2075 | LCC = 0.9543 | SRCC = 0.9553 ] +INFO:root:Average loss=0.04577871079742908 +INFO:root: +[Valid][28][UTT][ MSE = 0.1999 | LCC = 0.9542 | SRCC = 0.9556 ] +INFO:root:Average loss=0.043504351347684864 +INFO:root: +[Valid][29][UTT][ MSE = 0.1794 | LCC = 0.9520 | SRCC = 0.9539 ] +INFO:root:Average loss=0.04411762155592441 +INFO:root: +[Valid][30][UTT][ MSE = 0.1905 | LCC = 0.9541 | SRCC = 0.9544 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1936 | LCC = 0.9563 | SRCC = 0.9571 ] diff --git a/results_tencent/w2v2_base/layer1/run0/config.gin b/results_tencent/w2v2_base/layer1/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer1/run0/model.pt b/results_tencent/w2v2_base/layer1/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0b6495101ad362cd2f2c70e1e3070abc50c2ae21 --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:70fe29af6ace2a476ee3d25320386114995e253449762983a6d7c6e21ce132bf +size 2747979 diff --git a/results_tencent/w2v2_base/layer1/run0/model_best.pt b/results_tencent/w2v2_base/layer1/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..1bc18f3fa29200f53b23402d6a4b933edf314365 --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1dda050993497ab9189f4f2e8566a242aa35601749e568c861b6b1378771bb54 +size 2749702 diff --git a/results_tencent/w2v2_base/layer1/run0/train.log b/results_tencent/w2v2_base/layer1/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..11afa8299ed0c3dec768e61a546971ec93720474 --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.45640522229671476 +INFO:root: +[Valid][0][UTT][ MSE = 0.3149 | LCC = 0.9232 | SRCC = 0.9254 ] +INFO:root:Average loss=0.19531934517621993 +INFO:root: +[Valid][1][UTT][ MSE = 0.2471 | LCC = 0.9415 | SRCC = 0.9436 ] +INFO:root:Average loss=0.14997590404748917 +INFO:root: +[Valid][2][UTT][ MSE = 0.2145 | LCC = 0.9453 | SRCC = 0.9463 ] +INFO:root:Average loss=0.12921615374088288 +INFO:root: +[Valid][3][UTT][ MSE = 0.2267 | LCC = 0.9502 | SRCC = 0.9513 ] +INFO:root:Average loss=0.11758265376091004 +INFO:root: +[Valid][4][UTT][ MSE = 0.1841 | LCC = 0.9529 | SRCC = 0.9553 ] +INFO:root:Average loss=0.10765980544686317 +INFO:root: +[Valid][5][UTT][ MSE = 0.2034 | LCC = 0.9540 | SRCC = 0.9552 ] +INFO:root:Average loss=0.09764360904693603 +INFO:root: +[Valid][6][UTT][ MSE = 0.2252 | LCC = 0.9569 | SRCC = 0.9582 ] +INFO:root:Average loss=0.0905094248354435 +INFO:root: +[Valid][7][UTT][ MSE = 0.2058 | LCC = 0.9551 | SRCC = 0.9564 ] +INFO:root:Average loss=0.08585129803419113 +INFO:root: +[Valid][8][UTT][ MSE = 0.1867 | LCC = 0.9576 | SRCC = 0.9583 ] +INFO:root:Average loss=0.08117254489660262 +INFO:root: +[Valid][9][UTT][ MSE = 0.1937 | LCC = 0.9564 | SRCC = 0.9577 ] +INFO:root:Average loss=0.07328682920336724 +INFO:root: +[Valid][10][UTT][ MSE = 0.1746 | LCC = 0.9587 | SRCC = 0.9596 ] +INFO:root:Average loss=0.06986144065856933 +INFO:root: +[Valid][11][UTT][ MSE = 0.1953 | LCC = 0.9599 | SRCC = 0.9607 ] +INFO:root:Average loss=0.0647151803970337 +INFO:root: +[Valid][12][UTT][ MSE = 0.1836 | LCC = 0.9599 | SRCC = 0.9606 ] +INFO:root:Average loss=0.06323978868126869 +INFO:root: +[Valid][13][UTT][ MSE = 0.1712 | LCC = 0.9589 | SRCC = 0.9598 ] +INFO:root:Average loss=0.05977163076400757 +INFO:root: +[Valid][14][UTT][ MSE = 0.1675 | LCC = 0.9602 | SRCC = 0.9611 ] +INFO:root:Average loss=0.05677146202325821 +INFO:root: +[Valid][15][UTT][ MSE = 0.2023 | LCC = 0.9602 | SRCC = 0.9616 ] +INFO:root:Average loss=0.052898502856492995 +INFO:root: +[Valid][16][UTT][ MSE = 0.1818 | LCC = 0.9604 | SRCC = 0.9614 ] +INFO:root:Average loss=0.05103454594314098 +INFO:root: +[Valid][17][UTT][ MSE = 0.1792 | LCC = 0.9598 | SRCC = 0.9605 ] +INFO:root:Average loss=0.05008603894710541 +INFO:root: +[Valid][18][UTT][ MSE = 0.1857 | LCC = 0.9602 | SRCC = 0.9615 ] +INFO:root:Average loss=0.048748465433716776 +INFO:root: +[Valid][19][UTT][ MSE = 0.1656 | LCC = 0.9587 | SRCC = 0.9604 ] +INFO:root:Average loss=0.047746947899460795 +INFO:root: +[Valid][20][UTT][ MSE = 0.1680 | LCC = 0.9617 | SRCC = 0.9623 ] +INFO:root:Average loss=0.046190849006175996 +INFO:root: +[Valid][21][UTT][ MSE = 0.1937 | LCC = 0.9607 | SRCC = 0.9612 ] +INFO:root:Average loss=0.042409289106726644 +INFO:root: +[Valid][22][UTT][ MSE = 0.1743 | LCC = 0.9612 | SRCC = 0.9616 ] +INFO:root:Average loss=0.03816517156362534 +INFO:root: +[Valid][23][UTT][ MSE = 0.1948 | LCC = 0.9607 | SRCC = 0.9619 ] +INFO:root:Average loss=0.04200173392891884 +INFO:root: +[Valid][24][UTT][ MSE = 0.1996 | LCC = 0.9600 | SRCC = 0.9610 ] +INFO:root:Average loss=0.037020873710513116 +INFO:root: +[Valid][25][UTT][ MSE = 0.1714 | LCC = 0.9624 | SRCC = 0.9627 ] +INFO:root:Average loss=0.037491687849164006 +INFO:root: +[Valid][26][UTT][ MSE = 0.1625 | LCC = 0.9622 | SRCC = 0.9624 ] +INFO:root:Average loss=0.035738135039806364 +INFO:root: +[Valid][27][UTT][ MSE = 0.1807 | LCC = 0.9618 | SRCC = 0.9625 ] +INFO:root:Average loss=0.033393235355615614 +INFO:root: +[Valid][28][UTT][ MSE = 0.1616 | LCC = 0.9623 | SRCC = 0.9626 ] +INFO:root:Average loss=0.03314866365492344 +INFO:root: +[Valid][29][UTT][ MSE = 0.1619 | LCC = 0.9616 | SRCC = 0.9628 ] +INFO:root:Average loss=0.03191780799627304 +INFO:root: +[Valid][30][UTT][ MSE = 0.1881 | LCC = 0.9612 | SRCC = 0.9622 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1777 | LCC = 0.9620 | SRCC = 0.9630 ] diff --git a/results_tencent/w2v2_base/layer1/run1/config.gin b/results_tencent/w2v2_base/layer1/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer1/run1/model.pt b/results_tencent/w2v2_base/layer1/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f2c4d7ae5bd34db077cce9d7293b27760592f907 --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c14075e5e7da3710745842561653f214c347a1caea1093497d31cd9ce5043cd0 +size 2747979 diff --git a/results_tencent/w2v2_base/layer1/run1/model_best.pt b/results_tencent/w2v2_base/layer1/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a5c9bfe6f2f977308d4a778f3dd3f14379b2821a --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4011c4935d726c695be64524bd8bd4a668dd2f92d90a4b62ff43cf16947d54b1 +size 2749702 diff --git a/results_tencent/w2v2_base/layer1/run1/train.log b/results_tencent/w2v2_base/layer1/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d83672700d0b264f15a83e5d3eef553f20a4244c --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4867064747810364 +INFO:root: +[Valid][0][UTT][ MSE = 0.2929 | LCC = 0.9216 | SRCC = 0.9242 ] +INFO:root:Average loss=0.19056438982486726 +INFO:root: +[Valid][1][UTT][ MSE = 0.2402 | LCC = 0.9410 | SRCC = 0.9431 ] +INFO:root:Average loss=0.15692429357767104 +INFO:root: +[Valid][2][UTT][ MSE = 0.2270 | LCC = 0.9455 | SRCC = 0.9476 ] +INFO:root:Average loss=0.1349184358716011 +INFO:root: +[Valid][3][UTT][ MSE = 0.2066 | LCC = 0.9498 | SRCC = 0.9509 ] +INFO:root:Average loss=0.11766123795509338 +INFO:root: +[Valid][4][UTT][ MSE = 0.2322 | LCC = 0.9520 | SRCC = 0.9529 ] +INFO:root:Average loss=0.1132155711054802 +INFO:root: +[Valid][5][UTT][ MSE = 0.2244 | LCC = 0.9532 | SRCC = 0.9544 ] +INFO:root:Average loss=0.10223063004016876 +INFO:root: +[Valid][6][UTT][ MSE = 0.2312 | LCC = 0.9543 | SRCC = 0.9555 ] +INFO:root:Average loss=0.0932825830578804 +INFO:root: +[Valid][7][UTT][ MSE = 0.1883 | LCC = 0.9570 | SRCC = 0.9574 ] +INFO:root:Average loss=0.0914798103272915 +INFO:root: +[Valid][8][UTT][ MSE = 0.2002 | LCC = 0.9552 | SRCC = 0.9559 ] +INFO:root:Average loss=0.08182309168577194 +INFO:root: +[Valid][9][UTT][ MSE = 0.1973 | LCC = 0.9581 | SRCC = 0.9586 ] +INFO:root:Average loss=0.07968087750673294 +INFO:root: +[Valid][10][UTT][ MSE = 0.1921 | LCC = 0.9575 | SRCC = 0.9582 ] +INFO:root:Average loss=0.07370896899700165 +INFO:root: +[Valid][11][UTT][ MSE = 0.1800 | LCC = 0.9576 | SRCC = 0.9581 ] +INFO:root:Average loss=0.0689657855629921 +INFO:root: +[Valid][12][UTT][ MSE = 0.2183 | LCC = 0.9575 | SRCC = 0.9581 ] +INFO:root:Average loss=0.06536097088456154 +INFO:root: +[Valid][13][UTT][ MSE = 0.2166 | LCC = 0.9583 | SRCC = 0.9588 ] +INFO:root:Average loss=0.0624329988360405 +INFO:root: +[Valid][14][UTT][ MSE = 0.1856 | LCC = 0.9584 | SRCC = 0.9589 ] +INFO:root:Average loss=0.059975528806447985 +INFO:root: +[Valid][15][UTT][ MSE = 0.2048 | LCC = 0.9584 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05709325933456421 +INFO:root: +[Valid][16][UTT][ MSE = 0.1895 | LCC = 0.9577 | SRCC = 0.9581 ] +INFO:root:Average loss=0.053723325073719026 +INFO:root: +[Valid][17][UTT][ MSE = 0.2165 | LCC = 0.9598 | SRCC = 0.9600 ] +INFO:root:Average loss=0.051211911663413046 +INFO:root: +[Valid][18][UTT][ MSE = 0.1802 | LCC = 0.9589 | SRCC = 0.9593 ] +INFO:root:Average loss=0.0493109487593174 +INFO:root: +[Valid][19][UTT][ MSE = 0.1789 | LCC = 0.9586 | SRCC = 0.9589 ] +INFO:root:Average loss=0.04806926667690277 +INFO:root: +[Valid][20][UTT][ MSE = 0.1784 | LCC = 0.9587 | SRCC = 0.9593 ] +INFO:root:Average loss=0.045756668522953986 +INFO:root: +[Valid][21][UTT][ MSE = 0.1911 | LCC = 0.9606 | SRCC = 0.9610 ] +INFO:root:Average loss=0.043777047008275984 +INFO:root: +[Valid][22][UTT][ MSE = 0.1955 | LCC = 0.9585 | SRCC = 0.9592 ] +INFO:root:Average loss=0.04137430752813816 +INFO:root: +[Valid][23][UTT][ MSE = 0.1754 | LCC = 0.9603 | SRCC = 0.9607 ] +INFO:root:Average loss=0.038788695737719536 +INFO:root: +[Valid][24][UTT][ MSE = 0.1745 | LCC = 0.9605 | SRCC = 0.9606 ] +INFO:root:Average loss=0.03872829085588455 +INFO:root: +[Valid][25][UTT][ MSE = 0.1632 | LCC = 0.9604 | SRCC = 0.9606 ] +INFO:root:Average loss=0.03766575840115547 +INFO:root: +[Valid][26][UTT][ MSE = 0.1792 | LCC = 0.9593 | SRCC = 0.9596 ] +INFO:root:Average loss=0.03657369640469551 +INFO:root: +[Valid][27][UTT][ MSE = 0.1802 | LCC = 0.9611 | SRCC = 0.9613 ] +INFO:root:Average loss=0.03644911700487137 +INFO:root: +[Valid][28][UTT][ MSE = 0.1785 | LCC = 0.9610 | SRCC = 0.9616 ] +INFO:root:Average loss=0.03436491799354553 +INFO:root: +[Valid][29][UTT][ MSE = 0.1821 | LCC = 0.9610 | SRCC = 0.9616 ] +INFO:root:Average loss=0.033972616419196126 +INFO:root: +[Valid][30][UTT][ MSE = 0.1811 | LCC = 0.9606 | SRCC = 0.9610 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1854 | LCC = 0.9610 | SRCC = 0.9613 ] diff --git a/results_tencent/w2v2_base/layer1/run2/config.gin b/results_tencent/w2v2_base/layer1/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer1/run2/model.pt b/results_tencent/w2v2_base/layer1/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d17dca14f4245394029f1b39d72044486cb5fc9d --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:85ae336d729b5e654b15c51175aa94544dae1d56a1a392785c3bd23775e71bf3 +size 2747979 diff --git a/results_tencent/w2v2_base/layer1/run2/model_best.pt b/results_tencent/w2v2_base/layer1/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..887bf48b58b85a65f80a36f63ec2401e0a93ce2c --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fdae3a6da4e8aa8ff6e4fb97edcb010fb713fc3c65ed7b6f9b869464e45e9434 +size 2749702 diff --git a/results_tencent/w2v2_base/layer1/run2/train.log b/results_tencent/w2v2_base/layer1/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..49f02c0699cf093911ad86ffb1aaa5a6bf95b6cf --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4679189817905426 +INFO:root: +[Valid][0][UTT][ MSE = 0.3198 | LCC = 0.9107 | SRCC = 0.9154 ] +INFO:root:Average loss=0.1971645611524582 +INFO:root: +[Valid][1][UTT][ MSE = 0.2395 | LCC = 0.9335 | SRCC = 0.9365 ] +INFO:root:Average loss=0.1580924564599991 +INFO:root: +[Valid][2][UTT][ MSE = 0.2093 | LCC = 0.9420 | SRCC = 0.9443 ] +INFO:root:Average loss=0.1411811403632164 +INFO:root: +[Valid][3][UTT][ MSE = 0.2004 | LCC = 0.9480 | SRCC = 0.9500 ] +INFO:root:Average loss=0.12428836822509766 +INFO:root: +[Valid][4][UTT][ MSE = 0.2224 | LCC = 0.9515 | SRCC = 0.9531 ] +INFO:root:Average loss=0.1136659723520279 +INFO:root: +[Valid][5][UTT][ MSE = 0.2025 | LCC = 0.9521 | SRCC = 0.9541 ] +INFO:root:Average loss=0.10267656794190406 +INFO:root: +[Valid][6][UTT][ MSE = 0.1892 | LCC = 0.9542 | SRCC = 0.9554 ] +INFO:root:Average loss=0.09374416610598564 +INFO:root: +[Valid][7][UTT][ MSE = 0.1890 | LCC = 0.9567 | SRCC = 0.9576 ] +INFO:root:Average loss=0.0910769615471363 +INFO:root: +[Valid][8][UTT][ MSE = 0.2175 | LCC = 0.9494 | SRCC = 0.9532 ] +INFO:root:Average loss=0.08412077647447586 +INFO:root: +[Valid][9][UTT][ MSE = 0.2152 | LCC = 0.9558 | SRCC = 0.9572 ] +INFO:root:Average loss=0.0822318531870842 +INFO:root: +[Valid][10][UTT][ MSE = 0.1870 | LCC = 0.9569 | SRCC = 0.9579 ] +INFO:root:Average loss=0.07307481601834297 +INFO:root: +[Valid][11][UTT][ MSE = 0.1883 | LCC = 0.9575 | SRCC = 0.9578 ] +INFO:root:Average loss=0.06859948736429214 +INFO:root: +[Valid][12][UTT][ MSE = 0.1651 | LCC = 0.9564 | SRCC = 0.9575 ] +INFO:root:Average loss=0.06538766169548035 +INFO:root: +[Valid][13][UTT][ MSE = 0.2091 | LCC = 0.9578 | SRCC = 0.9579 ] +INFO:root:Average loss=0.0643108220398426 +INFO:root: +[Valid][14][UTT][ MSE = 0.1830 | LCC = 0.9586 | SRCC = 0.9592 ] +INFO:root:Average loss=0.06069331595301628 +INFO:root: +[Valid][15][UTT][ MSE = 0.1757 | LCC = 0.9589 | SRCC = 0.9595 ] +INFO:root:Average loss=0.05666586083173752 +INFO:root: +[Valid][16][UTT][ MSE = 0.1846 | LCC = 0.9585 | SRCC = 0.9593 ] +INFO:root:Average loss=0.05369610303640365 +INFO:root: +[Valid][17][UTT][ MSE = 0.1730 | LCC = 0.9568 | SRCC = 0.9580 ] +INFO:root:Average loss=0.053409736782312395 +INFO:root: +[Valid][18][UTT][ MSE = 0.1672 | LCC = 0.9589 | SRCC = 0.9596 ] +INFO:root:Average loss=0.05177490155398846 +INFO:root: +[Valid][19][UTT][ MSE = 0.1700 | LCC = 0.9600 | SRCC = 0.9604 ] +INFO:root:Average loss=0.04862731626629829 +INFO:root: +[Valid][20][UTT][ MSE = 0.1709 | LCC = 0.9597 | SRCC = 0.9608 ] +INFO:root:Average loss=0.04602697883546352 +INFO:root: +[Valid][21][UTT][ MSE = 0.1739 | LCC = 0.9590 | SRCC = 0.9602 ] +INFO:root:Average loss=0.045486462384462356 +INFO:root: +[Valid][22][UTT][ MSE = 0.1780 | LCC = 0.9600 | SRCC = 0.9604 ] +INFO:root:Average loss=0.04374189499020577 +INFO:root: +[Valid][23][UTT][ MSE = 0.1949 | LCC = 0.9598 | SRCC = 0.9601 ] +INFO:root:Average loss=0.043816741839051244 +INFO:root: +[Valid][24][UTT][ MSE = 0.1613 | LCC = 0.9590 | SRCC = 0.9593 ] +INFO:root:Average loss=0.04040864227712154 +INFO:root: +[Valid][25][UTT][ MSE = 0.1749 | LCC = 0.9607 | SRCC = 0.9610 ] +INFO:root:Average loss=0.040567083552479746 +INFO:root: +[Valid][26][UTT][ MSE = 0.1815 | LCC = 0.9592 | SRCC = 0.9597 ] +INFO:root:Average loss=0.037591126382350924 +INFO:root: +[Valid][27][UTT][ MSE = 0.1667 | LCC = 0.9595 | SRCC = 0.9601 ] +INFO:root:Average loss=0.03693207912147045 +INFO:root: +[Valid][28][UTT][ MSE = 0.1863 | LCC = 0.9610 | SRCC = 0.9617 ] +INFO:root:Average loss=0.03463270157575607 +INFO:root: +[Valid][29][UTT][ MSE = 0.1711 | LCC = 0.9596 | SRCC = 0.9607 ] +INFO:root:Average loss=0.034355081751942636 +INFO:root: +[Valid][30][UTT][ MSE = 0.1768 | LCC = 0.9612 | SRCC = 0.9615 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1835 | LCC = 0.9607 | SRCC = 0.9612 ] diff --git a/results_tencent/w2v2_base/layer1/run3/config.gin b/results_tencent/w2v2_base/layer1/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer1/run3/model.pt b/results_tencent/w2v2_base/layer1/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5fbcc1165ba2e3b193c6c55d5e6d3e0a6e902a2a --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b088806d6fb1d85e9b1801b640d74c07427ae4973ea7de88d5447b8e2a244d93 +size 2747979 diff --git a/results_tencent/w2v2_base/layer1/run3/model_best.pt b/results_tencent/w2v2_base/layer1/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ea7b6603d1448663d554b43b8694a8b8bd9e426b --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d643cb954b2c7f41ff2e3ebe7a40df524d6ba8de9c34298de08beaf55fdabcea +size 2749638 diff --git a/results_tencent/w2v2_base/layer1/run3/train.log b/results_tencent/w2v2_base/layer1/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9d6ccc1d77820ebcc30edfccbdebd7940e817be6 --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44060325968265535 +INFO:root: +[Valid][0][UTT][ MSE = 0.2683 | LCC = 0.9230 | SRCC = 0.9247 ] +INFO:root:Average loss=0.19441159582138062 +INFO:root: +[Valid][1][UTT][ MSE = 0.2499 | LCC = 0.9372 | SRCC = 0.9383 ] +INFO:root:Average loss=0.15161423456668854 +INFO:root: +[Valid][2][UTT][ MSE = 0.2442 | LCC = 0.9461 | SRCC = 0.9475 ] +INFO:root:Average loss=0.13187498086690902 +INFO:root: +[Valid][3][UTT][ MSE = 0.1965 | LCC = 0.9491 | SRCC = 0.9496 ] +INFO:root:Average loss=0.12364903029799461 +INFO:root: +[Valid][4][UTT][ MSE = 0.2063 | LCC = 0.9521 | SRCC = 0.9532 ] +INFO:root:Average loss=0.10937597727775573 +INFO:root: +[Valid][5][UTT][ MSE = 0.2051 | LCC = 0.9534 | SRCC = 0.9544 ] +INFO:root:Average loss=0.10086165308952331 +INFO:root: +[Valid][6][UTT][ MSE = 0.2002 | LCC = 0.9557 | SRCC = 0.9562 ] +INFO:root:Average loss=0.09423695343732834 +INFO:root: +[Valid][7][UTT][ MSE = 0.2177 | LCC = 0.9547 | SRCC = 0.9563 ] +INFO:root:Average loss=0.08699205350875855 +INFO:root: +[Valid][8][UTT][ MSE = 0.1786 | LCC = 0.9561 | SRCC = 0.9573 ] +INFO:root:Average loss=0.0836032447218895 +INFO:root: +[Valid][9][UTT][ MSE = 0.1761 | LCC = 0.9559 | SRCC = 0.9572 ] +INFO:root:Average loss=0.07866504128277302 +INFO:root: +[Valid][10][UTT][ MSE = 0.1706 | LCC = 0.9580 | SRCC = 0.9590 ] +INFO:root:Average loss=0.07340481725335121 +INFO:root: +[Valid][11][UTT][ MSE = 0.1823 | LCC = 0.9589 | SRCC = 0.9593 ] +INFO:root:Average loss=0.06985837554931641 +INFO:root: +[Valid][12][UTT][ MSE = 0.1729 | LCC = 0.9588 | SRCC = 0.9597 ] +INFO:root:Average loss=0.06545016875863076 +INFO:root: +[Valid][13][UTT][ MSE = 0.1857 | LCC = 0.9579 | SRCC = 0.9585 ] +INFO:root:Average loss=0.06639567077159882 +INFO:root: +[Valid][14][UTT][ MSE = 0.2006 | LCC = 0.9565 | SRCC = 0.9571 ] +INFO:root:Average loss=0.06101396816968918 +INFO:root: +[Valid][15][UTT][ MSE = 0.1856 | LCC = 0.9592 | SRCC = 0.9594 ] +INFO:root:Average loss=0.055424198895692826 +INFO:root: +[Valid][16][UTT][ MSE = 0.1774 | LCC = 0.9584 | SRCC = 0.9588 ] +INFO:root:Average loss=0.0541675043553114 +INFO:root: +[Valid][17][UTT][ MSE = 0.1733 | LCC = 0.9579 | SRCC = 0.9584 ] +INFO:root:Average loss=0.05191429939866066 +INFO:root: +[Valid][18][UTT][ MSE = 0.1693 | LCC = 0.9598 | SRCC = 0.9605 ] +INFO:root:Average loss=0.050232640534639356 +INFO:root: +[Valid][19][UTT][ MSE = 0.1742 | LCC = 0.9586 | SRCC = 0.9595 ] +INFO:root:Average loss=0.04653973065316677 +INFO:root: +[Valid][20][UTT][ MSE = 0.1596 | LCC = 0.9572 | SRCC = 0.9579 ] +INFO:root:Average loss=0.04624857169389725 +INFO:root: +[Valid][21][UTT][ MSE = 0.1779 | LCC = 0.9599 | SRCC = 0.9603 ] +INFO:root:Average loss=0.04753159523010254 +INFO:root: +[Valid][22][UTT][ MSE = 0.1700 | LCC = 0.9597 | SRCC = 0.9601 ] +INFO:root:Average loss=0.04339529018104076 +INFO:root: +[Valid][23][UTT][ MSE = 0.1723 | LCC = 0.9574 | SRCC = 0.9579 ] +INFO:root:Average loss=0.039999673679471014 +INFO:root: +[Valid][24][UTT][ MSE = 0.1658 | LCC = 0.9603 | SRCC = 0.9603 ] +INFO:root:Average loss=0.042033637762069705 +INFO:root: +[Valid][25][UTT][ MSE = 0.1836 | LCC = 0.9588 | SRCC = 0.9593 ] +INFO:root:Average loss=0.038541323125362396 +INFO:root: +[Valid][26][UTT][ MSE = 0.1659 | LCC = 0.9591 | SRCC = 0.9592 ] +INFO:root:Average loss=0.03680554014444351 +INFO:root: +[Valid][27][UTT][ MSE = 0.1599 | LCC = 0.9579 | SRCC = 0.9590 ] +INFO:root:Average loss=0.03647517862915993 +INFO:root: +[Valid][28][UTT][ MSE = 0.1758 | LCC = 0.9599 | SRCC = 0.9602 ] +INFO:root:Average loss=0.03584339645504951 +INFO:root: +[Valid][29][UTT][ MSE = 0.1491 | LCC = 0.9597 | SRCC = 0.9603 ] +INFO:root:Average loss=0.03338309651613235 +INFO:root: +[Valid][30][UTT][ MSE = 0.1724 | LCC = 0.9589 | SRCC = 0.9594 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1622 | LCC = 0.9622 | SRCC = 0.9630 ] diff --git a/results_tencent/w2v2_base/layer1/run4/config.gin b/results_tencent/w2v2_base/layer1/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer1/run4/model.pt b/results_tencent/w2v2_base/layer1/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..257ed9b0ba5c42bef9c851e01d278b47be16797b --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:38cdce1fa75f80e2f2ed5025b7768fcc933de77bd66e087769dba71220fffe63 +size 2747979 diff --git a/results_tencent/w2v2_base/layer1/run4/model_best.pt b/results_tencent/w2v2_base/layer1/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..768e3ee18dd20248e8e51e10ac4628d38fb6e5f8 --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:15d1fec66ac79780fee1b3a3c64bda6657e353eda8cd85a1bf476cba88718339 +size 2749638 diff --git a/results_tencent/w2v2_base/layer1/run4/train.log b/results_tencent/w2v2_base/layer1/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e1f9a974fc8b40a89500f05f7f28771c210423e6 --- /dev/null +++ b/results_tencent/w2v2_base/layer1/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5546344001293182 +INFO:root: +[Valid][0][UTT][ MSE = 0.3180 | LCC = 0.9142 | SRCC = 0.9152 ] +INFO:root:Average loss=0.2039961130619049 +INFO:root: +[Valid][1][UTT][ MSE = 0.2685 | LCC = 0.9347 | SRCC = 0.9359 ] +INFO:root:Average loss=0.15930748915672302 +INFO:root: +[Valid][2][UTT][ MSE = 0.2676 | LCC = 0.9433 | SRCC = 0.9446 ] +INFO:root:Average loss=0.14249431270360946 +INFO:root: +[Valid][3][UTT][ MSE = 0.2444 | LCC = 0.9476 | SRCC = 0.9488 ] +INFO:root:Average loss=0.12399727666378021 +INFO:root: +[Valid][4][UTT][ MSE = 0.2488 | LCC = 0.9494 | SRCC = 0.9507 ] +INFO:root:Average loss=0.1209043426513672 +INFO:root: +[Valid][5][UTT][ MSE = 0.2130 | LCC = 0.9521 | SRCC = 0.9532 ] +INFO:root:Average loss=0.10423259687423705 +INFO:root: +[Valid][6][UTT][ MSE = 0.2111 | LCC = 0.9529 | SRCC = 0.9538 ] +INFO:root:Average loss=0.09910017561912536 +INFO:root: +[Valid][7][UTT][ MSE = 0.1783 | LCC = 0.9528 | SRCC = 0.9548 ] +INFO:root:Average loss=0.09024073246121407 +INFO:root: +[Valid][8][UTT][ MSE = 0.1872 | LCC = 0.9518 | SRCC = 0.9545 ] +INFO:root:Average loss=0.08527677699923515 +INFO:root: +[Valid][9][UTT][ MSE = 0.1796 | LCC = 0.9539 | SRCC = 0.9563 ] +INFO:root:Average loss=0.08087776726484298 +INFO:root: +[Valid][10][UTT][ MSE = 0.2446 | LCC = 0.9548 | SRCC = 0.9564 ] +INFO:root:Average loss=0.07976939076185227 +INFO:root: +[Valid][11][UTT][ MSE = 0.2075 | LCC = 0.9565 | SRCC = 0.9577 ] +INFO:root:Average loss=0.06937154093384743 +INFO:root: +[Valid][12][UTT][ MSE = 0.1946 | LCC = 0.9561 | SRCC = 0.9568 ] +INFO:root:Average loss=0.06770364883542061 +INFO:root: +[Valid][13][UTT][ MSE = 0.2031 | LCC = 0.9572 | SRCC = 0.9578 ] +INFO:root:Average loss=0.060803672298789024 +INFO:root: +[Valid][14][UTT][ MSE = 0.1929 | LCC = 0.9558 | SRCC = 0.9571 ] +INFO:root:Average loss=0.061345260992646214 +INFO:root: +[Valid][15][UTT][ MSE = 0.2056 | LCC = 0.9567 | SRCC = 0.9574 ] +INFO:root:Average loss=0.056969826012849806 +INFO:root: +[Valid][16][UTT][ MSE = 0.1871 | LCC = 0.9549 | SRCC = 0.9561 ] +INFO:root:Average loss=0.054213152557611465 +INFO:root: +[Valid][17][UTT][ MSE = 0.2198 | LCC = 0.9580 | SRCC = 0.9586 ] +INFO:root:Average loss=0.053705320566892624 +INFO:root: +[Valid][18][UTT][ MSE = 0.1772 | LCC = 0.9582 | SRCC = 0.9591 ] +INFO:root:Average loss=0.04936779950559139 +INFO:root: +[Valid][19][UTT][ MSE = 0.1851 | LCC = 0.9570 | SRCC = 0.9585 ] +INFO:root:Average loss=0.04893452313542366 +INFO:root: +[Valid][20][UTT][ MSE = 0.1833 | LCC = 0.9580 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04843881517648697 +INFO:root: +[Valid][21][UTT][ MSE = 0.1911 | LCC = 0.9561 | SRCC = 0.9571 ] +INFO:root:Average loss=0.04449731239676476 +INFO:root: +[Valid][22][UTT][ MSE = 0.2141 | LCC = 0.9594 | SRCC = 0.9600 ] +INFO:root:Average loss=0.045031165316700934 +INFO:root: +[Valid][23][UTT][ MSE = 0.1878 | LCC = 0.9577 | SRCC = 0.9588 ] +INFO:root:Average loss=0.04221521429717541 +INFO:root: +[Valid][24][UTT][ MSE = 0.1947 | LCC = 0.9591 | SRCC = 0.9598 ] +INFO:root:Average loss=0.039197355270385745 +INFO:root: +[Valid][25][UTT][ MSE = 0.1944 | LCC = 0.9590 | SRCC = 0.9599 ] +INFO:root:Average loss=0.0376873839944601 +INFO:root: +[Valid][26][UTT][ MSE = 0.1744 | LCC = 0.9576 | SRCC = 0.9590 ] +INFO:root:Average loss=0.03618018989264965 +INFO:root: +[Valid][27][UTT][ MSE = 0.1855 | LCC = 0.9566 | SRCC = 0.9572 ] +INFO:root:Average loss=0.034882146790623667 +INFO:root: +[Valid][28][UTT][ MSE = 0.1792 | LCC = 0.9600 | SRCC = 0.9606 ] +INFO:root:Average loss=0.03542292536795139 +INFO:root: +[Valid][29][UTT][ MSE = 0.1847 | LCC = 0.9596 | SRCC = 0.9599 ] +INFO:root:Average loss=0.03429706537723541 +INFO:root: +[Valid][30][UTT][ MSE = 0.1821 | LCC = 0.9591 | SRCC = 0.9598 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1847 | LCC = 0.9601 | SRCC = 0.9604 ] diff --git a/results_tencent/w2v2_base/layer10/run0/config.gin b/results_tencent/w2v2_base/layer10/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer10/run0/model.pt b/results_tencent/w2v2_base/layer10/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ef56b821003de6eb4907d7e413801e8c2152ce5f --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:afa070331de7a837e58409f66bb477a7e0990d445ca316e0bebfea4abfb1b69e +size 2747979 diff --git a/results_tencent/w2v2_base/layer10/run0/model_best.pt b/results_tencent/w2v2_base/layer10/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ece2b25b8081fbf70a455329792bd960b4a4ec76 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d0425077eba79ff76aedc3076e0bf4cc63f2e792015e0259378c0d09127b12a5 +size 2749638 diff --git a/results_tencent/w2v2_base/layer10/run0/train.log b/results_tencent/w2v2_base/layer10/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..eef4f78537fb3c1f83f6211ade34ea84b1179797 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5014021155834198 +INFO:root: +[Valid][0][UTT][ MSE = 0.2814 | LCC = 0.9089 | SRCC = 0.9108 ] +INFO:root:Average loss=0.23497523510456086 +INFO:root: +[Valid][1][UTT][ MSE = 0.2952 | LCC = 0.9332 | SRCC = 0.9343 ] +INFO:root:Average loss=0.19476969850063325 +INFO:root: +[Valid][2][UTT][ MSE = 0.3028 | LCC = 0.9322 | SRCC = 0.9337 ] +INFO:root:Average loss=0.17707715559005738 +INFO:root: +[Valid][3][UTT][ MSE = 0.2412 | LCC = 0.9372 | SRCC = 0.9399 ] +INFO:root:Average loss=0.16550643891096115 +INFO:root: +[Valid][4][UTT][ MSE = 0.2010 | LCC = 0.9417 | SRCC = 0.9446 ] +INFO:root:Average loss=0.15192551070451737 +INFO:root: +[Valid][5][UTT][ MSE = 0.2803 | LCC = 0.9412 | SRCC = 0.9431 ] +INFO:root:Average loss=0.13927937626838685 +INFO:root: +[Valid][6][UTT][ MSE = 0.2259 | LCC = 0.9486 | SRCC = 0.9501 ] +INFO:root:Average loss=0.13411734068393708 +INFO:root: +[Valid][7][UTT][ MSE = 0.2010 | LCC = 0.9503 | SRCC = 0.9517 ] +INFO:root:Average loss=0.12675749254226684 +INFO:root: +[Valid][8][UTT][ MSE = 0.2649 | LCC = 0.9387 | SRCC = 0.9402 ] +INFO:root:Average loss=0.12143711763620377 +INFO:root: +[Valid][9][UTT][ MSE = 0.2906 | LCC = 0.9395 | SRCC = 0.9401 ] +INFO:root:Average loss=0.11363052973151207 +INFO:root: +[Valid][10][UTT][ MSE = 0.2271 | LCC = 0.9515 | SRCC = 0.9532 ] +INFO:root:Average loss=0.10836589297652245 +INFO:root: +[Valid][11][UTT][ MSE = 0.1969 | LCC = 0.9526 | SRCC = 0.9535 ] +INFO:root:Average loss=0.10760377436876296 +INFO:root: +[Valid][12][UTT][ MSE = 0.2092 | LCC = 0.9507 | SRCC = 0.9522 ] +INFO:root:Average loss=0.09664743265509605 +INFO:root: +[Valid][13][UTT][ MSE = 0.1998 | LCC = 0.9528 | SRCC = 0.9537 ] +INFO:root:Average loss=0.0932683499455452 +INFO:root: +[Valid][14][UTT][ MSE = 0.1895 | LCC = 0.9539 | SRCC = 0.9547 ] +INFO:root:Average loss=0.09422821041941642 +INFO:root: +[Valid][15][UTT][ MSE = 0.2353 | LCC = 0.9551 | SRCC = 0.9556 ] +INFO:root:Average loss=0.09149050816893578 +INFO:root: +[Valid][16][UTT][ MSE = 0.1953 | LCC = 0.9548 | SRCC = 0.9561 ] +INFO:root:Average loss=0.08612626639008522 +INFO:root: +[Valid][17][UTT][ MSE = 0.2042 | LCC = 0.9534 | SRCC = 0.9544 ] +INFO:root:Average loss=0.08552507150173187 +INFO:root: +[Valid][18][UTT][ MSE = 0.2113 | LCC = 0.9544 | SRCC = 0.9551 ] +INFO:root:Average loss=0.07905705881118774 +INFO:root: +[Valid][19][UTT][ MSE = 0.2151 | LCC = 0.9528 | SRCC = 0.9538 ] +INFO:root:Average loss=0.07765994456410408 +INFO:root: +[Valid][20][UTT][ MSE = 0.2389 | LCC = 0.9549 | SRCC = 0.9547 ] +INFO:root:Average loss=0.07329837024211884 +INFO:root: +[Valid][21][UTT][ MSE = 0.2039 | LCC = 0.9552 | SRCC = 0.9548 ] +INFO:root:Average loss=0.07235442161560059 +INFO:root: +[Valid][22][UTT][ MSE = 0.2140 | LCC = 0.9547 | SRCC = 0.9546 ] +INFO:root:Average loss=0.06871049961447716 +INFO:root: +[Valid][23][UTT][ MSE = 0.2144 | LCC = 0.9521 | SRCC = 0.9518 ] +INFO:root:Average loss=0.06815191397070884 +INFO:root: +[Valid][24][UTT][ MSE = 0.1948 | LCC = 0.9549 | SRCC = 0.9556 ] +INFO:root:Average loss=0.06692243510484695 +INFO:root: +[Valid][25][UTT][ MSE = 0.1893 | LCC = 0.9527 | SRCC = 0.9541 ] +INFO:root:Average loss=0.06423515036702156 +INFO:root: +[Valid][26][UTT][ MSE = 0.1875 | LCC = 0.9558 | SRCC = 0.9553 ] +INFO:root:Average loss=0.06291434067487717 +INFO:root: +[Valid][27][UTT][ MSE = 0.1881 | LCC = 0.9537 | SRCC = 0.9544 ] +INFO:root:Average loss=0.05889962214231491 +INFO:root: +[Valid][28][UTT][ MSE = 0.2347 | LCC = 0.9495 | SRCC = 0.9516 ] +INFO:root:Average loss=0.05749878364801407 +INFO:root: +[Valid][29][UTT][ MSE = 0.2272 | LCC = 0.9554 | SRCC = 0.9550 ] +INFO:root:Average loss=0.056704952150583264 +INFO:root: +[Valid][30][UTT][ MSE = 0.2277 | LCC = 0.9539 | SRCC = 0.9539 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1922 | LCC = 0.9562 | SRCC = 0.9572 ] diff --git a/results_tencent/w2v2_base/layer10/run1/config.gin b/results_tencent/w2v2_base/layer10/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer10/run1/model.pt b/results_tencent/w2v2_base/layer10/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3bb42eff63eb64031ce7ad995fc210a0f146cace --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f51350ff233f8bd72325b90eb5a2723f492de96ac7ef1b1cf17202a066bc8546 +size 2747979 diff --git a/results_tencent/w2v2_base/layer10/run1/model_best.pt b/results_tencent/w2v2_base/layer10/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..da342149814266e3b7b37aa70a51c0c1d4d3c9bc --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43b9ecaf7f1a2faa519ad260ae831a0b59bec116ebe55cd67e791b6d43302e65 +size 2749702 diff --git a/results_tencent/w2v2_base/layer10/run1/train.log b/results_tencent/w2v2_base/layer10/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0a2a6d07f2531dea23839b762ce1d0732d8450a5 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6833050657510757 +INFO:root: +[Valid][0][UTT][ MSE = 0.3379 | LCC = 0.8955 | SRCC = 0.8959 ] +INFO:root:Average loss=0.2358826175928116 +INFO:root: +[Valid][1][UTT][ MSE = 0.3249 | LCC = 0.9277 | SRCC = 0.9301 ] +INFO:root:Average loss=0.19705918312072754 +INFO:root: +[Valid][2][UTT][ MSE = 0.2471 | LCC = 0.9336 | SRCC = 0.9355 ] +INFO:root:Average loss=0.17107334196567536 +INFO:root: +[Valid][3][UTT][ MSE = 0.2433 | LCC = 0.9392 | SRCC = 0.9394 ] +INFO:root:Average loss=0.15237635657191276 +INFO:root: +[Valid][4][UTT][ MSE = 0.2521 | LCC = 0.9451 | SRCC = 0.9464 ] +INFO:root:Average loss=0.14148516488075255 +INFO:root: +[Valid][5][UTT][ MSE = 0.2310 | LCC = 0.9467 | SRCC = 0.9481 ] +INFO:root:Average loss=0.13179047554731368 +INFO:root: +[Valid][6][UTT][ MSE = 0.2360 | LCC = 0.9484 | SRCC = 0.9499 ] +INFO:root:Average loss=0.126170602619648 +INFO:root: +[Valid][7][UTT][ MSE = 0.2365 | LCC = 0.9477 | SRCC = 0.9487 ] +INFO:root:Average loss=0.12095378297567368 +INFO:root: +[Valid][8][UTT][ MSE = 0.2251 | LCC = 0.9505 | SRCC = 0.9525 ] +INFO:root:Average loss=0.11898388028144836 +INFO:root: +[Valid][9][UTT][ MSE = 0.2545 | LCC = 0.9505 | SRCC = 0.9531 ] +INFO:root:Average loss=0.10983263796567917 +INFO:root: +[Valid][10][UTT][ MSE = 0.2351 | LCC = 0.9450 | SRCC = 0.9469 ] +INFO:root:Average loss=0.10531623673439026 +INFO:root: +[Valid][11][UTT][ MSE = 0.2131 | LCC = 0.9521 | SRCC = 0.9530 ] +INFO:root:Average loss=0.10241852903366089 +INFO:root: +[Valid][12][UTT][ MSE = 0.2324 | LCC = 0.9502 | SRCC = 0.9519 ] +INFO:root:Average loss=0.09717350262403489 +INFO:root: +[Valid][13][UTT][ MSE = 0.2341 | LCC = 0.9517 | SRCC = 0.9530 ] +INFO:root:Average loss=0.09212541884183884 +INFO:root: +[Valid][14][UTT][ MSE = 0.2385 | LCC = 0.9509 | SRCC = 0.9527 ] +INFO:root:Average loss=0.09008664178848266 +INFO:root: +[Valid][15][UTT][ MSE = 0.2153 | LCC = 0.9541 | SRCC = 0.9550 ] +INFO:root:Average loss=0.08516580140590668 +INFO:root: +[Valid][16][UTT][ MSE = 0.2157 | LCC = 0.9476 | SRCC = 0.9480 ] +INFO:root:Average loss=0.08267290022969245 +INFO:root: +[Valid][17][UTT][ MSE = 0.2059 | LCC = 0.9528 | SRCC = 0.9538 ] +INFO:root:Average loss=0.07928339612483978 +INFO:root: +[Valid][18][UTT][ MSE = 0.2127 | LCC = 0.9498 | SRCC = 0.9510 ] +INFO:root:Average loss=0.07573144277930259 +INFO:root: +[Valid][19][UTT][ MSE = 0.2044 | LCC = 0.9523 | SRCC = 0.9530 ] +INFO:root:Average loss=0.07662017554044724 +INFO:root: +[Valid][20][UTT][ MSE = 0.2404 | LCC = 0.9535 | SRCC = 0.9553 ] +INFO:root:Average loss=0.06979743993282318 +INFO:root: +[Valid][21][UTT][ MSE = 0.2021 | LCC = 0.9529 | SRCC = 0.9547 ] +INFO:root:Average loss=0.07315372423827649 +INFO:root: +[Valid][22][UTT][ MSE = 0.1914 | LCC = 0.9488 | SRCC = 0.9504 ] +INFO:root:Average loss=0.06921289706230163 +INFO:root: +[Valid][23][UTT][ MSE = 0.1996 | LCC = 0.9520 | SRCC = 0.9533 ] +INFO:root:Average loss=0.06599494653940201 +INFO:root: +[Valid][24][UTT][ MSE = 0.2040 | LCC = 0.9538 | SRCC = 0.9547 ] +INFO:root:Average loss=0.06185012874007225 +INFO:root: +[Valid][25][UTT][ MSE = 0.2085 | LCC = 0.9535 | SRCC = 0.9545 ] +INFO:root:Average loss=0.060814731985330585 +INFO:root: +[Valid][26][UTT][ MSE = 0.2211 | LCC = 0.9536 | SRCC = 0.9551 ] +INFO:root:Average loss=0.0591035515666008 +INFO:root: +[Valid][27][UTT][ MSE = 0.1903 | LCC = 0.9537 | SRCC = 0.9540 ] +INFO:root:Average loss=0.05818577072024345 +INFO:root: +[Valid][28][UTT][ MSE = 0.2360 | LCC = 0.9531 | SRCC = 0.9553 ] +INFO:root:Average loss=0.05599719472229481 +INFO:root: +[Valid][29][UTT][ MSE = 0.2493 | LCC = 0.9522 | SRCC = 0.9541 ] +INFO:root:Average loss=0.05238624107837677 +INFO:root: +[Valid][30][UTT][ MSE = 0.2250 | LCC = 0.9536 | SRCC = 0.9550 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2314 | LCC = 0.9535 | SRCC = 0.9548 ] diff --git a/results_tencent/w2v2_base/layer10/run2/config.gin b/results_tencent/w2v2_base/layer10/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer10/run2/model.pt b/results_tencent/w2v2_base/layer10/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e52b00a999125f79050dfd873cfe3131928a3b30 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ef002ec25a0ae15b66455998e4f53209f3331a08f41088fc0a597dc964e18dd1 +size 2747979 diff --git a/results_tencent/w2v2_base/layer10/run2/model_best.pt b/results_tencent/w2v2_base/layer10/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9b675c8e09ba8663812bd9cdc93367ee91e42ec4 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:67ef86d477a3fe9b1a1fd8573872d9e5029cb4a397ea40b895dd13d7ca2cae8a +size 2749638 diff --git a/results_tencent/w2v2_base/layer10/run2/train.log b/results_tencent/w2v2_base/layer10/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9264f1aee1b7e01acb6b5409680f5cac5c8cf5e7 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6141754460334777 +INFO:root: +[Valid][0][UTT][ MSE = 0.3509 | LCC = 0.8983 | SRCC = 0.9004 ] +INFO:root:Average loss=0.23877217495441436 +INFO:root: +[Valid][1][UTT][ MSE = 0.2833 | LCC = 0.9232 | SRCC = 0.9237 ] +INFO:root:Average loss=0.19058691656589508 +INFO:root: +[Valid][2][UTT][ MSE = 0.2563 | LCC = 0.9381 | SRCC = 0.9407 ] +INFO:root:Average loss=0.1683231657743454 +INFO:root: +[Valid][3][UTT][ MSE = 0.2267 | LCC = 0.9402 | SRCC = 0.9415 ] +INFO:root:Average loss=0.14944227021932602 +INFO:root: +[Valid][4][UTT][ MSE = 0.2144 | LCC = 0.9439 | SRCC = 0.9460 ] +INFO:root:Average loss=0.14393416792154312 +INFO:root: +[Valid][5][UTT][ MSE = 0.2377 | LCC = 0.9461 | SRCC = 0.9483 ] +INFO:root:Average loss=0.1355625301003456 +INFO:root: +[Valid][6][UTT][ MSE = 0.2730 | LCC = 0.9462 | SRCC = 0.9490 ] +INFO:root:Average loss=0.1315132052898407 +INFO:root: +[Valid][7][UTT][ MSE = 0.2425 | LCC = 0.9490 | SRCC = 0.9507 ] +INFO:root:Average loss=0.13047779792547226 +INFO:root: +[Valid][8][UTT][ MSE = 0.2240 | LCC = 0.9481 | SRCC = 0.9499 ] +INFO:root:Average loss=0.11236211109161377 +INFO:root: +[Valid][9][UTT][ MSE = 0.2146 | LCC = 0.9508 | SRCC = 0.9523 ] +INFO:root:Average loss=0.11110353654623031 +INFO:root: +[Valid][10][UTT][ MSE = 0.2414 | LCC = 0.9515 | SRCC = 0.9535 ] +INFO:root:Average loss=0.10392388978600502 +INFO:root: +[Valid][11][UTT][ MSE = 0.2206 | LCC = 0.9514 | SRCC = 0.9530 ] +INFO:root:Average loss=0.09884071925282478 +INFO:root: +[Valid][12][UTT][ MSE = 0.1992 | LCC = 0.9496 | SRCC = 0.9517 ] +INFO:root:Average loss=0.09857537007331849 +INFO:root: +[Valid][13][UTT][ MSE = 0.2183 | LCC = 0.9524 | SRCC = 0.9537 ] +INFO:root:Average loss=0.09544353744387626 +INFO:root: +[Valid][14][UTT][ MSE = 0.2178 | LCC = 0.9531 | SRCC = 0.9541 ] +INFO:root:Average loss=0.0916655415892601 +INFO:root: +[Valid][15][UTT][ MSE = 0.2134 | LCC = 0.9530 | SRCC = 0.9540 ] +INFO:root:Average loss=0.089254007011652 +INFO:root: +[Valid][16][UTT][ MSE = 0.2151 | LCC = 0.9525 | SRCC = 0.9532 ] +INFO:root:Average loss=0.08481873297691345 +INFO:root: +[Valid][17][UTT][ MSE = 0.2230 | LCC = 0.9534 | SRCC = 0.9544 ] +INFO:root:Average loss=0.08673698452115058 +INFO:root: +[Valid][18][UTT][ MSE = 0.1958 | LCC = 0.9533 | SRCC = 0.9538 ] +INFO:root:Average loss=0.07711614578962327 +INFO:root: +[Valid][19][UTT][ MSE = 0.2149 | LCC = 0.9507 | SRCC = 0.9521 ] +INFO:root:Average loss=0.07706710079312325 +INFO:root: +[Valid][20][UTT][ MSE = 0.2117 | LCC = 0.9540 | SRCC = 0.9542 ] +INFO:root:Average loss=0.07350507205724716 +INFO:root: +[Valid][21][UTT][ MSE = 0.1907 | LCC = 0.9549 | SRCC = 0.9557 ] +INFO:root:Average loss=0.06959401127696037 +INFO:root: +[Valid][22][UTT][ MSE = 0.1901 | LCC = 0.9530 | SRCC = 0.9537 ] +INFO:root:Average loss=0.07059612539410591 +INFO:root: +[Valid][23][UTT][ MSE = 0.2137 | LCC = 0.9541 | SRCC = 0.9548 ] +INFO:root:Average loss=0.06548276641964912 +INFO:root: +[Valid][24][UTT][ MSE = 0.1933 | LCC = 0.9534 | SRCC = 0.9549 ] +INFO:root:Average loss=0.062284262865781787 +INFO:root: +[Valid][25][UTT][ MSE = 0.2054 | LCC = 0.9541 | SRCC = 0.9547 ] +INFO:root:Average loss=0.06255162945389747 +INFO:root: +[Valid][26][UTT][ MSE = 0.2003 | LCC = 0.9531 | SRCC = 0.9534 ] +INFO:root:Average loss=0.06105635243654251 +INFO:root: +[Valid][27][UTT][ MSE = 0.1795 | LCC = 0.9529 | SRCC = 0.9540 ] +INFO:root:Average loss=0.056437740162014964 +INFO:root: +[Valid][28][UTT][ MSE = 0.2013 | LCC = 0.9531 | SRCC = 0.9539 ] +INFO:root:Average loss=0.05399019080400467 +INFO:root: +[Valid][29][UTT][ MSE = 0.2069 | LCC = 0.9540 | SRCC = 0.9540 ] +INFO:root:Average loss=0.05466146217286587 +INFO:root: +[Valid][30][UTT][ MSE = 0.1928 | LCC = 0.9517 | SRCC = 0.9524 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1990 | LCC = 0.9551 | SRCC = 0.9558 ] diff --git a/results_tencent/w2v2_base/layer10/run3/config.gin b/results_tencent/w2v2_base/layer10/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer10/run3/model.pt b/results_tencent/w2v2_base/layer10/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9ad38d2808bf15f01e789cefa1d8ea341a8a4603 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c6b755e5be3c363d13c1ddfa451cb133643e36b00282365d0e871171bef9f462 +size 2747979 diff --git a/results_tencent/w2v2_base/layer10/run3/model_best.pt b/results_tencent/w2v2_base/layer10/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..fd69ee1c9c92e39d8f60698a7a278f774c17562b --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:87819d1ecfe82c655b3674cad0ac31d04f9ce65f038275512304db031b43a7c8 +size 2749702 diff --git a/results_tencent/w2v2_base/layer10/run3/train.log b/results_tencent/w2v2_base/layer10/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1f45a73dd8c5353e594f777952a58173068cf391 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4866420583724976 +INFO:root: +[Valid][0][UTT][ MSE = 0.2810 | LCC = 0.9121 | SRCC = 0.9145 ] +INFO:root:Average loss=0.2265995320081711 +INFO:root: +[Valid][1][UTT][ MSE = 0.2657 | LCC = 0.9302 | SRCC = 0.9313 ] +INFO:root:Average loss=0.18112735140323638 +INFO:root: +[Valid][2][UTT][ MSE = 0.2309 | LCC = 0.9401 | SRCC = 0.9413 ] +INFO:root:Average loss=0.16742794018983842 +INFO:root: +[Valid][3][UTT][ MSE = 0.2672 | LCC = 0.9319 | SRCC = 0.9357 ] +INFO:root:Average loss=0.15394514548778535 +INFO:root: +[Valid][4][UTT][ MSE = 0.2382 | LCC = 0.9423 | SRCC = 0.9441 ] +INFO:root:Average loss=0.14199903404712677 +INFO:root: +[Valid][5][UTT][ MSE = 0.2886 | LCC = 0.9469 | SRCC = 0.9477 ] +INFO:root:Average loss=0.13540122187137604 +INFO:root: +[Valid][6][UTT][ MSE = 0.2107 | LCC = 0.9476 | SRCC = 0.9482 ] +INFO:root:Average loss=0.1262679055929184 +INFO:root: +[Valid][7][UTT][ MSE = 0.1986 | LCC = 0.9484 | SRCC = 0.9507 ] +INFO:root:Average loss=0.11697091776132583 +INFO:root: +[Valid][8][UTT][ MSE = 0.2218 | LCC = 0.9513 | SRCC = 0.9516 ] +INFO:root:Average loss=0.1136043501496315 +INFO:root: +[Valid][9][UTT][ MSE = 0.2109 | LCC = 0.9485 | SRCC = 0.9513 ] +INFO:root:Average loss=0.11225579133629798 +INFO:root: +[Valid][10][UTT][ MSE = 0.1814 | LCC = 0.9519 | SRCC = 0.9535 ] +INFO:root:Average loss=0.10477510952949524 +INFO:root: +[Valid][11][UTT][ MSE = 0.2123 | LCC = 0.9534 | SRCC = 0.9543 ] +INFO:root:Average loss=0.10044978937506675 +INFO:root: +[Valid][12][UTT][ MSE = 0.2095 | LCC = 0.9506 | SRCC = 0.9530 ] +INFO:root:Average loss=0.10024743261933326 +INFO:root: +[Valid][13][UTT][ MSE = 0.2288 | LCC = 0.9533 | SRCC = 0.9541 ] +INFO:root:Average loss=0.09537120351195336 +INFO:root: +[Valid][14][UTT][ MSE = 0.1941 | LCC = 0.9542 | SRCC = 0.9542 ] +INFO:root:Average loss=0.08851299872994423 +INFO:root: +[Valid][15][UTT][ MSE = 0.2863 | LCC = 0.9521 | SRCC = 0.9529 ] +INFO:root:Average loss=0.08827878218889236 +INFO:root: +[Valid][16][UTT][ MSE = 0.2005 | LCC = 0.9533 | SRCC = 0.9537 ] +INFO:root:Average loss=0.08065296775102615 +INFO:root: +[Valid][17][UTT][ MSE = 0.1913 | LCC = 0.9545 | SRCC = 0.9541 ] +INFO:root:Average loss=0.0819734598994255 +INFO:root: +[Valid][18][UTT][ MSE = 0.1836 | LCC = 0.9563 | SRCC = 0.9563 ] +INFO:root:Average loss=0.0752915113568306 +INFO:root: +[Valid][19][UTT][ MSE = 0.1884 | LCC = 0.9549 | SRCC = 0.9558 ] +INFO:root:Average loss=0.07466227576136589 +INFO:root: +[Valid][20][UTT][ MSE = 0.2384 | LCC = 0.9547 | SRCC = 0.9550 ] +INFO:root:Average loss=0.07011279553174972 +INFO:root: +[Valid][21][UTT][ MSE = 0.1988 | LCC = 0.9551 | SRCC = 0.9554 ] +INFO:root:Average loss=0.0698498151898384 +INFO:root: +[Valid][22][UTT][ MSE = 0.2263 | LCC = 0.9547 | SRCC = 0.9553 ] +INFO:root:Average loss=0.06716396582126617 +INFO:root: +[Valid][23][UTT][ MSE = 0.1760 | LCC = 0.9550 | SRCC = 0.9553 ] +INFO:root:Average loss=0.06598066526651382 +INFO:root: +[Valid][24][UTT][ MSE = 0.2392 | LCC = 0.9542 | SRCC = 0.9553 ] +INFO:root:Average loss=0.06283867058157921 +INFO:root: +[Valid][25][UTT][ MSE = 0.2161 | LCC = 0.9538 | SRCC = 0.9538 ] +INFO:root:Average loss=0.06240953281521797 +INFO:root: +[Valid][26][UTT][ MSE = 0.2174 | LCC = 0.9541 | SRCC = 0.9535 ] +INFO:root:Average loss=0.06053435355424881 +INFO:root: +[Valid][27][UTT][ MSE = 0.2422 | LCC = 0.9511 | SRCC = 0.9508 ] +INFO:root:Average loss=0.0589320752620697 +INFO:root: +[Valid][28][UTT][ MSE = 0.1960 | LCC = 0.9558 | SRCC = 0.9557 ] +INFO:root:Average loss=0.058986039936542514 +INFO:root: +[Valid][29][UTT][ MSE = 0.1895 | LCC = 0.9552 | SRCC = 0.9555 ] +INFO:root:Average loss=0.054865549385547635 +INFO:root: +[Valid][30][UTT][ MSE = 0.1789 | LCC = 0.9548 | SRCC = 0.9551 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1952 | LCC = 0.9541 | SRCC = 0.9547 ] diff --git a/results_tencent/w2v2_base/layer10/run4/config.gin b/results_tencent/w2v2_base/layer10/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer10/run4/model.pt b/results_tencent/w2v2_base/layer10/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0fe90eb3a61a274c0e72f598753496b7ace81f44 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:540b832c708bbb790a5dc44e6655ae3537c80bf4b89485c4eed2fcf029b88753 +size 2747979 diff --git a/results_tencent/w2v2_base/layer10/run4/model_best.pt b/results_tencent/w2v2_base/layer10/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d38e5e53c8e20e882246d632ce60f82a9566f4a9 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:552078e318c4f9c2c4ed0da587141ce2eb17ae496a2b93baf8bfed759331949b +size 2749638 diff --git a/results_tencent/w2v2_base/layer10/run4/train.log b/results_tencent/w2v2_base/layer10/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..19f28fb74d46c91c11bb51af0708e4fd6ed505b3 --- /dev/null +++ b/results_tencent/w2v2_base/layer10/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5576680352687836 +INFO:root: +[Valid][0][UTT][ MSE = 0.3201 | LCC = 0.9143 | SRCC = 0.9153 ] +INFO:root:Average loss=0.23916537249088288 +INFO:root: +[Valid][1][UTT][ MSE = 0.2977 | LCC = 0.9238 | SRCC = 0.9262 ] +INFO:root:Average loss=0.1917428902387619 +INFO:root: +[Valid][2][UTT][ MSE = 0.2429 | LCC = 0.9383 | SRCC = 0.9397 ] +INFO:root:Average loss=0.16857908487319948 +INFO:root: +[Valid][3][UTT][ MSE = 0.2886 | LCC = 0.9366 | SRCC = 0.9392 ] +INFO:root:Average loss=0.15183681178092956 +INFO:root: +[Valid][4][UTT][ MSE = 0.2550 | LCC = 0.9437 | SRCC = 0.9459 ] +INFO:root:Average loss=0.15230360573530197 +INFO:root: +[Valid][5][UTT][ MSE = 0.1974 | LCC = 0.9458 | SRCC = 0.9486 ] +INFO:root:Average loss=0.13751437532901764 +INFO:root: +[Valid][6][UTT][ MSE = 0.2142 | LCC = 0.9455 | SRCC = 0.9469 ] +INFO:root:Average loss=0.12803515940904617 +INFO:root: +[Valid][7][UTT][ MSE = 0.2349 | LCC = 0.9494 | SRCC = 0.9501 ] +INFO:root:Average loss=0.1256012635231018 +INFO:root: +[Valid][8][UTT][ MSE = 0.2279 | LCC = 0.9506 | SRCC = 0.9520 ] +INFO:root:Average loss=0.1130274132490158 +INFO:root: +[Valid][9][UTT][ MSE = 0.2303 | LCC = 0.9506 | SRCC = 0.9521 ] +INFO:root:Average loss=0.1099078910946846 +INFO:root: +[Valid][10][UTT][ MSE = 0.1919 | LCC = 0.9515 | SRCC = 0.9531 ] +INFO:root:Average loss=0.1054424991607666 +INFO:root: +[Valid][11][UTT][ MSE = 0.2277 | LCC = 0.9394 | SRCC = 0.9426 ] +INFO:root:Average loss=0.10171599543094635 +INFO:root: +[Valid][12][UTT][ MSE = 0.1738 | LCC = 0.9527 | SRCC = 0.9537 ] +INFO:root:Average loss=0.09464134398102761 +INFO:root: +[Valid][13][UTT][ MSE = 0.2372 | LCC = 0.9526 | SRCC = 0.9545 ] +INFO:root:Average loss=0.09795195564627647 +INFO:root: +[Valid][14][UTT][ MSE = 0.1826 | LCC = 0.9524 | SRCC = 0.9536 ] +INFO:root:Average loss=0.0855869548022747 +INFO:root: +[Valid][15][UTT][ MSE = 0.2106 | LCC = 0.9506 | SRCC = 0.9524 ] +INFO:root:Average loss=0.08752905580401421 +INFO:root: +[Valid][16][UTT][ MSE = 0.2201 | LCC = 0.9475 | SRCC = 0.9490 ] +INFO:root:Average loss=0.08359409776329994 +INFO:root: +[Valid][17][UTT][ MSE = 0.1877 | LCC = 0.9513 | SRCC = 0.9530 ] +INFO:root:Average loss=0.07925798878073692 +INFO:root: +[Valid][18][UTT][ MSE = 0.1773 | LCC = 0.9528 | SRCC = 0.9526 ] +INFO:root:Average loss=0.08113798460364342 +INFO:root: +[Valid][19][UTT][ MSE = 0.1603 | LCC = 0.9526 | SRCC = 0.9543 ] +INFO:root:Average loss=0.08158670210838317 +INFO:root: +[Valid][20][UTT][ MSE = 0.1643 | LCC = 0.9542 | SRCC = 0.9549 ] +INFO:root:Average loss=0.07376670613884925 +INFO:root: +[Valid][21][UTT][ MSE = 0.1834 | LCC = 0.9535 | SRCC = 0.9542 ] +INFO:root:Average loss=0.06941380709409714 +INFO:root: +[Valid][22][UTT][ MSE = 0.2265 | LCC = 0.9469 | SRCC = 0.9498 ] +INFO:root:Average loss=0.07094217690825462 +INFO:root: +[Valid][23][UTT][ MSE = 0.1771 | LCC = 0.9545 | SRCC = 0.9545 ] +INFO:root:Average loss=0.06678099909424781 +INFO:root: +[Valid][24][UTT][ MSE = 0.1997 | LCC = 0.9507 | SRCC = 0.9509 ] +INFO:root:Average loss=0.06542498219013214 +INFO:root: +[Valid][25][UTT][ MSE = 0.1942 | LCC = 0.9526 | SRCC = 0.9534 ] +INFO:root:Average loss=0.06174080008268356 +INFO:root: +[Valid][26][UTT][ MSE = 0.1757 | LCC = 0.9504 | SRCC = 0.9513 ] +INFO:root:Average loss=0.057315763920545576 +INFO:root: +[Valid][27][UTT][ MSE = 0.1679 | LCC = 0.9547 | SRCC = 0.9554 ] +INFO:root:Average loss=0.060240774005651476 +INFO:root: +[Valid][28][UTT][ MSE = 0.2007 | LCC = 0.9540 | SRCC = 0.9546 ] +INFO:root:Average loss=0.059225142389535905 +INFO:root: +[Valid][29][UTT][ MSE = 0.1757 | LCC = 0.9544 | SRCC = 0.9548 ] +INFO:root:Average loss=0.05325778983533382 +INFO:root: +[Valid][30][UTT][ MSE = 0.1804 | LCC = 0.9549 | SRCC = 0.9549 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1804 | LCC = 0.9565 | SRCC = 0.9570 ] diff --git a/results_tencent/w2v2_base/layer11/run0/config.gin b/results_tencent/w2v2_base/layer11/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer11/run0/model.pt b/results_tencent/w2v2_base/layer11/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9e4f0e846edb681aa5466b9bf8a496b0e91c4ccb --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:50d2fd0f9ccbb14b901acb9c798e2d8e717c2800a9b24ef27e18dd65a066cde0 +size 2747979 diff --git a/results_tencent/w2v2_base/layer11/run0/model_best.pt b/results_tencent/w2v2_base/layer11/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..186ad0f850e90fde739b47e50ff8d0ad31cdd47a --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6e26565c6c51499fd661e3b5c88f3db94bfc44aee0bb632e5c8767baeed9ce82 +size 2749638 diff --git a/results_tencent/w2v2_base/layer11/run0/train.log b/results_tencent/w2v2_base/layer11/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e2ad2f603b5bded15e1b96df29ae0d0c4a107f2d --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.49719896042346956 +INFO:root: +[Valid][0][UTT][ MSE = 0.3307 | LCC = 0.9058 | SRCC = 0.9089 ] +INFO:root:Average loss=0.26075546753406526 +INFO:root: +[Valid][1][UTT][ MSE = 0.3223 | LCC = 0.9172 | SRCC = 0.9199 ] +INFO:root:Average loss=0.21068650543689726 +INFO:root: +[Valid][2][UTT][ MSE = 0.2394 | LCC = 0.9274 | SRCC = 0.9287 ] +INFO:root:Average loss=0.19064053159952163 +INFO:root: +[Valid][3][UTT][ MSE = 0.2634 | LCC = 0.9336 | SRCC = 0.9357 ] +INFO:root:Average loss=0.17178169274330138 +INFO:root: +[Valid][4][UTT][ MSE = 0.2865 | LCC = 0.9349 | SRCC = 0.9366 ] +INFO:root:Average loss=0.16664757269620895 +INFO:root: +[Valid][5][UTT][ MSE = 0.2732 | LCC = 0.9343 | SRCC = 0.9375 ] +INFO:root:Average loss=0.1569429606795311 +INFO:root: +[Valid][6][UTT][ MSE = 0.2368 | LCC = 0.9410 | SRCC = 0.9423 ] +INFO:root:Average loss=0.1489917061626911 +INFO:root: +[Valid][7][UTT][ MSE = 0.2308 | LCC = 0.9404 | SRCC = 0.9417 ] +INFO:root:Average loss=0.13810816860198974 +INFO:root: +[Valid][8][UTT][ MSE = 0.2557 | LCC = 0.9416 | SRCC = 0.9439 ] +INFO:root:Average loss=0.12948104637861252 +INFO:root: +[Valid][9][UTT][ MSE = 0.2096 | LCC = 0.9439 | SRCC = 0.9443 ] +INFO:root:Average loss=0.12598332929611206 +INFO:root: +[Valid][10][UTT][ MSE = 0.2251 | LCC = 0.9432 | SRCC = 0.9442 ] +INFO:root:Average loss=0.1189342452287674 +INFO:root: +[Valid][11][UTT][ MSE = 0.2515 | LCC = 0.9417 | SRCC = 0.9443 ] +INFO:root:Average loss=0.1185251974761486 +INFO:root: +[Valid][12][UTT][ MSE = 0.2303 | LCC = 0.9436 | SRCC = 0.9446 ] +INFO:root:Average loss=0.11081412455439568 +INFO:root: +[Valid][13][UTT][ MSE = 0.2047 | LCC = 0.9452 | SRCC = 0.9460 ] +INFO:root:Average loss=0.10631788021326065 +INFO:root: +[Valid][14][UTT][ MSE = 0.2233 | LCC = 0.9469 | SRCC = 0.9480 ] +INFO:root:Average loss=0.10689106065034866 +INFO:root: +[Valid][15][UTT][ MSE = 0.2188 | LCC = 0.9450 | SRCC = 0.9462 ] +INFO:root:Average loss=0.10398036295175553 +INFO:root: +[Valid][16][UTT][ MSE = 0.1958 | LCC = 0.9461 | SRCC = 0.9475 ] +INFO:root:Average loss=0.1024542707502842 +INFO:root: +[Valid][17][UTT][ MSE = 0.2178 | LCC = 0.9451 | SRCC = 0.9467 ] +INFO:root:Average loss=0.09336707809567452 +INFO:root: +[Valid][18][UTT][ MSE = 0.1926 | LCC = 0.9464 | SRCC = 0.9468 ] +INFO:root:Average loss=0.09253980126976967 +INFO:root: +[Valid][19][UTT][ MSE = 0.2289 | LCC = 0.9459 | SRCC = 0.9477 ] +INFO:root:Average loss=0.08933713737130165 +INFO:root: +[Valid][20][UTT][ MSE = 0.2357 | LCC = 0.9437 | SRCC = 0.9451 ] +INFO:root:Average loss=0.08568824476003647 +INFO:root: +[Valid][21][UTT][ MSE = 0.2103 | LCC = 0.9486 | SRCC = 0.9496 ] +INFO:root:Average loss=0.08413927388191224 +INFO:root: +[Valid][22][UTT][ MSE = 0.2417 | LCC = 0.9441 | SRCC = 0.9454 ] +INFO:root:Average loss=0.08195547980070114 +INFO:root: +[Valid][23][UTT][ MSE = 0.2249 | LCC = 0.9479 | SRCC = 0.9494 ] +INFO:root:Average loss=0.07950736719369889 +INFO:root: +[Valid][24][UTT][ MSE = 0.2198 | LCC = 0.9481 | SRCC = 0.9494 ] +INFO:root:Average loss=0.07958603677153588 +INFO:root: +[Valid][25][UTT][ MSE = 0.2021 | LCC = 0.9475 | SRCC = 0.9479 ] +INFO:root:Average loss=0.0764790431857109 +INFO:root: +[Valid][26][UTT][ MSE = 0.2067 | LCC = 0.9485 | SRCC = 0.9495 ] +INFO:root:Average loss=0.07549632453918458 +INFO:root: +[Valid][27][UTT][ MSE = 0.2229 | LCC = 0.9475 | SRCC = 0.9487 ] +INFO:root:Average loss=0.07541598144173622 +INFO:root: +[Valid][28][UTT][ MSE = 0.2324 | LCC = 0.9481 | SRCC = 0.9489 ] +INFO:root:Average loss=0.0725514925122261 +INFO:root: +[Valid][29][UTT][ MSE = 0.1838 | LCC = 0.9481 | SRCC = 0.9490 ] +INFO:root:Average loss=0.06761295741796494 +INFO:root: +[Valid][30][UTT][ MSE = 0.2151 | LCC = 0.9472 | SRCC = 0.9480 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2113 | LCC = 0.9516 | SRCC = 0.9526 ] diff --git a/results_tencent/w2v2_base/layer11/run1/config.gin b/results_tencent/w2v2_base/layer11/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer11/run1/model.pt b/results_tencent/w2v2_base/layer11/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f51aabb753ebf6de30072f8454f29c8231474807 --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:acf33bc8a794f63229af0f64ce1621c3167a6ae9ec5bfb04780054c5dd7e08de +size 2747979 diff --git a/results_tencent/w2v2_base/layer11/run1/model_best.pt b/results_tencent/w2v2_base/layer11/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2bef448a21f95070f886feb737330d91363c508a --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4f7c68ad2302d49d41eb379a7738fa1b3afe764d83c7de7514c8b4615c2179d +size 2749638 diff --git a/results_tencent/w2v2_base/layer11/run1/train.log b/results_tencent/w2v2_base/layer11/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0a0c10efd2d576b198d55fd7deeb137bc1520b8b --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5153049086332321 +INFO:root: +[Valid][0][UTT][ MSE = 0.3432 | LCC = 0.8999 | SRCC = 0.9020 ] +INFO:root:Average loss=0.2600029705166817 +INFO:root: +[Valid][1][UTT][ MSE = 0.2713 | LCC = 0.9207 | SRCC = 0.9233 ] +INFO:root:Average loss=0.21224652034044267 +INFO:root: +[Valid][2][UTT][ MSE = 0.4920 | LCC = 0.9167 | SRCC = 0.9169 ] +INFO:root:Average loss=0.1960317976474762 +INFO:root: +[Valid][3][UTT][ MSE = 0.2815 | LCC = 0.9332 | SRCC = 0.9352 ] +INFO:root:Average loss=0.17152852618694306 +INFO:root: +[Valid][4][UTT][ MSE = 0.2705 | LCC = 0.9364 | SRCC = 0.9372 ] +INFO:root:Average loss=0.15918887203931809 +INFO:root: +[Valid][5][UTT][ MSE = 0.2652 | LCC = 0.9369 | SRCC = 0.9380 ] +INFO:root:Average loss=0.150550035238266 +INFO:root: +[Valid][6][UTT][ MSE = 0.3146 | LCC = 0.9377 | SRCC = 0.9386 ] +INFO:root:Average loss=0.14639698910713195 +INFO:root: +[Valid][7][UTT][ MSE = 0.2669 | LCC = 0.9401 | SRCC = 0.9413 ] +INFO:root:Average loss=0.13843698209524155 +INFO:root: +[Valid][8][UTT][ MSE = 0.2451 | LCC = 0.9410 | SRCC = 0.9428 ] +INFO:root:Average loss=0.1297837014198303 +INFO:root: +[Valid][9][UTT][ MSE = 0.2173 | LCC = 0.9398 | SRCC = 0.9415 ] +INFO:root:Average loss=0.12590641820430756 +INFO:root: +[Valid][10][UTT][ MSE = 0.2386 | LCC = 0.9385 | SRCC = 0.9389 ] +INFO:root:Average loss=0.12396994981169701 +INFO:root: +[Valid][11][UTT][ MSE = 0.3286 | LCC = 0.9443 | SRCC = 0.9457 ] +INFO:root:Average loss=0.12278691399097443 +INFO:root: +[Valid][12][UTT][ MSE = 0.2457 | LCC = 0.9456 | SRCC = 0.9467 ] +INFO:root:Average loss=0.11017679554224015 +INFO:root: +[Valid][13][UTT][ MSE = 0.2843 | LCC = 0.9433 | SRCC = 0.9444 ] +INFO:root:Average loss=0.11282495740056038 +INFO:root: +[Valid][14][UTT][ MSE = 0.2244 | LCC = 0.9437 | SRCC = 0.9455 ] +INFO:root:Average loss=0.10435372266173362 +INFO:root: +[Valid][15][UTT][ MSE = 0.3187 | LCC = 0.9434 | SRCC = 0.9450 ] +INFO:root:Average loss=0.09892180892825127 +INFO:root: +[Valid][16][UTT][ MSE = 0.2559 | LCC = 0.9414 | SRCC = 0.9426 ] +INFO:root:Average loss=0.09569830083847046 +INFO:root: +[Valid][17][UTT][ MSE = 0.2451 | LCC = 0.9454 | SRCC = 0.9463 ] +INFO:root:Average loss=0.09723351562023162 +INFO:root: +[Valid][18][UTT][ MSE = 0.2441 | LCC = 0.9462 | SRCC = 0.9483 ] +INFO:root:Average loss=0.09001479974389076 +INFO:root: +[Valid][19][UTT][ MSE = 0.2659 | LCC = 0.9446 | SRCC = 0.9460 ] +INFO:root:Average loss=0.08853997665643693 +INFO:root: +[Valid][20][UTT][ MSE = 0.2483 | LCC = 0.9442 | SRCC = 0.9455 ] +INFO:root:Average loss=0.08636480781435966 +INFO:root: +[Valid][21][UTT][ MSE = 0.2406 | LCC = 0.9430 | SRCC = 0.9437 ] +INFO:root:Average loss=0.0853591431081295 +INFO:root: +[Valid][22][UTT][ MSE = 0.1848 | LCC = 0.9471 | SRCC = 0.9481 ] +INFO:root:Average loss=0.07886911505460739 +INFO:root: +[Valid][23][UTT][ MSE = 0.2057 | LCC = 0.9477 | SRCC = 0.9483 ] +INFO:root:Average loss=0.07711518025398255 +INFO:root: +[Valid][24][UTT][ MSE = 0.2393 | LCC = 0.9449 | SRCC = 0.9459 ] +INFO:root:Average loss=0.07359989982843398 +INFO:root: +[Valid][25][UTT][ MSE = 0.2116 | LCC = 0.9470 | SRCC = 0.9480 ] +INFO:root:Average loss=0.07674321922659874 +INFO:root: +[Valid][26][UTT][ MSE = 0.2405 | LCC = 0.9446 | SRCC = 0.9457 ] +INFO:root:Average loss=0.07353832542896271 +INFO:root: +[Valid][27][UTT][ MSE = 0.2210 | LCC = 0.9449 | SRCC = 0.9467 ] +INFO:root:Average loss=0.0705032505095005 +INFO:root: +[Valid][28][UTT][ MSE = 0.2061 | LCC = 0.9472 | SRCC = 0.9485 ] +INFO:root:Average loss=0.06755787909030914 +INFO:root: +[Valid][29][UTT][ MSE = 0.2419 | LCC = 0.9448 | SRCC = 0.9450 ] +INFO:root:Average loss=0.06437323698401451 +INFO:root: +[Valid][30][UTT][ MSE = 0.2194 | LCC = 0.9461 | SRCC = 0.9470 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2083 | LCC = 0.9494 | SRCC = 0.9501 ] diff --git a/results_tencent/w2v2_base/layer11/run2/config.gin b/results_tencent/w2v2_base/layer11/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer11/run2/model.pt b/results_tencent/w2v2_base/layer11/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9153963cb1d4a5762a689cfe469957e2b19c562f --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3fa630b7a659b9122a93cdde101c816bf5746822bd990a3f0f89921ed8d2c672 +size 2747979 diff --git a/results_tencent/w2v2_base/layer11/run2/model_best.pt b/results_tencent/w2v2_base/layer11/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..3c62b4c51bcc6cef34820f0d949f78f4faeec152 --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1c28b9989551d1b4d112243035170eb37ef021d481be5f0cd3671d283a8ab683 +size 2749638 diff --git a/results_tencent/w2v2_base/layer11/run2/train.log b/results_tencent/w2v2_base/layer11/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d6fa7185d742e86d134d749a8d65f9e82a4987c1 --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.49282935297489167 +INFO:root: +[Valid][0][UTT][ MSE = 0.3858 | LCC = 0.8978 | SRCC = 0.9002 ] +INFO:root:Average loss=0.23823538422584534 +INFO:root: +[Valid][1][UTT][ MSE = 0.2854 | LCC = 0.9248 | SRCC = 0.9256 ] +INFO:root:Average loss=0.20092974197864533 +INFO:root: +[Valid][2][UTT][ MSE = 0.3510 | LCC = 0.9158 | SRCC = 0.9186 ] +INFO:root:Average loss=0.18877190351486206 +INFO:root: +[Valid][3][UTT][ MSE = 0.2795 | LCC = 0.9346 | SRCC = 0.9354 ] +INFO:root:Average loss=0.16644815629720688 +INFO:root: +[Valid][4][UTT][ MSE = 0.2878 | LCC = 0.9360 | SRCC = 0.9384 ] +INFO:root:Average loss=0.15499469244480132 +INFO:root: +[Valid][5][UTT][ MSE = 0.2909 | LCC = 0.9390 | SRCC = 0.9421 ] +INFO:root:Average loss=0.15141322267055513 +INFO:root: +[Valid][6][UTT][ MSE = 0.2765 | LCC = 0.9407 | SRCC = 0.9430 ] +INFO:root:Average loss=0.13878727716207503 +INFO:root: +[Valid][7][UTT][ MSE = 0.2505 | LCC = 0.9383 | SRCC = 0.9394 ] +INFO:root:Average loss=0.13785524761676787 +INFO:root: +[Valid][8][UTT][ MSE = 0.2649 | LCC = 0.9391 | SRCC = 0.9422 ] +INFO:root:Average loss=0.1291708934903145 +INFO:root: +[Valid][9][UTT][ MSE = 0.2337 | LCC = 0.9446 | SRCC = 0.9454 ] +INFO:root:Average loss=0.13188169497251512 +INFO:root: +[Valid][10][UTT][ MSE = 0.2645 | LCC = 0.9421 | SRCC = 0.9452 ] +INFO:root:Average loss=0.1176688787639141 +INFO:root: +[Valid][11][UTT][ MSE = 0.2608 | LCC = 0.9434 | SRCC = 0.9461 ] +INFO:root:Average loss=0.11079785972833633 +INFO:root: +[Valid][12][UTT][ MSE = 0.2420 | LCC = 0.9409 | SRCC = 0.9421 ] +INFO:root:Average loss=0.10880159848928452 +INFO:root: +[Valid][13][UTT][ MSE = 0.2672 | LCC = 0.9411 | SRCC = 0.9445 ] +INFO:root:Average loss=0.10370312339067458 +INFO:root: +[Valid][14][UTT][ MSE = 0.2098 | LCC = 0.9457 | SRCC = 0.9467 ] +INFO:root:Average loss=0.1039863538146019 +INFO:root: +[Valid][15][UTT][ MSE = 0.2611 | LCC = 0.9417 | SRCC = 0.9420 ] +INFO:root:Average loss=0.1002749639749527 +INFO:root: +[Valid][16][UTT][ MSE = 0.2113 | LCC = 0.9455 | SRCC = 0.9464 ] +INFO:root:Average loss=0.09952416816353798 +INFO:root: +[Valid][17][UTT][ MSE = 0.2535 | LCC = 0.9447 | SRCC = 0.9474 ] +INFO:root:Average loss=0.09070604112744331 +INFO:root: +[Valid][18][UTT][ MSE = 0.2608 | LCC = 0.9459 | SRCC = 0.9486 ] +INFO:root:Average loss=0.09162835359573364 +INFO:root: +[Valid][19][UTT][ MSE = 0.2295 | LCC = 0.9479 | SRCC = 0.9493 ] +INFO:root:Average loss=0.08949027615785599 +INFO:root: +[Valid][20][UTT][ MSE = 0.2632 | LCC = 0.9428 | SRCC = 0.9445 ] +INFO:root:Average loss=0.08289537417888641 +INFO:root: +[Valid][21][UTT][ MSE = 0.2172 | LCC = 0.9474 | SRCC = 0.9481 ] +INFO:root:Average loss=0.07950579404830932 +INFO:root: +[Valid][22][UTT][ MSE = 0.2493 | LCC = 0.9446 | SRCC = 0.9470 ] +INFO:root:Average loss=0.07792687469720841 +INFO:root: +[Valid][23][UTT][ MSE = 0.2379 | LCC = 0.9461 | SRCC = 0.9472 ] +INFO:root:Average loss=0.08088292121887207 +INFO:root: +[Valid][24][UTT][ MSE = 0.2275 | LCC = 0.9469 | SRCC = 0.9486 ] +INFO:root:Average loss=0.07764821580052375 +INFO:root: +[Valid][25][UTT][ MSE = 0.2295 | LCC = 0.9466 | SRCC = 0.9483 ] +INFO:root:Average loss=0.0730533946454525 +INFO:root: +[Valid][26][UTT][ MSE = 0.2112 | LCC = 0.9457 | SRCC = 0.9464 ] +INFO:root:Average loss=0.07397855737805367 +INFO:root: +[Valid][27][UTT][ MSE = 0.2523 | LCC = 0.9360 | SRCC = 0.9369 ] +INFO:root:Average loss=0.06885792133212089 +INFO:root: +[Valid][28][UTT][ MSE = 0.2325 | LCC = 0.9461 | SRCC = 0.9484 ] +INFO:root:Average loss=0.0672453345656395 +INFO:root: +[Valid][29][UTT][ MSE = 0.2867 | LCC = 0.9437 | SRCC = 0.9485 ] +INFO:root:Average loss=0.06656914681196213 +INFO:root: +[Valid][30][UTT][ MSE = 0.2816 | LCC = 0.9424 | SRCC = 0.9464 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2340 | LCC = 0.9501 | SRCC = 0.9514 ] diff --git a/results_tencent/w2v2_base/layer11/run3/config.gin b/results_tencent/w2v2_base/layer11/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer11/run3/model.pt b/results_tencent/w2v2_base/layer11/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..30475927bb0bc2e519bd6fe013c06264f57edb2f --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f3e914de6ba3e88f896b88ae5b1580470b22da659ec5ecb1dedaf2c82d8a646d +size 2747979 diff --git a/results_tencent/w2v2_base/layer11/run3/model_best.pt b/results_tencent/w2v2_base/layer11/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..8258f9f541da5c12d07723699b453be1c814d610 --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d2eb40cf151c1e4992e0b40e8c886bfc197669573235d4030086fa4a158fc629 +size 2749638 diff --git a/results_tencent/w2v2_base/layer11/run3/train.log b/results_tencent/w2v2_base/layer11/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3ab829f3ac2eedc72e05bc1d2009ec1b2b9e5d10 --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.49507029640674594 +INFO:root: +[Valid][0][UTT][ MSE = 0.3523 | LCC = 0.9052 | SRCC = 0.9071 ] +INFO:root:Average loss=0.24515238904953002 +INFO:root: +[Valid][1][UTT][ MSE = 0.3945 | LCC = 0.9245 | SRCC = 0.9284 ] +INFO:root:Average loss=0.20210679119825364 +INFO:root: +[Valid][2][UTT][ MSE = 0.2944 | LCC = 0.9310 | SRCC = 0.9331 ] +INFO:root:Average loss=0.18246727085113526 +INFO:root: +[Valid][3][UTT][ MSE = 0.3022 | LCC = 0.9348 | SRCC = 0.9352 ] +INFO:root:Average loss=0.17159550511837005 +INFO:root: +[Valid][4][UTT][ MSE = 0.2726 | LCC = 0.9324 | SRCC = 0.9336 ] +INFO:root:Average loss=0.16500129014253617 +INFO:root: +[Valid][5][UTT][ MSE = 0.2559 | LCC = 0.9382 | SRCC = 0.9406 ] +INFO:root:Average loss=0.15160028767585754 +INFO:root: +[Valid][6][UTT][ MSE = 0.2542 | LCC = 0.9425 | SRCC = 0.9431 ] +INFO:root:Average loss=0.1403082319498062 +INFO:root: +[Valid][7][UTT][ MSE = 0.3020 | LCC = 0.9425 | SRCC = 0.9446 ] +INFO:root:Average loss=0.1346265340447426 +INFO:root: +[Valid][8][UTT][ MSE = 0.2656 | LCC = 0.9276 | SRCC = 0.9276 ] +INFO:root:Average loss=0.1319011937379837 +INFO:root: +[Valid][9][UTT][ MSE = 0.2325 | LCC = 0.9419 | SRCC = 0.9424 ] +INFO:root:Average loss=0.12495408433675766 +INFO:root: +[Valid][10][UTT][ MSE = 0.2623 | LCC = 0.9443 | SRCC = 0.9453 ] +INFO:root:Average loss=0.12400438794493675 +INFO:root: +[Valid][11][UTT][ MSE = 0.2257 | LCC = 0.9445 | SRCC = 0.9456 ] +INFO:root:Average loss=0.11412214070558548 +INFO:root: +[Valid][12][UTT][ MSE = 0.2402 | LCC = 0.9409 | SRCC = 0.9424 ] +INFO:root:Average loss=0.11210138261318207 +INFO:root: +[Valid][13][UTT][ MSE = 0.2326 | LCC = 0.9416 | SRCC = 0.9433 ] +INFO:root:Average loss=0.10748193097114563 +INFO:root: +[Valid][14][UTT][ MSE = 0.2508 | LCC = 0.9443 | SRCC = 0.9461 ] +INFO:root:Average loss=0.10413589492440224 +INFO:root: +[Valid][15][UTT][ MSE = 0.2250 | LCC = 0.9436 | SRCC = 0.9446 ] +INFO:root:Average loss=0.10705152305960655 +INFO:root: +[Valid][16][UTT][ MSE = 0.2620 | LCC = 0.9466 | SRCC = 0.9484 ] +INFO:root:Average loss=0.10192995154857636 +INFO:root: +[Valid][17][UTT][ MSE = 0.2605 | LCC = 0.9440 | SRCC = 0.9446 ] +INFO:root:Average loss=0.09732627904415131 +INFO:root: +[Valid][18][UTT][ MSE = 0.2081 | LCC = 0.9449 | SRCC = 0.9464 ] +INFO:root:Average loss=0.0955988620519638 +INFO:root: +[Valid][19][UTT][ MSE = 0.2571 | LCC = 0.9445 | SRCC = 0.9456 ] +INFO:root:Average loss=0.08671276345849037 +INFO:root: +[Valid][20][UTT][ MSE = 0.2399 | LCC = 0.9466 | SRCC = 0.9482 ] +INFO:root:Average loss=0.08636446091532707 +INFO:root: +[Valid][21][UTT][ MSE = 0.2299 | LCC = 0.9451 | SRCC = 0.9462 ] +INFO:root:Average loss=0.08479340046644211 +INFO:root: +[Valid][22][UTT][ MSE = 0.2355 | LCC = 0.9457 | SRCC = 0.9472 ] +INFO:root:Average loss=0.08113335210084915 +INFO:root: +[Valid][23][UTT][ MSE = 0.2687 | LCC = 0.9457 | SRCC = 0.9471 ] +INFO:root:Average loss=0.08288158065080643 +INFO:root: +[Valid][24][UTT][ MSE = 0.2281 | LCC = 0.9449 | SRCC = 0.9458 ] +INFO:root:Average loss=0.07927634730935097 +INFO:root: +[Valid][25][UTT][ MSE = 0.2382 | LCC = 0.9431 | SRCC = 0.9435 ] +INFO:root:Average loss=0.0756127254664898 +INFO:root: +[Valid][26][UTT][ MSE = 0.2313 | LCC = 0.9473 | SRCC = 0.9484 ] +INFO:root:Average loss=0.07488960403203965 +INFO:root: +[Valid][27][UTT][ MSE = 0.2411 | LCC = 0.9464 | SRCC = 0.9475 ] +INFO:root:Average loss=0.07102125999331474 +INFO:root: +[Valid][28][UTT][ MSE = 0.2613 | LCC = 0.9399 | SRCC = 0.9418 ] +INFO:root:Average loss=0.07354590624570846 +INFO:root: +[Valid][29][UTT][ MSE = 0.2253 | LCC = 0.9451 | SRCC = 0.9462 ] +INFO:root:Average loss=0.06978840255737305 +INFO:root: +[Valid][30][UTT][ MSE = 0.2480 | LCC = 0.9465 | SRCC = 0.9481 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2329 | LCC = 0.9502 | SRCC = 0.9511 ] diff --git a/results_tencent/w2v2_base/layer11/run4/config.gin b/results_tencent/w2v2_base/layer11/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer11/run4/model.pt b/results_tencent/w2v2_base/layer11/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..249ab24ce8388ebd1263addad4ff199e6f0c9201 --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ec709639c777442eaeddef7584adc79d677b6f6937043600d3a2961c70ade375 +size 2747979 diff --git a/results_tencent/w2v2_base/layer11/run4/model_best.pt b/results_tencent/w2v2_base/layer11/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d7dfa7e2608eff4b5da516afb8e2e47150618d99 --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d17f742948f4020d3a7a128f283e0e4610e38c5118e757cc4c44383f9aa4edc +size 2749638 diff --git a/results_tencent/w2v2_base/layer11/run4/train.log b/results_tencent/w2v2_base/layer11/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e4e1cf4538e72e7ca2665e0dfcfb8028162004a4 --- /dev/null +++ b/results_tencent/w2v2_base/layer11/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5161828171014786 +INFO:root: +[Valid][0][UTT][ MSE = 0.3590 | LCC = 0.9008 | SRCC = 0.9016 ] +INFO:root:Average loss=0.24595444989204407 +INFO:root: +[Valid][1][UTT][ MSE = 0.2994 | LCC = 0.9215 | SRCC = 0.9231 ] +INFO:root:Average loss=0.20394501787424088 +INFO:root: +[Valid][2][UTT][ MSE = 0.2804 | LCC = 0.9312 | SRCC = 0.9333 ] +INFO:root:Average loss=0.1884960209131241 +INFO:root: +[Valid][3][UTT][ MSE = 0.2259 | LCC = 0.9320 | SRCC = 0.9333 ] +INFO:root:Average loss=0.168979154586792 +INFO:root: +[Valid][4][UTT][ MSE = 0.2117 | LCC = 0.9365 | SRCC = 0.9388 ] +INFO:root:Average loss=0.16066139674186705 +INFO:root: +[Valid][5][UTT][ MSE = 0.2452 | LCC = 0.9375 | SRCC = 0.9382 ] +INFO:root:Average loss=0.14733073657751083 +INFO:root: +[Valid][6][UTT][ MSE = 0.2427 | LCC = 0.9419 | SRCC = 0.9432 ] +INFO:root:Average loss=0.13646503740549087 +INFO:root: +[Valid][7][UTT][ MSE = 0.2403 | LCC = 0.9429 | SRCC = 0.9440 ] +INFO:root:Average loss=0.13941759246587754 +INFO:root: +[Valid][8][UTT][ MSE = 0.2427 | LCC = 0.9437 | SRCC = 0.9447 ] +INFO:root:Average loss=0.12822344127297403 +INFO:root: +[Valid][9][UTT][ MSE = 0.2422 | LCC = 0.9439 | SRCC = 0.9446 ] +INFO:root:Average loss=0.12160122781991958 +INFO:root: +[Valid][10][UTT][ MSE = 0.2353 | LCC = 0.9441 | SRCC = 0.9454 ] +INFO:root:Average loss=0.11459461653232575 +INFO:root: +[Valid][11][UTT][ MSE = 0.2131 | LCC = 0.9430 | SRCC = 0.9448 ] +INFO:root:Average loss=0.11289322054386139 +INFO:root: +[Valid][12][UTT][ MSE = 0.2032 | LCC = 0.9437 | SRCC = 0.9459 ] +INFO:root:Average loss=0.10700430315732956 +INFO:root: +[Valid][13][UTT][ MSE = 0.2511 | LCC = 0.9432 | SRCC = 0.9435 ] +INFO:root:Average loss=0.10598276364803315 +INFO:root: +[Valid][14][UTT][ MSE = 0.2203 | LCC = 0.9470 | SRCC = 0.9479 ] +INFO:root:Average loss=0.10183336877822877 +INFO:root: +[Valid][15][UTT][ MSE = 0.2125 | LCC = 0.9475 | SRCC = 0.9480 ] +INFO:root:Average loss=0.09885310247540474 +INFO:root: +[Valid][16][UTT][ MSE = 0.2702 | LCC = 0.9451 | SRCC = 0.9458 ] +INFO:root:Average loss=0.09709569963812828 +INFO:root: +[Valid][17][UTT][ MSE = 0.2284 | LCC = 0.9462 | SRCC = 0.9470 ] +INFO:root:Average loss=0.09241523844003677 +INFO:root: +[Valid][18][UTT][ MSE = 0.2505 | LCC = 0.9438 | SRCC = 0.9449 ] +INFO:root:Average loss=0.0924131990969181 +INFO:root: +[Valid][19][UTT][ MSE = 0.2243 | LCC = 0.9413 | SRCC = 0.9422 ] +INFO:root:Average loss=0.0856989551782608 +INFO:root: +[Valid][20][UTT][ MSE = 0.2142 | LCC = 0.9460 | SRCC = 0.9471 ] +INFO:root:Average loss=0.08361427620053291 +INFO:root: +[Valid][21][UTT][ MSE = 0.2166 | LCC = 0.9449 | SRCC = 0.9466 ] +INFO:root:Average loss=0.08031192472577095 +INFO:root: +[Valid][22][UTT][ MSE = 0.2108 | LCC = 0.9451 | SRCC = 0.9467 ] +INFO:root:Average loss=0.07725468945503235 +INFO:root: +[Valid][23][UTT][ MSE = 0.2251 | LCC = 0.9432 | SRCC = 0.9437 ] +INFO:root:Average loss=0.07761018657684327 +INFO:root: +[Valid][24][UTT][ MSE = 0.2120 | LCC = 0.9458 | SRCC = 0.9466 ] +INFO:root:Average loss=0.07815147525072098 +INFO:root: +[Valid][25][UTT][ MSE = 0.1914 | LCC = 0.9458 | SRCC = 0.9470 ] +INFO:root:Average loss=0.07208443084359169 +INFO:root: +[Valid][26][UTT][ MSE = 0.2354 | LCC = 0.9458 | SRCC = 0.9470 ] +INFO:root:Average loss=0.07060180923342704 +INFO:root: +[Valid][27][UTT][ MSE = 0.2190 | LCC = 0.9441 | SRCC = 0.9447 ] +INFO:root:Average loss=0.07053875008225441 +INFO:root: +[Valid][28][UTT][ MSE = 0.2043 | LCC = 0.9449 | SRCC = 0.9459 ] +INFO:root:Average loss=0.07057834377884865 +INFO:root: +[Valid][29][UTT][ MSE = 0.2415 | LCC = 0.9454 | SRCC = 0.9471 ] +INFO:root:Average loss=0.06281989619135857 +INFO:root: +[Valid][30][UTT][ MSE = 0.2355 | LCC = 0.9448 | SRCC = 0.9464 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2134 | LCC = 0.9506 | SRCC = 0.9508 ] diff --git a/results_tencent/w2v2_base/layer2/run0/config.gin b/results_tencent/w2v2_base/layer2/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer2/run0/model.pt b/results_tencent/w2v2_base/layer2/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8d2270ccbd3c00c37b95fa3489f443133f9d0536 --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d89f6e4dea21a54b2cd66ece9a2c46b4910a09df9582c97b2683753b5a8694ec +size 2747979 diff --git a/results_tencent/w2v2_base/layer2/run0/model_best.pt b/results_tencent/w2v2_base/layer2/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..1111d9ad59d1fc9e30843e96f9c9f26c588f861b --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4fb5da44ca842a3d05b21104a92d32170638c2aa7c5cfead7afc443aaaf37455 +size 2749702 diff --git a/results_tencent/w2v2_base/layer2/run0/train.log b/results_tencent/w2v2_base/layer2/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a2ec3871685792ec58132d2615ef02a0fc0d9b03 --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5856948512792587 +INFO:root: +[Valid][0][UTT][ MSE = 0.3380 | LCC = 0.9183 | SRCC = 0.9183 ] +INFO:root:Average loss=0.18739027136564254 +INFO:root: +[Valid][1][UTT][ MSE = 0.2788 | LCC = 0.9424 | SRCC = 0.9443 ] +INFO:root:Average loss=0.15097570621967316 +INFO:root: +[Valid][2][UTT][ MSE = 0.2594 | LCC = 0.9502 | SRCC = 0.9517 ] +INFO:root:Average loss=0.12709319716691972 +INFO:root: +[Valid][3][UTT][ MSE = 0.2516 | LCC = 0.9501 | SRCC = 0.9516 ] +INFO:root:Average loss=0.11090694907307624 +INFO:root: +[Valid][4][UTT][ MSE = 0.2367 | LCC = 0.9541 | SRCC = 0.9547 ] +INFO:root:Average loss=0.10263495874404907 +INFO:root: +[Valid][5][UTT][ MSE = 0.2115 | LCC = 0.9577 | SRCC = 0.9584 ] +INFO:root:Average loss=0.09442538702487946 +INFO:root: +[Valid][6][UTT][ MSE = 0.2246 | LCC = 0.9584 | SRCC = 0.9593 ] +INFO:root:Average loss=0.08455735731124878 +INFO:root: +[Valid][7][UTT][ MSE = 0.2027 | LCC = 0.9599 | SRCC = 0.9606 ] +INFO:root:Average loss=0.0815282621383667 +INFO:root: +[Valid][8][UTT][ MSE = 0.2057 | LCC = 0.9600 | SRCC = 0.9613 ] +INFO:root:Average loss=0.07484030628204345 +INFO:root: +[Valid][9][UTT][ MSE = 0.2174 | LCC = 0.9587 | SRCC = 0.9595 ] +INFO:root:Average loss=0.07057100656628609 +INFO:root: +[Valid][10][UTT][ MSE = 0.2035 | LCC = 0.9591 | SRCC = 0.9605 ] +INFO:root:Average loss=0.06679676029086112 +INFO:root: +[Valid][11][UTT][ MSE = 0.1894 | LCC = 0.9612 | SRCC = 0.9620 ] +INFO:root:Average loss=0.060386383026838304 +INFO:root: +[Valid][12][UTT][ MSE = 0.1961 | LCC = 0.9612 | SRCC = 0.9621 ] +INFO:root:Average loss=0.06018797835707664 +INFO:root: +[Valid][13][UTT][ MSE = 0.1857 | LCC = 0.9623 | SRCC = 0.9629 ] +INFO:root:Average loss=0.054301320433616636 +INFO:root: +[Valid][14][UTT][ MSE = 0.1739 | LCC = 0.9615 | SRCC = 0.9617 ] +INFO:root:Average loss=0.05101581424474716 +INFO:root: +[Valid][15][UTT][ MSE = 0.1917 | LCC = 0.9610 | SRCC = 0.9618 ] +INFO:root:Average loss=0.050565781176090244 +INFO:root: +[Valid][16][UTT][ MSE = 0.2096 | LCC = 0.9612 | SRCC = 0.9614 ] +INFO:root:Average loss=0.04686493061482906 +INFO:root: +[Valid][17][UTT][ MSE = 0.1869 | LCC = 0.9618 | SRCC = 0.9627 ] +INFO:root:Average loss=0.044601531267166136 +INFO:root: +[Valid][18][UTT][ MSE = 0.1935 | LCC = 0.9617 | SRCC = 0.9621 ] +INFO:root:Average loss=0.042568092986941336 +INFO:root: +[Valid][19][UTT][ MSE = 0.2034 | LCC = 0.9603 | SRCC = 0.9622 ] +INFO:root:Average loss=0.04227137991786003 +INFO:root: +[Valid][20][UTT][ MSE = 0.1889 | LCC = 0.9614 | SRCC = 0.9620 ] +INFO:root:Average loss=0.03982332476973534 +INFO:root: +[Valid][21][UTT][ MSE = 0.2209 | LCC = 0.9595 | SRCC = 0.9624 ] +INFO:root:Average loss=0.03835466592013836 +INFO:root: +[Valid][22][UTT][ MSE = 0.1972 | LCC = 0.9596 | SRCC = 0.9617 ] +INFO:root:Average loss=0.035417172059416774 +INFO:root: +[Valid][23][UTT][ MSE = 0.1907 | LCC = 0.9613 | SRCC = 0.9624 ] +INFO:root:Average loss=0.03455271877348423 +INFO:root: +[Valid][24][UTT][ MSE = 0.1773 | LCC = 0.9602 | SRCC = 0.9608 ] +INFO:root:Average loss=0.03401912330091 +INFO:root: +[Valid][25][UTT][ MSE = 0.1900 | LCC = 0.9622 | SRCC = 0.9627 ] +INFO:root:Average loss=0.033823756024241446 +INFO:root: +[Valid][26][UTT][ MSE = 0.1866 | LCC = 0.9617 | SRCC = 0.9626 ] +INFO:root:Average loss=0.031205155283212663 +INFO:root: +[Valid][27][UTT][ MSE = 0.1875 | LCC = 0.9618 | SRCC = 0.9624 ] +INFO:root:Average loss=0.030642650932073592 +INFO:root: +[Valid][28][UTT][ MSE = 0.1879 | LCC = 0.9617 | SRCC = 0.9624 ] +INFO:root:Average loss=0.028120731234550476 +INFO:root: +[Valid][29][UTT][ MSE = 0.1938 | LCC = 0.9615 | SRCC = 0.9620 ] +INFO:root:Average loss=0.028246744468808174 +INFO:root: +[Valid][30][UTT][ MSE = 0.1774 | LCC = 0.9624 | SRCC = 0.9628 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1887 | LCC = 0.9599 | SRCC = 0.9615 ] diff --git a/results_tencent/w2v2_base/layer2/run1/config.gin b/results_tencent/w2v2_base/layer2/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer2/run1/model.pt b/results_tencent/w2v2_base/layer2/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..82155bfbbdaeae62b20439af2556a7c29269d92d --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d584a3c812a42fc0fff7294c04626023520da9d3d5a395ab0a32865940935a0b +size 2747979 diff --git a/results_tencent/w2v2_base/layer2/run1/model_best.pt b/results_tencent/w2v2_base/layer2/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..74c291d44d935da2b4b3fae984f492b9e3a7c5f5 --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e80ae0e2e2c39e660b4a3ae72e2db7030425524ba13edc26fb1f38424167673 +size 2749702 diff --git a/results_tencent/w2v2_base/layer2/run1/train.log b/results_tencent/w2v2_base/layer2/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..985779982f5d967a22d079eb7819f905f3aa988c --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4339019076824188 +INFO:root: +[Valid][0][UTT][ MSE = 0.3052 | LCC = 0.9202 | SRCC = 0.9221 ] +INFO:root:Average loss=0.18805820763111114 +INFO:root: +[Valid][1][UTT][ MSE = 0.2496 | LCC = 0.9369 | SRCC = 0.9384 ] +INFO:root:Average loss=0.1480264095067978 +INFO:root: +[Valid][2][UTT][ MSE = 0.2107 | LCC = 0.9462 | SRCC = 0.9464 ] +INFO:root:Average loss=0.12726880198717117 +INFO:root: +[Valid][3][UTT][ MSE = 0.2521 | LCC = 0.9489 | SRCC = 0.9487 ] +INFO:root:Average loss=0.11553645369410515 +INFO:root: +[Valid][4][UTT][ MSE = 0.2502 | LCC = 0.9534 | SRCC = 0.9539 ] +INFO:root:Average loss=0.10477597284317017 +INFO:root: +[Valid][5][UTT][ MSE = 0.2129 | LCC = 0.9569 | SRCC = 0.9570 ] +INFO:root:Average loss=0.09670401835441589 +INFO:root: +[Valid][6][UTT][ MSE = 0.2184 | LCC = 0.9554 | SRCC = 0.9561 ] +INFO:root:Average loss=0.09026175481081009 +INFO:root: +[Valid][7][UTT][ MSE = 0.2238 | LCC = 0.9574 | SRCC = 0.9581 ] +INFO:root:Average loss=0.08202378794550896 +INFO:root: +[Valid][8][UTT][ MSE = 0.2042 | LCC = 0.9572 | SRCC = 0.9581 ] +INFO:root:Average loss=0.0792367514371872 +INFO:root: +[Valid][9][UTT][ MSE = 0.2135 | LCC = 0.9589 | SRCC = 0.9594 ] +INFO:root:Average loss=0.07554311496019363 +INFO:root: +[Valid][10][UTT][ MSE = 0.2340 | LCC = 0.9587 | SRCC = 0.9599 ] +INFO:root:Average loss=0.06909546288847923 +INFO:root: +[Valid][11][UTT][ MSE = 0.2046 | LCC = 0.9597 | SRCC = 0.9606 ] +INFO:root:Average loss=0.06460147219896316 +INFO:root: +[Valid][12][UTT][ MSE = 0.2099 | LCC = 0.9591 | SRCC = 0.9601 ] +INFO:root:Average loss=0.060111500293016436 +INFO:root: +[Valid][13][UTT][ MSE = 0.1980 | LCC = 0.9599 | SRCC = 0.9600 ] +INFO:root:Average loss=0.05922685393691063 +INFO:root: +[Valid][14][UTT][ MSE = 0.2273 | LCC = 0.9606 | SRCC = 0.9608 ] +INFO:root:Average loss=0.053945865899324416 +INFO:root: +[Valid][15][UTT][ MSE = 0.1871 | LCC = 0.9615 | SRCC = 0.9617 ] +INFO:root:Average loss=0.05256804770231247 +INFO:root: +[Valid][16][UTT][ MSE = 0.2018 | LCC = 0.9611 | SRCC = 0.9612 ] +INFO:root:Average loss=0.05317214018106461 +INFO:root: +[Valid][17][UTT][ MSE = 0.2179 | LCC = 0.9588 | SRCC = 0.9593 ] +INFO:root:Average loss=0.04875285659730434 +INFO:root: +[Valid][18][UTT][ MSE = 0.1689 | LCC = 0.9610 | SRCC = 0.9606 ] +INFO:root:Average loss=0.04853769356012344 +INFO:root: +[Valid][19][UTT][ MSE = 0.1831 | LCC = 0.9599 | SRCC = 0.9599 ] +INFO:root:Average loss=0.046294996589422224 +INFO:root: +[Valid][20][UTT][ MSE = 0.1870 | LCC = 0.9620 | SRCC = 0.9619 ] +INFO:root:Average loss=0.04236216396093369 +INFO:root: +[Valid][21][UTT][ MSE = 0.1853 | LCC = 0.9616 | SRCC = 0.9616 ] +INFO:root:Average loss=0.039178299278020856 +INFO:root: +[Valid][22][UTT][ MSE = 0.1794 | LCC = 0.9609 | SRCC = 0.9610 ] +INFO:root:Average loss=0.03975175330042839 +INFO:root: +[Valid][23][UTT][ MSE = 0.1847 | LCC = 0.9620 | SRCC = 0.9618 ] +INFO:root:Average loss=0.03790787336230278 +INFO:root: +[Valid][24][UTT][ MSE = 0.1869 | LCC = 0.9620 | SRCC = 0.9617 ] +INFO:root:Average loss=0.03518472297489643 +INFO:root: +[Valid][25][UTT][ MSE = 0.1991 | LCC = 0.9615 | SRCC = 0.9615 ] +INFO:root:Average loss=0.035642568141222 +INFO:root: +[Valid][26][UTT][ MSE = 0.1870 | LCC = 0.9616 | SRCC = 0.9614 ] +INFO:root:Average loss=0.03558756808936596 +INFO:root: +[Valid][27][UTT][ MSE = 0.1809 | LCC = 0.9617 | SRCC = 0.9619 ] +INFO:root:Average loss=0.03251959536969662 +INFO:root: +[Valid][28][UTT][ MSE = 0.1730 | LCC = 0.9615 | SRCC = 0.9613 ] +INFO:root:Average loss=0.03229303640127182 +INFO:root: +[Valid][29][UTT][ MSE = 0.1991 | LCC = 0.9614 | SRCC = 0.9616 ] +INFO:root:Average loss=0.029922035433351993 +INFO:root: +[Valid][30][UTT][ MSE = 0.1798 | LCC = 0.9614 | SRCC = 0.9614 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1933 | LCC = 0.9619 | SRCC = 0.9626 ] diff --git a/results_tencent/w2v2_base/layer2/run2/config.gin b/results_tencent/w2v2_base/layer2/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer2/run2/model.pt b/results_tencent/w2v2_base/layer2/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..324a36a7258b09a32c903c473ca7700e90d34f92 --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9fabea5b229d9778f8d23437561e36e20693eabf1dcb67f6051fd859b4c30072 +size 2747979 diff --git a/results_tencent/w2v2_base/layer2/run2/model_best.pt b/results_tencent/w2v2_base/layer2/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e230311ae24ba29331a5e8f9495441db07442429 --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6150bd70f354df0daae01f76bef41e757fc9825729931228a9401f9f0e001a45 +size 2749638 diff --git a/results_tencent/w2v2_base/layer2/run2/train.log b/results_tencent/w2v2_base/layer2/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3eef6b227cad8e0047efb2e32f7960c3fe6c361d --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.47560793995857237 +INFO:root: +[Valid][0][UTT][ MSE = 0.2795 | LCC = 0.9167 | SRCC = 0.9193 ] +INFO:root:Average loss=0.19367209142446518 +INFO:root: +[Valid][1][UTT][ MSE = 0.2651 | LCC = 0.9379 | SRCC = 0.9387 ] +INFO:root:Average loss=0.15128176397085188 +INFO:root: +[Valid][2][UTT][ MSE = 0.2381 | LCC = 0.9479 | SRCC = 0.9487 ] +INFO:root:Average loss=0.13211281961202623 +INFO:root: +[Valid][3][UTT][ MSE = 0.2104 | LCC = 0.9511 | SRCC = 0.9517 ] +INFO:root:Average loss=0.11954992890357971 +INFO:root: +[Valid][4][UTT][ MSE = 0.2161 | LCC = 0.9520 | SRCC = 0.9528 ] +INFO:root:Average loss=0.10772420191764831 +INFO:root: +[Valid][5][UTT][ MSE = 0.1985 | LCC = 0.9555 | SRCC = 0.9566 ] +INFO:root:Average loss=0.09877047142386436 +INFO:root: +[Valid][6][UTT][ MSE = 0.2169 | LCC = 0.9560 | SRCC = 0.9565 ] +INFO:root:Average loss=0.09312293750047684 +INFO:root: +[Valid][7][UTT][ MSE = 0.2035 | LCC = 0.9555 | SRCC = 0.9559 ] +INFO:root:Average loss=0.08587145152688026 +INFO:root: +[Valid][8][UTT][ MSE = 0.1965 | LCC = 0.9583 | SRCC = 0.9590 ] +INFO:root:Average loss=0.0784718444943428 +INFO:root: +[Valid][9][UTT][ MSE = 0.1882 | LCC = 0.9587 | SRCC = 0.9593 ] +INFO:root:Average loss=0.07784395298361778 +INFO:root: +[Valid][10][UTT][ MSE = 0.2222 | LCC = 0.9582 | SRCC = 0.9594 ] +INFO:root:Average loss=0.07092516186833382 +INFO:root: +[Valid][11][UTT][ MSE = 0.2103 | LCC = 0.9604 | SRCC = 0.9611 ] +INFO:root:Average loss=0.0703880267739296 +INFO:root: +[Valid][12][UTT][ MSE = 0.1951 | LCC = 0.9607 | SRCC = 0.9612 ] +INFO:root:Average loss=0.06300024205446243 +INFO:root: +[Valid][13][UTT][ MSE = 0.1912 | LCC = 0.9601 | SRCC = 0.9609 ] +INFO:root:Average loss=0.06124608716368675 +INFO:root: +[Valid][14][UTT][ MSE = 0.2079 | LCC = 0.9588 | SRCC = 0.9590 ] +INFO:root:Average loss=0.057909306213259695 +INFO:root: +[Valid][15][UTT][ MSE = 0.1978 | LCC = 0.9607 | SRCC = 0.9612 ] +INFO:root:Average loss=0.054226934149861336 +INFO:root: +[Valid][16][UTT][ MSE = 0.1856 | LCC = 0.9614 | SRCC = 0.9621 ] +INFO:root:Average loss=0.05045130151510239 +INFO:root: +[Valid][17][UTT][ MSE = 0.1996 | LCC = 0.9600 | SRCC = 0.9612 ] +INFO:root:Average loss=0.04966398403048515 +INFO:root: +[Valid][18][UTT][ MSE = 0.1869 | LCC = 0.9611 | SRCC = 0.9612 ] +INFO:root:Average loss=0.04738026760518551 +INFO:root: +[Valid][19][UTT][ MSE = 0.1969 | LCC = 0.9595 | SRCC = 0.9601 ] +INFO:root:Average loss=0.0453522901982069 +INFO:root: +[Valid][20][UTT][ MSE = 0.1942 | LCC = 0.9591 | SRCC = 0.9591 ] +INFO:root:Average loss=0.04357967771589756 +INFO:root: +[Valid][21][UTT][ MSE = 0.1593 | LCC = 0.9620 | SRCC = 0.9623 ] +INFO:root:Average loss=0.042852017298340796 +INFO:root: +[Valid][22][UTT][ MSE = 0.1881 | LCC = 0.9600 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04380917733907699 +INFO:root: +[Valid][23][UTT][ MSE = 0.1771 | LCC = 0.9619 | SRCC = 0.9621 ] +INFO:root:Average loss=0.03859485417604446 +INFO:root: +[Valid][24][UTT][ MSE = 0.1797 | LCC = 0.9619 | SRCC = 0.9620 ] +INFO:root:Average loss=0.03794382807612419 +INFO:root: +[Valid][25][UTT][ MSE = 0.1799 | LCC = 0.9606 | SRCC = 0.9606 ] +INFO:root:Average loss=0.03536686600744724 +INFO:root: +[Valid][26][UTT][ MSE = 0.1833 | LCC = 0.9612 | SRCC = 0.9611 ] +INFO:root:Average loss=0.03576806637644768 +INFO:root: +[Valid][27][UTT][ MSE = 0.1850 | LCC = 0.9607 | SRCC = 0.9607 ] +INFO:root:Average loss=0.03303347520530224 +INFO:root: +[Valid][28][UTT][ MSE = 0.1890 | LCC = 0.9617 | SRCC = 0.9623 ] +INFO:root:Average loss=0.03393215462565422 +INFO:root: +[Valid][29][UTT][ MSE = 0.1631 | LCC = 0.9621 | SRCC = 0.9624 ] +INFO:root:Average loss=0.03191566719114781 +INFO:root: +[Valid][30][UTT][ MSE = 0.1832 | LCC = 0.9610 | SRCC = 0.9610 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1609 | LCC = 0.9633 | SRCC = 0.9634 ] diff --git a/results_tencent/w2v2_base/layer2/run3/config.gin b/results_tencent/w2v2_base/layer2/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer2/run3/model.pt b/results_tencent/w2v2_base/layer2/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bd2a9c516c61c67911e21fc9db1cb65ccfccae84 --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6512ab0ed89b1bccff476595d67f6bbbfe9f07f4bf07e11765024b7c39ad6186 +size 2747979 diff --git a/results_tencent/w2v2_base/layer2/run3/model_best.pt b/results_tencent/w2v2_base/layer2/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f6f8727655915c3a65402b73b3531363cd8836bc --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7faf1c7b02c2bd6cb44f81af0a12048aecad9639371ef391ce1c06f4e6b1422f +size 2749638 diff --git a/results_tencent/w2v2_base/layer2/run3/train.log b/results_tencent/w2v2_base/layer2/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..4850214cf3609f84aabacc49e765e5f47e671b62 --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46191040313243864 +INFO:root: +[Valid][0][UTT][ MSE = 0.3068 | LCC = 0.9253 | SRCC = 0.9270 ] +INFO:root:Average loss=0.18527924764156342 +INFO:root: +[Valid][1][UTT][ MSE = 0.2272 | LCC = 0.9410 | SRCC = 0.9414 ] +INFO:root:Average loss=0.14219368916749955 +INFO:root: +[Valid][2][UTT][ MSE = 0.2510 | LCC = 0.9474 | SRCC = 0.9477 ] +INFO:root:Average loss=0.13026730442047119 +INFO:root: +[Valid][3][UTT][ MSE = 0.2226 | LCC = 0.9528 | SRCC = 0.9530 ] +INFO:root:Average loss=0.11396081733703613 +INFO:root: +[Valid][4][UTT][ MSE = 0.2285 | LCC = 0.9562 | SRCC = 0.9562 ] +INFO:root:Average loss=0.10492490509152412 +INFO:root: +[Valid][5][UTT][ MSE = 0.2217 | LCC = 0.9546 | SRCC = 0.9552 ] +INFO:root:Average loss=0.0963984449505806 +INFO:root: +[Valid][6][UTT][ MSE = 0.2268 | LCC = 0.9568 | SRCC = 0.9564 ] +INFO:root:Average loss=0.09050767958164214 +INFO:root: +[Valid][7][UTT][ MSE = 0.2138 | LCC = 0.9588 | SRCC = 0.9592 ] +INFO:root:Average loss=0.08210362276434899 +INFO:root: +[Valid][8][UTT][ MSE = 0.2273 | LCC = 0.9594 | SRCC = 0.9592 ] +INFO:root:Average loss=0.07670258501172066 +INFO:root: +[Valid][9][UTT][ MSE = 0.1961 | LCC = 0.9609 | SRCC = 0.9612 ] +INFO:root:Average loss=0.07420992913842202 +INFO:root: +[Valid][10][UTT][ MSE = 0.1849 | LCC = 0.9613 | SRCC = 0.9614 ] +INFO:root:Average loss=0.0672793140411377 +INFO:root: +[Valid][11][UTT][ MSE = 0.1937 | LCC = 0.9604 | SRCC = 0.9604 ] +INFO:root:Average loss=0.06365387436747551 +INFO:root: +[Valid][12][UTT][ MSE = 0.2089 | LCC = 0.9614 | SRCC = 0.9617 ] +INFO:root:Average loss=0.060393993109464646 +INFO:root: +[Valid][13][UTT][ MSE = 0.2090 | LCC = 0.9615 | SRCC = 0.9615 ] +INFO:root:Average loss=0.05927133056521416 +INFO:root: +[Valid][14][UTT][ MSE = 0.1949 | LCC = 0.9614 | SRCC = 0.9612 ] +INFO:root:Average loss=0.05575791072845459 +INFO:root: +[Valid][15][UTT][ MSE = 0.1943 | LCC = 0.9614 | SRCC = 0.9611 ] +INFO:root:Average loss=0.05222832447290421 +INFO:root: +[Valid][16][UTT][ MSE = 0.1840 | LCC = 0.9607 | SRCC = 0.9606 ] +INFO:root:Average loss=0.05078308400511742 +INFO:root: +[Valid][17][UTT][ MSE = 0.1993 | LCC = 0.9614 | SRCC = 0.9608 ] +INFO:root:Average loss=0.04974817803502083 +INFO:root: +[Valid][18][UTT][ MSE = 0.1719 | LCC = 0.9616 | SRCC = 0.9610 ] +INFO:root:Average loss=0.04548726609349251 +INFO:root: +[Valid][19][UTT][ MSE = 0.1891 | LCC = 0.9622 | SRCC = 0.9619 ] +INFO:root:Average loss=0.04223978641629219 +INFO:root: +[Valid][20][UTT][ MSE = 0.1986 | LCC = 0.9615 | SRCC = 0.9609 ] +INFO:root:Average loss=0.04020724767446518 +INFO:root: +[Valid][21][UTT][ MSE = 0.1795 | LCC = 0.9621 | SRCC = 0.9615 ] +INFO:root:Average loss=0.04076806566119194 +INFO:root: +[Valid][22][UTT][ MSE = 0.2078 | LCC = 0.9594 | SRCC = 0.9586 ] +INFO:root:Average loss=0.03789367724955082 +INFO:root: +[Valid][23][UTT][ MSE = 0.1832 | LCC = 0.9616 | SRCC = 0.9617 ] +INFO:root:Average loss=0.039518030375242236 +INFO:root: +[Valid][24][UTT][ MSE = 0.2110 | LCC = 0.9608 | SRCC = 0.9600 ] +INFO:root:Average loss=0.037134180024266245 +INFO:root: +[Valid][25][UTT][ MSE = 0.1864 | LCC = 0.9621 | SRCC = 0.9619 ] +INFO:root:Average loss=0.035176919490098955 +INFO:root: +[Valid][26][UTT][ MSE = 0.1988 | LCC = 0.9618 | SRCC = 0.9615 ] +INFO:root:Average loss=0.033833885252475736 +INFO:root: +[Valid][27][UTT][ MSE = 0.1926 | LCC = 0.9628 | SRCC = 0.9622 ] +INFO:root:Average loss=0.032123164400458336 +INFO:root: +[Valid][28][UTT][ MSE = 0.1667 | LCC = 0.9622 | SRCC = 0.9624 ] +INFO:root:Average loss=0.030204799339175224 +INFO:root: +[Valid][29][UTT][ MSE = 0.2031 | LCC = 0.9620 | SRCC = 0.9615 ] +INFO:root:Average loss=0.030019140228629112 +INFO:root: +[Valid][30][UTT][ MSE = 0.2036 | LCC = 0.9613 | SRCC = 0.9611 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1877 | LCC = 0.9632 | SRCC = 0.9637 ] diff --git a/results_tencent/w2v2_base/layer2/run4/config.gin b/results_tencent/w2v2_base/layer2/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer2/run4/model.pt b/results_tencent/w2v2_base/layer2/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8644b2ef8ced0716ad1cd17e8c8b3f3f60a4b231 --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1bd16e3a163938ffd589e1828e898c05b14c9d542b9212a7e3588d9e98bc6196 +size 2747979 diff --git a/results_tencent/w2v2_base/layer2/run4/model_best.pt b/results_tencent/w2v2_base/layer2/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f5e6953712630fa5edb7f77d9375ac9180972b3e --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:129b052fe900161086fd7a42da5072e293479f88961d61e741bd90ec55a4d63f +size 2749638 diff --git a/results_tencent/w2v2_base/layer2/run4/train.log b/results_tencent/w2v2_base/layer2/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a9abe4ce95bda450e6b5bec596e63fe383fa1112 --- /dev/null +++ b/results_tencent/w2v2_base/layer2/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4645985828638077 +INFO:root: +[Valid][0][UTT][ MSE = 0.3052 | LCC = 0.9220 | SRCC = 0.9237 ] +INFO:root:Average loss=0.18214418107271194 +INFO:root: +[Valid][1][UTT][ MSE = 0.2934 | LCC = 0.9415 | SRCC = 0.9438 ] +INFO:root:Average loss=0.15089934545755387 +INFO:root: +[Valid][2][UTT][ MSE = 0.2435 | LCC = 0.9497 | SRCC = 0.9518 ] +INFO:root:Average loss=0.13097171902656554 +INFO:root: +[Valid][3][UTT][ MSE = 0.2389 | LCC = 0.9544 | SRCC = 0.9556 ] +INFO:root:Average loss=0.11437022006511688 +INFO:root: +[Valid][4][UTT][ MSE = 0.2076 | LCC = 0.9540 | SRCC = 0.9546 ] +INFO:root:Average loss=0.10576553863286972 +INFO:root: +[Valid][5][UTT][ MSE = 0.1906 | LCC = 0.9556 | SRCC = 0.9565 ] +INFO:root:Average loss=0.09931270289421082 +INFO:root: +[Valid][6][UTT][ MSE = 0.2158 | LCC = 0.9564 | SRCC = 0.9571 ] +INFO:root:Average loss=0.09062260353565216 +INFO:root: +[Valid][7][UTT][ MSE = 0.1927 | LCC = 0.9579 | SRCC = 0.9589 ] +INFO:root:Average loss=0.08089686706662177 +INFO:root: +[Valid][8][UTT][ MSE = 0.1940 | LCC = 0.9597 | SRCC = 0.9602 ] +INFO:root:Average loss=0.08014169484376907 +INFO:root: +[Valid][9][UTT][ MSE = 0.2045 | LCC = 0.9571 | SRCC = 0.9577 ] +INFO:root:Average loss=0.07742647889256478 +INFO:root: +[Valid][10][UTT][ MSE = 0.2235 | LCC = 0.9587 | SRCC = 0.9604 ] +INFO:root:Average loss=0.06970648774504662 +INFO:root: +[Valid][11][UTT][ MSE = 0.2122 | LCC = 0.9605 | SRCC = 0.9619 ] +INFO:root:Average loss=0.0660446771979332 +INFO:root: +[Valid][12][UTT][ MSE = 0.1958 | LCC = 0.9596 | SRCC = 0.9607 ] +INFO:root:Average loss=0.06020088791847229 +INFO:root: +[Valid][13][UTT][ MSE = 0.2028 | LCC = 0.9606 | SRCC = 0.9621 ] +INFO:root:Average loss=0.0590966804176569 +INFO:root: +[Valid][14][UTT][ MSE = 0.1712 | LCC = 0.9607 | SRCC = 0.9616 ] +INFO:root:Average loss=0.05691158555448055 +INFO:root: +[Valid][15][UTT][ MSE = 0.2059 | LCC = 0.9590 | SRCC = 0.9599 ] +INFO:root:Average loss=0.053330937817692756 +INFO:root: +[Valid][16][UTT][ MSE = 0.1880 | LCC = 0.9611 | SRCC = 0.9621 ] +INFO:root:Average loss=0.049969565153121946 +INFO:root: +[Valid][17][UTT][ MSE = 0.1789 | LCC = 0.9614 | SRCC = 0.9625 ] +INFO:root:Average loss=0.049618667662143706 +INFO:root: +[Valid][18][UTT][ MSE = 0.1747 | LCC = 0.9618 | SRCC = 0.9624 ] +INFO:root:Average loss=0.048569891780614854 +INFO:root: +[Valid][19][UTT][ MSE = 0.1939 | LCC = 0.9615 | SRCC = 0.9623 ] +INFO:root:Average loss=0.0434344831854105 +INFO:root: +[Valid][20][UTT][ MSE = 0.1734 | LCC = 0.9614 | SRCC = 0.9615 ] +INFO:root:Average loss=0.042706018030643465 +INFO:root: +[Valid][21][UTT][ MSE = 0.1843 | LCC = 0.9604 | SRCC = 0.9611 ] +INFO:root:Average loss=0.04048334690928459 +INFO:root: +[Valid][22][UTT][ MSE = 0.1643 | LCC = 0.9600 | SRCC = 0.9610 ] +INFO:root:Average loss=0.0423705004453659 +INFO:root: +[Valid][23][UTT][ MSE = 0.1915 | LCC = 0.9605 | SRCC = 0.9611 ] +INFO:root:Average loss=0.03636538136005402 +INFO:root: +[Valid][24][UTT][ MSE = 0.1899 | LCC = 0.9619 | SRCC = 0.9623 ] +INFO:root:Average loss=0.03608750015497208 +INFO:root: +[Valid][25][UTT][ MSE = 0.1892 | LCC = 0.9609 | SRCC = 0.9614 ] +INFO:root:Average loss=0.034511686503887175 +INFO:root: +[Valid][26][UTT][ MSE = 0.1986 | LCC = 0.9618 | SRCC = 0.9624 ] +INFO:root:Average loss=0.03264368757605553 +INFO:root: +[Valid][27][UTT][ MSE = 0.1830 | LCC = 0.9618 | SRCC = 0.9620 ] +INFO:root:Average loss=0.031401074856519696 +INFO:root: +[Valid][28][UTT][ MSE = 0.2006 | LCC = 0.9617 | SRCC = 0.9627 ] +INFO:root:Average loss=0.03184669382870197 +INFO:root: +[Valid][29][UTT][ MSE = 0.1856 | LCC = 0.9618 | SRCC = 0.9625 ] +INFO:root:Average loss=0.030551816031336783 +INFO:root: +[Valid][30][UTT][ MSE = 0.1895 | LCC = 0.9613 | SRCC = 0.9616 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1900 | LCC = 0.9629 | SRCC = 0.9637 ] diff --git a/results_tencent/w2v2_base/layer3/run0/config.gin b/results_tencent/w2v2_base/layer3/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer3/run0/model.pt b/results_tencent/w2v2_base/layer3/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..01b1998e1baca34d243394918b6ff683f0e01754 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3188569bfbba99a8812f4a5e844be63b67cafe4379fcbd8030b320cfb945050c +size 2747979 diff --git a/results_tencent/w2v2_base/layer3/run0/model_best.pt b/results_tencent/w2v2_base/layer3/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ddb613b761c1f845a1c00d4da6e8fa47241a85c3 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:da18d3cc4998178822a47b74237c671683e867367b7d16fd7b3c7a9905e0c0ed +size 2749638 diff --git a/results_tencent/w2v2_base/layer3/run0/train.log b/results_tencent/w2v2_base/layer3/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b55a4c2a5e486b668ce0f0c7c9d6840fd3084e93 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.38318222653865813 +INFO:root: +[Valid][0][UTT][ MSE = 0.2886 | LCC = 0.9252 | SRCC = 0.9238 ] +INFO:root:Average loss=0.17581135243177415 +INFO:root: +[Valid][1][UTT][ MSE = 0.2385 | LCC = 0.9433 | SRCC = 0.9447 ] +INFO:root:Average loss=0.1409878050684929 +INFO:root: +[Valid][2][UTT][ MSE = 0.2422 | LCC = 0.9494 | SRCC = 0.9503 ] +INFO:root:Average loss=0.12424791052937507 +INFO:root: +[Valid][3][UTT][ MSE = 0.1879 | LCC = 0.9504 | SRCC = 0.9522 ] +INFO:root:Average loss=0.11060295969247819 +INFO:root: +[Valid][4][UTT][ MSE = 0.1711 | LCC = 0.9528 | SRCC = 0.9539 ] +INFO:root:Average loss=0.1039771226644516 +INFO:root: +[Valid][5][UTT][ MSE = 0.1944 | LCC = 0.9553 | SRCC = 0.9552 ] +INFO:root:Average loss=0.09728397178649903 +INFO:root: +[Valid][6][UTT][ MSE = 0.2015 | LCC = 0.9544 | SRCC = 0.9545 ] +INFO:root:Average loss=0.087167300760746 +INFO:root: +[Valid][7][UTT][ MSE = 0.1939 | LCC = 0.9587 | SRCC = 0.9586 ] +INFO:root:Average loss=0.080818345785141 +INFO:root: +[Valid][8][UTT][ MSE = 0.1845 | LCC = 0.9575 | SRCC = 0.9578 ] +INFO:root:Average loss=0.07896076345443725 +INFO:root: +[Valid][9][UTT][ MSE = 0.1910 | LCC = 0.9594 | SRCC = 0.9595 ] +INFO:root:Average loss=0.07247159573435784 +INFO:root: +[Valid][10][UTT][ MSE = 0.1981 | LCC = 0.9597 | SRCC = 0.9596 ] +INFO:root:Average loss=0.06764095997810364 +INFO:root: +[Valid][11][UTT][ MSE = 0.2083 | LCC = 0.9596 | SRCC = 0.9596 ] +INFO:root:Average loss=0.06175232470035553 +INFO:root: +[Valid][12][UTT][ MSE = 0.2191 | LCC = 0.9595 | SRCC = 0.9597 ] +INFO:root:Average loss=0.061960301488637924 +INFO:root: +[Valid][13][UTT][ MSE = 0.1824 | LCC = 0.9602 | SRCC = 0.9599 ] +INFO:root:Average loss=0.056701491177082064 +INFO:root: +[Valid][14][UTT][ MSE = 0.2042 | LCC = 0.9580 | SRCC = 0.9578 ] +INFO:root:Average loss=0.054073020815849304 +INFO:root: +[Valid][15][UTT][ MSE = 0.1599 | LCC = 0.9586 | SRCC = 0.9594 ] +INFO:root:Average loss=0.0511134392619133 +INFO:root: +[Valid][16][UTT][ MSE = 0.1931 | LCC = 0.9596 | SRCC = 0.9594 ] +INFO:root:Average loss=0.04693780291080475 +INFO:root: +[Valid][17][UTT][ MSE = 0.1790 | LCC = 0.9588 | SRCC = 0.9596 ] +INFO:root:Average loss=0.04600025235116482 +INFO:root: +[Valid][18][UTT][ MSE = 0.1877 | LCC = 0.9575 | SRCC = 0.9581 ] +INFO:root:Average loss=0.045125697433948515 +INFO:root: +[Valid][19][UTT][ MSE = 0.1843 | LCC = 0.9589 | SRCC = 0.9591 ] +INFO:root:Average loss=0.04546265582740307 +INFO:root: +[Valid][20][UTT][ MSE = 0.1727 | LCC = 0.9599 | SRCC = 0.9601 ] +INFO:root:Average loss=0.041723196640610694 +INFO:root: +[Valid][21][UTT][ MSE = 0.1823 | LCC = 0.9589 | SRCC = 0.9594 ] +INFO:root:Average loss=0.038818366542458536 +INFO:root: +[Valid][22][UTT][ MSE = 0.1692 | LCC = 0.9598 | SRCC = 0.9597 ] +INFO:root:Average loss=0.04027046486735344 +INFO:root: +[Valid][23][UTT][ MSE = 0.1818 | LCC = 0.9605 | SRCC = 0.9605 ] +INFO:root:Average loss=0.03677663484215737 +INFO:root: +[Valid][24][UTT][ MSE = 0.1771 | LCC = 0.9590 | SRCC = 0.9591 ] +INFO:root:Average loss=0.03421598662436008 +INFO:root: +[Valid][25][UTT][ MSE = 0.1809 | LCC = 0.9594 | SRCC = 0.9599 ] +INFO:root:Average loss=0.03416808708012104 +INFO:root: +[Valid][26][UTT][ MSE = 0.1879 | LCC = 0.9605 | SRCC = 0.9607 ] +INFO:root:Average loss=0.03232380958646536 +INFO:root: +[Valid][27][UTT][ MSE = 0.1768 | LCC = 0.9601 | SRCC = 0.9604 ] +INFO:root:Average loss=0.032865375995635986 +INFO:root: +[Valid][28][UTT][ MSE = 0.1774 | LCC = 0.9603 | SRCC = 0.9608 ] +INFO:root:Average loss=0.03175116848945618 +INFO:root: +[Valid][29][UTT][ MSE = 0.1935 | LCC = 0.9610 | SRCC = 0.9611 ] +INFO:root:Average loss=0.02983792743831873 +INFO:root: +[Valid][30][UTT][ MSE = 0.1850 | LCC = 0.9607 | SRCC = 0.9606 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1921 | LCC = 0.9623 | SRCC = 0.9628 ] diff --git a/results_tencent/w2v2_base/layer3/run1/config.gin b/results_tencent/w2v2_base/layer3/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer3/run1/model.pt b/results_tencent/w2v2_base/layer3/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a577e5ed6921a7138236bac66dbda5173512b999 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:516c06a67c638738e038c16e6ce72c8887b99a3f7090590e0c80cf790ecd5686 +size 2747979 diff --git a/results_tencent/w2v2_base/layer3/run1/model_best.pt b/results_tencent/w2v2_base/layer3/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2707a2722e0ec0342342218aad3ef55a981cb705 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:edfeb91a0139a826aadc04e92cf0a20bd6da8745a2679c95aa1e29f37cc86ccb +size 2749638 diff --git a/results_tencent/w2v2_base/layer3/run1/train.log b/results_tencent/w2v2_base/layer3/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0f2fc0bd5dca67c8ec4aa09ecd842ff5dded6644 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4212976461648941 +INFO:root: +[Valid][0][UTT][ MSE = 0.2527 | LCC = 0.9315 | SRCC = 0.9330 ] +INFO:root:Average loss=0.1695575966835022 +INFO:root: +[Valid][1][UTT][ MSE = 0.2496 | LCC = 0.9431 | SRCC = 0.9430 ] +INFO:root:Average loss=0.1382922561764717 +INFO:root: +[Valid][2][UTT][ MSE = 0.2367 | LCC = 0.9513 | SRCC = 0.9520 ] +INFO:root:Average loss=0.11922707712650299 +INFO:root: +[Valid][3][UTT][ MSE = 0.2024 | LCC = 0.9527 | SRCC = 0.9543 ] +INFO:root:Average loss=0.10871079304814339 +INFO:root: +[Valid][4][UTT][ MSE = 0.2454 | LCC = 0.9545 | SRCC = 0.9562 ] +INFO:root:Average loss=0.1016488582789898 +INFO:root: +[Valid][5][UTT][ MSE = 0.2021 | LCC = 0.9543 | SRCC = 0.9553 ] +INFO:root:Average loss=0.09377503618597985 +INFO:root: +[Valid][6][UTT][ MSE = 0.1870 | LCC = 0.9569 | SRCC = 0.9576 ] +INFO:root:Average loss=0.0851477279663086 +INFO:root: +[Valid][7][UTT][ MSE = 0.1759 | LCC = 0.9580 | SRCC = 0.9583 ] +INFO:root:Average loss=0.08034432637691498 +INFO:root: +[Valid][8][UTT][ MSE = 0.1933 | LCC = 0.9579 | SRCC = 0.9584 ] +INFO:root:Average loss=0.07616471591591835 +INFO:root: +[Valid][9][UTT][ MSE = 0.1947 | LCC = 0.9590 | SRCC = 0.9596 ] +INFO:root:Average loss=0.07119247677922248 +INFO:root: +[Valid][10][UTT][ MSE = 0.1982 | LCC = 0.9596 | SRCC = 0.9600 ] +INFO:root:Average loss=0.06779711812734604 +INFO:root: +[Valid][11][UTT][ MSE = 0.1686 | LCC = 0.9604 | SRCC = 0.9607 ] +INFO:root:Average loss=0.06270005291700363 +INFO:root: +[Valid][12][UTT][ MSE = 0.1789 | LCC = 0.9596 | SRCC = 0.9599 ] +INFO:root:Average loss=0.05891247773170471 +INFO:root: +[Valid][13][UTT][ MSE = 0.2044 | LCC = 0.9597 | SRCC = 0.9600 ] +INFO:root:Average loss=0.05577555574476719 +INFO:root: +[Valid][14][UTT][ MSE = 0.2028 | LCC = 0.9599 | SRCC = 0.9597 ] +INFO:root:Average loss=0.0530716752409935 +INFO:root: +[Valid][15][UTT][ MSE = 0.2003 | LCC = 0.9605 | SRCC = 0.9607 ] +INFO:root:Average loss=0.050267931699752806 +INFO:root: +[Valid][16][UTT][ MSE = 0.1973 | LCC = 0.9590 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04879061551392078 +INFO:root: +[Valid][17][UTT][ MSE = 0.2026 | LCC = 0.9607 | SRCC = 0.9618 ] +INFO:root:Average loss=0.04639125648140907 +INFO:root: +[Valid][18][UTT][ MSE = 0.1853 | LCC = 0.9605 | SRCC = 0.9607 ] +INFO:root:Average loss=0.04262359811365604 +INFO:root: +[Valid][19][UTT][ MSE = 0.1728 | LCC = 0.9612 | SRCC = 0.9612 ] +INFO:root:Average loss=0.04163797552883625 +INFO:root: +[Valid][20][UTT][ MSE = 0.1827 | LCC = 0.9598 | SRCC = 0.9596 ] +INFO:root:Average loss=0.03979134894907475 +INFO:root: +[Valid][21][UTT][ MSE = 0.2044 | LCC = 0.9582 | SRCC = 0.9589 ] +INFO:root:Average loss=0.041045928612351415 +INFO:root: +[Valid][22][UTT][ MSE = 0.1886 | LCC = 0.9610 | SRCC = 0.9609 ] +INFO:root:Average loss=0.03764885853230953 +INFO:root: +[Valid][23][UTT][ MSE = 0.2101 | LCC = 0.9603 | SRCC = 0.9600 ] +INFO:root:Average loss=0.03710040834546089 +INFO:root: +[Valid][24][UTT][ MSE = 0.1882 | LCC = 0.9608 | SRCC = 0.9605 ] +INFO:root:Average loss=0.03360492289066315 +INFO:root: +[Valid][25][UTT][ MSE = 0.1968 | LCC = 0.9596 | SRCC = 0.9594 ] +INFO:root:Average loss=0.036543464869260786 +INFO:root: +[Valid][26][UTT][ MSE = 0.1770 | LCC = 0.9608 | SRCC = 0.9608 ] +INFO:root:Average loss=0.03078585609793663 +INFO:root: +[Valid][27][UTT][ MSE = 0.1895 | LCC = 0.9604 | SRCC = 0.9604 ] +INFO:root:Average loss=0.028988080121576786 +INFO:root: +[Valid][28][UTT][ MSE = 0.1828 | LCC = 0.9592 | SRCC = 0.9594 ] +INFO:root:Average loss=0.030287978261709213 +INFO:root: +[Valid][29][UTT][ MSE = 0.1965 | LCC = 0.9598 | SRCC = 0.9598 ] +INFO:root:Average loss=0.028728166207671164 +INFO:root: +[Valid][30][UTT][ MSE = 0.2081 | LCC = 0.9609 | SRCC = 0.9611 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1738 | LCC = 0.9612 | SRCC = 0.9623 ] diff --git a/results_tencent/w2v2_base/layer3/run2/config.gin b/results_tencent/w2v2_base/layer3/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer3/run2/model.pt b/results_tencent/w2v2_base/layer3/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..52a5f309728d96f9d53c370081af248bb31f5f5c --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dcfbf04287f5c8ab126713bca732ab0e8675f0afa79f72a646b00195022945b2 +size 2747979 diff --git a/results_tencent/w2v2_base/layer3/run2/model_best.pt b/results_tencent/w2v2_base/layer3/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..dab7e041c3f290eeada8f5517e8cf41819c1842e --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:47e260da417e17da76660462dd546bf1d19e5a6c33630a0a7ca28ac20bb434df +size 2749702 diff --git a/results_tencent/w2v2_base/layer3/run2/train.log b/results_tencent/w2v2_base/layer3/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..083bc217704c2808756914d7b7baacd9e2329639 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5103735047578811 +INFO:root: +[Valid][0][UTT][ MSE = 0.2611 | LCC = 0.9273 | SRCC = 0.9282 ] +INFO:root:Average loss=0.18033341872692107 +INFO:root: +[Valid][1][UTT][ MSE = 0.2301 | LCC = 0.9439 | SRCC = 0.9448 ] +INFO:root:Average loss=0.14787631231546403 +INFO:root: +[Valid][2][UTT][ MSE = 0.1970 | LCC = 0.9475 | SRCC = 0.9485 ] +INFO:root:Average loss=0.12972269314527513 +INFO:root: +[Valid][3][UTT][ MSE = 0.2115 | LCC = 0.9530 | SRCC = 0.9531 ] +INFO:root:Average loss=0.1150755015015602 +INFO:root: +[Valid][4][UTT][ MSE = 0.2126 | LCC = 0.9538 | SRCC = 0.9540 ] +INFO:root:Average loss=0.10387728607654571 +INFO:root: +[Valid][5][UTT][ MSE = 0.2109 | LCC = 0.9534 | SRCC = 0.9544 ] +INFO:root:Average loss=0.10121607851982117 +INFO:root: +[Valid][6][UTT][ MSE = 0.1961 | LCC = 0.9567 | SRCC = 0.9575 ] +INFO:root:Average loss=0.09204978424310684 +INFO:root: +[Valid][7][UTT][ MSE = 0.1792 | LCC = 0.9557 | SRCC = 0.9569 ] +INFO:root:Average loss=0.08319276294112206 +INFO:root: +[Valid][8][UTT][ MSE = 0.2146 | LCC = 0.9565 | SRCC = 0.9568 ] +INFO:root:Average loss=0.07887856203317642 +INFO:root: +[Valid][9][UTT][ MSE = 0.1765 | LCC = 0.9589 | SRCC = 0.9595 ] +INFO:root:Average loss=0.07455550992488862 +INFO:root: +[Valid][10][UTT][ MSE = 0.1736 | LCC = 0.9573 | SRCC = 0.9595 ] +INFO:root:Average loss=0.07036864578723907 +INFO:root: +[Valid][11][UTT][ MSE = 0.2169 | LCC = 0.9581 | SRCC = 0.9594 ] +INFO:root:Average loss=0.06654268181324005 +INFO:root: +[Valid][12][UTT][ MSE = 0.1989 | LCC = 0.9602 | SRCC = 0.9603 ] +INFO:root:Average loss=0.06351102110743523 +INFO:root: +[Valid][13][UTT][ MSE = 0.1787 | LCC = 0.9591 | SRCC = 0.9602 ] +INFO:root:Average loss=0.059201607465744016 +INFO:root: +[Valid][14][UTT][ MSE = 0.1813 | LCC = 0.9602 | SRCC = 0.9608 ] +INFO:root:Average loss=0.054926185891032216 +INFO:root: +[Valid][15][UTT][ MSE = 0.1945 | LCC = 0.9597 | SRCC = 0.9596 ] +INFO:root:Average loss=0.05355389641225338 +INFO:root: +[Valid][16][UTT][ MSE = 0.1838 | LCC = 0.9584 | SRCC = 0.9584 ] +INFO:root:Average loss=0.050967462986707686 +INFO:root: +[Valid][17][UTT][ MSE = 0.1749 | LCC = 0.9616 | SRCC = 0.9626 ] +INFO:root:Average loss=0.048184314146637916 +INFO:root: +[Valid][18][UTT][ MSE = 0.1701 | LCC = 0.9598 | SRCC = 0.9603 ] +INFO:root:Average loss=0.04801710398495197 +INFO:root: +[Valid][19][UTT][ MSE = 0.1647 | LCC = 0.9605 | SRCC = 0.9615 ] +INFO:root:Average loss=0.04545899100601673 +INFO:root: +[Valid][20][UTT][ MSE = 0.1839 | LCC = 0.9602 | SRCC = 0.9609 ] +INFO:root:Average loss=0.04191277877986431 +INFO:root: +[Valid][21][UTT][ MSE = 0.1555 | LCC = 0.9582 | SRCC = 0.9600 ] +INFO:root:Average loss=0.040037925586104396 +INFO:root: +[Valid][22][UTT][ MSE = 0.1771 | LCC = 0.9608 | SRCC = 0.9618 ] +INFO:root:Average loss=0.04138298164308071 +INFO:root: +[Valid][23][UTT][ MSE = 0.1693 | LCC = 0.9605 | SRCC = 0.9610 ] +INFO:root:Average loss=0.03852628193795681 +INFO:root: +[Valid][24][UTT][ MSE = 0.1637 | LCC = 0.9601 | SRCC = 0.9607 ] +INFO:root:Average loss=0.036160078689455984 +INFO:root: +[Valid][25][UTT][ MSE = 0.1724 | LCC = 0.9611 | SRCC = 0.9618 ] +INFO:root:Average loss=0.03615627747774124 +INFO:root: +[Valid][26][UTT][ MSE = 0.1763 | LCC = 0.9620 | SRCC = 0.9624 ] +INFO:root:Average loss=0.033690467849373816 +INFO:root: +[Valid][27][UTT][ MSE = 0.1746 | LCC = 0.9599 | SRCC = 0.9609 ] +INFO:root:Average loss=0.0342740553021431 +INFO:root: +[Valid][28][UTT][ MSE = 0.1880 | LCC = 0.9584 | SRCC = 0.9592 ] +INFO:root:Average loss=0.034505881056189534 +INFO:root: +[Valid][29][UTT][ MSE = 0.1642 | LCC = 0.9614 | SRCC = 0.9620 ] +INFO:root:Average loss=0.03168781504034996 +INFO:root: +[Valid][30][UTT][ MSE = 0.1726 | LCC = 0.9615 | SRCC = 0.9617 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1843 | LCC = 0.9602 | SRCC = 0.9608 ] diff --git a/results_tencent/w2v2_base/layer3/run3/config.gin b/results_tencent/w2v2_base/layer3/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer3/run3/model.pt b/results_tencent/w2v2_base/layer3/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2cf694c4db54d6342c989f5ec374295ce6b931b6 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e6c003730ffb1a8adf486bb7bccd195d46bba6d3e929db33fc04abf0ffffe20f +size 2747979 diff --git a/results_tencent/w2v2_base/layer3/run3/model_best.pt b/results_tencent/w2v2_base/layer3/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f6aaca532707f175b205ca7a68e24d3d9fc5518e --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f1664d7e6cb274b8b51f9106c7b0f0c9f83cf372ba438bfbef503793c3e0d62c +size 2749638 diff --git a/results_tencent/w2v2_base/layer3/run3/train.log b/results_tencent/w2v2_base/layer3/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..55669c94b4e6c845482599123053d6024a53ca07 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.43969398880004884 +INFO:root: +[Valid][0][UTT][ MSE = 0.2979 | LCC = 0.9276 | SRCC = 0.9300 ] +INFO:root:Average loss=0.18096865272521973 +INFO:root: +[Valid][1][UTT][ MSE = 0.2286 | LCC = 0.9459 | SRCC = 0.9468 ] +INFO:root:Average loss=0.14295930254459382 +INFO:root: +[Valid][2][UTT][ MSE = 0.1946 | LCC = 0.9501 | SRCC = 0.9515 ] +INFO:root:Average loss=0.12679281717538834 +INFO:root: +[Valid][3][UTT][ MSE = 0.1817 | LCC = 0.9535 | SRCC = 0.9548 ] +INFO:root:Average loss=0.11519403147697449 +INFO:root: +[Valid][4][UTT][ MSE = 0.1895 | LCC = 0.9548 | SRCC = 0.9551 ] +INFO:root:Average loss=0.10360042876005172 +INFO:root: +[Valid][5][UTT][ MSE = 0.1819 | LCC = 0.9579 | SRCC = 0.9582 ] +INFO:root:Average loss=0.09500545239448548 +INFO:root: +[Valid][6][UTT][ MSE = 0.1742 | LCC = 0.9576 | SRCC = 0.9579 ] +INFO:root:Average loss=0.086694907695055 +INFO:root: +[Valid][7][UTT][ MSE = 0.1978 | LCC = 0.9580 | SRCC = 0.9581 ] +INFO:root:Average loss=0.08038668259978295 +INFO:root: +[Valid][8][UTT][ MSE = 0.1750 | LCC = 0.9588 | SRCC = 0.9591 ] +INFO:root:Average loss=0.07724956235289573 +INFO:root: +[Valid][9][UTT][ MSE = 0.1972 | LCC = 0.9588 | SRCC = 0.9591 ] +INFO:root:Average loss=0.07065532487630845 +INFO:root: +[Valid][10][UTT][ MSE = 0.1727 | LCC = 0.9588 | SRCC = 0.9591 ] +INFO:root:Average loss=0.06661067202687264 +INFO:root: +[Valid][11][UTT][ MSE = 0.1877 | LCC = 0.9583 | SRCC = 0.9578 ] +INFO:root:Average loss=0.06445700138807296 +INFO:root: +[Valid][12][UTT][ MSE = 0.1734 | LCC = 0.9606 | SRCC = 0.9601 ] +INFO:root:Average loss=0.0628869343996048 +INFO:root: +[Valid][13][UTT][ MSE = 0.1905 | LCC = 0.9584 | SRCC = 0.9578 ] +INFO:root:Average loss=0.05476364862918854 +INFO:root: +[Valid][14][UTT][ MSE = 0.1843 | LCC = 0.9584 | SRCC = 0.9579 ] +INFO:root:Average loss=0.058644708260893824 +INFO:root: +[Valid][15][UTT][ MSE = 0.1887 | LCC = 0.9601 | SRCC = 0.9601 ] +INFO:root:Average loss=0.052595284074544904 +INFO:root: +[Valid][16][UTT][ MSE = 0.1574 | LCC = 0.9595 | SRCC = 0.9588 ] +INFO:root:Average loss=0.05040139308571816 +INFO:root: +[Valid][17][UTT][ MSE = 0.1688 | LCC = 0.9599 | SRCC = 0.9593 ] +INFO:root:Average loss=0.0466309716552496 +INFO:root: +[Valid][18][UTT][ MSE = 0.1620 | LCC = 0.9603 | SRCC = 0.9598 ] +INFO:root:Average loss=0.044140950351953506 +INFO:root: +[Valid][19][UTT][ MSE = 0.1868 | LCC = 0.9601 | SRCC = 0.9598 ] +INFO:root:Average loss=0.044259860873222354 +INFO:root: +[Valid][20][UTT][ MSE = 0.1777 | LCC = 0.9611 | SRCC = 0.9608 ] +INFO:root:Average loss=0.042076033726334575 +INFO:root: +[Valid][21][UTT][ MSE = 0.1686 | LCC = 0.9607 | SRCC = 0.9606 ] +INFO:root:Average loss=0.04258312675356865 +INFO:root: +[Valid][22][UTT][ MSE = 0.1697 | LCC = 0.9604 | SRCC = 0.9603 ] +INFO:root:Average loss=0.039617718189954756 +INFO:root: +[Valid][23][UTT][ MSE = 0.1750 | LCC = 0.9591 | SRCC = 0.9587 ] +INFO:root:Average loss=0.035701968848705294 +INFO:root: +[Valid][24][UTT][ MSE = 0.1761 | LCC = 0.9589 | SRCC = 0.9579 ] +INFO:root:Average loss=0.035998524755239486 +INFO:root: +[Valid][25][UTT][ MSE = 0.1670 | LCC = 0.9602 | SRCC = 0.9595 ] +INFO:root:Average loss=0.03552182102203369 +INFO:root: +[Valid][26][UTT][ MSE = 0.1679 | LCC = 0.9603 | SRCC = 0.9597 ] +INFO:root:Average loss=0.03434444235265255 +INFO:root: +[Valid][27][UTT][ MSE = 0.1599 | LCC = 0.9608 | SRCC = 0.9603 ] +INFO:root:Average loss=0.032505589008331295 +INFO:root: +[Valid][28][UTT][ MSE = 0.1706 | LCC = 0.9607 | SRCC = 0.9603 ] +INFO:root:Average loss=0.03063434699177742 +INFO:root: +[Valid][29][UTT][ MSE = 0.1585 | LCC = 0.9598 | SRCC = 0.9598 ] +INFO:root:Average loss=0.03008310776948929 +INFO:root: +[Valid][30][UTT][ MSE = 0.1635 | LCC = 0.9604 | SRCC = 0.9601 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1770 | LCC = 0.9628 | SRCC = 0.9633 ] diff --git a/results_tencent/w2v2_base/layer3/run4/config.gin b/results_tencent/w2v2_base/layer3/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer3/run4/model.pt b/results_tencent/w2v2_base/layer3/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f17f1d30786eb15e15ddc139de8bba24c94788e2 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2b5439e0acf59dc9bed1413c896f6019bd3f180390757e3f68472d6cff95d362 +size 2747979 diff --git a/results_tencent/w2v2_base/layer3/run4/model_best.pt b/results_tencent/w2v2_base/layer3/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..70cf184899e15a088aefacb9fb2fa4c9f410ed15 --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b61fb4ff1b710d65bfa7ced7156afb1b84465712160da477eac15cff22663831 +size 2749638 diff --git a/results_tencent/w2v2_base/layer3/run4/train.log b/results_tencent/w2v2_base/layer3/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..36e63d6b2333320a533e3b235e959f6e43a36f7d --- /dev/null +++ b/results_tencent/w2v2_base/layer3/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5297788614034653 +INFO:root: +[Valid][0][UTT][ MSE = 0.2841 | LCC = 0.9246 | SRCC = 0.9253 ] +INFO:root:Average loss=0.1778841906785965 +INFO:root: +[Valid][1][UTT][ MSE = 0.2358 | LCC = 0.9396 | SRCC = 0.9404 ] +INFO:root:Average loss=0.1435697311758995 +INFO:root: +[Valid][2][UTT][ MSE = 0.2554 | LCC = 0.9476 | SRCC = 0.9485 ] +INFO:root:Average loss=0.12729280978441238 +INFO:root: +[Valid][3][UTT][ MSE = 0.2397 | LCC = 0.9527 | SRCC = 0.9527 ] +INFO:root:Average loss=0.11133299908041953 +INFO:root: +[Valid][4][UTT][ MSE = 0.2062 | LCC = 0.9546 | SRCC = 0.9547 ] +INFO:root:Average loss=0.10197651755809783 +INFO:root: +[Valid][5][UTT][ MSE = 0.2173 | LCC = 0.9565 | SRCC = 0.9568 ] +INFO:root:Average loss=0.09267050126194953 +INFO:root: +[Valid][6][UTT][ MSE = 0.1981 | LCC = 0.9555 | SRCC = 0.9552 ] +INFO:root:Average loss=0.0890592525601387 +INFO:root: +[Valid][7][UTT][ MSE = 0.2085 | LCC = 0.9587 | SRCC = 0.9590 ] +INFO:root:Average loss=0.080743253916502 +INFO:root: +[Valid][8][UTT][ MSE = 0.1986 | LCC = 0.9583 | SRCC = 0.9584 ] +INFO:root:Average loss=0.07664622649550439 +INFO:root: +[Valid][9][UTT][ MSE = 0.2185 | LCC = 0.9586 | SRCC = 0.9594 ] +INFO:root:Average loss=0.07088005435466767 +INFO:root: +[Valid][10][UTT][ MSE = 0.1899 | LCC = 0.9587 | SRCC = 0.9585 ] +INFO:root:Average loss=0.06642591094970703 +INFO:root: +[Valid][11][UTT][ MSE = 0.2005 | LCC = 0.9596 | SRCC = 0.9594 ] +INFO:root:Average loss=0.0662348998785019 +INFO:root: +[Valid][12][UTT][ MSE = 0.2148 | LCC = 0.9609 | SRCC = 0.9612 ] +INFO:root:Average loss=0.05985616348683834 +INFO:root: +[Valid][13][UTT][ MSE = 0.2135 | LCC = 0.9603 | SRCC = 0.9606 ] +INFO:root:Average loss=0.06072604063153267 +INFO:root: +[Valid][14][UTT][ MSE = 0.2106 | LCC = 0.9612 | SRCC = 0.9615 ] +INFO:root:Average loss=0.05400150594115257 +INFO:root: +[Valid][15][UTT][ MSE = 0.1886 | LCC = 0.9600 | SRCC = 0.9598 ] +INFO:root:Average loss=0.05104274997115135 +INFO:root: +[Valid][16][UTT][ MSE = 0.1886 | LCC = 0.9609 | SRCC = 0.9610 ] +INFO:root:Average loss=0.048573012962937354 +INFO:root: +[Valid][17][UTT][ MSE = 0.2052 | LCC = 0.9607 | SRCC = 0.9614 ] +INFO:root:Average loss=0.04822887933254242 +INFO:root: +[Valid][18][UTT][ MSE = 0.2089 | LCC = 0.9612 | SRCC = 0.9612 ] +INFO:root:Average loss=0.046430713891983035 +INFO:root: +[Valid][19][UTT][ MSE = 0.2219 | LCC = 0.9616 | SRCC = 0.9619 ] +INFO:root:Average loss=0.04207477368414402 +INFO:root: +[Valid][20][UTT][ MSE = 0.1943 | LCC = 0.9622 | SRCC = 0.9622 ] +INFO:root:Average loss=0.0419696082174778 +INFO:root: +[Valid][21][UTT][ MSE = 0.2013 | LCC = 0.9613 | SRCC = 0.9613 ] +INFO:root:Average loss=0.040305506885051724 +INFO:root: +[Valid][22][UTT][ MSE = 0.2061 | LCC = 0.9617 | SRCC = 0.9620 ] +INFO:root:Average loss=0.0401180704087019 +INFO:root: +[Valid][23][UTT][ MSE = 0.1791 | LCC = 0.9623 | SRCC = 0.9623 ] +INFO:root:Average loss=0.03615359164774418 +INFO:root: +[Valid][24][UTT][ MSE = 0.1827 | LCC = 0.9589 | SRCC = 0.9591 ] +INFO:root:Average loss=0.037566984355449674 +INFO:root: +[Valid][25][UTT][ MSE = 0.1991 | LCC = 0.9621 | SRCC = 0.9620 ] +INFO:root:Average loss=0.0351417605727911 +INFO:root: +[Valid][26][UTT][ MSE = 0.1889 | LCC = 0.9614 | SRCC = 0.9618 ] +INFO:root:Average loss=0.03239202770590782 +INFO:root: +[Valid][27][UTT][ MSE = 0.1934 | LCC = 0.9619 | SRCC = 0.9622 ] +INFO:root:Average loss=0.03085931387543678 +INFO:root: +[Valid][28][UTT][ MSE = 0.2084 | LCC = 0.9623 | SRCC = 0.9623 ] +INFO:root:Average loss=0.03136058874428272 +INFO:root: +[Valid][29][UTT][ MSE = 0.2152 | LCC = 0.9621 | SRCC = 0.9623 ] +INFO:root:Average loss=0.0310584734082222 +INFO:root: +[Valid][30][UTT][ MSE = 0.1951 | LCC = 0.9607 | SRCC = 0.9605 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2170 | LCC = 0.9624 | SRCC = 0.9632 ] diff --git a/results_tencent/w2v2_base/layer4/run0/config.gin b/results_tencent/w2v2_base/layer4/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer4/run0/model.pt b/results_tencent/w2v2_base/layer4/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8f002fb131d0a7dd7d4bd8591798a935012dd9f0 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8d44cc5b40eac5061d4d1a88b03e32e682575352c9f352c6a3d259f7c0581685 +size 2747979 diff --git a/results_tencent/w2v2_base/layer4/run0/model_best.pt b/results_tencent/w2v2_base/layer4/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..794639d4adfcd446d5b6b71e26d03434f02fb372 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9d00315dd85e10a1149602258eb441a958cdf0279500a87032bfd96a5af22c7c +size 2749702 diff --git a/results_tencent/w2v2_base/layer4/run0/train.log b/results_tencent/w2v2_base/layer4/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..901eb31eba40831018761c6e50456b825aa5b833 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44398763358592985 +INFO:root: +[Valid][0][UTT][ MSE = 0.2908 | LCC = 0.9263 | SRCC = 0.9279 ] +INFO:root:Average loss=0.1795945090651512 +INFO:root: +[Valid][1][UTT][ MSE = 0.2547 | LCC = 0.9427 | SRCC = 0.9441 ] +INFO:root:Average loss=0.1448367499113083 +INFO:root: +[Valid][2][UTT][ MSE = 0.2220 | LCC = 0.9492 | SRCC = 0.9501 ] +INFO:root:Average loss=0.12515450644493104 +INFO:root: +[Valid][3][UTT][ MSE = 0.2151 | LCC = 0.9513 | SRCC = 0.9513 ] +INFO:root:Average loss=0.11439409095048904 +INFO:root: +[Valid][4][UTT][ MSE = 0.2202 | LCC = 0.9530 | SRCC = 0.9532 ] +INFO:root:Average loss=0.10433179646730423 +INFO:root: +[Valid][5][UTT][ MSE = 0.2131 | LCC = 0.9552 | SRCC = 0.9563 ] +INFO:root:Average loss=0.09305171501636506 +INFO:root: +[Valid][6][UTT][ MSE = 0.1934 | LCC = 0.9562 | SRCC = 0.9557 ] +INFO:root:Average loss=0.08599352714419364 +INFO:root: +[Valid][7][UTT][ MSE = 0.2045 | LCC = 0.9565 | SRCC = 0.9562 ] +INFO:root:Average loss=0.0802842719256878 +INFO:root: +[Valid][8][UTT][ MSE = 0.1915 | LCC = 0.9560 | SRCC = 0.9560 ] +INFO:root:Average loss=0.0741444533765316 +INFO:root: +[Valid][9][UTT][ MSE = 0.1853 | LCC = 0.9586 | SRCC = 0.9589 ] +INFO:root:Average loss=0.07241530653834342 +INFO:root: +[Valid][10][UTT][ MSE = 0.1865 | LCC = 0.9582 | SRCC = 0.9579 ] +INFO:root:Average loss=0.06547068512439727 +INFO:root: +[Valid][11][UTT][ MSE = 0.1860 | LCC = 0.9586 | SRCC = 0.9588 ] +INFO:root:Average loss=0.06150847414135933 +INFO:root: +[Valid][12][UTT][ MSE = 0.1818 | LCC = 0.9577 | SRCC = 0.9581 ] +INFO:root:Average loss=0.05668385031819344 +INFO:root: +[Valid][13][UTT][ MSE = 0.1935 | LCC = 0.9587 | SRCC = 0.9581 ] +INFO:root:Average loss=0.052272444665431975 +INFO:root: +[Valid][14][UTT][ MSE = 0.1795 | LCC = 0.9585 | SRCC = 0.9584 ] +INFO:root:Average loss=0.05182192173600197 +INFO:root: +[Valid][15][UTT][ MSE = 0.1825 | LCC = 0.9599 | SRCC = 0.9594 ] +INFO:root:Average loss=0.04899405887722969 +INFO:root: +[Valid][16][UTT][ MSE = 0.1741 | LCC = 0.9600 | SRCC = 0.9598 ] +INFO:root:Average loss=0.045754702791571615 +INFO:root: +[Valid][17][UTT][ MSE = 0.1844 | LCC = 0.9600 | SRCC = 0.9605 ] +INFO:root:Average loss=0.044034603759646414 +INFO:root: +[Valid][18][UTT][ MSE = 0.1839 | LCC = 0.9598 | SRCC = 0.9601 ] +INFO:root:Average loss=0.04272802817821503 +INFO:root: +[Valid][19][UTT][ MSE = 0.1878 | LCC = 0.9595 | SRCC = 0.9587 ] +INFO:root:Average loss=0.039388107925653455 +INFO:root: +[Valid][20][UTT][ MSE = 0.1948 | LCC = 0.9592 | SRCC = 0.9593 ] +INFO:root:Average loss=0.037795317858457564 +INFO:root: +[Valid][21][UTT][ MSE = 0.1948 | LCC = 0.9589 | SRCC = 0.9592 ] +INFO:root:Average loss=0.03639946512877941 +INFO:root: +[Valid][22][UTT][ MSE = 0.1808 | LCC = 0.9596 | SRCC = 0.9595 ] +INFO:root:Average loss=0.034336156249046325 +INFO:root: +[Valid][23][UTT][ MSE = 0.1863 | LCC = 0.9604 | SRCC = 0.9601 ] +INFO:root:Average loss=0.033413119822740554 +INFO:root: +[Valid][24][UTT][ MSE = 0.1809 | LCC = 0.9598 | SRCC = 0.9596 ] +INFO:root:Average loss=0.03328752911090851 +INFO:root: +[Valid][25][UTT][ MSE = 0.1880 | LCC = 0.9597 | SRCC = 0.9589 ] +INFO:root:Average loss=0.031046655789017677 +INFO:root: +[Valid][26][UTT][ MSE = 0.1950 | LCC = 0.9593 | SRCC = 0.9594 ] +INFO:root:Average loss=0.02859292646497488 +INFO:root: +[Valid][27][UTT][ MSE = 0.1805 | LCC = 0.9608 | SRCC = 0.9607 ] +INFO:root:Average loss=0.027490422010421754 +INFO:root: +[Valid][28][UTT][ MSE = 0.1853 | LCC = 0.9595 | SRCC = 0.9588 ] +INFO:root:Average loss=0.02641232459247112 +INFO:root: +[Valid][29][UTT][ MSE = 0.1717 | LCC = 0.9606 | SRCC = 0.9601 ] +INFO:root:Average loss=0.0277088091224432 +INFO:root: +[Valid][30][UTT][ MSE = 0.1815 | LCC = 0.9610 | SRCC = 0.9602 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1882 | LCC = 0.9594 | SRCC = 0.9596 ] diff --git a/results_tencent/w2v2_base/layer4/run1/config.gin b/results_tencent/w2v2_base/layer4/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer4/run1/model.pt b/results_tencent/w2v2_base/layer4/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..26f5732aee35e5f6fbb47c3704fc962e34d9db24 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2a57812f5de6e56cdeceb010ba2b01c24e45fc7fdec3ce1e0584621d77890ce4 +size 2747979 diff --git a/results_tencent/w2v2_base/layer4/run1/model_best.pt b/results_tencent/w2v2_base/layer4/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e466d01b38836eb0f2361db26c829095669877e2 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:29764ae06c4115829f96c1df5671d4db2fd4ede6b59d984855e1c8a4cab152ed +size 2749702 diff --git a/results_tencent/w2v2_base/layer4/run1/train.log b/results_tencent/w2v2_base/layer4/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1e8908efb1aa73663f73c3b63e6cf3ca15fac1ca --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4333301765918732 +INFO:root: +[Valid][0][UTT][ MSE = 0.2567 | LCC = 0.9310 | SRCC = 0.9326 ] +INFO:root:Average loss=0.17783973556756974 +INFO:root: +[Valid][1][UTT][ MSE = 0.2206 | LCC = 0.9453 | SRCC = 0.9455 ] +INFO:root:Average loss=0.14344788032770156 +INFO:root: +[Valid][2][UTT][ MSE = 0.2401 | LCC = 0.9505 | SRCC = 0.9504 ] +INFO:root:Average loss=0.1312152786254883 +INFO:root: +[Valid][3][UTT][ MSE = 0.2095 | LCC = 0.9526 | SRCC = 0.9533 ] +INFO:root:Average loss=0.11333716666698455 +INFO:root: +[Valid][4][UTT][ MSE = 0.2193 | LCC = 0.9541 | SRCC = 0.9545 ] +INFO:root:Average loss=0.10626679879426956 +INFO:root: +[Valid][5][UTT][ MSE = 0.1946 | LCC = 0.9555 | SRCC = 0.9558 ] +INFO:root:Average loss=0.09939905497431754 +INFO:root: +[Valid][6][UTT][ MSE = 0.1901 | LCC = 0.9564 | SRCC = 0.9558 ] +INFO:root:Average loss=0.08503128832578659 +INFO:root: +[Valid][7][UTT][ MSE = 0.2178 | LCC = 0.9561 | SRCC = 0.9572 ] +INFO:root:Average loss=0.08137960490584374 +INFO:root: +[Valid][8][UTT][ MSE = 0.2125 | LCC = 0.9561 | SRCC = 0.9561 ] +INFO:root:Average loss=0.07317584881186485 +INFO:root: +[Valid][9][UTT][ MSE = 0.2006 | LCC = 0.9578 | SRCC = 0.9576 ] +INFO:root:Average loss=0.07086038911342621 +INFO:root: +[Valid][10][UTT][ MSE = 0.2120 | LCC = 0.9573 | SRCC = 0.9577 ] +INFO:root:Average loss=0.06648324227333069 +INFO:root: +[Valid][11][UTT][ MSE = 0.2146 | LCC = 0.9569 | SRCC = 0.9574 ] +INFO:root:Average loss=0.06315417632460595 +INFO:root: +[Valid][12][UTT][ MSE = 0.2026 | LCC = 0.9582 | SRCC = 0.9579 ] +INFO:root:Average loss=0.055746525973081586 +INFO:root: +[Valid][13][UTT][ MSE = 0.1972 | LCC = 0.9574 | SRCC = 0.9560 ] +INFO:root:Average loss=0.05227048508822918 +INFO:root: +[Valid][14][UTT][ MSE = 0.2017 | LCC = 0.9567 | SRCC = 0.9557 ] +INFO:root:Average loss=0.05052034328877926 +INFO:root: +[Valid][15][UTT][ MSE = 0.1849 | LCC = 0.9578 | SRCC = 0.9569 ] +INFO:root:Average loss=0.046346463188529016 +INFO:root: +[Valid][16][UTT][ MSE = 0.2110 | LCC = 0.9579 | SRCC = 0.9581 ] +INFO:root:Average loss=0.045148750379681586 +INFO:root: +[Valid][17][UTT][ MSE = 0.2068 | LCC = 0.9562 | SRCC = 0.9564 ] +INFO:root:Average loss=0.043039487451314924 +INFO:root: +[Valid][18][UTT][ MSE = 0.2084 | LCC = 0.9573 | SRCC = 0.9575 ] +INFO:root:Average loss=0.03939817348122597 +INFO:root: +[Valid][19][UTT][ MSE = 0.1950 | LCC = 0.9582 | SRCC = 0.9574 ] +INFO:root:Average loss=0.03914067390561104 +INFO:root: +[Valid][20][UTT][ MSE = 0.2101 | LCC = 0.9578 | SRCC = 0.9578 ] +INFO:root:Average loss=0.03629158698022365 +INFO:root: +[Valid][21][UTT][ MSE = 0.1985 | LCC = 0.9579 | SRCC = 0.9581 ] +INFO:root:Average loss=0.035033461049199105 +INFO:root: +[Valid][22][UTT][ MSE = 0.2035 | LCC = 0.9591 | SRCC = 0.9585 ] +INFO:root:Average loss=0.035568484902381896 +INFO:root: +[Valid][23][UTT][ MSE = 0.1912 | LCC = 0.9570 | SRCC = 0.9572 ] +INFO:root:Average loss=0.0334855684787035 +INFO:root: +[Valid][24][UTT][ MSE = 0.2072 | LCC = 0.9579 | SRCC = 0.9582 ] +INFO:root:Average loss=0.03195412167906761 +INFO:root: +[Valid][25][UTT][ MSE = 0.1995 | LCC = 0.9582 | SRCC = 0.9580 ] +INFO:root:Average loss=0.03032459209859371 +INFO:root: +[Valid][26][UTT][ MSE = 0.2103 | LCC = 0.9579 | SRCC = 0.9587 ] +INFO:root:Average loss=0.027878949880599975 +INFO:root: +[Valid][27][UTT][ MSE = 0.1915 | LCC = 0.9587 | SRCC = 0.9586 ] +INFO:root:Average loss=0.027846487894654274 +INFO:root: +[Valid][28][UTT][ MSE = 0.2009 | LCC = 0.9590 | SRCC = 0.9584 ] +INFO:root:Average loss=0.027260970771312714 +INFO:root: +[Valid][29][UTT][ MSE = 0.1851 | LCC = 0.9591 | SRCC = 0.9589 ] +INFO:root:Average loss=0.026887282386422157 +INFO:root: +[Valid][30][UTT][ MSE = 0.1895 | LCC = 0.9588 | SRCC = 0.9583 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2090 | LCC = 0.9578 | SRCC = 0.9590 ] diff --git a/results_tencent/w2v2_base/layer4/run2/config.gin b/results_tencent/w2v2_base/layer4/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer4/run2/model.pt b/results_tencent/w2v2_base/layer4/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0ba089dffe16a048eef8110a7bac102dea38fbee --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2bac7e53bd9d9900668339798e45d3e7bd8874c4e25e4a38b3984ec3e757c6de +size 2747979 diff --git a/results_tencent/w2v2_base/layer4/run2/model_best.pt b/results_tencent/w2v2_base/layer4/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b80576d2d3d40f665ea550da9694c0a315b7bed6 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:52015ae3d954d103a3a852dd5935a1f256a2ba5dcab5c181140669340b57f790 +size 2749702 diff --git a/results_tencent/w2v2_base/layer4/run2/train.log b/results_tencent/w2v2_base/layer4/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c7fae66dd696cf3d6710edf3f902e7eb1b3b7cd8 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.43584087216854095 +INFO:root: +[Valid][0][UTT][ MSE = 0.2599 | LCC = 0.9309 | SRCC = 0.9329 ] +INFO:root:Average loss=0.17968033581972123 +INFO:root: +[Valid][1][UTT][ MSE = 0.2509 | LCC = 0.9462 | SRCC = 0.9473 ] +INFO:root:Average loss=0.14478528076410294 +INFO:root: +[Valid][2][UTT][ MSE = 0.2256 | LCC = 0.9523 | SRCC = 0.9535 ] +INFO:root:Average loss=0.12234343630075455 +INFO:root: +[Valid][3][UTT][ MSE = 0.2153 | LCC = 0.9550 | SRCC = 0.9557 ] +INFO:root:Average loss=0.11198893374204635 +INFO:root: +[Valid][4][UTT][ MSE = 0.2198 | LCC = 0.9571 | SRCC = 0.9580 ] +INFO:root:Average loss=0.10161325815320014 +INFO:root: +[Valid][5][UTT][ MSE = 0.2150 | LCC = 0.9581 | SRCC = 0.9582 ] +INFO:root:Average loss=0.09832503813505172 +INFO:root: +[Valid][6][UTT][ MSE = 0.1767 | LCC = 0.9576 | SRCC = 0.9572 ] +INFO:root:Average loss=0.08653784576058388 +INFO:root: +[Valid][7][UTT][ MSE = 0.2419 | LCC = 0.9561 | SRCC = 0.9584 ] +INFO:root:Average loss=0.07747087210416793 +INFO:root: +[Valid][8][UTT][ MSE = 0.2006 | LCC = 0.9594 | SRCC = 0.9591 ] +INFO:root:Average loss=0.0739674321115017 +INFO:root: +[Valid][9][UTT][ MSE = 0.2020 | LCC = 0.9583 | SRCC = 0.9576 ] +INFO:root:Average loss=0.07097343099117279 +INFO:root: +[Valid][10][UTT][ MSE = 0.1944 | LCC = 0.9595 | SRCC = 0.9603 ] +INFO:root:Average loss=0.06194250482320786 +INFO:root: +[Valid][11][UTT][ MSE = 0.2009 | LCC = 0.9581 | SRCC = 0.9583 ] +INFO:root:Average loss=0.059822224855422976 +INFO:root: +[Valid][12][UTT][ MSE = 0.1909 | LCC = 0.9599 | SRCC = 0.9597 ] +INFO:root:Average loss=0.056567489564418795 +INFO:root: +[Valid][13][UTT][ MSE = 0.2055 | LCC = 0.9580 | SRCC = 0.9585 ] +INFO:root:Average loss=0.053915883272886274 +INFO:root: +[Valid][14][UTT][ MSE = 0.2104 | LCC = 0.9589 | SRCC = 0.9587 ] +INFO:root:Average loss=0.05354737243056297 +INFO:root: +[Valid][15][UTT][ MSE = 0.1965 | LCC = 0.9596 | SRCC = 0.9595 ] +INFO:root:Average loss=0.04551299597322941 +INFO:root: +[Valid][16][UTT][ MSE = 0.1873 | LCC = 0.9604 | SRCC = 0.9599 ] +INFO:root:Average loss=0.045054707482457164 +INFO:root: +[Valid][17][UTT][ MSE = 0.2108 | LCC = 0.9596 | SRCC = 0.9598 ] +INFO:root:Average loss=0.044418891310691835 +INFO:root: +[Valid][18][UTT][ MSE = 0.2223 | LCC = 0.9595 | SRCC = 0.9602 ] +INFO:root:Average loss=0.04146296185255051 +INFO:root: +[Valid][19][UTT][ MSE = 0.2326 | LCC = 0.9577 | SRCC = 0.9586 ] +INFO:root:Average loss=0.039357109010219576 +INFO:root: +[Valid][20][UTT][ MSE = 0.1912 | LCC = 0.9590 | SRCC = 0.9579 ] +INFO:root:Average loss=0.037435766965150837 +INFO:root: +[Valid][21][UTT][ MSE = 0.1957 | LCC = 0.9584 | SRCC = 0.9583 ] +INFO:root:Average loss=0.03505666510760784 +INFO:root: +[Valid][22][UTT][ MSE = 0.1955 | LCC = 0.9601 | SRCC = 0.9591 ] +INFO:root:Average loss=0.03396441902220249 +INFO:root: +[Valid][23][UTT][ MSE = 0.2162 | LCC = 0.9558 | SRCC = 0.9562 ] +INFO:root:Average loss=0.03292750287055969 +INFO:root: +[Valid][24][UTT][ MSE = 0.1794 | LCC = 0.9603 | SRCC = 0.9592 ] +INFO:root:Average loss=0.03210844051837921 +INFO:root: +[Valid][25][UTT][ MSE = 0.2185 | LCC = 0.9579 | SRCC = 0.9589 ] +INFO:root:Average loss=0.02958159427344799 +INFO:root: +[Valid][26][UTT][ MSE = 0.2080 | LCC = 0.9583 | SRCC = 0.9585 ] +INFO:root:Average loss=0.029069089233875276 +INFO:root: +[Valid][27][UTT][ MSE = 0.1907 | LCC = 0.9582 | SRCC = 0.9577 ] +INFO:root:Average loss=0.02738276480138302 +INFO:root: +[Valid][28][UTT][ MSE = 0.1940 | LCC = 0.9587 | SRCC = 0.9581 ] +INFO:root:Average loss=0.026595400243997574 +INFO:root: +[Valid][29][UTT][ MSE = 0.2316 | LCC = 0.9571 | SRCC = 0.9586 ] +INFO:root:Average loss=0.026522634416818618 +INFO:root: +[Valid][30][UTT][ MSE = 0.1997 | LCC = 0.9591 | SRCC = 0.9586 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1978 | LCC = 0.9596 | SRCC = 0.9601 ] diff --git a/results_tencent/w2v2_base/layer4/run3/config.gin b/results_tencent/w2v2_base/layer4/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer4/run3/model.pt b/results_tencent/w2v2_base/layer4/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..22519ba17557ae7e6cb818028b3e68ad32ba0c50 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a2b3d24071d5baf7d036c2ec207df0cb411eae18a96c6ad7f8a184e4f3b4af71 +size 2747979 diff --git a/results_tencent/w2v2_base/layer4/run3/model_best.pt b/results_tencent/w2v2_base/layer4/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..281a38bb1461788c50625c38d7715f444278ff42 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:972ab99f1c2026b15630d0715bfe8398b9f358c38a70f19632cd32a670cdbfec +size 2749638 diff --git a/results_tencent/w2v2_base/layer4/run3/train.log b/results_tencent/w2v2_base/layer4/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e84b9f1d99066ec68a82aba56f459a2178f4df0a --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3980644030570984 +INFO:root: +[Valid][0][UTT][ MSE = 0.2765 | LCC = 0.9297 | SRCC = 0.9292 ] +INFO:root:Average loss=0.16906231713294984 +INFO:root: +[Valid][1][UTT][ MSE = 0.2635 | LCC = 0.9450 | SRCC = 0.9465 ] +INFO:root:Average loss=0.1366833911538124 +INFO:root: +[Valid][2][UTT][ MSE = 0.2687 | LCC = 0.9499 | SRCC = 0.9511 ] +INFO:root:Average loss=0.123332426071167 +INFO:root: +[Valid][3][UTT][ MSE = 0.2716 | LCC = 0.9490 | SRCC = 0.9505 ] +INFO:root:Average loss=0.11036894410848617 +INFO:root: +[Valid][4][UTT][ MSE = 0.2358 | LCC = 0.9531 | SRCC = 0.9537 ] +INFO:root:Average loss=0.10111603137850761 +INFO:root: +[Valid][5][UTT][ MSE = 0.1991 | LCC = 0.9554 | SRCC = 0.9554 ] +INFO:root:Average loss=0.0926556601524353 +INFO:root: +[Valid][6][UTT][ MSE = 0.2010 | LCC = 0.9560 | SRCC = 0.9555 ] +INFO:root:Average loss=0.0837196162045002 +INFO:root: +[Valid][7][UTT][ MSE = 0.2140 | LCC = 0.9558 | SRCC = 0.9558 ] +INFO:root:Average loss=0.08094075626134872 +INFO:root: +[Valid][8][UTT][ MSE = 0.1891 | LCC = 0.9568 | SRCC = 0.9569 ] +INFO:root:Average loss=0.07236809846758842 +INFO:root: +[Valid][9][UTT][ MSE = 0.2013 | LCC = 0.9571 | SRCC = 0.9565 ] +INFO:root:Average loss=0.06604772561788559 +INFO:root: +[Valid][10][UTT][ MSE = 0.1925 | LCC = 0.9589 | SRCC = 0.9586 ] +INFO:root:Average loss=0.06498740062117576 +INFO:root: +[Valid][11][UTT][ MSE = 0.2067 | LCC = 0.9566 | SRCC = 0.9563 ] +INFO:root:Average loss=0.05690423193573952 +INFO:root: +[Valid][12][UTT][ MSE = 0.2301 | LCC = 0.9563 | SRCC = 0.9568 ] +INFO:root:Average loss=0.053891088485717775 +INFO:root: +[Valid][13][UTT][ MSE = 0.2135 | LCC = 0.9577 | SRCC = 0.9573 ] +INFO:root:Average loss=0.052665045514702795 +INFO:root: +[Valid][14][UTT][ MSE = 0.2033 | LCC = 0.9573 | SRCC = 0.9571 ] +INFO:root:Average loss=0.04896819831430912 +INFO:root: +[Valid][15][UTT][ MSE = 0.1950 | LCC = 0.9579 | SRCC = 0.9571 ] +INFO:root:Average loss=0.049505623430013655 +INFO:root: +[Valid][16][UTT][ MSE = 0.1869 | LCC = 0.9579 | SRCC = 0.9576 ] +INFO:root:Average loss=0.04526092340052128 +INFO:root: +[Valid][17][UTT][ MSE = 0.1929 | LCC = 0.9584 | SRCC = 0.9578 ] +INFO:root:Average loss=0.042413923740386966 +INFO:root: +[Valid][18][UTT][ MSE = 0.2133 | LCC = 0.9578 | SRCC = 0.9577 ] +INFO:root:Average loss=0.040653034567832945 +INFO:root: +[Valid][19][UTT][ MSE = 0.2032 | LCC = 0.9576 | SRCC = 0.9577 ] +INFO:root:Average loss=0.038730909913778304 +INFO:root: +[Valid][20][UTT][ MSE = 0.2095 | LCC = 0.9571 | SRCC = 0.9580 ] +INFO:root:Average loss=0.03644523656368256 +INFO:root: +[Valid][21][UTT][ MSE = 0.1953 | LCC = 0.9583 | SRCC = 0.9579 ] +INFO:root:Average loss=0.04019929300248623 +INFO:root: +[Valid][22][UTT][ MSE = 0.1942 | LCC = 0.9595 | SRCC = 0.9596 ] +INFO:root:Average loss=0.035748537465929986 +INFO:root: +[Valid][23][UTT][ MSE = 0.2046 | LCC = 0.9581 | SRCC = 0.9580 ] +INFO:root:Average loss=0.0338275157213211 +INFO:root: +[Valid][24][UTT][ MSE = 0.2120 | LCC = 0.9565 | SRCC = 0.9570 ] +INFO:root:Average loss=0.03145446452498436 +INFO:root: +[Valid][25][UTT][ MSE = 0.2045 | LCC = 0.9580 | SRCC = 0.9580 ] +INFO:root:Average loss=0.030427433118224143 +INFO:root: +[Valid][26][UTT][ MSE = 0.2248 | LCC = 0.9572 | SRCC = 0.9581 ] +INFO:root:Average loss=0.031806381866335866 +INFO:root: +[Valid][27][UTT][ MSE = 0.2229 | LCC = 0.9569 | SRCC = 0.9578 ] +INFO:root:Average loss=0.028356815844774248 +INFO:root: +[Valid][28][UTT][ MSE = 0.2036 | LCC = 0.9588 | SRCC = 0.9580 ] +INFO:root:Average loss=0.026959722518920898 +INFO:root: +[Valid][29][UTT][ MSE = 0.2167 | LCC = 0.9580 | SRCC = 0.9581 ] +INFO:root:Average loss=0.027020460329949855 +INFO:root: +[Valid][30][UTT][ MSE = 0.1986 | LCC = 0.9581 | SRCC = 0.9579 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2044 | LCC = 0.9596 | SRCC = 0.9602 ] diff --git a/results_tencent/w2v2_base/layer4/run4/config.gin b/results_tencent/w2v2_base/layer4/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer4/run4/model.pt b/results_tencent/w2v2_base/layer4/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d566af9e2bd99c102b6e41a980b2488ed5befc98 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5c3ef93b0f014d096934c5ce71a677a6ecbcbbe9fb3667860d8d8cf20ff8b7be +size 2747979 diff --git a/results_tencent/w2v2_base/layer4/run4/model_best.pt b/results_tencent/w2v2_base/layer4/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..11df2176780e76bfdb0ecd4bff0c37f9410de0e7 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:85a4f97b19a7aaf7705b43c89d3df266d0673f88e1de44b27d105b4bcc1eb78a +size 2749702 diff --git a/results_tencent/w2v2_base/layer4/run4/train.log b/results_tencent/w2v2_base/layer4/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3a5133a572ed1d9315bd2d954511d9b314e54c18 --- /dev/null +++ b/results_tencent/w2v2_base/layer4/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4106907190084457 +INFO:root: +[Valid][0][UTT][ MSE = 0.2920 | LCC = 0.9337 | SRCC = 0.9356 ] +INFO:root:Average loss=0.17627969139814376 +INFO:root: +[Valid][1][UTT][ MSE = 0.2380 | LCC = 0.9456 | SRCC = 0.9472 ] +INFO:root:Average loss=0.14279057830572128 +INFO:root: +[Valid][2][UTT][ MSE = 0.2375 | LCC = 0.9502 | SRCC = 0.9518 ] +INFO:root:Average loss=0.12179717004299163 +INFO:root: +[Valid][3][UTT][ MSE = 0.2215 | LCC = 0.9530 | SRCC = 0.9530 ] +INFO:root:Average loss=0.11081579875946045 +INFO:root: +[Valid][4][UTT][ MSE = 0.2328 | LCC = 0.9556 | SRCC = 0.9559 ] +INFO:root:Average loss=0.10054363912343979 +INFO:root: +[Valid][5][UTT][ MSE = 0.1782 | LCC = 0.9561 | SRCC = 0.9564 ] +INFO:root:Average loss=0.0920693496465683 +INFO:root: +[Valid][6][UTT][ MSE = 0.2038 | LCC = 0.9578 | SRCC = 0.9573 ] +INFO:root:Average loss=0.08308769768476486 +INFO:root: +[Valid][7][UTT][ MSE = 0.2075 | LCC = 0.9579 | SRCC = 0.9579 ] +INFO:root:Average loss=0.07683025762438773 +INFO:root: +[Valid][8][UTT][ MSE = 0.1773 | LCC = 0.9592 | SRCC = 0.9586 ] +INFO:root:Average loss=0.07507341706752776 +INFO:root: +[Valid][9][UTT][ MSE = 0.1698 | LCC = 0.9577 | SRCC = 0.9577 ] +INFO:root:Average loss=0.06873861548304558 +INFO:root: +[Valid][10][UTT][ MSE = 0.1765 | LCC = 0.9592 | SRCC = 0.9587 ] +INFO:root:Average loss=0.0630535401403904 +INFO:root: +[Valid][11][UTT][ MSE = 0.1844 | LCC = 0.9600 | SRCC = 0.9606 ] +INFO:root:Average loss=0.06061263653635979 +INFO:root: +[Valid][12][UTT][ MSE = 0.1775 | LCC = 0.9590 | SRCC = 0.9583 ] +INFO:root:Average loss=0.05842108257114887 +INFO:root: +[Valid][13][UTT][ MSE = 0.1924 | LCC = 0.9596 | SRCC = 0.9592 ] +INFO:root:Average loss=0.05198372937738895 +INFO:root: +[Valid][14][UTT][ MSE = 0.1753 | LCC = 0.9601 | SRCC = 0.9598 ] +INFO:root:Average loss=0.04939606675505638 +INFO:root: +[Valid][15][UTT][ MSE = 0.1896 | LCC = 0.9601 | SRCC = 0.9600 ] +INFO:root:Average loss=0.048591174229979515 +INFO:root: +[Valid][16][UTT][ MSE = 0.1856 | LCC = 0.9601 | SRCC = 0.9598 ] +INFO:root:Average loss=0.04434145724773407 +INFO:root: +[Valid][17][UTT][ MSE = 0.2045 | LCC = 0.9589 | SRCC = 0.9597 ] +INFO:root:Average loss=0.044526171281933785 +INFO:root: +[Valid][18][UTT][ MSE = 0.2122 | LCC = 0.9590 | SRCC = 0.9595 ] +INFO:root:Average loss=0.043359251230955125 +INFO:root: +[Valid][19][UTT][ MSE = 0.1850 | LCC = 0.9607 | SRCC = 0.9598 ] +INFO:root:Average loss=0.03968668858706951 +INFO:root: +[Valid][20][UTT][ MSE = 0.1772 | LCC = 0.9613 | SRCC = 0.9610 ] +INFO:root:Average loss=0.03763818971812725 +INFO:root: +[Valid][21][UTT][ MSE = 0.1852 | LCC = 0.9612 | SRCC = 0.9607 ] +INFO:root:Average loss=0.035656862899661065 +INFO:root: +[Valid][22][UTT][ MSE = 0.1852 | LCC = 0.9611 | SRCC = 0.9608 ] +INFO:root:Average loss=0.03335111139714718 +INFO:root: +[Valid][23][UTT][ MSE = 0.1962 | LCC = 0.9605 | SRCC = 0.9613 ] +INFO:root:Average loss=0.03389063455164432 +INFO:root: +[Valid][24][UTT][ MSE = 0.1997 | LCC = 0.9611 | SRCC = 0.9616 ] +INFO:root:Average loss=0.032967375367879866 +INFO:root: +[Valid][25][UTT][ MSE = 0.1920 | LCC = 0.9613 | SRCC = 0.9612 ] +INFO:root:Average loss=0.02997587387263775 +INFO:root: +[Valid][26][UTT][ MSE = 0.1848 | LCC = 0.9620 | SRCC = 0.9617 ] +INFO:root:Average loss=0.028613951951265336 +INFO:root: +[Valid][27][UTT][ MSE = 0.1823 | LCC = 0.9607 | SRCC = 0.9602 ] +INFO:root:Average loss=0.0285871951431036 +INFO:root: +[Valid][28][UTT][ MSE = 0.1829 | LCC = 0.9610 | SRCC = 0.9612 ] +INFO:root:Average loss=0.029168396547436713 +INFO:root: +[Valid][29][UTT][ MSE = 0.1985 | LCC = 0.9609 | SRCC = 0.9617 ] +INFO:root:Average loss=0.028176488995552064 +INFO:root: +[Valid][30][UTT][ MSE = 0.1702 | LCC = 0.9620 | SRCC = 0.9616 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1916 | LCC = 0.9587 | SRCC = 0.9593 ] diff --git a/results_tencent/w2v2_base/layer5/run0/config.gin b/results_tencent/w2v2_base/layer5/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer5/run0/model.pt b/results_tencent/w2v2_base/layer5/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4c5c2da8d47ef89c7550e7e2cf4f847f1aa15643 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e2c239c2993ff1a89d385fa9b5c78fabb312ba91795f72b1c87123f7edcf0d3b +size 2747979 diff --git a/results_tencent/w2v2_base/layer5/run0/model_best.pt b/results_tencent/w2v2_base/layer5/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..fe84e6562405da32caf81206a97dc240e5c61950 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b0e74df1c81462f65f5a0c58cc5bcc079a8e2eefb2ef941770534db15e6de4a7 +size 2749702 diff --git a/results_tencent/w2v2_base/layer5/run0/train.log b/results_tencent/w2v2_base/layer5/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2b3faa6560267f54dae67b4adb48a3557b36e184 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.41125085496902464 +INFO:root: +[Valid][0][UTT][ MSE = 0.2476 | LCC = 0.9281 | SRCC = 0.9289 ] +INFO:root:Average loss=0.18181199580430984 +INFO:root: +[Valid][1][UTT][ MSE = 0.2025 | LCC = 0.9419 | SRCC = 0.9437 ] +INFO:root:Average loss=0.1466971235871315 +INFO:root: +[Valid][2][UTT][ MSE = 0.2236 | LCC = 0.9486 | SRCC = 0.9501 ] +INFO:root:Average loss=0.12683400028944017 +INFO:root: +[Valid][3][UTT][ MSE = 0.2039 | LCC = 0.9507 | SRCC = 0.9514 ] +INFO:root:Average loss=0.11624941855669021 +INFO:root: +[Valid][4][UTT][ MSE = 0.2082 | LCC = 0.9528 | SRCC = 0.9536 ] +INFO:root:Average loss=0.10504457160830498 +INFO:root: +[Valid][5][UTT][ MSE = 0.2186 | LCC = 0.9521 | SRCC = 0.9535 ] +INFO:root:Average loss=0.09724783420562744 +INFO:root: +[Valid][6][UTT][ MSE = 0.1992 | LCC = 0.9533 | SRCC = 0.9535 ] +INFO:root:Average loss=0.09186559733748435 +INFO:root: +[Valid][7][UTT][ MSE = 0.1918 | LCC = 0.9545 | SRCC = 0.9553 ] +INFO:root:Average loss=0.08156689924001694 +INFO:root: +[Valid][8][UTT][ MSE = 0.1971 | LCC = 0.9541 | SRCC = 0.9550 ] +INFO:root:Average loss=0.07898528549075126 +INFO:root: +[Valid][9][UTT][ MSE = 0.1959 | LCC = 0.9550 | SRCC = 0.9549 ] +INFO:root:Average loss=0.0706961980164051 +INFO:root: +[Valid][10][UTT][ MSE = 0.1893 | LCC = 0.9549 | SRCC = 0.9549 ] +INFO:root:Average loss=0.06440725269913673 +INFO:root: +[Valid][11][UTT][ MSE = 0.1748 | LCC = 0.9556 | SRCC = 0.9559 ] +INFO:root:Average loss=0.06060295736789703 +INFO:root: +[Valid][12][UTT][ MSE = 0.1781 | LCC = 0.9546 | SRCC = 0.9548 ] +INFO:root:Average loss=0.05682353144884109 +INFO:root: +[Valid][13][UTT][ MSE = 0.1716 | LCC = 0.9555 | SRCC = 0.9557 ] +INFO:root:Average loss=0.05389153614640236 +INFO:root: +[Valid][14][UTT][ MSE = 0.1742 | LCC = 0.9557 | SRCC = 0.9553 ] +INFO:root:Average loss=0.05057576304674149 +INFO:root: +[Valid][15][UTT][ MSE = 0.1804 | LCC = 0.9553 | SRCC = 0.9558 ] +INFO:root:Average loss=0.04906993654370308 +INFO:root: +[Valid][16][UTT][ MSE = 0.1956 | LCC = 0.9542 | SRCC = 0.9540 ] +INFO:root:Average loss=0.04419212420284748 +INFO:root: +[Valid][17][UTT][ MSE = 0.1999 | LCC = 0.9570 | SRCC = 0.9575 ] +INFO:root:Average loss=0.04472943688929081 +INFO:root: +[Valid][18][UTT][ MSE = 0.1719 | LCC = 0.9560 | SRCC = 0.9560 ] +INFO:root:Average loss=0.041492516040802 +INFO:root: +[Valid][19][UTT][ MSE = 0.1833 | LCC = 0.9573 | SRCC = 0.9576 ] +INFO:root:Average loss=0.03969432026147843 +INFO:root: +[Valid][20][UTT][ MSE = 0.1869 | LCC = 0.9559 | SRCC = 0.9555 ] +INFO:root:Average loss=0.03946386666595936 +INFO:root: +[Valid][21][UTT][ MSE = 0.1860 | LCC = 0.9546 | SRCC = 0.9546 ] +INFO:root:Average loss=0.035930150553584096 +INFO:root: +[Valid][22][UTT][ MSE = 0.1721 | LCC = 0.9568 | SRCC = 0.9570 ] +INFO:root:Average loss=0.03500489833950996 +INFO:root: +[Valid][23][UTT][ MSE = 0.2016 | LCC = 0.9560 | SRCC = 0.9561 ] +INFO:root:Average loss=0.03417266200482845 +INFO:root: +[Valid][24][UTT][ MSE = 0.1780 | LCC = 0.9586 | SRCC = 0.9582 ] +INFO:root:Average loss=0.03212958846986294 +INFO:root: +[Valid][25][UTT][ MSE = 0.1633 | LCC = 0.9577 | SRCC = 0.9577 ] +INFO:root:Average loss=0.03197493596374989 +INFO:root: +[Valid][26][UTT][ MSE = 0.1834 | LCC = 0.9576 | SRCC = 0.9578 ] +INFO:root:Average loss=0.030046206310391427 +INFO:root: +[Valid][27][UTT][ MSE = 0.1933 | LCC = 0.9584 | SRCC = 0.9581 ] +INFO:root:Average loss=0.03261821237206459 +INFO:root: +[Valid][28][UTT][ MSE = 0.1860 | LCC = 0.9581 | SRCC = 0.9579 ] +INFO:root:Average loss=0.027694196984171866 +INFO:root: +[Valid][29][UTT][ MSE = 0.1764 | LCC = 0.9574 | SRCC = 0.9572 ] +INFO:root:Average loss=0.027325836181640625 +INFO:root: +[Valid][30][UTT][ MSE = 0.1775 | LCC = 0.9575 | SRCC = 0.9579 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1840 | LCC = 0.9561 | SRCC = 0.9568 ] diff --git a/results_tencent/w2v2_base/layer5/run1/config.gin b/results_tencent/w2v2_base/layer5/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer5/run1/model.pt b/results_tencent/w2v2_base/layer5/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..fd195798831d6c5298f8a0d73685f60949667e46 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a18a2f27b898268f5f1a09ac2f859eb10762d8fc6f1467fa1b127bc6d5d22a3 +size 2747979 diff --git a/results_tencent/w2v2_base/layer5/run1/model_best.pt b/results_tencent/w2v2_base/layer5/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..8aee8412b702b9367c4cf78dd0edaa518dec2648 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ef38e5bf8f88efe5875849b72cb11b582d75797187b3b07ddbc3a6a53fc1893e +size 2749702 diff --git a/results_tencent/w2v2_base/layer5/run1/train.log b/results_tencent/w2v2_base/layer5/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..fc577748c3205cd86ba0d486298c2b2c9ca0dd74 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.47719403624534606 +INFO:root: +[Valid][0][UTT][ MSE = 0.2621 | LCC = 0.9284 | SRCC = 0.9309 ] +INFO:root:Average loss=0.17938822597265244 +INFO:root: +[Valid][1][UTT][ MSE = 0.2772 | LCC = 0.9418 | SRCC = 0.9439 ] +INFO:root:Average loss=0.146253966152668 +INFO:root: +[Valid][2][UTT][ MSE = 0.2439 | LCC = 0.9451 | SRCC = 0.9460 ] +INFO:root:Average loss=0.12563653188943863 +INFO:root: +[Valid][3][UTT][ MSE = 0.2166 | LCC = 0.9490 | SRCC = 0.9492 ] +INFO:root:Average loss=0.11242063981294632 +INFO:root: +[Valid][4][UTT][ MSE = 0.2147 | LCC = 0.9525 | SRCC = 0.9540 ] +INFO:root:Average loss=0.1040298863351345 +INFO:root: +[Valid][5][UTT][ MSE = 0.2201 | LCC = 0.9519 | SRCC = 0.9527 ] +INFO:root:Average loss=0.09326359882950783 +INFO:root: +[Valid][6][UTT][ MSE = 0.2214 | LCC = 0.9530 | SRCC = 0.9541 ] +INFO:root:Average loss=0.08865910887718201 +INFO:root: +[Valid][7][UTT][ MSE = 0.2379 | LCC = 0.9479 | SRCC = 0.9490 ] +INFO:root:Average loss=0.08274820470809936 +INFO:root: +[Valid][8][UTT][ MSE = 0.2233 | LCC = 0.9541 | SRCC = 0.9551 ] +INFO:root:Average loss=0.0733099988102913 +INFO:root: +[Valid][9][UTT][ MSE = 0.2158 | LCC = 0.9543 | SRCC = 0.9550 ] +INFO:root:Average loss=0.07033597657084464 +INFO:root: +[Valid][10][UTT][ MSE = 0.2152 | LCC = 0.9540 | SRCC = 0.9538 ] +INFO:root:Average loss=0.06458013245463372 +INFO:root: +[Valid][11][UTT][ MSE = 0.1966 | LCC = 0.9545 | SRCC = 0.9550 ] +INFO:root:Average loss=0.06276470246911049 +INFO:root: +[Valid][12][UTT][ MSE = 0.2059 | LCC = 0.9554 | SRCC = 0.9550 ] +INFO:root:Average loss=0.055358204558491705 +INFO:root: +[Valid][13][UTT][ MSE = 0.2145 | LCC = 0.9537 | SRCC = 0.9540 ] +INFO:root:Average loss=0.0532461750805378 +INFO:root: +[Valid][14][UTT][ MSE = 0.1927 | LCC = 0.9522 | SRCC = 0.9523 ] +INFO:root:Average loss=0.049486609742045405 +INFO:root: +[Valid][15][UTT][ MSE = 0.2267 | LCC = 0.9542 | SRCC = 0.9544 ] +INFO:root:Average loss=0.046633185118436817 +INFO:root: +[Valid][16][UTT][ MSE = 0.2085 | LCC = 0.9556 | SRCC = 0.9556 ] +INFO:root:Average loss=0.046611439645290374 +INFO:root: +[Valid][17][UTT][ MSE = 0.1982 | LCC = 0.9535 | SRCC = 0.9533 ] +INFO:root:Average loss=0.04152176697552204 +INFO:root: +[Valid][18][UTT][ MSE = 0.2036 | LCC = 0.9538 | SRCC = 0.9538 ] +INFO:root:Average loss=0.04183908239006996 +INFO:root: +[Valid][19][UTT][ MSE = 0.2108 | LCC = 0.9526 | SRCC = 0.9527 ] +INFO:root:Average loss=0.039678726881742475 +INFO:root: +[Valid][20][UTT][ MSE = 0.1965 | LCC = 0.9557 | SRCC = 0.9557 ] +INFO:root:Average loss=0.03735539293289185 +INFO:root: +[Valid][21][UTT][ MSE = 0.1988 | LCC = 0.9543 | SRCC = 0.9540 ] +INFO:root:Average loss=0.03860775329172611 +INFO:root: +[Valid][22][UTT][ MSE = 0.2097 | LCC = 0.9526 | SRCC = 0.9524 ] +INFO:root:Average loss=0.034629773452878 +INFO:root: +[Valid][23][UTT][ MSE = 0.1843 | LCC = 0.9534 | SRCC = 0.9532 ] +INFO:root:Average loss=0.03444469772279263 +INFO:root: +[Valid][24][UTT][ MSE = 0.2042 | LCC = 0.9558 | SRCC = 0.9556 ] +INFO:root:Average loss=0.03121941705048084 +INFO:root: +[Valid][25][UTT][ MSE = 0.1966 | LCC = 0.9553 | SRCC = 0.9553 ] +INFO:root:Average loss=0.03042904157936573 +INFO:root: +[Valid][26][UTT][ MSE = 0.2209 | LCC = 0.9553 | SRCC = 0.9554 ] +INFO:root:Average loss=0.029986432507634164 +INFO:root: +[Valid][27][UTT][ MSE = 0.2055 | LCC = 0.9534 | SRCC = 0.9531 ] +INFO:root:Average loss=0.02825859074294567 +INFO:root: +[Valid][28][UTT][ MSE = 0.2027 | LCC = 0.9539 | SRCC = 0.9541 ] +INFO:root:Average loss=0.029834769666194916 +INFO:root: +[Valid][29][UTT][ MSE = 0.2066 | LCC = 0.9559 | SRCC = 0.9563 ] +INFO:root:Average loss=0.02610857004672289 +INFO:root: +[Valid][30][UTT][ MSE = 0.1924 | LCC = 0.9550 | SRCC = 0.9546 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2172 | LCC = 0.9557 | SRCC = 0.9569 ] diff --git a/results_tencent/w2v2_base/layer5/run2/config.gin b/results_tencent/w2v2_base/layer5/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer5/run2/model.pt b/results_tencent/w2v2_base/layer5/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c576e660c6db5f0a90998543e1f65e9a06453f5d --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a578e79b0fad2551b98e08c354ebd140dbe06614e585548fd91c7f1e10345d73 +size 2747979 diff --git a/results_tencent/w2v2_base/layer5/run2/model_best.pt b/results_tencent/w2v2_base/layer5/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..bacf98f67a028d6b8c6a174ca47d8f70cb0695e6 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7121a99478cb9540881fa09e1cb01f3802b3ca2d5d19f1ee82241bd6ee8d9471 +size 2749702 diff --git a/results_tencent/w2v2_base/layer5/run2/train.log b/results_tencent/w2v2_base/layer5/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..562ce66d777ca407a6b9ea3e7855e1eb1bfebe34 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6080906347036362 +INFO:root: +[Valid][0][UTT][ MSE = 0.2772 | LCC = 0.9288 | SRCC = 0.9313 ] +INFO:root:Average loss=0.1781305087208748 +INFO:root: +[Valid][1][UTT][ MSE = 0.2123 | LCC = 0.9463 | SRCC = 0.9483 ] +INFO:root:Average loss=0.13992599487304688 +INFO:root: +[Valid][2][UTT][ MSE = 0.2140 | LCC = 0.9501 | SRCC = 0.9520 ] +INFO:root:Average loss=0.12519819208979607 +INFO:root: +[Valid][3][UTT][ MSE = 0.1999 | LCC = 0.9528 | SRCC = 0.9542 ] +INFO:root:Average loss=0.10941404241323471 +INFO:root: +[Valid][4][UTT][ MSE = 0.2079 | LCC = 0.9507 | SRCC = 0.9526 ] +INFO:root:Average loss=0.09915468609333038 +INFO:root: +[Valid][5][UTT][ MSE = 0.1900 | LCC = 0.9551 | SRCC = 0.9558 ] +INFO:root:Average loss=0.08919636651873589 +INFO:root: +[Valid][6][UTT][ MSE = 0.1972 | LCC = 0.9528 | SRCC = 0.9545 ] +INFO:root:Average loss=0.08147384747862815 +INFO:root: +[Valid][7][UTT][ MSE = 0.1876 | LCC = 0.9546 | SRCC = 0.9559 ] +INFO:root:Average loss=0.07457343348860741 +INFO:root: +[Valid][8][UTT][ MSE = 0.2274 | LCC = 0.9529 | SRCC = 0.9539 ] +INFO:root:Average loss=0.0695228773355484 +INFO:root: +[Valid][9][UTT][ MSE = 0.1851 | LCC = 0.9551 | SRCC = 0.9557 ] +INFO:root:Average loss=0.06426863959431649 +INFO:root: +[Valid][10][UTT][ MSE = 0.2169 | LCC = 0.9532 | SRCC = 0.9536 ] +INFO:root:Average loss=0.060570698082447055 +INFO:root: +[Valid][11][UTT][ MSE = 0.1816 | LCC = 0.9555 | SRCC = 0.9562 ] +INFO:root:Average loss=0.054283338725566864 +INFO:root: +[Valid][12][UTT][ MSE = 0.2101 | LCC = 0.9515 | SRCC = 0.9523 ] +INFO:root:Average loss=0.05283642356097698 +INFO:root: +[Valid][13][UTT][ MSE = 0.1918 | LCC = 0.9566 | SRCC = 0.9571 ] +INFO:root:Average loss=0.0491203960031271 +INFO:root: +[Valid][14][UTT][ MSE = 0.1720 | LCC = 0.9552 | SRCC = 0.9556 ] +INFO:root:Average loss=0.045963629260659215 +INFO:root: +[Valid][15][UTT][ MSE = 0.1828 | LCC = 0.9549 | SRCC = 0.9566 ] +INFO:root:Average loss=0.04464930640161038 +INFO:root: +[Valid][16][UTT][ MSE = 0.1771 | LCC = 0.9555 | SRCC = 0.9562 ] +INFO:root:Average loss=0.04107583263516426 +INFO:root: +[Valid][17][UTT][ MSE = 0.1716 | LCC = 0.9554 | SRCC = 0.9557 ] +INFO:root:Average loss=0.03998103232681751 +INFO:root: +[Valid][18][UTT][ MSE = 0.1759 | LCC = 0.9567 | SRCC = 0.9574 ] +INFO:root:Average loss=0.03712421092391014 +INFO:root: +[Valid][19][UTT][ MSE = 0.1880 | LCC = 0.9557 | SRCC = 0.9562 ] +INFO:root:Average loss=0.03451367020606995 +INFO:root: +[Valid][20][UTT][ MSE = 0.2112 | LCC = 0.9530 | SRCC = 0.9541 ] +INFO:root:Average loss=0.03419963110983372 +INFO:root: +[Valid][21][UTT][ MSE = 0.1963 | LCC = 0.9561 | SRCC = 0.9558 ] +INFO:root:Average loss=0.03469160322844982 +INFO:root: +[Valid][22][UTT][ MSE = 0.1873 | LCC = 0.9547 | SRCC = 0.9549 ] +INFO:root:Average loss=0.03213440825045109 +INFO:root: +[Valid][23][UTT][ MSE = 0.2100 | LCC = 0.9518 | SRCC = 0.9538 ] +INFO:root:Average loss=0.03091093309223652 +INFO:root: +[Valid][24][UTT][ MSE = 0.1769 | LCC = 0.9556 | SRCC = 0.9559 ] +INFO:root:Average loss=0.028306058585643768 +INFO:root: +[Valid][25][UTT][ MSE = 0.1880 | LCC = 0.9555 | SRCC = 0.9567 ] +INFO:root:Average loss=0.02849062967300415 +INFO:root: +[Valid][26][UTT][ MSE = 0.1913 | LCC = 0.9552 | SRCC = 0.9568 ] +INFO:root:Average loss=0.027015159264206886 +INFO:root: +[Valid][27][UTT][ MSE = 0.1740 | LCC = 0.9560 | SRCC = 0.9561 ] +INFO:root:Average loss=0.025740687265992164 +INFO:root: +[Valid][28][UTT][ MSE = 0.1782 | LCC = 0.9563 | SRCC = 0.9561 ] +INFO:root:Average loss=0.02645353774726391 +INFO:root: +[Valid][29][UTT][ MSE = 0.1923 | LCC = 0.9545 | SRCC = 0.9553 ] +INFO:root:Average loss=0.024068560920655728 +INFO:root: +[Valid][30][UTT][ MSE = 0.1857 | LCC = 0.9561 | SRCC = 0.9571 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1775 | LCC = 0.9554 | SRCC = 0.9560 ] diff --git a/results_tencent/w2v2_base/layer5/run3/config.gin b/results_tencent/w2v2_base/layer5/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer5/run3/model.pt b/results_tencent/w2v2_base/layer5/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1548bf3493168ce95087bea9b599b54907c89e4c --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a2b0bd0ff3574db641242c7a2918a885438a01cffa5e3dd3bbb69ec2931fc085 +size 2747979 diff --git a/results_tencent/w2v2_base/layer5/run3/model_best.pt b/results_tencent/w2v2_base/layer5/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..81ec7e8ade51446f41d519a53b131dfcc81f5c52 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:db40dbd757f926f6a05da3f98db90ade2683c3aa153949ca35b2a1d5d5a5be55 +size 2749702 diff --git a/results_tencent/w2v2_base/layer5/run3/train.log b/results_tencent/w2v2_base/layer5/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3e04221d25a126178fc671a065f0b4d337bc276e --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.45256095921993256 +INFO:root: +[Valid][0][UTT][ MSE = 0.2780 | LCC = 0.9293 | SRCC = 0.9307 ] +INFO:root:Average loss=0.18117979919910432 +INFO:root: +[Valid][1][UTT][ MSE = 0.2330 | LCC = 0.9423 | SRCC = 0.9431 ] +INFO:root:Average loss=0.14618470883369447 +INFO:root: +[Valid][2][UTT][ MSE = 0.2375 | LCC = 0.9485 | SRCC = 0.9492 ] +INFO:root:Average loss=0.12841790670156478 +INFO:root: +[Valid][3][UTT][ MSE = 0.2334 | LCC = 0.9528 | SRCC = 0.9529 ] +INFO:root:Average loss=0.11076931166648865 +INFO:root: +[Valid][4][UTT][ MSE = 0.2099 | LCC = 0.9545 | SRCC = 0.9549 ] +INFO:root:Average loss=0.10039441868662834 +INFO:root: +[Valid][5][UTT][ MSE = 0.2277 | LCC = 0.9547 | SRCC = 0.9550 ] +INFO:root:Average loss=0.0960552702844143 +INFO:root: +[Valid][6][UTT][ MSE = 0.2114 | LCC = 0.9548 | SRCC = 0.9554 ] +INFO:root:Average loss=0.08963952684402465 +INFO:root: +[Valid][7][UTT][ MSE = 0.1899 | LCC = 0.9550 | SRCC = 0.9556 ] +INFO:root:Average loss=0.08083869385719299 +INFO:root: +[Valid][8][UTT][ MSE = 0.1916 | LCC = 0.9554 | SRCC = 0.9548 ] +INFO:root:Average loss=0.07376405650377274 +INFO:root: +[Valid][9][UTT][ MSE = 0.1964 | LCC = 0.9565 | SRCC = 0.9555 ] +INFO:root:Average loss=0.0688879260122776 +INFO:root: +[Valid][10][UTT][ MSE = 0.1905 | LCC = 0.9572 | SRCC = 0.9569 ] +INFO:root:Average loss=0.06313383972644807 +INFO:root: +[Valid][11][UTT][ MSE = 0.1830 | LCC = 0.9562 | SRCC = 0.9558 ] +INFO:root:Average loss=0.06120883822441101 +INFO:root: +[Valid][12][UTT][ MSE = 0.1902 | LCC = 0.9561 | SRCC = 0.9561 ] +INFO:root:Average loss=0.05664561685919762 +INFO:root: +[Valid][13][UTT][ MSE = 0.2066 | LCC = 0.9569 | SRCC = 0.9565 ] +INFO:root:Average loss=0.05430577111244202 +INFO:root: +[Valid][14][UTT][ MSE = 0.1814 | LCC = 0.9558 | SRCC = 0.9556 ] +INFO:root:Average loss=0.04824578599631786 +INFO:root: +[Valid][15][UTT][ MSE = 0.2036 | LCC = 0.9557 | SRCC = 0.9555 ] +INFO:root:Average loss=0.05025681120157242 +INFO:root: +[Valid][16][UTT][ MSE = 0.1879 | LCC = 0.9555 | SRCC = 0.9545 ] +INFO:root:Average loss=0.04360196682810783 +INFO:root: +[Valid][17][UTT][ MSE = 0.1952 | LCC = 0.9558 | SRCC = 0.9554 ] +INFO:root:Average loss=0.04360712516307831 +INFO:root: +[Valid][18][UTT][ MSE = 0.2042 | LCC = 0.9549 | SRCC = 0.9542 ] +INFO:root:Average loss=0.041914040192961695 +INFO:root: +[Valid][19][UTT][ MSE = 0.2050 | LCC = 0.9558 | SRCC = 0.9552 ] +INFO:root:Average loss=0.038886642739176754 +INFO:root: +[Valid][20][UTT][ MSE = 0.1981 | LCC = 0.9557 | SRCC = 0.9548 ] +INFO:root:Average loss=0.03694001625478267 +INFO:root: +[Valid][21][UTT][ MSE = 0.2030 | LCC = 0.9568 | SRCC = 0.9561 ] +INFO:root:Average loss=0.034886650934815404 +INFO:root: +[Valid][22][UTT][ MSE = 0.2139 | LCC = 0.9564 | SRCC = 0.9561 ] +INFO:root:Average loss=0.034440542832016946 +INFO:root: +[Valid][23][UTT][ MSE = 0.1986 | LCC = 0.9564 | SRCC = 0.9556 ] +INFO:root:Average loss=0.0317358725219965 +INFO:root: +[Valid][24][UTT][ MSE = 0.1944 | LCC = 0.9560 | SRCC = 0.9554 ] +INFO:root:Average loss=0.030387851625680924 +INFO:root: +[Valid][25][UTT][ MSE = 0.1921 | LCC = 0.9569 | SRCC = 0.9568 ] +INFO:root:Average loss=0.02952957034111023 +INFO:root: +[Valid][26][UTT][ MSE = 0.1837 | LCC = 0.9568 | SRCC = 0.9568 ] +INFO:root:Average loss=0.028762717515230178 +INFO:root: +[Valid][27][UTT][ MSE = 0.2018 | LCC = 0.9561 | SRCC = 0.9559 ] +INFO:root:Average loss=0.027249329373240472 +INFO:root: +[Valid][28][UTT][ MSE = 0.1912 | LCC = 0.9573 | SRCC = 0.9568 ] +INFO:root:Average loss=0.027297958627343178 +INFO:root: +[Valid][29][UTT][ MSE = 0.1972 | LCC = 0.9570 | SRCC = 0.9567 ] +INFO:root:Average loss=0.026269529938697814 +INFO:root: +[Valid][30][UTT][ MSE = 0.1795 | LCC = 0.9570 | SRCC = 0.9569 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2004 | LCC = 0.9557 | SRCC = 0.9560 ] diff --git a/results_tencent/w2v2_base/layer5/run4/config.gin b/results_tencent/w2v2_base/layer5/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer5/run4/model.pt b/results_tencent/w2v2_base/layer5/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7df58585f292d17f51e324ca10847a6fe5111ffb --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19264ad7c892fb6cfe2567a1088bfe413ec219bba0796a353399a574ef863f57 +size 2747979 diff --git a/results_tencent/w2v2_base/layer5/run4/model_best.pt b/results_tencent/w2v2_base/layer5/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..baf2b03c2413db0b6e3838ef2455552417f92172 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:02cf4de9bbd6968a7260a51d18deebd0560b19cfa50c4a7a1490d91b63ebb100 +size 2749638 diff --git a/results_tencent/w2v2_base/layer5/run4/train.log b/results_tencent/w2v2_base/layer5/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b9830bf2b36618be8a51984de66ee5dbf6ce0f38 --- /dev/null +++ b/results_tencent/w2v2_base/layer5/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.47299851083755495 +INFO:root: +[Valid][0][UTT][ MSE = 0.3118 | LCC = 0.9167 | SRCC = 0.9187 ] +INFO:root:Average loss=0.19005402171611785 +INFO:root: +[Valid][1][UTT][ MSE = 0.2519 | LCC = 0.9383 | SRCC = 0.9381 ] +INFO:root:Average loss=0.1572931528687477 +INFO:root: +[Valid][2][UTT][ MSE = 0.2387 | LCC = 0.9447 | SRCC = 0.9446 ] +INFO:root:Average loss=0.13248370212316513 +INFO:root: +[Valid][3][UTT][ MSE = 0.2567 | LCC = 0.9469 | SRCC = 0.9479 ] +INFO:root:Average loss=0.1195727113187313 +INFO:root: +[Valid][4][UTT][ MSE = 0.2127 | LCC = 0.9514 | SRCC = 0.9514 ] +INFO:root:Average loss=0.10745625805854797 +INFO:root: +[Valid][5][UTT][ MSE = 0.2033 | LCC = 0.9527 | SRCC = 0.9528 ] +INFO:root:Average loss=0.09958443766832352 +INFO:root: +[Valid][6][UTT][ MSE = 0.1881 | LCC = 0.9539 | SRCC = 0.9540 ] +INFO:root:Average loss=0.0893568762242794 +INFO:root: +[Valid][7][UTT][ MSE = 0.1988 | LCC = 0.9537 | SRCC = 0.9539 ] +INFO:root:Average loss=0.08432120504975318 +INFO:root: +[Valid][8][UTT][ MSE = 0.2036 | LCC = 0.9551 | SRCC = 0.9554 ] +INFO:root:Average loss=0.07735562825202942 +INFO:root: +[Valid][9][UTT][ MSE = 0.2173 | LCC = 0.9547 | SRCC = 0.9545 ] +INFO:root:Average loss=0.07179782313108445 +INFO:root: +[Valid][10][UTT][ MSE = 0.1992 | LCC = 0.9548 | SRCC = 0.9550 ] +INFO:root:Average loss=0.07041533002257347 +INFO:root: +[Valid][11][UTT][ MSE = 0.1833 | LCC = 0.9554 | SRCC = 0.9557 ] +INFO:root:Average loss=0.0636598884165287 +INFO:root: +[Valid][12][UTT][ MSE = 0.2026 | LCC = 0.9546 | SRCC = 0.9542 ] +INFO:root:Average loss=0.058074014097452165 +INFO:root: +[Valid][13][UTT][ MSE = 0.1818 | LCC = 0.9548 | SRCC = 0.9546 ] +INFO:root:Average loss=0.05729126305878162 +INFO:root: +[Valid][14][UTT][ MSE = 0.2116 | LCC = 0.9544 | SRCC = 0.9541 ] +INFO:root:Average loss=0.054557086557149886 +INFO:root: +[Valid][15][UTT][ MSE = 0.1888 | LCC = 0.9541 | SRCC = 0.9536 ] +INFO:root:Average loss=0.05060158434510231 +INFO:root: +[Valid][16][UTT][ MSE = 0.2015 | LCC = 0.9543 | SRCC = 0.9540 ] +INFO:root:Average loss=0.045160741329193115 +INFO:root: +[Valid][17][UTT][ MSE = 0.1933 | LCC = 0.9557 | SRCC = 0.9557 ] +INFO:root:Average loss=0.0437779576331377 +INFO:root: +[Valid][18][UTT][ MSE = 0.1722 | LCC = 0.9553 | SRCC = 0.9556 ] +INFO:root:Average loss=0.04244110947847366 +INFO:root: +[Valid][19][UTT][ MSE = 0.1863 | LCC = 0.9552 | SRCC = 0.9548 ] +INFO:root:Average loss=0.040198873713612554 +INFO:root: +[Valid][20][UTT][ MSE = 0.2001 | LCC = 0.9560 | SRCC = 0.9559 ] +INFO:root:Average loss=0.0384520613849163 +INFO:root: +[Valid][21][UTT][ MSE = 0.2026 | LCC = 0.9533 | SRCC = 0.9530 ] +INFO:root:Average loss=0.03862393806874752 +INFO:root: +[Valid][22][UTT][ MSE = 0.1878 | LCC = 0.9563 | SRCC = 0.9564 ] +INFO:root:Average loss=0.03395559033751488 +INFO:root: +[Valid][23][UTT][ MSE = 0.1827 | LCC = 0.9554 | SRCC = 0.9552 ] +INFO:root:Average loss=0.03401939733326435 +INFO:root: +[Valid][24][UTT][ MSE = 0.1910 | LCC = 0.9555 | SRCC = 0.9554 ] +INFO:root:Average loss=0.03266850678622723 +INFO:root: +[Valid][25][UTT][ MSE = 0.2030 | LCC = 0.9564 | SRCC = 0.9560 ] +INFO:root:Average loss=0.03322021059691906 +INFO:root: +[Valid][26][UTT][ MSE = 0.2022 | LCC = 0.9561 | SRCC = 0.9558 ] +INFO:root:Average loss=0.030895111918449403 +INFO:root: +[Valid][27][UTT][ MSE = 0.2014 | LCC = 0.9565 | SRCC = 0.9567 ] +INFO:root:Average loss=0.030161238744854928 +INFO:root: +[Valid][28][UTT][ MSE = 0.1857 | LCC = 0.9566 | SRCC = 0.9565 ] +INFO:root:Average loss=0.02914444998651743 +INFO:root: +[Valid][29][UTT][ MSE = 0.1852 | LCC = 0.9565 | SRCC = 0.9565 ] +INFO:root:Average loss=0.027649776250123978 +INFO:root: +[Valid][30][UTT][ MSE = 0.1924 | LCC = 0.9561 | SRCC = 0.9560 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1841 | LCC = 0.9582 | SRCC = 0.9587 ] diff --git a/results_tencent/w2v2_base/layer6/run0/config.gin b/results_tencent/w2v2_base/layer6/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer6/run0/model.pt b/results_tencent/w2v2_base/layer6/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9cdd7b6bf848c33a3574c1e7f5ff159e226c864a --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:237cf5dd23ae4b2de742313f5cbc5117d766083f89f285ee2d021b71c6601e78 +size 2747979 diff --git a/results_tencent/w2v2_base/layer6/run0/model_best.pt b/results_tencent/w2v2_base/layer6/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..eee0b6add9d7a53991b46436a631ec7579fa1487 --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fb583c93bfb50d4e65d3301dd3939e50a29cd536cb2c29dc9a252d6f10cc3fbd +size 2749702 diff --git a/results_tencent/w2v2_base/layer6/run0/train.log b/results_tencent/w2v2_base/layer6/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..bef93c6f004ba38f96a14d90919a3f6be8e0e76e --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4964496808052063 +INFO:root: +[Valid][0][UTT][ MSE = 0.2820 | LCC = 0.9246 | SRCC = 0.9271 ] +INFO:root:Average loss=0.18211031502485275 +INFO:root: +[Valid][1][UTT][ MSE = 0.2377 | LCC = 0.9406 | SRCC = 0.9416 ] +INFO:root:Average loss=0.14759923696517943 +INFO:root: +[Valid][2][UTT][ MSE = 0.2644 | LCC = 0.9437 | SRCC = 0.9437 ] +INFO:root:Average loss=0.13221753388643265 +INFO:root: +[Valid][3][UTT][ MSE = 0.2338 | LCC = 0.9433 | SRCC = 0.9452 ] +INFO:root:Average loss=0.11816493666172027 +INFO:root: +[Valid][4][UTT][ MSE = 0.1986 | LCC = 0.9494 | SRCC = 0.9500 ] +INFO:root:Average loss=0.10603754213452339 +INFO:root: +[Valid][5][UTT][ MSE = 0.2298 | LCC = 0.9503 | SRCC = 0.9509 ] +INFO:root:Average loss=0.1000047864317894 +INFO:root: +[Valid][6][UTT][ MSE = 0.2322 | LCC = 0.9514 | SRCC = 0.9518 ] +INFO:root:Average loss=0.08806965875625611 +INFO:root: +[Valid][7][UTT][ MSE = 0.2343 | LCC = 0.9519 | SRCC = 0.9522 ] +INFO:root:Average loss=0.08229750254750252 +INFO:root: +[Valid][8][UTT][ MSE = 0.2265 | LCC = 0.9530 | SRCC = 0.9534 ] +INFO:root:Average loss=0.07558203938603401 +INFO:root: +[Valid][9][UTT][ MSE = 0.1978 | LCC = 0.9525 | SRCC = 0.9525 ] +INFO:root:Average loss=0.07190240490436554 +INFO:root: +[Valid][10][UTT][ MSE = 0.2336 | LCC = 0.9525 | SRCC = 0.9531 ] +INFO:root:Average loss=0.06518563145399094 +INFO:root: +[Valid][11][UTT][ MSE = 0.2331 | LCC = 0.9536 | SRCC = 0.9542 ] +INFO:root:Average loss=0.06128095060586929 +INFO:root: +[Valid][12][UTT][ MSE = 0.2069 | LCC = 0.9550 | SRCC = 0.9547 ] +INFO:root:Average loss=0.05622939178347588 +INFO:root: +[Valid][13][UTT][ MSE = 0.2174 | LCC = 0.9530 | SRCC = 0.9535 ] +INFO:root:Average loss=0.05209077848494053 +INFO:root: +[Valid][14][UTT][ MSE = 0.2043 | LCC = 0.9540 | SRCC = 0.9540 ] +INFO:root:Average loss=0.05028011167049408 +INFO:root: +[Valid][15][UTT][ MSE = 0.2188 | LCC = 0.9533 | SRCC = 0.9536 ] +INFO:root:Average loss=0.04791378608345986 +INFO:root: +[Valid][16][UTT][ MSE = 0.1923 | LCC = 0.9551 | SRCC = 0.9547 ] +INFO:root:Average loss=0.045989278703927994 +INFO:root: +[Valid][17][UTT][ MSE = 0.2271 | LCC = 0.9546 | SRCC = 0.9550 ] +INFO:root:Average loss=0.043443485096096995 +INFO:root: +[Valid][18][UTT][ MSE = 0.2211 | LCC = 0.9554 | SRCC = 0.9558 ] +INFO:root:Average loss=0.04040505690872669 +INFO:root: +[Valid][19][UTT][ MSE = 0.2212 | LCC = 0.9548 | SRCC = 0.9544 ] +INFO:root:Average loss=0.038696258008480075 +INFO:root: +[Valid][20][UTT][ MSE = 0.2120 | LCC = 0.9541 | SRCC = 0.9543 ] +INFO:root:Average loss=0.03879747723042965 +INFO:root: +[Valid][21][UTT][ MSE = 0.2149 | LCC = 0.9555 | SRCC = 0.9555 ] +INFO:root:Average loss=0.03526605105400085 +INFO:root: +[Valid][22][UTT][ MSE = 0.2010 | LCC = 0.9552 | SRCC = 0.9555 ] +INFO:root:Average loss=0.03439477989077568 +INFO:root: +[Valid][23][UTT][ MSE = 0.2296 | LCC = 0.9548 | SRCC = 0.9555 ] +INFO:root:Average loss=0.0334092925041914 +INFO:root: +[Valid][24][UTT][ MSE = 0.2149 | LCC = 0.9556 | SRCC = 0.9555 ] +INFO:root:Average loss=0.030763095408678054 +INFO:root: +[Valid][25][UTT][ MSE = 0.2164 | LCC = 0.9548 | SRCC = 0.9544 ] +INFO:root:Average loss=0.029839814856648446 +INFO:root: +[Valid][26][UTT][ MSE = 0.2184 | LCC = 0.9556 | SRCC = 0.9561 ] +INFO:root:Average loss=0.029158102542161942 +INFO:root: +[Valid][27][UTT][ MSE = 0.2096 | LCC = 0.9553 | SRCC = 0.9552 ] +INFO:root:Average loss=0.028182941995561123 +INFO:root: +[Valid][28][UTT][ MSE = 0.2154 | LCC = 0.9547 | SRCC = 0.9542 ] +INFO:root:Average loss=0.026787277713418007 +INFO:root: +[Valid][29][UTT][ MSE = 0.1998 | LCC = 0.9547 | SRCC = 0.9544 ] +INFO:root:Average loss=0.026905025899410247 +INFO:root: +[Valid][30][UTT][ MSE = 0.2144 | LCC = 0.9537 | SRCC = 0.9540 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2224 | LCC = 0.9531 | SRCC = 0.9551 ] diff --git a/results_tencent/w2v2_base/layer6/run1/config.gin b/results_tencent/w2v2_base/layer6/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer6/run1/model.pt b/results_tencent/w2v2_base/layer6/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..67ce42e19e68e84257901153a7baebb447b80606 --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5621346843958a931bcc871cd446cd2d96902459aeb6e2bd2058ebeb930c82df +size 2747979 diff --git a/results_tencent/w2v2_base/layer6/run1/model_best.pt b/results_tencent/w2v2_base/layer6/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a35caaeaa5ce025b444edc91340fa05988102cf1 --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:695e4c0d874fadf2578235976100a5aa1b9782305f47f595db1ab337f48ea571 +size 2749702 diff --git a/results_tencent/w2v2_base/layer6/run1/train.log b/results_tencent/w2v2_base/layer6/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b94ac8e971abeb45cdaa5280446dafcd65645f32 --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.509103485584259 +INFO:root: +[Valid][0][UTT][ MSE = 0.3035 | LCC = 0.9197 | SRCC = 0.9230 ] +INFO:root:Average loss=0.18778616148233412 +INFO:root: +[Valid][1][UTT][ MSE = 0.2285 | LCC = 0.9412 | SRCC = 0.9429 ] +INFO:root:Average loss=0.1519696004986763 +INFO:root: +[Valid][2][UTT][ MSE = 0.2733 | LCC = 0.9452 | SRCC = 0.9464 ] +INFO:root:Average loss=0.13142936462163926 +INFO:root: +[Valid][3][UTT][ MSE = 0.2428 | LCC = 0.9486 | SRCC = 0.9491 ] +INFO:root:Average loss=0.11809164315462112 +INFO:root: +[Valid][4][UTT][ MSE = 0.2048 | LCC = 0.9501 | SRCC = 0.9512 ] +INFO:root:Average loss=0.10726064753532409 +INFO:root: +[Valid][5][UTT][ MSE = 0.2520 | LCC = 0.9505 | SRCC = 0.9523 ] +INFO:root:Average loss=0.09633173960447311 +INFO:root: +[Valid][6][UTT][ MSE = 0.1955 | LCC = 0.9508 | SRCC = 0.9515 ] +INFO:root:Average loss=0.08971412187814712 +INFO:root: +[Valid][7][UTT][ MSE = 0.2470 | LCC = 0.9529 | SRCC = 0.9536 ] +INFO:root:Average loss=0.08240783962607384 +INFO:root: +[Valid][8][UTT][ MSE = 0.2027 | LCC = 0.9507 | SRCC = 0.9519 ] +INFO:root:Average loss=0.0773831635415554 +INFO:root: +[Valid][9][UTT][ MSE = 0.2333 | LCC = 0.9526 | SRCC = 0.9535 ] +INFO:root:Average loss=0.07354312473535538 +INFO:root: +[Valid][10][UTT][ MSE = 0.2225 | LCC = 0.9524 | SRCC = 0.9530 ] +INFO:root:Average loss=0.06403081440925598 +INFO:root: +[Valid][11][UTT][ MSE = 0.2171 | LCC = 0.9533 | SRCC = 0.9534 ] +INFO:root:Average loss=0.06375087305903435 +INFO:root: +[Valid][12][UTT][ MSE = 0.1970 | LCC = 0.9519 | SRCC = 0.9517 ] +INFO:root:Average loss=0.0599582602083683 +INFO:root: +[Valid][13][UTT][ MSE = 0.2160 | LCC = 0.9536 | SRCC = 0.9543 ] +INFO:root:Average loss=0.054252991542220116 +INFO:root: +[Valid][14][UTT][ MSE = 0.1939 | LCC = 0.9526 | SRCC = 0.9526 ] +INFO:root:Average loss=0.054074354469776154 +INFO:root: +[Valid][15][UTT][ MSE = 0.1951 | LCC = 0.9546 | SRCC = 0.9549 ] +INFO:root:Average loss=0.05069985020160675 +INFO:root: +[Valid][16][UTT][ MSE = 0.2018 | LCC = 0.9536 | SRCC = 0.9537 ] +INFO:root:Average loss=0.04582629232108593 +INFO:root: +[Valid][17][UTT][ MSE = 0.1985 | LCC = 0.9528 | SRCC = 0.9530 ] +INFO:root:Average loss=0.044897915929555894 +INFO:root: +[Valid][18][UTT][ MSE = 0.2233 | LCC = 0.9528 | SRCC = 0.9535 ] +INFO:root:Average loss=0.04371457774937153 +INFO:root: +[Valid][19][UTT][ MSE = 0.2108 | LCC = 0.9531 | SRCC = 0.9534 ] +INFO:root:Average loss=0.03908770583570004 +INFO:root: +[Valid][20][UTT][ MSE = 0.2355 | LCC = 0.9514 | SRCC = 0.9523 ] +INFO:root:Average loss=0.039222451701760294 +INFO:root: +[Valid][21][UTT][ MSE = 0.2112 | LCC = 0.9543 | SRCC = 0.9548 ] +INFO:root:Average loss=0.03637759356200695 +INFO:root: +[Valid][22][UTT][ MSE = 0.2352 | LCC = 0.9522 | SRCC = 0.9525 ] +INFO:root:Average loss=0.03639232502877712 +INFO:root: +[Valid][23][UTT][ MSE = 0.2067 | LCC = 0.9526 | SRCC = 0.9526 ] +INFO:root:Average loss=0.032903242066502573 +INFO:root: +[Valid][24][UTT][ MSE = 0.2022 | LCC = 0.9535 | SRCC = 0.9538 ] +INFO:root:Average loss=0.03250853615999222 +INFO:root: +[Valid][25][UTT][ MSE = 0.2187 | LCC = 0.9519 | SRCC = 0.9531 ] +INFO:root:Average loss=0.03229016450047493 +INFO:root: +[Valid][26][UTT][ MSE = 0.1873 | LCC = 0.9548 | SRCC = 0.9543 ] +INFO:root:Average loss=0.030635146707296373 +INFO:root: +[Valid][27][UTT][ MSE = 0.1908 | LCC = 0.9535 | SRCC = 0.9532 ] +INFO:root:Average loss=0.02959622633457184 +INFO:root: +[Valid][28][UTT][ MSE = 0.2189 | LCC = 0.9538 | SRCC = 0.9540 ] +INFO:root:Average loss=0.027742041766643525 +INFO:root: +[Valid][29][UTT][ MSE = 0.2095 | LCC = 0.9545 | SRCC = 0.9553 ] +INFO:root:Average loss=0.026790095791220665 +INFO:root: +[Valid][30][UTT][ MSE = 0.2283 | LCC = 0.9528 | SRCC = 0.9538 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1888 | LCC = 0.9542 | SRCC = 0.9551 ] diff --git a/results_tencent/w2v2_base/layer6/run2/config.gin b/results_tencent/w2v2_base/layer6/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer6/run2/model.pt b/results_tencent/w2v2_base/layer6/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..aa30d727ba4f3c6bc02a21d7296ab75f48b83dfc --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b0283bd7aa6d01cd7784244931c2d340e314c315abb20f8666a5a8385ac119c +size 2747979 diff --git a/results_tencent/w2v2_base/layer6/run2/model_best.pt b/results_tencent/w2v2_base/layer6/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..20acda6e4e18407727facd8444c713402ae11f4c --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:249f412831385cf4230a82227a65f922c521e42c7a351689f8ed363e12358339 +size 2749702 diff --git a/results_tencent/w2v2_base/layer6/run2/train.log b/results_tencent/w2v2_base/layer6/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2ce89242936beecea8e54624a8c4c53855572c89 --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.573265992641449 +INFO:root: +[Valid][0][UTT][ MSE = 0.3367 | LCC = 0.9080 | SRCC = 0.9136 ] +INFO:root:Average loss=0.20607798331975938 +INFO:root: +[Valid][1][UTT][ MSE = 0.2355 | LCC = 0.9361 | SRCC = 0.9387 ] +INFO:root:Average loss=0.15718668711185455 +INFO:root: +[Valid][2][UTT][ MSE = 0.2335 | LCC = 0.9436 | SRCC = 0.9458 ] +INFO:root:Average loss=0.1320381688475609 +INFO:root: +[Valid][3][UTT][ MSE = 0.2015 | LCC = 0.9488 | SRCC = 0.9490 ] +INFO:root:Average loss=0.11728354740142823 +INFO:root: +[Valid][4][UTT][ MSE = 0.2311 | LCC = 0.9498 | SRCC = 0.9495 ] +INFO:root:Average loss=0.10689728105068207 +INFO:root: +[Valid][5][UTT][ MSE = 0.2028 | LCC = 0.9527 | SRCC = 0.9521 ] +INFO:root:Average loss=0.1000913252234459 +INFO:root: +[Valid][6][UTT][ MSE = 0.1980 | LCC = 0.9523 | SRCC = 0.9518 ] +INFO:root:Average loss=0.09223453378677368 +INFO:root: +[Valid][7][UTT][ MSE = 0.1964 | LCC = 0.9534 | SRCC = 0.9532 ] +INFO:root:Average loss=0.08162043878436089 +INFO:root: +[Valid][8][UTT][ MSE = 0.1963 | LCC = 0.9528 | SRCC = 0.9525 ] +INFO:root:Average loss=0.07535241118073463 +INFO:root: +[Valid][9][UTT][ MSE = 0.1957 | LCC = 0.9545 | SRCC = 0.9540 ] +INFO:root:Average loss=0.07067690920829774 +INFO:root: +[Valid][10][UTT][ MSE = 0.1930 | LCC = 0.9525 | SRCC = 0.9527 ] +INFO:root:Average loss=0.06457604441046715 +INFO:root: +[Valid][11][UTT][ MSE = 0.1958 | LCC = 0.9536 | SRCC = 0.9536 ] +INFO:root:Average loss=0.05913432374596596 +INFO:root: +[Valid][12][UTT][ MSE = 0.1847 | LCC = 0.9532 | SRCC = 0.9528 ] +INFO:root:Average loss=0.056198416650295256 +INFO:root: +[Valid][13][UTT][ MSE = 0.1958 | LCC = 0.9547 | SRCC = 0.9546 ] +INFO:root:Average loss=0.053317849189043046 +INFO:root: +[Valid][14][UTT][ MSE = 0.1889 | LCC = 0.9540 | SRCC = 0.9541 ] +INFO:root:Average loss=0.050411739498376844 +INFO:root: +[Valid][15][UTT][ MSE = 0.1891 | LCC = 0.9531 | SRCC = 0.9530 ] +INFO:root:Average loss=0.048320881247520445 +INFO:root: +[Valid][16][UTT][ MSE = 0.2082 | LCC = 0.9540 | SRCC = 0.9541 ] +INFO:root:Average loss=0.046570097133517264 +INFO:root: +[Valid][17][UTT][ MSE = 0.1987 | LCC = 0.9528 | SRCC = 0.9530 ] +INFO:root:Average loss=0.04428727953135967 +INFO:root: +[Valid][18][UTT][ MSE = 0.1865 | LCC = 0.9531 | SRCC = 0.9531 ] +INFO:root:Average loss=0.041313169091939925 +INFO:root: +[Valid][19][UTT][ MSE = 0.1837 | LCC = 0.9528 | SRCC = 0.9527 ] +INFO:root:Average loss=0.03895676036179066 +INFO:root: +[Valid][20][UTT][ MSE = 0.1843 | LCC = 0.9553 | SRCC = 0.9551 ] +INFO:root:Average loss=0.039370365142822265 +INFO:root: +[Valid][21][UTT][ MSE = 0.1972 | LCC = 0.9508 | SRCC = 0.9504 ] +INFO:root:Average loss=0.03646983928978443 +INFO:root: +[Valid][22][UTT][ MSE = 0.1832 | LCC = 0.9559 | SRCC = 0.9557 ] +INFO:root:Average loss=0.03431620278954506 +INFO:root: +[Valid][23][UTT][ MSE = 0.1943 | LCC = 0.9540 | SRCC = 0.9539 ] +INFO:root:Average loss=0.034378441035747526 +INFO:root: +[Valid][24][UTT][ MSE = 0.1935 | LCC = 0.9544 | SRCC = 0.9541 ] +INFO:root:Average loss=0.031143992885947228 +INFO:root: +[Valid][25][UTT][ MSE = 0.1953 | LCC = 0.9540 | SRCC = 0.9539 ] +INFO:root:Average loss=0.03009871070086956 +INFO:root: +[Valid][26][UTT][ MSE = 0.1873 | LCC = 0.9548 | SRCC = 0.9546 ] +INFO:root:Average loss=0.029844417467713356 +INFO:root: +[Valid][27][UTT][ MSE = 0.1986 | LCC = 0.9558 | SRCC = 0.9555 ] +INFO:root:Average loss=0.028812220126390457 +INFO:root: +[Valid][28][UTT][ MSE = 0.1792 | LCC = 0.9540 | SRCC = 0.9538 ] +INFO:root:Average loss=0.026282853916287423 +INFO:root: +[Valid][29][UTT][ MSE = 0.1966 | LCC = 0.9544 | SRCC = 0.9541 ] +INFO:root:Average loss=0.025273734517395497 +INFO:root: +[Valid][30][UTT][ MSE = 0.2006 | LCC = 0.9532 | SRCC = 0.9530 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1905 | LCC = 0.9542 | SRCC = 0.9551 ] diff --git a/results_tencent/w2v2_base/layer6/run3/config.gin b/results_tencent/w2v2_base/layer6/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer6/run3/model.pt b/results_tencent/w2v2_base/layer6/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4001b672029454b20e311d97d31eca41ecf26211 --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7c70150b7dfb7f64bee0c03ae9f5a98e081c00ee4c43f700d0a5b16db3f2ff39 +size 2747979 diff --git a/results_tencent/w2v2_base/layer6/run3/model_best.pt b/results_tencent/w2v2_base/layer6/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..1c67ccdfb04cc088ac3fc34dd7ca19a9509f897f --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6d336809d304d40a9a0d9451259bb16ace08ccd416abfa06765ea9a77377930b +size 2749702 diff --git a/results_tencent/w2v2_base/layer6/run3/train.log b/results_tencent/w2v2_base/layer6/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b9c0a37ab0e20ea89d595d613b6a104d0aff28b0 --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46138644456863404 +INFO:root: +[Valid][0][UTT][ MSE = 0.2788 | LCC = 0.9241 | SRCC = 0.9260 ] +INFO:root:Average loss=0.18598020565509796 +INFO:root: +[Valid][1][UTT][ MSE = 0.2448 | LCC = 0.9394 | SRCC = 0.9402 ] +INFO:root:Average loss=0.15053254973888397 +INFO:root: +[Valid][2][UTT][ MSE = 0.2193 | LCC = 0.9435 | SRCC = 0.9439 ] +INFO:root:Average loss=0.13450923854112626 +INFO:root: +[Valid][3][UTT][ MSE = 0.2487 | LCC = 0.9474 | SRCC = 0.9477 ] +INFO:root:Average loss=0.11599486231803895 +INFO:root: +[Valid][4][UTT][ MSE = 0.2291 | LCC = 0.9515 | SRCC = 0.9516 ] +INFO:root:Average loss=0.1029166253209114 +INFO:root: +[Valid][5][UTT][ MSE = 0.2495 | LCC = 0.9513 | SRCC = 0.9513 ] +INFO:root:Average loss=0.09526389467716218 +INFO:root: +[Valid][6][UTT][ MSE = 0.2082 | LCC = 0.9531 | SRCC = 0.9531 ] +INFO:root:Average loss=0.08855937957763672 +INFO:root: +[Valid][7][UTT][ MSE = 0.2032 | LCC = 0.9542 | SRCC = 0.9539 ] +INFO:root:Average loss=0.07882406648993492 +INFO:root: +[Valid][8][UTT][ MSE = 0.1899 | LCC = 0.9539 | SRCC = 0.9534 ] +INFO:root:Average loss=0.07362402230501175 +INFO:root: +[Valid][9][UTT][ MSE = 0.2256 | LCC = 0.9543 | SRCC = 0.9543 ] +INFO:root:Average loss=0.06819047096371651 +INFO:root: +[Valid][10][UTT][ MSE = 0.2136 | LCC = 0.9541 | SRCC = 0.9537 ] +INFO:root:Average loss=0.06566660180687904 +INFO:root: +[Valid][11][UTT][ MSE = 0.2078 | LCC = 0.9548 | SRCC = 0.9545 ] +INFO:root:Average loss=0.06298906055092812 +INFO:root: +[Valid][12][UTT][ MSE = 0.1934 | LCC = 0.9558 | SRCC = 0.9550 ] +INFO:root:Average loss=0.05580642569065094 +INFO:root: +[Valid][13][UTT][ MSE = 0.1964 | LCC = 0.9565 | SRCC = 0.9559 ] +INFO:root:Average loss=0.05106067173182964 +INFO:root: +[Valid][14][UTT][ MSE = 0.2256 | LCC = 0.9559 | SRCC = 0.9559 ] +INFO:root:Average loss=0.05053350432217121 +INFO:root: +[Valid][15][UTT][ MSE = 0.2140 | LCC = 0.9551 | SRCC = 0.9551 ] +INFO:root:Average loss=0.047647893846035 +INFO:root: +[Valid][16][UTT][ MSE = 0.2170 | LCC = 0.9554 | SRCC = 0.9555 ] +INFO:root:Average loss=0.0440691457092762 +INFO:root: +[Valid][17][UTT][ MSE = 0.1981 | LCC = 0.9554 | SRCC = 0.9547 ] +INFO:root:Average loss=0.04104357202351093 +INFO:root: +[Valid][18][UTT][ MSE = 0.1898 | LCC = 0.9552 | SRCC = 0.9544 ] +INFO:root:Average loss=0.03936381536722183 +INFO:root: +[Valid][19][UTT][ MSE = 0.2001 | LCC = 0.9551 | SRCC = 0.9542 ] +INFO:root:Average loss=0.038452261656522754 +INFO:root: +[Valid][20][UTT][ MSE = 0.2138 | LCC = 0.9559 | SRCC = 0.9553 ] +INFO:root:Average loss=0.03720599995553493 +INFO:root: +[Valid][21][UTT][ MSE = 0.2035 | LCC = 0.9554 | SRCC = 0.9548 ] +INFO:root:Average loss=0.035613310515880583 +INFO:root: +[Valid][22][UTT][ MSE = 0.1903 | LCC = 0.9555 | SRCC = 0.9548 ] +INFO:root:Average loss=0.03509629175066948 +INFO:root: +[Valid][23][UTT][ MSE = 0.1785 | LCC = 0.9546 | SRCC = 0.9543 ] +INFO:root:Average loss=0.032934789806604386 +INFO:root: +[Valid][24][UTT][ MSE = 0.2089 | LCC = 0.9555 | SRCC = 0.9549 ] +INFO:root:Average loss=0.031883284389972684 +INFO:root: +[Valid][25][UTT][ MSE = 0.1900 | LCC = 0.9539 | SRCC = 0.9528 ] +INFO:root:Average loss=0.034952421620488165 +INFO:root: +[Valid][26][UTT][ MSE = 0.2096 | LCC = 0.9555 | SRCC = 0.9548 ] +INFO:root:Average loss=0.030749918177723885 +INFO:root: +[Valid][27][UTT][ MSE = 0.2019 | LCC = 0.9550 | SRCC = 0.9547 ] +INFO:root:Average loss=0.028994192138314247 +INFO:root: +[Valid][28][UTT][ MSE = 0.1749 | LCC = 0.9556 | SRCC = 0.9547 ] +INFO:root:Average loss=0.02696235366165638 +INFO:root: +[Valid][29][UTT][ MSE = 0.1860 | LCC = 0.9557 | SRCC = 0.9544 ] +INFO:root:Average loss=0.024771322898566723 +INFO:root: +[Valid][30][UTT][ MSE = 0.1994 | LCC = 0.9565 | SRCC = 0.9554 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2031 | LCC = 0.9553 | SRCC = 0.9562 ] diff --git a/results_tencent/w2v2_base/layer6/run4/config.gin b/results_tencent/w2v2_base/layer6/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer6/run4/model.pt b/results_tencent/w2v2_base/layer6/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0234b74fe707d2334f1922b0b945e0c80692794d --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:233324ca299bc8e0d0ee632b8bfd1998257148e02981aadf48a3ad013bd47175 +size 2747979 diff --git a/results_tencent/w2v2_base/layer6/run4/model_best.pt b/results_tencent/w2v2_base/layer6/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..bae606fa2fe513c4d31a3f484d4dc9e4a79f845b --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:599798183f1b1b4313b11bc60e85141ff009d0b1c13efbe2bc637209de05f411 +size 2749638 diff --git a/results_tencent/w2v2_base/layer6/run4/train.log b/results_tencent/w2v2_base/layer6/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a824d72d231badc8441a8c0da8bc984785f99f19 --- /dev/null +++ b/results_tencent/w2v2_base/layer6/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5029064471721649 +INFO:root: +[Valid][0][UTT][ MSE = 0.2824 | LCC = 0.9180 | SRCC = 0.9179 ] +INFO:root:Average loss=0.19389624577760697 +INFO:root: +[Valid][1][UTT][ MSE = 0.2313 | LCC = 0.9371 | SRCC = 0.9389 ] +INFO:root:Average loss=0.1542189890742302 +INFO:root: +[Valid][2][UTT][ MSE = 0.2281 | LCC = 0.9444 | SRCC = 0.9457 ] +INFO:root:Average loss=0.13572586485743524 +INFO:root: +[Valid][3][UTT][ MSE = 0.2209 | LCC = 0.9475 | SRCC = 0.9488 ] +INFO:root:Average loss=0.12059253579378128 +INFO:root: +[Valid][4][UTT][ MSE = 0.2164 | LCC = 0.9502 | SRCC = 0.9508 ] +INFO:root:Average loss=0.10706348979473114 +INFO:root: +[Valid][5][UTT][ MSE = 0.2043 | LCC = 0.9519 | SRCC = 0.9522 ] +INFO:root:Average loss=0.0964868126809597 +INFO:root: +[Valid][6][UTT][ MSE = 0.1972 | LCC = 0.9523 | SRCC = 0.9521 ] +INFO:root:Average loss=0.09050163877010345 +INFO:root: +[Valid][7][UTT][ MSE = 0.2261 | LCC = 0.9526 | SRCC = 0.9523 ] +INFO:root:Average loss=0.08639875429868699 +INFO:root: +[Valid][8][UTT][ MSE = 0.2179 | LCC = 0.9517 | SRCC = 0.9520 ] +INFO:root:Average loss=0.07822945144772529 +INFO:root: +[Valid][9][UTT][ MSE = 0.1952 | LCC = 0.9541 | SRCC = 0.9542 ] +INFO:root:Average loss=0.07189642462134362 +INFO:root: +[Valid][10][UTT][ MSE = 0.1996 | LCC = 0.9527 | SRCC = 0.9532 ] +INFO:root:Average loss=0.06770532232522965 +INFO:root: +[Valid][11][UTT][ MSE = 0.1811 | LCC = 0.9536 | SRCC = 0.9535 ] +INFO:root:Average loss=0.06469581693410874 +INFO:root: +[Valid][12][UTT][ MSE = 0.1889 | LCC = 0.9488 | SRCC = 0.9489 ] +INFO:root:Average loss=0.0617309036552906 +INFO:root: +[Valid][13][UTT][ MSE = 0.1856 | LCC = 0.9532 | SRCC = 0.9530 ] +INFO:root:Average loss=0.056156175702810285 +INFO:root: +[Valid][14][UTT][ MSE = 0.1979 | LCC = 0.9534 | SRCC = 0.9526 ] +INFO:root:Average loss=0.054342541590332985 +INFO:root: +[Valid][15][UTT][ MSE = 0.2120 | LCC = 0.9548 | SRCC = 0.9539 ] +INFO:root:Average loss=0.05017782360315323 +INFO:root: +[Valid][16][UTT][ MSE = 0.1930 | LCC = 0.9546 | SRCC = 0.9543 ] +INFO:root:Average loss=0.04647701631486416 +INFO:root: +[Valid][17][UTT][ MSE = 0.2081 | LCC = 0.9529 | SRCC = 0.9526 ] +INFO:root:Average loss=0.0456433279812336 +INFO:root: +[Valid][18][UTT][ MSE = 0.2088 | LCC = 0.9537 | SRCC = 0.9534 ] +INFO:root:Average loss=0.04161232647299767 +INFO:root: +[Valid][19][UTT][ MSE = 0.2136 | LCC = 0.9527 | SRCC = 0.9521 ] +INFO:root:Average loss=0.03873299416899681 +INFO:root: +[Valid][20][UTT][ MSE = 0.2143 | LCC = 0.9529 | SRCC = 0.9531 ] +INFO:root:Average loss=0.03930717596411705 +INFO:root: +[Valid][21][UTT][ MSE = 0.2089 | LCC = 0.9532 | SRCC = 0.9525 ] +INFO:root:Average loss=0.03459497202932835 +INFO:root: +[Valid][22][UTT][ MSE = 0.1776 | LCC = 0.9517 | SRCC = 0.9516 ] +INFO:root:Average loss=0.03429144786298275 +INFO:root: +[Valid][23][UTT][ MSE = 0.1957 | LCC = 0.9534 | SRCC = 0.9535 ] +INFO:root:Average loss=0.03374573300778866 +INFO:root: +[Valid][24][UTT][ MSE = 0.1859 | LCC = 0.9521 | SRCC = 0.9520 ] +INFO:root:Average loss=0.033696296140551564 +INFO:root: +[Valid][25][UTT][ MSE = 0.1873 | LCC = 0.9517 | SRCC = 0.9515 ] +INFO:root:Average loss=0.0318011344075203 +INFO:root: +[Valid][26][UTT][ MSE = 0.1903 | LCC = 0.9544 | SRCC = 0.9541 ] +INFO:root:Average loss=0.02915976729989052 +INFO:root: +[Valid][27][UTT][ MSE = 0.1989 | LCC = 0.9540 | SRCC = 0.9537 ] +INFO:root:Average loss=0.028887270033359527 +INFO:root: +[Valid][28][UTT][ MSE = 0.1962 | LCC = 0.9539 | SRCC = 0.9536 ] +INFO:root:Average loss=0.028946391828358172 +INFO:root: +[Valid][29][UTT][ MSE = 0.1886 | LCC = 0.9520 | SRCC = 0.9517 ] +INFO:root:Average loss=0.029933612957596778 +INFO:root: +[Valid][30][UTT][ MSE = 0.2096 | LCC = 0.9529 | SRCC = 0.9528 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2107 | LCC = 0.9558 | SRCC = 0.9573 ] diff --git a/results_tencent/w2v2_base/layer7/run0/config.gin b/results_tencent/w2v2_base/layer7/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer7/run0/model.pt b/results_tencent/w2v2_base/layer7/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..104448eae83e64cf48712ec5833f4f171436252e --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a089fc56f1beba10fcb13a6aa5e8b62b4d47679e3a0c4c754fc7d9c1e9a23fa7 +size 2747979 diff --git a/results_tencent/w2v2_base/layer7/run0/model_best.pt b/results_tencent/w2v2_base/layer7/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..eea1597271db96f8ac62531a0e0bcb854dd44828 --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6dd1855fa99c0bf543dabda4d75f24c63f161e967f87fde60136660fdb4bb2ed +size 2749702 diff --git a/results_tencent/w2v2_base/layer7/run0/train.log b/results_tencent/w2v2_base/layer7/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..8b06156ee238ef62dea01f1c1f8fe8ee23fc273f --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5193981934785843 +INFO:root: +[Valid][0][UTT][ MSE = 0.2919 | LCC = 0.9221 | SRCC = 0.9219 ] +INFO:root:Average loss=0.18046985644102095 +INFO:root: +[Valid][1][UTT][ MSE = 0.2447 | LCC = 0.9410 | SRCC = 0.9417 ] +INFO:root:Average loss=0.1504060403108597 +INFO:root: +[Valid][2][UTT][ MSE = 0.2471 | LCC = 0.9395 | SRCC = 0.9398 ] +INFO:root:Average loss=0.12993736129999162 +INFO:root: +[Valid][3][UTT][ MSE = 0.2327 | LCC = 0.9514 | SRCC = 0.9526 ] +INFO:root:Average loss=0.11544721567630768 +INFO:root: +[Valid][4][UTT][ MSE = 0.2123 | LCC = 0.9509 | SRCC = 0.9524 ] +INFO:root:Average loss=0.1025715210735798 +INFO:root: +[Valid][5][UTT][ MSE = 0.2138 | LCC = 0.9509 | SRCC = 0.9531 ] +INFO:root:Average loss=0.09418592384457589 +INFO:root: +[Valid][6][UTT][ MSE = 0.2235 | LCC = 0.9530 | SRCC = 0.9538 ] +INFO:root:Average loss=0.085387898594141 +INFO:root: +[Valid][7][UTT][ MSE = 0.1924 | LCC = 0.9536 | SRCC = 0.9545 ] +INFO:root:Average loss=0.0838015812933445 +INFO:root: +[Valid][8][UTT][ MSE = 0.2159 | LCC = 0.9505 | SRCC = 0.9510 ] +INFO:root:Average loss=0.07674246954917907 +INFO:root: +[Valid][9][UTT][ MSE = 0.2063 | LCC = 0.9552 | SRCC = 0.9553 ] +INFO:root:Average loss=0.07107310143113137 +INFO:root: +[Valid][10][UTT][ MSE = 0.2295 | LCC = 0.9526 | SRCC = 0.9539 ] +INFO:root:Average loss=0.0651664582490921 +INFO:root: +[Valid][11][UTT][ MSE = 0.2133 | LCC = 0.9530 | SRCC = 0.9533 ] +INFO:root:Average loss=0.06176995629072189 +INFO:root: +[Valid][12][UTT][ MSE = 0.2104 | LCC = 0.9535 | SRCC = 0.9534 ] +INFO:root:Average loss=0.05706806516647339 +INFO:root: +[Valid][13][UTT][ MSE = 0.2180 | LCC = 0.9530 | SRCC = 0.9530 ] +INFO:root:Average loss=0.053303583815693854 +INFO:root: +[Valid][14][UTT][ MSE = 0.2067 | LCC = 0.9542 | SRCC = 0.9540 ] +INFO:root:Average loss=0.04934834590554237 +INFO:root: +[Valid][15][UTT][ MSE = 0.1834 | LCC = 0.9540 | SRCC = 0.9546 ] +INFO:root:Average loss=0.04791363044083118 +INFO:root: +[Valid][16][UTT][ MSE = 0.1975 | LCC = 0.9543 | SRCC = 0.9535 ] +INFO:root:Average loss=0.04658899243175983 +INFO:root: +[Valid][17][UTT][ MSE = 0.1931 | LCC = 0.9561 | SRCC = 0.9563 ] +INFO:root:Average loss=0.042296958476305005 +INFO:root: +[Valid][18][UTT][ MSE = 0.1860 | LCC = 0.9524 | SRCC = 0.9532 ] +INFO:root:Average loss=0.03989433294534683 +INFO:root: +[Valid][19][UTT][ MSE = 0.2069 | LCC = 0.9552 | SRCC = 0.9559 ] +INFO:root:Average loss=0.03864786168932915 +INFO:root: +[Valid][20][UTT][ MSE = 0.2122 | LCC = 0.9539 | SRCC = 0.9540 ] +INFO:root:Average loss=0.03752205330133438 +INFO:root: +[Valid][21][UTT][ MSE = 0.2015 | LCC = 0.9553 | SRCC = 0.9558 ] +INFO:root:Average loss=0.0375347962230444 +INFO:root: +[Valid][22][UTT][ MSE = 0.2083 | LCC = 0.9536 | SRCC = 0.9539 ] +INFO:root:Average loss=0.032888634607195856 +INFO:root: +[Valid][23][UTT][ MSE = 0.1921 | LCC = 0.9526 | SRCC = 0.9528 ] +INFO:root:Average loss=0.03240425074100494 +INFO:root: +[Valid][24][UTT][ MSE = 0.2047 | LCC = 0.9556 | SRCC = 0.9558 ] +INFO:root:Average loss=0.031971426665782926 +INFO:root: +[Valid][25][UTT][ MSE = 0.2003 | LCC = 0.9544 | SRCC = 0.9545 ] +INFO:root:Average loss=0.030381831392645837 +INFO:root: +[Valid][26][UTT][ MSE = 0.2145 | LCC = 0.9555 | SRCC = 0.9558 ] +INFO:root:Average loss=0.030992392584681512 +INFO:root: +[Valid][27][UTT][ MSE = 0.2095 | LCC = 0.9547 | SRCC = 0.9553 ] +INFO:root:Average loss=0.029569377586245536 +INFO:root: +[Valid][28][UTT][ MSE = 0.2161 | LCC = 0.9556 | SRCC = 0.9560 ] +INFO:root:Average loss=0.027308474227786064 +INFO:root: +[Valid][29][UTT][ MSE = 0.2175 | LCC = 0.9547 | SRCC = 0.9546 ] +INFO:root:Average loss=0.026212678037583828 +INFO:root: +[Valid][30][UTT][ MSE = 0.2035 | LCC = 0.9520 | SRCC = 0.9521 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2015 | LCC = 0.9535 | SRCC = 0.9546 ] diff --git a/results_tencent/w2v2_base/layer7/run1/config.gin b/results_tencent/w2v2_base/layer7/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer7/run1/model.pt b/results_tencent/w2v2_base/layer7/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..794b33768d3dbe0544d7ce84810412bbbb7f2d6d --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:49a4ba4130fb24bab44d14f5900b111528f5a1e8d32c66d1b2925c4ebaf87530 +size 2747979 diff --git a/results_tencent/w2v2_base/layer7/run1/model_best.pt b/results_tencent/w2v2_base/layer7/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d43e1dcc05191eb2713ee4f054815b6f72f791e1 --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:55594a1aabb3d18636c907502adf0e1fc204106ae18c5d1deea98de07b306775 +size 2749702 diff --git a/results_tencent/w2v2_base/layer7/run1/train.log b/results_tencent/w2v2_base/layer7/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a253902b3568dc37467320c9877d2e29cc74dd41 --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5014873633384704 +INFO:root: +[Valid][0][UTT][ MSE = 0.2446 | LCC = 0.9222 | SRCC = 0.9248 ] +INFO:root:Average loss=0.18183583563566208 +INFO:root: +[Valid][1][UTT][ MSE = 0.2216 | LCC = 0.9413 | SRCC = 0.9423 ] +INFO:root:Average loss=0.1467885720729828 +INFO:root: +[Valid][2][UTT][ MSE = 0.2367 | LCC = 0.9480 | SRCC = 0.9480 ] +INFO:root:Average loss=0.12564155006408692 +INFO:root: +[Valid][3][UTT][ MSE = 0.2403 | LCC = 0.9491 | SRCC = 0.9497 ] +INFO:root:Average loss=0.11159128904342651 +INFO:root: +[Valid][4][UTT][ MSE = 0.2124 | LCC = 0.9500 | SRCC = 0.9507 ] +INFO:root:Average loss=0.1039645994901657 +INFO:root: +[Valid][5][UTT][ MSE = 0.2123 | LCC = 0.9506 | SRCC = 0.9513 ] +INFO:root:Average loss=0.09127648892998695 +INFO:root: +[Valid][6][UTT][ MSE = 0.1775 | LCC = 0.9510 | SRCC = 0.9516 ] +INFO:root:Average loss=0.08477612400054932 +INFO:root: +[Valid][7][UTT][ MSE = 0.2231 | LCC = 0.9513 | SRCC = 0.9528 ] +INFO:root:Average loss=0.07875199270248413 +INFO:root: +[Valid][8][UTT][ MSE = 0.1896 | LCC = 0.9514 | SRCC = 0.9514 ] +INFO:root:Average loss=0.07521219989657402 +INFO:root: +[Valid][9][UTT][ MSE = 0.1902 | LCC = 0.9522 | SRCC = 0.9527 ] +INFO:root:Average loss=0.06602723610401154 +INFO:root: +[Valid][10][UTT][ MSE = 0.1979 | LCC = 0.9525 | SRCC = 0.9526 ] +INFO:root:Average loss=0.06471650367975235 +INFO:root: +[Valid][11][UTT][ MSE = 0.1733 | LCC = 0.9528 | SRCC = 0.9526 ] +INFO:root:Average loss=0.05958950811624527 +INFO:root: +[Valid][12][UTT][ MSE = 0.1937 | LCC = 0.9536 | SRCC = 0.9535 ] +INFO:root:Average loss=0.057283385306596754 +INFO:root: +[Valid][13][UTT][ MSE = 0.2122 | LCC = 0.9515 | SRCC = 0.9523 ] +INFO:root:Average loss=0.05336844003200531 +INFO:root: +[Valid][14][UTT][ MSE = 0.1731 | LCC = 0.9537 | SRCC = 0.9531 ] +INFO:root:Average loss=0.04959870471060276 +INFO:root: +[Valid][15][UTT][ MSE = 0.1809 | LCC = 0.9520 | SRCC = 0.9518 ] +INFO:root:Average loss=0.04540753574669361 +INFO:root: +[Valid][16][UTT][ MSE = 0.1860 | LCC = 0.9529 | SRCC = 0.9523 ] +INFO:root:Average loss=0.04446196693181992 +INFO:root: +[Valid][17][UTT][ MSE = 0.1847 | LCC = 0.9522 | SRCC = 0.9521 ] +INFO:root:Average loss=0.04222408092021942 +INFO:root: +[Valid][18][UTT][ MSE = 0.1700 | LCC = 0.9538 | SRCC = 0.9529 ] +INFO:root:Average loss=0.040167486920952795 +INFO:root: +[Valid][19][UTT][ MSE = 0.1783 | LCC = 0.9544 | SRCC = 0.9539 ] +INFO:root:Average loss=0.0375723213404417 +INFO:root: +[Valid][20][UTT][ MSE = 0.1747 | LCC = 0.9548 | SRCC = 0.9542 ] +INFO:root:Average loss=0.03688623209297657 +INFO:root: +[Valid][21][UTT][ MSE = 0.1678 | LCC = 0.9538 | SRCC = 0.9534 ] +INFO:root:Average loss=0.03490481805801392 +INFO:root: +[Valid][22][UTT][ MSE = 0.1929 | LCC = 0.9535 | SRCC = 0.9533 ] +INFO:root:Average loss=0.0331564249843359 +INFO:root: +[Valid][23][UTT][ MSE = 0.1761 | LCC = 0.9538 | SRCC = 0.9534 ] +INFO:root:Average loss=0.03279092384874821 +INFO:root: +[Valid][24][UTT][ MSE = 0.1689 | LCC = 0.9530 | SRCC = 0.9523 ] +INFO:root:Average loss=0.030277934685349464 +INFO:root: +[Valid][25][UTT][ MSE = 0.1807 | LCC = 0.9533 | SRCC = 0.9527 ] +INFO:root:Average loss=0.03008510956168175 +INFO:root: +[Valid][26][UTT][ MSE = 0.1753 | LCC = 0.9542 | SRCC = 0.9533 ] +INFO:root:Average loss=0.029418001368641852 +INFO:root: +[Valid][27][UTT][ MSE = 0.1910 | LCC = 0.9537 | SRCC = 0.9537 ] +INFO:root:Average loss=0.028241884618997574 +INFO:root: +[Valid][28][UTT][ MSE = 0.1882 | LCC = 0.9533 | SRCC = 0.9525 ] +INFO:root:Average loss=0.027142094656825065 +INFO:root: +[Valid][29][UTT][ MSE = 0.1759 | LCC = 0.9530 | SRCC = 0.9524 ] +INFO:root:Average loss=0.026340259850025177 +INFO:root: +[Valid][30][UTT][ MSE = 0.1730 | LCC = 0.9531 | SRCC = 0.9523 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1791 | LCC = 0.9537 | SRCC = 0.9544 ] diff --git a/results_tencent/w2v2_base/layer7/run2/config.gin b/results_tencent/w2v2_base/layer7/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer7/run2/model.pt b/results_tencent/w2v2_base/layer7/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..58b3b35201d49fca034739e64277e1556a31a1e6 --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3f5ce5a12ef4aba1df126d39e27a1e5c07c2a9c36028ee26556082c114711247 +size 2747979 diff --git a/results_tencent/w2v2_base/layer7/run2/model_best.pt b/results_tencent/w2v2_base/layer7/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..150c5f698cbabc2d517394c797a3c918ca7a0e3c --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dec9735da699bff1a0d6d6734662d36d17539d1fc5232e3614e4302fe8892f3b +size 2749702 diff --git a/results_tencent/w2v2_base/layer7/run2/train.log b/results_tencent/w2v2_base/layer7/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..4234aa0f7df588f4ee0ee81ffa0852b91466414e --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4769088163375855 +INFO:root: +[Valid][0][UTT][ MSE = 0.2714 | LCC = 0.9200 | SRCC = 0.9200 ] +INFO:root:Average loss=0.18432400411367417 +INFO:root: +[Valid][1][UTT][ MSE = 0.2408 | LCC = 0.9432 | SRCC = 0.9446 ] +INFO:root:Average loss=0.14638750731945038 +INFO:root: +[Valid][2][UTT][ MSE = 0.2191 | LCC = 0.9491 | SRCC = 0.9502 ] +INFO:root:Average loss=0.12824191337823868 +INFO:root: +[Valid][3][UTT][ MSE = 0.1968 | LCC = 0.9499 | SRCC = 0.9503 ] +INFO:root:Average loss=0.11317472219467163 +INFO:root: +[Valid][4][UTT][ MSE = 0.2084 | LCC = 0.9528 | SRCC = 0.9531 ] +INFO:root:Average loss=0.10381424474716186 +INFO:root: +[Valid][5][UTT][ MSE = 0.2013 | LCC = 0.9516 | SRCC = 0.9518 ] +INFO:root:Average loss=0.09636339032649995 +INFO:root: +[Valid][6][UTT][ MSE = 0.1928 | LCC = 0.9516 | SRCC = 0.9518 ] +INFO:root:Average loss=0.08700429818034172 +INFO:root: +[Valid][7][UTT][ MSE = 0.2164 | LCC = 0.9515 | SRCC = 0.9517 ] +INFO:root:Average loss=0.08100976103544236 +INFO:root: +[Valid][8][UTT][ MSE = 0.2012 | LCC = 0.9536 | SRCC = 0.9538 ] +INFO:root:Average loss=0.07417283681035042 +INFO:root: +[Valid][9][UTT][ MSE = 0.2116 | LCC = 0.9527 | SRCC = 0.9535 ] +INFO:root:Average loss=0.07153912982344628 +INFO:root: +[Valid][10][UTT][ MSE = 0.2037 | LCC = 0.9531 | SRCC = 0.9536 ] +INFO:root:Average loss=0.06462233510613441 +INFO:root: +[Valid][11][UTT][ MSE = 0.2263 | LCC = 0.9494 | SRCC = 0.9488 ] +INFO:root:Average loss=0.057913041323423384 +INFO:root: +[Valid][12][UTT][ MSE = 0.1969 | LCC = 0.9528 | SRCC = 0.9532 ] +INFO:root:Average loss=0.05770887705683708 +INFO:root: +[Valid][13][UTT][ MSE = 0.1990 | LCC = 0.9521 | SRCC = 0.9517 ] +INFO:root:Average loss=0.05253556141257286 +INFO:root: +[Valid][14][UTT][ MSE = 0.1969 | LCC = 0.9538 | SRCC = 0.9532 ] +INFO:root:Average loss=0.04745512790977955 +INFO:root: +[Valid][15][UTT][ MSE = 0.1950 | LCC = 0.9536 | SRCC = 0.9536 ] +INFO:root:Average loss=0.04589810912311077 +INFO:root: +[Valid][16][UTT][ MSE = 0.2100 | LCC = 0.9524 | SRCC = 0.9519 ] +INFO:root:Average loss=0.04578298930823803 +INFO:root: +[Valid][17][UTT][ MSE = 0.1948 | LCC = 0.9542 | SRCC = 0.9539 ] +INFO:root:Average loss=0.042795859038829806 +INFO:root: +[Valid][18][UTT][ MSE = 0.2192 | LCC = 0.9524 | SRCC = 0.9528 ] +INFO:root:Average loss=0.04044201326370239 +INFO:root: +[Valid][19][UTT][ MSE = 0.2238 | LCC = 0.9525 | SRCC = 0.9519 ] +INFO:root:Average loss=0.040515239864587785 +INFO:root: +[Valid][20][UTT][ MSE = 0.2007 | LCC = 0.9542 | SRCC = 0.9543 ] +INFO:root:Average loss=0.035534782722592355 +INFO:root: +[Valid][21][UTT][ MSE = 0.1993 | LCC = 0.9546 | SRCC = 0.9549 ] +INFO:root:Average loss=0.035052238389849666 +INFO:root: +[Valid][22][UTT][ MSE = 0.2121 | LCC = 0.9531 | SRCC = 0.9542 ] +INFO:root:Average loss=0.03301692436635494 +INFO:root: +[Valid][23][UTT][ MSE = 0.1934 | LCC = 0.9546 | SRCC = 0.9543 ] +INFO:root:Average loss=0.03143006950616836 +INFO:root: +[Valid][24][UTT][ MSE = 0.1990 | LCC = 0.9548 | SRCC = 0.9551 ] +INFO:root:Average loss=0.03021784929931164 +INFO:root: +[Valid][25][UTT][ MSE = 0.2032 | LCC = 0.9556 | SRCC = 0.9554 ] +INFO:root:Average loss=0.030638257026672364 +INFO:root: +[Valid][26][UTT][ MSE = 0.1889 | LCC = 0.9533 | SRCC = 0.9534 ] +INFO:root:Average loss=0.03163348629325628 +INFO:root: +[Valid][27][UTT][ MSE = 0.1907 | LCC = 0.9554 | SRCC = 0.9558 ] +INFO:root:Average loss=0.028707700200378893 +INFO:root: +[Valid][28][UTT][ MSE = 0.1913 | LCC = 0.9542 | SRCC = 0.9540 ] +INFO:root:Average loss=0.02709523904323578 +INFO:root: +[Valid][29][UTT][ MSE = 0.1975 | LCC = 0.9548 | SRCC = 0.9550 ] +INFO:root:Average loss=0.0258325876891613 +INFO:root: +[Valid][30][UTT][ MSE = 0.2122 | LCC = 0.9545 | SRCC = 0.9548 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1994 | LCC = 0.9552 | SRCC = 0.9562 ] diff --git a/results_tencent/w2v2_base/layer7/run3/config.gin b/results_tencent/w2v2_base/layer7/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer7/run3/model.pt b/results_tencent/w2v2_base/layer7/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..12c812a2ce972074a212b88496d107b05874635d --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19d8d5c2482ee1606652e9664fde8ef63a386380417f46cebd770f6aaab62d0c +size 2747979 diff --git a/results_tencent/w2v2_base/layer7/run3/model_best.pt b/results_tencent/w2v2_base/layer7/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..821fec9f034eb707fed45522f22e6d0c94284068 --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:504934eddfe24b01b3918b6bb6c88a7893fcc8753a476b358bf8383a6d612dd7 +size 2749702 diff --git a/results_tencent/w2v2_base/layer7/run3/train.log b/results_tencent/w2v2_base/layer7/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f1901970e4d1667466419e8da841a260b6d86261 --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4406893171072006 +INFO:root: +[Valid][0][UTT][ MSE = 0.3287 | LCC = 0.9309 | SRCC = 0.9335 ] +INFO:root:Average loss=0.17560335856676101 +INFO:root: +[Valid][1][UTT][ MSE = 0.2701 | LCC = 0.9441 | SRCC = 0.9462 ] +INFO:root:Average loss=0.1519721880555153 +INFO:root: +[Valid][2][UTT][ MSE = 0.2280 | LCC = 0.9440 | SRCC = 0.9450 ] +INFO:root:Average loss=0.1361408366560936 +INFO:root: +[Valid][3][UTT][ MSE = 0.3587 | LCC = 0.9442 | SRCC = 0.9500 ] +INFO:root:Average loss=0.11595060175657272 +INFO:root: +[Valid][4][UTT][ MSE = 0.2320 | LCC = 0.9521 | SRCC = 0.9531 ] +INFO:root:Average loss=0.10603981885313987 +INFO:root: +[Valid][5][UTT][ MSE = 0.2093 | LCC = 0.9532 | SRCC = 0.9539 ] +INFO:root:Average loss=0.09867864143848419 +INFO:root: +[Valid][6][UTT][ MSE = 0.2259 | LCC = 0.9539 | SRCC = 0.9544 ] +INFO:root:Average loss=0.09016418758034705 +INFO:root: +[Valid][7][UTT][ MSE = 0.2356 | LCC = 0.9528 | SRCC = 0.9534 ] +INFO:root:Average loss=0.08094170281291008 +INFO:root: +[Valid][8][UTT][ MSE = 0.2382 | LCC = 0.9530 | SRCC = 0.9544 ] +INFO:root:Average loss=0.07695983800292015 +INFO:root: +[Valid][9][UTT][ MSE = 0.1918 | LCC = 0.9543 | SRCC = 0.9542 ] +INFO:root:Average loss=0.07245738351345063 +INFO:root: +[Valid][10][UTT][ MSE = 0.2355 | LCC = 0.9531 | SRCC = 0.9539 ] +INFO:root:Average loss=0.06589436814188958 +INFO:root: +[Valid][11][UTT][ MSE = 0.2051 | LCC = 0.9538 | SRCC = 0.9536 ] +INFO:root:Average loss=0.06329853758215905 +INFO:root: +[Valid][12][UTT][ MSE = 0.2076 | LCC = 0.9551 | SRCC = 0.9554 ] +INFO:root:Average loss=0.0586172348856926 +INFO:root: +[Valid][13][UTT][ MSE = 0.2350 | LCC = 0.9544 | SRCC = 0.9554 ] +INFO:root:Average loss=0.05812856721878052 +INFO:root: +[Valid][14][UTT][ MSE = 0.1999 | LCC = 0.9544 | SRCC = 0.9540 ] +INFO:root:Average loss=0.05389929839968682 +INFO:root: +[Valid][15][UTT][ MSE = 0.1841 | LCC = 0.9544 | SRCC = 0.9544 ] +INFO:root:Average loss=0.05016003900766373 +INFO:root: +[Valid][16][UTT][ MSE = 0.2127 | LCC = 0.9532 | SRCC = 0.9535 ] +INFO:root:Average loss=0.04601322814822197 +INFO:root: +[Valid][17][UTT][ MSE = 0.1753 | LCC = 0.9535 | SRCC = 0.9526 ] +INFO:root:Average loss=0.04578034481406212 +INFO:root: +[Valid][18][UTT][ MSE = 0.2208 | LCC = 0.9522 | SRCC = 0.9532 ] +INFO:root:Average loss=0.04403518083691597 +INFO:root: +[Valid][19][UTT][ MSE = 0.2367 | LCC = 0.9532 | SRCC = 0.9547 ] +INFO:root:Average loss=0.04051751600205898 +INFO:root: +[Valid][20][UTT][ MSE = 0.2060 | LCC = 0.9543 | SRCC = 0.9537 ] +INFO:root:Average loss=0.04137501484155655 +INFO:root: +[Valid][21][UTT][ MSE = 0.1959 | LCC = 0.9537 | SRCC = 0.9535 ] +INFO:root:Average loss=0.036845344454050064 +INFO:root: +[Valid][22][UTT][ MSE = 0.2244 | LCC = 0.9534 | SRCC = 0.9542 ] +INFO:root:Average loss=0.03752871409058571 +INFO:root: +[Valid][23][UTT][ MSE = 0.2095 | LCC = 0.9526 | SRCC = 0.9525 ] +INFO:root:Average loss=0.03276751585304737 +INFO:root: +[Valid][24][UTT][ MSE = 0.2016 | LCC = 0.9546 | SRCC = 0.9541 ] +INFO:root:Average loss=0.03240685963630676 +INFO:root: +[Valid][25][UTT][ MSE = 0.2202 | LCC = 0.9534 | SRCC = 0.9541 ] +INFO:root:Average loss=0.030457565531134604 +INFO:root: +[Valid][26][UTT][ MSE = 0.2170 | LCC = 0.9526 | SRCC = 0.9533 ] +INFO:root:Average loss=0.02987356451153755 +INFO:root: +[Valid][27][UTT][ MSE = 0.2422 | LCC = 0.9528 | SRCC = 0.9548 ] +INFO:root:Average loss=0.02946445609629154 +INFO:root: +[Valid][28][UTT][ MSE = 0.2198 | LCC = 0.9544 | SRCC = 0.9541 ] +INFO:root:Average loss=0.02678786487877369 +INFO:root: +[Valid][29][UTT][ MSE = 0.2154 | LCC = 0.9548 | SRCC = 0.9551 ] +INFO:root:Average loss=0.028733922883868217 +INFO:root: +[Valid][30][UTT][ MSE = 0.2307 | LCC = 0.9536 | SRCC = 0.9548 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2186 | LCC = 0.9523 | SRCC = 0.9542 ] diff --git a/results_tencent/w2v2_base/layer7/run4/config.gin b/results_tencent/w2v2_base/layer7/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer7/run4/model.pt b/results_tencent/w2v2_base/layer7/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9e4f54da835e9c9de6d456eb1e7b60509696a1bc --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0367e03bb16bf36bc2c7424ac36c6ad7d78d78878fdb7c6a6866ebd1c2a4f26a +size 2747979 diff --git a/results_tencent/w2v2_base/layer7/run4/model_best.pt b/results_tencent/w2v2_base/layer7/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..27e6b4913585e7fa27b05a2382aac5048d6b2957 --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af9666f9a3030b8852a4519dba5dd14b30c504a8dda6f2c4b510bdef0aa29b7d +size 2749702 diff --git a/results_tencent/w2v2_base/layer7/run4/train.log b/results_tencent/w2v2_base/layer7/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..8a3757799b155a093e1f1b4f9a29733c10342143 --- /dev/null +++ b/results_tencent/w2v2_base/layer7/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4809232792854309 +INFO:root: +[Valid][0][UTT][ MSE = 0.3130 | LCC = 0.9238 | SRCC = 0.9247 ] +INFO:root:Average loss=0.1844312646985054 +INFO:root: +[Valid][1][UTT][ MSE = 0.2481 | LCC = 0.9412 | SRCC = 0.9429 ] +INFO:root:Average loss=0.14870794475078583 +INFO:root: +[Valid][2][UTT][ MSE = 0.2742 | LCC = 0.9471 | SRCC = 0.9488 ] +INFO:root:Average loss=0.1318001142144203 +INFO:root: +[Valid][3][UTT][ MSE = 0.2248 | LCC = 0.9465 | SRCC = 0.9491 ] +INFO:root:Average loss=0.11555240005254745 +INFO:root: +[Valid][4][UTT][ MSE = 0.2260 | LCC = 0.9505 | SRCC = 0.9521 ] +INFO:root:Average loss=0.10465169924497604 +INFO:root: +[Valid][5][UTT][ MSE = 0.2208 | LCC = 0.9505 | SRCC = 0.9510 ] +INFO:root:Average loss=0.09424411118030548 +INFO:root: +[Valid][6][UTT][ MSE = 0.2171 | LCC = 0.9516 | SRCC = 0.9524 ] +INFO:root:Average loss=0.08702895280718803 +INFO:root: +[Valid][7][UTT][ MSE = 0.2193 | LCC = 0.9508 | SRCC = 0.9513 ] +INFO:root:Average loss=0.08302406790852547 +INFO:root: +[Valid][8][UTT][ MSE = 0.2246 | LCC = 0.9537 | SRCC = 0.9546 ] +INFO:root:Average loss=0.07632618150115013 +INFO:root: +[Valid][9][UTT][ MSE = 0.2379 | LCC = 0.9538 | SRCC = 0.9547 ] +INFO:root:Average loss=0.06997029420733451 +INFO:root: +[Valid][10][UTT][ MSE = 0.2287 | LCC = 0.9534 | SRCC = 0.9541 ] +INFO:root:Average loss=0.06463650313019753 +INFO:root: +[Valid][11][UTT][ MSE = 0.2062 | LCC = 0.9543 | SRCC = 0.9549 ] +INFO:root:Average loss=0.06314782711863517 +INFO:root: +[Valid][12][UTT][ MSE = 0.1965 | LCC = 0.9539 | SRCC = 0.9541 ] +INFO:root:Average loss=0.05961649397015572 +INFO:root: +[Valid][13][UTT][ MSE = 0.2034 | LCC = 0.9519 | SRCC = 0.9524 ] +INFO:root:Average loss=0.056063036799430845 +INFO:root: +[Valid][14][UTT][ MSE = 0.1843 | LCC = 0.9527 | SRCC = 0.9538 ] +INFO:root:Average loss=0.050948335081338884 +INFO:root: +[Valid][15][UTT][ MSE = 0.2011 | LCC = 0.9550 | SRCC = 0.9558 ] +INFO:root:Average loss=0.04730568416416645 +INFO:root: +[Valid][16][UTT][ MSE = 0.2005 | LCC = 0.9545 | SRCC = 0.9543 ] +INFO:root:Average loss=0.0467319258749485 +INFO:root: +[Valid][17][UTT][ MSE = 0.2121 | LCC = 0.9559 | SRCC = 0.9561 ] +INFO:root:Average loss=0.04379575257003308 +INFO:root: +[Valid][18][UTT][ MSE = 0.2409 | LCC = 0.9540 | SRCC = 0.9555 ] +INFO:root:Average loss=0.040937366157770154 +INFO:root: +[Valid][19][UTT][ MSE = 0.2157 | LCC = 0.9554 | SRCC = 0.9557 ] +INFO:root:Average loss=0.03987592108547688 +INFO:root: +[Valid][20][UTT][ MSE = 0.2156 | LCC = 0.9558 | SRCC = 0.9561 ] +INFO:root:Average loss=0.037456769078969956 +INFO:root: +[Valid][21][UTT][ MSE = 0.2045 | LCC = 0.9555 | SRCC = 0.9560 ] +INFO:root:Average loss=0.03658824333548546 +INFO:root: +[Valid][22][UTT][ MSE = 0.1998 | LCC = 0.9559 | SRCC = 0.9560 ] +INFO:root:Average loss=0.036822371274232865 +INFO:root: +[Valid][23][UTT][ MSE = 0.2142 | LCC = 0.9558 | SRCC = 0.9563 ] +INFO:root:Average loss=0.0326299899071455 +INFO:root: +[Valid][24][UTT][ MSE = 0.2086 | LCC = 0.9560 | SRCC = 0.9562 ] +INFO:root:Average loss=0.03188420332968235 +INFO:root: +[Valid][25][UTT][ MSE = 0.1947 | LCC = 0.9552 | SRCC = 0.9556 ] +INFO:root:Average loss=0.03133970728516579 +INFO:root: +[Valid][26][UTT][ MSE = 0.1860 | LCC = 0.9531 | SRCC = 0.9537 ] +INFO:root:Average loss=0.0311632065474987 +INFO:root: +[Valid][27][UTT][ MSE = 0.2169 | LCC = 0.9555 | SRCC = 0.9561 ] +INFO:root:Average loss=0.029340460374951364 +INFO:root: +[Valid][28][UTT][ MSE = 0.2011 | LCC = 0.9551 | SRCC = 0.9554 ] +INFO:root:Average loss=0.028608019188046455 +INFO:root: +[Valid][29][UTT][ MSE = 0.2143 | LCC = 0.9529 | SRCC = 0.9529 ] +INFO:root:Average loss=0.02772654299438 +INFO:root: +[Valid][30][UTT][ MSE = 0.2257 | LCC = 0.9553 | SRCC = 0.9555 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2187 | LCC = 0.9539 | SRCC = 0.9555 ] diff --git a/results_tencent/w2v2_base/layer8/run0/config.gin b/results_tencent/w2v2_base/layer8/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer8/run0/model.pt b/results_tencent/w2v2_base/layer8/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d36e71999f0404435564baccf6e741cd92da83ab --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9d021f74fb5c2753e4a51220c0b224b6b899e1b9052c31ee8ed9dda44ae80f3a +size 2747979 diff --git a/results_tencent/w2v2_base/layer8/run0/model_best.pt b/results_tencent/w2v2_base/layer8/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..82339fb6c3ab26ff1663330e19130abdd020e5ce --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e2f32df570accfd7e8e6f184f769d9e0e825d564bda58b6c2d206251f81bd19 +size 2749702 diff --git a/results_tencent/w2v2_base/layer8/run0/train.log b/results_tencent/w2v2_base/layer8/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..24caf3b9251a2422d54ce7ef9c59278552d6a5b9 --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4289683861732483 +INFO:root: +[Valid][0][UTT][ MSE = 0.2858 | LCC = 0.9252 | SRCC = 0.9260 ] +INFO:root:Average loss=0.17718233221769333 +INFO:root: +[Valid][1][UTT][ MSE = 0.2628 | LCC = 0.9447 | SRCC = 0.9462 ] +INFO:root:Average loss=0.14373777765035628 +INFO:root: +[Valid][2][UTT][ MSE = 0.2641 | LCC = 0.9503 | SRCC = 0.9519 ] +INFO:root:Average loss=0.12599315190315247 +INFO:root: +[Valid][3][UTT][ MSE = 0.2322 | LCC = 0.9510 | SRCC = 0.9520 ] +INFO:root:Average loss=0.11577539730072021 +INFO:root: +[Valid][4][UTT][ MSE = 0.2475 | LCC = 0.9526 | SRCC = 0.9535 ] +INFO:root:Average loss=0.10802377831935883 +INFO:root: +[Valid][5][UTT][ MSE = 0.2465 | LCC = 0.9513 | SRCC = 0.9521 ] +INFO:root:Average loss=0.09810008764266968 +INFO:root: +[Valid][6][UTT][ MSE = 0.2038 | LCC = 0.9546 | SRCC = 0.9556 ] +INFO:root:Average loss=0.08975399890542031 +INFO:root: +[Valid][7][UTT][ MSE = 0.2244 | LCC = 0.9548 | SRCC = 0.9567 ] +INFO:root:Average loss=0.0906244504749775 +INFO:root: +[Valid][8][UTT][ MSE = 0.2070 | LCC = 0.9565 | SRCC = 0.9566 ] +INFO:root:Average loss=0.08100704994797707 +INFO:root: +[Valid][9][UTT][ MSE = 0.1976 | LCC = 0.9574 | SRCC = 0.9577 ] +INFO:root:Average loss=0.07615425863862038 +INFO:root: +[Valid][10][UTT][ MSE = 0.2097 | LCC = 0.9577 | SRCC = 0.9582 ] +INFO:root:Average loss=0.07391305539011955 +INFO:root: +[Valid][11][UTT][ MSE = 0.1884 | LCC = 0.9561 | SRCC = 0.9561 ] +INFO:root:Average loss=0.06644571506977082 +INFO:root: +[Valid][12][UTT][ MSE = 0.1816 | LCC = 0.9579 | SRCC = 0.9577 ] +INFO:root:Average loss=0.06169328954815865 +INFO:root: +[Valid][13][UTT][ MSE = 0.2063 | LCC = 0.9584 | SRCC = 0.9584 ] +INFO:root:Average loss=0.06132322189211845 +INFO:root: +[Valid][14][UTT][ MSE = 0.1894 | LCC = 0.9541 | SRCC = 0.9538 ] +INFO:root:Average loss=0.05688429492712021 +INFO:root: +[Valid][15][UTT][ MSE = 0.1958 | LCC = 0.9581 | SRCC = 0.9577 ] +INFO:root:Average loss=0.05667714941501618 +INFO:root: +[Valid][16][UTT][ MSE = 0.2084 | LCC = 0.9570 | SRCC = 0.9585 ] +INFO:root:Average loss=0.05223903775215149 +INFO:root: +[Valid][17][UTT][ MSE = 0.1950 | LCC = 0.9567 | SRCC = 0.9570 ] +INFO:root:Average loss=0.051306282848119736 +INFO:root: +[Valid][18][UTT][ MSE = 0.1961 | LCC = 0.9597 | SRCC = 0.9597 ] +INFO:root:Average loss=0.05022509175539017 +INFO:root: +[Valid][19][UTT][ MSE = 0.1974 | LCC = 0.9584 | SRCC = 0.9579 ] +INFO:root:Average loss=0.045317441210150716 +INFO:root: +[Valid][20][UTT][ MSE = 0.1933 | LCC = 0.9579 | SRCC = 0.9572 ] +INFO:root:Average loss=0.04357644686102867 +INFO:root: +[Valid][21][UTT][ MSE = 0.1791 | LCC = 0.9579 | SRCC = 0.9577 ] +INFO:root:Average loss=0.042120617017149926 +INFO:root: +[Valid][22][UTT][ MSE = 0.1943 | LCC = 0.9579 | SRCC = 0.9576 ] +INFO:root:Average loss=0.03970662182569504 +INFO:root: +[Valid][23][UTT][ MSE = 0.2037 | LCC = 0.9571 | SRCC = 0.9576 ] +INFO:root:Average loss=0.03844494639337063 +INFO:root: +[Valid][24][UTT][ MSE = 0.1858 | LCC = 0.9581 | SRCC = 0.9577 ] +INFO:root:Average loss=0.03819364975392819 +INFO:root: +[Valid][25][UTT][ MSE = 0.1998 | LCC = 0.9580 | SRCC = 0.9580 ] +INFO:root:Average loss=0.035747096732258794 +INFO:root: +[Valid][26][UTT][ MSE = 0.2004 | LCC = 0.9571 | SRCC = 0.9568 ] +INFO:root:Average loss=0.034217322155833244 +INFO:root: +[Valid][27][UTT][ MSE = 0.1740 | LCC = 0.9571 | SRCC = 0.9572 ] +INFO:root:Average loss=0.034713639706373214 +INFO:root: +[Valid][28][UTT][ MSE = 0.1909 | LCC = 0.9568 | SRCC = 0.9563 ] +INFO:root:Average loss=0.031156548023223877 +INFO:root: +[Valid][29][UTT][ MSE = 0.2040 | LCC = 0.9595 | SRCC = 0.9598 ] +INFO:root:Average loss=0.02966235712170601 +INFO:root: +[Valid][30][UTT][ MSE = 0.1991 | LCC = 0.9576 | SRCC = 0.9574 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2039 | LCC = 0.9573 | SRCC = 0.9586 ] diff --git a/results_tencent/w2v2_base/layer8/run1/config.gin b/results_tencent/w2v2_base/layer8/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer8/run1/model.pt b/results_tencent/w2v2_base/layer8/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a1327b567e424bf6a8a6c039a5110abadfb61789 --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b18b6b80987fa20affb506d4394c6bfe1e7e58f90db212f8c5e80bfe8551dd7d +size 2747979 diff --git a/results_tencent/w2v2_base/layer8/run1/model_best.pt b/results_tencent/w2v2_base/layer8/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..df1f587c46e4da29ecc7ce8618cd8a7b25d6d5f5 --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7fc0f21f07ed7a2f87bd3857bf131ebd5647b0a01e3dccf42119f1fa30b90d18 +size 2749702 diff --git a/results_tencent/w2v2_base/layer8/run1/train.log b/results_tencent/w2v2_base/layer8/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..08d230cc1fdc1237b6c0d0537945a11c8a7472f5 --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4618189264535904 +INFO:root: +[Valid][0][UTT][ MSE = 0.3503 | LCC = 0.9282 | SRCC = 0.9302 ] +INFO:root:Average loss=0.18059584134817122 +INFO:root: +[Valid][1][UTT][ MSE = 0.2433 | LCC = 0.9432 | SRCC = 0.9450 ] +INFO:root:Average loss=0.14645611250400542 +INFO:root: +[Valid][2][UTT][ MSE = 0.2502 | LCC = 0.9476 | SRCC = 0.9487 ] +INFO:root:Average loss=0.1266143301129341 +INFO:root: +[Valid][3][UTT][ MSE = 0.2268 | LCC = 0.9518 | SRCC = 0.9531 ] +INFO:root:Average loss=0.11477037143707275 +INFO:root: +[Valid][4][UTT][ MSE = 0.2225 | LCC = 0.9513 | SRCC = 0.9528 ] +INFO:root:Average loss=0.1084843248128891 +INFO:root: +[Valid][5][UTT][ MSE = 0.2242 | LCC = 0.9562 | SRCC = 0.9568 ] +INFO:root:Average loss=0.09747142109274864 +INFO:root: +[Valid][6][UTT][ MSE = 0.2081 | LCC = 0.9561 | SRCC = 0.9569 ] +INFO:root:Average loss=0.09050379201769829 +INFO:root: +[Valid][7][UTT][ MSE = 0.2208 | LCC = 0.9568 | SRCC = 0.9577 ] +INFO:root:Average loss=0.08913270750641823 +INFO:root: +[Valid][8][UTT][ MSE = 0.2283 | LCC = 0.9583 | SRCC = 0.9594 ] +INFO:root:Average loss=0.08098553860187531 +INFO:root: +[Valid][9][UTT][ MSE = 0.2244 | LCC = 0.9576 | SRCC = 0.9580 ] +INFO:root:Average loss=0.07676892986893653 +INFO:root: +[Valid][10][UTT][ MSE = 0.2075 | LCC = 0.9587 | SRCC = 0.9593 ] +INFO:root:Average loss=0.07117529448866844 +INFO:root: +[Valid][11][UTT][ MSE = 0.1851 | LCC = 0.9575 | SRCC = 0.9579 ] +INFO:root:Average loss=0.06660675704479217 +INFO:root: +[Valid][12][UTT][ MSE = 0.2216 | LCC = 0.9575 | SRCC = 0.9577 ] +INFO:root:Average loss=0.06289535439014435 +INFO:root: +[Valid][13][UTT][ MSE = 0.1783 | LCC = 0.9582 | SRCC = 0.9585 ] +INFO:root:Average loss=0.05966844272613525 +INFO:root: +[Valid][14][UTT][ MSE = 0.1989 | LCC = 0.9584 | SRCC = 0.9588 ] +INFO:root:Average loss=0.0592323534488678 +INFO:root: +[Valid][15][UTT][ MSE = 0.2157 | LCC = 0.9553 | SRCC = 0.9556 ] +INFO:root:Average loss=0.05886334756016731 +INFO:root: +[Valid][16][UTT][ MSE = 0.1992 | LCC = 0.9584 | SRCC = 0.9590 ] +INFO:root:Average loss=0.051437876969575885 +INFO:root: +[Valid][17][UTT][ MSE = 0.1776 | LCC = 0.9588 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04998076981306076 +INFO:root: +[Valid][18][UTT][ MSE = 0.1917 | LCC = 0.9593 | SRCC = 0.9598 ] +INFO:root:Average loss=0.04623785658180714 +INFO:root: +[Valid][19][UTT][ MSE = 0.1901 | LCC = 0.9583 | SRCC = 0.9585 ] +INFO:root:Average loss=0.045879228338599205 +INFO:root: +[Valid][20][UTT][ MSE = 0.1892 | LCC = 0.9587 | SRCC = 0.9589 ] +INFO:root:Average loss=0.04323547948896885 +INFO:root: +[Valid][21][UTT][ MSE = 0.1989 | LCC = 0.9572 | SRCC = 0.9576 ] +INFO:root:Average loss=0.04243971291184425 +INFO:root: +[Valid][22][UTT][ MSE = 0.1840 | LCC = 0.9597 | SRCC = 0.9596 ] +INFO:root:Average loss=0.04153646364808083 +INFO:root: +[Valid][23][UTT][ MSE = 0.1720 | LCC = 0.9576 | SRCC = 0.9581 ] +INFO:root:Average loss=0.03878722910583019 +INFO:root: +[Valid][24][UTT][ MSE = 0.2216 | LCC = 0.9574 | SRCC = 0.9573 ] +INFO:root:Average loss=0.03700771646201611 +INFO:root: +[Valid][25][UTT][ MSE = 0.2118 | LCC = 0.9581 | SRCC = 0.9577 ] +INFO:root:Average loss=0.03762073455750942 +INFO:root: +[Valid][26][UTT][ MSE = 0.2020 | LCC = 0.9589 | SRCC = 0.9594 ] +INFO:root:Average loss=0.037272025302052494 +INFO:root: +[Valid][27][UTT][ MSE = 0.2048 | LCC = 0.9591 | SRCC = 0.9588 ] +INFO:root:Average loss=0.03272280806303024 +INFO:root: +[Valid][28][UTT][ MSE = 0.1845 | LCC = 0.9572 | SRCC = 0.9581 ] +INFO:root:Average loss=0.03116894793510437 +INFO:root: +[Valid][29][UTT][ MSE = 0.1911 | LCC = 0.9573 | SRCC = 0.9574 ] +INFO:root:Average loss=0.030406037852168083 +INFO:root: +[Valid][30][UTT][ MSE = 0.2163 | LCC = 0.9589 | SRCC = 0.9587 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1894 | LCC = 0.9578 | SRCC = 0.9584 ] diff --git a/results_tencent/w2v2_base/layer8/run2/config.gin b/results_tencent/w2v2_base/layer8/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer8/run2/model.pt b/results_tencent/w2v2_base/layer8/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6a445384d5000eabb3a95767519359bcdc132aa3 --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:893f997aee904382c97c09ac0cb6c5b850d82368ae90d3d00d2d8692585cbfbd +size 2747979 diff --git a/results_tencent/w2v2_base/layer8/run2/model_best.pt b/results_tencent/w2v2_base/layer8/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..befcb02ae412d1c3cd24d337480bf317419e9c98 --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f2a54cf89538dcb9b6de00173d81dcd2835acb6cba9e29160f4a322114f5031b +size 2749702 diff --git a/results_tencent/w2v2_base/layer8/run2/train.log b/results_tencent/w2v2_base/layer8/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a525ec13870780d4dd926f9ad06c193a0fbc4702 --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4571918870210648 +INFO:root: +[Valid][0][UTT][ MSE = 0.3367 | LCC = 0.9243 | SRCC = 0.9256 ] +INFO:root:Average loss=0.1923685957789421 +INFO:root: +[Valid][1][UTT][ MSE = 0.2409 | LCC = 0.9420 | SRCC = 0.9427 ] +INFO:root:Average loss=0.15242068696022035 +INFO:root: +[Valid][2][UTT][ MSE = 0.2591 | LCC = 0.9495 | SRCC = 0.9504 ] +INFO:root:Average loss=0.1315249847173691 +INFO:root: +[Valid][3][UTT][ MSE = 0.2388 | LCC = 0.9520 | SRCC = 0.9526 ] +INFO:root:Average loss=0.11755440819263459 +INFO:root: +[Valid][4][UTT][ MSE = 0.2470 | LCC = 0.9532 | SRCC = 0.9549 ] +INFO:root:Average loss=0.10987194654345513 +INFO:root: +[Valid][5][UTT][ MSE = 0.2246 | LCC = 0.9560 | SRCC = 0.9567 ] +INFO:root:Average loss=0.09856676465272904 +INFO:root: +[Valid][6][UTT][ MSE = 0.2074 | LCC = 0.9570 | SRCC = 0.9574 ] +INFO:root:Average loss=0.09061972412467002 +INFO:root: +[Valid][7][UTT][ MSE = 0.2043 | LCC = 0.9575 | SRCC = 0.9579 ] +INFO:root:Average loss=0.08364086383581161 +INFO:root: +[Valid][8][UTT][ MSE = 0.2352 | LCC = 0.9557 | SRCC = 0.9559 ] +INFO:root:Average loss=0.07972354233264924 +INFO:root: +[Valid][9][UTT][ MSE = 0.2317 | LCC = 0.9574 | SRCC = 0.9595 ] +INFO:root:Average loss=0.07436887785792351 +INFO:root: +[Valid][10][UTT][ MSE = 0.2107 | LCC = 0.9584 | SRCC = 0.9585 ] +INFO:root:Average loss=0.07215715453028679 +INFO:root: +[Valid][11][UTT][ MSE = 0.2445 | LCC = 0.9581 | SRCC = 0.9594 ] +INFO:root:Average loss=0.06735939311981201 +INFO:root: +[Valid][12][UTT][ MSE = 0.1927 | LCC = 0.9587 | SRCC = 0.9593 ] +INFO:root:Average loss=0.06352572214603425 +INFO:root: +[Valid][13][UTT][ MSE = 0.2017 | LCC = 0.9585 | SRCC = 0.9590 ] +INFO:root:Average loss=0.058492394030094146 +INFO:root: +[Valid][14][UTT][ MSE = 0.2092 | LCC = 0.9564 | SRCC = 0.9572 ] +INFO:root:Average loss=0.05622906669974327 +INFO:root: +[Valid][15][UTT][ MSE = 0.2034 | LCC = 0.9586 | SRCC = 0.9584 ] +INFO:root:Average loss=0.051605295151472094 +INFO:root: +[Valid][16][UTT][ MSE = 0.2106 | LCC = 0.9581 | SRCC = 0.9589 ] +INFO:root:Average loss=0.05433893357217312 +INFO:root: +[Valid][17][UTT][ MSE = 0.2016 | LCC = 0.9560 | SRCC = 0.9565 ] +INFO:root:Average loss=0.050246870175004 +INFO:root: +[Valid][18][UTT][ MSE = 0.1834 | LCC = 0.9581 | SRCC = 0.9586 ] +INFO:root:Average loss=0.04742002008855343 +INFO:root: +[Valid][19][UTT][ MSE = 0.2329 | LCC = 0.9562 | SRCC = 0.9583 ] +INFO:root:Average loss=0.04636184099316597 +INFO:root: +[Valid][20][UTT][ MSE = 0.1990 | LCC = 0.9584 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04210781982541084 +INFO:root: +[Valid][21][UTT][ MSE = 0.2059 | LCC = 0.9585 | SRCC = 0.9593 ] +INFO:root:Average loss=0.042988675892353057 +INFO:root: +[Valid][22][UTT][ MSE = 0.2210 | LCC = 0.9583 | SRCC = 0.9589 ] +INFO:root:Average loss=0.03993924188613891 +INFO:root: +[Valid][23][UTT][ MSE = 0.1922 | LCC = 0.9594 | SRCC = 0.9594 ] +INFO:root:Average loss=0.03664241774380207 +INFO:root: +[Valid][24][UTT][ MSE = 0.1940 | LCC = 0.9582 | SRCC = 0.9587 ] +INFO:root:Average loss=0.03750536341965199 +INFO:root: +[Valid][25][UTT][ MSE = 0.2124 | LCC = 0.9569 | SRCC = 0.9574 ] +INFO:root:Average loss=0.035882012993097304 +INFO:root: +[Valid][26][UTT][ MSE = 0.2003 | LCC = 0.9581 | SRCC = 0.9583 ] +INFO:root:Average loss=0.03407710583508015 +INFO:root: +[Valid][27][UTT][ MSE = 0.2171 | LCC = 0.9572 | SRCC = 0.9582 ] +INFO:root:Average loss=0.033856821551918985 +INFO:root: +[Valid][28][UTT][ MSE = 0.2048 | LCC = 0.9568 | SRCC = 0.9575 ] +INFO:root:Average loss=0.03176658008992672 +INFO:root: +[Valid][29][UTT][ MSE = 0.2018 | LCC = 0.9564 | SRCC = 0.9567 ] +INFO:root:Average loss=0.03179532271623611 +INFO:root: +[Valid][30][UTT][ MSE = 0.2020 | LCC = 0.9578 | SRCC = 0.9588 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2020 | LCC = 0.9554 | SRCC = 0.9561 ] diff --git a/results_tencent/w2v2_base/layer8/run3/config.gin b/results_tencent/w2v2_base/layer8/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer8/run3/model.pt b/results_tencent/w2v2_base/layer8/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bc0f310041bc8dee71f2a88a3f26ce2ef6276c45 --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f975f4e351a05d5712d9105cfeff008f63ce0a8b76db6833d03a87c306bf6ab9 +size 2747979 diff --git a/results_tencent/w2v2_base/layer8/run3/model_best.pt b/results_tencent/w2v2_base/layer8/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..76d88e6492f820a7c3d3b269c8b0ec4058ee9263 --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1ce733307ab242723c5aaac8aa13e97dfd274d8cadaa1e3df8407c2a3c1e1e76 +size 2749702 diff --git a/results_tencent/w2v2_base/layer8/run3/train.log b/results_tencent/w2v2_base/layer8/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..266c4fa7ed10033b5c3fbe81e0569f51749a0cec --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.48033323836326597 +INFO:root: +[Valid][0][UTT][ MSE = 0.2902 | LCC = 0.9263 | SRCC = 0.9267 ] +INFO:root:Average loss=0.18676458847522737 +INFO:root: +[Valid][1][UTT][ MSE = 0.2259 | LCC = 0.9407 | SRCC = 0.9435 ] +INFO:root:Average loss=0.14681020760536193 +INFO:root: +[Valid][2][UTT][ MSE = 0.2078 | LCC = 0.9509 | SRCC = 0.9519 ] +INFO:root:Average loss=0.12661688017845155 +INFO:root: +[Valid][3][UTT][ MSE = 0.2267 | LCC = 0.9522 | SRCC = 0.9528 ] +INFO:root:Average loss=0.11214682722091675 +INFO:root: +[Valid][4][UTT][ MSE = 0.2066 | LCC = 0.9548 | SRCC = 0.9553 ] +INFO:root:Average loss=0.10369019836187363 +INFO:root: +[Valid][5][UTT][ MSE = 0.1793 | LCC = 0.9548 | SRCC = 0.9556 ] +INFO:root:Average loss=0.09879981750249862 +INFO:root: +[Valid][6][UTT][ MSE = 0.2409 | LCC = 0.9564 | SRCC = 0.9566 ] +INFO:root:Average loss=0.0915748071372509 +INFO:root: +[Valid][7][UTT][ MSE = 0.1978 | LCC = 0.9575 | SRCC = 0.9593 ] +INFO:root:Average loss=0.08450850710272789 +INFO:root: +[Valid][8][UTT][ MSE = 0.1806 | LCC = 0.9577 | SRCC = 0.9580 ] +INFO:root:Average loss=0.07952978119254112 +INFO:root: +[Valid][9][UTT][ MSE = 0.1863 | LCC = 0.9576 | SRCC = 0.9575 ] +INFO:root:Average loss=0.07532200601696969 +INFO:root: +[Valid][10][UTT][ MSE = 0.2033 | LCC = 0.9586 | SRCC = 0.9593 ] +INFO:root:Average loss=0.07098658660054206 +INFO:root: +[Valid][11][UTT][ MSE = 0.2060 | LCC = 0.9578 | SRCC = 0.9580 ] +INFO:root:Average loss=0.06727003839612007 +INFO:root: +[Valid][12][UTT][ MSE = 0.2090 | LCC = 0.9582 | SRCC = 0.9580 ] +INFO:root:Average loss=0.06162134736776352 +INFO:root: +[Valid][13][UTT][ MSE = 0.2058 | LCC = 0.9568 | SRCC = 0.9570 ] +INFO:root:Average loss=0.05789002111554146 +INFO:root: +[Valid][14][UTT][ MSE = 0.1866 | LCC = 0.9588 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05829712530970573 +INFO:root: +[Valid][15][UTT][ MSE = 0.1914 | LCC = 0.9594 | SRCC = 0.9598 ] +INFO:root:Average loss=0.05503104443848133 +INFO:root: +[Valid][16][UTT][ MSE = 0.1886 | LCC = 0.9590 | SRCC = 0.9588 ] +INFO:root:Average loss=0.05325146661698818 +INFO:root: +[Valid][17][UTT][ MSE = 0.2021 | LCC = 0.9566 | SRCC = 0.9572 ] +INFO:root:Average loss=0.051130649790167806 +INFO:root: +[Valid][18][UTT][ MSE = 0.1642 | LCC = 0.9557 | SRCC = 0.9573 ] +INFO:root:Average loss=0.04669809259474277 +INFO:root: +[Valid][19][UTT][ MSE = 0.1874 | LCC = 0.9586 | SRCC = 0.9582 ] +INFO:root:Average loss=0.04469840393960476 +INFO:root: +[Valid][20][UTT][ MSE = 0.1747 | LCC = 0.9577 | SRCC = 0.9577 ] +INFO:root:Average loss=0.042196161642670635 +INFO:root: +[Valid][21][UTT][ MSE = 0.1870 | LCC = 0.9560 | SRCC = 0.9567 ] +INFO:root:Average loss=0.04159903322160244 +INFO:root: +[Valid][22][UTT][ MSE = 0.1921 | LCC = 0.9559 | SRCC = 0.9556 ] +INFO:root:Average loss=0.041575494840741155 +INFO:root: +[Valid][23][UTT][ MSE = 0.1808 | LCC = 0.9587 | SRCC = 0.9590 ] +INFO:root:Average loss=0.038859107673168185 +INFO:root: +[Valid][24][UTT][ MSE = 0.1879 | LCC = 0.9585 | SRCC = 0.9588 ] +INFO:root:Average loss=0.03808003666996956 +INFO:root: +[Valid][25][UTT][ MSE = 0.1791 | LCC = 0.9586 | SRCC = 0.9591 ] +INFO:root:Average loss=0.03570889054238796 +INFO:root: +[Valid][26][UTT][ MSE = 0.1882 | LCC = 0.9574 | SRCC = 0.9574 ] +INFO:root:Average loss=0.034045323953032496 +INFO:root: +[Valid][27][UTT][ MSE = 0.1958 | LCC = 0.9584 | SRCC = 0.9580 ] +INFO:root:Average loss=0.034114502280950545 +INFO:root: +[Valid][28][UTT][ MSE = 0.1857 | LCC = 0.9581 | SRCC = 0.9583 ] +INFO:root:Average loss=0.03154521468281746 +INFO:root: +[Valid][29][UTT][ MSE = 0.2009 | LCC = 0.9579 | SRCC = 0.9577 ] +INFO:root:Average loss=0.030369196161627768 +INFO:root: +[Valid][30][UTT][ MSE = 0.1906 | LCC = 0.9584 | SRCC = 0.9582 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1996 | LCC = 0.9553 | SRCC = 0.9571 ] diff --git a/results_tencent/w2v2_base/layer8/run4/config.gin b/results_tencent/w2v2_base/layer8/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer8/run4/model.pt b/results_tencent/w2v2_base/layer8/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7381fb43e981a75dd36710f03038131f2722e87d --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ab441be99b05f9d3abb1600e8a0794e77b4cabe0665a2f3db127b0c80f941fda +size 2747979 diff --git a/results_tencent/w2v2_base/layer8/run4/model_best.pt b/results_tencent/w2v2_base/layer8/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9dc98304f70fa65e3b1fcab757fcdb75c7c1304b --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7c8adf2795583e9cd3461d872b4e279cde2c16997ba479d6c6eae119d35ccb3f +size 2749702 diff --git a/results_tencent/w2v2_base/layer8/run4/train.log b/results_tencent/w2v2_base/layer8/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c16bd80c71bc792b125d920297af3c3cba3f9848 --- /dev/null +++ b/results_tencent/w2v2_base/layer8/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.457505961894989 +INFO:root: +[Valid][0][UTT][ MSE = 0.2770 | LCC = 0.9237 | SRCC = 0.9268 ] +INFO:root:Average loss=0.17953538393974305 +INFO:root: +[Valid][1][UTT][ MSE = 0.2331 | LCC = 0.9356 | SRCC = 0.9381 ] +INFO:root:Average loss=0.1465079179406166 +INFO:root: +[Valid][2][UTT][ MSE = 0.2256 | LCC = 0.9475 | SRCC = 0.9480 ] +INFO:root:Average loss=0.12579864954948425 +INFO:root: +[Valid][3][UTT][ MSE = 0.2058 | LCC = 0.9514 | SRCC = 0.9530 ] +INFO:root:Average loss=0.113373714864254 +INFO:root: +[Valid][4][UTT][ MSE = 0.2000 | LCC = 0.9536 | SRCC = 0.9545 ] +INFO:root:Average loss=0.10284956896305084 +INFO:root: +[Valid][5][UTT][ MSE = 0.1986 | LCC = 0.9551 | SRCC = 0.9558 ] +INFO:root:Average loss=0.09599802404642105 +INFO:root: +[Valid][6][UTT][ MSE = 0.2538 | LCC = 0.9553 | SRCC = 0.9573 ] +INFO:root:Average loss=0.09065581744909286 +INFO:root: +[Valid][7][UTT][ MSE = 0.2012 | LCC = 0.9574 | SRCC = 0.9584 ] +INFO:root:Average loss=0.08363529375195503 +INFO:root: +[Valid][8][UTT][ MSE = 0.1861 | LCC = 0.9531 | SRCC = 0.9546 ] +INFO:root:Average loss=0.07970955124497414 +INFO:root: +[Valid][9][UTT][ MSE = 0.1797 | LCC = 0.9574 | SRCC = 0.9581 ] +INFO:root:Average loss=0.07218611907958984 +INFO:root: +[Valid][10][UTT][ MSE = 0.1929 | LCC = 0.9573 | SRCC = 0.9577 ] +INFO:root:Average loss=0.07023585256934166 +INFO:root: +[Valid][11][UTT][ MSE = 0.1996 | LCC = 0.9556 | SRCC = 0.9563 ] +INFO:root:Average loss=0.06552653476595878 +INFO:root: +[Valid][12][UTT][ MSE = 0.1820 | LCC = 0.9570 | SRCC = 0.9576 ] +INFO:root:Average loss=0.06268525475263595 +INFO:root: +[Valid][13][UTT][ MSE = 0.2029 | LCC = 0.9563 | SRCC = 0.9568 ] +INFO:root:Average loss=0.05810241976380348 +INFO:root: +[Valid][14][UTT][ MSE = 0.1868 | LCC = 0.9584 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05421293756365776 +INFO:root: +[Valid][15][UTT][ MSE = 0.1929 | LCC = 0.9567 | SRCC = 0.9573 ] +INFO:root:Average loss=0.05182148888707161 +INFO:root: +[Valid][16][UTT][ MSE = 0.2027 | LCC = 0.9590 | SRCC = 0.9590 ] +INFO:root:Average loss=0.05279571658372879 +INFO:root: +[Valid][17][UTT][ MSE = 0.2085 | LCC = 0.9562 | SRCC = 0.9570 ] +INFO:root:Average loss=0.0476153354793787 +INFO:root: +[Valid][18][UTT][ MSE = 0.2203 | LCC = 0.9585 | SRCC = 0.9589 ] +INFO:root:Average loss=0.04454228998720646 +INFO:root: +[Valid][19][UTT][ MSE = 0.1811 | LCC = 0.9562 | SRCC = 0.9570 ] +INFO:root:Average loss=0.04465470764040947 +INFO:root: +[Valid][20][UTT][ MSE = 0.1989 | LCC = 0.9586 | SRCC = 0.9584 ] +INFO:root:Average loss=0.041959594517946244 +INFO:root: +[Valid][21][UTT][ MSE = 0.2022 | LCC = 0.9585 | SRCC = 0.9588 ] +INFO:root:Average loss=0.04005330044031143 +INFO:root: +[Valid][22][UTT][ MSE = 0.2009 | LCC = 0.9561 | SRCC = 0.9561 ] +INFO:root:Average loss=0.03974200914800167 +INFO:root: +[Valid][23][UTT][ MSE = 0.2294 | LCC = 0.9524 | SRCC = 0.9532 ] +INFO:root:Average loss=0.037543199732899664 +INFO:root: +[Valid][24][UTT][ MSE = 0.1902 | LCC = 0.9593 | SRCC = 0.9595 ] +INFO:root:Average loss=0.03526245822012424 +INFO:root: +[Valid][25][UTT][ MSE = 0.2148 | LCC = 0.9566 | SRCC = 0.9572 ] +INFO:root:Average loss=0.03359608905017376 +INFO:root: +[Valid][26][UTT][ MSE = 0.1950 | LCC = 0.9588 | SRCC = 0.9587 ] +INFO:root:Average loss=0.03380897691845894 +INFO:root: +[Valid][27][UTT][ MSE = 0.1949 | LCC = 0.9583 | SRCC = 0.9577 ] +INFO:root:Average loss=0.03283349588513374 +INFO:root: +[Valid][28][UTT][ MSE = 0.2009 | LCC = 0.9581 | SRCC = 0.9579 ] +INFO:root:Average loss=0.03203093089163303 +INFO:root: +[Valid][29][UTT][ MSE = 0.2007 | LCC = 0.9582 | SRCC = 0.9580 ] +INFO:root:Average loss=0.03265168417990208 +INFO:root: +[Valid][30][UTT][ MSE = 0.2041 | LCC = 0.9556 | SRCC = 0.9557 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2004 | LCC = 0.9557 | SRCC = 0.9564 ] diff --git a/results_tencent/w2v2_base/layer9/run0/config.gin b/results_tencent/w2v2_base/layer9/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer9/run0/model.pt b/results_tencent/w2v2_base/layer9/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..474bb843213219e57314b49383bd7bcbc257f969 --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:107005fe5f1cea0f9192a6764d40db21fbe1591cb69848d86882f15944b89c4e +size 2747979 diff --git a/results_tencent/w2v2_base/layer9/run0/model_best.pt b/results_tencent/w2v2_base/layer9/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..18d8038e6d2afbd588454ffad775b87bab9c5939 --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:96a9b3580c37d75de07dacb08bba5ec2478ada2be3103b46a554e6b47abf5555 +size 2749702 diff --git a/results_tencent/w2v2_base/layer9/run0/train.log b/results_tencent/w2v2_base/layer9/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c65151686b60ffcd40b520836dabce0a9cf4b7ad --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4600423321723938 +INFO:root: +[Valid][0][UTT][ MSE = 0.2985 | LCC = 0.9275 | SRCC = 0.9285 ] +INFO:root:Average loss=0.18118676549196244 +INFO:root: +[Valid][1][UTT][ MSE = 0.2463 | LCC = 0.9450 | SRCC = 0.9462 ] +INFO:root:Average loss=0.1439670158624649 +INFO:root: +[Valid][2][UTT][ MSE = 0.2470 | LCC = 0.9512 | SRCC = 0.9527 ] +INFO:root:Average loss=0.13061812376976012 +INFO:root: +[Valid][3][UTT][ MSE = 0.2051 | LCC = 0.9531 | SRCC = 0.9543 ] +INFO:root:Average loss=0.11605198878049851 +INFO:root: +[Valid][4][UTT][ MSE = 0.2621 | LCC = 0.9526 | SRCC = 0.9569 ] +INFO:root:Average loss=0.107225516051054 +INFO:root: +[Valid][5][UTT][ MSE = 0.1953 | LCC = 0.9568 | SRCC = 0.9578 ] +INFO:root:Average loss=0.09663335159420967 +INFO:root: +[Valid][6][UTT][ MSE = 0.1830 | LCC = 0.9576 | SRCC = 0.9580 ] +INFO:root:Average loss=0.09250523510575294 +INFO:root: +[Valid][7][UTT][ MSE = 0.1833 | LCC = 0.9581 | SRCC = 0.9581 ] +INFO:root:Average loss=0.08591616481542587 +INFO:root: +[Valid][8][UTT][ MSE = 0.1871 | LCC = 0.9576 | SRCC = 0.9586 ] +INFO:root:Average loss=0.08305582082271576 +INFO:root: +[Valid][9][UTT][ MSE = 0.1858 | LCC = 0.9592 | SRCC = 0.9594 ] +INFO:root:Average loss=0.07767535173892975 +INFO:root: +[Valid][10][UTT][ MSE = 0.1813 | LCC = 0.9601 | SRCC = 0.9608 ] +INFO:root:Average loss=0.07130129280686379 +INFO:root: +[Valid][11][UTT][ MSE = 0.2114 | LCC = 0.9595 | SRCC = 0.9607 ] +INFO:root:Average loss=0.07040292260050773 +INFO:root: +[Valid][12][UTT][ MSE = 0.1947 | LCC = 0.9592 | SRCC = 0.9596 ] +INFO:root:Average loss=0.06572153013944626 +INFO:root: +[Valid][13][UTT][ MSE = 0.1950 | LCC = 0.9594 | SRCC = 0.9604 ] +INFO:root:Average loss=0.06286424788832665 +INFO:root: +[Valid][14][UTT][ MSE = 0.1889 | LCC = 0.9606 | SRCC = 0.9606 ] +INFO:root:Average loss=0.06064805203676224 +INFO:root: +[Valid][15][UTT][ MSE = 0.1806 | LCC = 0.9615 | SRCC = 0.9617 ] +INFO:root:Average loss=0.055247776716947555 +INFO:root: +[Valid][16][UTT][ MSE = 0.1968 | LCC = 0.9605 | SRCC = 0.9600 ] +INFO:root:Average loss=0.05572229601442814 +INFO:root: +[Valid][17][UTT][ MSE = 0.1737 | LCC = 0.9614 | SRCC = 0.9611 ] +INFO:root:Average loss=0.051520288228988645 +INFO:root: +[Valid][18][UTT][ MSE = 0.1813 | LCC = 0.9608 | SRCC = 0.9611 ] +INFO:root:Average loss=0.04979335106909275 +INFO:root: +[Valid][19][UTT][ MSE = 0.2054 | LCC = 0.9608 | SRCC = 0.9616 ] +INFO:root:Average loss=0.04712054497003555 +INFO:root: +[Valid][20][UTT][ MSE = 0.1808 | LCC = 0.9612 | SRCC = 0.9609 ] +INFO:root:Average loss=0.0442212571054697 +INFO:root: +[Valid][21][UTT][ MSE = 0.1698 | LCC = 0.9598 | SRCC = 0.9602 ] +INFO:root:Average loss=0.04579475358128548 +INFO:root: +[Valid][22][UTT][ MSE = 0.1606 | LCC = 0.9613 | SRCC = 0.9616 ] +INFO:root:Average loss=0.04372332695126534 +INFO:root: +[Valid][23][UTT][ MSE = 0.1846 | LCC = 0.9606 | SRCC = 0.9611 ] +INFO:root:Average loss=0.039272620737552645 +INFO:root: +[Valid][24][UTT][ MSE = 0.1951 | LCC = 0.9603 | SRCC = 0.9610 ] +INFO:root:Average loss=0.0391554779112339 +INFO:root: +[Valid][25][UTT][ MSE = 0.1674 | LCC = 0.9608 | SRCC = 0.9609 ] +INFO:root:Average loss=0.03752056443691253 +INFO:root: +[Valid][26][UTT][ MSE = 0.1920 | LCC = 0.9620 | SRCC = 0.9622 ] +INFO:root:Average loss=0.03868252055346966 +INFO:root: +[Valid][27][UTT][ MSE = 0.1814 | LCC = 0.9608 | SRCC = 0.9609 ] +INFO:root:Average loss=0.03841265422105789 +INFO:root: +[Valid][28][UTT][ MSE = 0.1707 | LCC = 0.9613 | SRCC = 0.9615 ] +INFO:root:Average loss=0.033679188087582586 +INFO:root: +[Valid][29][UTT][ MSE = 0.2048 | LCC = 0.9600 | SRCC = 0.9605 ] +INFO:root:Average loss=0.03225672456622124 +INFO:root: +[Valid][30][UTT][ MSE = 0.1875 | LCC = 0.9610 | SRCC = 0.9612 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2022 | LCC = 0.9579 | SRCC = 0.9591 ] diff --git a/results_tencent/w2v2_base/layer9/run1/config.gin b/results_tencent/w2v2_base/layer9/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer9/run1/model.pt b/results_tencent/w2v2_base/layer9/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d5efef4d1b68d4735c881f8185433525ee3b94b3 --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6b71654604a1f0b2db9d2039335edeb7240d30a6d9e99402a9309c609b49d9fc +size 2747979 diff --git a/results_tencent/w2v2_base/layer9/run1/model_best.pt b/results_tencent/w2v2_base/layer9/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c9998855af69fdae5420b3000923530cbec5ebda --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16914915a8ecd48e8a390c6c3c7ae89b24abc20b28671d6dfb8844f6af66af4d +size 2749702 diff --git a/results_tencent/w2v2_base/layer9/run1/train.log b/results_tencent/w2v2_base/layer9/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9818dd63fd669d877507ccdce973125607141222 --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.38488785791397095 +INFO:root: +[Valid][0][UTT][ MSE = 0.2901 | LCC = 0.9327 | SRCC = 0.9349 ] +INFO:root:Average loss=0.17206081461906433 +INFO:root: +[Valid][1][UTT][ MSE = 0.2494 | LCC = 0.9404 | SRCC = 0.9432 ] +INFO:root:Average loss=0.14681835222244263 +INFO:root: +[Valid][2][UTT][ MSE = 0.2080 | LCC = 0.9508 | SRCC = 0.9517 ] +INFO:root:Average loss=0.1255141116976738 +INFO:root: +[Valid][3][UTT][ MSE = 0.2459 | LCC = 0.9525 | SRCC = 0.9541 ] +INFO:root:Average loss=0.11464775949716569 +INFO:root: +[Valid][4][UTT][ MSE = 0.2054 | LCC = 0.9556 | SRCC = 0.9565 ] +INFO:root:Average loss=0.10337488296627999 +INFO:root: +[Valid][5][UTT][ MSE = 0.2242 | LCC = 0.9545 | SRCC = 0.9551 ] +INFO:root:Average loss=0.09650979402661324 +INFO:root: +[Valid][6][UTT][ MSE = 0.2410 | LCC = 0.9554 | SRCC = 0.9576 ] +INFO:root:Average loss=0.08894150346517563 +INFO:root: +[Valid][7][UTT][ MSE = 0.1685 | LCC = 0.9571 | SRCC = 0.9574 ] +INFO:root:Average loss=0.08704590222239494 +INFO:root: +[Valid][8][UTT][ MSE = 0.2097 | LCC = 0.9573 | SRCC = 0.9578 ] +INFO:root:Average loss=0.08049719414114952 +INFO:root: +[Valid][9][UTT][ MSE = 0.2074 | LCC = 0.9568 | SRCC = 0.9579 ] +INFO:root:Average loss=0.0757464699447155 +INFO:root: +[Valid][10][UTT][ MSE = 0.1964 | LCC = 0.9592 | SRCC = 0.9599 ] +INFO:root:Average loss=0.07311320155858994 +INFO:root: +[Valid][11][UTT][ MSE = 0.1864 | LCC = 0.9590 | SRCC = 0.9589 ] +INFO:root:Average loss=0.06876605302095413 +INFO:root: +[Valid][12][UTT][ MSE = 0.2150 | LCC = 0.9585 | SRCC = 0.9592 ] +INFO:root:Average loss=0.0666842770576477 +INFO:root: +[Valid][13][UTT][ MSE = 0.1960 | LCC = 0.9588 | SRCC = 0.9588 ] +INFO:root:Average loss=0.06130083879828453 +INFO:root: +[Valid][14][UTT][ MSE = 0.1785 | LCC = 0.9594 | SRCC = 0.9593 ] +INFO:root:Average loss=0.058882549196481705 +INFO:root: +[Valid][15][UTT][ MSE = 0.2084 | LCC = 0.9586 | SRCC = 0.9591 ] +INFO:root:Average loss=0.05571475276350975 +INFO:root: +[Valid][16][UTT][ MSE = 0.1849 | LCC = 0.9598 | SRCC = 0.9597 ] +INFO:root:Average loss=0.05321459691226482 +INFO:root: +[Valid][17][UTT][ MSE = 0.2107 | LCC = 0.9581 | SRCC = 0.9580 ] +INFO:root:Average loss=0.05143951198458672 +INFO:root: +[Valid][18][UTT][ MSE = 0.1887 | LCC = 0.9584 | SRCC = 0.9583 ] +INFO:root:Average loss=0.049135391265153885 +INFO:root: +[Valid][19][UTT][ MSE = 0.2120 | LCC = 0.9583 | SRCC = 0.9589 ] +INFO:root:Average loss=0.04646381948888302 +INFO:root: +[Valid][20][UTT][ MSE = 0.1822 | LCC = 0.9595 | SRCC = 0.9594 ] +INFO:root:Average loss=0.04406776562333107 +INFO:root: +[Valid][21][UTT][ MSE = 0.2047 | LCC = 0.9586 | SRCC = 0.9593 ] +INFO:root:Average loss=0.045837701082229614 +INFO:root: +[Valid][22][UTT][ MSE = 0.1954 | LCC = 0.9587 | SRCC = 0.9582 ] +INFO:root:Average loss=0.04483763238787651 +INFO:root: +[Valid][23][UTT][ MSE = 0.1925 | LCC = 0.9592 | SRCC = 0.9593 ] +INFO:root:Average loss=0.040482610166072845 +INFO:root: +[Valid][24][UTT][ MSE = 0.1942 | LCC = 0.9596 | SRCC = 0.9589 ] +INFO:root:Average loss=0.0400471902936697 +INFO:root: +[Valid][25][UTT][ MSE = 0.1759 | LCC = 0.9593 | SRCC = 0.9589 ] +INFO:root:Average loss=0.037572915241122246 +INFO:root: +[Valid][26][UTT][ MSE = 0.2026 | LCC = 0.9593 | SRCC = 0.9593 ] +INFO:root:Average loss=0.03798970732092857 +INFO:root: +[Valid][27][UTT][ MSE = 0.1799 | LCC = 0.9598 | SRCC = 0.9596 ] +INFO:root:Average loss=0.03563271501660347 +INFO:root: +[Valid][28][UTT][ MSE = 0.1941 | LCC = 0.9586 | SRCC = 0.9590 ] +INFO:root:Average loss=0.03561373944580555 +INFO:root: +[Valid][29][UTT][ MSE = 0.2040 | LCC = 0.9592 | SRCC = 0.9594 ] +INFO:root:Average loss=0.0322020618468523 +INFO:root: +[Valid][30][UTT][ MSE = 0.1798 | LCC = 0.9586 | SRCC = 0.9587 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1989 | LCC = 0.9569 | SRCC = 0.9575 ] diff --git a/results_tencent/w2v2_base/layer9/run2/config.gin b/results_tencent/w2v2_base/layer9/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer9/run2/model.pt b/results_tencent/w2v2_base/layer9/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bf1aa1b382842406c0e4dad6f1a64dc0768d9242 --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:76c87c46368cf26a48cae9967c492af841c47fb55dcbdea4a32408c8a0359b5a +size 2747979 diff --git a/results_tencent/w2v2_base/layer9/run2/model_best.pt b/results_tencent/w2v2_base/layer9/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9a269c437296296862ee402ea730f28e698774c9 --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2a83501c40ee9d23ee78b6c30075c883b244da7b61f94203e849bccb7f3c0d6d +size 2749702 diff --git a/results_tencent/w2v2_base/layer9/run2/train.log b/results_tencent/w2v2_base/layer9/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d58a24a505ee96c2e9e61ae8814887300d440a61 --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44134008622169496 +INFO:root: +[Valid][0][UTT][ MSE = 0.2872 | LCC = 0.9245 | SRCC = 0.9257 ] +INFO:root:Average loss=0.1845103096961975 +INFO:root: +[Valid][1][UTT][ MSE = 0.2094 | LCC = 0.9424 | SRCC = 0.9444 ] +INFO:root:Average loss=0.14963708490133285 +INFO:root: +[Valid][2][UTT][ MSE = 0.2181 | LCC = 0.9497 | SRCC = 0.9510 ] +INFO:root:Average loss=0.13257612210512162 +INFO:root: +[Valid][3][UTT][ MSE = 0.2328 | LCC = 0.9488 | SRCC = 0.9496 ] +INFO:root:Average loss=0.11762724012136459 +INFO:root: +[Valid][4][UTT][ MSE = 0.2466 | LCC = 0.9533 | SRCC = 0.9538 ] +INFO:root:Average loss=0.11056965696811676 +INFO:root: +[Valid][5][UTT][ MSE = 0.2202 | LCC = 0.9553 | SRCC = 0.9573 ] +INFO:root:Average loss=0.10181708103418351 +INFO:root: +[Valid][6][UTT][ MSE = 0.1761 | LCC = 0.9562 | SRCC = 0.9573 ] +INFO:root:Average loss=0.09477315881848335 +INFO:root: +[Valid][7][UTT][ MSE = 0.2161 | LCC = 0.9572 | SRCC = 0.9586 ] +INFO:root:Average loss=0.09185017845034599 +INFO:root: +[Valid][8][UTT][ MSE = 0.2090 | LCC = 0.9588 | SRCC = 0.9592 ] +INFO:root:Average loss=0.08429942828416824 +INFO:root: +[Valid][9][UTT][ MSE = 0.2010 | LCC = 0.9592 | SRCC = 0.9595 ] +INFO:root:Average loss=0.07947642156481743 +INFO:root: +[Valid][10][UTT][ MSE = 0.1968 | LCC = 0.9592 | SRCC = 0.9600 ] +INFO:root:Average loss=0.07902738842368126 +INFO:root: +[Valid][11][UTT][ MSE = 0.1937 | LCC = 0.9578 | SRCC = 0.9583 ] +INFO:root:Average loss=0.07183388531208039 +INFO:root: +[Valid][12][UTT][ MSE = 0.2043 | LCC = 0.9593 | SRCC = 0.9595 ] +INFO:root:Average loss=0.06914781644940377 +INFO:root: +[Valid][13][UTT][ MSE = 0.1882 | LCC = 0.9594 | SRCC = 0.9599 ] +INFO:root:Average loss=0.06601070901751518 +INFO:root: +[Valid][14][UTT][ MSE = 0.1942 | LCC = 0.9583 | SRCC = 0.9598 ] +INFO:root:Average loss=0.06231301289796829 +INFO:root: +[Valid][15][UTT][ MSE = 0.1698 | LCC = 0.9602 | SRCC = 0.9611 ] +INFO:root:Average loss=0.0685509783923626 +INFO:root: +[Valid][16][UTT][ MSE = 0.1837 | LCC = 0.9612 | SRCC = 0.9614 ] +INFO:root:Average loss=0.05680865913629532 +INFO:root: +[Valid][17][UTT][ MSE = 0.1916 | LCC = 0.9613 | SRCC = 0.9615 ] +INFO:root:Average loss=0.05644666057825089 +INFO:root: +[Valid][18][UTT][ MSE = 0.1872 | LCC = 0.9562 | SRCC = 0.9572 ] +INFO:root:Average loss=0.05210241198539734 +INFO:root: +[Valid][19][UTT][ MSE = 0.1901 | LCC = 0.9584 | SRCC = 0.9589 ] +INFO:root:Average loss=0.050447841063141825 +INFO:root: +[Valid][20][UTT][ MSE = 0.1907 | LCC = 0.9598 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04776925602555275 +INFO:root: +[Valid][21][UTT][ MSE = 0.1979 | LCC = 0.9599 | SRCC = 0.9601 ] +INFO:root:Average loss=0.04804039500653744 +INFO:root: +[Valid][22][UTT][ MSE = 0.1892 | LCC = 0.9589 | SRCC = 0.9589 ] +INFO:root:Average loss=0.04340790332853794 +INFO:root: +[Valid][23][UTT][ MSE = 0.1804 | LCC = 0.9602 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04497475910186768 +INFO:root: +[Valid][24][UTT][ MSE = 0.2007 | LCC = 0.9595 | SRCC = 0.9598 ] +INFO:root:Average loss=0.04432575222849846 +INFO:root: +[Valid][25][UTT][ MSE = 0.1955 | LCC = 0.9600 | SRCC = 0.9603 ] +INFO:root:Average loss=0.03934980608522892 +INFO:root: +[Valid][26][UTT][ MSE = 0.2016 | LCC = 0.9587 | SRCC = 0.9592 ] +INFO:root:Average loss=0.0396113938242197 +INFO:root: +[Valid][27][UTT][ MSE = 0.1937 | LCC = 0.9609 | SRCC = 0.9607 ] +INFO:root:Average loss=0.037328398883342745 +INFO:root: +[Valid][28][UTT][ MSE = 0.1776 | LCC = 0.9601 | SRCC = 0.9607 ] +INFO:root:Average loss=0.03701386910676956 +INFO:root: +[Valid][29][UTT][ MSE = 0.1815 | LCC = 0.9603 | SRCC = 0.9601 ] +INFO:root:Average loss=0.03513243196904659 +INFO:root: +[Valid][30][UTT][ MSE = 0.1909 | LCC = 0.9593 | SRCC = 0.9596 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1984 | LCC = 0.9590 | SRCC = 0.9598 ] diff --git a/results_tencent/w2v2_base/layer9/run3/config.gin b/results_tencent/w2v2_base/layer9/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer9/run3/model.pt b/results_tencent/w2v2_base/layer9/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..910b71e75d419b69e1dd3435af3a02bc531efc7d --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:52a0beef2bc3500e6657741a8cfd9e97a954a22fbf9477cda6204a93cd358b6b +size 2747979 diff --git a/results_tencent/w2v2_base/layer9/run3/model_best.pt b/results_tencent/w2v2_base/layer9/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..632b89c235fcda9cc8054ce13a603ea7293fba58 --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ad3a42a322a9361b477b0f6bace1812c15ff177da8ccc97d827c65b72ab8708b +size 2749702 diff --git a/results_tencent/w2v2_base/layer9/run3/train.log b/results_tencent/w2v2_base/layer9/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..508c9502b475efac1fec7e602f7617bfcb874e2a --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4620986734032631 +INFO:root: +[Valid][0][UTT][ MSE = 0.2932 | LCC = 0.9274 | SRCC = 0.9274 ] +INFO:root:Average loss=0.18642554777860643 +INFO:root: +[Valid][1][UTT][ MSE = 0.2342 | LCC = 0.9419 | SRCC = 0.9432 ] +INFO:root:Average loss=0.14669875675439834 +INFO:root: +[Valid][2][UTT][ MSE = 0.2609 | LCC = 0.9470 | SRCC = 0.9478 ] +INFO:root:Average loss=0.12615760403871537 +INFO:root: +[Valid][3][UTT][ MSE = 0.2334 | LCC = 0.9478 | SRCC = 0.9486 ] +INFO:root:Average loss=0.12149242234230041 +INFO:root: +[Valid][4][UTT][ MSE = 0.2231 | LCC = 0.9527 | SRCC = 0.9531 ] +INFO:root:Average loss=0.10639131453633309 +INFO:root: +[Valid][5][UTT][ MSE = 0.2403 | LCC = 0.9547 | SRCC = 0.9554 ] +INFO:root:Average loss=0.10387178903818131 +INFO:root: +[Valid][6][UTT][ MSE = 0.2224 | LCC = 0.9534 | SRCC = 0.9545 ] +INFO:root:Average loss=0.09296718662977219 +INFO:root: +[Valid][7][UTT][ MSE = 0.1987 | LCC = 0.9569 | SRCC = 0.9575 ] +INFO:root:Average loss=0.08847703847289086 +INFO:root: +[Valid][8][UTT][ MSE = 0.1843 | LCC = 0.9579 | SRCC = 0.9589 ] +INFO:root:Average loss=0.08371629625558853 +INFO:root: +[Valid][9][UTT][ MSE = 0.2184 | LCC = 0.9567 | SRCC = 0.9567 ] +INFO:root:Average loss=0.07920476227998734 +INFO:root: +[Valid][10][UTT][ MSE = 0.1812 | LCC = 0.9576 | SRCC = 0.9578 ] +INFO:root:Average loss=0.07437439209222793 +INFO:root: +[Valid][11][UTT][ MSE = 0.2059 | LCC = 0.9565 | SRCC = 0.9563 ] +INFO:root:Average loss=0.0709079700410366 +INFO:root: +[Valid][12][UTT][ MSE = 0.1980 | LCC = 0.9593 | SRCC = 0.9590 ] +INFO:root:Average loss=0.06738332030177116 +INFO:root: +[Valid][13][UTT][ MSE = 0.1643 | LCC = 0.9577 | SRCC = 0.9578 ] +INFO:root:Average loss=0.06527387207746506 +INFO:root: +[Valid][14][UTT][ MSE = 0.2048 | LCC = 0.9580 | SRCC = 0.9581 ] +INFO:root:Average loss=0.06099698489904404 +INFO:root: +[Valid][15][UTT][ MSE = 0.1944 | LCC = 0.9590 | SRCC = 0.9586 ] +INFO:root:Average loss=0.060574112311005596 +INFO:root: +[Valid][16][UTT][ MSE = 0.2053 | LCC = 0.9564 | SRCC = 0.9560 ] +INFO:root:Average loss=0.05622568660974502 +INFO:root: +[Valid][17][UTT][ MSE = 0.1944 | LCC = 0.9582 | SRCC = 0.9581 ] +INFO:root:Average loss=0.05406207545101643 +INFO:root: +[Valid][18][UTT][ MSE = 0.1926 | LCC = 0.9585 | SRCC = 0.9585 ] +INFO:root:Average loss=0.05133905026316643 +INFO:root: +[Valid][19][UTT][ MSE = 0.1864 | LCC = 0.9600 | SRCC = 0.9599 ] +INFO:root:Average loss=0.05177675841748714 +INFO:root: +[Valid][20][UTT][ MSE = 0.1947 | LCC = 0.9602 | SRCC = 0.9600 ] +INFO:root:Average loss=0.048810413718223575 +INFO:root: +[Valid][21][UTT][ MSE = 0.1859 | LCC = 0.9572 | SRCC = 0.9568 ] +INFO:root:Average loss=0.04540929448604584 +INFO:root: +[Valid][22][UTT][ MSE = 0.1787 | LCC = 0.9582 | SRCC = 0.9581 ] +INFO:root:Average loss=0.046539432883262635 +INFO:root: +[Valid][23][UTT][ MSE = 0.2085 | LCC = 0.9578 | SRCC = 0.9572 ] +INFO:root:Average loss=0.04221009178459644 +INFO:root: +[Valid][24][UTT][ MSE = 0.2098 | LCC = 0.9574 | SRCC = 0.9568 ] +INFO:root:Average loss=0.0393694875985384 +INFO:root: +[Valid][25][UTT][ MSE = 0.1969 | LCC = 0.9577 | SRCC = 0.9571 ] +INFO:root:Average loss=0.040776963606476785 +INFO:root: +[Valid][26][UTT][ MSE = 0.1983 | LCC = 0.9573 | SRCC = 0.9572 ] +INFO:root:Average loss=0.03914239470660687 +INFO:root: +[Valid][27][UTT][ MSE = 0.1974 | LCC = 0.9581 | SRCC = 0.9576 ] +INFO:root:Average loss=0.03750082740187645 +INFO:root: +[Valid][28][UTT][ MSE = 0.2200 | LCC = 0.9594 | SRCC = 0.9588 ] +INFO:root:Average loss=0.03821298025548458 +INFO:root: +[Valid][29][UTT][ MSE = 0.1995 | LCC = 0.9596 | SRCC = 0.9598 ] +INFO:root:Average loss=0.034670513778924945 +INFO:root: +[Valid][30][UTT][ MSE = 0.1956 | LCC = 0.9604 | SRCC = 0.9603 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2027 | LCC = 0.9568 | SRCC = 0.9578 ] diff --git a/results_tencent/w2v2_base/layer9/run4/config.gin b/results_tencent/w2v2_base/layer9/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..a18a82b65aa5381b0292abb520636378e93f86d6 --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base/layer9/run4/model.pt b/results_tencent/w2v2_base/layer9/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..583c19a94a90dfcf83ef93377ba36a4e02959c2d --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:833e08304f0b032194268b166b206bc1373072931dab9059d3e30683b3e392fd +size 2747979 diff --git a/results_tencent/w2v2_base/layer9/run4/model_best.pt b/results_tencent/w2v2_base/layer9/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..57dc25fc28e87137f0055e1e45fa5ffaa62147e6 --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:334d36e335c7a5eee45777cae8f122a1e66cb96b0de34aefe4475325e12120cb +size 2749702 diff --git a/results_tencent/w2v2_base/layer9/run4/train.log b/results_tencent/w2v2_base/layer9/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c1fc7377bc62a6d9654fc7b8120c65ab7b8c4ae6 --- /dev/null +++ b/results_tencent/w2v2_base/layer9/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.47990059208869934 +INFO:root: +[Valid][0][UTT][ MSE = 0.3218 | LCC = 0.9176 | SRCC = 0.9190 ] +INFO:root:Average loss=0.19599879443645477 +INFO:root: +[Valid][1][UTT][ MSE = 0.2248 | LCC = 0.9393 | SRCC = 0.9411 ] +INFO:root:Average loss=0.1501146567463875 +INFO:root: +[Valid][2][UTT][ MSE = 0.2135 | LCC = 0.9425 | SRCC = 0.9454 ] +INFO:root:Average loss=0.13485183715820312 +INFO:root: +[Valid][3][UTT][ MSE = 0.2263 | LCC = 0.9507 | SRCC = 0.9519 ] +INFO:root:Average loss=0.11866378903388977 +INFO:root: +[Valid][4][UTT][ MSE = 0.2038 | LCC = 0.9533 | SRCC = 0.9549 ] +INFO:root:Average loss=0.1107051301896572 +INFO:root: +[Valid][5][UTT][ MSE = 0.2240 | LCC = 0.9557 | SRCC = 0.9576 ] +INFO:root:Average loss=0.10336738443374634 +INFO:root: +[Valid][6][UTT][ MSE = 0.1702 | LCC = 0.9566 | SRCC = 0.9584 ] +INFO:root:Average loss=0.09492393353581428 +INFO:root: +[Valid][7][UTT][ MSE = 0.2083 | LCC = 0.9565 | SRCC = 0.9579 ] +INFO:root:Average loss=0.08919140142202377 +INFO:root: +[Valid][8][UTT][ MSE = 0.2117 | LCC = 0.9567 | SRCC = 0.9579 ] +INFO:root:Average loss=0.087555315554142 +INFO:root: +[Valid][9][UTT][ MSE = 0.1961 | LCC = 0.9584 | SRCC = 0.9589 ] +INFO:root:Average loss=0.08199657294154167 +INFO:root: +[Valid][10][UTT][ MSE = 0.2114 | LCC = 0.9583 | SRCC = 0.9597 ] +INFO:root:Average loss=0.07435027471184731 +INFO:root: +[Valid][11][UTT][ MSE = 0.2071 | LCC = 0.9579 | SRCC = 0.9589 ] +INFO:root:Average loss=0.07369155099987984 +INFO:root: +[Valid][12][UTT][ MSE = 0.1870 | LCC = 0.9575 | SRCC = 0.9577 ] +INFO:root:Average loss=0.06767214635014535 +INFO:root: +[Valid][13][UTT][ MSE = 0.1808 | LCC = 0.9566 | SRCC = 0.9580 ] +INFO:root:Average loss=0.0650751205086708 +INFO:root: +[Valid][14][UTT][ MSE = 0.2106 | LCC = 0.9586 | SRCC = 0.9590 ] +INFO:root:Average loss=0.06020495760440826 +INFO:root: +[Valid][15][UTT][ MSE = 0.1845 | LCC = 0.9581 | SRCC = 0.9588 ] +INFO:root:Average loss=0.05973772253096104 +INFO:root: +[Valid][16][UTT][ MSE = 0.2031 | LCC = 0.9607 | SRCC = 0.9613 ] +INFO:root:Average loss=0.0581268929541111 +INFO:root: +[Valid][17][UTT][ MSE = 0.1937 | LCC = 0.9607 | SRCC = 0.9608 ] +INFO:root:Average loss=0.0521228911280632 +INFO:root: +[Valid][18][UTT][ MSE = 0.2003 | LCC = 0.9595 | SRCC = 0.9599 ] +INFO:root:Average loss=0.05008061718940735 +INFO:root: +[Valid][19][UTT][ MSE = 0.1968 | LCC = 0.9603 | SRCC = 0.9605 ] +INFO:root:Average loss=0.05028889785706997 +INFO:root: +[Valid][20][UTT][ MSE = 0.2025 | LCC = 0.9598 | SRCC = 0.9597 ] +INFO:root:Average loss=0.04728607541322708 +INFO:root: +[Valid][21][UTT][ MSE = 0.1774 | LCC = 0.9604 | SRCC = 0.9606 ] +INFO:root:Average loss=0.045512697115540506 +INFO:root: +[Valid][22][UTT][ MSE = 0.1968 | LCC = 0.9582 | SRCC = 0.9582 ] +INFO:root:Average loss=0.04404397442936897 +INFO:root: +[Valid][23][UTT][ MSE = 0.1898 | LCC = 0.9608 | SRCC = 0.9609 ] +INFO:root:Average loss=0.04267848554253578 +INFO:root: +[Valid][24][UTT][ MSE = 0.2038 | LCC = 0.9601 | SRCC = 0.9598 ] +INFO:root:Average loss=0.04194591556489467 +INFO:root: +[Valid][25][UTT][ MSE = 0.1932 | LCC = 0.9601 | SRCC = 0.9607 ] +INFO:root:Average loss=0.0403424691259861 +INFO:root: +[Valid][26][UTT][ MSE = 0.2024 | LCC = 0.9591 | SRCC = 0.9588 ] +INFO:root:Average loss=0.04022276416420936 +INFO:root: +[Valid][27][UTT][ MSE = 0.1803 | LCC = 0.9604 | SRCC = 0.9603 ] +INFO:root:Average loss=0.037275801479816434 +INFO:root: +[Valid][28][UTT][ MSE = 0.2094 | LCC = 0.9594 | SRCC = 0.9594 ] +INFO:root:Average loss=0.03804623945057392 +INFO:root: +[Valid][29][UTT][ MSE = 0.1821 | LCC = 0.9603 | SRCC = 0.9606 ] +INFO:root:Average loss=0.03623796030879021 +INFO:root: +[Valid][30][UTT][ MSE = 0.1817 | LCC = 0.9600 | SRCC = 0.9604 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2012 | LCC = 0.9576 | SRCC = 0.9587 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b5b85e7d6a98aafc1b8e07da3ad38e5d02418290 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ab9e3aaae69c7e751303b8ffd5d2d40a7c236aed20ee641f5ddfa5b60e3b7998 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e6d280dbc69faf72cb5fc044caf667fd2bd8341f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af75323900ffa1f22400d30bf21f1856d0ef260278f8668da098b7927e5d3950 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..28b0c32857a4a93736bfa538083ef994312ec49e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5498214404582977 +INFO:root: +[Valid][0][UTT][ MSE = 0.3163 | LCC = 0.9026 | SRCC = 0.9062 ] +INFO:root:Average loss=0.23513825619220732 +INFO:root: +[Valid][1][UTT][ MSE = 0.2970 | LCC = 0.9243 | SRCC = 0.9253 ] +INFO:root:Average loss=0.19283570021390914 +INFO:root: +[Valid][2][UTT][ MSE = 0.2568 | LCC = 0.9324 | SRCC = 0.9343 ] +INFO:root:Average loss=0.17435431510210037 +INFO:root: +[Valid][3][UTT][ MSE = 0.2552 | LCC = 0.9357 | SRCC = 0.9367 ] +INFO:root:Average loss=0.1582770220041275 +INFO:root: +[Valid][4][UTT][ MSE = 0.2052 | LCC = 0.9413 | SRCC = 0.9430 ] +INFO:root:Average loss=0.14278519040346146 +INFO:root: +[Valid][5][UTT][ MSE = 0.2147 | LCC = 0.9430 | SRCC = 0.9446 ] +INFO:root:Average loss=0.1311248115003109 +INFO:root: +[Valid][6][UTT][ MSE = 0.2272 | LCC = 0.9461 | SRCC = 0.9472 ] +INFO:root:Average loss=0.1241414077281952 +INFO:root: +[Valid][7][UTT][ MSE = 0.2056 | LCC = 0.9476 | SRCC = 0.9490 ] +INFO:root:Average loss=0.1159870363175869 +INFO:root: +[Valid][8][UTT][ MSE = 0.2249 | LCC = 0.9465 | SRCC = 0.9479 ] +INFO:root:Average loss=0.10892801147699356 +INFO:root: +[Valid][9][UTT][ MSE = 0.1840 | LCC = 0.9492 | SRCC = 0.9498 ] +INFO:root:Average loss=0.10259622955322266 +INFO:root: +[Valid][10][UTT][ MSE = 0.2095 | LCC = 0.9507 | SRCC = 0.9510 ] +INFO:root:Average loss=0.09578531259298324 +INFO:root: +[Valid][11][UTT][ MSE = 0.2021 | LCC = 0.9506 | SRCC = 0.9512 ] +INFO:root:Average loss=0.09502453908324242 +INFO:root: +[Valid][12][UTT][ MSE = 0.2088 | LCC = 0.9517 | SRCC = 0.9526 ] +INFO:root:Average loss=0.09010354968905449 +INFO:root: +[Valid][13][UTT][ MSE = 0.1915 | LCC = 0.9503 | SRCC = 0.9512 ] +INFO:root:Average loss=0.08493196594715119 +INFO:root: +[Valid][14][UTT][ MSE = 0.2014 | LCC = 0.9518 | SRCC = 0.9526 ] +INFO:root:Average loss=0.07915147551894187 +INFO:root: +[Valid][15][UTT][ MSE = 0.2092 | LCC = 0.9529 | SRCC = 0.9532 ] +INFO:root:Average loss=0.07977811160683632 +INFO:root: +[Valid][16][UTT][ MSE = 0.2045 | LCC = 0.9499 | SRCC = 0.9507 ] +INFO:root:Average loss=0.07245570403337478 +INFO:root: +[Valid][17][UTT][ MSE = 0.1926 | LCC = 0.9535 | SRCC = 0.9539 ] +INFO:root:Average loss=0.07553564113378525 +INFO:root: +[Valid][18][UTT][ MSE = 0.1840 | LCC = 0.9545 | SRCC = 0.9551 ] +INFO:root:Average loss=0.06683241000771523 +INFO:root: +[Valid][19][UTT][ MSE = 0.1875 | LCC = 0.9529 | SRCC = 0.9536 ] +INFO:root:Average loss=0.06553673842549324 +INFO:root: +[Valid][20][UTT][ MSE = 0.1876 | LCC = 0.9542 | SRCC = 0.9550 ] +INFO:root:Average loss=0.06448753762245178 +INFO:root: +[Valid][21][UTT][ MSE = 0.2035 | LCC = 0.9553 | SRCC = 0.9560 ] +INFO:root:Average loss=0.06093893346190453 +INFO:root: +[Valid][22][UTT][ MSE = 0.2018 | LCC = 0.9554 | SRCC = 0.9560 ] +INFO:root:Average loss=0.061921474501490596 +INFO:root: +[Valid][23][UTT][ MSE = 0.2028 | LCC = 0.9533 | SRCC = 0.9539 ] +INFO:root:Average loss=0.05902229157090187 +INFO:root: +[Valid][24][UTT][ MSE = 0.1913 | LCC = 0.9543 | SRCC = 0.9552 ] +INFO:root:Average loss=0.056587240278720856 +INFO:root: +[Valid][25][UTT][ MSE = 0.1905 | LCC = 0.9567 | SRCC = 0.9572 ] +INFO:root:Average loss=0.05569815531373024 +INFO:root: +[Valid][26][UTT][ MSE = 0.2050 | LCC = 0.9542 | SRCC = 0.9550 ] +INFO:root:Average loss=0.052767016440629956 +INFO:root: +[Valid][27][UTT][ MSE = 0.1880 | LCC = 0.9557 | SRCC = 0.9564 ] +INFO:root:Average loss=0.051841654017567636 +INFO:root: +[Valid][28][UTT][ MSE = 0.1970 | LCC = 0.9560 | SRCC = 0.9568 ] +INFO:root:Average loss=0.050541588515043256 +INFO:root: +[Valid][29][UTT][ MSE = 0.1818 | LCC = 0.9554 | SRCC = 0.9559 ] +INFO:root:Average loss=0.04729851919412613 +INFO:root: +[Valid][30][UTT][ MSE = 0.1718 | LCC = 0.9561 | SRCC = 0.9568 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1997 | LCC = 0.9563 | SRCC = 0.9572 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bfa0a6e36a612fbcbb92e00dbf636a09f9e63d0d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f32b078243f2303add1cbfceddbb99c12903ddbdad93fe6d079d8d90f21759ab +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7e68d830a5c4fcdebe1f2a4f4a560a1ac1947c69 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4e3809fa6440c6df5493d75938439137c07cb795906d96f7e1a531887bb78791 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f6f4119a9fe3151756b649a6bc4d72a19de0c33d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5874765741825104 +INFO:root: +[Valid][0][UTT][ MSE = 0.3271 | LCC = 0.9030 | SRCC = 0.9049 ] +INFO:root:Average loss=0.24464523220062256 +INFO:root: +[Valid][1][UTT][ MSE = 0.2778 | LCC = 0.9230 | SRCC = 0.9246 ] +INFO:root:Average loss=0.19600290709733964 +INFO:root: +[Valid][2][UTT][ MSE = 0.2540 | LCC = 0.9313 | SRCC = 0.9332 ] +INFO:root:Average loss=0.1761953086256981 +INFO:root: +[Valid][3][UTT][ MSE = 0.2288 | LCC = 0.9356 | SRCC = 0.9378 ] +INFO:root:Average loss=0.1577732150554657 +INFO:root: +[Valid][4][UTT][ MSE = 0.2171 | LCC = 0.9379 | SRCC = 0.9398 ] +INFO:root:Average loss=0.14477181798219682 +INFO:root: +[Valid][5][UTT][ MSE = 0.2340 | LCC = 0.9429 | SRCC = 0.9444 ] +INFO:root:Average loss=0.13032295566797256 +INFO:root: +[Valid][6][UTT][ MSE = 0.2350 | LCC = 0.9426 | SRCC = 0.9443 ] +INFO:root:Average loss=0.12380033767223358 +INFO:root: +[Valid][7][UTT][ MSE = 0.2081 | LCC = 0.9440 | SRCC = 0.9453 ] +INFO:root:Average loss=0.11605560159683227 +INFO:root: +[Valid][8][UTT][ MSE = 0.1969 | LCC = 0.9480 | SRCC = 0.9484 ] +INFO:root:Average loss=0.104717090010643 +INFO:root: +[Valid][9][UTT][ MSE = 0.2019 | LCC = 0.9484 | SRCC = 0.9502 ] +INFO:root:Average loss=0.1062909565269947 +INFO:root: +[Valid][10][UTT][ MSE = 0.2055 | LCC = 0.9484 | SRCC = 0.9493 ] +INFO:root:Average loss=0.09421191796660423 +INFO:root: +[Valid][11][UTT][ MSE = 0.1929 | LCC = 0.9498 | SRCC = 0.9503 ] +INFO:root:Average loss=0.0945673130452633 +INFO:root: +[Valid][12][UTT][ MSE = 0.2019 | LCC = 0.9518 | SRCC = 0.9524 ] +INFO:root:Average loss=0.08683719849586487 +INFO:root: +[Valid][13][UTT][ MSE = 0.1937 | LCC = 0.9506 | SRCC = 0.9507 ] +INFO:root:Average loss=0.08333367651700974 +INFO:root: +[Valid][14][UTT][ MSE = 0.1864 | LCC = 0.9515 | SRCC = 0.9520 ] +INFO:root:Average loss=0.08322999158501625 +INFO:root: +[Valid][15][UTT][ MSE = 0.1978 | LCC = 0.9487 | SRCC = 0.9489 ] +INFO:root:Average loss=0.0774692026078701 +INFO:root: +[Valid][16][UTT][ MSE = 0.2029 | LCC = 0.9491 | SRCC = 0.9493 ] +INFO:root:Average loss=0.07411796507239342 +INFO:root: +[Valid][17][UTT][ MSE = 0.1798 | LCC = 0.9526 | SRCC = 0.9529 ] +INFO:root:Average loss=0.07235254284739494 +INFO:root: +[Valid][18][UTT][ MSE = 0.2208 | LCC = 0.9518 | SRCC = 0.9527 ] +INFO:root:Average loss=0.0691223354935646 +INFO:root: +[Valid][19][UTT][ MSE = 0.1904 | LCC = 0.9539 | SRCC = 0.9543 ] +INFO:root:Average loss=0.0643552868962288 +INFO:root: +[Valid][20][UTT][ MSE = 0.2044 | LCC = 0.9532 | SRCC = 0.9535 ] +INFO:root:Average loss=0.06485883176326751 +INFO:root: +[Valid][21][UTT][ MSE = 0.2072 | LCC = 0.9546 | SRCC = 0.9546 ] +INFO:root:Average loss=0.060619796872138976 +INFO:root: +[Valid][22][UTT][ MSE = 0.1794 | LCC = 0.9545 | SRCC = 0.9549 ] +INFO:root:Average loss=0.06467466327548027 +INFO:root: +[Valid][23][UTT][ MSE = 0.1771 | LCC = 0.9537 | SRCC = 0.9547 ] +INFO:root:Average loss=0.05618610548973083 +INFO:root: +[Valid][24][UTT][ MSE = 0.2001 | LCC = 0.9540 | SRCC = 0.9548 ] +INFO:root:Average loss=0.05709683150053024 +INFO:root: +[Valid][25][UTT][ MSE = 0.1910 | LCC = 0.9550 | SRCC = 0.9553 ] +INFO:root:Average loss=0.05693872693181038 +INFO:root: +[Valid][26][UTT][ MSE = 0.2051 | LCC = 0.9547 | SRCC = 0.9562 ] +INFO:root:Average loss=0.05623861721158028 +INFO:root: +[Valid][27][UTT][ MSE = 0.2078 | LCC = 0.9547 | SRCC = 0.9551 ] +INFO:root:Average loss=0.05116684919595718 +INFO:root: +[Valid][28][UTT][ MSE = 0.1803 | LCC = 0.9552 | SRCC = 0.9559 ] +INFO:root:Average loss=0.05030637915432453 +INFO:root: +[Valid][29][UTT][ MSE = 0.1888 | LCC = 0.9555 | SRCC = 0.9559 ] +INFO:root:Average loss=0.04800549738109112 +INFO:root: +[Valid][30][UTT][ MSE = 0.1904 | LCC = 0.9534 | SRCC = 0.9541 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1968 | LCC = 0.9553 | SRCC = 0.9559 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4d89672d7b434515645d9935c7d05aca0929051f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aa1c1695360d1d303785b90a6623e8bb02e06a71cf97d0ab4ca3b0ae44012eb1 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..1f773b7a5287aee30ceae05ef40a1f4762f0dcce --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5c72bf983c4e35bf4a52514d13d761f47a7c9a582c002174f6003bc5fad74445 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b4ed1379f48baf918601ce0870b4e88db450a283 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5352451982498169 +INFO:root: +[Valid][0][UTT][ MSE = 0.3510 | LCC = 0.9021 | SRCC = 0.9035 ] +INFO:root:Average loss=0.2290169314146042 +INFO:root: +[Valid][1][UTT][ MSE = 0.2736 | LCC = 0.9229 | SRCC = 0.9243 ] +INFO:root:Average loss=0.19572282141447067 +INFO:root: +[Valid][2][UTT][ MSE = 0.2848 | LCC = 0.9283 | SRCC = 0.9302 ] +INFO:root:Average loss=0.1767163901925087 +INFO:root: +[Valid][3][UTT][ MSE = 0.2231 | LCC = 0.9328 | SRCC = 0.9339 ] +INFO:root:Average loss=0.15441066348552704 +INFO:root: +[Valid][4][UTT][ MSE = 0.2397 | LCC = 0.9373 | SRCC = 0.9383 ] +INFO:root:Average loss=0.13550329184532164 +INFO:root: +[Valid][5][UTT][ MSE = 0.2285 | LCC = 0.9438 | SRCC = 0.9449 ] +INFO:root:Average loss=0.12557087510824202 +INFO:root: +[Valid][6][UTT][ MSE = 0.2404 | LCC = 0.9441 | SRCC = 0.9459 ] +INFO:root:Average loss=0.12241460484266281 +INFO:root: +[Valid][7][UTT][ MSE = 0.2414 | LCC = 0.9433 | SRCC = 0.9451 ] +INFO:root:Average loss=0.1169785521030426 +INFO:root: +[Valid][8][UTT][ MSE = 0.2293 | LCC = 0.9472 | SRCC = 0.9481 ] +INFO:root:Average loss=0.1091008649468422 +INFO:root: +[Valid][9][UTT][ MSE = 0.2052 | LCC = 0.9477 | SRCC = 0.9491 ] +INFO:root:Average loss=0.10073007863759995 +INFO:root: +[Valid][10][UTT][ MSE = 0.2100 | LCC = 0.9486 | SRCC = 0.9491 ] +INFO:root:Average loss=0.0941714423596859 +INFO:root: +[Valid][11][UTT][ MSE = 0.2099 | LCC = 0.9484 | SRCC = 0.9491 ] +INFO:root:Average loss=0.0965525329709053 +INFO:root: +[Valid][12][UTT][ MSE = 0.2004 | LCC = 0.9484 | SRCC = 0.9495 ] +INFO:root:Average loss=0.08801886668801308 +INFO:root: +[Valid][13][UTT][ MSE = 0.2077 | LCC = 0.9497 | SRCC = 0.9509 ] +INFO:root:Average loss=0.0843541396856308 +INFO:root: +[Valid][14][UTT][ MSE = 0.2052 | LCC = 0.9510 | SRCC = 0.9518 ] +INFO:root:Average loss=0.08056368604302407 +INFO:root: +[Valid][15][UTT][ MSE = 0.1917 | LCC = 0.9524 | SRCC = 0.9533 ] +INFO:root:Average loss=0.07705343174934387 +INFO:root: +[Valid][16][UTT][ MSE = 0.1925 | LCC = 0.9520 | SRCC = 0.9524 ] +INFO:root:Average loss=0.07166802647709847 +INFO:root: +[Valid][17][UTT][ MSE = 0.1880 | LCC = 0.9522 | SRCC = 0.9529 ] +INFO:root:Average loss=0.06850264108180999 +INFO:root: +[Valid][18][UTT][ MSE = 0.1937 | LCC = 0.9523 | SRCC = 0.9529 ] +INFO:root:Average loss=0.06501003423333168 +INFO:root: +[Valid][19][UTT][ MSE = 0.2016 | LCC = 0.9540 | SRCC = 0.9545 ] +INFO:root:Average loss=0.06986146032810212 +INFO:root: +[Valid][20][UTT][ MSE = 0.2086 | LCC = 0.9506 | SRCC = 0.9521 ] +INFO:root:Average loss=0.06207995092868805 +INFO:root: +[Valid][21][UTT][ MSE = 0.1826 | LCC = 0.9541 | SRCC = 0.9548 ] +INFO:root:Average loss=0.06358936415612698 +INFO:root: +[Valid][22][UTT][ MSE = 0.1928 | LCC = 0.9532 | SRCC = 0.9537 ] +INFO:root:Average loss=0.05912833827733994 +INFO:root: +[Valid][23][UTT][ MSE = 0.1862 | LCC = 0.9545 | SRCC = 0.9551 ] +INFO:root:Average loss=0.05657418620586395 +INFO:root: +[Valid][24][UTT][ MSE = 0.1870 | LCC = 0.9549 | SRCC = 0.9558 ] +INFO:root:Average loss=0.05757870107889176 +INFO:root: +[Valid][25][UTT][ MSE = 0.2081 | LCC = 0.9542 | SRCC = 0.9553 ] +INFO:root:Average loss=0.05701523244380951 +INFO:root: +[Valid][26][UTT][ MSE = 0.1926 | LCC = 0.9549 | SRCC = 0.9552 ] +INFO:root:Average loss=0.050716456577181816 +INFO:root: +[Valid][27][UTT][ MSE = 0.1886 | LCC = 0.9528 | SRCC = 0.9539 ] +INFO:root:Average loss=0.051143169730901716 +INFO:root: +[Valid][28][UTT][ MSE = 0.2047 | LCC = 0.9536 | SRCC = 0.9544 ] +INFO:root:Average loss=0.04999509765207767 +INFO:root: +[Valid][29][UTT][ MSE = 0.1993 | LCC = 0.9534 | SRCC = 0.9535 ] +INFO:root:Average loss=0.04810072605311871 +INFO:root: +[Valid][30][UTT][ MSE = 0.1936 | LCC = 0.9545 | SRCC = 0.9548 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1935 | LCC = 0.9547 | SRCC = 0.9556 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ba8c2c3c0dc34907ff89ad7b8cc3e2d55a13b12c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:794ff2eff3c5775bce65584e66b0a24b2981637459f357c695bf9b9271c5de18 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9ceeda9af99d4ab0e98f135284ca4d9bd9160d7d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e1ee9d7544610f930a91cbf5a6a97e0dc267c3debc4de5ca6c96bab0cb1d4dc +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1051af6e03eb540567dfa8a60f474e18bbec8e55 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.48609879553318025 +INFO:root: +[Valid][0][UTT][ MSE = 0.3012 | LCC = 0.9093 | SRCC = 0.9109 ] +INFO:root:Average loss=0.23013638591766358 +INFO:root: +[Valid][1][UTT][ MSE = 0.2946 | LCC = 0.9237 | SRCC = 0.9242 ] +INFO:root:Average loss=0.19458530592918397 +INFO:root: +[Valid][2][UTT][ MSE = 0.2759 | LCC = 0.9329 | SRCC = 0.9345 ] +INFO:root:Average loss=0.166081090092659 +INFO:root: +[Valid][3][UTT][ MSE = 0.2317 | LCC = 0.9334 | SRCC = 0.9379 ] +INFO:root:Average loss=0.15042913836240768 +INFO:root: +[Valid][4][UTT][ MSE = 0.2396 | LCC = 0.9373 | SRCC = 0.9390 ] +INFO:root:Average loss=0.14053946721553803 +INFO:root: +[Valid][5][UTT][ MSE = 0.2338 | LCC = 0.9439 | SRCC = 0.9452 ] +INFO:root:Average loss=0.1283691084086895 +INFO:root: +[Valid][6][UTT][ MSE = 0.2044 | LCC = 0.9414 | SRCC = 0.9453 ] +INFO:root:Average loss=0.11789745569229126 +INFO:root: +[Valid][7][UTT][ MSE = 0.2004 | LCC = 0.9455 | SRCC = 0.9470 ] +INFO:root:Average loss=0.11291191133856773 +INFO:root: +[Valid][8][UTT][ MSE = 0.2220 | LCC = 0.9471 | SRCC = 0.9482 ] +INFO:root:Average loss=0.10893614745140076 +INFO:root: +[Valid][9][UTT][ MSE = 0.2000 | LCC = 0.9474 | SRCC = 0.9496 ] +INFO:root:Average loss=0.09919457510113716 +INFO:root: +[Valid][10][UTT][ MSE = 0.2180 | LCC = 0.9487 | SRCC = 0.9501 ] +INFO:root:Average loss=0.0954367390871048 +INFO:root: +[Valid][11][UTT][ MSE = 0.2150 | LCC = 0.9507 | SRCC = 0.9515 ] +INFO:root:Average loss=0.09157994011044503 +INFO:root: +[Valid][12][UTT][ MSE = 0.1882 | LCC = 0.9505 | SRCC = 0.9519 ] +INFO:root:Average loss=0.08376541969180107 +INFO:root: +[Valid][13][UTT][ MSE = 0.2080 | LCC = 0.9510 | SRCC = 0.9529 ] +INFO:root:Average loss=0.07998515406250954 +INFO:root: +[Valid][14][UTT][ MSE = 0.1922 | LCC = 0.9526 | SRCC = 0.9532 ] +INFO:root:Average loss=0.07746104452013969 +INFO:root: +[Valid][15][UTT][ MSE = 0.2149 | LCC = 0.9533 | SRCC = 0.9541 ] +INFO:root:Average loss=0.0739516843855381 +INFO:root: +[Valid][16][UTT][ MSE = 0.2148 | LCC = 0.9540 | SRCC = 0.9548 ] +INFO:root:Average loss=0.07269917249679565 +INFO:root: +[Valid][17][UTT][ MSE = 0.2061 | LCC = 0.9518 | SRCC = 0.9529 ] +INFO:root:Average loss=0.07102503174543381 +INFO:root: +[Valid][18][UTT][ MSE = 0.1913 | LCC = 0.9535 | SRCC = 0.9546 ] +INFO:root:Average loss=0.06689457455277444 +INFO:root: +[Valid][19][UTT][ MSE = 0.1840 | LCC = 0.9522 | SRCC = 0.9533 ] +INFO:root:Average loss=0.061735625743865964 +INFO:root: +[Valid][20][UTT][ MSE = 0.1769 | LCC = 0.9536 | SRCC = 0.9548 ] +INFO:root:Average loss=0.06207034243643284 +INFO:root: +[Valid][21][UTT][ MSE = 0.1927 | LCC = 0.9545 | SRCC = 0.9547 ] +INFO:root:Average loss=0.06016211724281311 +INFO:root: +[Valid][22][UTT][ MSE = 0.1810 | LCC = 0.9552 | SRCC = 0.9565 ] +INFO:root:Average loss=0.056521793842315676 +INFO:root: +[Valid][23][UTT][ MSE = 0.2180 | LCC = 0.9558 | SRCC = 0.9569 ] +INFO:root:Average loss=0.055744273692369464 +INFO:root: +[Valid][24][UTT][ MSE = 0.1898 | LCC = 0.9554 | SRCC = 0.9561 ] +INFO:root:Average loss=0.05494032338261604 +INFO:root: +[Valid][25][UTT][ MSE = 0.1852 | LCC = 0.9549 | SRCC = 0.9566 ] +INFO:root:Average loss=0.05170549313724041 +INFO:root: +[Valid][26][UTT][ MSE = 0.1680 | LCC = 0.9553 | SRCC = 0.9563 ] +INFO:root:Average loss=0.05176024588942528 +INFO:root: +[Valid][27][UTT][ MSE = 0.1901 | LCC = 0.9561 | SRCC = 0.9569 ] +INFO:root:Average loss=0.05233112198114395 +INFO:root: +[Valid][28][UTT][ MSE = 0.1874 | LCC = 0.9564 | SRCC = 0.9572 ] +INFO:root:Average loss=0.048079297810792926 +INFO:root: +[Valid][29][UTT][ MSE = 0.1772 | LCC = 0.9563 | SRCC = 0.9569 ] +INFO:root:Average loss=0.04482141077518463 +INFO:root: +[Valid][30][UTT][ MSE = 0.1914 | LCC = 0.9565 | SRCC = 0.9573 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1959 | LCC = 0.9558 | SRCC = 0.9570 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e229aed3b7f7bae4e126ab93bf57e7d24d26d852 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fdbf286013b473da348b280cdfb609249946ca8288e10da0cf1fb3d279ed197d +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..340e0d9ff529c64815fb53da94e56ac503a53015 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b0f04fbd70f47903e4ff64b9f821e4ffaf5d09bd59a9a7fb07d2f8e7559a9fc +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ea11aca85ebc0d5c65cc911cbc4213b79160de7f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer0/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4772605905532837 +INFO:root: +[Valid][0][UTT][ MSE = 0.3082 | LCC = 0.9094 | SRCC = 0.9109 ] +INFO:root:Average loss=0.22504019314050674 +INFO:root: +[Valid][1][UTT][ MSE = 0.2836 | LCC = 0.9277 | SRCC = 0.9298 ] +INFO:root:Average loss=0.1846683020591736 +INFO:root: +[Valid][2][UTT][ MSE = 0.2925 | LCC = 0.9311 | SRCC = 0.9319 ] +INFO:root:Average loss=0.16405193334817886 +INFO:root: +[Valid][3][UTT][ MSE = 0.2882 | LCC = 0.9347 | SRCC = 0.9361 ] +INFO:root:Average loss=0.15065780055522918 +INFO:root: +[Valid][4][UTT][ MSE = 0.2402 | LCC = 0.9418 | SRCC = 0.9436 ] +INFO:root:Average loss=0.14236305910348893 +INFO:root: +[Valid][5][UTT][ MSE = 0.2422 | LCC = 0.9466 | SRCC = 0.9477 ] +INFO:root:Average loss=0.1246984764933586 +INFO:root: +[Valid][6][UTT][ MSE = 0.2277 | LCC = 0.9463 | SRCC = 0.9485 ] +INFO:root:Average loss=0.11651446953415871 +INFO:root: +[Valid][7][UTT][ MSE = 0.2687 | LCC = 0.9460 | SRCC = 0.9486 ] +INFO:root:Average loss=0.11018335476517677 +INFO:root: +[Valid][8][UTT][ MSE = 0.2363 | LCC = 0.9500 | SRCC = 0.9506 ] +INFO:root:Average loss=0.10530276924371719 +INFO:root: +[Valid][9][UTT][ MSE = 0.2328 | LCC = 0.9507 | SRCC = 0.9510 ] +INFO:root:Average loss=0.10099353104829788 +INFO:root: +[Valid][10][UTT][ MSE = 0.2389 | LCC = 0.9510 | SRCC = 0.9517 ] +INFO:root:Average loss=0.09486592125892639 +INFO:root: +[Valid][11][UTT][ MSE = 0.2169 | LCC = 0.9522 | SRCC = 0.9529 ] +INFO:root:Average loss=0.08866133490204811 +INFO:root: +[Valid][12][UTT][ MSE = 0.2196 | LCC = 0.9517 | SRCC = 0.9526 ] +INFO:root:Average loss=0.08628150519728661 +INFO:root: +[Valid][13][UTT][ MSE = 0.2028 | LCC = 0.9541 | SRCC = 0.9549 ] +INFO:root:Average loss=0.0822636761367321 +INFO:root: +[Valid][14][UTT][ MSE = 0.2041 | LCC = 0.9516 | SRCC = 0.9538 ] +INFO:root:Average loss=0.07835948354005813 +INFO:root: +[Valid][15][UTT][ MSE = 0.1778 | LCC = 0.9541 | SRCC = 0.9550 ] +INFO:root:Average loss=0.07694206964969635 +INFO:root: +[Valid][16][UTT][ MSE = 0.1810 | LCC = 0.9539 | SRCC = 0.9551 ] +INFO:root:Average loss=0.0757912979722023 +INFO:root: +[Valid][17][UTT][ MSE = 0.1925 | LCC = 0.9545 | SRCC = 0.9555 ] +INFO:root:Average loss=0.06967092986404896 +INFO:root: +[Valid][18][UTT][ MSE = 0.2076 | LCC = 0.9544 | SRCC = 0.9551 ] +INFO:root:Average loss=0.06928732758760452 +INFO:root: +[Valid][19][UTT][ MSE = 0.2063 | LCC = 0.9528 | SRCC = 0.9540 ] +INFO:root:Average loss=0.06516571989655495 +INFO:root: +[Valid][20][UTT][ MSE = 0.2189 | LCC = 0.9568 | SRCC = 0.9578 ] +INFO:root:Average loss=0.060450150102376934 +INFO:root: +[Valid][21][UTT][ MSE = 0.1943 | LCC = 0.9555 | SRCC = 0.9564 ] +INFO:root:Average loss=0.060271236896514896 +INFO:root: +[Valid][22][UTT][ MSE = 0.1850 | LCC = 0.9566 | SRCC = 0.9571 ] +INFO:root:Average loss=0.057836766988039016 +INFO:root: +[Valid][23][UTT][ MSE = 0.2084 | LCC = 0.9573 | SRCC = 0.9579 ] +INFO:root:Average loss=0.05773171335458756 +INFO:root: +[Valid][24][UTT][ MSE = 0.1918 | LCC = 0.9565 | SRCC = 0.9573 ] +INFO:root:Average loss=0.05259528355300427 +INFO:root: +[Valid][25][UTT][ MSE = 0.1991 | LCC = 0.9568 | SRCC = 0.9576 ] +INFO:root:Average loss=0.05175529520213604 +INFO:root: +[Valid][26][UTT][ MSE = 0.1885 | LCC = 0.9526 | SRCC = 0.9546 ] +INFO:root:Average loss=0.050316621914505956 +INFO:root: +[Valid][27][UTT][ MSE = 0.1891 | LCC = 0.9567 | SRCC = 0.9578 ] +INFO:root:Average loss=0.050547399833798405 +INFO:root: +[Valid][28][UTT][ MSE = 0.1982 | LCC = 0.9564 | SRCC = 0.9571 ] +INFO:root:Average loss=0.049238622665405275 +INFO:root: +[Valid][29][UTT][ MSE = 0.1805 | LCC = 0.9567 | SRCC = 0.9581 ] +INFO:root:Average loss=0.04629709769785404 +INFO:root: +[Valid][30][UTT][ MSE = 0.1886 | LCC = 0.9570 | SRCC = 0.9574 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2046 | LCC = 0.9579 | SRCC = 0.9587 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4964edba9e8cb77945dc5aa61cd8bfd2086cc770 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2ac5304884a6f6dfd0a8c4802631fdf1ed3c7f899126bf86a466db3486f0afa4 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..aec3b72f60d66e39675a97a6050d4bcc70059328 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ccc9c5c53b63bd79cdd061e1b46a8021063e3a587db798782de579dbe60c8182 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3b0333deb243895bd764156e7295bab4ac9d49d1 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4500472741127014 +INFO:root: +[Valid][0][UTT][ MSE = 0.2789 | LCC = 0.9197 | SRCC = 0.9203 ] +INFO:root:Average loss=0.20433527833223342 +INFO:root: +[Valid][1][UTT][ MSE = 0.2797 | LCC = 0.9318 | SRCC = 0.9326 ] +INFO:root:Average loss=0.16176993358135222 +INFO:root: +[Valid][2][UTT][ MSE = 0.2310 | LCC = 0.9398 | SRCC = 0.9407 ] +INFO:root:Average loss=0.1428378913998604 +INFO:root: +[Valid][3][UTT][ MSE = 0.2475 | LCC = 0.9447 | SRCC = 0.9467 ] +INFO:root:Average loss=0.12786625027656556 +INFO:root: +[Valid][4][UTT][ MSE = 0.2463 | LCC = 0.9472 | SRCC = 0.9484 ] +INFO:root:Average loss=0.11541675335168838 +INFO:root: +[Valid][5][UTT][ MSE = 0.2294 | LCC = 0.9498 | SRCC = 0.9505 ] +INFO:root:Average loss=0.10985802453756333 +INFO:root: +[Valid][6][UTT][ MSE = 0.2076 | LCC = 0.9504 | SRCC = 0.9522 ] +INFO:root:Average loss=0.10493680426478386 +INFO:root: +[Valid][7][UTT][ MSE = 0.2324 | LCC = 0.9531 | SRCC = 0.9540 ] +INFO:root:Average loss=0.09596617126464843 +INFO:root: +[Valid][8][UTT][ MSE = 0.2146 | LCC = 0.9517 | SRCC = 0.9521 ] +INFO:root:Average loss=0.08954203376173973 +INFO:root: +[Valid][9][UTT][ MSE = 0.2104 | LCC = 0.9536 | SRCC = 0.9546 ] +INFO:root:Average loss=0.08462690344452858 +INFO:root: +[Valid][10][UTT][ MSE = 0.2074 | LCC = 0.9545 | SRCC = 0.9549 ] +INFO:root:Average loss=0.07801691818237305 +INFO:root: +[Valid][11][UTT][ MSE = 0.1972 | LCC = 0.9528 | SRCC = 0.9534 ] +INFO:root:Average loss=0.0745689300596714 +INFO:root: +[Valid][12][UTT][ MSE = 0.2349 | LCC = 0.9550 | SRCC = 0.9554 ] +INFO:root:Average loss=0.07305534061789512 +INFO:root: +[Valid][13][UTT][ MSE = 0.2492 | LCC = 0.9563 | SRCC = 0.9574 ] +INFO:root:Average loss=0.06769965994358063 +INFO:root: +[Valid][14][UTT][ MSE = 0.2446 | LCC = 0.9541 | SRCC = 0.9550 ] +INFO:root:Average loss=0.06492629688978195 +INFO:root: +[Valid][15][UTT][ MSE = 0.2156 | LCC = 0.9560 | SRCC = 0.9565 ] +INFO:root:Average loss=0.06270226949453354 +INFO:root: +[Valid][16][UTT][ MSE = 0.2212 | LCC = 0.9563 | SRCC = 0.9570 ] +INFO:root:Average loss=0.06066493573784828 +INFO:root: +[Valid][17][UTT][ MSE = 0.2369 | LCC = 0.9552 | SRCC = 0.9563 ] +INFO:root:Average loss=0.06000835597515106 +INFO:root: +[Valid][18][UTT][ MSE = 0.2397 | LCC = 0.9576 | SRCC = 0.9582 ] +INFO:root:Average loss=0.05547741276025772 +INFO:root: +[Valid][19][UTT][ MSE = 0.2159 | LCC = 0.9572 | SRCC = 0.9574 ] +INFO:root:Average loss=0.05259480139613151 +INFO:root: +[Valid][20][UTT][ MSE = 0.1928 | LCC = 0.9571 | SRCC = 0.9571 ] +INFO:root:Average loss=0.047629774898290636 +INFO:root: +[Valid][21][UTT][ MSE = 0.2032 | LCC = 0.9577 | SRCC = 0.9578 ] +INFO:root:Average loss=0.04865344867110252 +INFO:root: +[Valid][22][UTT][ MSE = 0.2129 | LCC = 0.9567 | SRCC = 0.9564 ] +INFO:root:Average loss=0.04823714351654053 +INFO:root: +[Valid][23][UTT][ MSE = 0.2089 | LCC = 0.9570 | SRCC = 0.9575 ] +INFO:root:Average loss=0.04681071637570858 +INFO:root: +[Valid][24][UTT][ MSE = 0.2348 | LCC = 0.9580 | SRCC = 0.9589 ] +INFO:root:Average loss=0.04272075526416302 +INFO:root: +[Valid][25][UTT][ MSE = 0.1979 | LCC = 0.9567 | SRCC = 0.9574 ] +INFO:root:Average loss=0.04095559187233448 +INFO:root: +[Valid][26][UTT][ MSE = 0.2286 | LCC = 0.9585 | SRCC = 0.9591 ] +INFO:root:Average loss=0.03948648536205292 +INFO:root: +[Valid][27][UTT][ MSE = 0.1999 | LCC = 0.9571 | SRCC = 0.9575 ] +INFO:root:Average loss=0.04199828180670738 +INFO:root: +[Valid][28][UTT][ MSE = 0.1821 | LCC = 0.9572 | SRCC = 0.9578 ] +INFO:root:Average loss=0.0392700120806694 +INFO:root: +[Valid][29][UTT][ MSE = 0.1943 | LCC = 0.9586 | SRCC = 0.9587 ] +INFO:root:Average loss=0.03773244695365429 +INFO:root: +[Valid][30][UTT][ MSE = 0.2070 | LCC = 0.9569 | SRCC = 0.9570 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2041 | LCC = 0.9588 | SRCC = 0.9596 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d1a1faaa60eef1ad8cc5f0fea2155b1d4e0fa598 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd1ef69ca021071b1c8e24f2174127b0a21890748003e2220d5dccb24b6ecc3b +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..59e8a4ab216b2d5bdd64f508e6917360a5341dcb --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0c113b9de92b0adac5b9a9c82f94b5711484c18ee52f72d8d5ad15bb32cd4fca +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..38cd2263a8c027265c5816e82961793f4c3a2c22 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4635176169872284 +INFO:root: +[Valid][0][UTT][ MSE = 0.3620 | LCC = 0.9219 | SRCC = 0.9232 ] +INFO:root:Average loss=0.19217748439311982 +INFO:root: +[Valid][1][UTT][ MSE = 0.2819 | LCC = 0.9379 | SRCC = 0.9400 ] +INFO:root:Average loss=0.1595679584145546 +INFO:root: +[Valid][2][UTT][ MSE = 0.2676 | LCC = 0.9413 | SRCC = 0.9423 ] +INFO:root:Average loss=0.13566535902023316 +INFO:root: +[Valid][3][UTT][ MSE = 0.2623 | LCC = 0.9467 | SRCC = 0.9477 ] +INFO:root:Average loss=0.12256775915622711 +INFO:root: +[Valid][4][UTT][ MSE = 0.2266 | LCC = 0.9500 | SRCC = 0.9514 ] +INFO:root:Average loss=0.116958043217659 +INFO:root: +[Valid][5][UTT][ MSE = 0.2248 | LCC = 0.9524 | SRCC = 0.9538 ] +INFO:root:Average loss=0.10920168888568878 +INFO:root: +[Valid][6][UTT][ MSE = 0.2455 | LCC = 0.9498 | SRCC = 0.9512 ] +INFO:root:Average loss=0.10146687352657319 +INFO:root: +[Valid][7][UTT][ MSE = 0.2248 | LCC = 0.9543 | SRCC = 0.9546 ] +INFO:root:Average loss=0.09929269450902939 +INFO:root: +[Valid][8][UTT][ MSE = 0.1909 | LCC = 0.9534 | SRCC = 0.9549 ] +INFO:root:Average loss=0.08665375411510468 +INFO:root: +[Valid][9][UTT][ MSE = 0.2518 | LCC = 0.9515 | SRCC = 0.9524 ] +INFO:root:Average loss=0.08430674141645432 +INFO:root: +[Valid][10][UTT][ MSE = 0.2321 | LCC = 0.9551 | SRCC = 0.9558 ] +INFO:root:Average loss=0.08371666437387466 +INFO:root: +[Valid][11][UTT][ MSE = 0.2035 | LCC = 0.9542 | SRCC = 0.9559 ] +INFO:root:Average loss=0.07672618541121483 +INFO:root: +[Valid][12][UTT][ MSE = 0.2021 | LCC = 0.9557 | SRCC = 0.9564 ] +INFO:root:Average loss=0.07150312146544456 +INFO:root: +[Valid][13][UTT][ MSE = 0.2027 | LCC = 0.9544 | SRCC = 0.9547 ] +INFO:root:Average loss=0.07065327090024948 +INFO:root: +[Valid][14][UTT][ MSE = 0.2038 | LCC = 0.9572 | SRCC = 0.9579 ] +INFO:root:Average loss=0.06390759727358818 +INFO:root: +[Valid][15][UTT][ MSE = 0.2132 | LCC = 0.9573 | SRCC = 0.9579 ] +INFO:root:Average loss=0.06396513548493385 +INFO:root: +[Valid][16][UTT][ MSE = 0.1791 | LCC = 0.9582 | SRCC = 0.9586 ] +INFO:root:Average loss=0.061278361827135086 +INFO:root: +[Valid][17][UTT][ MSE = 0.1955 | LCC = 0.9559 | SRCC = 0.9561 ] +INFO:root:Average loss=0.05802006432414055 +INFO:root: +[Valid][18][UTT][ MSE = 0.1759 | LCC = 0.9575 | SRCC = 0.9584 ] +INFO:root:Average loss=0.05406341490149498 +INFO:root: +[Valid][19][UTT][ MSE = 0.1815 | LCC = 0.9590 | SRCC = 0.9599 ] +INFO:root:Average loss=0.052382520988583564 +INFO:root: +[Valid][20][UTT][ MSE = 0.2011 | LCC = 0.9588 | SRCC = 0.9592 ] +INFO:root:Average loss=0.049562449678778646 +INFO:root: +[Valid][21][UTT][ MSE = 0.2059 | LCC = 0.9589 | SRCC = 0.9595 ] +INFO:root:Average loss=0.048826723337173465 +INFO:root: +[Valid][22][UTT][ MSE = 0.1862 | LCC = 0.9589 | SRCC = 0.9588 ] +INFO:root:Average loss=0.04799194410443306 +INFO:root: +[Valid][23][UTT][ MSE = 0.1948 | LCC = 0.9588 | SRCC = 0.9597 ] +INFO:root:Average loss=0.04488111518323421 +INFO:root: +[Valid][24][UTT][ MSE = 0.1788 | LCC = 0.9597 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04297950111329556 +INFO:root: +[Valid][25][UTT][ MSE = 0.1785 | LCC = 0.9592 | SRCC = 0.9599 ] +INFO:root:Average loss=0.041352280691266056 +INFO:root: +[Valid][26][UTT][ MSE = 0.1859 | LCC = 0.9590 | SRCC = 0.9593 ] +INFO:root:Average loss=0.041041877955198286 +INFO:root: +[Valid][27][UTT][ MSE = 0.2314 | LCC = 0.9588 | SRCC = 0.9600 ] +INFO:root:Average loss=0.04062087400257587 +INFO:root: +[Valid][28][UTT][ MSE = 0.2003 | LCC = 0.9593 | SRCC = 0.9599 ] +INFO:root:Average loss=0.03715365779399872 +INFO:root: +[Valid][29][UTT][ MSE = 0.1796 | LCC = 0.9603 | SRCC = 0.9609 ] +INFO:root:Average loss=0.03784128598868847 +INFO:root: +[Valid][30][UTT][ MSE = 0.1965 | LCC = 0.9598 | SRCC = 0.9599 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1861 | LCC = 0.9586 | SRCC = 0.9592 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e9eeecfbf9cf974737fbf33abf421fc5857bc8ca --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a9207ae68eea91f1081380c1749a55a250e35d9abd42486aec9e29c147b5c55f +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0bd9bd7635a8dfc6041d0937468b1e26a7c92fdf --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:feb6e2825d6ee8323c2f50a42aa0c300f772bbb0ead8cd6ae246fa2ea0f217ea +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..8338c89603aa71820fc08c5da354d94de5aa2e43 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6177280603647232 +INFO:root: +[Valid][0][UTT][ MSE = 0.3333 | LCC = 0.9179 | SRCC = 0.9203 ] +INFO:root:Average loss=0.19369201219081877 +INFO:root: +[Valid][1][UTT][ MSE = 0.2629 | LCC = 0.9361 | SRCC = 0.9376 ] +INFO:root:Average loss=0.1556881166100502 +INFO:root: +[Valid][2][UTT][ MSE = 0.2472 | LCC = 0.9429 | SRCC = 0.9443 ] +INFO:root:Average loss=0.13654927444458007 +INFO:root: +[Valid][3][UTT][ MSE = 0.2471 | LCC = 0.9454 | SRCC = 0.9456 ] +INFO:root:Average loss=0.12032974070310593 +INFO:root: +[Valid][4][UTT][ MSE = 0.2125 | LCC = 0.9446 | SRCC = 0.9445 ] +INFO:root:Average loss=0.11512114125490189 +INFO:root: +[Valid][5][UTT][ MSE = 0.2066 | LCC = 0.9509 | SRCC = 0.9508 ] +INFO:root:Average loss=0.106413579672575 +INFO:root: +[Valid][6][UTT][ MSE = 0.2196 | LCC = 0.9515 | SRCC = 0.9518 ] +INFO:root:Average loss=0.10049265229701995 +INFO:root: +[Valid][7][UTT][ MSE = 0.2097 | LCC = 0.9529 | SRCC = 0.9528 ] +INFO:root:Average loss=0.09091598555445671 +INFO:root: +[Valid][8][UTT][ MSE = 0.2040 | LCC = 0.9523 | SRCC = 0.9528 ] +INFO:root:Average loss=0.08523266077041626 +INFO:root: +[Valid][9][UTT][ MSE = 0.2590 | LCC = 0.9523 | SRCC = 0.9544 ] +INFO:root:Average loss=0.08158923119306564 +INFO:root: +[Valid][10][UTT][ MSE = 0.1823 | LCC = 0.9529 | SRCC = 0.9539 ] +INFO:root:Average loss=0.07743859994411469 +INFO:root: +[Valid][11][UTT][ MSE = 0.2141 | LCC = 0.9554 | SRCC = 0.9555 ] +INFO:root:Average loss=0.07415051403641701 +INFO:root: +[Valid][12][UTT][ MSE = 0.2004 | LCC = 0.9549 | SRCC = 0.9552 ] +INFO:root:Average loss=0.07020535933971404 +INFO:root: +[Valid][13][UTT][ MSE = 0.2066 | LCC = 0.9546 | SRCC = 0.9541 ] +INFO:root:Average loss=0.06480990904569627 +INFO:root: +[Valid][14][UTT][ MSE = 0.2095 | LCC = 0.9533 | SRCC = 0.9531 ] +INFO:root:Average loss=0.0634517513513565 +INFO:root: +[Valid][15][UTT][ MSE = 0.2075 | LCC = 0.9552 | SRCC = 0.9551 ] +INFO:root:Average loss=0.059521265625953675 +INFO:root: +[Valid][16][UTT][ MSE = 0.2152 | LCC = 0.9564 | SRCC = 0.9570 ] +INFO:root:Average loss=0.05698331767320633 +INFO:root: +[Valid][17][UTT][ MSE = 0.1911 | LCC = 0.9570 | SRCC = 0.9570 ] +INFO:root:Average loss=0.053992869734764096 +INFO:root: +[Valid][18][UTT][ MSE = 0.2389 | LCC = 0.9558 | SRCC = 0.9573 ] +INFO:root:Average loss=0.05465609467029572 +INFO:root: +[Valid][19][UTT][ MSE = 0.1847 | LCC = 0.9586 | SRCC = 0.9582 ] +INFO:root:Average loss=0.051099047273397445 +INFO:root: +[Valid][20][UTT][ MSE = 0.1915 | LCC = 0.9589 | SRCC = 0.9582 ] +INFO:root:Average loss=0.048738373339176175 +INFO:root: +[Valid][21][UTT][ MSE = 0.2048 | LCC = 0.9590 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04653130292892456 +INFO:root: +[Valid][22][UTT][ MSE = 0.1879 | LCC = 0.9576 | SRCC = 0.9573 ] +INFO:root:Average loss=0.04436153754591942 +INFO:root: +[Valid][23][UTT][ MSE = 0.2092 | LCC = 0.9586 | SRCC = 0.9585 ] +INFO:root:Average loss=0.04412468591332436 +INFO:root: +[Valid][24][UTT][ MSE = 0.2006 | LCC = 0.9568 | SRCC = 0.9570 ] +INFO:root:Average loss=0.04204258669912815 +INFO:root: +[Valid][25][UTT][ MSE = 0.1935 | LCC = 0.9595 | SRCC = 0.9594 ] +INFO:root:Average loss=0.03908932501077652 +INFO:root: +[Valid][26][UTT][ MSE = 0.1936 | LCC = 0.9558 | SRCC = 0.9557 ] +INFO:root:Average loss=0.0396658326536417 +INFO:root: +[Valid][27][UTT][ MSE = 0.1949 | LCC = 0.9588 | SRCC = 0.9586 ] +INFO:root:Average loss=0.037008364289999006 +INFO:root: +[Valid][28][UTT][ MSE = 0.1810 | LCC = 0.9580 | SRCC = 0.9577 ] +INFO:root:Average loss=0.037497466996312144 +INFO:root: +[Valid][29][UTT][ MSE = 0.1912 | LCC = 0.9598 | SRCC = 0.9596 ] +INFO:root:Average loss=0.03604000946879387 +INFO:root: +[Valid][30][UTT][ MSE = 0.2042 | LCC = 0.9592 | SRCC = 0.9591 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1993 | LCC = 0.9582 | SRCC = 0.9586 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..67f4b843e2f367e3582da9ce7d9e87d0fe06c5a1 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e20b42d3e1ded7470770e294b1a83ad97834d70c14746e6f3adfe9102a42d1b +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b96c1dec9710b7971fe2000a8a59da21bbed380f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:430d57c17f1e3c56725074204f1089635bd2359f949c885e337169cdb862a53b +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..087a199570cf97db17bdde631dd94f184b4e53c6 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.45262454760074616 +INFO:root: +[Valid][0][UTT][ MSE = 0.3032 | LCC = 0.9178 | SRCC = 0.9195 ] +INFO:root:Average loss=0.18623159247636795 +INFO:root: +[Valid][1][UTT][ MSE = 0.2657 | LCC = 0.9358 | SRCC = 0.9375 ] +INFO:root:Average loss=0.1524810455441475 +INFO:root: +[Valid][2][UTT][ MSE = 0.2398 | LCC = 0.9432 | SRCC = 0.9442 ] +INFO:root:Average loss=0.1343418537378311 +INFO:root: +[Valid][3][UTT][ MSE = 0.2291 | LCC = 0.9474 | SRCC = 0.9482 ] +INFO:root:Average loss=0.12516494524478913 +INFO:root: +[Valid][4][UTT][ MSE = 0.2624 | LCC = 0.9454 | SRCC = 0.9460 ] +INFO:root:Average loss=0.11201304593682289 +INFO:root: +[Valid][5][UTT][ MSE = 0.2255 | LCC = 0.9506 | SRCC = 0.9524 ] +INFO:root:Average loss=0.10731234589219094 +INFO:root: +[Valid][6][UTT][ MSE = 0.2128 | LCC = 0.9524 | SRCC = 0.9528 ] +INFO:root:Average loss=0.0954464130103588 +INFO:root: +[Valid][7][UTT][ MSE = 0.2422 | LCC = 0.9527 | SRCC = 0.9534 ] +INFO:root:Average loss=0.09017503899335862 +INFO:root: +[Valid][8][UTT][ MSE = 0.2658 | LCC = 0.9546 | SRCC = 0.9552 ] +INFO:root:Average loss=0.08469965368509293 +INFO:root: +[Valid][9][UTT][ MSE = 0.2320 | LCC = 0.9533 | SRCC = 0.9540 ] +INFO:root:Average loss=0.08093800154328347 +INFO:root: +[Valid][10][UTT][ MSE = 0.2176 | LCC = 0.9520 | SRCC = 0.9534 ] +INFO:root:Average loss=0.07656704732775688 +INFO:root: +[Valid][11][UTT][ MSE = 0.2237 | LCC = 0.9555 | SRCC = 0.9559 ] +INFO:root:Average loss=0.06991558119654656 +INFO:root: +[Valid][12][UTT][ MSE = 0.2440 | LCC = 0.9532 | SRCC = 0.9538 ] +INFO:root:Average loss=0.06866298642754555 +INFO:root: +[Valid][13][UTT][ MSE = 0.2144 | LCC = 0.9557 | SRCC = 0.9558 ] +INFO:root:Average loss=0.06716896757483483 +INFO:root: +[Valid][14][UTT][ MSE = 0.1998 | LCC = 0.9573 | SRCC = 0.9574 ] +INFO:root:Average loss=0.061635677829384804 +INFO:root: +[Valid][15][UTT][ MSE = 0.2140 | LCC = 0.9567 | SRCC = 0.9573 ] +INFO:root:Average loss=0.056281558960676194 +INFO:root: +[Valid][16][UTT][ MSE = 0.2028 | LCC = 0.9567 | SRCC = 0.9567 ] +INFO:root:Average loss=0.05691093796491623 +INFO:root: +[Valid][17][UTT][ MSE = 0.1973 | LCC = 0.9561 | SRCC = 0.9563 ] +INFO:root:Average loss=0.0516981550604105 +INFO:root: +[Valid][18][UTT][ MSE = 0.2043 | LCC = 0.9572 | SRCC = 0.9576 ] +INFO:root:Average loss=0.05003721210360527 +INFO:root: +[Valid][19][UTT][ MSE = 0.1947 | LCC = 0.9578 | SRCC = 0.9581 ] +INFO:root:Average loss=0.049761462792754174 +INFO:root: +[Valid][20][UTT][ MSE = 0.2280 | LCC = 0.9573 | SRCC = 0.9570 ] +INFO:root:Average loss=0.047563161954283714 +INFO:root: +[Valid][21][UTT][ MSE = 0.2247 | LCC = 0.9580 | SRCC = 0.9584 ] +INFO:root:Average loss=0.0481893759071827 +INFO:root: +[Valid][22][UTT][ MSE = 0.2183 | LCC = 0.9579 | SRCC = 0.9581 ] +INFO:root:Average loss=0.04366876816749573 +INFO:root: +[Valid][23][UTT][ MSE = 0.2280 | LCC = 0.9585 | SRCC = 0.9586 ] +INFO:root:Average loss=0.04312713938951492 +INFO:root: +[Valid][24][UTT][ MSE = 0.1885 | LCC = 0.9589 | SRCC = 0.9595 ] +INFO:root:Average loss=0.0403843564838171 +INFO:root: +[Valid][25][UTT][ MSE = 0.2203 | LCC = 0.9584 | SRCC = 0.9583 ] +INFO:root:Average loss=0.039315524622797964 +INFO:root: +[Valid][26][UTT][ MSE = 0.2069 | LCC = 0.9589 | SRCC = 0.9589 ] +INFO:root:Average loss=0.038045583084225655 +INFO:root: +[Valid][27][UTT][ MSE = 0.1929 | LCC = 0.9601 | SRCC = 0.9605 ] +INFO:root:Average loss=0.0374175681322813 +INFO:root: +[Valid][28][UTT][ MSE = 0.2082 | LCC = 0.9589 | SRCC = 0.9593 ] +INFO:root:Average loss=0.0360670952796936 +INFO:root: +[Valid][29][UTT][ MSE = 0.2061 | LCC = 0.9598 | SRCC = 0.9604 ] +INFO:root:Average loss=0.03521760991215706 +INFO:root: +[Valid][30][UTT][ MSE = 0.1969 | LCC = 0.9591 | SRCC = 0.9594 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1942 | LCC = 0.9591 | SRCC = 0.9598 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f56d87d1fb77dfaaba2a21784cbe348d653a2577 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:45b42544a68c2ce9056f794e28e25ddc315145110cabb49f22aa9cf64dc9b6f9 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..8dcaa73eebe5f58d9247153380e7842f9008d33d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b675e156230a18a5a0589ace070b68311d4a2ab83b7b62920b6c964454b8f0b9 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a3490739b465b9ebf19e993d0037d3221dfca244 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer1/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3936773397922516 +INFO:root: +[Valid][0][UTT][ MSE = 0.3155 | LCC = 0.9258 | SRCC = 0.9262 ] +INFO:root:Average loss=0.1838404302597046 +INFO:root: +[Valid][1][UTT][ MSE = 0.2702 | LCC = 0.9399 | SRCC = 0.9410 ] +INFO:root:Average loss=0.15183659535646438 +INFO:root: +[Valid][2][UTT][ MSE = 0.2513 | LCC = 0.9439 | SRCC = 0.9446 ] +INFO:root:Average loss=0.13422144782543183 +INFO:root: +[Valid][3][UTT][ MSE = 0.2476 | LCC = 0.9442 | SRCC = 0.9449 ] +INFO:root:Average loss=0.12462042087316513 +INFO:root: +[Valid][4][UTT][ MSE = 0.2597 | LCC = 0.9484 | SRCC = 0.9490 ] +INFO:root:Average loss=0.11024944385886193 +INFO:root: +[Valid][5][UTT][ MSE = 0.2239 | LCC = 0.9521 | SRCC = 0.9523 ] +INFO:root:Average loss=0.10433466479182243 +INFO:root: +[Valid][6][UTT][ MSE = 0.2348 | LCC = 0.9525 | SRCC = 0.9528 ] +INFO:root:Average loss=0.0977769292294979 +INFO:root: +[Valid][7][UTT][ MSE = 0.2030 | LCC = 0.9534 | SRCC = 0.9542 ] +INFO:root:Average loss=0.09213896712660789 +INFO:root: +[Valid][8][UTT][ MSE = 0.2006 | LCC = 0.9549 | SRCC = 0.9552 ] +INFO:root:Average loss=0.085471471875906 +INFO:root: +[Valid][9][UTT][ MSE = 0.1725 | LCC = 0.9557 | SRCC = 0.9559 ] +INFO:root:Average loss=0.08093336790800094 +INFO:root: +[Valid][10][UTT][ MSE = 0.1957 | LCC = 0.9540 | SRCC = 0.9540 ] +INFO:root:Average loss=0.0817233034670353 +INFO:root: +[Valid][11][UTT][ MSE = 0.1918 | LCC = 0.9560 | SRCC = 0.9561 ] +INFO:root:Average loss=0.07112137520313264 +INFO:root: +[Valid][12][UTT][ MSE = 0.2006 | LCC = 0.9555 | SRCC = 0.9553 ] +INFO:root:Average loss=0.06960062298178672 +INFO:root: +[Valid][13][UTT][ MSE = 0.2219 | LCC = 0.9512 | SRCC = 0.9552 ] +INFO:root:Average loss=0.06663702896237374 +INFO:root: +[Valid][14][UTT][ MSE = 0.2065 | LCC = 0.9573 | SRCC = 0.9574 ] +INFO:root:Average loss=0.06290504771471024 +INFO:root: +[Valid][15][UTT][ MSE = 0.1828 | LCC = 0.9567 | SRCC = 0.9574 ] +INFO:root:Average loss=0.061189590513706205 +INFO:root: +[Valid][16][UTT][ MSE = 0.1981 | LCC = 0.9568 | SRCC = 0.9576 ] +INFO:root:Average loss=0.05451443275809288 +INFO:root: +[Valid][17][UTT][ MSE = 0.1910 | LCC = 0.9563 | SRCC = 0.9562 ] +INFO:root:Average loss=0.055239839166402814 +INFO:root: +[Valid][18][UTT][ MSE = 0.1812 | LCC = 0.9570 | SRCC = 0.9569 ] +INFO:root:Average loss=0.05165509669482708 +INFO:root: +[Valid][19][UTT][ MSE = 0.1886 | LCC = 0.9583 | SRCC = 0.9580 ] +INFO:root:Average loss=0.048803542584180834 +INFO:root: +[Valid][20][UTT][ MSE = 0.1958 | LCC = 0.9583 | SRCC = 0.9578 ] +INFO:root:Average loss=0.04708941204845905 +INFO:root: +[Valid][21][UTT][ MSE = 0.1935 | LCC = 0.9564 | SRCC = 0.9561 ] +INFO:root:Average loss=0.04686663772165775 +INFO:root: +[Valid][22][UTT][ MSE = 0.1917 | LCC = 0.9583 | SRCC = 0.9579 ] +INFO:root:Average loss=0.04644606120884418 +INFO:root: +[Valid][23][UTT][ MSE = 0.2033 | LCC = 0.9580 | SRCC = 0.9586 ] +INFO:root:Average loss=0.042809940010309216 +INFO:root: +[Valid][24][UTT][ MSE = 0.1773 | LCC = 0.9584 | SRCC = 0.9583 ] +INFO:root:Average loss=0.045390093833208085 +INFO:root: +[Valid][25][UTT][ MSE = 0.1877 | LCC = 0.9576 | SRCC = 0.9575 ] +INFO:root:Average loss=0.04056744278967381 +INFO:root: +[Valid][26][UTT][ MSE = 0.1683 | LCC = 0.9565 | SRCC = 0.9563 ] +INFO:root:Average loss=0.042159137442708015 +INFO:root: +[Valid][27][UTT][ MSE = 0.1878 | LCC = 0.9580 | SRCC = 0.9579 ] +INFO:root:Average loss=0.037718096181750295 +INFO:root: +[Valid][28][UTT][ MSE = 0.1757 | LCC = 0.9583 | SRCC = 0.9579 ] +INFO:root:Average loss=0.03638336130976677 +INFO:root: +[Valid][29][UTT][ MSE = 0.1882 | LCC = 0.9580 | SRCC = 0.9576 ] +INFO:root:Average loss=0.03774034285545349 +INFO:root: +[Valid][30][UTT][ MSE = 0.2188 | LCC = 0.9563 | SRCC = 0.9561 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1747 | LCC = 0.9588 | SRCC = 0.9587 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f91718be801bcc560c11cf979249c14a9752ff02 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4599f5e4d2e0cc8960eb6b02a478aee389adf2812715633a99088eaaf598947d +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d6f5a060fcb83dcba07e1f15d3b41feaad61ad6b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:29625584726c2dd2795f1c4016c07033bac15934db6200482e1ce881a410528a +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0c5a3cc657aba2c35d2e62106f27510756fb7f08 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.24830776649713515 +INFO:root: +[Valid][0][UTT][ MSE = 0.3060 | LCC = 0.9419 | SRCC = 0.9436 ] +INFO:root:Average loss=0.15613619023561479 +INFO:root: +[Valid][1][UTT][ MSE = 0.2745 | LCC = 0.9457 | SRCC = 0.9469 ] +INFO:root:Average loss=0.1482595028281212 +INFO:root: +[Valid][2][UTT][ MSE = 0.2732 | LCC = 0.9445 | SRCC = 0.9492 ] +INFO:root:Average loss=0.14049203443527222 +INFO:root: +[Valid][3][UTT][ MSE = 0.2744 | LCC = 0.9491 | SRCC = 0.9516 ] +INFO:root:Average loss=0.13548963648080825 +INFO:root: +[Valid][4][UTT][ MSE = 0.2559 | LCC = 0.9502 | SRCC = 0.9517 ] +INFO:root:Average loss=0.13349572116136552 +INFO:root: +[Valid][5][UTT][ MSE = 0.2668 | LCC = 0.9515 | SRCC = 0.9528 ] +INFO:root:Average loss=0.12811619687080383 +INFO:root: +[Valid][6][UTT][ MSE = 0.3074 | LCC = 0.9465 | SRCC = 0.9505 ] +INFO:root:Average loss=0.12503455901145935 +INFO:root: +[Valid][7][UTT][ MSE = 0.2749 | LCC = 0.9526 | SRCC = 0.9531 ] +INFO:root:Average loss=0.12353295558691024 +INFO:root: +[Valid][8][UTT][ MSE = 0.2813 | LCC = 0.9540 | SRCC = 0.9543 ] +INFO:root:Average loss=0.12149932014942169 +INFO:root: +[Valid][9][UTT][ MSE = 0.2625 | LCC = 0.9537 | SRCC = 0.9544 ] +INFO:root:Average loss=0.11779827076196671 +INFO:root: +[Valid][10][UTT][ MSE = 0.2923 | LCC = 0.9539 | SRCC = 0.9542 ] +INFO:root:Average loss=0.11868463197350503 +INFO:root: +[Valid][11][UTT][ MSE = 0.3008 | LCC = 0.9489 | SRCC = 0.9535 ] +INFO:root:Average loss=0.11520671457052231 +INFO:root: +[Valid][12][UTT][ MSE = 0.2820 | LCC = 0.9519 | SRCC = 0.9533 ] +INFO:root:Average loss=0.11120937150716782 +INFO:root: +[Valid][13][UTT][ MSE = 0.2653 | LCC = 0.9528 | SRCC = 0.9549 ] +INFO:root:Average loss=0.11379290491342545 +INFO:root: +[Valid][14][UTT][ MSE = 0.2670 | LCC = 0.9538 | SRCC = 0.9557 ] +INFO:root:Average loss=0.1098162373304367 +INFO:root: +[Valid][15][UTT][ MSE = 0.2595 | LCC = 0.9539 | SRCC = 0.9548 ] +INFO:root:Average loss=0.11053543657064438 +INFO:root: +[Valid][16][UTT][ MSE = 0.2465 | LCC = 0.9540 | SRCC = 0.9555 ] +INFO:root:Average loss=0.10345384952425957 +INFO:root: +[Valid][17][UTT][ MSE = 0.2514 | LCC = 0.9534 | SRCC = 0.9546 ] +INFO:root:Average loss=0.10253187066316605 +INFO:root: +[Valid][18][UTT][ MSE = 0.2277 | LCC = 0.9523 | SRCC = 0.9546 ] +INFO:root:Average loss=0.10199417585134506 +INFO:root: +[Valid][19][UTT][ MSE = 0.2409 | LCC = 0.9525 | SRCC = 0.9555 ] +INFO:root:Average loss=0.1003849160373211 +INFO:root: +[Valid][20][UTT][ MSE = 0.2316 | LCC = 0.9538 | SRCC = 0.9553 ] +INFO:root:Average loss=0.0973091898560524 +INFO:root: +[Valid][21][UTT][ MSE = 0.2457 | LCC = 0.9526 | SRCC = 0.9541 ] +INFO:root:Average loss=0.09991709366440774 +INFO:root: +[Valid][22][UTT][ MSE = 0.2734 | LCC = 0.9503 | SRCC = 0.9536 ] +INFO:root:Average loss=0.09511888837814331 +INFO:root: +[Valid][23][UTT][ MSE = 0.2443 | LCC = 0.9529 | SRCC = 0.9549 ] +INFO:root:Average loss=0.09266128474473953 +INFO:root: +[Valid][24][UTT][ MSE = 0.2541 | LCC = 0.9527 | SRCC = 0.9549 ] +INFO:root:Average loss=0.09591928419470787 +INFO:root: +[Valid][25][UTT][ MSE = 0.2402 | LCC = 0.9523 | SRCC = 0.9537 ] +INFO:root:Average loss=0.09461772322654724 +INFO:root: +[Valid][26][UTT][ MSE = 0.2469 | LCC = 0.9521 | SRCC = 0.9545 ] +INFO:root:Average loss=0.08747722911834717 +INFO:root: +[Valid][27][UTT][ MSE = 0.2399 | LCC = 0.9514 | SRCC = 0.9536 ] +INFO:root:Average loss=0.08808277258276939 +INFO:root: +[Valid][28][UTT][ MSE = 0.2514 | LCC = 0.9520 | SRCC = 0.9529 ] +INFO:root:Average loss=0.08473639470338822 +INFO:root: +[Valid][29][UTT][ MSE = 0.2423 | LCC = 0.9525 | SRCC = 0.9538 ] +INFO:root:Average loss=0.0852522642314434 +INFO:root: +[Valid][30][UTT][ MSE = 0.2342 | LCC = 0.9519 | SRCC = 0.9542 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2530 | LCC = 0.9559 | SRCC = 0.9575 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4662f62f128a83d626ffb3cbceb8dd0712ff7225 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:53100f630d85808ee65e0aec7140b80841537d5d05c99f81be26e6d42b251978 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..5e0ff7a809f4c6df58a162974f90b2168e7014a0 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bc4be86b34ebb9320f9e6b2a81cd5992c76b3fbb15f93906807ea673a820507c +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..667b5bd4a5a35cd06903597a493071cbe98ea1fd --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.2737233681082726 +INFO:root: +[Valid][0][UTT][ MSE = 0.2964 | LCC = 0.9421 | SRCC = 0.9427 ] +INFO:root:Average loss=0.15617733877897264 +INFO:root: +[Valid][1][UTT][ MSE = 0.2734 | LCC = 0.9466 | SRCC = 0.9465 ] +INFO:root:Average loss=0.14722304379940032 +INFO:root: +[Valid][2][UTT][ MSE = 0.2976 | LCC = 0.9483 | SRCC = 0.9486 ] +INFO:root:Average loss=0.13902530092000961 +INFO:root: +[Valid][3][UTT][ MSE = 0.2602 | LCC = 0.9501 | SRCC = 0.9517 ] +INFO:root:Average loss=0.13231035679578781 +INFO:root: +[Valid][4][UTT][ MSE = 0.2599 | LCC = 0.9499 | SRCC = 0.9525 ] +INFO:root:Average loss=0.12944638049602508 +INFO:root: +[Valid][5][UTT][ MSE = 0.2536 | LCC = 0.9504 | SRCC = 0.9529 ] +INFO:root:Average loss=0.12917381209135056 +INFO:root: +[Valid][6][UTT][ MSE = 0.2547 | LCC = 0.9513 | SRCC = 0.9535 ] +INFO:root:Average loss=0.1234755374789238 +INFO:root: +[Valid][7][UTT][ MSE = 0.2799 | LCC = 0.9528 | SRCC = 0.9536 ] +INFO:root:Average loss=0.12109400635957718 +INFO:root: +[Valid][8][UTT][ MSE = 0.2348 | LCC = 0.9532 | SRCC = 0.9537 ] +INFO:root:Average loss=0.11889956891536713 +INFO:root: +[Valid][9][UTT][ MSE = 0.2422 | LCC = 0.9515 | SRCC = 0.9547 ] +INFO:root:Average loss=0.11813472127914429 +INFO:root: +[Valid][10][UTT][ MSE = 0.2794 | LCC = 0.9530 | SRCC = 0.9550 ] +INFO:root:Average loss=0.11462802025675774 +INFO:root: +[Valid][11][UTT][ MSE = 0.2914 | LCC = 0.9521 | SRCC = 0.9541 ] +INFO:root:Average loss=0.11601899340748786 +INFO:root: +[Valid][12][UTT][ MSE = 0.2539 | LCC = 0.9500 | SRCC = 0.9540 ] +INFO:root:Average loss=0.11404969751834869 +INFO:root: +[Valid][13][UTT][ MSE = 0.2491 | LCC = 0.9511 | SRCC = 0.9537 ] +INFO:root:Average loss=0.10840177634358406 +INFO:root: +[Valid][14][UTT][ MSE = 0.2506 | LCC = 0.9534 | SRCC = 0.9539 ] +INFO:root:Average loss=0.11118825632333755 +INFO:root: +[Valid][15][UTT][ MSE = 0.2561 | LCC = 0.9536 | SRCC = 0.9551 ] +INFO:root:Average loss=0.10387786296010017 +INFO:root: +[Valid][16][UTT][ MSE = 0.2511 | LCC = 0.9514 | SRCC = 0.9540 ] +INFO:root:Average loss=0.10369600325822831 +INFO:root: +[Valid][17][UTT][ MSE = 0.2851 | LCC = 0.9534 | SRCC = 0.9545 ] +INFO:root:Average loss=0.10356753012537956 +INFO:root: +[Valid][18][UTT][ MSE = 0.2260 | LCC = 0.9536 | SRCC = 0.9550 ] +INFO:root:Average loss=0.10072359296679496 +INFO:root: +[Valid][19][UTT][ MSE = 0.2599 | LCC = 0.9532 | SRCC = 0.9542 ] +INFO:root:Average loss=0.10073469120264053 +INFO:root: +[Valid][20][UTT][ MSE = 0.2530 | LCC = 0.9542 | SRCC = 0.9558 ] +INFO:root:Average loss=0.09792698216438293 +INFO:root: +[Valid][21][UTT][ MSE = 0.2546 | LCC = 0.9522 | SRCC = 0.9544 ] +INFO:root:Average loss=0.09730604267120362 +INFO:root: +[Valid][22][UTT][ MSE = 0.2671 | LCC = 0.9518 | SRCC = 0.9520 ] +INFO:root:Average loss=0.09430684992671012 +INFO:root: +[Valid][23][UTT][ MSE = 0.2677 | LCC = 0.9535 | SRCC = 0.9545 ] +INFO:root:Average loss=0.09196612733602524 +INFO:root: +[Valid][24][UTT][ MSE = 0.2284 | LCC = 0.9521 | SRCC = 0.9541 ] +INFO:root:Average loss=0.09236472225189209 +INFO:root: +[Valid][25][UTT][ MSE = 0.2530 | LCC = 0.9531 | SRCC = 0.9544 ] +INFO:root:Average loss=0.09066559624671935 +INFO:root: +[Valid][26][UTT][ MSE = 0.2286 | LCC = 0.9532 | SRCC = 0.9550 ] +INFO:root:Average loss=0.09077227047085762 +INFO:root: +[Valid][27][UTT][ MSE = 0.2457 | LCC = 0.9536 | SRCC = 0.9544 ] +INFO:root:Average loss=0.08698692944645882 +INFO:root: +[Valid][28][UTT][ MSE = 0.2448 | LCC = 0.9516 | SRCC = 0.9541 ] +INFO:root:Average loss=0.08533721524477005 +INFO:root: +[Valid][29][UTT][ MSE = 0.2479 | LCC = 0.9530 | SRCC = 0.9534 ] +INFO:root:Average loss=0.08582079228758811 +INFO:root: +[Valid][30][UTT][ MSE = 0.2548 | LCC = 0.9537 | SRCC = 0.9546 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2565 | LCC = 0.9567 | SRCC = 0.9585 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6c233d1823018fce18e91504c88cbff57f3dd834 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:13311ac8aa502cb082cae8f51502e6558a8da5ce184ae87ff2ff2e438925c2ab +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..95fad9f473dd58f5f43de5fdfb07abae6bdaae20 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6eafdff34af8079394e25a02ac81cf9f5bc2749eeb78695f1ccd747b99643c7a +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d175924718cf3f807f8fd0b12c3372a0d579dc5b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.2659680091738701 +INFO:root: +[Valid][0][UTT][ MSE = 0.3193 | LCC = 0.9416 | SRCC = 0.9416 ] +INFO:root:Average loss=0.15606255984306336 +INFO:root: +[Valid][1][UTT][ MSE = 0.2549 | LCC = 0.9404 | SRCC = 0.9460 ] +INFO:root:Average loss=0.14666409316658974 +INFO:root: +[Valid][2][UTT][ MSE = 0.2623 | LCC = 0.9490 | SRCC = 0.9498 ] +INFO:root:Average loss=0.14100806361436843 +INFO:root: +[Valid][3][UTT][ MSE = 0.2743 | LCC = 0.9472 | SRCC = 0.9503 ] +INFO:root:Average loss=0.13458540320396423 +INFO:root: +[Valid][4][UTT][ MSE = 0.2590 | LCC = 0.9499 | SRCC = 0.9519 ] +INFO:root:Average loss=0.1337956710457802 +INFO:root: +[Valid][5][UTT][ MSE = 0.2356 | LCC = 0.9513 | SRCC = 0.9520 ] +INFO:root:Average loss=0.1267787446975708 +INFO:root: +[Valid][6][UTT][ MSE = 0.2350 | LCC = 0.9519 | SRCC = 0.9531 ] +INFO:root:Average loss=0.12794397574663163 +INFO:root: +[Valid][7][UTT][ MSE = 0.2227 | LCC = 0.9516 | SRCC = 0.9524 ] +INFO:root:Average loss=0.120183580160141 +INFO:root: +[Valid][8][UTT][ MSE = 0.2807 | LCC = 0.9507 | SRCC = 0.9539 ] +INFO:root:Average loss=0.120121322453022 +INFO:root: +[Valid][9][UTT][ MSE = 0.2411 | LCC = 0.9527 | SRCC = 0.9538 ] +INFO:root:Average loss=0.12246333768963814 +INFO:root: +[Valid][10][UTT][ MSE = 0.2501 | LCC = 0.9529 | SRCC = 0.9537 ] +INFO:root:Average loss=0.1142940809726715 +INFO:root: +[Valid][11][UTT][ MSE = 0.2531 | LCC = 0.9496 | SRCC = 0.9520 ] +INFO:root:Average loss=0.11419139382243157 +INFO:root: +[Valid][12][UTT][ MSE = 0.2285 | LCC = 0.9508 | SRCC = 0.9530 ] +INFO:root:Average loss=0.1105129443705082 +INFO:root: +[Valid][13][UTT][ MSE = 0.2509 | LCC = 0.9508 | SRCC = 0.9531 ] +INFO:root:Average loss=0.11114811086654663 +INFO:root: +[Valid][14][UTT][ MSE = 0.2588 | LCC = 0.9525 | SRCC = 0.9537 ] +INFO:root:Average loss=0.10660567760467529 +INFO:root: +[Valid][15][UTT][ MSE = 0.2334 | LCC = 0.9536 | SRCC = 0.9546 ] +INFO:root:Average loss=0.10552438700199127 +INFO:root: +[Valid][16][UTT][ MSE = 0.2340 | LCC = 0.9518 | SRCC = 0.9537 ] +INFO:root:Average loss=0.10417621490359306 +INFO:root: +[Valid][17][UTT][ MSE = 0.2238 | LCC = 0.9533 | SRCC = 0.9538 ] +INFO:root:Average loss=0.10194973587989807 +INFO:root: +[Valid][18][UTT][ MSE = 0.2527 | LCC = 0.9517 | SRCC = 0.9549 ] +INFO:root:Average loss=0.1022991962581873 +INFO:root: +[Valid][19][UTT][ MSE = 0.2517 | LCC = 0.9530 | SRCC = 0.9535 ] +INFO:root:Average loss=0.09747041448950768 +INFO:root: +[Valid][20][UTT][ MSE = 0.2396 | LCC = 0.9529 | SRCC = 0.9543 ] +INFO:root:Average loss=0.09477328971028327 +INFO:root: +[Valid][21][UTT][ MSE = 0.2297 | LCC = 0.9515 | SRCC = 0.9543 ] +INFO:root:Average loss=0.09711089080572129 +INFO:root: +[Valid][22][UTT][ MSE = 0.2306 | LCC = 0.9541 | SRCC = 0.9550 ] +INFO:root:Average loss=0.09334062704443932 +INFO:root: +[Valid][23][UTT][ MSE = 0.2217 | LCC = 0.9516 | SRCC = 0.9536 ] +INFO:root:Average loss=0.09319785106182098 +INFO:root: +[Valid][24][UTT][ MSE = 0.2435 | LCC = 0.9516 | SRCC = 0.9530 ] +INFO:root:Average loss=0.0907546828687191 +INFO:root: +[Valid][25][UTT][ MSE = 0.2200 | LCC = 0.9498 | SRCC = 0.9534 ] +INFO:root:Average loss=0.08782442998886109 +INFO:root: +[Valid][26][UTT][ MSE = 0.2439 | LCC = 0.9533 | SRCC = 0.9538 ] +INFO:root:Average loss=0.08890553817152977 +INFO:root: +[Valid][27][UTT][ MSE = 0.2029 | LCC = 0.9534 | SRCC = 0.9548 ] +INFO:root:Average loss=0.08517673176527023 +INFO:root: +[Valid][28][UTT][ MSE = 0.2260 | LCC = 0.9523 | SRCC = 0.9548 ] +INFO:root:Average loss=0.0847964029610157 +INFO:root: +[Valid][29][UTT][ MSE = 0.2246 | LCC = 0.9514 | SRCC = 0.9543 ] +INFO:root:Average loss=0.08415468809008599 +INFO:root: +[Valid][30][UTT][ MSE = 0.2321 | LCC = 0.9543 | SRCC = 0.9546 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2310 | LCC = 0.9571 | SRCC = 0.9582 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f6e48c54d97da5cf1a908a721cb4b82d790234a7 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:524243a3bfba67fe820a3821c0ab8feec2625dddf1a5b923c984b5b676be892f +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d1c13848036288ce534558dbcaa061e5b755eee5 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4c81b8c3390b16f6916c5c7415b897596f1c36c8569bd3aab09e70ce7abca528 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1ffc6663b3ebeccd138b204de144a373296da841 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.24486669754981993 +INFO:root: +[Valid][0][UTT][ MSE = 0.3287 | LCC = 0.9429 | SRCC = 0.9435 ] +INFO:root:Average loss=0.1564360436797142 +INFO:root: +[Valid][1][UTT][ MSE = 0.2948 | LCC = 0.9459 | SRCC = 0.9461 ] +INFO:root:Average loss=0.14884252482652663 +INFO:root: +[Valid][2][UTT][ MSE = 0.2934 | LCC = 0.9466 | SRCC = 0.9477 ] +INFO:root:Average loss=0.13851956444978714 +INFO:root: +[Valid][3][UTT][ MSE = 0.2683 | LCC = 0.9455 | SRCC = 0.9492 ] +INFO:root:Average loss=0.13867349046468735 +INFO:root: +[Valid][4][UTT][ MSE = 0.2531 | LCC = 0.9500 | SRCC = 0.9509 ] +INFO:root:Average loss=0.1327685831785202 +INFO:root: +[Valid][5][UTT][ MSE = 0.2689 | LCC = 0.9498 | SRCC = 0.9518 ] +INFO:root:Average loss=0.1265624817609787 +INFO:root: +[Valid][6][UTT][ MSE = 0.2463 | LCC = 0.9500 | SRCC = 0.9532 ] +INFO:root:Average loss=0.1272378454208374 +INFO:root: +[Valid][7][UTT][ MSE = 0.2917 | LCC = 0.9504 | SRCC = 0.9518 ] +INFO:root:Average loss=0.12293823176622391 +INFO:root: +[Valid][8][UTT][ MSE = 0.2453 | LCC = 0.9466 | SRCC = 0.9533 ] +INFO:root:Average loss=0.11929519414901733 +INFO:root: +[Valid][9][UTT][ MSE = 0.2473 | LCC = 0.9512 | SRCC = 0.9543 ] +INFO:root:Average loss=0.11941675910353661 +INFO:root: +[Valid][10][UTT][ MSE = 0.3026 | LCC = 0.9517 | SRCC = 0.9533 ] +INFO:root:Average loss=0.11277472853660583 +INFO:root: +[Valid][11][UTT][ MSE = 0.2764 | LCC = 0.9532 | SRCC = 0.9539 ] +INFO:root:Average loss=0.11344250148534775 +INFO:root: +[Valid][12][UTT][ MSE = 0.3226 | LCC = 0.9519 | SRCC = 0.9537 ] +INFO:root:Average loss=0.11281519854068756 +INFO:root: +[Valid][13][UTT][ MSE = 0.2832 | LCC = 0.9494 | SRCC = 0.9534 ] +INFO:root:Average loss=0.11079167822003365 +INFO:root: +[Valid][14][UTT][ MSE = 0.2605 | LCC = 0.9536 | SRCC = 0.9542 ] +INFO:root:Average loss=0.10708553010225295 +INFO:root: +[Valid][15][UTT][ MSE = 0.2578 | LCC = 0.9526 | SRCC = 0.9543 ] +INFO:root:Average loss=0.10375850728154182 +INFO:root: +[Valid][16][UTT][ MSE = 0.3054 | LCC = 0.9526 | SRCC = 0.9539 ] +INFO:root:Average loss=0.10255330428481102 +INFO:root: +[Valid][17][UTT][ MSE = 0.2694 | LCC = 0.9506 | SRCC = 0.9535 ] +INFO:root:Average loss=0.10159090802073478 +INFO:root: +[Valid][18][UTT][ MSE = 0.2488 | LCC = 0.9509 | SRCC = 0.9543 ] +INFO:root:Average loss=0.10189673632383346 +INFO:root: +[Valid][19][UTT][ MSE = 0.2549 | LCC = 0.9526 | SRCC = 0.9548 ] +INFO:root:Average loss=0.09782864505052566 +INFO:root: +[Valid][20][UTT][ MSE = 0.2501 | LCC = 0.9532 | SRCC = 0.9536 ] +INFO:root:Average loss=0.09930760139226913 +INFO:root: +[Valid][21][UTT][ MSE = 0.2205 | LCC = 0.9522 | SRCC = 0.9541 ] +INFO:root:Average loss=0.09422397845983506 +INFO:root: +[Valid][22][UTT][ MSE = 0.2552 | LCC = 0.9520 | SRCC = 0.9530 ] +INFO:root:Average loss=0.09322182646393776 +INFO:root: +[Valid][23][UTT][ MSE = 0.2455 | LCC = 0.9522 | SRCC = 0.9523 ] +INFO:root:Average loss=0.09255443462729454 +INFO:root: +[Valid][24][UTT][ MSE = 0.2755 | LCC = 0.9515 | SRCC = 0.9526 ] +INFO:root:Average loss=0.09207073467969894 +INFO:root: +[Valid][25][UTT][ MSE = 0.2285 | LCC = 0.9522 | SRCC = 0.9550 ] +INFO:root:Average loss=0.08958420386910439 +INFO:root: +[Valid][26][UTT][ MSE = 0.2566 | LCC = 0.9527 | SRCC = 0.9537 ] +INFO:root:Average loss=0.08868339020013809 +INFO:root: +[Valid][27][UTT][ MSE = 0.2595 | LCC = 0.9508 | SRCC = 0.9533 ] +INFO:root:Average loss=0.08524419930577277 +INFO:root: +[Valid][28][UTT][ MSE = 0.2592 | LCC = 0.9523 | SRCC = 0.9538 ] +INFO:root:Average loss=0.08550067684054374 +INFO:root: +[Valid][29][UTT][ MSE = 0.2540 | LCC = 0.9514 | SRCC = 0.9539 ] +INFO:root:Average loss=0.08482162287831306 +INFO:root: +[Valid][30][UTT][ MSE = 0.2437 | LCC = 0.9499 | SRCC = 0.9534 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2638 | LCC = 0.9573 | SRCC = 0.9579 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..45cfc0ac7086b962dff4222ac7a07189edf272ff --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:33d77125d24a4efec57fa47e8cbd08db7561ec786987c150b3692549bcd0075f +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..1bf4bad8277ec1de4a7cb76b7448d1f95ee5a2f6 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b3ab1aed20cdb729bd26935aab93467cfb39b8e53f804518ce45b8806ba6bd08 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3d034205f446abc5d15c7c54a81ccc48e8bef0be --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer10/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3080876564383507 +INFO:root: +[Valid][0][UTT][ MSE = 0.3530 | LCC = 0.9415 | SRCC = 0.9452 ] +INFO:root:Average loss=0.15528575211763382 +INFO:root: +[Valid][1][UTT][ MSE = 0.3196 | LCC = 0.9462 | SRCC = 0.9466 ] +INFO:root:Average loss=0.15049249702692033 +INFO:root: +[Valid][2][UTT][ MSE = 0.3384 | LCC = 0.9469 | SRCC = 0.9491 ] +INFO:root:Average loss=0.14122160971164704 +INFO:root: +[Valid][3][UTT][ MSE = 0.3194 | LCC = 0.9491 | SRCC = 0.9520 ] +INFO:root:Average loss=0.13677442997694014 +INFO:root: +[Valid][4][UTT][ MSE = 0.2865 | LCC = 0.9490 | SRCC = 0.9509 ] +INFO:root:Average loss=0.13104517054557802 +INFO:root: +[Valid][5][UTT][ MSE = 0.3082 | LCC = 0.9520 | SRCC = 0.9528 ] +INFO:root:Average loss=0.13092611759901046 +INFO:root: +[Valid][6][UTT][ MSE = 0.2928 | LCC = 0.9506 | SRCC = 0.9531 ] +INFO:root:Average loss=0.12387714004516602 +INFO:root: +[Valid][7][UTT][ MSE = 0.2892 | LCC = 0.9510 | SRCC = 0.9530 ] +INFO:root:Average loss=0.12087551659345627 +INFO:root: +[Valid][8][UTT][ MSE = 0.2908 | LCC = 0.9526 | SRCC = 0.9536 ] +INFO:root:Average loss=0.12015684551000595 +INFO:root: +[Valid][9][UTT][ MSE = 0.3449 | LCC = 0.9525 | SRCC = 0.9535 ] +INFO:root:Average loss=0.11915102541446686 +INFO:root: +[Valid][10][UTT][ MSE = 0.2982 | LCC = 0.9515 | SRCC = 0.9540 ] +INFO:root:Average loss=0.113944190710783 +INFO:root: +[Valid][11][UTT][ MSE = 0.2823 | LCC = 0.9515 | SRCC = 0.9540 ] +INFO:root:Average loss=0.1152996262907982 +INFO:root: +[Valid][12][UTT][ MSE = 0.2930 | LCC = 0.9524 | SRCC = 0.9545 ] +INFO:root:Average loss=0.11091017675399781 +INFO:root: +[Valid][13][UTT][ MSE = 0.2820 | LCC = 0.9504 | SRCC = 0.9533 ] +INFO:root:Average loss=0.10958422982692718 +INFO:root: +[Valid][14][UTT][ MSE = 0.2608 | LCC = 0.9515 | SRCC = 0.9545 ] +INFO:root:Average loss=0.10395783549547195 +INFO:root: +[Valid][15][UTT][ MSE = 0.2836 | LCC = 0.9494 | SRCC = 0.9522 ] +INFO:root:Average loss=0.10786350664496422 +INFO:root: +[Valid][16][UTT][ MSE = 0.2901 | LCC = 0.9530 | SRCC = 0.9541 ] +INFO:root:Average loss=0.10385571074485779 +INFO:root: +[Valid][17][UTT][ MSE = 0.2602 | LCC = 0.9540 | SRCC = 0.9549 ] +INFO:root:Average loss=0.10013984158635139 +INFO:root: +[Valid][18][UTT][ MSE = 0.2661 | LCC = 0.9513 | SRCC = 0.9550 ] +INFO:root:Average loss=0.09963467341661453 +INFO:root: +[Valid][19][UTT][ MSE = 0.2603 | LCC = 0.9520 | SRCC = 0.9551 ] +INFO:root:Average loss=0.09645124325156212 +INFO:root: +[Valid][20][UTT][ MSE = 0.2456 | LCC = 0.9519 | SRCC = 0.9537 ] +INFO:root:Average loss=0.09695808482170105 +INFO:root: +[Valid][21][UTT][ MSE = 0.2568 | LCC = 0.9535 | SRCC = 0.9550 ] +INFO:root:Average loss=0.09400217798352241 +INFO:root: +[Valid][22][UTT][ MSE = 0.2772 | LCC = 0.9516 | SRCC = 0.9533 ] +INFO:root:Average loss=0.09283907932043076 +INFO:root: +[Valid][23][UTT][ MSE = 0.2928 | LCC = 0.9519 | SRCC = 0.9526 ] +INFO:root:Average loss=0.09254033634066582 +INFO:root: +[Valid][24][UTT][ MSE = 0.2682 | LCC = 0.9530 | SRCC = 0.9552 ] +INFO:root:Average loss=0.09324570623040199 +INFO:root: +[Valid][25][UTT][ MSE = 0.2623 | LCC = 0.9545 | SRCC = 0.9551 ] +INFO:root:Average loss=0.09041057646274567 +INFO:root: +[Valid][26][UTT][ MSE = 0.2606 | LCC = 0.9529 | SRCC = 0.9544 ] +INFO:root:Average loss=0.08623898005485535 +INFO:root: +[Valid][27][UTT][ MSE = 0.2707 | LCC = 0.9523 | SRCC = 0.9538 ] +INFO:root:Average loss=0.08675179713964462 +INFO:root: +[Valid][28][UTT][ MSE = 0.2692 | LCC = 0.9517 | SRCC = 0.9534 ] +INFO:root:Average loss=0.08271222984790802 +INFO:root: +[Valid][29][UTT][ MSE = 0.2673 | LCC = 0.9511 | SRCC = 0.9544 ] +INFO:root:Average loss=0.08377949637174606 +INFO:root: +[Valid][30][UTT][ MSE = 0.2573 | LCC = 0.9526 | SRCC = 0.9555 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2700 | LCC = 0.9571 | SRCC = 0.9578 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a4fc0b56795e5678de960bf2fb3ea2307fec5213 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0154fe7c4c4269fdfef9a1dff2e2ab69fc308a66acd5a1ea24f670d754d736de +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..221a26faec0dff7e1ed676dd569091272ccd7374 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e50638530112c2a9a443dd8b29357c19ae18192ec993d9d71f64c48a516ad13 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..eb193d54e1e21d1cfeff09bdbe477822ccb922ec --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.2814579069018364 +INFO:root: +[Valid][0][UTT][ MSE = 0.3319 | LCC = 0.9399 | SRCC = 0.9407 ] +INFO:root:Average loss=0.1681163548231125 +INFO:root: +[Valid][1][UTT][ MSE = 0.2811 | LCC = 0.9428 | SRCC = 0.9436 ] +INFO:root:Average loss=0.1587605374455452 +INFO:root: +[Valid][2][UTT][ MSE = 0.3296 | LCC = 0.9413 | SRCC = 0.9435 ] +INFO:root:Average loss=0.15672245758771897 +INFO:root: +[Valid][3][UTT][ MSE = 0.2907 | LCC = 0.9443 | SRCC = 0.9464 ] +INFO:root:Average loss=0.14909898579120637 +INFO:root: +[Valid][4][UTT][ MSE = 0.2893 | LCC = 0.9461 | SRCC = 0.9478 ] +INFO:root:Average loss=0.14467973446846008 +INFO:root: +[Valid][5][UTT][ MSE = 0.2857 | LCC = 0.9427 | SRCC = 0.9484 ] +INFO:root:Average loss=0.14321539884805679 +INFO:root: +[Valid][6][UTT][ MSE = 0.2876 | LCC = 0.9461 | SRCC = 0.9481 ] +INFO:root:Average loss=0.14112593787908553 +INFO:root: +[Valid][7][UTT][ MSE = 0.2800 | LCC = 0.9476 | SRCC = 0.9491 ] +INFO:root:Average loss=0.1387976431250572 +INFO:root: +[Valid][8][UTT][ MSE = 0.3049 | LCC = 0.9498 | SRCC = 0.9496 ] +INFO:root:Average loss=0.13608652496337892 +INFO:root: +[Valid][9][UTT][ MSE = 0.2798 | LCC = 0.9478 | SRCC = 0.9499 ] +INFO:root:Average loss=0.13494518673419953 +INFO:root: +[Valid][10][UTT][ MSE = 0.2608 | LCC = 0.9502 | SRCC = 0.9508 ] +INFO:root:Average loss=0.13293912762403487 +INFO:root: +[Valid][11][UTT][ MSE = 0.2675 | LCC = 0.9494 | SRCC = 0.9509 ] +INFO:root:Average loss=0.13143137091398238 +INFO:root: +[Valid][12][UTT][ MSE = 0.2892 | LCC = 0.9481 | SRCC = 0.9517 ] +INFO:root:Average loss=0.13007173389196397 +INFO:root: +[Valid][13][UTT][ MSE = 0.2436 | LCC = 0.9490 | SRCC = 0.9512 ] +INFO:root:Average loss=0.12983617621660232 +INFO:root: +[Valid][14][UTT][ MSE = 0.2980 | LCC = 0.9526 | SRCC = 0.9528 ] +INFO:root:Average loss=0.12755935254693032 +INFO:root: +[Valid][15][UTT][ MSE = 0.2354 | LCC = 0.9519 | SRCC = 0.9526 ] +INFO:root:Average loss=0.12442345941066742 +INFO:root: +[Valid][16][UTT][ MSE = 0.2683 | LCC = 0.9527 | SRCC = 0.9527 ] +INFO:root:Average loss=0.1226974201798439 +INFO:root: +[Valid][17][UTT][ MSE = 0.2438 | LCC = 0.9503 | SRCC = 0.9530 ] +INFO:root:Average loss=0.12161212295293808 +INFO:root: +[Valid][18][UTT][ MSE = 0.2403 | LCC = 0.9500 | SRCC = 0.9517 ] +INFO:root:Average loss=0.12279547661542893 +INFO:root: +[Valid][19][UTT][ MSE = 0.2538 | LCC = 0.9517 | SRCC = 0.9528 ] +INFO:root:Average loss=0.12112716269493103 +INFO:root: +[Valid][20][UTT][ MSE = 0.2464 | LCC = 0.9505 | SRCC = 0.9531 ] +INFO:root:Average loss=0.11933672231435775 +INFO:root: +[Valid][21][UTT][ MSE = 0.2589 | LCC = 0.9515 | SRCC = 0.9538 ] +INFO:root:Average loss=0.11738722330331802 +INFO:root: +[Valid][22][UTT][ MSE = 0.2428 | LCC = 0.9502 | SRCC = 0.9513 ] +INFO:root:Average loss=0.11611400640010834 +INFO:root: +[Valid][23][UTT][ MSE = 0.2408 | LCC = 0.9517 | SRCC = 0.9537 ] +INFO:root:Average loss=0.11838986563682556 +INFO:root: +[Valid][24][UTT][ MSE = 0.2630 | LCC = 0.9530 | SRCC = 0.9539 ] +INFO:root:Average loss=0.11163491669297218 +INFO:root: +[Valid][25][UTT][ MSE = 0.2619 | LCC = 0.9538 | SRCC = 0.9540 ] +INFO:root:Average loss=0.11341638958454132 +INFO:root: +[Valid][26][UTT][ MSE = 0.2345 | LCC = 0.9533 | SRCC = 0.9535 ] +INFO:root:Average loss=0.11265718096494674 +INFO:root: +[Valid][27][UTT][ MSE = 0.2385 | LCC = 0.9519 | SRCC = 0.9531 ] +INFO:root:Average loss=0.1161238203048706 +INFO:root: +[Valid][28][UTT][ MSE = 0.2474 | LCC = 0.9518 | SRCC = 0.9529 ] +INFO:root:Average loss=0.10881737300753594 +INFO:root: +[Valid][29][UTT][ MSE = 0.2686 | LCC = 0.9534 | SRCC = 0.9541 ] +INFO:root:Average loss=0.1080647459924221 +INFO:root: +[Valid][30][UTT][ MSE = 0.2264 | LCC = 0.9522 | SRCC = 0.9542 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2667 | LCC = 0.9542 | SRCC = 0.9552 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0523bc44e16cfa01e92d71e015501f20533559d6 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:41f7a384a4a8a3df7c0b93bc0990ccd666c77245e12e3936a6813d7de013218a +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..50904d5fd522dbce3e77bdecd3e4c0d70dfd8734 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d51cac1962be61efe05d9ccbaeaf29c99b199e5e9a45000de40db5980983a270 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6253f2ca67e2d9feed3e3bfeb6e08622e547f47b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.26752584624290465 +INFO:root: +[Valid][0][UTT][ MSE = 0.3694 | LCC = 0.9379 | SRCC = 0.9409 ] +INFO:root:Average loss=0.1714120327234268 +INFO:root: +[Valid][1][UTT][ MSE = 0.2982 | LCC = 0.9406 | SRCC = 0.9427 ] +INFO:root:Average loss=0.1582636525630951 +INFO:root: +[Valid][2][UTT][ MSE = 0.3295 | LCC = 0.9398 | SRCC = 0.9445 ] +INFO:root:Average loss=0.15711607474088668 +INFO:root: +[Valid][3][UTT][ MSE = 0.3360 | LCC = 0.9440 | SRCC = 0.9452 ] +INFO:root:Average loss=0.14941147410869599 +INFO:root: +[Valid][4][UTT][ MSE = 0.3415 | LCC = 0.9470 | SRCC = 0.9474 ] +INFO:root:Average loss=0.14750538003444671 +INFO:root: +[Valid][5][UTT][ MSE = 0.3390 | LCC = 0.9464 | SRCC = 0.9478 ] +INFO:root:Average loss=0.14630106765031814 +INFO:root: +[Valid][6][UTT][ MSE = 0.3142 | LCC = 0.9470 | SRCC = 0.9477 ] +INFO:root:Average loss=0.1440025275349617 +INFO:root: +[Valid][7][UTT][ MSE = 0.3347 | LCC = 0.9504 | SRCC = 0.9501 ] +INFO:root:Average loss=0.14344898641109466 +INFO:root: +[Valid][8][UTT][ MSE = 0.3095 | LCC = 0.9499 | SRCC = 0.9494 ] +INFO:root:Average loss=0.13908492386341095 +INFO:root: +[Valid][9][UTT][ MSE = 0.3390 | LCC = 0.9492 | SRCC = 0.9496 ] +INFO:root:Average loss=0.13530862963199616 +INFO:root: +[Valid][10][UTT][ MSE = 0.3035 | LCC = 0.9515 | SRCC = 0.9512 ] +INFO:root:Average loss=0.13540555948019029 +INFO:root: +[Valid][11][UTT][ MSE = 0.3034 | LCC = 0.9507 | SRCC = 0.9514 ] +INFO:root:Average loss=0.13611926335096358 +INFO:root: +[Valid][12][UTT][ MSE = 0.2849 | LCC = 0.9491 | SRCC = 0.9495 ] +INFO:root:Average loss=0.1294824341535568 +INFO:root: +[Valid][13][UTT][ MSE = 0.3018 | LCC = 0.9511 | SRCC = 0.9510 ] +INFO:root:Average loss=0.12979632291197776 +INFO:root: +[Valid][14][UTT][ MSE = 0.2751 | LCC = 0.9500 | SRCC = 0.9500 ] +INFO:root:Average loss=0.12813393047451974 +INFO:root: +[Valid][15][UTT][ MSE = 0.3039 | LCC = 0.9514 | SRCC = 0.9522 ] +INFO:root:Average loss=0.1261380226612091 +INFO:root: +[Valid][16][UTT][ MSE = 0.2733 | LCC = 0.9516 | SRCC = 0.9527 ] +INFO:root:Average loss=0.12329052561521531 +INFO:root: +[Valid][17][UTT][ MSE = 0.2713 | LCC = 0.9513 | SRCC = 0.9521 ] +INFO:root:Average loss=0.12180546694993973 +INFO:root: +[Valid][18][UTT][ MSE = 0.2890 | LCC = 0.9526 | SRCC = 0.9524 ] +INFO:root:Average loss=0.12136321759223938 +INFO:root: +[Valid][19][UTT][ MSE = 0.3175 | LCC = 0.9484 | SRCC = 0.9492 ] +INFO:root:Average loss=0.12028357744216919 +INFO:root: +[Valid][20][UTT][ MSE = 0.2688 | LCC = 0.9516 | SRCC = 0.9523 ] +INFO:root:Average loss=0.1187065219283104 +INFO:root: +[Valid][21][UTT][ MSE = 0.2904 | LCC = 0.9509 | SRCC = 0.9517 ] +INFO:root:Average loss=0.11998590114712715 +INFO:root: +[Valid][22][UTT][ MSE = 0.2606 | LCC = 0.9522 | SRCC = 0.9533 ] +INFO:root:Average loss=0.11649193394184112 +INFO:root: +[Valid][23][UTT][ MSE = 0.3010 | LCC = 0.9524 | SRCC = 0.9532 ] +INFO:root:Average loss=0.11456608593463898 +INFO:root: +[Valid][24][UTT][ MSE = 0.2952 | LCC = 0.9532 | SRCC = 0.9540 ] +INFO:root:Average loss=0.11226453202962876 +INFO:root: +[Valid][25][UTT][ MSE = 0.2818 | LCC = 0.9530 | SRCC = 0.9531 ] +INFO:root:Average loss=0.11334878280758857 +INFO:root: +[Valid][26][UTT][ MSE = 0.2599 | LCC = 0.9492 | SRCC = 0.9515 ] +INFO:root:Average loss=0.11547447800636292 +INFO:root: +[Valid][27][UTT][ MSE = 0.2651 | LCC = 0.9527 | SRCC = 0.9531 ] +INFO:root:Average loss=0.11301700872182846 +INFO:root: +[Valid][28][UTT][ MSE = 0.2592 | LCC = 0.9546 | SRCC = 0.9545 ] +INFO:root:Average loss=0.11093720984458923 +INFO:root: +[Valid][29][UTT][ MSE = 0.2673 | LCC = 0.9490 | SRCC = 0.9511 ] +INFO:root:Average loss=0.10909995344281197 +INFO:root: +[Valid][30][UTT][ MSE = 0.2595 | LCC = 0.9518 | SRCC = 0.9537 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2628 | LCC = 0.9555 | SRCC = 0.9564 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a3ff24998a14213d635830bf3c2232e34b9a4143 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cf0b7aa2050c0ab04e3f9e1d179c3380cc72e1ff380f2a4be030ddddf0b14958 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b4d665fc40c487a7a23e231adac1a43dc486e40a --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:72d718629cbaa3c8406f740c55be65acc56ca7f6cd037ff94eb1faca6f035bce +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9fb281ce5e2a13aba9cfcc035630220cc215c767 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.2810729897022247 +INFO:root: +[Valid][0][UTT][ MSE = 0.3980 | LCC = 0.9381 | SRCC = 0.9397 ] +INFO:root:Average loss=0.16840830880403518 +INFO:root: +[Valid][1][UTT][ MSE = 0.3811 | LCC = 0.9419 | SRCC = 0.9429 ] +INFO:root:Average loss=0.15785593211650847 +INFO:root: +[Valid][2][UTT][ MSE = 0.3552 | LCC = 0.9449 | SRCC = 0.9451 ] +INFO:root:Average loss=0.152396779358387 +INFO:root: +[Valid][3][UTT][ MSE = 0.2871 | LCC = 0.9465 | SRCC = 0.9475 ] +INFO:root:Average loss=0.14861381489038467 +INFO:root: +[Valid][4][UTT][ MSE = 0.3657 | LCC = 0.9485 | SRCC = 0.9486 ] +INFO:root:Average loss=0.14705237901210785 +INFO:root: +[Valid][5][UTT][ MSE = 0.3503 | LCC = 0.9484 | SRCC = 0.9485 ] +INFO:root:Average loss=0.14486832880973816 +INFO:root: +[Valid][6][UTT][ MSE = 0.3294 | LCC = 0.9474 | SRCC = 0.9496 ] +INFO:root:Average loss=0.1399860918521881 +INFO:root: +[Valid][7][UTT][ MSE = 0.3453 | LCC = 0.9488 | SRCC = 0.9492 ] +INFO:root:Average loss=0.13733734411001206 +INFO:root: +[Valid][8][UTT][ MSE = 0.3123 | LCC = 0.9495 | SRCC = 0.9505 ] +INFO:root:Average loss=0.1371632850766182 +INFO:root: +[Valid][9][UTT][ MSE = 0.3538 | LCC = 0.9510 | SRCC = 0.9507 ] +INFO:root:Average loss=0.13350439816713333 +INFO:root: +[Valid][10][UTT][ MSE = 0.3118 | LCC = 0.9497 | SRCC = 0.9512 ] +INFO:root:Average loss=0.13415789794921876 +INFO:root: +[Valid][11][UTT][ MSE = 0.2804 | LCC = 0.9491 | SRCC = 0.9506 ] +INFO:root:Average loss=0.13312641334533692 +INFO:root: +[Valid][12][UTT][ MSE = 0.2897 | LCC = 0.9523 | SRCC = 0.9523 ] +INFO:root:Average loss=0.13308629369735717 +INFO:root: +[Valid][13][UTT][ MSE = 0.2707 | LCC = 0.9516 | SRCC = 0.9523 ] +INFO:root:Average loss=0.1298388592004776 +INFO:root: +[Valid][14][UTT][ MSE = 0.2854 | LCC = 0.9507 | SRCC = 0.9524 ] +INFO:root:Average loss=0.12757051622867585 +INFO:root: +[Valid][15][UTT][ MSE = 0.3029 | LCC = 0.9535 | SRCC = 0.9535 ] +INFO:root:Average loss=0.12301386564970017 +INFO:root: +[Valid][16][UTT][ MSE = 0.2865 | LCC = 0.9533 | SRCC = 0.9539 ] +INFO:root:Average loss=0.12534945210814477 +INFO:root: +[Valid][17][UTT][ MSE = 0.3262 | LCC = 0.9464 | SRCC = 0.9482 ] +INFO:root:Average loss=0.12152049434185028 +INFO:root: +[Valid][18][UTT][ MSE = 0.2784 | LCC = 0.9523 | SRCC = 0.9530 ] +INFO:root:Average loss=0.12336529141664505 +INFO:root: +[Valid][19][UTT][ MSE = 0.3394 | LCC = 0.9515 | SRCC = 0.9539 ] +INFO:root:Average loss=0.12293041896820069 +INFO:root: +[Valid][20][UTT][ MSE = 0.2389 | LCC = 0.9525 | SRCC = 0.9538 ] +INFO:root:Average loss=0.1177524955868721 +INFO:root: +[Valid][21][UTT][ MSE = 0.2803 | LCC = 0.9530 | SRCC = 0.9534 ] +INFO:root:Average loss=0.1181523651778698 +INFO:root: +[Valid][22][UTT][ MSE = 0.2715 | LCC = 0.9540 | SRCC = 0.9546 ] +INFO:root:Average loss=0.11745476230978966 +INFO:root: +[Valid][23][UTT][ MSE = 0.2987 | LCC = 0.9530 | SRCC = 0.9545 ] +INFO:root:Average loss=0.1142566346526146 +INFO:root: +[Valid][24][UTT][ MSE = 0.2745 | LCC = 0.9520 | SRCC = 0.9542 ] +INFO:root:Average loss=0.11410136380791665 +INFO:root: +[Valid][25][UTT][ MSE = 0.2480 | LCC = 0.9537 | SRCC = 0.9543 ] +INFO:root:Average loss=0.11344864404201507 +INFO:root: +[Valid][26][UTT][ MSE = 0.2696 | LCC = 0.9546 | SRCC = 0.9549 ] +INFO:root:Average loss=0.11257067209482192 +INFO:root: +[Valid][27][UTT][ MSE = 0.2968 | LCC = 0.9535 | SRCC = 0.9543 ] +INFO:root:Average loss=0.11191925364732742 +INFO:root: +[Valid][28][UTT][ MSE = 0.2635 | LCC = 0.9518 | SRCC = 0.9535 ] +INFO:root:Average loss=0.11590092623233796 +INFO:root: +[Valid][29][UTT][ MSE = 0.2758 | LCC = 0.9532 | SRCC = 0.9543 ] +INFO:root:Average loss=0.10834644377231598 +INFO:root: +[Valid][30][UTT][ MSE = 0.2737 | LCC = 0.9543 | SRCC = 0.9547 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2673 | LCC = 0.9564 | SRCC = 0.9571 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..21dc06a5928077ca8c391c3decd315ef84807b67 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:95923f2c86f406df8150450bcc8a045e8dac71cc301b6ce33335b2a805b786fa +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..07771f6b2d75da56f97c71408a28c84307f0d045 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7f74963943f1332725a822193e37474a2bb422bb70b39039af48c50c2aee7c13 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..82f0030acc78e7888c06756de7648c49cfa2f71e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.2491286446452141 +INFO:root: +[Valid][0][UTT][ MSE = 0.3676 | LCC = 0.9398 | SRCC = 0.9394 ] +INFO:root:Average loss=0.16993608343601227 +INFO:root: +[Valid][1][UTT][ MSE = 0.3843 | LCC = 0.9390 | SRCC = 0.9420 ] +INFO:root:Average loss=0.15856459772586823 +INFO:root: +[Valid][2][UTT][ MSE = 0.3190 | LCC = 0.9436 | SRCC = 0.9456 ] +INFO:root:Average loss=0.15166519194841385 +INFO:root: +[Valid][3][UTT][ MSE = 0.3316 | LCC = 0.9464 | SRCC = 0.9473 ] +INFO:root:Average loss=0.15046386659145355 +INFO:root: +[Valid][4][UTT][ MSE = 0.3467 | LCC = 0.9422 | SRCC = 0.9448 ] +INFO:root:Average loss=0.14578987342119218 +INFO:root: +[Valid][5][UTT][ MSE = 0.2914 | LCC = 0.9473 | SRCC = 0.9470 ] +INFO:root:Average loss=0.14307096749544143 +INFO:root: +[Valid][6][UTT][ MSE = 0.2823 | LCC = 0.9474 | SRCC = 0.9494 ] +INFO:root:Average loss=0.1427698029279709 +INFO:root: +[Valid][7][UTT][ MSE = 0.2835 | LCC = 0.9498 | SRCC = 0.9500 ] +INFO:root:Average loss=0.13771494495868683 +INFO:root: +[Valid][8][UTT][ MSE = 0.2521 | LCC = 0.9493 | SRCC = 0.9503 ] +INFO:root:Average loss=0.13901798686385156 +INFO:root: +[Valid][9][UTT][ MSE = 0.2659 | LCC = 0.9480 | SRCC = 0.9497 ] +INFO:root:Average loss=0.13444769942760468 +INFO:root: +[Valid][10][UTT][ MSE = 0.2933 | LCC = 0.9486 | SRCC = 0.9489 ] +INFO:root:Average loss=0.13447156268358232 +INFO:root: +[Valid][11][UTT][ MSE = 0.3025 | LCC = 0.9494 | SRCC = 0.9504 ] +INFO:root:Average loss=0.1292623587846756 +INFO:root: +[Valid][12][UTT][ MSE = 0.2955 | LCC = 0.9501 | SRCC = 0.9506 ] +INFO:root:Average loss=0.12814094746112822 +INFO:root: +[Valid][13][UTT][ MSE = 0.2779 | LCC = 0.9508 | SRCC = 0.9516 ] +INFO:root:Average loss=0.12741958248615265 +INFO:root: +[Valid][14][UTT][ MSE = 0.2565 | LCC = 0.9509 | SRCC = 0.9516 ] +INFO:root:Average loss=0.12686253565549852 +INFO:root: +[Valid][15][UTT][ MSE = 0.2741 | LCC = 0.9511 | SRCC = 0.9514 ] +INFO:root:Average loss=0.12480669343471527 +INFO:root: +[Valid][16][UTT][ MSE = 0.2860 | LCC = 0.9510 | SRCC = 0.9520 ] +INFO:root:Average loss=0.12438437730073928 +INFO:root: +[Valid][17][UTT][ MSE = 0.2716 | LCC = 0.9518 | SRCC = 0.9522 ] +INFO:root:Average loss=0.12209893009066582 +INFO:root: +[Valid][18][UTT][ MSE = 0.2644 | LCC = 0.9526 | SRCC = 0.9528 ] +INFO:root:Average loss=0.12122797948122024 +INFO:root: +[Valid][19][UTT][ MSE = 0.2888 | LCC = 0.9504 | SRCC = 0.9503 ] +INFO:root:Average loss=0.12109525394439698 +INFO:root: +[Valid][20][UTT][ MSE = 0.2838 | LCC = 0.9507 | SRCC = 0.9505 ] +INFO:root:Average loss=0.11912874621152877 +INFO:root: +[Valid][21][UTT][ MSE = 0.2716 | LCC = 0.9534 | SRCC = 0.9533 ] +INFO:root:Average loss=0.12002099484205246 +INFO:root: +[Valid][22][UTT][ MSE = 0.3070 | LCC = 0.9492 | SRCC = 0.9524 ] +INFO:root:Average loss=0.11758092260360718 +INFO:root: +[Valid][23][UTT][ MSE = 0.2635 | LCC = 0.9535 | SRCC = 0.9538 ] +INFO:root:Average loss=0.11564086580276489 +INFO:root: +[Valid][24][UTT][ MSE = 0.3040 | LCC = 0.9515 | SRCC = 0.9527 ] +INFO:root:Average loss=0.1152178338766098 +INFO:root: +[Valid][25][UTT][ MSE = 0.2731 | LCC = 0.9529 | SRCC = 0.9536 ] +INFO:root:Average loss=0.11391247186064721 +INFO:root: +[Valid][26][UTT][ MSE = 0.2275 | LCC = 0.9526 | SRCC = 0.9532 ] +INFO:root:Average loss=0.11633453410863877 +INFO:root: +[Valid][27][UTT][ MSE = 0.2600 | LCC = 0.9524 | SRCC = 0.9531 ] +INFO:root:Average loss=0.11343312075734138 +INFO:root: +[Valid][28][UTT][ MSE = 0.2582 | LCC = 0.9486 | SRCC = 0.9516 ] +INFO:root:Average loss=0.11281874525547028 +INFO:root: +[Valid][29][UTT][ MSE = 0.2464 | LCC = 0.9526 | SRCC = 0.9534 ] +INFO:root:Average loss=0.11113272628188134 +INFO:root: +[Valid][30][UTT][ MSE = 0.2652 | LCC = 0.9515 | SRCC = 0.9519 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2660 | LCC = 0.9539 | SRCC = 0.9549 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f18932b4fb9513a724d8e81f2f5c98ee548a7958 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:08979d6d1a8a63685b49e48018cddd3823e1f58fdb2be09d73d426e2e984309b +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..57342beb1ff9bdf45d02a2d3bb43028502503d09 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:556eb892af811d5f9a5411c33009fccebb6bee3958d5ad76f010503a8a5902ae +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..23a04e055d8fb784fe89619237f2f5008db9fee1 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer11/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.294693432033062 +INFO:root: +[Valid][0][UTT][ MSE = 0.3416 | LCC = 0.9398 | SRCC = 0.9399 ] +INFO:root:Average loss=0.16454577815532684 +INFO:root: +[Valid][1][UTT][ MSE = 0.3162 | LCC = 0.9445 | SRCC = 0.9446 ] +INFO:root:Average loss=0.15942982065677644 +INFO:root: +[Valid][2][UTT][ MSE = 0.3079 | LCC = 0.9448 | SRCC = 0.9445 ] +INFO:root:Average loss=0.15693715012073517 +INFO:root: +[Valid][3][UTT][ MSE = 0.3588 | LCC = 0.9377 | SRCC = 0.9457 ] +INFO:root:Average loss=0.15180404329299926 +INFO:root: +[Valid][4][UTT][ MSE = 0.3363 | LCC = 0.9419 | SRCC = 0.9447 ] +INFO:root:Average loss=0.14878379338979722 +INFO:root: +[Valid][5][UTT][ MSE = 0.2659 | LCC = 0.9484 | SRCC = 0.9484 ] +INFO:root:Average loss=0.14256107634305953 +INFO:root: +[Valid][6][UTT][ MSE = 0.3013 | LCC = 0.9450 | SRCC = 0.9481 ] +INFO:root:Average loss=0.14149716252088546 +INFO:root: +[Valid][7][UTT][ MSE = 0.3279 | LCC = 0.9459 | SRCC = 0.9472 ] +INFO:root:Average loss=0.13851482653617858 +INFO:root: +[Valid][8][UTT][ MSE = 0.3037 | LCC = 0.9445 | SRCC = 0.9488 ] +INFO:root:Average loss=0.13857754611968995 +INFO:root: +[Valid][9][UTT][ MSE = 0.2901 | LCC = 0.9502 | SRCC = 0.9506 ] +INFO:root:Average loss=0.1335086454153061 +INFO:root: +[Valid][10][UTT][ MSE = 0.2749 | LCC = 0.9495 | SRCC = 0.9505 ] +INFO:root:Average loss=0.13379533085227013 +INFO:root: +[Valid][11][UTT][ MSE = 0.3019 | LCC = 0.9491 | SRCC = 0.9509 ] +INFO:root:Average loss=0.1337599179148674 +INFO:root: +[Valid][12][UTT][ MSE = 0.3077 | LCC = 0.9500 | SRCC = 0.9506 ] +INFO:root:Average loss=0.13220434832572936 +INFO:root: +[Valid][13][UTT][ MSE = 0.2993 | LCC = 0.9501 | SRCC = 0.9512 ] +INFO:root:Average loss=0.129887038230896 +INFO:root: +[Valid][14][UTT][ MSE = 0.2765 | LCC = 0.9509 | SRCC = 0.9513 ] +INFO:root:Average loss=0.12550398874282837 +INFO:root: +[Valid][15][UTT][ MSE = 0.3124 | LCC = 0.9455 | SRCC = 0.9508 ] +INFO:root:Average loss=0.1267671024799347 +INFO:root: +[Valid][16][UTT][ MSE = 0.2899 | LCC = 0.9508 | SRCC = 0.9514 ] +INFO:root:Average loss=0.12446775436401367 +INFO:root: +[Valid][17][UTT][ MSE = 0.2968 | LCC = 0.9487 | SRCC = 0.9515 ] +INFO:root:Average loss=0.12256696653366089 +INFO:root: +[Valid][18][UTT][ MSE = 0.2601 | LCC = 0.9504 | SRCC = 0.9518 ] +INFO:root:Average loss=0.12226933419704437 +INFO:root: +[Valid][19][UTT][ MSE = 0.2893 | LCC = 0.9483 | SRCC = 0.9517 ] +INFO:root:Average loss=0.11927892526984214 +INFO:root: +[Valid][20][UTT][ MSE = 0.2650 | LCC = 0.9522 | SRCC = 0.9530 ] +INFO:root:Average loss=0.12019276398420334 +INFO:root: +[Valid][21][UTT][ MSE = 0.2995 | LCC = 0.9508 | SRCC = 0.9513 ] +INFO:root:Average loss=0.11862098675966262 +INFO:root: +[Valid][22][UTT][ MSE = 0.2744 | LCC = 0.9526 | SRCC = 0.9534 ] +INFO:root:Average loss=0.11847129878401756 +INFO:root: +[Valid][23][UTT][ MSE = 0.2668 | LCC = 0.9489 | SRCC = 0.9517 ] +INFO:root:Average loss=0.11586503410339355 +INFO:root: +[Valid][24][UTT][ MSE = 0.2721 | LCC = 0.9510 | SRCC = 0.9535 ] +INFO:root:Average loss=0.11564524805545807 +INFO:root: +[Valid][25][UTT][ MSE = 0.2415 | LCC = 0.9535 | SRCC = 0.9536 ] +INFO:root:Average loss=0.11104352656006813 +INFO:root: +[Valid][26][UTT][ MSE = 0.2874 | LCC = 0.9499 | SRCC = 0.9525 ] +INFO:root:Average loss=0.1120197041630745 +INFO:root: +[Valid][27][UTT][ MSE = 0.3027 | LCC = 0.9464 | SRCC = 0.9501 ] +INFO:root:Average loss=0.11102058473229408 +INFO:root: +[Valid][28][UTT][ MSE = 0.2608 | LCC = 0.9517 | SRCC = 0.9532 ] +INFO:root:Average loss=0.11328127974271775 +INFO:root: +[Valid][29][UTT][ MSE = 0.2869 | LCC = 0.9525 | SRCC = 0.9532 ] +INFO:root:Average loss=0.1081151987016201 +INFO:root: +[Valid][30][UTT][ MSE = 0.2830 | LCC = 0.9512 | SRCC = 0.9531 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2418 | LCC = 0.9547 | SRCC = 0.9550 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6d78bdd9cfbfc5eaa26f39f20718bebbc87dbf46 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:880a6487bbf4e408bf03a36e1b56e7c2341c3be647cfbd551c9d7b25f788884b +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d749690c82ace8eb1f25e6c79d6129712d4ea1a2 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:76be47bb58a87eb9d4db39d1f2e07ac08020ebc6fb1dcd9b530163849d883e8c +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..178180fbfe71561978d544fab35ccbde5fb18845 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.39776571452617643 +INFO:root: +[Valid][0][UTT][ MSE = 0.2849 | LCC = 0.9272 | SRCC = 0.9305 ] +INFO:root:Average loss=0.1776919149160385 +INFO:root: +[Valid][1][UTT][ MSE = 0.2627 | LCC = 0.9435 | SRCC = 0.9439 ] +INFO:root:Average loss=0.14179367589950562 +INFO:root: +[Valid][2][UTT][ MSE = 0.1959 | LCC = 0.9479 | SRCC = 0.9499 ] +INFO:root:Average loss=0.12560794693231583 +INFO:root: +[Valid][3][UTT][ MSE = 0.2132 | LCC = 0.9522 | SRCC = 0.9527 ] +INFO:root:Average loss=0.11401467913389206 +INFO:root: +[Valid][4][UTT][ MSE = 0.2101 | LCC = 0.9545 | SRCC = 0.9546 ] +INFO:root:Average loss=0.10625395318865775 +INFO:root: +[Valid][5][UTT][ MSE = 0.2004 | LCC = 0.9562 | SRCC = 0.9567 ] +INFO:root:Average loss=0.09361294296383858 +INFO:root: +[Valid][6][UTT][ MSE = 0.1896 | LCC = 0.9564 | SRCC = 0.9570 ] +INFO:root:Average loss=0.090866629332304 +INFO:root: +[Valid][7][UTT][ MSE = 0.1908 | LCC = 0.9575 | SRCC = 0.9584 ] +INFO:root:Average loss=0.08507742893695831 +INFO:root: +[Valid][8][UTT][ MSE = 0.1923 | LCC = 0.9580 | SRCC = 0.9588 ] +INFO:root:Average loss=0.0806419829428196 +INFO:root: +[Valid][9][UTT][ MSE = 0.2178 | LCC = 0.9590 | SRCC = 0.9595 ] +INFO:root:Average loss=0.07358412340283393 +INFO:root: +[Valid][10][UTT][ MSE = 0.1938 | LCC = 0.9570 | SRCC = 0.9575 ] +INFO:root:Average loss=0.06900703671574593 +INFO:root: +[Valid][11][UTT][ MSE = 0.1802 | LCC = 0.9598 | SRCC = 0.9599 ] +INFO:root:Average loss=0.06648158988356591 +INFO:root: +[Valid][12][UTT][ MSE = 0.1810 | LCC = 0.9573 | SRCC = 0.9578 ] +INFO:root:Average loss=0.062286653473973275 +INFO:root: +[Valid][13][UTT][ MSE = 0.1829 | LCC = 0.9577 | SRCC = 0.9577 ] +INFO:root:Average loss=0.059788195997476576 +INFO:root: +[Valid][14][UTT][ MSE = 0.2058 | LCC = 0.9595 | SRCC = 0.9597 ] +INFO:root:Average loss=0.05922142452001572 +INFO:root: +[Valid][15][UTT][ MSE = 0.1951 | LCC = 0.9591 | SRCC = 0.9595 ] +INFO:root:Average loss=0.05288430547714233 +INFO:root: +[Valid][16][UTT][ MSE = 0.1884 | LCC = 0.9598 | SRCC = 0.9601 ] +INFO:root:Average loss=0.05112069357931614 +INFO:root: +[Valid][17][UTT][ MSE = 0.1787 | LCC = 0.9601 | SRCC = 0.9606 ] +INFO:root:Average loss=0.05063433602452278 +INFO:root: +[Valid][18][UTT][ MSE = 0.1872 | LCC = 0.9591 | SRCC = 0.9593 ] +INFO:root:Average loss=0.047983339577913287 +INFO:root: +[Valid][19][UTT][ MSE = 0.1797 | LCC = 0.9585 | SRCC = 0.9583 ] +INFO:root:Average loss=0.047228816881775854 +INFO:root: +[Valid][20][UTT][ MSE = 0.1888 | LCC = 0.9598 | SRCC = 0.9601 ] +INFO:root:Average loss=0.044366488128900525 +INFO:root: +[Valid][21][UTT][ MSE = 0.1766 | LCC = 0.9588 | SRCC = 0.9596 ] +INFO:root:Average loss=0.0429242409914732 +INFO:root: +[Valid][22][UTT][ MSE = 0.2011 | LCC = 0.9594 | SRCC = 0.9593 ] +INFO:root:Average loss=0.04086180020868778 +INFO:root: +[Valid][23][UTT][ MSE = 0.2032 | LCC = 0.9583 | SRCC = 0.9584 ] +INFO:root:Average loss=0.038362254530191424 +INFO:root: +[Valid][24][UTT][ MSE = 0.1828 | LCC = 0.9607 | SRCC = 0.9608 ] +INFO:root:Average loss=0.036279375046491626 +INFO:root: +[Valid][25][UTT][ MSE = 0.1948 | LCC = 0.9600 | SRCC = 0.9605 ] +INFO:root:Average loss=0.03563812455534935 +INFO:root: +[Valid][26][UTT][ MSE = 0.1854 | LCC = 0.9600 | SRCC = 0.9603 ] +INFO:root:Average loss=0.035077432468533516 +INFO:root: +[Valid][27][UTT][ MSE = 0.1722 | LCC = 0.9604 | SRCC = 0.9603 ] +INFO:root:Average loss=0.03151769176125527 +INFO:root: +[Valid][28][UTT][ MSE = 0.1831 | LCC = 0.9593 | SRCC = 0.9595 ] +INFO:root:Average loss=0.03193476229906082 +INFO:root: +[Valid][29][UTT][ MSE = 0.1832 | LCC = 0.9604 | SRCC = 0.9604 ] +INFO:root:Average loss=0.030853326335549356 +INFO:root: +[Valid][30][UTT][ MSE = 0.1784 | LCC = 0.9599 | SRCC = 0.9601 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1778 | LCC = 0.9642 | SRCC = 0.9645 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3893c60f674ef76eb21f70f715405daa5a78f82b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0075eb71c51b1e3cf056271e335c3b34904159925592e863a13cf9dc60a45097 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..8defc03c6d45dbddd1e5a82db552c1e725a14066 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3c8eb3c31d54b8b86ca9c7829852b69159c2f6210af539229ca3423ccb528717 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f9c6e71fa9bf880a075babbceb1e8a129ca11e13 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.43586616206169126 +INFO:root: +[Valid][0][UTT][ MSE = 0.2921 | LCC = 0.9246 | SRCC = 0.9256 ] +INFO:root:Average loss=0.1772779755592346 +INFO:root: +[Valid][1][UTT][ MSE = 0.2644 | LCC = 0.9425 | SRCC = 0.9432 ] +INFO:root:Average loss=0.1456099750995636 +INFO:root: +[Valid][2][UTT][ MSE = 0.2525 | LCC = 0.9480 | SRCC = 0.9489 ] +INFO:root:Average loss=0.13076415741443634 +INFO:root: +[Valid][3][UTT][ MSE = 0.1971 | LCC = 0.9504 | SRCC = 0.9518 ] +INFO:root:Average loss=0.1163666644692421 +INFO:root: +[Valid][4][UTT][ MSE = 0.2397 | LCC = 0.9515 | SRCC = 0.9517 ] +INFO:root:Average loss=0.10612465563416482 +INFO:root: +[Valid][5][UTT][ MSE = 0.2061 | LCC = 0.9539 | SRCC = 0.9543 ] +INFO:root:Average loss=0.0981700849533081 +INFO:root: +[Valid][6][UTT][ MSE = 0.2158 | LCC = 0.9559 | SRCC = 0.9560 ] +INFO:root:Average loss=0.09404602470993996 +INFO:root: +[Valid][7][UTT][ MSE = 0.2289 | LCC = 0.9545 | SRCC = 0.9543 ] +INFO:root:Average loss=0.08625945094227791 +INFO:root: +[Valid][8][UTT][ MSE = 0.1970 | LCC = 0.9543 | SRCC = 0.9541 ] +INFO:root:Average loss=0.07899524611234665 +INFO:root: +[Valid][9][UTT][ MSE = 0.1995 | LCC = 0.9577 | SRCC = 0.9580 ] +INFO:root:Average loss=0.0779505233168602 +INFO:root: +[Valid][10][UTT][ MSE = 0.2249 | LCC = 0.9576 | SRCC = 0.9582 ] +INFO:root:Average loss=0.07282762634754181 +INFO:root: +[Valid][11][UTT][ MSE = 0.2076 | LCC = 0.9595 | SRCC = 0.9597 ] +INFO:root:Average loss=0.06856367626786232 +INFO:root: +[Valid][12][UTT][ MSE = 0.2162 | LCC = 0.9591 | SRCC = 0.9597 ] +INFO:root:Average loss=0.06509211620688439 +INFO:root: +[Valid][13][UTT][ MSE = 0.2014 | LCC = 0.9568 | SRCC = 0.9566 ] +INFO:root:Average loss=0.061082896023988725 +INFO:root: +[Valid][14][UTT][ MSE = 0.2025 | LCC = 0.9600 | SRCC = 0.9605 ] +INFO:root:Average loss=0.058813192635774614 +INFO:root: +[Valid][15][UTT][ MSE = 0.2095 | LCC = 0.9589 | SRCC = 0.9591 ] +INFO:root:Average loss=0.0538287286311388 +INFO:root: +[Valid][16][UTT][ MSE = 0.2085 | LCC = 0.9597 | SRCC = 0.9601 ] +INFO:root:Average loss=0.05311391732096672 +INFO:root: +[Valid][17][UTT][ MSE = 0.1995 | LCC = 0.9577 | SRCC = 0.9578 ] +INFO:root:Average loss=0.05123066186904907 +INFO:root: +[Valid][18][UTT][ MSE = 0.1860 | LCC = 0.9593 | SRCC = 0.9600 ] +INFO:root:Average loss=0.050158897161483765 +INFO:root: +[Valid][19][UTT][ MSE = 0.1967 | LCC = 0.9587 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04656707169115543 +INFO:root: +[Valid][20][UTT][ MSE = 0.1916 | LCC = 0.9588 | SRCC = 0.9592 ] +INFO:root:Average loss=0.04404191240668297 +INFO:root: +[Valid][21][UTT][ MSE = 0.1902 | LCC = 0.9589 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04321418865025044 +INFO:root: +[Valid][22][UTT][ MSE = 0.2021 | LCC = 0.9592 | SRCC = 0.9596 ] +INFO:root:Average loss=0.04252463681995869 +INFO:root: +[Valid][23][UTT][ MSE = 0.1839 | LCC = 0.9604 | SRCC = 0.9605 ] +INFO:root:Average loss=0.04021894545853138 +INFO:root: +[Valid][24][UTT][ MSE = 0.1904 | LCC = 0.9596 | SRCC = 0.9597 ] +INFO:root:Average loss=0.03864576449990272 +INFO:root: +[Valid][25][UTT][ MSE = 0.2030 | LCC = 0.9599 | SRCC = 0.9607 ] +INFO:root:Average loss=0.03659179072082043 +INFO:root: +[Valid][26][UTT][ MSE = 0.1965 | LCC = 0.9592 | SRCC = 0.9591 ] +INFO:root:Average loss=0.03481927110254764 +INFO:root: +[Valid][27][UTT][ MSE = 0.1772 | LCC = 0.9592 | SRCC = 0.9592 ] +INFO:root:Average loss=0.03590240357816219 +INFO:root: +[Valid][28][UTT][ MSE = 0.1810 | LCC = 0.9596 | SRCC = 0.9592 ] +INFO:root:Average loss=0.033336896657943725 +INFO:root: +[Valid][29][UTT][ MSE = 0.1821 | LCC = 0.9592 | SRCC = 0.9590 ] +INFO:root:Average loss=0.03092334695160389 +INFO:root: +[Valid][30][UTT][ MSE = 0.1832 | LCC = 0.9600 | SRCC = 0.9599 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1902 | LCC = 0.9631 | SRCC = 0.9631 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..65c48bf68f1dca42bf9bdad2a50b53e77f6576eb --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8018c1fd24b36966d4180903a4f30fb475d4f0baec664868c7275963bfc70c77 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c82afdf9434a25b56d78c85694313c49e0c52cbf --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:825f51580cec103e17444e97902e6468557a967eefd8ed8408fb81b52efcc5f5 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..326a2fd9fe5055ca362011238fb7ada70f4aef0c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.37320385277271273 +INFO:root: +[Valid][0][UTT][ MSE = 0.2817 | LCC = 0.9304 | SRCC = 0.9303 ] +INFO:root:Average loss=0.17113502311706544 +INFO:root: +[Valid][1][UTT][ MSE = 0.2323 | LCC = 0.9428 | SRCC = 0.9435 ] +INFO:root:Average loss=0.14274987852573395 +INFO:root: +[Valid][2][UTT][ MSE = 0.2443 | LCC = 0.9498 | SRCC = 0.9509 ] +INFO:root:Average loss=0.12415663832426072 +INFO:root: +[Valid][3][UTT][ MSE = 0.2029 | LCC = 0.9519 | SRCC = 0.9525 ] +INFO:root:Average loss=0.11983593112230301 +INFO:root: +[Valid][4][UTT][ MSE = 0.1989 | LCC = 0.9545 | SRCC = 0.9554 ] +INFO:root:Average loss=0.10409631901979446 +INFO:root: +[Valid][5][UTT][ MSE = 0.2099 | LCC = 0.9550 | SRCC = 0.9551 ] +INFO:root:Average loss=0.09635118225216865 +INFO:root: +[Valid][6][UTT][ MSE = 0.1975 | LCC = 0.9563 | SRCC = 0.9569 ] +INFO:root:Average loss=0.08751782801747322 +INFO:root: +[Valid][7][UTT][ MSE = 0.2016 | LCC = 0.9575 | SRCC = 0.9582 ] +INFO:root:Average loss=0.08714604005217552 +INFO:root: +[Valid][8][UTT][ MSE = 0.1757 | LCC = 0.9589 | SRCC = 0.9597 ] +INFO:root:Average loss=0.0784608890414238 +INFO:root: +[Valid][9][UTT][ MSE = 0.1904 | LCC = 0.9572 | SRCC = 0.9586 ] +INFO:root:Average loss=0.07311865967512131 +INFO:root: +[Valid][10][UTT][ MSE = 0.1891 | LCC = 0.9591 | SRCC = 0.9602 ] +INFO:root:Average loss=0.07120506975054741 +INFO:root: +[Valid][11][UTT][ MSE = 0.1747 | LCC = 0.9593 | SRCC = 0.9603 ] +INFO:root:Average loss=0.07355224645137787 +INFO:root: +[Valid][12][UTT][ MSE = 0.1843 | LCC = 0.9564 | SRCC = 0.9581 ] +INFO:root:Average loss=0.06705085849761963 +INFO:root: +[Valid][13][UTT][ MSE = 0.1787 | LCC = 0.9585 | SRCC = 0.9596 ] +INFO:root:Average loss=0.05786131936311722 +INFO:root: +[Valid][14][UTT][ MSE = 0.2100 | LCC = 0.9599 | SRCC = 0.9605 ] +INFO:root:Average loss=0.05668758666515351 +INFO:root: +[Valid][15][UTT][ MSE = 0.1917 | LCC = 0.9606 | SRCC = 0.9608 ] +INFO:root:Average loss=0.05558195179700851 +INFO:root: +[Valid][16][UTT][ MSE = 0.1922 | LCC = 0.9605 | SRCC = 0.9610 ] +INFO:root:Average loss=0.05286922112107277 +INFO:root: +[Valid][17][UTT][ MSE = 0.1708 | LCC = 0.9604 | SRCC = 0.9612 ] +INFO:root:Average loss=0.0494537270963192 +INFO:root: +[Valid][18][UTT][ MSE = 0.1787 | LCC = 0.9584 | SRCC = 0.9587 ] +INFO:root:Average loss=0.04636689195036888 +INFO:root: +[Valid][19][UTT][ MSE = 0.1972 | LCC = 0.9604 | SRCC = 0.9608 ] +INFO:root:Average loss=0.047773210316896436 +INFO:root: +[Valid][20][UTT][ MSE = 0.1810 | LCC = 0.9593 | SRCC = 0.9596 ] +INFO:root:Average loss=0.044091486439108846 +INFO:root: +[Valid][21][UTT][ MSE = 0.1824 | LCC = 0.9603 | SRCC = 0.9610 ] +INFO:root:Average loss=0.044911709904670714 +INFO:root: +[Valid][22][UTT][ MSE = 0.1854 | LCC = 0.9570 | SRCC = 0.9579 ] +INFO:root:Average loss=0.04138122633099556 +INFO:root: +[Valid][23][UTT][ MSE = 0.1932 | LCC = 0.9580 | SRCC = 0.9584 ] +INFO:root:Average loss=0.04073394365608692 +INFO:root: +[Valid][24][UTT][ MSE = 0.1875 | LCC = 0.9589 | SRCC = 0.9590 ] +INFO:root:Average loss=0.040643860012292865 +INFO:root: +[Valid][25][UTT][ MSE = 0.1869 | LCC = 0.9606 | SRCC = 0.9608 ] +INFO:root:Average loss=0.036474543601274494 +INFO:root: +[Valid][26][UTT][ MSE = 0.1943 | LCC = 0.9581 | SRCC = 0.9584 ] +INFO:root:Average loss=0.0345177134424448 +INFO:root: +[Valid][27][UTT][ MSE = 0.1790 | LCC = 0.9594 | SRCC = 0.9597 ] +INFO:root:Average loss=0.033321273237466816 +INFO:root: +[Valid][28][UTT][ MSE = 0.1813 | LCC = 0.9596 | SRCC = 0.9599 ] +INFO:root:Average loss=0.03217491203546524 +INFO:root: +[Valid][29][UTT][ MSE = 0.1896 | LCC = 0.9589 | SRCC = 0.9592 ] +INFO:root:Average loss=0.03179540793597698 +INFO:root: +[Valid][30][UTT][ MSE = 0.1811 | LCC = 0.9583 | SRCC = 0.9586 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1924 | LCC = 0.9621 | SRCC = 0.9627 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e0c31166ddfea55db52f967e0a44fb44f06fe2de --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e4b3d3ecee51dbce08ab13cb6b04d8eed671cce49aa265331f2f9da16a43c272 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a2f546bb68ec69c7eb0e5dd54d11d790ee620404 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c5d8fe3335d466cb955c8558b679240affe86896a780b163b4573e62eb06b721 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b454fa9fe123d51586efb9ca974e6b009c13bea6 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4046903829574585 +INFO:root: +[Valid][0][UTT][ MSE = 0.3160 | LCC = 0.9295 | SRCC = 0.9300 ] +INFO:root:Average loss=0.1712173318862915 +INFO:root: +[Valid][1][UTT][ MSE = 0.2570 | LCC = 0.9414 | SRCC = 0.9416 ] +INFO:root:Average loss=0.14171188044548036 +INFO:root: +[Valid][2][UTT][ MSE = 0.2131 | LCC = 0.9482 | SRCC = 0.9502 ] +INFO:root:Average loss=0.12154416543245315 +INFO:root: +[Valid][3][UTT][ MSE = 0.2062 | LCC = 0.9533 | SRCC = 0.9534 ] +INFO:root:Average loss=0.1129860919713974 +INFO:root: +[Valid][4][UTT][ MSE = 0.1943 | LCC = 0.9551 | SRCC = 0.9555 ] +INFO:root:Average loss=0.10168216860294342 +INFO:root: +[Valid][5][UTT][ MSE = 0.2054 | LCC = 0.9564 | SRCC = 0.9568 ] +INFO:root:Average loss=0.09613592007756233 +INFO:root: +[Valid][6][UTT][ MSE = 0.2096 | LCC = 0.9580 | SRCC = 0.9583 ] +INFO:root:Average loss=0.08780994591116906 +INFO:root: +[Valid][7][UTT][ MSE = 0.2190 | LCC = 0.9581 | SRCC = 0.9584 ] +INFO:root:Average loss=0.08332859134674073 +INFO:root: +[Valid][8][UTT][ MSE = 0.2010 | LCC = 0.9580 | SRCC = 0.9580 ] +INFO:root:Average loss=0.07680934992432595 +INFO:root: +[Valid][9][UTT][ MSE = 0.1959 | LCC = 0.9592 | SRCC = 0.9593 ] +INFO:root:Average loss=0.07740027567744255 +INFO:root: +[Valid][10][UTT][ MSE = 0.2226 | LCC = 0.9587 | SRCC = 0.9588 ] +INFO:root:Average loss=0.07134138929843903 +INFO:root: +[Valid][11][UTT][ MSE = 0.2023 | LCC = 0.9590 | SRCC = 0.9594 ] +INFO:root:Average loss=0.06296368461847306 +INFO:root: +[Valid][12][UTT][ MSE = 0.1941 | LCC = 0.9604 | SRCC = 0.9609 ] +INFO:root:Average loss=0.06369222086668015 +INFO:root: +[Valid][13][UTT][ MSE = 0.1886 | LCC = 0.9604 | SRCC = 0.9611 ] +INFO:root:Average loss=0.061657356917858126 +INFO:root: +[Valid][14][UTT][ MSE = 0.1924 | LCC = 0.9590 | SRCC = 0.9598 ] +INFO:root:Average loss=0.059952604353427884 +INFO:root: +[Valid][15][UTT][ MSE = 0.1739 | LCC = 0.9605 | SRCC = 0.9609 ] +INFO:root:Average loss=0.05505753901600838 +INFO:root: +[Valid][16][UTT][ MSE = 0.1826 | LCC = 0.9611 | SRCC = 0.9617 ] +INFO:root:Average loss=0.05041541653871536 +INFO:root: +[Valid][17][UTT][ MSE = 0.1876 | LCC = 0.9604 | SRCC = 0.9609 ] +INFO:root:Average loss=0.04922955647110939 +INFO:root: +[Valid][18][UTT][ MSE = 0.1782 | LCC = 0.9598 | SRCC = 0.9605 ] +INFO:root:Average loss=0.04730920681357384 +INFO:root: +[Valid][19][UTT][ MSE = 0.2122 | LCC = 0.9592 | SRCC = 0.9606 ] +INFO:root:Average loss=0.044819448471069334 +INFO:root: +[Valid][20][UTT][ MSE = 0.1923 | LCC = 0.9612 | SRCC = 0.9616 ] +INFO:root:Average loss=0.04418468557298184 +INFO:root: +[Valid][21][UTT][ MSE = 0.1768 | LCC = 0.9613 | SRCC = 0.9615 ] +INFO:root:Average loss=0.04188068924844265 +INFO:root: +[Valid][22][UTT][ MSE = 0.1698 | LCC = 0.9602 | SRCC = 0.9606 ] +INFO:root:Average loss=0.03913301755487919 +INFO:root: +[Valid][23][UTT][ MSE = 0.1739 | LCC = 0.9614 | SRCC = 0.9615 ] +INFO:root:Average loss=0.0396536867916584 +INFO:root: +[Valid][24][UTT][ MSE = 0.1902 | LCC = 0.9617 | SRCC = 0.9618 ] +INFO:root:Average loss=0.03688461880385876 +INFO:root: +[Valid][25][UTT][ MSE = 0.1865 | LCC = 0.9619 | SRCC = 0.9620 ] +INFO:root:Average loss=0.03623818454146385 +INFO:root: +[Valid][26][UTT][ MSE = 0.2180 | LCC = 0.9610 | SRCC = 0.9611 ] +INFO:root:Average loss=0.03471230265498161 +INFO:root: +[Valid][27][UTT][ MSE = 0.1856 | LCC = 0.9613 | SRCC = 0.9617 ] +INFO:root:Average loss=0.033661818727850915 +INFO:root: +[Valid][28][UTT][ MSE = 0.1670 | LCC = 0.9619 | SRCC = 0.9621 ] +INFO:root:Average loss=0.03230272567272186 +INFO:root: +[Valid][29][UTT][ MSE = 0.1864 | LCC = 0.9619 | SRCC = 0.9616 ] +INFO:root:Average loss=0.03175543296337128 +INFO:root: +[Valid][30][UTT][ MSE = 0.1858 | LCC = 0.9613 | SRCC = 0.9612 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1877 | LCC = 0.9620 | SRCC = 0.9622 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..555bc6a701a60480a28862a3d2a21b03f75c07a5 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5465e993be24c389a8e9527b54af18e38dcc48bc2932569d8529eec2cefd0d12 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7485100633a9c358421b43c003e638f36e872584 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c34f19c28a0bcdf4928eb4f20dbcf12418557c38d50422eb395b2785933d7403 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e5e01c44eb4cb9e1369b90b06c37fd021efdbf6f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer2/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.39208995032310484 +INFO:root: +[Valid][0][UTT][ MSE = 0.2519 | LCC = 0.9262 | SRCC = 0.9277 ] +INFO:root:Average loss=0.17426038843393327 +INFO:root: +[Valid][1][UTT][ MSE = 0.2462 | LCC = 0.9443 | SRCC = 0.9444 ] +INFO:root:Average loss=0.14091511327028275 +INFO:root: +[Valid][2][UTT][ MSE = 0.2328 | LCC = 0.9510 | SRCC = 0.9517 ] +INFO:root:Average loss=0.126977580845356 +INFO:root: +[Valid][3][UTT][ MSE = 0.2067 | LCC = 0.9534 | SRCC = 0.9537 ] +INFO:root:Average loss=0.11063832676410675 +INFO:root: +[Valid][4][UTT][ MSE = 0.2175 | LCC = 0.9554 | SRCC = 0.9559 ] +INFO:root:Average loss=0.10218813267350196 +INFO:root: +[Valid][5][UTT][ MSE = 0.2098 | LCC = 0.9564 | SRCC = 0.9564 ] +INFO:root:Average loss=0.09333678689599037 +INFO:root: +[Valid][6][UTT][ MSE = 0.2125 | LCC = 0.9584 | SRCC = 0.9588 ] +INFO:root:Average loss=0.08822806122899056 +INFO:root: +[Valid][7][UTT][ MSE = 0.1854 | LCC = 0.9589 | SRCC = 0.9596 ] +INFO:root:Average loss=0.08209411412477494 +INFO:root: +[Valid][8][UTT][ MSE = 0.1833 | LCC = 0.9577 | SRCC = 0.9579 ] +INFO:root:Average loss=0.07938002079725265 +INFO:root: +[Valid][9][UTT][ MSE = 0.2161 | LCC = 0.9592 | SRCC = 0.9597 ] +INFO:root:Average loss=0.07464420926570893 +INFO:root: +[Valid][10][UTT][ MSE = 0.2121 | LCC = 0.9594 | SRCC = 0.9598 ] +INFO:root:Average loss=0.06893349623680115 +INFO:root: +[Valid][11][UTT][ MSE = 0.1840 | LCC = 0.9588 | SRCC = 0.9590 ] +INFO:root:Average loss=0.06552802953124047 +INFO:root: +[Valid][12][UTT][ MSE = 0.2149 | LCC = 0.9573 | SRCC = 0.9569 ] +INFO:root:Average loss=0.06428874057531357 +INFO:root: +[Valid][13][UTT][ MSE = 0.1923 | LCC = 0.9609 | SRCC = 0.9611 ] +INFO:root:Average loss=0.06134999135136604 +INFO:root: +[Valid][14][UTT][ MSE = 0.1970 | LCC = 0.9606 | SRCC = 0.9609 ] +INFO:root:Average loss=0.05633388343453407 +INFO:root: +[Valid][15][UTT][ MSE = 0.1910 | LCC = 0.9604 | SRCC = 0.9606 ] +INFO:root:Average loss=0.0559159991145134 +INFO:root: +[Valid][16][UTT][ MSE = 0.1802 | LCC = 0.9590 | SRCC = 0.9591 ] +INFO:root:Average loss=0.05236103381216526 +INFO:root: +[Valid][17][UTT][ MSE = 0.1930 | LCC = 0.9576 | SRCC = 0.9577 ] +INFO:root:Average loss=0.04926888267695904 +INFO:root: +[Valid][18][UTT][ MSE = 0.2104 | LCC = 0.9561 | SRCC = 0.9562 ] +INFO:root:Average loss=0.05004231402277946 +INFO:root: +[Valid][19][UTT][ MSE = 0.1960 | LCC = 0.9599 | SRCC = 0.9601 ] +INFO:root:Average loss=0.04630968050658703 +INFO:root: +[Valid][20][UTT][ MSE = 0.1873 | LCC = 0.9608 | SRCC = 0.9607 ] +INFO:root:Average loss=0.04456302855908871 +INFO:root: +[Valid][21][UTT][ MSE = 0.1806 | LCC = 0.9605 | SRCC = 0.9603 ] +INFO:root:Average loss=0.04326164166629314 +INFO:root: +[Valid][22][UTT][ MSE = 0.1859 | LCC = 0.9605 | SRCC = 0.9607 ] +INFO:root:Average loss=0.039171313494443895 +INFO:root: +[Valid][23][UTT][ MSE = 0.1695 | LCC = 0.9615 | SRCC = 0.9616 ] +INFO:root:Average loss=0.0399609866142273 +INFO:root: +[Valid][24][UTT][ MSE = 0.1723 | LCC = 0.9604 | SRCC = 0.9607 ] +INFO:root:Average loss=0.038154055535793305 +INFO:root: +[Valid][25][UTT][ MSE = 0.1785 | LCC = 0.9622 | SRCC = 0.9617 ] +INFO:root:Average loss=0.03592310260236263 +INFO:root: +[Valid][26][UTT][ MSE = 0.1844 | LCC = 0.9614 | SRCC = 0.9615 ] +INFO:root:Average loss=0.034783662125468254 +INFO:root: +[Valid][27][UTT][ MSE = 0.1897 | LCC = 0.9621 | SRCC = 0.9621 ] +INFO:root:Average loss=0.03360560680925846 +INFO:root: +[Valid][28][UTT][ MSE = 0.1707 | LCC = 0.9617 | SRCC = 0.9615 ] +INFO:root:Average loss=0.032955035835504534 +INFO:root: +[Valid][29][UTT][ MSE = 0.1825 | LCC = 0.9608 | SRCC = 0.9605 ] +INFO:root:Average loss=0.031597038984298706 +INFO:root: +[Valid][30][UTT][ MSE = 0.1631 | LCC = 0.9614 | SRCC = 0.9612 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1800 | LCC = 0.9632 | SRCC = 0.9628 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..61d68c324a344f331b8b3c8a1d7d5266dd20fa24 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b10564ed8b81de351954602d3aa10af0ea3ea4b98c73a64b24d6909829b6c92a +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ad5be455a95001c03c43765bf298f0e95e52ef9f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e8c84a158a2c51b29c1f674a951d6e4636b75f96c29fc08ea973da3887cccd4 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..136ad8f90c04632030aca1080d35fd17417d23d2 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4287740668058395 +INFO:root: +[Valid][0][UTT][ MSE = 0.2454 | LCC = 0.9364 | SRCC = 0.9379 ] +INFO:root:Average loss=0.15769613540172578 +INFO:root: +[Valid][1][UTT][ MSE = 0.2102 | LCC = 0.9473 | SRCC = 0.9487 ] +INFO:root:Average loss=0.1418088579773903 +INFO:root: +[Valid][2][UTT][ MSE = 0.2359 | LCC = 0.9506 | SRCC = 0.9513 ] +INFO:root:Average loss=0.12154997730255127 +INFO:root: +[Valid][3][UTT][ MSE = 0.2167 | LCC = 0.9546 | SRCC = 0.9550 ] +INFO:root:Average loss=0.11224934154748917 +INFO:root: +[Valid][4][UTT][ MSE = 0.2368 | LCC = 0.9564 | SRCC = 0.9573 ] +INFO:root:Average loss=0.10211182263493537 +INFO:root: +[Valid][5][UTT][ MSE = 0.2106 | LCC = 0.9582 | SRCC = 0.9586 ] +INFO:root:Average loss=0.09602621856331825 +INFO:root: +[Valid][6][UTT][ MSE = 0.1977 | LCC = 0.9569 | SRCC = 0.9585 ] +INFO:root:Average loss=0.08676207058131695 +INFO:root: +[Valid][7][UTT][ MSE = 0.1820 | LCC = 0.9583 | SRCC = 0.9590 ] +INFO:root:Average loss=0.08388190540671349 +INFO:root: +[Valid][8][UTT][ MSE = 0.1993 | LCC = 0.9561 | SRCC = 0.9569 ] +INFO:root:Average loss=0.07696693760156631 +INFO:root: +[Valid][9][UTT][ MSE = 0.2066 | LCC = 0.9596 | SRCC = 0.9599 ] +INFO:root:Average loss=0.07100330141186714 +INFO:root: +[Valid][10][UTT][ MSE = 0.2165 | LCC = 0.9581 | SRCC = 0.9599 ] +INFO:root:Average loss=0.06864057728648186 +INFO:root: +[Valid][11][UTT][ MSE = 0.2121 | LCC = 0.9574 | SRCC = 0.9579 ] +INFO:root:Average loss=0.06946399945020676 +INFO:root: +[Valid][12][UTT][ MSE = 0.1935 | LCC = 0.9602 | SRCC = 0.9610 ] +INFO:root:Average loss=0.06109640210866928 +INFO:root: +[Valid][13][UTT][ MSE = 0.2121 | LCC = 0.9585 | SRCC = 0.9592 ] +INFO:root:Average loss=0.05817800614237785 +INFO:root: +[Valid][14][UTT][ MSE = 0.1731 | LCC = 0.9583 | SRCC = 0.9589 ] +INFO:root:Average loss=0.05380546846985817 +INFO:root: +[Valid][15][UTT][ MSE = 0.2130 | LCC = 0.9571 | SRCC = 0.9570 ] +INFO:root:Average loss=0.05157685036957264 +INFO:root: +[Valid][16][UTT][ MSE = 0.1680 | LCC = 0.9594 | SRCC = 0.9602 ] +INFO:root:Average loss=0.05326346117258072 +INFO:root: +[Valid][17][UTT][ MSE = 0.1875 | LCC = 0.9603 | SRCC = 0.9599 ] +INFO:root:Average loss=0.049640038937330246 +INFO:root: +[Valid][18][UTT][ MSE = 0.1954 | LCC = 0.9600 | SRCC = 0.9605 ] +INFO:root:Average loss=0.04815455523133278 +INFO:root: +[Valid][19][UTT][ MSE = 0.1799 | LCC = 0.9607 | SRCC = 0.9610 ] +INFO:root:Average loss=0.046536705389618876 +INFO:root: +[Valid][20][UTT][ MSE = 0.2090 | LCC = 0.9600 | SRCC = 0.9603 ] +INFO:root:Average loss=0.04267659217119217 +INFO:root: +[Valid][21][UTT][ MSE = 0.1648 | LCC = 0.9591 | SRCC = 0.9596 ] +INFO:root:Average loss=0.04190800760686397 +INFO:root: +[Valid][22][UTT][ MSE = 0.1845 | LCC = 0.9605 | SRCC = 0.9607 ] +INFO:root:Average loss=0.04087987804412842 +INFO:root: +[Valid][23][UTT][ MSE = 0.1905 | LCC = 0.9591 | SRCC = 0.9588 ] +INFO:root:Average loss=0.03742321023344994 +INFO:root: +[Valid][24][UTT][ MSE = 0.1984 | LCC = 0.9576 | SRCC = 0.9577 ] +INFO:root:Average loss=0.03586369432508946 +INFO:root: +[Valid][25][UTT][ MSE = 0.1973 | LCC = 0.9584 | SRCC = 0.9580 ] +INFO:root:Average loss=0.03660762448608875 +INFO:root: +[Valid][26][UTT][ MSE = 0.1960 | LCC = 0.9595 | SRCC = 0.9593 ] +INFO:root:Average loss=0.033148269832134244 +INFO:root: +[Valid][27][UTT][ MSE = 0.1901 | LCC = 0.9578 | SRCC = 0.9582 ] +INFO:root:Average loss=0.033859216809272764 +INFO:root: +[Valid][28][UTT][ MSE = 0.1984 | LCC = 0.9593 | SRCC = 0.9591 ] +INFO:root:Average loss=0.032301793873310086 +INFO:root: +[Valid][29][UTT][ MSE = 0.1997 | LCC = 0.9595 | SRCC = 0.9594 ] +INFO:root:Average loss=0.03018148408830166 +INFO:root: +[Valid][30][UTT][ MSE = 0.1723 | LCC = 0.9589 | SRCC = 0.9590 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1835 | LCC = 0.9617 | SRCC = 0.9623 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..075fb749b5a448cb1801fd4c88d23ee5b3fb08c9 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7a1ba82b3b31e85c45d0adf316df0a09e8f48e6aa5ce58dc28adf2faec23e553 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4af643210def3b56b759d7dc859abae59346dce9 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e361dd27183c4562cdca1f685af8277a98c7306017a49f025d90d634cda1b1f9 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f3f736f58b85a486383d8e26864a3414011207e8 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4359325876235962 +INFO:root: +[Valid][0][UTT][ MSE = 0.2859 | LCC = 0.9300 | SRCC = 0.9315 ] +INFO:root:Average loss=0.16952040612697603 +INFO:root: +[Valid][1][UTT][ MSE = 0.2270 | LCC = 0.9441 | SRCC = 0.9455 ] +INFO:root:Average loss=0.1370247420668602 +INFO:root: +[Valid][2][UTT][ MSE = 0.2454 | LCC = 0.9467 | SRCC = 0.9467 ] +INFO:root:Average loss=0.12448304069042206 +INFO:root: +[Valid][3][UTT][ MSE = 0.2377 | LCC = 0.9529 | SRCC = 0.9543 ] +INFO:root:Average loss=0.10794764894247055 +INFO:root: +[Valid][4][UTT][ MSE = 0.2060 | LCC = 0.9541 | SRCC = 0.9556 ] +INFO:root:Average loss=0.09618583169579506 +INFO:root: +[Valid][5][UTT][ MSE = 0.2212 | LCC = 0.9570 | SRCC = 0.9574 ] +INFO:root:Average loss=0.09476396849751473 +INFO:root: +[Valid][6][UTT][ MSE = 0.1938 | LCC = 0.9585 | SRCC = 0.9587 ] +INFO:root:Average loss=0.08779235339164734 +INFO:root: +[Valid][7][UTT][ MSE = 0.1995 | LCC = 0.9580 | SRCC = 0.9583 ] +INFO:root:Average loss=0.08186386877298354 +INFO:root: +[Valid][8][UTT][ MSE = 0.2288 | LCC = 0.9582 | SRCC = 0.9595 ] +INFO:root:Average loss=0.07949888059496879 +INFO:root: +[Valid][9][UTT][ MSE = 0.1985 | LCC = 0.9585 | SRCC = 0.9590 ] +INFO:root:Average loss=0.07280512607097625 +INFO:root: +[Valid][10][UTT][ MSE = 0.1887 | LCC = 0.9568 | SRCC = 0.9573 ] +INFO:root:Average loss=0.06926583144068718 +INFO:root: +[Valid][11][UTT][ MSE = 0.1979 | LCC = 0.9603 | SRCC = 0.9601 ] +INFO:root:Average loss=0.06615937101840973 +INFO:root: +[Valid][12][UTT][ MSE = 0.2019 | LCC = 0.9606 | SRCC = 0.9609 ] +INFO:root:Average loss=0.06139174392819405 +INFO:root: +[Valid][13][UTT][ MSE = 0.1981 | LCC = 0.9607 | SRCC = 0.9608 ] +INFO:root:Average loss=0.059495743930339816 +INFO:root: +[Valid][14][UTT][ MSE = 0.1979 | LCC = 0.9608 | SRCC = 0.9608 ] +INFO:root:Average loss=0.056687614500522614 +INFO:root: +[Valid][15][UTT][ MSE = 0.1789 | LCC = 0.9613 | SRCC = 0.9612 ] +INFO:root:Average loss=0.05491618767380715 +INFO:root: +[Valid][16][UTT][ MSE = 0.2123 | LCC = 0.9589 | SRCC = 0.9588 ] +INFO:root:Average loss=0.051394963771104814 +INFO:root: +[Valid][17][UTT][ MSE = 0.2194 | LCC = 0.9581 | SRCC = 0.9579 ] +INFO:root:Average loss=0.05043763069808483 +INFO:root: +[Valid][18][UTT][ MSE = 0.1991 | LCC = 0.9595 | SRCC = 0.9592 ] +INFO:root:Average loss=0.04581816354393959 +INFO:root: +[Valid][19][UTT][ MSE = 0.2020 | LCC = 0.9591 | SRCC = 0.9589 ] +INFO:root:Average loss=0.044705509901046755 +INFO:root: +[Valid][20][UTT][ MSE = 0.1985 | LCC = 0.9578 | SRCC = 0.9579 ] +INFO:root:Average loss=0.0441566329151392 +INFO:root: +[Valid][21][UTT][ MSE = 0.2120 | LCC = 0.9597 | SRCC = 0.9598 ] +INFO:root:Average loss=0.04128260526061058 +INFO:root: +[Valid][22][UTT][ MSE = 0.1896 | LCC = 0.9604 | SRCC = 0.9602 ] +INFO:root:Average loss=0.039908301144838336 +INFO:root: +[Valid][23][UTT][ MSE = 0.2013 | LCC = 0.9581 | SRCC = 0.9573 ] +INFO:root:Average loss=0.03906964795291424 +INFO:root: +[Valid][24][UTT][ MSE = 0.1943 | LCC = 0.9592 | SRCC = 0.9584 ] +INFO:root:Average loss=0.036511901140213014 +INFO:root: +[Valid][25][UTT][ MSE = 0.1856 | LCC = 0.9611 | SRCC = 0.9608 ] +INFO:root:Average loss=0.03675111357867718 +INFO:root: +[Valid][26][UTT][ MSE = 0.1936 | LCC = 0.9598 | SRCC = 0.9590 ] +INFO:root:Average loss=0.03490931595861912 +INFO:root: +[Valid][27][UTT][ MSE = 0.1897 | LCC = 0.9605 | SRCC = 0.9602 ] +INFO:root:Average loss=0.034124983951449396 +INFO:root: +[Valid][28][UTT][ MSE = 0.1991 | LCC = 0.9602 | SRCC = 0.9606 ] +INFO:root:Average loss=0.03351705484092236 +INFO:root: +[Valid][29][UTT][ MSE = 0.1924 | LCC = 0.9603 | SRCC = 0.9597 ] +INFO:root:Average loss=0.0337140655964613 +INFO:root: +[Valid][30][UTT][ MSE = 0.1900 | LCC = 0.9604 | SRCC = 0.9598 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1845 | LCC = 0.9619 | SRCC = 0.9620 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..603cd87f6f25b82dc152bb70df868d0722954546 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7a9df82138567d772b5f5b8597839edacf9f7d34f4234d38de00efa67ab16fcb +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c83924bc2caf7a9d875cc41e5101889c3e22eef9 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:35daab40eafa78d3a26922d7b20d60bac31c0219d541f1fe12b62bd2c90d1dbe +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..7398ae1040070fe3e5d89c445aff098876decbdc --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3452635897397995 +INFO:root: +[Valid][0][UTT][ MSE = 0.3052 | LCC = 0.9337 | SRCC = 0.9363 ] +INFO:root:Average loss=0.15610538631677628 +INFO:root: +[Valid][1][UTT][ MSE = 0.2113 | LCC = 0.9503 | SRCC = 0.9506 ] +INFO:root:Average loss=0.1373106309771538 +INFO:root: +[Valid][2][UTT][ MSE = 0.2325 | LCC = 0.9545 | SRCC = 0.9547 ] +INFO:root:Average loss=0.11487879118323326 +INFO:root: +[Valid][3][UTT][ MSE = 0.2046 | LCC = 0.9562 | SRCC = 0.9574 ] +INFO:root:Average loss=0.10793872728943825 +INFO:root: +[Valid][4][UTT][ MSE = 0.2196 | LCC = 0.9585 | SRCC = 0.9596 ] +INFO:root:Average loss=0.0982695409655571 +INFO:root: +[Valid][5][UTT][ MSE = 0.2114 | LCC = 0.9591 | SRCC = 0.9598 ] +INFO:root:Average loss=0.09338906615972518 +INFO:root: +[Valid][6][UTT][ MSE = 0.1938 | LCC = 0.9590 | SRCC = 0.9601 ] +INFO:root:Average loss=0.08491008853912353 +INFO:root: +[Valid][7][UTT][ MSE = 0.2072 | LCC = 0.9604 | SRCC = 0.9606 ] +INFO:root:Average loss=0.07927806824445724 +INFO:root: +[Valid][8][UTT][ MSE = 0.2030 | LCC = 0.9612 | SRCC = 0.9621 ] +INFO:root:Average loss=0.07635519781708718 +INFO:root: +[Valid][9][UTT][ MSE = 0.2136 | LCC = 0.9605 | SRCC = 0.9612 ] +INFO:root:Average loss=0.07320140868425369 +INFO:root: +[Valid][10][UTT][ MSE = 0.1796 | LCC = 0.9574 | SRCC = 0.9582 ] +INFO:root:Average loss=0.06705818137526512 +INFO:root: +[Valid][11][UTT][ MSE = 0.2180 | LCC = 0.9614 | SRCC = 0.9627 ] +INFO:root:Average loss=0.06411692291498185 +INFO:root: +[Valid][12][UTT][ MSE = 0.2183 | LCC = 0.9611 | SRCC = 0.9619 ] +INFO:root:Average loss=0.06326044836640358 +INFO:root: +[Valid][13][UTT][ MSE = 0.1906 | LCC = 0.9600 | SRCC = 0.9606 ] +INFO:root:Average loss=0.05860505047440529 +INFO:root: +[Valid][14][UTT][ MSE = 0.1932 | LCC = 0.9619 | SRCC = 0.9619 ] +INFO:root:Average loss=0.0550775461345911 +INFO:root: +[Valid][15][UTT][ MSE = 0.2265 | LCC = 0.9607 | SRCC = 0.9614 ] +INFO:root:Average loss=0.054033829912543296 +INFO:root: +[Valid][16][UTT][ MSE = 0.2014 | LCC = 0.9612 | SRCC = 0.9617 ] +INFO:root:Average loss=0.05056327031552792 +INFO:root: +[Valid][17][UTT][ MSE = 0.2081 | LCC = 0.9592 | SRCC = 0.9597 ] +INFO:root:Average loss=0.049560311794281005 +INFO:root: +[Valid][18][UTT][ MSE = 0.2070 | LCC = 0.9594 | SRCC = 0.9594 ] +INFO:root:Average loss=0.044368858337402346 +INFO:root: +[Valid][19][UTT][ MSE = 0.2050 | LCC = 0.9604 | SRCC = 0.9604 ] +INFO:root:Average loss=0.04245235143601894 +INFO:root: +[Valid][20][UTT][ MSE = 0.2173 | LCC = 0.9602 | SRCC = 0.9604 ] +INFO:root:Average loss=0.04336317339539528 +INFO:root: +[Valid][21][UTT][ MSE = 0.2007 | LCC = 0.9605 | SRCC = 0.9609 ] +INFO:root:Average loss=0.03984508626163006 +INFO:root: +[Valid][22][UTT][ MSE = 0.2015 | LCC = 0.9613 | SRCC = 0.9613 ] +INFO:root:Average loss=0.041448670357465744 +INFO:root: +[Valid][23][UTT][ MSE = 0.2080 | LCC = 0.9597 | SRCC = 0.9596 ] +INFO:root:Average loss=0.038118569299578665 +INFO:root: +[Valid][24][UTT][ MSE = 0.2127 | LCC = 0.9609 | SRCC = 0.9610 ] +INFO:root:Average loss=0.0366579931974411 +INFO:root: +[Valid][25][UTT][ MSE = 0.2036 | LCC = 0.9600 | SRCC = 0.9608 ] +INFO:root:Average loss=0.0353623004257679 +INFO:root: +[Valid][26][UTT][ MSE = 0.1912 | LCC = 0.9607 | SRCC = 0.9605 ] +INFO:root:Average loss=0.03330779393017292 +INFO:root: +[Valid][27][UTT][ MSE = 0.1913 | LCC = 0.9587 | SRCC = 0.9591 ] +INFO:root:Average loss=0.03417316311597824 +INFO:root: +[Valid][28][UTT][ MSE = 0.1931 | LCC = 0.9613 | SRCC = 0.9611 ] +INFO:root:Average loss=0.033114988788962366 +INFO:root: +[Valid][29][UTT][ MSE = 0.1933 | LCC = 0.9605 | SRCC = 0.9608 ] +INFO:root:Average loss=0.02989492008090019 +INFO:root: +[Valid][30][UTT][ MSE = 0.1820 | LCC = 0.9607 | SRCC = 0.9610 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1940 | LCC = 0.9634 | SRCC = 0.9633 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..886919daa03833788ec64205d80488f1eb26637b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a4731adaf662f15a41726a9ddd81af2a873656bac8725299a1786d41d676e05a +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a542c544ce6446a3323d9ca3671e1befa94550be --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:970fb251466396424116a94abb33fc24323a072e93822262cd2e2a49ed537bdb +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..511883d4d15ce17de58596d2a5914c62ee14ddab --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3524146439433098 +INFO:root: +[Valid][0][UTT][ MSE = 0.3262 | LCC = 0.9377 | SRCC = 0.9393 ] +INFO:root:Average loss=0.15546491688489913 +INFO:root: +[Valid][1][UTT][ MSE = 0.2344 | LCC = 0.9499 | SRCC = 0.9507 ] +INFO:root:Average loss=0.1325764063000679 +INFO:root: +[Valid][2][UTT][ MSE = 0.2119 | LCC = 0.9534 | SRCC = 0.9539 ] +INFO:root:Average loss=0.1141132264137268 +INFO:root: +[Valid][3][UTT][ MSE = 0.2206 | LCC = 0.9555 | SRCC = 0.9558 ] +INFO:root:Average loss=0.10672926193475724 +INFO:root: +[Valid][4][UTT][ MSE = 0.2380 | LCC = 0.9542 | SRCC = 0.9575 ] +INFO:root:Average loss=0.09860274550318718 +INFO:root: +[Valid][5][UTT][ MSE = 0.2257 | LCC = 0.9566 | SRCC = 0.9595 ] +INFO:root:Average loss=0.09159068113565445 +INFO:root: +[Valid][6][UTT][ MSE = 0.2289 | LCC = 0.9572 | SRCC = 0.9573 ] +INFO:root:Average loss=0.08562250941991806 +INFO:root: +[Valid][7][UTT][ MSE = 0.1949 | LCC = 0.9605 | SRCC = 0.9613 ] +INFO:root:Average loss=0.08019080033898354 +INFO:root: +[Valid][8][UTT][ MSE = 0.1862 | LCC = 0.9602 | SRCC = 0.9605 ] +INFO:root:Average loss=0.07524429440498352 +INFO:root: +[Valid][9][UTT][ MSE = 0.2130 | LCC = 0.9595 | SRCC = 0.9604 ] +INFO:root:Average loss=0.07086783024668694 +INFO:root: +[Valid][10][UTT][ MSE = 0.1841 | LCC = 0.9605 | SRCC = 0.9604 ] +INFO:root:Average loss=0.06925582110881806 +INFO:root: +[Valid][11][UTT][ MSE = 0.1841 | LCC = 0.9604 | SRCC = 0.9606 ] +INFO:root:Average loss=0.06443569231033325 +INFO:root: +[Valid][12][UTT][ MSE = 0.2247 | LCC = 0.9604 | SRCC = 0.9621 ] +INFO:root:Average loss=0.06166176778078079 +INFO:root: +[Valid][13][UTT][ MSE = 0.2128 | LCC = 0.9598 | SRCC = 0.9614 ] +INFO:root:Average loss=0.05775927959382534 +INFO:root: +[Valid][14][UTT][ MSE = 0.2155 | LCC = 0.9604 | SRCC = 0.9614 ] +INFO:root:Average loss=0.05611680719256401 +INFO:root: +[Valid][15][UTT][ MSE = 0.1793 | LCC = 0.9605 | SRCC = 0.9607 ] +INFO:root:Average loss=0.05226528899371624 +INFO:root: +[Valid][16][UTT][ MSE = 0.1910 | LCC = 0.9612 | SRCC = 0.9616 ] +INFO:root:Average loss=0.051237036138772965 +INFO:root: +[Valid][17][UTT][ MSE = 0.2091 | LCC = 0.9596 | SRCC = 0.9609 ] +INFO:root:Average loss=0.04969743891060353 +INFO:root: +[Valid][18][UTT][ MSE = 0.1978 | LCC = 0.9596 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04619717156887054 +INFO:root: +[Valid][19][UTT][ MSE = 0.1827 | LCC = 0.9602 | SRCC = 0.9610 ] +INFO:root:Average loss=0.04382316246628761 +INFO:root: +[Valid][20][UTT][ MSE = 0.2102 | LCC = 0.9599 | SRCC = 0.9614 ] +INFO:root:Average loss=0.04264224810898304 +INFO:root: +[Valid][21][UTT][ MSE = 0.2017 | LCC = 0.9603 | SRCC = 0.9602 ] +INFO:root:Average loss=0.04064175923168659 +INFO:root: +[Valid][22][UTT][ MSE = 0.2030 | LCC = 0.9620 | SRCC = 0.9620 ] +INFO:root:Average loss=0.03921466708183288 +INFO:root: +[Valid][23][UTT][ MSE = 0.1983 | LCC = 0.9601 | SRCC = 0.9601 ] +INFO:root:Average loss=0.03802167086303234 +INFO:root: +[Valid][24][UTT][ MSE = 0.1822 | LCC = 0.9612 | SRCC = 0.9611 ] +INFO:root:Average loss=0.035626949444413185 +INFO:root: +[Valid][25][UTT][ MSE = 0.1858 | LCC = 0.9610 | SRCC = 0.9607 ] +INFO:root:Average loss=0.03855367739498615 +INFO:root: +[Valid][26][UTT][ MSE = 0.2229 | LCC = 0.9581 | SRCC = 0.9596 ] +INFO:root:Average loss=0.035598322451114654 +INFO:root: +[Valid][27][UTT][ MSE = 0.1916 | LCC = 0.9594 | SRCC = 0.9598 ] +INFO:root:Average loss=0.0331553463190794 +INFO:root: +[Valid][28][UTT][ MSE = 0.2053 | LCC = 0.9599 | SRCC = 0.9607 ] +INFO:root:Average loss=0.03247456903755665 +INFO:root: +[Valid][29][UTT][ MSE = 0.2116 | LCC = 0.9613 | SRCC = 0.9618 ] +INFO:root:Average loss=0.03178111781179905 +INFO:root: +[Valid][30][UTT][ MSE = 0.1893 | LCC = 0.9615 | SRCC = 0.9615 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2083 | LCC = 0.9620 | SRCC = 0.9624 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..92c34e97cb770c3b687e79a09f7da451c303d360 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1eb7750b96d9a585caf3ccec15caced815edbab09a1cc8990a7b29e98e48a465 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..931300f7c97be0ced64a10b9c6fb6832793ec055 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e4a4a72eeaaff0bc14e6058385a64cc36b5c90cbd3272633854acc2125f0412 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c3b1f7a9689090dbeb0df7a3af4bfae84bf36cbb --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer3/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42577731347084047 +INFO:root: +[Valid][0][UTT][ MSE = 0.2252 | LCC = 0.9346 | SRCC = 0.9371 ] +INFO:root:Average loss=0.16224783146381377 +INFO:root: +[Valid][1][UTT][ MSE = 0.2532 | LCC = 0.9473 | SRCC = 0.9477 ] +INFO:root:Average loss=0.1350227999687195 +INFO:root: +[Valid][2][UTT][ MSE = 0.2275 | LCC = 0.9537 | SRCC = 0.9546 ] +INFO:root:Average loss=0.11712997275590896 +INFO:root: +[Valid][3][UTT][ MSE = 0.2321 | LCC = 0.9572 | SRCC = 0.9581 ] +INFO:root:Average loss=0.10698774909973144 +INFO:root: +[Valid][4][UTT][ MSE = 0.2236 | LCC = 0.9572 | SRCC = 0.9580 ] +INFO:root:Average loss=0.1050637735426426 +INFO:root: +[Valid][5][UTT][ MSE = 0.1940 | LCC = 0.9579 | SRCC = 0.9581 ] +INFO:root:Average loss=0.0924404633641243 +INFO:root: +[Valid][6][UTT][ MSE = 0.1957 | LCC = 0.9591 | SRCC = 0.9594 ] +INFO:root:Average loss=0.08576106098294257 +INFO:root: +[Valid][7][UTT][ MSE = 0.2037 | LCC = 0.9602 | SRCC = 0.9605 ] +INFO:root:Average loss=0.08065507814288139 +INFO:root: +[Valid][8][UTT][ MSE = 0.1815 | LCC = 0.9605 | SRCC = 0.9610 ] +INFO:root:Average loss=0.08056152877211571 +INFO:root: +[Valid][9][UTT][ MSE = 0.2006 | LCC = 0.9604 | SRCC = 0.9611 ] +INFO:root:Average loss=0.0704458135664463 +INFO:root: +[Valid][10][UTT][ MSE = 0.1831 | LCC = 0.9611 | SRCC = 0.9617 ] +INFO:root:Average loss=0.06652435180544854 +INFO:root: +[Valid][11][UTT][ MSE = 0.1918 | LCC = 0.9605 | SRCC = 0.9611 ] +INFO:root:Average loss=0.06372630721330642 +INFO:root: +[Valid][12][UTT][ MSE = 0.1533 | LCC = 0.9607 | SRCC = 0.9612 ] +INFO:root:Average loss=0.06040510360896587 +INFO:root: +[Valid][13][UTT][ MSE = 0.1813 | LCC = 0.9594 | SRCC = 0.9599 ] +INFO:root:Average loss=0.05636016407608986 +INFO:root: +[Valid][14][UTT][ MSE = 0.1839 | LCC = 0.9606 | SRCC = 0.9609 ] +INFO:root:Average loss=0.05303274828195572 +INFO:root: +[Valid][15][UTT][ MSE = 0.2006 | LCC = 0.9586 | SRCC = 0.9597 ] +INFO:root:Average loss=0.05353266625106335 +INFO:root: +[Valid][16][UTT][ MSE = 0.1844 | LCC = 0.9613 | SRCC = 0.9617 ] +INFO:root:Average loss=0.04911442618072033 +INFO:root: +[Valid][17][UTT][ MSE = 0.1717 | LCC = 0.9604 | SRCC = 0.9606 ] +INFO:root:Average loss=0.05135999692976475 +INFO:root: +[Valid][18][UTT][ MSE = 0.1839 | LCC = 0.9611 | SRCC = 0.9615 ] +INFO:root:Average loss=0.04550264637172222 +INFO:root: +[Valid][19][UTT][ MSE = 0.2000 | LCC = 0.9586 | SRCC = 0.9601 ] +INFO:root:Average loss=0.04469928625226021 +INFO:root: +[Valid][20][UTT][ MSE = 0.1840 | LCC = 0.9607 | SRCC = 0.9609 ] +INFO:root:Average loss=0.041932889312505724 +INFO:root: +[Valid][21][UTT][ MSE = 0.2076 | LCC = 0.9603 | SRCC = 0.9608 ] +INFO:root:Average loss=0.04101724828779697 +INFO:root: +[Valid][22][UTT][ MSE = 0.1756 | LCC = 0.9607 | SRCC = 0.9609 ] +INFO:root:Average loss=0.038743870839476585 +INFO:root: +[Valid][23][UTT][ MSE = 0.1803 | LCC = 0.9608 | SRCC = 0.9608 ] +INFO:root:Average loss=0.0367754320576787 +INFO:root: +[Valid][24][UTT][ MSE = 0.1860 | LCC = 0.9605 | SRCC = 0.9608 ] +INFO:root:Average loss=0.03815360900759697 +INFO:root: +[Valid][25][UTT][ MSE = 0.1909 | LCC = 0.9607 | SRCC = 0.9611 ] +INFO:root:Average loss=0.03445761893689633 +INFO:root: +[Valid][26][UTT][ MSE = 0.1790 | LCC = 0.9600 | SRCC = 0.9598 ] +INFO:root:Average loss=0.03321198364347219 +INFO:root: +[Valid][27][UTT][ MSE = 0.1765 | LCC = 0.9597 | SRCC = 0.9598 ] +INFO:root:Average loss=0.03283246059715748 +INFO:root: +[Valid][28][UTT][ MSE = 0.2039 | LCC = 0.9592 | SRCC = 0.9610 ] +INFO:root:Average loss=0.03216901668906212 +INFO:root: +[Valid][29][UTT][ MSE = 0.1721 | LCC = 0.9602 | SRCC = 0.9602 ] +INFO:root:Average loss=0.031386203214526176 +INFO:root: +[Valid][30][UTT][ MSE = 0.1966 | LCC = 0.9599 | SRCC = 0.9599 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1893 | LCC = 0.9624 | SRCC = 0.9626 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8a142d66573671d4c12dfa50caa52fd2afde585b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc337d84cd08c54268d048793ed026b82f1bc20a0c263e39b72f9a1e4355fe62 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7ef591719fe5cb415b7e136bafdc65014d8f1e13 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:032367e701a0662e600966bbe92308dfc408d6222eec1b43555d62c4c6eac0f3 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..75528ca44f61d45d76e6c57eab477665521041bd --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.30360024309158323 +INFO:root: +[Valid][0][UTT][ MSE = 0.2933 | LCC = 0.9385 | SRCC = 0.9389 ] +INFO:root:Average loss=0.1520119521021843 +INFO:root: +[Valid][1][UTT][ MSE = 0.2373 | LCC = 0.9476 | SRCC = 0.9488 ] +INFO:root:Average loss=0.13129016435146332 +INFO:root: +[Valid][2][UTT][ MSE = 0.2035 | LCC = 0.9525 | SRCC = 0.9537 ] +INFO:root:Average loss=0.12032624101638793 +INFO:root: +[Valid][3][UTT][ MSE = 0.1999 | LCC = 0.9539 | SRCC = 0.9556 ] +INFO:root:Average loss=0.10887437763810158 +INFO:root: +[Valid][4][UTT][ MSE = 0.2363 | LCC = 0.9540 | SRCC = 0.9555 ] +INFO:root:Average loss=0.10401451966166496 +INFO:root: +[Valid][5][UTT][ MSE = 0.2144 | LCC = 0.9563 | SRCC = 0.9578 ] +INFO:root:Average loss=0.09547675824165344 +INFO:root: +[Valid][6][UTT][ MSE = 0.2310 | LCC = 0.9541 | SRCC = 0.9569 ] +INFO:root:Average loss=0.0901850102543831 +INFO:root: +[Valid][7][UTT][ MSE = 0.1974 | LCC = 0.9573 | SRCC = 0.9586 ] +INFO:root:Average loss=0.08683418446779251 +INFO:root: +[Valid][8][UTT][ MSE = 0.2486 | LCC = 0.9559 | SRCC = 0.9585 ] +INFO:root:Average loss=0.08053539967536927 +INFO:root: +[Valid][9][UTT][ MSE = 0.2181 | LCC = 0.9579 | SRCC = 0.9598 ] +INFO:root:Average loss=0.07407044339179993 +INFO:root: +[Valid][10][UTT][ MSE = 0.2080 | LCC = 0.9579 | SRCC = 0.9582 ] +INFO:root:Average loss=0.07099031433463096 +INFO:root: +[Valid][11][UTT][ MSE = 0.2021 | LCC = 0.9580 | SRCC = 0.9585 ] +INFO:root:Average loss=0.06786281943321228 +INFO:root: +[Valid][12][UTT][ MSE = 0.2474 | LCC = 0.9498 | SRCC = 0.9507 ] +INFO:root:Average loss=0.06745550784468651 +INFO:root: +[Valid][13][UTT][ MSE = 0.2114 | LCC = 0.9564 | SRCC = 0.9560 ] +INFO:root:Average loss=0.060906559497118 +INFO:root: +[Valid][14][UTT][ MSE = 0.1935 | LCC = 0.9559 | SRCC = 0.9561 ] +INFO:root:Average loss=0.059003563493490216 +INFO:root: +[Valid][15][UTT][ MSE = 0.1993 | LCC = 0.9598 | SRCC = 0.9602 ] +INFO:root:Average loss=0.056502995163202285 +INFO:root: +[Valid][16][UTT][ MSE = 0.2195 | LCC = 0.9587 | SRCC = 0.9597 ] +INFO:root:Average loss=0.05284862950444222 +INFO:root: +[Valid][17][UTT][ MSE = 0.1861 | LCC = 0.9599 | SRCC = 0.9599 ] +INFO:root:Average loss=0.05009556309878826 +INFO:root: +[Valid][18][UTT][ MSE = 0.2140 | LCC = 0.9578 | SRCC = 0.9586 ] +INFO:root:Average loss=0.04860097803175449 +INFO:root: +[Valid][19][UTT][ MSE = 0.1979 | LCC = 0.9581 | SRCC = 0.9580 ] +INFO:root:Average loss=0.047395546734333036 +INFO:root: +[Valid][20][UTT][ MSE = 0.1944 | LCC = 0.9579 | SRCC = 0.9579 ] +INFO:root:Average loss=0.043817380771040915 +INFO:root: +[Valid][21][UTT][ MSE = 0.2193 | LCC = 0.9576 | SRCC = 0.9576 ] +INFO:root:Average loss=0.04257627759873867 +INFO:root: +[Valid][22][UTT][ MSE = 0.2030 | LCC = 0.9563 | SRCC = 0.9559 ] +INFO:root:Average loss=0.03984389521181583 +INFO:root: +[Valid][23][UTT][ MSE = 0.1970 | LCC = 0.9582 | SRCC = 0.9579 ] +INFO:root:Average loss=0.038699860513210296 +INFO:root: +[Valid][24][UTT][ MSE = 0.1958 | LCC = 0.9580 | SRCC = 0.9576 ] +INFO:root:Average loss=0.0371279361397028 +INFO:root: +[Valid][25][UTT][ MSE = 0.1889 | LCC = 0.9586 | SRCC = 0.9581 ] +INFO:root:Average loss=0.036144756942987445 +INFO:root: +[Valid][26][UTT][ MSE = 0.1817 | LCC = 0.9581 | SRCC = 0.9580 ] +INFO:root:Average loss=0.03440969733893871 +INFO:root: +[Valid][27][UTT][ MSE = 0.1859 | LCC = 0.9586 | SRCC = 0.9579 ] +INFO:root:Average loss=0.033863270953297615 +INFO:root: +[Valid][28][UTT][ MSE = 0.1977 | LCC = 0.9596 | SRCC = 0.9598 ] +INFO:root:Average loss=0.03136920571327209 +INFO:root: +[Valid][29][UTT][ MSE = 0.2154 | LCC = 0.9562 | SRCC = 0.9561 ] +INFO:root:Average loss=0.03183744175732136 +INFO:root: +[Valid][30][UTT][ MSE = 0.1907 | LCC = 0.9581 | SRCC = 0.9579 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1867 | LCC = 0.9619 | SRCC = 0.9628 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..87b202910bcada4b7b2096b3a83b52403524b826 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9c0787d87d3eccccf4efbe66ba6b55486b4e4f1318b5819008d43ddad60539e4 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e403426bfbc25e12fbe04862bfc47143956a3e5f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2fa77e8871b3d7f67954b32860cdb38af54d31ff46a241b144158a30a3146467 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0dba677cd9f92d2469ea25a5b586dcd791c1d500 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3490954117178917 +INFO:root: +[Valid][0][UTT][ MSE = 0.2801 | LCC = 0.9379 | SRCC = 0.9393 ] +INFO:root:Average loss=0.15184027123451233 +INFO:root: +[Valid][1][UTT][ MSE = 0.2856 | LCC = 0.9485 | SRCC = 0.9495 ] +INFO:root:Average loss=0.1303409726023674 +INFO:root: +[Valid][2][UTT][ MSE = 0.2198 | LCC = 0.9512 | SRCC = 0.9527 ] +INFO:root:Average loss=0.12031117498874665 +INFO:root: +[Valid][3][UTT][ MSE = 0.2948 | LCC = 0.9494 | SRCC = 0.9504 ] +INFO:root:Average loss=0.11299333894252778 +INFO:root: +[Valid][4][UTT][ MSE = 0.3144 | LCC = 0.9534 | SRCC = 0.9552 ] +INFO:root:Average loss=0.10194392436742783 +INFO:root: +[Valid][5][UTT][ MSE = 0.2185 | LCC = 0.9564 | SRCC = 0.9570 ] +INFO:root:Average loss=0.09444976916909217 +INFO:root: +[Valid][6][UTT][ MSE = 0.2157 | LCC = 0.9557 | SRCC = 0.9557 ] +INFO:root:Average loss=0.08982504537701606 +INFO:root: +[Valid][7][UTT][ MSE = 0.2269 | LCC = 0.9577 | SRCC = 0.9587 ] +INFO:root:Average loss=0.08272979420423508 +INFO:root: +[Valid][8][UTT][ MSE = 0.1991 | LCC = 0.9583 | SRCC = 0.9584 ] +INFO:root:Average loss=0.08048034906387329 +INFO:root: +[Valid][9][UTT][ MSE = 0.2151 | LCC = 0.9576 | SRCC = 0.9578 ] +INFO:root:Average loss=0.0740503724515438 +INFO:root: +[Valid][10][UTT][ MSE = 0.2214 | LCC = 0.9588 | SRCC = 0.9595 ] +INFO:root:Average loss=0.06980373081564903 +INFO:root: +[Valid][11][UTT][ MSE = 0.2171 | LCC = 0.9592 | SRCC = 0.9594 ] +INFO:root:Average loss=0.06364359536767006 +INFO:root: +[Valid][12][UTT][ MSE = 0.2192 | LCC = 0.9584 | SRCC = 0.9583 ] +INFO:root:Average loss=0.0634286650121212 +INFO:root: +[Valid][13][UTT][ MSE = 0.2090 | LCC = 0.9591 | SRCC = 0.9588 ] +INFO:root:Average loss=0.05953037196397781 +INFO:root: +[Valid][14][UTT][ MSE = 0.2175 | LCC = 0.9591 | SRCC = 0.9598 ] +INFO:root:Average loss=0.055343692749738695 +INFO:root: +[Valid][15][UTT][ MSE = 0.2394 | LCC = 0.9598 | SRCC = 0.9606 ] +INFO:root:Average loss=0.055121919065713884 +INFO:root: +[Valid][16][UTT][ MSE = 0.2109 | LCC = 0.9595 | SRCC = 0.9596 ] +INFO:root:Average loss=0.04996191003918648 +INFO:root: +[Valid][17][UTT][ MSE = 0.2182 | LCC = 0.9599 | SRCC = 0.9605 ] +INFO:root:Average loss=0.047590723797678945 +INFO:root: +[Valid][18][UTT][ MSE = 0.2387 | LCC = 0.9571 | SRCC = 0.9579 ] +INFO:root:Average loss=0.04507879278063774 +INFO:root: +[Valid][19][UTT][ MSE = 0.2166 | LCC = 0.9597 | SRCC = 0.9597 ] +INFO:root:Average loss=0.042524634554982184 +INFO:root: +[Valid][20][UTT][ MSE = 0.2133 | LCC = 0.9600 | SRCC = 0.9602 ] +INFO:root:Average loss=0.042242441698908804 +INFO:root: +[Valid][21][UTT][ MSE = 0.2206 | LCC = 0.9578 | SRCC = 0.9586 ] +INFO:root:Average loss=0.039637404948472975 +INFO:root: +[Valid][22][UTT][ MSE = 0.2407 | LCC = 0.9583 | SRCC = 0.9597 ] +INFO:root:Average loss=0.03739105635881424 +INFO:root: +[Valid][23][UTT][ MSE = 0.2279 | LCC = 0.9592 | SRCC = 0.9591 ] +INFO:root:Average loss=0.03617840723693371 +INFO:root: +[Valid][24][UTT][ MSE = 0.2219 | LCC = 0.9592 | SRCC = 0.9593 ] +INFO:root:Average loss=0.03571388091146946 +INFO:root: +[Valid][25][UTT][ MSE = 0.2208 | LCC = 0.9592 | SRCC = 0.9593 ] +INFO:root:Average loss=0.03496771539747715 +INFO:root: +[Valid][26][UTT][ MSE = 0.2487 | LCC = 0.9589 | SRCC = 0.9599 ] +INFO:root:Average loss=0.03302478235960007 +INFO:root: +[Valid][27][UTT][ MSE = 0.2218 | LCC = 0.9601 | SRCC = 0.9601 ] +INFO:root:Average loss=0.03201964601874351 +INFO:root: +[Valid][28][UTT][ MSE = 0.2150 | LCC = 0.9599 | SRCC = 0.9599 ] +INFO:root:Average loss=0.030538536846637727 +INFO:root: +[Valid][29][UTT][ MSE = 0.2029 | LCC = 0.9596 | SRCC = 0.9592 ] +INFO:root:Average loss=0.028853406891226768 +INFO:root: +[Valid][30][UTT][ MSE = 0.2122 | LCC = 0.9600 | SRCC = 0.9601 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2299 | LCC = 0.9610 | SRCC = 0.9622 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..19cf86a20ab82fe0c3f64bd940a0676adf585b91 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:79a9a47f5c463fac23de7de94dd2a2dd16ea61a29255d531526cc52e356741bd +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..202b684cb04c3e091081000f7268d9c8ebb53358 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e0a4f3b04edd750578e87dd8995035598c2dbaec3e7baae7eea19587240a4722 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..7050591a21142314953caf9660297d7d08e60e50 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.34652160757780076 +INFO:root: +[Valid][0][UTT][ MSE = 0.2644 | LCC = 0.9400 | SRCC = 0.9413 ] +INFO:root:Average loss=0.1562773073911667 +INFO:root: +[Valid][1][UTT][ MSE = 0.2775 | LCC = 0.9478 | SRCC = 0.9481 ] +INFO:root:Average loss=0.13257338660955428 +INFO:root: +[Valid][2][UTT][ MSE = 0.2191 | LCC = 0.9526 | SRCC = 0.9530 ] +INFO:root:Average loss=0.11995602878928184 +INFO:root: +[Valid][3][UTT][ MSE = 0.2228 | LCC = 0.9547 | SRCC = 0.9547 ] +INFO:root:Average loss=0.11188725903630256 +INFO:root: +[Valid][4][UTT][ MSE = 0.2054 | LCC = 0.9567 | SRCC = 0.9570 ] +INFO:root:Average loss=0.10273432970046997 +INFO:root: +[Valid][5][UTT][ MSE = 0.2129 | LCC = 0.9568 | SRCC = 0.9571 ] +INFO:root:Average loss=0.0957745446562767 +INFO:root: +[Valid][6][UTT][ MSE = 0.2109 | LCC = 0.9562 | SRCC = 0.9565 ] +INFO:root:Average loss=0.0921597076356411 +INFO:root: +[Valid][7][UTT][ MSE = 0.2015 | LCC = 0.9560 | SRCC = 0.9579 ] +INFO:root:Average loss=0.08479813358187675 +INFO:root: +[Valid][8][UTT][ MSE = 0.2079 | LCC = 0.9589 | SRCC = 0.9594 ] +INFO:root:Average loss=0.07779620549082757 +INFO:root: +[Valid][9][UTT][ MSE = 0.2020 | LCC = 0.9589 | SRCC = 0.9595 ] +INFO:root:Average loss=0.07507120206952095 +INFO:root: +[Valid][10][UTT][ MSE = 0.1820 | LCC = 0.9583 | SRCC = 0.9585 ] +INFO:root:Average loss=0.06891658115386963 +INFO:root: +[Valid][11][UTT][ MSE = 0.1841 | LCC = 0.9586 | SRCC = 0.9588 ] +INFO:root:Average loss=0.06492325684428216 +INFO:root: +[Valid][12][UTT][ MSE = 0.2204 | LCC = 0.9568 | SRCC = 0.9566 ] +INFO:root:Average loss=0.06176918968558311 +INFO:root: +[Valid][13][UTT][ MSE = 0.1953 | LCC = 0.9593 | SRCC = 0.9589 ] +INFO:root:Average loss=0.05730765509605408 +INFO:root: +[Valid][14][UTT][ MSE = 0.2384 | LCC = 0.9570 | SRCC = 0.9593 ] +INFO:root:Average loss=0.05585192322731018 +INFO:root: +[Valid][15][UTT][ MSE = 0.2136 | LCC = 0.9596 | SRCC = 0.9599 ] +INFO:root:Average loss=0.05236889110505581 +INFO:root: +[Valid][16][UTT][ MSE = 0.1902 | LCC = 0.9585 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05398068195581436 +INFO:root: +[Valid][17][UTT][ MSE = 0.1763 | LCC = 0.9607 | SRCC = 0.9603 ] +INFO:root:Average loss=0.05215719719231129 +INFO:root: +[Valid][18][UTT][ MSE = 0.2004 | LCC = 0.9596 | SRCC = 0.9589 ] +INFO:root:Average loss=0.045830249160528184 +INFO:root: +[Valid][19][UTT][ MSE = 0.1930 | LCC = 0.9608 | SRCC = 0.9609 ] +INFO:root:Average loss=0.04459753593802452 +INFO:root: +[Valid][20][UTT][ MSE = 0.1877 | LCC = 0.9593 | SRCC = 0.9592 ] +INFO:root:Average loss=0.042448630079627035 +INFO:root: +[Valid][21][UTT][ MSE = 0.2141 | LCC = 0.9585 | SRCC = 0.9599 ] +INFO:root:Average loss=0.042732096284627916 +INFO:root: +[Valid][22][UTT][ MSE = 0.1851 | LCC = 0.9601 | SRCC = 0.9597 ] +INFO:root:Average loss=0.03994652278721333 +INFO:root: +[Valid][23][UTT][ MSE = 0.1672 | LCC = 0.9606 | SRCC = 0.9603 ] +INFO:root:Average loss=0.036794823586940764 +INFO:root: +[Valid][24][UTT][ MSE = 0.1952 | LCC = 0.9594 | SRCC = 0.9593 ] +INFO:root:Average loss=0.03547748671472072 +INFO:root: +[Valid][25][UTT][ MSE = 0.1953 | LCC = 0.9597 | SRCC = 0.9590 ] +INFO:root:Average loss=0.034268740728497504 +INFO:root: +[Valid][26][UTT][ MSE = 0.1850 | LCC = 0.9601 | SRCC = 0.9598 ] +INFO:root:Average loss=0.03526370877027512 +INFO:root: +[Valid][27][UTT][ MSE = 0.1844 | LCC = 0.9602 | SRCC = 0.9599 ] +INFO:root:Average loss=0.033075018525123595 +INFO:root: +[Valid][28][UTT][ MSE = 0.2052 | LCC = 0.9606 | SRCC = 0.9606 ] +INFO:root:Average loss=0.03108982026576996 +INFO:root: +[Valid][29][UTT][ MSE = 0.1752 | LCC = 0.9607 | SRCC = 0.9603 ] +INFO:root:Average loss=0.030149685434997082 +INFO:root: +[Valid][30][UTT][ MSE = 0.1768 | LCC = 0.9600 | SRCC = 0.9596 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1922 | LCC = 0.9630 | SRCC = 0.9631 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b829862f4c80aaf56efa39771ff8f800f5bf9b66 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a9d7013e9e890097d75bc42ea5ba4ebc00bcf8f06bdd1c292a6a7a94a6044fbf +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f85a6e3bf6b20150e1d80a4e9b1a89f0492e02fc --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1762996a368b097163977ea8154e3812264e3d5940d6e6d59f1d52bd116bdfcf +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..fda8f4030dff6a04b5e1e364adb72da01320f2a5 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.36297896862030027 +INFO:root: +[Valid][0][UTT][ MSE = 0.2468 | LCC = 0.9414 | SRCC = 0.9431 ] +INFO:root:Average loss=0.15071746045351028 +INFO:root: +[Valid][1][UTT][ MSE = 0.2306 | LCC = 0.9515 | SRCC = 0.9523 ] +INFO:root:Average loss=0.13283072048425673 +INFO:root: +[Valid][2][UTT][ MSE = 0.2134 | LCC = 0.9533 | SRCC = 0.9539 ] +INFO:root:Average loss=0.116373430788517 +INFO:root: +[Valid][3][UTT][ MSE = 0.2249 | LCC = 0.9546 | SRCC = 0.9552 ] +INFO:root:Average loss=0.10793334218859672 +INFO:root: +[Valid][4][UTT][ MSE = 0.2547 | LCC = 0.9559 | SRCC = 0.9583 ] +INFO:root:Average loss=0.10074497884511947 +INFO:root: +[Valid][5][UTT][ MSE = 0.2425 | LCC = 0.9546 | SRCC = 0.9568 ] +INFO:root:Average loss=0.09407487779855728 +INFO:root: +[Valid][6][UTT][ MSE = 0.2137 | LCC = 0.9571 | SRCC = 0.9584 ] +INFO:root:Average loss=0.08821859750151634 +INFO:root: +[Valid][7][UTT][ MSE = 0.2049 | LCC = 0.9583 | SRCC = 0.9590 ] +INFO:root:Average loss=0.08248506367206573 +INFO:root: +[Valid][8][UTT][ MSE = 0.2022 | LCC = 0.9590 | SRCC = 0.9595 ] +INFO:root:Average loss=0.07889951032400132 +INFO:root: +[Valid][9][UTT][ MSE = 0.1909 | LCC = 0.9593 | SRCC = 0.9601 ] +INFO:root:Average loss=0.07489894524216652 +INFO:root: +[Valid][10][UTT][ MSE = 0.1749 | LCC = 0.9599 | SRCC = 0.9604 ] +INFO:root:Average loss=0.07117710047960281 +INFO:root: +[Valid][11][UTT][ MSE = 0.1947 | LCC = 0.9589 | SRCC = 0.9601 ] +INFO:root:Average loss=0.0649281147122383 +INFO:root: +[Valid][12][UTT][ MSE = 0.1992 | LCC = 0.9581 | SRCC = 0.9607 ] +INFO:root:Average loss=0.062373973533511165 +INFO:root: +[Valid][13][UTT][ MSE = 0.2090 | LCC = 0.9592 | SRCC = 0.9597 ] +INFO:root:Average loss=0.06292976140975952 +INFO:root: +[Valid][14][UTT][ MSE = 0.2156 | LCC = 0.9585 | SRCC = 0.9592 ] +INFO:root:Average loss=0.05485524463653564 +INFO:root: +[Valid][15][UTT][ MSE = 0.1847 | LCC = 0.9592 | SRCC = 0.9597 ] +INFO:root:Average loss=0.054311452627182004 +INFO:root: +[Valid][16][UTT][ MSE = 0.1939 | LCC = 0.9606 | SRCC = 0.9612 ] +INFO:root:Average loss=0.050597572669386866 +INFO:root: +[Valid][17][UTT][ MSE = 0.1885 | LCC = 0.9599 | SRCC = 0.9597 ] +INFO:root:Average loss=0.04929787132143974 +INFO:root: +[Valid][18][UTT][ MSE = 0.2309 | LCC = 0.9580 | SRCC = 0.9601 ] +INFO:root:Average loss=0.049284088224172594 +INFO:root: +[Valid][19][UTT][ MSE = 0.1838 | LCC = 0.9599 | SRCC = 0.9601 ] +INFO:root:Average loss=0.04488600336015224 +INFO:root: +[Valid][20][UTT][ MSE = 0.1845 | LCC = 0.9588 | SRCC = 0.9591 ] +INFO:root:Average loss=0.042219607040286065 +INFO:root: +[Valid][21][UTT][ MSE = 0.1822 | LCC = 0.9597 | SRCC = 0.9597 ] +INFO:root:Average loss=0.04125615657866001 +INFO:root: +[Valid][22][UTT][ MSE = 0.1962 | LCC = 0.9595 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04167598760128021 +INFO:root: +[Valid][23][UTT][ MSE = 0.1997 | LCC = 0.9585 | SRCC = 0.9588 ] +INFO:root:Average loss=0.03881159715354442 +INFO:root: +[Valid][24][UTT][ MSE = 0.1895 | LCC = 0.9597 | SRCC = 0.9607 ] +INFO:root:Average loss=0.03793550632894039 +INFO:root: +[Valid][25][UTT][ MSE = 0.1964 | LCC = 0.9571 | SRCC = 0.9568 ] +INFO:root:Average loss=0.03663704474270344 +INFO:root: +[Valid][26][UTT][ MSE = 0.2033 | LCC = 0.9594 | SRCC = 0.9593 ] +INFO:root:Average loss=0.036675572797656056 +INFO:root: +[Valid][27][UTT][ MSE = 0.2125 | LCC = 0.9593 | SRCC = 0.9604 ] +INFO:root:Average loss=0.034380490496754644 +INFO:root: +[Valid][28][UTT][ MSE = 0.1913 | LCC = 0.9593 | SRCC = 0.9595 ] +INFO:root:Average loss=0.032176850393414495 +INFO:root: +[Valid][29][UTT][ MSE = 0.2013 | LCC = 0.9590 | SRCC = 0.9593 ] +INFO:root:Average loss=0.030882035538554193 +INFO:root: +[Valid][30][UTT][ MSE = 0.1970 | LCC = 0.9589 | SRCC = 0.9595 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1969 | LCC = 0.9618 | SRCC = 0.9626 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4cd37be808cf617e25b03b35f45453d7dfb008e7 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3a0bf725d2f89ae7e117e4fed62f6f84c165f7895b2d7ed6b399e54754c22e9a +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d3f596db5308d7b7a31041d478674b9415825431 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:734256970ef01e82f884e45c1128c7ebb332335a13c4816d3bba6d531b5578cc +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0f727d5938b2316871fbb835b2a8bedea15019c3 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer4/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.341235601067543 +INFO:root: +[Valid][0][UTT][ MSE = 0.2674 | LCC = 0.9420 | SRCC = 0.9428 ] +INFO:root:Average loss=0.15052006393671036 +INFO:root: +[Valid][1][UTT][ MSE = 0.2015 | LCC = 0.9485 | SRCC = 0.9487 ] +INFO:root:Average loss=0.12818254417181016 +INFO:root: +[Valid][2][UTT][ MSE = 0.2207 | LCC = 0.9530 | SRCC = 0.9536 ] +INFO:root:Average loss=0.11378926020860672 +INFO:root: +[Valid][3][UTT][ MSE = 0.2267 | LCC = 0.9546 | SRCC = 0.9556 ] +INFO:root:Average loss=0.10910151767730712 +INFO:root: +[Valid][4][UTT][ MSE = 0.1965 | LCC = 0.9553 | SRCC = 0.9564 ] +INFO:root:Average loss=0.09981915268301964 +INFO:root: +[Valid][5][UTT][ MSE = 0.1879 | LCC = 0.9563 | SRCC = 0.9573 ] +INFO:root:Average loss=0.09362156230211258 +INFO:root: +[Valid][6][UTT][ MSE = 0.1932 | LCC = 0.9583 | SRCC = 0.9585 ] +INFO:root:Average loss=0.08771464106440544 +INFO:root: +[Valid][7][UTT][ MSE = 0.2006 | LCC = 0.9544 | SRCC = 0.9558 ] +INFO:root:Average loss=0.08241606491804122 +INFO:root: +[Valid][8][UTT][ MSE = 0.2007 | LCC = 0.9557 | SRCC = 0.9560 ] +INFO:root:Average loss=0.07820027750730514 +INFO:root: +[Valid][9][UTT][ MSE = 0.1981 | LCC = 0.9581 | SRCC = 0.9588 ] +INFO:root:Average loss=0.07197282168269158 +INFO:root: +[Valid][10][UTT][ MSE = 0.2154 | LCC = 0.9594 | SRCC = 0.9601 ] +INFO:root:Average loss=0.06961002510786056 +INFO:root: +[Valid][11][UTT][ MSE = 0.2104 | LCC = 0.9573 | SRCC = 0.9579 ] +INFO:root:Average loss=0.06706754219532013 +INFO:root: +[Valid][12][UTT][ MSE = 0.2008 | LCC = 0.9589 | SRCC = 0.9594 ] +INFO:root:Average loss=0.06203213295340538 +INFO:root: +[Valid][13][UTT][ MSE = 0.1915 | LCC = 0.9566 | SRCC = 0.9577 ] +INFO:root:Average loss=0.05797907593846321 +INFO:root: +[Valid][14][UTT][ MSE = 0.1979 | LCC = 0.9582 | SRCC = 0.9582 ] +INFO:root:Average loss=0.05619348853826523 +INFO:root: +[Valid][15][UTT][ MSE = 0.1849 | LCC = 0.9598 | SRCC = 0.9598 ] +INFO:root:Average loss=0.05528194597363472 +INFO:root: +[Valid][16][UTT][ MSE = 0.1970 | LCC = 0.9590 | SRCC = 0.9587 ] +INFO:root:Average loss=0.049988295286893844 +INFO:root: +[Valid][17][UTT][ MSE = 0.1846 | LCC = 0.9589 | SRCC = 0.9588 ] +INFO:root:Average loss=0.04977930746972561 +INFO:root: +[Valid][18][UTT][ MSE = 0.1902 | LCC = 0.9583 | SRCC = 0.9584 ] +INFO:root:Average loss=0.0464349847137928 +INFO:root: +[Valid][19][UTT][ MSE = 0.1894 | LCC = 0.9588 | SRCC = 0.9588 ] +INFO:root:Average loss=0.045088540911674496 +INFO:root: +[Valid][20][UTT][ MSE = 0.1886 | LCC = 0.9577 | SRCC = 0.9579 ] +INFO:root:Average loss=0.04283284294605255 +INFO:root: +[Valid][21][UTT][ MSE = 0.1929 | LCC = 0.9597 | SRCC = 0.9593 ] +INFO:root:Average loss=0.04080114966630936 +INFO:root: +[Valid][22][UTT][ MSE = 0.1828 | LCC = 0.9589 | SRCC = 0.9587 ] +INFO:root:Average loss=0.040459719702601434 +INFO:root: +[Valid][23][UTT][ MSE = 0.1709 | LCC = 0.9585 | SRCC = 0.9583 ] +INFO:root:Average loss=0.038415065869688986 +INFO:root: +[Valid][24][UTT][ MSE = 0.1981 | LCC = 0.9590 | SRCC = 0.9590 ] +INFO:root:Average loss=0.03655851484835148 +INFO:root: +[Valid][25][UTT][ MSE = 0.1829 | LCC = 0.9582 | SRCC = 0.9579 ] +INFO:root:Average loss=0.0352251258045435 +INFO:root: +[Valid][26][UTT][ MSE = 0.1967 | LCC = 0.9600 | SRCC = 0.9596 ] +INFO:root:Average loss=0.03346323011815548 +INFO:root: +[Valid][27][UTT][ MSE = 0.1862 | LCC = 0.9573 | SRCC = 0.9574 ] +INFO:root:Average loss=0.033777901709079745 +INFO:root: +[Valid][28][UTT][ MSE = 0.1762 | LCC = 0.9594 | SRCC = 0.9591 ] +INFO:root:Average loss=0.03139345297217369 +INFO:root: +[Valid][29][UTT][ MSE = 0.1836 | LCC = 0.9592 | SRCC = 0.9587 ] +INFO:root:Average loss=0.029340354457497598 +INFO:root: +[Valid][30][UTT][ MSE = 0.1920 | LCC = 0.9599 | SRCC = 0.9595 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2018 | LCC = 0.9607 | SRCC = 0.9616 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1c49fa83f112c326c84e7639e270125ba7445b25 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bc4d92be5ccb2e8b46c1c8ed96613b85ce95375369a9f2fab6740db224533738 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..90af6ade37d7c65f38c220493745eb0f0ae3e613 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:db1b7c4a60ef65a6c4483b4e52944ac2ebcda510251819ba6a40f726abe0507c +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b54a321a6acaca8b83e9e8e559ec352c83f1f8c5 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.32627365440130235 +INFO:root: +[Valid][0][UTT][ MSE = 0.2967 | LCC = 0.9438 | SRCC = 0.9433 ] +INFO:root:Average loss=0.1438968030810356 +INFO:root: +[Valid][1][UTT][ MSE = 0.2144 | LCC = 0.9505 | SRCC = 0.9507 ] +INFO:root:Average loss=0.12814959728717804 +INFO:root: +[Valid][2][UTT][ MSE = 0.2251 | LCC = 0.9522 | SRCC = 0.9528 ] +INFO:root:Average loss=0.11470777332782746 +INFO:root: +[Valid][3][UTT][ MSE = 0.2892 | LCC = 0.9524 | SRCC = 0.9541 ] +INFO:root:Average loss=0.10515965586900711 +INFO:root: +[Valid][4][UTT][ MSE = 0.2535 | LCC = 0.9546 | SRCC = 0.9556 ] +INFO:root:Average loss=0.10048742634057999 +INFO:root: +[Valid][5][UTT][ MSE = 0.2124 | LCC = 0.9560 | SRCC = 0.9556 ] +INFO:root:Average loss=0.09115569016337395 +INFO:root: +[Valid][6][UTT][ MSE = 0.2310 | LCC = 0.9557 | SRCC = 0.9567 ] +INFO:root:Average loss=0.08922135108709335 +INFO:root: +[Valid][7][UTT][ MSE = 0.2368 | LCC = 0.9564 | SRCC = 0.9568 ] +INFO:root:Average loss=0.08030216664075851 +INFO:root: +[Valid][8][UTT][ MSE = 0.1970 | LCC = 0.9565 | SRCC = 0.9577 ] +INFO:root:Average loss=0.07691459220647812 +INFO:root: +[Valid][9][UTT][ MSE = 0.2470 | LCC = 0.9546 | SRCC = 0.9561 ] +INFO:root:Average loss=0.07218553158640861 +INFO:root: +[Valid][10][UTT][ MSE = 0.2457 | LCC = 0.9550 | SRCC = 0.9564 ] +INFO:root:Average loss=0.06983205357193947 +INFO:root: +[Valid][11][UTT][ MSE = 0.2374 | LCC = 0.9569 | SRCC = 0.9575 ] +INFO:root:Average loss=0.06663227608799935 +INFO:root: +[Valid][12][UTT][ MSE = 0.2086 | LCC = 0.9581 | SRCC = 0.9583 ] +INFO:root:Average loss=0.06146671566367149 +INFO:root: +[Valid][13][UTT][ MSE = 0.2146 | LCC = 0.9564 | SRCC = 0.9565 ] +INFO:root:Average loss=0.05926440528035164 +INFO:root: +[Valid][14][UTT][ MSE = 0.1977 | LCC = 0.9588 | SRCC = 0.9593 ] +INFO:root:Average loss=0.054625833421945574 +INFO:root: +[Valid][15][UTT][ MSE = 0.2237 | LCC = 0.9576 | SRCC = 0.9578 ] +INFO:root:Average loss=0.053203936144709585 +INFO:root: +[Valid][16][UTT][ MSE = 0.2072 | LCC = 0.9577 | SRCC = 0.9576 ] +INFO:root:Average loss=0.04916460084915161 +INFO:root: +[Valid][17][UTT][ MSE = 0.2212 | LCC = 0.9579 | SRCC = 0.9587 ] +INFO:root:Average loss=0.04973052859306335 +INFO:root: +[Valid][18][UTT][ MSE = 0.2266 | LCC = 0.9588 | SRCC = 0.9594 ] +INFO:root:Average loss=0.04734068885445595 +INFO:root: +[Valid][19][UTT][ MSE = 0.2330 | LCC = 0.9567 | SRCC = 0.9569 ] +INFO:root:Average loss=0.04764915220439434 +INFO:root: +[Valid][20][UTT][ MSE = 0.2079 | LCC = 0.9564 | SRCC = 0.9574 ] +INFO:root:Average loss=0.04424103218317032 +INFO:root: +[Valid][21][UTT][ MSE = 0.1821 | LCC = 0.9587 | SRCC = 0.9581 ] +INFO:root:Average loss=0.04094070702791214 +INFO:root: +[Valid][22][UTT][ MSE = 0.2403 | LCC = 0.9553 | SRCC = 0.9561 ] +INFO:root:Average loss=0.040149343237280845 +INFO:root: +[Valid][23][UTT][ MSE = 0.2131 | LCC = 0.9584 | SRCC = 0.9582 ] +INFO:root:Average loss=0.03650422964990139 +INFO:root: +[Valid][24][UTT][ MSE = 0.1937 | LCC = 0.9576 | SRCC = 0.9574 ] +INFO:root:Average loss=0.03705155239999294 +INFO:root: +[Valid][25][UTT][ MSE = 0.2063 | LCC = 0.9583 | SRCC = 0.9583 ] +INFO:root:Average loss=0.03508858525753021 +INFO:root: +[Valid][26][UTT][ MSE = 0.1961 | LCC = 0.9583 | SRCC = 0.9585 ] +INFO:root:Average loss=0.03589435185492039 +INFO:root: +[Valid][27][UTT][ MSE = 0.2214 | LCC = 0.9571 | SRCC = 0.9573 ] +INFO:root:Average loss=0.032194247633218764 +INFO:root: +[Valid][28][UTT][ MSE = 0.2009 | LCC = 0.9579 | SRCC = 0.9580 ] +INFO:root:Average loss=0.03200893084704876 +INFO:root: +[Valid][29][UTT][ MSE = 0.2027 | LCC = 0.9558 | SRCC = 0.9550 ] +INFO:root:Average loss=0.030707904085516928 +INFO:root: +[Valid][30][UTT][ MSE = 0.1988 | LCC = 0.9584 | SRCC = 0.9581 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2214 | LCC = 0.9612 | SRCC = 0.9622 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b2e7e37e86ee1735a886763e77a2c1e324f8d48e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1ee6d8b53de60a82cbe30d91a01d8e5d0db22030d7ab7ec3f86b528b0485c645 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0ea813e4643a2b5ff45f7aeb878c55c710e7ddd7 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5996eb1fe0e00202da1a5126d81a79e450f9bb31612d4d9b151f64b2f96918d4 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9f8a947b0fc5ab0291bd480cf7f3b760d50ec965 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4051126863360405 +INFO:root: +[Valid][0][UTT][ MSE = 0.2536 | LCC = 0.9374 | SRCC = 0.9418 ] +INFO:root:Average loss=0.1481834117770195 +INFO:root: +[Valid][1][UTT][ MSE = 0.2236 | LCC = 0.9513 | SRCC = 0.9523 ] +INFO:root:Average loss=0.12796098119020463 +INFO:root: +[Valid][2][UTT][ MSE = 0.2539 | LCC = 0.9521 | SRCC = 0.9523 ] +INFO:root:Average loss=0.11835080885887146 +INFO:root: +[Valid][3][UTT][ MSE = 0.2172 | LCC = 0.9566 | SRCC = 0.9572 ] +INFO:root:Average loss=0.10557284933328628 +INFO:root: +[Valid][4][UTT][ MSE = 0.2187 | LCC = 0.9567 | SRCC = 0.9567 ] +INFO:root:Average loss=0.10022731786966324 +INFO:root: +[Valid][5][UTT][ MSE = 0.2133 | LCC = 0.9563 | SRCC = 0.9582 ] +INFO:root:Average loss=0.09152496647834778 +INFO:root: +[Valid][6][UTT][ MSE = 0.2319 | LCC = 0.9557 | SRCC = 0.9567 ] +INFO:root:Average loss=0.08342859449982643 +INFO:root: +[Valid][7][UTT][ MSE = 0.2210 | LCC = 0.9584 | SRCC = 0.9593 ] +INFO:root:Average loss=0.08231085339188576 +INFO:root: +[Valid][8][UTT][ MSE = 0.2040 | LCC = 0.9564 | SRCC = 0.9570 ] +INFO:root:Average loss=0.07753730201721191 +INFO:root: +[Valid][9][UTT][ MSE = 0.2114 | LCC = 0.9580 | SRCC = 0.9591 ] +INFO:root:Average loss=0.0750473027229309 +INFO:root: +[Valid][10][UTT][ MSE = 0.2095 | LCC = 0.9588 | SRCC = 0.9595 ] +INFO:root:Average loss=0.06672670203447342 +INFO:root: +[Valid][11][UTT][ MSE = 0.1898 | LCC = 0.9563 | SRCC = 0.9570 ] +INFO:root:Average loss=0.0660080473124981 +INFO:root: +[Valid][12][UTT][ MSE = 0.2096 | LCC = 0.9580 | SRCC = 0.9584 ] +INFO:root:Average loss=0.05846017771959305 +INFO:root: +[Valid][13][UTT][ MSE = 0.1891 | LCC = 0.9594 | SRCC = 0.9602 ] +INFO:root:Average loss=0.05881543046236038 +INFO:root: +[Valid][14][UTT][ MSE = 0.1926 | LCC = 0.9576 | SRCC = 0.9578 ] +INFO:root:Average loss=0.05514155027270317 +INFO:root: +[Valid][15][UTT][ MSE = 0.2053 | LCC = 0.9578 | SRCC = 0.9592 ] +INFO:root:Average loss=0.051235197111964226 +INFO:root: +[Valid][16][UTT][ MSE = 0.2215 | LCC = 0.9572 | SRCC = 0.9574 ] +INFO:root:Average loss=0.04915694454312324 +INFO:root: +[Valid][17][UTT][ MSE = 0.2061 | LCC = 0.9575 | SRCC = 0.9592 ] +INFO:root:Average loss=0.047320010930299757 +INFO:root: +[Valid][18][UTT][ MSE = 0.1742 | LCC = 0.9573 | SRCC = 0.9577 ] +INFO:root:Average loss=0.04473039335012436 +INFO:root: +[Valid][19][UTT][ MSE = 0.2022 | LCC = 0.9589 | SRCC = 0.9590 ] +INFO:root:Average loss=0.043510037079453466 +INFO:root: +[Valid][20][UTT][ MSE = 0.1944 | LCC = 0.9593 | SRCC = 0.9597 ] +INFO:root:Average loss=0.04108585818111896 +INFO:root: +[Valid][21][UTT][ MSE = 0.1933 | LCC = 0.9598 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04140052849054337 +INFO:root: +[Valid][22][UTT][ MSE = 0.1978 | LCC = 0.9608 | SRCC = 0.9608 ] +INFO:root:Average loss=0.03621869860589504 +INFO:root: +[Valid][23][UTT][ MSE = 0.1946 | LCC = 0.9583 | SRCC = 0.9586 ] +INFO:root:Average loss=0.03659888008236885 +INFO:root: +[Valid][24][UTT][ MSE = 0.1993 | LCC = 0.9576 | SRCC = 0.9576 ] +INFO:root:Average loss=0.036180024430155756 +INFO:root: +[Valid][25][UTT][ MSE = 0.2016 | LCC = 0.9592 | SRCC = 0.9590 ] +INFO:root:Average loss=0.033362088203430174 +INFO:root: +[Valid][26][UTT][ MSE = 0.1935 | LCC = 0.9590 | SRCC = 0.9586 ] +INFO:root:Average loss=0.031403139531612397 +INFO:root: +[Valid][27][UTT][ MSE = 0.1813 | LCC = 0.9577 | SRCC = 0.9577 ] +INFO:root:Average loss=0.03149606953561306 +INFO:root: +[Valid][28][UTT][ MSE = 0.1894 | LCC = 0.9590 | SRCC = 0.9590 ] +INFO:root:Average loss=0.03030565136671066 +INFO:root: +[Valid][29][UTT][ MSE = 0.2032 | LCC = 0.9577 | SRCC = 0.9577 ] +INFO:root:Average loss=0.027772716000676156 +INFO:root: +[Valid][30][UTT][ MSE = 0.2039 | LCC = 0.9577 | SRCC = 0.9574 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1964 | LCC = 0.9607 | SRCC = 0.9614 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1c1960c430200f2fc8ea34257f7aadf07f897113 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f65962030d40a1ac39aea12bc2028eae2a3e00650d098c1beacaf2f371cd0d1f +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..02b2715a08850fe847be5f5aca7e355338772591 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6d33173a47445ebdb277a218d24800a47880c7096d01bc6ba4272a1b7eba0d17 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2ee97d0546e8196e3359f5f55da1cbd947145229 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.33673186337947847 +INFO:root: +[Valid][0][UTT][ MSE = 0.3184 | LCC = 0.9395 | SRCC = 0.9383 ] +INFO:root:Average loss=0.1506358675956726 +INFO:root: +[Valid][1][UTT][ MSE = 0.2325 | LCC = 0.9478 | SRCC = 0.9471 ] +INFO:root:Average loss=0.12987169128656387 +INFO:root: +[Valid][2][UTT][ MSE = 0.2522 | LCC = 0.9488 | SRCC = 0.9486 ] +INFO:root:Average loss=0.11454474103450775 +INFO:root: +[Valid][3][UTT][ MSE = 0.2543 | LCC = 0.9525 | SRCC = 0.9539 ] +INFO:root:Average loss=0.10610054954886436 +INFO:root: +[Valid][4][UTT][ MSE = 0.2664 | LCC = 0.9510 | SRCC = 0.9514 ] +INFO:root:Average loss=0.09855341517925262 +INFO:root: +[Valid][5][UTT][ MSE = 0.1908 | LCC = 0.9549 | SRCC = 0.9556 ] +INFO:root:Average loss=0.09660678800940514 +INFO:root: +[Valid][6][UTT][ MSE = 0.2075 | LCC = 0.9559 | SRCC = 0.9555 ] +INFO:root:Average loss=0.0867241505086422 +INFO:root: +[Valid][7][UTT][ MSE = 0.2369 | LCC = 0.9572 | SRCC = 0.9575 ] +INFO:root:Average loss=0.07958874437212944 +INFO:root: +[Valid][8][UTT][ MSE = 0.2029 | LCC = 0.9561 | SRCC = 0.9561 ] +INFO:root:Average loss=0.07599846667051316 +INFO:root: +[Valid][9][UTT][ MSE = 0.2066 | LCC = 0.9571 | SRCC = 0.9572 ] +INFO:root:Average loss=0.07194361928105354 +INFO:root: +[Valid][10][UTT][ MSE = 0.2270 | LCC = 0.9569 | SRCC = 0.9573 ] +INFO:root:Average loss=0.06855438029766082 +INFO:root: +[Valid][11][UTT][ MSE = 0.2152 | LCC = 0.9576 | SRCC = 0.9576 ] +INFO:root:Average loss=0.0655221548974514 +INFO:root: +[Valid][12][UTT][ MSE = 0.2516 | LCC = 0.9559 | SRCC = 0.9573 ] +INFO:root:Average loss=0.06461082628369332 +INFO:root: +[Valid][13][UTT][ MSE = 0.2171 | LCC = 0.9576 | SRCC = 0.9579 ] +INFO:root:Average loss=0.05696059456467629 +INFO:root: +[Valid][14][UTT][ MSE = 0.2022 | LCC = 0.9570 | SRCC = 0.9573 ] +INFO:root:Average loss=0.055527139276266096 +INFO:root: +[Valid][15][UTT][ MSE = 0.2096 | LCC = 0.9578 | SRCC = 0.9579 ] +INFO:root:Average loss=0.05171019223332405 +INFO:root: +[Valid][16][UTT][ MSE = 0.2206 | LCC = 0.9569 | SRCC = 0.9571 ] +INFO:root:Average loss=0.05254049725830555 +INFO:root: +[Valid][17][UTT][ MSE = 0.2053 | LCC = 0.9577 | SRCC = 0.9581 ] +INFO:root:Average loss=0.049200872480869295 +INFO:root: +[Valid][18][UTT][ MSE = 0.2406 | LCC = 0.9571 | SRCC = 0.9575 ] +INFO:root:Average loss=0.04593975773453712 +INFO:root: +[Valid][19][UTT][ MSE = 0.2227 | LCC = 0.9571 | SRCC = 0.9578 ] +INFO:root:Average loss=0.0422774047255516 +INFO:root: +[Valid][20][UTT][ MSE = 0.1976 | LCC = 0.9580 | SRCC = 0.9584 ] +INFO:root:Average loss=0.04232733392715454 +INFO:root: +[Valid][21][UTT][ MSE = 0.2098 | LCC = 0.9567 | SRCC = 0.9570 ] +INFO:root:Average loss=0.04193034461140633 +INFO:root: +[Valid][22][UTT][ MSE = 0.2026 | LCC = 0.9574 | SRCC = 0.9568 ] +INFO:root:Average loss=0.040354332819581035 +INFO:root: +[Valid][23][UTT][ MSE = 0.2347 | LCC = 0.9553 | SRCC = 0.9550 ] +INFO:root:Average loss=0.03707298173010349 +INFO:root: +[Valid][24][UTT][ MSE = 0.2305 | LCC = 0.9546 | SRCC = 0.9548 ] +INFO:root:Average loss=0.03574846038222313 +INFO:root: +[Valid][25][UTT][ MSE = 0.1961 | LCC = 0.9587 | SRCC = 0.9584 ] +INFO:root:Average loss=0.03558618815243244 +INFO:root: +[Valid][26][UTT][ MSE = 0.2212 | LCC = 0.9568 | SRCC = 0.9566 ] +INFO:root:Average loss=0.03479737015068531 +INFO:root: +[Valid][27][UTT][ MSE = 0.2336 | LCC = 0.9572 | SRCC = 0.9578 ] +INFO:root:Average loss=0.031972869098186495 +INFO:root: +[Valid][28][UTT][ MSE = 0.2170 | LCC = 0.9577 | SRCC = 0.9580 ] +INFO:root:Average loss=0.030352113276720046 +INFO:root: +[Valid][29][UTT][ MSE = 0.1939 | LCC = 0.9566 | SRCC = 0.9569 ] +INFO:root:Average loss=0.03037506926059723 +INFO:root: +[Valid][30][UTT][ MSE = 0.2195 | LCC = 0.9569 | SRCC = 0.9568 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1919 | LCC = 0.9611 | SRCC = 0.9613 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..107f571e6ee758b001a8a25e14e744cdcbff7267 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7624a01de347cd5fc5fa0106d63798009b3c444c66648bb1949e9be1f21cbddc +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..81ed65398c56f9ed24f0a7edc41373b66b8278aa --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ead3d37e1bfa7ae30e5bd6d43d1d5fe5d1bc1a219ddb3833967321c8d865ad74 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..39e94e9c02a90af5bc5a2bf78b1c1c26cda0475a --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3193078647851944 +INFO:root: +[Valid][0][UTT][ MSE = 0.2865 | LCC = 0.9381 | SRCC = 0.9392 ] +INFO:root:Average loss=0.14614689821004867 +INFO:root: +[Valid][1][UTT][ MSE = 0.2371 | LCC = 0.9478 | SRCC = 0.9506 ] +INFO:root:Average loss=0.1334929983615875 +INFO:root: +[Valid][2][UTT][ MSE = 0.2537 | LCC = 0.9521 | SRCC = 0.9531 ] +INFO:root:Average loss=0.11957506197690963 +INFO:root: +[Valid][3][UTT][ MSE = 0.2576 | LCC = 0.9529 | SRCC = 0.9540 ] +INFO:root:Average loss=0.10773164516687393 +INFO:root: +[Valid][4][UTT][ MSE = 0.2403 | LCC = 0.9545 | SRCC = 0.9562 ] +INFO:root:Average loss=0.09957037705183029 +INFO:root: +[Valid][5][UTT][ MSE = 0.2275 | LCC = 0.9556 | SRCC = 0.9570 ] +INFO:root:Average loss=0.09070865687727928 +INFO:root: +[Valid][6][UTT][ MSE = 0.2234 | LCC = 0.9556 | SRCC = 0.9564 ] +INFO:root:Average loss=0.08845935645699501 +INFO:root: +[Valid][7][UTT][ MSE = 0.2099 | LCC = 0.9555 | SRCC = 0.9569 ] +INFO:root:Average loss=0.08173885643482208 +INFO:root: +[Valid][8][UTT][ MSE = 0.2247 | LCC = 0.9566 | SRCC = 0.9579 ] +INFO:root:Average loss=0.07597270992398263 +INFO:root: +[Valid][9][UTT][ MSE = 0.2112 | LCC = 0.9562 | SRCC = 0.9575 ] +INFO:root:Average loss=0.07383167952299118 +INFO:root: +[Valid][10][UTT][ MSE = 0.2154 | LCC = 0.9556 | SRCC = 0.9568 ] +INFO:root:Average loss=0.07013013938069343 +INFO:root: +[Valid][11][UTT][ MSE = 0.1998 | LCC = 0.9551 | SRCC = 0.9568 ] +INFO:root:Average loss=0.06781005045771599 +INFO:root: +[Valid][12][UTT][ MSE = 0.1972 | LCC = 0.9517 | SRCC = 0.9563 ] +INFO:root:Average loss=0.06531688937544823 +INFO:root: +[Valid][13][UTT][ MSE = 0.1922 | LCC = 0.9534 | SRCC = 0.9537 ] +INFO:root:Average loss=0.0609672309756279 +INFO:root: +[Valid][14][UTT][ MSE = 0.2447 | LCC = 0.9561 | SRCC = 0.9567 ] +INFO:root:Average loss=0.05498715135455132 +INFO:root: +[Valid][15][UTT][ MSE = 0.2269 | LCC = 0.9547 | SRCC = 0.9567 ] +INFO:root:Average loss=0.054558214545249936 +INFO:root: +[Valid][16][UTT][ MSE = 0.2325 | LCC = 0.9556 | SRCC = 0.9563 ] +INFO:root:Average loss=0.05244516535103321 +INFO:root: +[Valid][17][UTT][ MSE = 0.2179 | LCC = 0.9549 | SRCC = 0.9553 ] +INFO:root:Average loss=0.04686738005280495 +INFO:root: +[Valid][18][UTT][ MSE = 0.2122 | LCC = 0.9544 | SRCC = 0.9560 ] +INFO:root:Average loss=0.046296169757843016 +INFO:root: +[Valid][19][UTT][ MSE = 0.2124 | LCC = 0.9559 | SRCC = 0.9562 ] +INFO:root:Average loss=0.0456297754496336 +INFO:root: +[Valid][20][UTT][ MSE = 0.2038 | LCC = 0.9551 | SRCC = 0.9560 ] +INFO:root:Average loss=0.04617303939163685 +INFO:root: +[Valid][21][UTT][ MSE = 0.2299 | LCC = 0.9544 | SRCC = 0.9546 ] +INFO:root:Average loss=0.04108772724866867 +INFO:root: +[Valid][22][UTT][ MSE = 0.2008 | LCC = 0.9551 | SRCC = 0.9560 ] +INFO:root:Average loss=0.039033295229077336 +INFO:root: +[Valid][23][UTT][ MSE = 0.2176 | LCC = 0.9554 | SRCC = 0.9557 ] +INFO:root:Average loss=0.03910186928510666 +INFO:root: +[Valid][24][UTT][ MSE = 0.2065 | LCC = 0.9548 | SRCC = 0.9560 ] +INFO:root:Average loss=0.03713284358382225 +INFO:root: +[Valid][25][UTT][ MSE = 0.2176 | LCC = 0.9531 | SRCC = 0.9532 ] +INFO:root:Average loss=0.03573847168684006 +INFO:root: +[Valid][26][UTT][ MSE = 0.2110 | LCC = 0.9551 | SRCC = 0.9556 ] +INFO:root:Average loss=0.03419796559214592 +INFO:root: +[Valid][27][UTT][ MSE = 0.1972 | LCC = 0.9544 | SRCC = 0.9550 ] +INFO:root:Average loss=0.0322996196448803 +INFO:root: +[Valid][28][UTT][ MSE = 0.2103 | LCC = 0.9546 | SRCC = 0.9555 ] +INFO:root:Average loss=0.030879612512886526 +INFO:root: +[Valid][29][UTT][ MSE = 0.2101 | LCC = 0.9544 | SRCC = 0.9550 ] +INFO:root:Average loss=0.030267291814088822 +INFO:root: +[Valid][30][UTT][ MSE = 0.2136 | LCC = 0.9540 | SRCC = 0.9544 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2200 | LCC = 0.9618 | SRCC = 0.9631 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..62f306e03f54e4dfe3da820796707bad617bef2a --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2dcc6addfe7fa74b04ca520b8e3e68e235b405c2f609aa0b84913c6f92e659e9 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b4336a62903139b12cc7fa5ab0d8daefed2587ba --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4c9eb2323c87e5c8ddb56028d3e9da400046b7ca87dafe45dae78722c1b0433d +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d87d011b116cddd0a4c91c58497ba229fe005865 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer5/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.35047619104385375 +INFO:root: +[Valid][0][UTT][ MSE = 0.2612 | LCC = 0.9422 | SRCC = 0.9431 ] +INFO:root:Average loss=0.14705853134393693 +INFO:root: +[Valid][1][UTT][ MSE = 0.2285 | LCC = 0.9494 | SRCC = 0.9505 ] +INFO:root:Average loss=0.1266117984354496 +INFO:root: +[Valid][2][UTT][ MSE = 0.2612 | LCC = 0.9532 | SRCC = 0.9540 ] +INFO:root:Average loss=0.12042512983083725 +INFO:root: +[Valid][3][UTT][ MSE = 0.2499 | LCC = 0.9548 | SRCC = 0.9554 ] +INFO:root:Average loss=0.11004531455039979 +INFO:root: +[Valid][4][UTT][ MSE = 0.1921 | LCC = 0.9548 | SRCC = 0.9554 ] +INFO:root:Average loss=0.09878658884763718 +INFO:root: +[Valid][5][UTT][ MSE = 0.2000 | LCC = 0.9558 | SRCC = 0.9561 ] +INFO:root:Average loss=0.09282484304904938 +INFO:root: +[Valid][6][UTT][ MSE = 0.2033 | LCC = 0.9569 | SRCC = 0.9577 ] +INFO:root:Average loss=0.08763456678390504 +INFO:root: +[Valid][7][UTT][ MSE = 0.2224 | LCC = 0.9548 | SRCC = 0.9552 ] +INFO:root:Average loss=0.08199020057916641 +INFO:root: +[Valid][8][UTT][ MSE = 0.2191 | LCC = 0.9562 | SRCC = 0.9562 ] +INFO:root:Average loss=0.0825715366601944 +INFO:root: +[Valid][9][UTT][ MSE = 0.2033 | LCC = 0.9559 | SRCC = 0.9572 ] +INFO:root:Average loss=0.07397867700457574 +INFO:root: +[Valid][10][UTT][ MSE = 0.2451 | LCC = 0.9552 | SRCC = 0.9554 ] +INFO:root:Average loss=0.07229868525266647 +INFO:root: +[Valid][11][UTT][ MSE = 0.1851 | LCC = 0.9555 | SRCC = 0.9560 ] +INFO:root:Average loss=0.06663109788298607 +INFO:root: +[Valid][12][UTT][ MSE = 0.2183 | LCC = 0.9569 | SRCC = 0.9569 ] +INFO:root:Average loss=0.062111250549554825 +INFO:root: +[Valid][13][UTT][ MSE = 0.2247 | LCC = 0.9560 | SRCC = 0.9558 ] +INFO:root:Average loss=0.059780418276786804 +INFO:root: +[Valid][14][UTT][ MSE = 0.2094 | LCC = 0.9563 | SRCC = 0.9569 ] +INFO:root:Average loss=0.05649308413267136 +INFO:root: +[Valid][15][UTT][ MSE = 0.2036 | LCC = 0.9565 | SRCC = 0.9566 ] +INFO:root:Average loss=0.05402046567201614 +INFO:root: +[Valid][16][UTT][ MSE = 0.2198 | LCC = 0.9556 | SRCC = 0.9558 ] +INFO:root:Average loss=0.04929806555807591 +INFO:root: +[Valid][17][UTT][ MSE = 0.1994 | LCC = 0.9565 | SRCC = 0.9568 ] +INFO:root:Average loss=0.049537268459796904 +INFO:root: +[Valid][18][UTT][ MSE = 0.2075 | LCC = 0.9560 | SRCC = 0.9561 ] +INFO:root:Average loss=0.04567935809493065 +INFO:root: +[Valid][19][UTT][ MSE = 0.1866 | LCC = 0.9563 | SRCC = 0.9564 ] +INFO:root:Average loss=0.04524998800456524 +INFO:root: +[Valid][20][UTT][ MSE = 0.1850 | LCC = 0.9564 | SRCC = 0.9561 ] +INFO:root:Average loss=0.04214710658788681 +INFO:root: +[Valid][21][UTT][ MSE = 0.1921 | LCC = 0.9562 | SRCC = 0.9563 ] +INFO:root:Average loss=0.04413691847026348 +INFO:root: +[Valid][22][UTT][ MSE = 0.1952 | LCC = 0.9564 | SRCC = 0.9564 ] +INFO:root:Average loss=0.03876747554540634 +INFO:root: +[Valid][23][UTT][ MSE = 0.1839 | LCC = 0.9568 | SRCC = 0.9564 ] +INFO:root:Average loss=0.0372189384996891 +INFO:root: +[Valid][24][UTT][ MSE = 0.2029 | LCC = 0.9567 | SRCC = 0.9570 ] +INFO:root:Average loss=0.03628586407005787 +INFO:root: +[Valid][25][UTT][ MSE = 0.1904 | LCC = 0.9560 | SRCC = 0.9559 ] +INFO:root:Average loss=0.03566468881070614 +INFO:root: +[Valid][26][UTT][ MSE = 0.1998 | LCC = 0.9562 | SRCC = 0.9561 ] +INFO:root:Average loss=0.03462847128510475 +INFO:root: +[Valid][27][UTT][ MSE = 0.1914 | LCC = 0.9561 | SRCC = 0.9558 ] +INFO:root:Average loss=0.03148554496467113 +INFO:root: +[Valid][28][UTT][ MSE = 0.1996 | LCC = 0.9564 | SRCC = 0.9556 ] +INFO:root:Average loss=0.030584373503923415 +INFO:root: +[Valid][29][UTT][ MSE = 0.1731 | LCC = 0.9548 | SRCC = 0.9553 ] +INFO:root:Average loss=0.02950389391183853 +INFO:root: +[Valid][30][UTT][ MSE = 0.1907 | LCC = 0.9557 | SRCC = 0.9554 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2010 | LCC = 0.9602 | SRCC = 0.9614 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ddbd42493e7b5d85a3cfd08edc8ec460b3ae3eb0 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:14cd860a86f4881df54b55a85cab5125c4992efb72ca6b3be2eabe6273cc43b6 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a8ea471a0398ff858c9ebbde899a3557a5a1703c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a30dc0867b1a06f79a6ecb4a3cbebf9e2b85031093cf014cec1aedc8ab0abe25 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..7885ceab458936f925f1b68f2f0790840550acc8 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.36164612346887587 +INFO:root: +[Valid][0][UTT][ MSE = 0.2785 | LCC = 0.9433 | SRCC = 0.9436 ] +INFO:root:Average loss=0.147511084318161 +INFO:root: +[Valid][1][UTT][ MSE = 0.2656 | LCC = 0.9483 | SRCC = 0.9496 ] +INFO:root:Average loss=0.13036566054821014 +INFO:root: +[Valid][2][UTT][ MSE = 0.2295 | LCC = 0.9526 | SRCC = 0.9536 ] +INFO:root:Average loss=0.11707021763920784 +INFO:root: +[Valid][3][UTT][ MSE = 0.2550 | LCC = 0.9547 | SRCC = 0.9553 ] +INFO:root:Average loss=0.10638764390349388 +INFO:root: +[Valid][4][UTT][ MSE = 0.2285 | LCC = 0.9565 | SRCC = 0.9568 ] +INFO:root:Average loss=0.09995857152342796 +INFO:root: +[Valid][5][UTT][ MSE = 0.2300 | LCC = 0.9537 | SRCC = 0.9548 ] +INFO:root:Average loss=0.09270221146941185 +INFO:root: +[Valid][6][UTT][ MSE = 0.2401 | LCC = 0.9570 | SRCC = 0.9580 ] +INFO:root:Average loss=0.08590665489435195 +INFO:root: +[Valid][7][UTT][ MSE = 0.2086 | LCC = 0.9550 | SRCC = 0.9572 ] +INFO:root:Average loss=0.08175588181614876 +INFO:root: +[Valid][8][UTT][ MSE = 0.2088 | LCC = 0.9575 | SRCC = 0.9581 ] +INFO:root:Average loss=0.07567291074991227 +INFO:root: +[Valid][9][UTT][ MSE = 0.2082 | LCC = 0.9574 | SRCC = 0.9571 ] +INFO:root:Average loss=0.07044751992821693 +INFO:root: +[Valid][10][UTT][ MSE = 0.2162 | LCC = 0.9587 | SRCC = 0.9593 ] +INFO:root:Average loss=0.06891674128174782 +INFO:root: +[Valid][11][UTT][ MSE = 0.2159 | LCC = 0.9589 | SRCC = 0.9588 ] +INFO:root:Average loss=0.06451411470770836 +INFO:root: +[Valid][12][UTT][ MSE = 0.2346 | LCC = 0.9575 | SRCC = 0.9586 ] +INFO:root:Average loss=0.06208140751719475 +INFO:root: +[Valid][13][UTT][ MSE = 0.1953 | LCC = 0.9579 | SRCC = 0.9586 ] +INFO:root:Average loss=0.0580930435359478 +INFO:root: +[Valid][14][UTT][ MSE = 0.2122 | LCC = 0.9588 | SRCC = 0.9585 ] +INFO:root:Average loss=0.0559641110599041 +INFO:root: +[Valid][15][UTT][ MSE = 0.2001 | LCC = 0.9550 | SRCC = 0.9571 ] +INFO:root:Average loss=0.05231441086530685 +INFO:root: +[Valid][16][UTT][ MSE = 0.2300 | LCC = 0.9552 | SRCC = 0.9560 ] +INFO:root:Average loss=0.050842049196362496 +INFO:root: +[Valid][17][UTT][ MSE = 0.2172 | LCC = 0.9580 | SRCC = 0.9587 ] +INFO:root:Average loss=0.04735422511398792 +INFO:root: +[Valid][18][UTT][ MSE = 0.1998 | LCC = 0.9571 | SRCC = 0.9577 ] +INFO:root:Average loss=0.04396275317668915 +INFO:root: +[Valid][19][UTT][ MSE = 0.1931 | LCC = 0.9575 | SRCC = 0.9576 ] +INFO:root:Average loss=0.04588114148378372 +INFO:root: +[Valid][20][UTT][ MSE = 0.1990 | LCC = 0.9570 | SRCC = 0.9576 ] +INFO:root:Average loss=0.044409206345677375 +INFO:root: +[Valid][21][UTT][ MSE = 0.2298 | LCC = 0.9571 | SRCC = 0.9576 ] +INFO:root:Average loss=0.042248033612966536 +INFO:root: +[Valid][22][UTT][ MSE = 0.2170 | LCC = 0.9576 | SRCC = 0.9577 ] +INFO:root:Average loss=0.037435688748955726 +INFO:root: +[Valid][23][UTT][ MSE = 0.2196 | LCC = 0.9524 | SRCC = 0.9538 ] +INFO:root:Average loss=0.036608822792768475 +INFO:root: +[Valid][24][UTT][ MSE = 0.1978 | LCC = 0.9571 | SRCC = 0.9566 ] +INFO:root:Average loss=0.036259615436196324 +INFO:root: +[Valid][25][UTT][ MSE = 0.2011 | LCC = 0.9563 | SRCC = 0.9570 ] +INFO:root:Average loss=0.03343905363976955 +INFO:root: +[Valid][26][UTT][ MSE = 0.2111 | LCC = 0.9557 | SRCC = 0.9558 ] +INFO:root:Average loss=0.03355132400989533 +INFO:root: +[Valid][27][UTT][ MSE = 0.2134 | LCC = 0.9564 | SRCC = 0.9565 ] +INFO:root:Average loss=0.033236321851611135 +INFO:root: +[Valid][28][UTT][ MSE = 0.2230 | LCC = 0.9581 | SRCC = 0.9582 ] +INFO:root:Average loss=0.03284964564442634 +INFO:root: +[Valid][29][UTT][ MSE = 0.2306 | LCC = 0.9564 | SRCC = 0.9571 ] +INFO:root:Average loss=0.03055306017398834 +INFO:root: +[Valid][30][UTT][ MSE = 0.2051 | LCC = 0.9566 | SRCC = 0.9567 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2200 | LCC = 0.9596 | SRCC = 0.9603 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bfa88316ce67ccd49f65a058e8a4e68144a4bb0b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c8c394d342b2faac37113e0b4fe76ae15b7668523c96c5c3aa994f15770f7631 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..3e24725c767145d53ca926f7315fc70dd36644fb --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:de3885cbbd50430ff2c2a1e2753dd7f6456602ea06eea26aa48f3446e3b2ee3e +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..fc3df7ec759bb4bbdcf5fd5bde215dcf998c5821 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.30748978769779206 +INFO:root: +[Valid][0][UTT][ MSE = 0.2611 | LCC = 0.9409 | SRCC = 0.9415 ] +INFO:root:Average loss=0.14552893996238708 +INFO:root: +[Valid][1][UTT][ MSE = 0.2606 | LCC = 0.9472 | SRCC = 0.9492 ] +INFO:root:Average loss=0.12687842494249343 +INFO:root: +[Valid][2][UTT][ MSE = 0.2447 | LCC = 0.9503 | SRCC = 0.9516 ] +INFO:root:Average loss=0.11665557962656022 +INFO:root: +[Valid][3][UTT][ MSE = 0.2550 | LCC = 0.9533 | SRCC = 0.9554 ] +INFO:root:Average loss=0.10743808230757713 +INFO:root: +[Valid][4][UTT][ MSE = 0.2448 | LCC = 0.9532 | SRCC = 0.9536 ] +INFO:root:Average loss=0.10134608313441276 +INFO:root: +[Valid][5][UTT][ MSE = 0.2003 | LCC = 0.9539 | SRCC = 0.9556 ] +INFO:root:Average loss=0.09337500789761544 +INFO:root: +[Valid][6][UTT][ MSE = 0.2147 | LCC = 0.9562 | SRCC = 0.9569 ] +INFO:root:Average loss=0.0904866308271885 +INFO:root: +[Valid][7][UTT][ MSE = 0.2301 | LCC = 0.9560 | SRCC = 0.9566 ] +INFO:root:Average loss=0.08317287209630013 +INFO:root: +[Valid][8][UTT][ MSE = 0.2214 | LCC = 0.9541 | SRCC = 0.9562 ] +INFO:root:Average loss=0.07908796399831772 +INFO:root: +[Valid][9][UTT][ MSE = 0.2288 | LCC = 0.9562 | SRCC = 0.9568 ] +INFO:root:Average loss=0.0733034915626049 +INFO:root: +[Valid][10][UTT][ MSE = 0.2266 | LCC = 0.9549 | SRCC = 0.9555 ] +INFO:root:Average loss=0.07519827196002006 +INFO:root: +[Valid][11][UTT][ MSE = 0.2585 | LCC = 0.9545 | SRCC = 0.9574 ] +INFO:root:Average loss=0.06936877873539925 +INFO:root: +[Valid][12][UTT][ MSE = 0.1996 | LCC = 0.9576 | SRCC = 0.9583 ] +INFO:root:Average loss=0.06350462812185287 +INFO:root: +[Valid][13][UTT][ MSE = 0.2271 | LCC = 0.9571 | SRCC = 0.9565 ] +INFO:root:Average loss=0.06145478481054306 +INFO:root: +[Valid][14][UTT][ MSE = 0.2002 | LCC = 0.9566 | SRCC = 0.9562 ] +INFO:root:Average loss=0.05856886008381844 +INFO:root: +[Valid][15][UTT][ MSE = 0.2094 | LCC = 0.9563 | SRCC = 0.9566 ] +INFO:root:Average loss=0.05613876891136169 +INFO:root: +[Valid][16][UTT][ MSE = 0.1993 | LCC = 0.9557 | SRCC = 0.9559 ] +INFO:root:Average loss=0.05089227195084095 +INFO:root: +[Valid][17][UTT][ MSE = 0.2037 | LCC = 0.9564 | SRCC = 0.9569 ] +INFO:root:Average loss=0.049743518263101576 +INFO:root: +[Valid][18][UTT][ MSE = 0.2297 | LCC = 0.9558 | SRCC = 0.9562 ] +INFO:root:Average loss=0.04699956284463406 +INFO:root: +[Valid][19][UTT][ MSE = 0.2432 | LCC = 0.9549 | SRCC = 0.9556 ] +INFO:root:Average loss=0.04884118518233299 +INFO:root: +[Valid][20][UTT][ MSE = 0.2028 | LCC = 0.9563 | SRCC = 0.9562 ] +INFO:root:Average loss=0.04609034755825996 +INFO:root: +[Valid][21][UTT][ MSE = 0.2103 | LCC = 0.9553 | SRCC = 0.9556 ] +INFO:root:Average loss=0.043546176724135875 +INFO:root: +[Valid][22][UTT][ MSE = 0.2059 | LCC = 0.9565 | SRCC = 0.9566 ] +INFO:root:Average loss=0.039370008558034894 +INFO:root: +[Valid][23][UTT][ MSE = 0.2135 | LCC = 0.9566 | SRCC = 0.9566 ] +INFO:root:Average loss=0.03960313528776169 +INFO:root: +[Valid][24][UTT][ MSE = 0.2326 | LCC = 0.9557 | SRCC = 0.9568 ] +INFO:root:Average loss=0.03811305330693722 +INFO:root: +[Valid][25][UTT][ MSE = 0.2220 | LCC = 0.9558 | SRCC = 0.9566 ] +INFO:root:Average loss=0.03592284019291401 +INFO:root: +[Valid][26][UTT][ MSE = 0.2255 | LCC = 0.9568 | SRCC = 0.9572 ] +INFO:root:Average loss=0.035173287138342856 +INFO:root: +[Valid][27][UTT][ MSE = 0.2014 | LCC = 0.9555 | SRCC = 0.9554 ] +INFO:root:Average loss=0.032870600134134294 +INFO:root: +[Valid][28][UTT][ MSE = 0.2114 | LCC = 0.9558 | SRCC = 0.9554 ] +INFO:root:Average loss=0.033045374721288684 +INFO:root: +[Valid][29][UTT][ MSE = 0.2318 | LCC = 0.9561 | SRCC = 0.9564 ] +INFO:root:Average loss=0.031751164421439174 +INFO:root: +[Valid][30][UTT][ MSE = 0.2248 | LCC = 0.9558 | SRCC = 0.9565 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2006 | LCC = 0.9589 | SRCC = 0.9599 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b2201d97775dd58838123723d889b9157b79d46e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f81b32b4d31b5ffb87373dc6f88b685398c64e2ee6400100e9af71e758c52408 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..dc8053f2c8de809b6b5ecd9589204bae29a8d8f1 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c359013fbdf6a016afc55d3f32846ea71280f28b71338cfcfd87f29b3741b079 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..01e899f534c063050f5f419625f7fc7578a4bd0f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3269629694223404 +INFO:root: +[Valid][0][UTT][ MSE = 0.2737 | LCC = 0.9376 | SRCC = 0.9398 ] +INFO:root:Average loss=0.14909662783145905 +INFO:root: +[Valid][1][UTT][ MSE = 0.2352 | LCC = 0.9454 | SRCC = 0.9471 ] +INFO:root:Average loss=0.13158055931329726 +INFO:root: +[Valid][2][UTT][ MSE = 0.2354 | LCC = 0.9479 | SRCC = 0.9499 ] +INFO:root:Average loss=0.11529004117846489 +INFO:root: +[Valid][3][UTT][ MSE = 0.2349 | LCC = 0.9531 | SRCC = 0.9536 ] +INFO:root:Average loss=0.10945475324988366 +INFO:root: +[Valid][4][UTT][ MSE = 0.2460 | LCC = 0.9538 | SRCC = 0.9552 ] +INFO:root:Average loss=0.10719930690526962 +INFO:root: +[Valid][5][UTT][ MSE = 0.2252 | LCC = 0.9521 | SRCC = 0.9544 ] +INFO:root:Average loss=0.0977038896381855 +INFO:root: +[Valid][6][UTT][ MSE = 0.2259 | LCC = 0.9543 | SRCC = 0.9551 ] +INFO:root:Average loss=0.08917248910665512 +INFO:root: +[Valid][7][UTT][ MSE = 0.2361 | LCC = 0.9552 | SRCC = 0.9563 ] +INFO:root:Average loss=0.08561380288004874 +INFO:root: +[Valid][8][UTT][ MSE = 0.2276 | LCC = 0.9554 | SRCC = 0.9565 ] +INFO:root:Average loss=0.07979120087623597 +INFO:root: +[Valid][9][UTT][ MSE = 0.2422 | LCC = 0.9543 | SRCC = 0.9548 ] +INFO:root:Average loss=0.07649511635303498 +INFO:root: +[Valid][10][UTT][ MSE = 0.2066 | LCC = 0.9540 | SRCC = 0.9564 ] +INFO:root:Average loss=0.07279786220192909 +INFO:root: +[Valid][11][UTT][ MSE = 0.2200 | LCC = 0.9556 | SRCC = 0.9557 ] +INFO:root:Average loss=0.0679960608780384 +INFO:root: +[Valid][12][UTT][ MSE = 0.1984 | LCC = 0.9558 | SRCC = 0.9558 ] +INFO:root:Average loss=0.06590570190548897 +INFO:root: +[Valid][13][UTT][ MSE = 0.2037 | LCC = 0.9532 | SRCC = 0.9558 ] +INFO:root:Average loss=0.06164687684178352 +INFO:root: +[Valid][14][UTT][ MSE = 0.2246 | LCC = 0.9555 | SRCC = 0.9566 ] +INFO:root:Average loss=0.05960640203952789 +INFO:root: +[Valid][15][UTT][ MSE = 0.2223 | LCC = 0.9549 | SRCC = 0.9558 ] +INFO:root:Average loss=0.055171439051628114 +INFO:root: +[Valid][16][UTT][ MSE = 0.2408 | LCC = 0.9546 | SRCC = 0.9554 ] +INFO:root:Average loss=0.054617216140031816 +INFO:root: +[Valid][17][UTT][ MSE = 0.2119 | LCC = 0.9534 | SRCC = 0.9547 ] +INFO:root:Average loss=0.0534062277674675 +INFO:root: +[Valid][18][UTT][ MSE = 0.2075 | LCC = 0.9543 | SRCC = 0.9556 ] +INFO:root:Average loss=0.04906862150132656 +INFO:root: +[Valid][19][UTT][ MSE = 0.2185 | LCC = 0.9548 | SRCC = 0.9551 ] +INFO:root:Average loss=0.046708398133516314 +INFO:root: +[Valid][20][UTT][ MSE = 0.2159 | LCC = 0.9549 | SRCC = 0.9556 ] +INFO:root:Average loss=0.04501734770834446 +INFO:root: +[Valid][21][UTT][ MSE = 0.2155 | LCC = 0.9551 | SRCC = 0.9551 ] +INFO:root:Average loss=0.0427704903036356 +INFO:root: +[Valid][22][UTT][ MSE = 0.2037 | LCC = 0.9547 | SRCC = 0.9555 ] +INFO:root:Average loss=0.04193524633347988 +INFO:root: +[Valid][23][UTT][ MSE = 0.2006 | LCC = 0.9547 | SRCC = 0.9550 ] +INFO:root:Average loss=0.041833166033029556 +INFO:root: +[Valid][24][UTT][ MSE = 0.1970 | LCC = 0.9545 | SRCC = 0.9548 ] +INFO:root:Average loss=0.037694771990180016 +INFO:root: +[Valid][25][UTT][ MSE = 0.2049 | LCC = 0.9531 | SRCC = 0.9540 ] +INFO:root:Average loss=0.03771403874456882 +INFO:root: +[Valid][26][UTT][ MSE = 0.2197 | LCC = 0.9554 | SRCC = 0.9556 ] +INFO:root:Average loss=0.034953573226928714 +INFO:root: +[Valid][27][UTT][ MSE = 0.2077 | LCC = 0.9528 | SRCC = 0.9534 ] +INFO:root:Average loss=0.03288366769254208 +INFO:root: +[Valid][28][UTT][ MSE = 0.1987 | LCC = 0.9542 | SRCC = 0.9548 ] +INFO:root:Average loss=0.035096725091338155 +INFO:root: +[Valid][29][UTT][ MSE = 0.2247 | LCC = 0.9529 | SRCC = 0.9539 ] +INFO:root:Average loss=0.031401543140411374 +INFO:root: +[Valid][30][UTT][ MSE = 0.1999 | LCC = 0.9549 | SRCC = 0.9553 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2006 | LCC = 0.9589 | SRCC = 0.9588 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f887ed5985f420a60d9f44cb43524aef1ec060ff --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5d5bd07e6c29886526e743b0d61f519e8e9776f6fee6676e10107e4c0cf65fb5 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6d0bfe0d6797be16c93e06e64f2da962d223e2eb --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e3535baf0295464376ac2b36577c91dbb17af0158a971854b243b213bcefc6b9 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..704390824f6473a026879e44df6b38bb90ce22ba --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3326600753068924 +INFO:root: +[Valid][0][UTT][ MSE = 0.2750 | LCC = 0.9418 | SRCC = 0.9416 ] +INFO:root:Average loss=0.15628031468391418 +INFO:root: +[Valid][1][UTT][ MSE = 0.2389 | LCC = 0.9475 | SRCC = 0.9482 ] +INFO:root:Average loss=0.1309358567595482 +INFO:root: +[Valid][2][UTT][ MSE = 0.2422 | LCC = 0.9505 | SRCC = 0.9518 ] +INFO:root:Average loss=0.11690038454532624 +INFO:root: +[Valid][3][UTT][ MSE = 0.2449 | LCC = 0.9519 | SRCC = 0.9533 ] +INFO:root:Average loss=0.11236624661087989 +INFO:root: +[Valid][4][UTT][ MSE = 0.2420 | LCC = 0.9542 | SRCC = 0.9555 ] +INFO:root:Average loss=0.1010169178545475 +INFO:root: +[Valid][5][UTT][ MSE = 0.2523 | LCC = 0.9531 | SRCC = 0.9543 ] +INFO:root:Average loss=0.0973352013528347 +INFO:root: +[Valid][6][UTT][ MSE = 0.2487 | LCC = 0.9554 | SRCC = 0.9571 ] +INFO:root:Average loss=0.09149822422862053 +INFO:root: +[Valid][7][UTT][ MSE = 0.2273 | LCC = 0.9558 | SRCC = 0.9570 ] +INFO:root:Average loss=0.08593838784098626 +INFO:root: +[Valid][8][UTT][ MSE = 0.2355 | LCC = 0.9558 | SRCC = 0.9569 ] +INFO:root:Average loss=0.08068263736367226 +INFO:root: +[Valid][9][UTT][ MSE = 0.2053 | LCC = 0.9564 | SRCC = 0.9575 ] +INFO:root:Average loss=0.07529582896828652 +INFO:root: +[Valid][10][UTT][ MSE = 0.1943 | LCC = 0.9561 | SRCC = 0.9572 ] +INFO:root:Average loss=0.07089119905233383 +INFO:root: +[Valid][11][UTT][ MSE = 0.2181 | LCC = 0.9555 | SRCC = 0.9571 ] +INFO:root:Average loss=0.06751432055234909 +INFO:root: +[Valid][12][UTT][ MSE = 0.2048 | LCC = 0.9559 | SRCC = 0.9566 ] +INFO:root:Average loss=0.06894676274061202 +INFO:root: +[Valid][13][UTT][ MSE = 0.2111 | LCC = 0.9531 | SRCC = 0.9552 ] +INFO:root:Average loss=0.06352665975689888 +INFO:root: +[Valid][14][UTT][ MSE = 0.2137 | LCC = 0.9570 | SRCC = 0.9582 ] +INFO:root:Average loss=0.05827030390501022 +INFO:root: +[Valid][15][UTT][ MSE = 0.2244 | LCC = 0.9546 | SRCC = 0.9555 ] +INFO:root:Average loss=0.05840131562948227 +INFO:root: +[Valid][16][UTT][ MSE = 0.2139 | LCC = 0.9574 | SRCC = 0.9582 ] +INFO:root:Average loss=0.05425757849216461 +INFO:root: +[Valid][17][UTT][ MSE = 0.2233 | LCC = 0.9564 | SRCC = 0.9567 ] +INFO:root:Average loss=0.051161944285035135 +INFO:root: +[Valid][18][UTT][ MSE = 0.2011 | LCC = 0.9571 | SRCC = 0.9581 ] +INFO:root:Average loss=0.049501235976815224 +INFO:root: +[Valid][19][UTT][ MSE = 0.2414 | LCC = 0.9554 | SRCC = 0.9576 ] +INFO:root:Average loss=0.04604356896877289 +INFO:root: +[Valid][20][UTT][ MSE = 0.2096 | LCC = 0.9561 | SRCC = 0.9574 ] +INFO:root:Average loss=0.04481786760687828 +INFO:root: +[Valid][21][UTT][ MSE = 0.2269 | LCC = 0.9555 | SRCC = 0.9570 ] +INFO:root:Average loss=0.045749466508626935 +INFO:root: +[Valid][22][UTT][ MSE = 0.2141 | LCC = 0.9553 | SRCC = 0.9561 ] +INFO:root:Average loss=0.04135171182453632 +INFO:root: +[Valid][23][UTT][ MSE = 0.1923 | LCC = 0.9555 | SRCC = 0.9556 ] +INFO:root:Average loss=0.040235247671604156 +INFO:root: +[Valid][24][UTT][ MSE = 0.2333 | LCC = 0.9538 | SRCC = 0.9562 ] +INFO:root:Average loss=0.04145850935578346 +INFO:root: +[Valid][25][UTT][ MSE = 0.2226 | LCC = 0.9557 | SRCC = 0.9564 ] +INFO:root:Average loss=0.03977352221310139 +INFO:root: +[Valid][26][UTT][ MSE = 0.2278 | LCC = 0.9540 | SRCC = 0.9557 ] +INFO:root:Average loss=0.03829728551208973 +INFO:root: +[Valid][27][UTT][ MSE = 0.2304 | LCC = 0.9540 | SRCC = 0.9551 ] +INFO:root:Average loss=0.03721414828300476 +INFO:root: +[Valid][28][UTT][ MSE = 0.2048 | LCC = 0.9554 | SRCC = 0.9558 ] +INFO:root:Average loss=0.03428089794516563 +INFO:root: +[Valid][29][UTT][ MSE = 0.2011 | LCC = 0.9564 | SRCC = 0.9569 ] +INFO:root:Average loss=0.03439929533004761 +INFO:root: +[Valid][30][UTT][ MSE = 0.2196 | LCC = 0.9557 | SRCC = 0.9558 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2199 | LCC = 0.9569 | SRCC = 0.9580 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..89d4db07a3f5b4fdc251bcba0b5d005c2a59c6a5 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:251eb7aaf00a773c0556696e109fd3eb6310f059191b8c3e7e95612ac2f37abc +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c3905d6fe94b38e5724a0ba692e3bd24253dddf1 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cbe392faac558fce64229dc39c2226397de60b02841ba70311a055e2b625f8c3 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ef979e4a4a53245238005e80d956e3158e85c76f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer6/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.32349775910377504 +INFO:root: +[Valid][0][UTT][ MSE = 0.2737 | LCC = 0.9384 | SRCC = 0.9406 ] +INFO:root:Average loss=0.1487037494778633 +INFO:root: +[Valid][1][UTT][ MSE = 0.2689 | LCC = 0.9485 | SRCC = 0.9500 ] +INFO:root:Average loss=0.1273988019824028 +INFO:root: +[Valid][2][UTT][ MSE = 0.2487 | LCC = 0.9478 | SRCC = 0.9497 ] +INFO:root:Average loss=0.12232420033216476 +INFO:root: +[Valid][3][UTT][ MSE = 0.2426 | LCC = 0.9495 | SRCC = 0.9510 ] +INFO:root:Average loss=0.11414646849036217 +INFO:root: +[Valid][4][UTT][ MSE = 0.2260 | LCC = 0.9538 | SRCC = 0.9558 ] +INFO:root:Average loss=0.10313904035091401 +INFO:root: +[Valid][5][UTT][ MSE = 0.2192 | LCC = 0.9557 | SRCC = 0.9569 ] +INFO:root:Average loss=0.09753788104653359 +INFO:root: +[Valid][6][UTT][ MSE = 0.2323 | LCC = 0.9571 | SRCC = 0.9578 ] +INFO:root:Average loss=0.08950006806850433 +INFO:root: +[Valid][7][UTT][ MSE = 0.2015 | LCC = 0.9531 | SRCC = 0.9571 ] +INFO:root:Average loss=0.08438062259554863 +INFO:root: +[Valid][8][UTT][ MSE = 0.2281 | LCC = 0.9553 | SRCC = 0.9571 ] +INFO:root:Average loss=0.08044395026564598 +INFO:root: +[Valid][9][UTT][ MSE = 0.2273 | LCC = 0.9584 | SRCC = 0.9594 ] +INFO:root:Average loss=0.0758035848736763 +INFO:root: +[Valid][10][UTT][ MSE = 0.2365 | LCC = 0.9572 | SRCC = 0.9583 ] +INFO:root:Average loss=0.07190624272823334 +INFO:root: +[Valid][11][UTT][ MSE = 0.2014 | LCC = 0.9556 | SRCC = 0.9579 ] +INFO:root:Average loss=0.06865391981601715 +INFO:root: +[Valid][12][UTT][ MSE = 0.2138 | LCC = 0.9546 | SRCC = 0.9564 ] +INFO:root:Average loss=0.06440997803211213 +INFO:root: +[Valid][13][UTT][ MSE = 0.2224 | LCC = 0.9569 | SRCC = 0.9581 ] +INFO:root:Average loss=0.0605299137532711 +INFO:root: +[Valid][14][UTT][ MSE = 0.2268 | LCC = 0.9508 | SRCC = 0.9542 ] +INFO:root:Average loss=0.05998674747347832 +INFO:root: +[Valid][15][UTT][ MSE = 0.2346 | LCC = 0.9567 | SRCC = 0.9581 ] +INFO:root:Average loss=0.05503417943418026 +INFO:root: +[Valid][16][UTT][ MSE = 0.1962 | LCC = 0.9545 | SRCC = 0.9571 ] +INFO:root:Average loss=0.05356888744235039 +INFO:root: +[Valid][17][UTT][ MSE = 0.2282 | LCC = 0.9552 | SRCC = 0.9566 ] +INFO:root:Average loss=0.05128421983122826 +INFO:root: +[Valid][18][UTT][ MSE = 0.2371 | LCC = 0.9568 | SRCC = 0.9582 ] +INFO:root:Average loss=0.05071501068770885 +INFO:root: +[Valid][19][UTT][ MSE = 0.2210 | LCC = 0.9569 | SRCC = 0.9575 ] +INFO:root:Average loss=0.04689657472074032 +INFO:root: +[Valid][20][UTT][ MSE = 0.2089 | LCC = 0.9530 | SRCC = 0.9540 ] +INFO:root:Average loss=0.04483867461979389 +INFO:root: +[Valid][21][UTT][ MSE = 0.2108 | LCC = 0.9571 | SRCC = 0.9573 ] +INFO:root:Average loss=0.042382901474833486 +INFO:root: +[Valid][22][UTT][ MSE = 0.2172 | LCC = 0.9569 | SRCC = 0.9576 ] +INFO:root:Average loss=0.04279503960907459 +INFO:root: +[Valid][23][UTT][ MSE = 0.2081 | LCC = 0.9539 | SRCC = 0.9545 ] +INFO:root:Average loss=0.0421861622184515 +INFO:root: +[Valid][24][UTT][ MSE = 0.2393 | LCC = 0.9575 | SRCC = 0.9577 ] +INFO:root:Average loss=0.040351855427026746 +INFO:root: +[Valid][25][UTT][ MSE = 0.2142 | LCC = 0.9552 | SRCC = 0.9557 ] +INFO:root:Average loss=0.037436325684189795 +INFO:root: +[Valid][26][UTT][ MSE = 0.2211 | LCC = 0.9572 | SRCC = 0.9580 ] +INFO:root:Average loss=0.03507441794872284 +INFO:root: +[Valid][27][UTT][ MSE = 0.2028 | LCC = 0.9553 | SRCC = 0.9562 ] +INFO:root:Average loss=0.03409595124423504 +INFO:root: +[Valid][28][UTT][ MSE = 0.2375 | LCC = 0.9564 | SRCC = 0.9573 ] +INFO:root:Average loss=0.033593574732542035 +INFO:root: +[Valid][29][UTT][ MSE = 0.1979 | LCC = 0.9531 | SRCC = 0.9551 ] +INFO:root:Average loss=0.03225208745151758 +INFO:root: +[Valid][30][UTT][ MSE = 0.2187 | LCC = 0.9567 | SRCC = 0.9569 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2354 | LCC = 0.9590 | SRCC = 0.9600 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..da30cbe970e299baaa1d7ef5c387442193f38f89 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5ccf302ebb56cc99d2e2d0677e685354f957e70dcf084fb6b0f07594510c4f52 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ee2647bb99fbf41ddd2a4ae3906e88253a8d6151 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e1ef370787526955f2b6dad62ab18358fce4abdcd0e28f43aac5f0e99a959154 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..4b58c2af03623789532321c83bff0e49abf8fc73 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3071185450553894 +INFO:root: +[Valid][0][UTT][ MSE = 0.3364 | LCC = 0.9390 | SRCC = 0.9413 ] +INFO:root:Average loss=0.15208939754962922 +INFO:root: +[Valid][1][UTT][ MSE = 0.3059 | LCC = 0.9482 | SRCC = 0.9490 ] +INFO:root:Average loss=0.1348381357192993 +INFO:root: +[Valid][2][UTT][ MSE = 0.2819 | LCC = 0.9499 | SRCC = 0.9506 ] +INFO:root:Average loss=0.1252950499355793 +INFO:root: +[Valid][3][UTT][ MSE = 0.2297 | LCC = 0.9496 | SRCC = 0.9535 ] +INFO:root:Average loss=0.11662052726745606 +INFO:root: +[Valid][4][UTT][ MSE = 0.2237 | LCC = 0.9526 | SRCC = 0.9540 ] +INFO:root:Average loss=0.10802914604544639 +INFO:root: +[Valid][5][UTT][ MSE = 0.2264 | LCC = 0.9523 | SRCC = 0.9541 ] +INFO:root:Average loss=0.10130780190229416 +INFO:root: +[Valid][6][UTT][ MSE = 0.2219 | LCC = 0.9507 | SRCC = 0.9542 ] +INFO:root:Average loss=0.09880457544326782 +INFO:root: +[Valid][7][UTT][ MSE = 0.2466 | LCC = 0.9550 | SRCC = 0.9559 ] +INFO:root:Average loss=0.08957595178484917 +INFO:root: +[Valid][8][UTT][ MSE = 0.2480 | LCC = 0.9539 | SRCC = 0.9549 ] +INFO:root:Average loss=0.08504068839550018 +INFO:root: +[Valid][9][UTT][ MSE = 0.2130 | LCC = 0.9514 | SRCC = 0.9533 ] +INFO:root:Average loss=0.07878849402070046 +INFO:root: +[Valid][10][UTT][ MSE = 0.2246 | LCC = 0.9547 | SRCC = 0.9549 ] +INFO:root:Average loss=0.07617045962810516 +INFO:root: +[Valid][11][UTT][ MSE = 0.2216 | LCC = 0.9541 | SRCC = 0.9549 ] +INFO:root:Average loss=0.07399026390910149 +INFO:root: +[Valid][12][UTT][ MSE = 0.2274 | LCC = 0.9538 | SRCC = 0.9556 ] +INFO:root:Average loss=0.06841027155518532 +INFO:root: +[Valid][13][UTT][ MSE = 0.2329 | LCC = 0.9495 | SRCC = 0.9520 ] +INFO:root:Average loss=0.06445404960215091 +INFO:root: +[Valid][14][UTT][ MSE = 0.2464 | LCC = 0.9535 | SRCC = 0.9551 ] +INFO:root:Average loss=0.06342428804934025 +INFO:root: +[Valid][15][UTT][ MSE = 0.2090 | LCC = 0.9524 | SRCC = 0.9534 ] +INFO:root:Average loss=0.06093594294786453 +INFO:root: +[Valid][16][UTT][ MSE = 0.2344 | LCC = 0.9531 | SRCC = 0.9548 ] +INFO:root:Average loss=0.06165218847990036 +INFO:root: +[Valid][17][UTT][ MSE = 0.2308 | LCC = 0.9542 | SRCC = 0.9544 ] +INFO:root:Average loss=0.05435122752189636 +INFO:root: +[Valid][18][UTT][ MSE = 0.2352 | LCC = 0.9540 | SRCC = 0.9546 ] +INFO:root:Average loss=0.053847511321306225 +INFO:root: +[Valid][19][UTT][ MSE = 0.2407 | LCC = 0.9519 | SRCC = 0.9522 ] +INFO:root:Average loss=0.05134607657790184 +INFO:root: +[Valid][20][UTT][ MSE = 0.2165 | LCC = 0.9509 | SRCC = 0.9526 ] +INFO:root:Average loss=0.05283810438215732 +INFO:root: +[Valid][21][UTT][ MSE = 0.2564 | LCC = 0.9534 | SRCC = 0.9554 ] +INFO:root:Average loss=0.04766807574033737 +INFO:root: +[Valid][22][UTT][ MSE = 0.2318 | LCC = 0.9533 | SRCC = 0.9557 ] +INFO:root:Average loss=0.0459422916918993 +INFO:root: +[Valid][23][UTT][ MSE = 0.2131 | LCC = 0.9535 | SRCC = 0.9536 ] +INFO:root:Average loss=0.04482821346819401 +INFO:root: +[Valid][24][UTT][ MSE = 0.2268 | LCC = 0.9536 | SRCC = 0.9542 ] +INFO:root:Average loss=0.046041647389531135 +INFO:root: +[Valid][25][UTT][ MSE = 0.2331 | LCC = 0.9536 | SRCC = 0.9531 ] +INFO:root:Average loss=0.04218719543516636 +INFO:root: +[Valid][26][UTT][ MSE = 0.2144 | LCC = 0.9542 | SRCC = 0.9539 ] +INFO:root:Average loss=0.04008609035611153 +INFO:root: +[Valid][27][UTT][ MSE = 0.2300 | LCC = 0.9490 | SRCC = 0.9511 ] +INFO:root:Average loss=0.039685267478227614 +INFO:root: +[Valid][28][UTT][ MSE = 0.2144 | LCC = 0.9531 | SRCC = 0.9537 ] +INFO:root:Average loss=0.039271630883216856 +INFO:root: +[Valid][29][UTT][ MSE = 0.2385 | LCC = 0.9507 | SRCC = 0.9515 ] +INFO:root:Average loss=0.03886643026769161 +INFO:root: +[Valid][30][UTT][ MSE = 0.2700 | LCC = 0.9500 | SRCC = 0.9516 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2454 | LCC = 0.9595 | SRCC = 0.9609 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8beb21894ff9c942446f15b98377fca82bfb907f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:976030ab7724f7d8362eb7083b95b793ff79a0c9ff3dff2b57899210fa9a3ff1 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..81b025a338a00c0550e7ab7ba1dea027d5fa9170 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:23ef9c01cb95ed9c2777d7b230c48d747070c2c0405307694d51bd609cbaf45a +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..4373ce20c40310d3c1a48c48a20ea0e0f8002fcc --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.30847410184144974 +INFO:root: +[Valid][0][UTT][ MSE = 0.3264 | LCC = 0.9389 | SRCC = 0.9403 ] +INFO:root:Average loss=0.15280600774288178 +INFO:root: +[Valid][1][UTT][ MSE = 0.3000 | LCC = 0.9436 | SRCC = 0.9440 ] +INFO:root:Average loss=0.13385916221141816 +INFO:root: +[Valid][2][UTT][ MSE = 0.2756 | LCC = 0.9504 | SRCC = 0.9515 ] +INFO:root:Average loss=0.12040250623226166 +INFO:root: +[Valid][3][UTT][ MSE = 0.3202 | LCC = 0.9510 | SRCC = 0.9529 ] +INFO:root:Average loss=0.1147640918493271 +INFO:root: +[Valid][4][UTT][ MSE = 0.2566 | LCC = 0.9518 | SRCC = 0.9537 ] +INFO:root:Average loss=0.10406961965560912 +INFO:root: +[Valid][5][UTT][ MSE = 0.2759 | LCC = 0.9529 | SRCC = 0.9534 ] +INFO:root:Average loss=0.10214550268650055 +INFO:root: +[Valid][6][UTT][ MSE = 0.2942 | LCC = 0.9544 | SRCC = 0.9550 ] +INFO:root:Average loss=0.09446702888607979 +INFO:root: +[Valid][7][UTT][ MSE = 0.2418 | LCC = 0.9519 | SRCC = 0.9534 ] +INFO:root:Average loss=0.09022290033102036 +INFO:root: +[Valid][8][UTT][ MSE = 0.2573 | LCC = 0.9543 | SRCC = 0.9547 ] +INFO:root:Average loss=0.0858368440568447 +INFO:root: +[Valid][9][UTT][ MSE = 0.2390 | LCC = 0.9531 | SRCC = 0.9539 ] +INFO:root:Average loss=0.08000860139727592 +INFO:root: +[Valid][10][UTT][ MSE = 0.2433 | LCC = 0.9531 | SRCC = 0.9537 ] +INFO:root:Average loss=0.07352007490396499 +INFO:root: +[Valid][11][UTT][ MSE = 0.2709 | LCC = 0.9541 | SRCC = 0.9548 ] +INFO:root:Average loss=0.07450019612908364 +INFO:root: +[Valid][12][UTT][ MSE = 0.2554 | LCC = 0.9537 | SRCC = 0.9545 ] +INFO:root:Average loss=0.06916494417190552 +INFO:root: +[Valid][13][UTT][ MSE = 0.2535 | LCC = 0.9527 | SRCC = 0.9532 ] +INFO:root:Average loss=0.06495236304402352 +INFO:root: +[Valid][14][UTT][ MSE = 0.2719 | LCC = 0.9535 | SRCC = 0.9543 ] +INFO:root:Average loss=0.060464295640587806 +INFO:root: +[Valid][15][UTT][ MSE = 0.2753 | LCC = 0.9548 | SRCC = 0.9552 ] +INFO:root:Average loss=0.057160533398389814 +INFO:root: +[Valid][16][UTT][ MSE = 0.2431 | LCC = 0.9532 | SRCC = 0.9529 ] +INFO:root:Average loss=0.057098071053624154 +INFO:root: +[Valid][17][UTT][ MSE = 0.2276 | LCC = 0.9542 | SRCC = 0.9548 ] +INFO:root:Average loss=0.0561094585955143 +INFO:root: +[Valid][18][UTT][ MSE = 0.2501 | LCC = 0.9543 | SRCC = 0.9543 ] +INFO:root:Average loss=0.05095202107727528 +INFO:root: +[Valid][19][UTT][ MSE = 0.2625 | LCC = 0.9542 | SRCC = 0.9541 ] +INFO:root:Average loss=0.050529106721282004 +INFO:root: +[Valid][20][UTT][ MSE = 0.2507 | LCC = 0.9537 | SRCC = 0.9534 ] +INFO:root:Average loss=0.049283692479133605 +INFO:root: +[Valid][21][UTT][ MSE = 0.2416 | LCC = 0.9533 | SRCC = 0.9534 ] +INFO:root:Average loss=0.04743722987174988 +INFO:root: +[Valid][22][UTT][ MSE = 0.2859 | LCC = 0.9499 | SRCC = 0.9513 ] +INFO:root:Average loss=0.04523732696473599 +INFO:root: +[Valid][23][UTT][ MSE = 0.2748 | LCC = 0.9538 | SRCC = 0.9536 ] +INFO:root:Average loss=0.046494645297527316 +INFO:root: +[Valid][24][UTT][ MSE = 0.2307 | LCC = 0.9530 | SRCC = 0.9529 ] +INFO:root:Average loss=0.0426791632771492 +INFO:root: +[Valid][25][UTT][ MSE = 0.2826 | LCC = 0.9514 | SRCC = 0.9519 ] +INFO:root:Average loss=0.04154197160899639 +INFO:root: +[Valid][26][UTT][ MSE = 0.2224 | LCC = 0.9522 | SRCC = 0.9527 ] +INFO:root:Average loss=0.0385200377702713 +INFO:root: +[Valid][27][UTT][ MSE = 0.2476 | LCC = 0.9530 | SRCC = 0.9531 ] +INFO:root:Average loss=0.0368787686675787 +INFO:root: +[Valid][28][UTT][ MSE = 0.2617 | LCC = 0.9527 | SRCC = 0.9528 ] +INFO:root:Average loss=0.03610721354186535 +INFO:root: +[Valid][29][UTT][ MSE = 0.2740 | LCC = 0.9532 | SRCC = 0.9536 ] +INFO:root:Average loss=0.035530262529850004 +INFO:root: +[Valid][30][UTT][ MSE = 0.2337 | LCC = 0.9530 | SRCC = 0.9519 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2840 | LCC = 0.9566 | SRCC = 0.9582 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9e5ad06f42e63a696a71a3d9332c607b4d48d33d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b7fadd1559aad09f1681e17f6831b9dbeb1f6418a1fd2241daa5ad82eeb8f6be +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e848d7bb5395170d1a306ba969c027762452e533 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c232c6911a4c5ad9148af6e6dd58fa5cfad5be48b7d790297b8453a29b268ccb +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d157191299959b8e9999a43a1426a775f5d55c75 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.32814219695329666 +INFO:root: +[Valid][0][UTT][ MSE = 0.3007 | LCC = 0.9379 | SRCC = 0.9405 ] +INFO:root:Average loss=0.15723401361703873 +INFO:root: +[Valid][1][UTT][ MSE = 0.2814 | LCC = 0.9477 | SRCC = 0.9479 ] +INFO:root:Average loss=0.1376397372484207 +INFO:root: +[Valid][2][UTT][ MSE = 0.2240 | LCC = 0.9492 | SRCC = 0.9518 ] +INFO:root:Average loss=0.1267694907784462 +INFO:root: +[Valid][3][UTT][ MSE = 0.2877 | LCC = 0.9497 | SRCC = 0.9514 ] +INFO:root:Average loss=0.11793879240751266 +INFO:root: +[Valid][4][UTT][ MSE = 0.2348 | LCC = 0.9474 | SRCC = 0.9501 ] +INFO:root:Average loss=0.10966292306780816 +INFO:root: +[Valid][5][UTT][ MSE = 0.2574 | LCC = 0.9511 | SRCC = 0.9531 ] +INFO:root:Average loss=0.10597797906398773 +INFO:root: +[Valid][6][UTT][ MSE = 0.2564 | LCC = 0.9539 | SRCC = 0.9556 ] +INFO:root:Average loss=0.09764265564084053 +INFO:root: +[Valid][7][UTT][ MSE = 0.2328 | LCC = 0.9536 | SRCC = 0.9545 ] +INFO:root:Average loss=0.09045343378186226 +INFO:root: +[Valid][8][UTT][ MSE = 0.2722 | LCC = 0.9540 | SRCC = 0.9556 ] +INFO:root:Average loss=0.08801506209373475 +INFO:root: +[Valid][9][UTT][ MSE = 0.2317 | LCC = 0.9519 | SRCC = 0.9540 ] +INFO:root:Average loss=0.08329158079624176 +INFO:root: +[Valid][10][UTT][ MSE = 0.2428 | LCC = 0.9517 | SRCC = 0.9536 ] +INFO:root:Average loss=0.07950962662696838 +INFO:root: +[Valid][11][UTT][ MSE = 0.2292 | LCC = 0.9516 | SRCC = 0.9542 ] +INFO:root:Average loss=0.07555354967713356 +INFO:root: +[Valid][12][UTT][ MSE = 0.2424 | LCC = 0.9541 | SRCC = 0.9546 ] +INFO:root:Average loss=0.07393871024250984 +INFO:root: +[Valid][13][UTT][ MSE = 0.2339 | LCC = 0.9535 | SRCC = 0.9545 ] +INFO:root:Average loss=0.06880267548561096 +INFO:root: +[Valid][14][UTT][ MSE = 0.2337 | LCC = 0.9523 | SRCC = 0.9531 ] +INFO:root:Average loss=0.0659366729259491 +INFO:root: +[Valid][15][UTT][ MSE = 0.2409 | LCC = 0.9539 | SRCC = 0.9546 ] +INFO:root:Average loss=0.06287454730272293 +INFO:root: +[Valid][16][UTT][ MSE = 0.2225 | LCC = 0.9533 | SRCC = 0.9538 ] +INFO:root:Average loss=0.06102655684947968 +INFO:root: +[Valid][17][UTT][ MSE = 0.2579 | LCC = 0.9531 | SRCC = 0.9539 ] +INFO:root:Average loss=0.05780705735087395 +INFO:root: +[Valid][18][UTT][ MSE = 0.2192 | LCC = 0.9504 | SRCC = 0.9525 ] +INFO:root:Average loss=0.05771105340123177 +INFO:root: +[Valid][19][UTT][ MSE = 0.2474 | LCC = 0.9533 | SRCC = 0.9537 ] +INFO:root:Average loss=0.053209788888692856 +INFO:root: +[Valid][20][UTT][ MSE = 0.2376 | LCC = 0.9520 | SRCC = 0.9524 ] +INFO:root:Average loss=0.052242047622799874 +INFO:root: +[Valid][21][UTT][ MSE = 0.2463 | LCC = 0.9520 | SRCC = 0.9523 ] +INFO:root:Average loss=0.05197446389496326 +INFO:root: +[Valid][22][UTT][ MSE = 0.2589 | LCC = 0.9501 | SRCC = 0.9508 ] +INFO:root:Average loss=0.04781760449707508 +INFO:root: +[Valid][23][UTT][ MSE = 0.2207 | LCC = 0.9531 | SRCC = 0.9529 ] +INFO:root:Average loss=0.048198169127106665 +INFO:root: +[Valid][24][UTT][ MSE = 0.2484 | LCC = 0.9520 | SRCC = 0.9522 ] +INFO:root:Average loss=0.04650917276740074 +INFO:root: +[Valid][25][UTT][ MSE = 0.2424 | LCC = 0.9517 | SRCC = 0.9521 ] +INFO:root:Average loss=0.043508746981620786 +INFO:root: +[Valid][26][UTT][ MSE = 0.2172 | LCC = 0.9514 | SRCC = 0.9515 ] +INFO:root:Average loss=0.041616315737366676 +INFO:root: +[Valid][27][UTT][ MSE = 0.2301 | LCC = 0.9527 | SRCC = 0.9527 ] +INFO:root:Average loss=0.04121719247102738 +INFO:root: +[Valid][28][UTT][ MSE = 0.2483 | LCC = 0.9516 | SRCC = 0.9518 ] +INFO:root:Average loss=0.040538819268345835 +INFO:root: +[Valid][29][UTT][ MSE = 0.2241 | LCC = 0.9508 | SRCC = 0.9514 ] +INFO:root:Average loss=0.040499822035431864 +INFO:root: +[Valid][30][UTT][ MSE = 0.2176 | LCC = 0.9537 | SRCC = 0.9531 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2425 | LCC = 0.9589 | SRCC = 0.9593 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8b215154ad0d2c008e19edaff999cb938d4012af --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2d8ae745cffa5c495634ee3134692abf280cac44dde8a4a0c7da5d05997feaa4 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9d850eb2813c5518c39bbfb3dabb14e6097778c8 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f539598b9dcdb3743332728fed7cd29ffa65a9fa2a6dbc694b1b1bc50e7fd104 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2db1207e5d78c3f80a246c76846c76727c25e870 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.319490697145462 +INFO:root: +[Valid][0][UTT][ MSE = 0.2669 | LCC = 0.9387 | SRCC = 0.9392 ] +INFO:root:Average loss=0.15273748582601548 +INFO:root: +[Valid][1][UTT][ MSE = 0.2623 | LCC = 0.9466 | SRCC = 0.9464 ] +INFO:root:Average loss=0.13509550201892853 +INFO:root: +[Valid][2][UTT][ MSE = 0.2577 | LCC = 0.9489 | SRCC = 0.9499 ] +INFO:root:Average loss=0.12548893475532533 +INFO:root: +[Valid][3][UTT][ MSE = 0.2432 | LCC = 0.9499 | SRCC = 0.9507 ] +INFO:root:Average loss=0.11435694444179535 +INFO:root: +[Valid][4][UTT][ MSE = 0.2080 | LCC = 0.9473 | SRCC = 0.9500 ] +INFO:root:Average loss=0.10775166708230972 +INFO:root: +[Valid][5][UTT][ MSE = 0.2371 | LCC = 0.9526 | SRCC = 0.9533 ] +INFO:root:Average loss=0.09861659580469131 +INFO:root: +[Valid][6][UTT][ MSE = 0.1934 | LCC = 0.9529 | SRCC = 0.9539 ] +INFO:root:Average loss=0.09687559840083122 +INFO:root: +[Valid][7][UTT][ MSE = 0.2559 | LCC = 0.9510 | SRCC = 0.9521 ] +INFO:root:Average loss=0.09057757046818733 +INFO:root: +[Valid][8][UTT][ MSE = 0.2219 | LCC = 0.9541 | SRCC = 0.9545 ] +INFO:root:Average loss=0.08396152329444885 +INFO:root: +[Valid][9][UTT][ MSE = 0.2112 | LCC = 0.9529 | SRCC = 0.9547 ] +INFO:root:Average loss=0.07929067695140839 +INFO:root: +[Valid][10][UTT][ MSE = 0.2366 | LCC = 0.9526 | SRCC = 0.9533 ] +INFO:root:Average loss=0.07505992770195008 +INFO:root: +[Valid][11][UTT][ MSE = 0.2179 | LCC = 0.9517 | SRCC = 0.9532 ] +INFO:root:Average loss=0.07441780418157577 +INFO:root: +[Valid][12][UTT][ MSE = 0.2046 | LCC = 0.9523 | SRCC = 0.9528 ] +INFO:root:Average loss=0.06981167501211166 +INFO:root: +[Valid][13][UTT][ MSE = 0.2264 | LCC = 0.9521 | SRCC = 0.9523 ] +INFO:root:Average loss=0.06466648891568184 +INFO:root: +[Valid][14][UTT][ MSE = 0.2457 | LCC = 0.9515 | SRCC = 0.9524 ] +INFO:root:Average loss=0.06357144400477409 +INFO:root: +[Valid][15][UTT][ MSE = 0.2137 | LCC = 0.9529 | SRCC = 0.9526 ] +INFO:root:Average loss=0.060262638568878174 +INFO:root: +[Valid][16][UTT][ MSE = 0.2288 | LCC = 0.9516 | SRCC = 0.9533 ] +INFO:root:Average loss=0.05724249836802483 +INFO:root: +[Valid][17][UTT][ MSE = 0.2127 | LCC = 0.9515 | SRCC = 0.9521 ] +INFO:root:Average loss=0.05527810843288899 +INFO:root: +[Valid][18][UTT][ MSE = 0.2183 | LCC = 0.9522 | SRCC = 0.9517 ] +INFO:root:Average loss=0.05255786678195 +INFO:root: +[Valid][19][UTT][ MSE = 0.1978 | LCC = 0.9525 | SRCC = 0.9516 ] +INFO:root:Average loss=0.050133297443389895 +INFO:root: +[Valid][20][UTT][ MSE = 0.2146 | LCC = 0.9526 | SRCC = 0.9525 ] +INFO:root:Average loss=0.048646175548434256 +INFO:root: +[Valid][21][UTT][ MSE = 0.2010 | LCC = 0.9525 | SRCC = 0.9517 ] +INFO:root:Average loss=0.04769981828331947 +INFO:root: +[Valid][22][UTT][ MSE = 0.2230 | LCC = 0.9520 | SRCC = 0.9523 ] +INFO:root:Average loss=0.04536271917819977 +INFO:root: +[Valid][23][UTT][ MSE = 0.2012 | LCC = 0.9531 | SRCC = 0.9527 ] +INFO:root:Average loss=0.04403653356432915 +INFO:root: +[Valid][24][UTT][ MSE = 0.2171 | LCC = 0.9526 | SRCC = 0.9523 ] +INFO:root:Average loss=0.04160919363796711 +INFO:root: +[Valid][25][UTT][ MSE = 0.2082 | LCC = 0.9533 | SRCC = 0.9531 ] +INFO:root:Average loss=0.04252136099338531 +INFO:root: +[Valid][26][UTT][ MSE = 0.2111 | LCC = 0.9521 | SRCC = 0.9516 ] +INFO:root:Average loss=0.04099906630814076 +INFO:root: +[Valid][27][UTT][ MSE = 0.2355 | LCC = 0.9514 | SRCC = 0.9521 ] +INFO:root:Average loss=0.03828198892623186 +INFO:root: +[Valid][28][UTT][ MSE = 0.2126 | LCC = 0.9535 | SRCC = 0.9527 ] +INFO:root:Average loss=0.03833235250413418 +INFO:root: +[Valid][29][UTT][ MSE = 0.1983 | LCC = 0.9523 | SRCC = 0.9519 ] +INFO:root:Average loss=0.03624026170372963 +INFO:root: +[Valid][30][UTT][ MSE = 0.1977 | LCC = 0.9521 | SRCC = 0.9521 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2126 | LCC = 0.9605 | SRCC = 0.9617 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a5a8a287668a4c154368dfd2b7570716d05e77db --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c976232fb3fb75147222341aa91e9d30a8d11fb3451b80daa19a1e89a41881fd +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2099c6d11668ee4705c0fecbec1036902cde85af --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c71b15c739e1fe8fba6585fcab7aa4f67e971495e28603e9c017eb74cf7e7ba8 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6c5fd16c98a2fdd06180a29d4a78a2f768acec72 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer7/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.338375807762146 +INFO:root: +[Valid][0][UTT][ MSE = 0.3055 | LCC = 0.9412 | SRCC = 0.9402 ] +INFO:root:Average loss=0.15143426966667176 +INFO:root: +[Valid][1][UTT][ MSE = 0.2763 | LCC = 0.9436 | SRCC = 0.9443 ] +INFO:root:Average loss=0.13527364617586135 +INFO:root: +[Valid][2][UTT][ MSE = 0.2979 | LCC = 0.9492 | SRCC = 0.9508 ] +INFO:root:Average loss=0.12299202483892441 +INFO:root: +[Valid][3][UTT][ MSE = 0.2843 | LCC = 0.9510 | SRCC = 0.9527 ] +INFO:root:Average loss=0.11617883720993996 +INFO:root: +[Valid][4][UTT][ MSE = 0.2398 | LCC = 0.9545 | SRCC = 0.9550 ] +INFO:root:Average loss=0.10493625536561012 +INFO:root: +[Valid][5][UTT][ MSE = 0.2730 | LCC = 0.9543 | SRCC = 0.9556 ] +INFO:root:Average loss=0.09823452731966972 +INFO:root: +[Valid][6][UTT][ MSE = 0.2433 | LCC = 0.9510 | SRCC = 0.9521 ] +INFO:root:Average loss=0.09474132412672043 +INFO:root: +[Valid][7][UTT][ MSE = 0.2549 | LCC = 0.9546 | SRCC = 0.9557 ] +INFO:root:Average loss=0.0883436162173748 +INFO:root: +[Valid][8][UTT][ MSE = 0.2529 | LCC = 0.9548 | SRCC = 0.9562 ] +INFO:root:Average loss=0.08497751134634018 +INFO:root: +[Valid][9][UTT][ MSE = 0.2191 | LCC = 0.9524 | SRCC = 0.9537 ] +INFO:root:Average loss=0.07910806518793106 +INFO:root: +[Valid][10][UTT][ MSE = 0.2259 | LCC = 0.9554 | SRCC = 0.9554 ] +INFO:root:Average loss=0.07903225433826447 +INFO:root: +[Valid][11][UTT][ MSE = 0.2280 | LCC = 0.9532 | SRCC = 0.9542 ] +INFO:root:Average loss=0.07219799846410752 +INFO:root: +[Valid][12][UTT][ MSE = 0.2065 | LCC = 0.9544 | SRCC = 0.9549 ] +INFO:root:Average loss=0.06824723720550536 +INFO:root: +[Valid][13][UTT][ MSE = 0.2232 | LCC = 0.9529 | SRCC = 0.9535 ] +INFO:root:Average loss=0.0644503261744976 +INFO:root: +[Valid][14][UTT][ MSE = 0.2378 | LCC = 0.9547 | SRCC = 0.9552 ] +INFO:root:Average loss=0.06298120158910751 +INFO:root: +[Valid][15][UTT][ MSE = 0.2509 | LCC = 0.9547 | SRCC = 0.9558 ] +INFO:root:Average loss=0.06061821123957634 +INFO:root: +[Valid][16][UTT][ MSE = 0.2190 | LCC = 0.9536 | SRCC = 0.9538 ] +INFO:root:Average loss=0.05725909429788589 +INFO:root: +[Valid][17][UTT][ MSE = 0.2166 | LCC = 0.9542 | SRCC = 0.9544 ] +INFO:root:Average loss=0.05859625358879566 +INFO:root: +[Valid][18][UTT][ MSE = 0.2142 | LCC = 0.9556 | SRCC = 0.9556 ] +INFO:root:Average loss=0.05281237855553627 +INFO:root: +[Valid][19][UTT][ MSE = 0.2153 | LCC = 0.9527 | SRCC = 0.9525 ] +INFO:root:Average loss=0.05266924470663071 +INFO:root: +[Valid][20][UTT][ MSE = 0.2447 | LCC = 0.9533 | SRCC = 0.9539 ] +INFO:root:Average loss=0.05060779197514057 +INFO:root: +[Valid][21][UTT][ MSE = 0.2150 | LCC = 0.9541 | SRCC = 0.9544 ] +INFO:root:Average loss=0.04620484921336174 +INFO:root: +[Valid][22][UTT][ MSE = 0.2065 | LCC = 0.9532 | SRCC = 0.9531 ] +INFO:root:Average loss=0.045024503171443936 +INFO:root: +[Valid][23][UTT][ MSE = 0.2448 | LCC = 0.9510 | SRCC = 0.9523 ] +INFO:root:Average loss=0.04435450224578381 +INFO:root: +[Valid][24][UTT][ MSE = 0.2253 | LCC = 0.9537 | SRCC = 0.9537 ] +INFO:root:Average loss=0.04367011873424053 +INFO:root: +[Valid][25][UTT][ MSE = 0.2261 | LCC = 0.9537 | SRCC = 0.9543 ] +INFO:root:Average loss=0.0445161866992712 +INFO:root: +[Valid][26][UTT][ MSE = 0.2227 | LCC = 0.9519 | SRCC = 0.9527 ] +INFO:root:Average loss=0.03996937818825245 +INFO:root: +[Valid][27][UTT][ MSE = 0.2185 | LCC = 0.9522 | SRCC = 0.9518 ] +INFO:root:Average loss=0.03823926717042923 +INFO:root: +[Valid][28][UTT][ MSE = 0.2502 | LCC = 0.9500 | SRCC = 0.9516 ] +INFO:root:Average loss=0.040532510638237 +INFO:root: +[Valid][29][UTT][ MSE = 0.2206 | LCC = 0.9533 | SRCC = 0.9531 ] +INFO:root:Average loss=0.037788561955094335 +INFO:root: +[Valid][30][UTT][ MSE = 0.2206 | LCC = 0.9534 | SRCC = 0.9536 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2147 | LCC = 0.9591 | SRCC = 0.9598 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e73c080d845759d1bf30356a243faddfe459b430 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fa5781cac9a603e5423d95702677a90fd54a457a17e6b1e341f0c46cb336836b +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..41504c1cda0fbefa50758ec0db172f8fdcc1fa94 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c23b42ed2dbd2a047fd86f615f2045d5412d00f26eeca750489496bc960f0cc3 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..4dad2364b28ba34160963d997337aaf60774b385 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.24733014464378356 +INFO:root: +[Valid][0][UTT][ MSE = 0.2798 | LCC = 0.9405 | SRCC = 0.9420 ] +INFO:root:Average loss=0.14890363484621047 +INFO:root: +[Valid][1][UTT][ MSE = 0.2659 | LCC = 0.9439 | SRCC = 0.9491 ] +INFO:root:Average loss=0.1352456362247467 +INFO:root: +[Valid][2][UTT][ MSE = 0.2465 | LCC = 0.9517 | SRCC = 0.9528 ] +INFO:root:Average loss=0.13165163797140123 +INFO:root: +[Valid][3][UTT][ MSE = 0.2428 | LCC = 0.9502 | SRCC = 0.9526 ] +INFO:root:Average loss=0.11935136985778809 +INFO:root: +[Valid][4][UTT][ MSE = 0.2278 | LCC = 0.9519 | SRCC = 0.9548 ] +INFO:root:Average loss=0.1151485867202282 +INFO:root: +[Valid][5][UTT][ MSE = 0.2461 | LCC = 0.9539 | SRCC = 0.9554 ] +INFO:root:Average loss=0.1089752487540245 +INFO:root: +[Valid][6][UTT][ MSE = 0.2655 | LCC = 0.9522 | SRCC = 0.9553 ] +INFO:root:Average loss=0.10650761500000953 +INFO:root: +[Valid][7][UTT][ MSE = 0.2312 | LCC = 0.9533 | SRCC = 0.9549 ] +INFO:root:Average loss=0.10165597161650658 +INFO:root: +[Valid][8][UTT][ MSE = 0.2015 | LCC = 0.9545 | SRCC = 0.9568 ] +INFO:root:Average loss=0.09582463499903679 +INFO:root: +[Valid][9][UTT][ MSE = 0.2314 | LCC = 0.9536 | SRCC = 0.9559 ] +INFO:root:Average loss=0.08998153650760651 +INFO:root: +[Valid][10][UTT][ MSE = 0.2070 | LCC = 0.9530 | SRCC = 0.9550 ] +INFO:root:Average loss=0.08944647753238678 +INFO:root: +[Valid][11][UTT][ MSE = 0.2182 | LCC = 0.9566 | SRCC = 0.9570 ] +INFO:root:Average loss=0.08859243911504745 +INFO:root: +[Valid][12][UTT][ MSE = 0.2394 | LCC = 0.9553 | SRCC = 0.9577 ] +INFO:root:Average loss=0.08489649814367295 +INFO:root: +[Valid][13][UTT][ MSE = 0.2499 | LCC = 0.9543 | SRCC = 0.9566 ] +INFO:root:Average loss=0.08138299354910851 +INFO:root: +[Valid][14][UTT][ MSE = 0.1934 | LCC = 0.9516 | SRCC = 0.9543 ] +INFO:root:Average loss=0.07700020214915275 +INFO:root: +[Valid][15][UTT][ MSE = 0.2107 | LCC = 0.9555 | SRCC = 0.9574 ] +INFO:root:Average loss=0.07369148299098015 +INFO:root: +[Valid][16][UTT][ MSE = 0.2359 | LCC = 0.9567 | SRCC = 0.9578 ] +INFO:root:Average loss=0.07485935017466545 +INFO:root: +[Valid][17][UTT][ MSE = 0.2436 | LCC = 0.9524 | SRCC = 0.9548 ] +INFO:root:Average loss=0.06918028923869134 +INFO:root: +[Valid][18][UTT][ MSE = 0.2103 | LCC = 0.9565 | SRCC = 0.9579 ] +INFO:root:Average loss=0.06731000536680222 +INFO:root: +[Valid][19][UTT][ MSE = 0.2336 | LCC = 0.9542 | SRCC = 0.9556 ] +INFO:root:Average loss=0.0653413378894329 +INFO:root: +[Valid][20][UTT][ MSE = 0.2292 | LCC = 0.9545 | SRCC = 0.9557 ] +INFO:root:Average loss=0.06527551278471946 +INFO:root: +[Valid][21][UTT][ MSE = 0.2157 | LCC = 0.9537 | SRCC = 0.9563 ] +INFO:root:Average loss=0.06399988725781441 +INFO:root: +[Valid][22][UTT][ MSE = 0.2017 | LCC = 0.9575 | SRCC = 0.9577 ] +INFO:root:Average loss=0.06033149859309196 +INFO:root: +[Valid][23][UTT][ MSE = 0.2105 | LCC = 0.9578 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05884064263105392 +INFO:root: +[Valid][24][UTT][ MSE = 0.2077 | LCC = 0.9557 | SRCC = 0.9570 ] +INFO:root:Average loss=0.05808066339790821 +INFO:root: +[Valid][25][UTT][ MSE = 0.2410 | LCC = 0.9550 | SRCC = 0.9571 ] +INFO:root:Average loss=0.05858028689026833 +INFO:root: +[Valid][26][UTT][ MSE = 0.2169 | LCC = 0.9531 | SRCC = 0.9557 ] +INFO:root:Average loss=0.05421239596605301 +INFO:root: +[Valid][27][UTT][ MSE = 0.2001 | LCC = 0.9553 | SRCC = 0.9562 ] +INFO:root:Average loss=0.05320944878458977 +INFO:root: +[Valid][28][UTT][ MSE = 0.2226 | LCC = 0.9520 | SRCC = 0.9549 ] +INFO:root:Average loss=0.05191692048311233 +INFO:root: +[Valid][29][UTT][ MSE = 0.2056 | LCC = 0.9541 | SRCC = 0.9550 ] +INFO:root:Average loss=0.04940225975215435 +INFO:root: +[Valid][30][UTT][ MSE = 0.2160 | LCC = 0.9570 | SRCC = 0.9576 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2129 | LCC = 0.9589 | SRCC = 0.9593 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a043e526cad9a100405556596d9898b8f1ba57f9 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d2a5755497a06befd23c66bc562f61f2e1aa11cef8c7b71f731f8fce1c9dcbb5 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..780b7de60407f8e3d98d827ed09c29bbebcd58ed --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4a92bae58177da390eeb7f6e4bbd813155f279f7b5f615b420fca4e396ab3a22 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..60225bfcfff2bd83d8d3ac3cec9d198eea773db4 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.28448558354377745 +INFO:root: +[Valid][0][UTT][ MSE = 0.3048 | LCC = 0.9417 | SRCC = 0.9418 ] +INFO:root:Average loss=0.15424219554662705 +INFO:root: +[Valid][1][UTT][ MSE = 0.3477 | LCC = 0.9435 | SRCC = 0.9462 ] +INFO:root:Average loss=0.14052646297216415 +INFO:root: +[Valid][2][UTT][ MSE = 0.2855 | LCC = 0.9502 | SRCC = 0.9515 ] +INFO:root:Average loss=0.12764964473247528 +INFO:root: +[Valid][3][UTT][ MSE = 0.2976 | LCC = 0.9510 | SRCC = 0.9535 ] +INFO:root:Average loss=0.12093314781785011 +INFO:root: +[Valid][4][UTT][ MSE = 0.2649 | LCC = 0.9528 | SRCC = 0.9545 ] +INFO:root:Average loss=0.11781243711709977 +INFO:root: +[Valid][5][UTT][ MSE = 0.2409 | LCC = 0.9541 | SRCC = 0.9555 ] +INFO:root:Average loss=0.10863607582449913 +INFO:root: +[Valid][6][UTT][ MSE = 0.2840 | LCC = 0.9548 | SRCC = 0.9567 ] +INFO:root:Average loss=0.11265245166420937 +INFO:root: +[Valid][7][UTT][ MSE = 0.2628 | LCC = 0.9550 | SRCC = 0.9570 ] +INFO:root:Average loss=0.09995205470919609 +INFO:root: +[Valid][8][UTT][ MSE = 0.2411 | LCC = 0.9551 | SRCC = 0.9560 ] +INFO:root:Average loss=0.09592345663905144 +INFO:root: +[Valid][9][UTT][ MSE = 0.2203 | LCC = 0.9545 | SRCC = 0.9558 ] +INFO:root:Average loss=0.09270992940664291 +INFO:root: +[Valid][10][UTT][ MSE = 0.2332 | LCC = 0.9543 | SRCC = 0.9565 ] +INFO:root:Average loss=0.08964308884739876 +INFO:root: +[Valid][11][UTT][ MSE = 0.2343 | LCC = 0.9527 | SRCC = 0.9534 ] +INFO:root:Average loss=0.08550577047467232 +INFO:root: +[Valid][12][UTT][ MSE = 0.2248 | LCC = 0.9554 | SRCC = 0.9558 ] +INFO:root:Average loss=0.08312442648410798 +INFO:root: +[Valid][13][UTT][ MSE = 0.2485 | LCC = 0.9552 | SRCC = 0.9562 ] +INFO:root:Average loss=0.07597479408979416 +INFO:root: +[Valid][14][UTT][ MSE = 0.2413 | LCC = 0.9549 | SRCC = 0.9560 ] +INFO:root:Average loss=0.07607378050684929 +INFO:root: +[Valid][15][UTT][ MSE = 0.2302 | LCC = 0.9543 | SRCC = 0.9555 ] +INFO:root:Average loss=0.07260325220227241 +INFO:root: +[Valid][16][UTT][ MSE = 0.2561 | LCC = 0.9545 | SRCC = 0.9555 ] +INFO:root:Average loss=0.0710518094599247 +INFO:root: +[Valid][17][UTT][ MSE = 0.2632 | LCC = 0.9539 | SRCC = 0.9548 ] +INFO:root:Average loss=0.0685799890011549 +INFO:root: +[Valid][18][UTT][ MSE = 0.2718 | LCC = 0.9535 | SRCC = 0.9551 ] +INFO:root:Average loss=0.0672579285055399 +INFO:root: +[Valid][19][UTT][ MSE = 0.2119 | LCC = 0.9530 | SRCC = 0.9531 ] +INFO:root:Average loss=0.0633063032925129 +INFO:root: +[Valid][20][UTT][ MSE = 0.2502 | LCC = 0.9529 | SRCC = 0.9557 ] +INFO:root:Average loss=0.06361642819643021 +INFO:root: +[Valid][21][UTT][ MSE = 0.2325 | LCC = 0.9531 | SRCC = 0.9535 ] +INFO:root:Average loss=0.06076620492339134 +INFO:root: +[Valid][22][UTT][ MSE = 0.2168 | LCC = 0.9553 | SRCC = 0.9562 ] +INFO:root:Average loss=0.05915972071886062 +INFO:root: +[Valid][23][UTT][ MSE = 0.2064 | LCC = 0.9523 | SRCC = 0.9537 ] +INFO:root:Average loss=0.05801529349386692 +INFO:root: +[Valid][24][UTT][ MSE = 0.2264 | LCC = 0.9558 | SRCC = 0.9567 ] +INFO:root:Average loss=0.05690932542085648 +INFO:root: +[Valid][25][UTT][ MSE = 0.2476 | LCC = 0.9494 | SRCC = 0.9502 ] +INFO:root:Average loss=0.057671471387147906 +INFO:root: +[Valid][26][UTT][ MSE = 0.2257 | LCC = 0.9539 | SRCC = 0.9548 ] +INFO:root:Average loss=0.053126771658658985 +INFO:root: +[Valid][27][UTT][ MSE = 0.2443 | LCC = 0.9538 | SRCC = 0.9553 ] +INFO:root:Average loss=0.04960929700732231 +INFO:root: +[Valid][28][UTT][ MSE = 0.2248 | LCC = 0.9555 | SRCC = 0.9558 ] +INFO:root:Average loss=0.048864162713289264 +INFO:root: +[Valid][29][UTT][ MSE = 0.2290 | LCC = 0.9547 | SRCC = 0.9558 ] +INFO:root:Average loss=0.047834970638155935 +INFO:root: +[Valid][30][UTT][ MSE = 0.2256 | LCC = 0.9554 | SRCC = 0.9557 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2320 | LCC = 0.9566 | SRCC = 0.9572 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c26a9ce658aa6461d5026d3e37f251580ac4490f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f955b84618639c8427cbc68ccfb95f5faf397bc4ecd20730d8bfec31a63ccd71 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c34f5145065bdd4e54a3e6cca8c76e8aaadf6b3a --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af8c6aa234608bb28c5acb38ea660476169e451bd5258299e1118a2d3cbebdfb +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c6682e0e2682ad487e9d2ddea34778105f2f8d9d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.31773817509412766 +INFO:root: +[Valid][0][UTT][ MSE = 0.2675 | LCC = 0.9382 | SRCC = 0.9404 ] +INFO:root:Average loss=0.15226690351963043 +INFO:root: +[Valid][1][UTT][ MSE = 0.2570 | LCC = 0.9480 | SRCC = 0.9494 ] +INFO:root:Average loss=0.13926805210113524 +INFO:root: +[Valid][2][UTT][ MSE = 0.2253 | LCC = 0.9472 | SRCC = 0.9514 ] +INFO:root:Average loss=0.1288311832547188 +INFO:root: +[Valid][3][UTT][ MSE = 0.2552 | LCC = 0.9526 | SRCC = 0.9544 ] +INFO:root:Average loss=0.12132719135284424 +INFO:root: +[Valid][4][UTT][ MSE = 0.2190 | LCC = 0.9533 | SRCC = 0.9547 ] +INFO:root:Average loss=0.1155809422135353 +INFO:root: +[Valid][5][UTT][ MSE = 0.2209 | LCC = 0.9526 | SRCC = 0.9545 ] +INFO:root:Average loss=0.10854007908701897 +INFO:root: +[Valid][6][UTT][ MSE = 0.2387 | LCC = 0.9558 | SRCC = 0.9573 ] +INFO:root:Average loss=0.10414539217948914 +INFO:root: +[Valid][7][UTT][ MSE = 0.2249 | LCC = 0.9538 | SRCC = 0.9560 ] +INFO:root:Average loss=0.09917936730384827 +INFO:root: +[Valid][8][UTT][ MSE = 0.2273 | LCC = 0.9564 | SRCC = 0.9570 ] +INFO:root:Average loss=0.09695577597618103 +INFO:root: +[Valid][9][UTT][ MSE = 0.2393 | LCC = 0.9552 | SRCC = 0.9571 ] +INFO:root:Average loss=0.09000667744874954 +INFO:root: +[Valid][10][UTT][ MSE = 0.2162 | LCC = 0.9555 | SRCC = 0.9580 ] +INFO:root:Average loss=0.08987167838215827 +INFO:root: +[Valid][11][UTT][ MSE = 0.2411 | LCC = 0.9572 | SRCC = 0.9574 ] +INFO:root:Average loss=0.0857013199031353 +INFO:root: +[Valid][12][UTT][ MSE = 0.2204 | LCC = 0.9547 | SRCC = 0.9551 ] +INFO:root:Average loss=0.07976656267046929 +INFO:root: +[Valid][13][UTT][ MSE = 0.1971 | LCC = 0.9567 | SRCC = 0.9575 ] +INFO:root:Average loss=0.08046184042096138 +INFO:root: +[Valid][14][UTT][ MSE = 0.2132 | LCC = 0.9570 | SRCC = 0.9572 ] +INFO:root:Average loss=0.07409888604283332 +INFO:root: +[Valid][15][UTT][ MSE = 0.2078 | LCC = 0.9552 | SRCC = 0.9559 ] +INFO:root:Average loss=0.07278017729520797 +INFO:root: +[Valid][16][UTT][ MSE = 0.2026 | LCC = 0.9553 | SRCC = 0.9576 ] +INFO:root:Average loss=0.07501217603683472 +INFO:root: +[Valid][17][UTT][ MSE = 0.2311 | LCC = 0.9568 | SRCC = 0.9583 ] +INFO:root:Average loss=0.07059222319722176 +INFO:root: +[Valid][18][UTT][ MSE = 0.2416 | LCC = 0.9543 | SRCC = 0.9560 ] +INFO:root:Average loss=0.06904644247889519 +INFO:root: +[Valid][19][UTT][ MSE = 0.2037 | LCC = 0.9559 | SRCC = 0.9572 ] +INFO:root:Average loss=0.06518424516916276 +INFO:root: +[Valid][20][UTT][ MSE = 0.2190 | LCC = 0.9542 | SRCC = 0.9561 ] +INFO:root:Average loss=0.06213113649189472 +INFO:root: +[Valid][21][UTT][ MSE = 0.2012 | LCC = 0.9569 | SRCC = 0.9574 ] +INFO:root:Average loss=0.059777047991752624 +INFO:root: +[Valid][22][UTT][ MSE = 0.1980 | LCC = 0.9568 | SRCC = 0.9577 ] +INFO:root:Average loss=0.05820715236663818 +INFO:root: +[Valid][23][UTT][ MSE = 0.2009 | LCC = 0.9548 | SRCC = 0.9576 ] +INFO:root:Average loss=0.05696473172307014 +INFO:root: +[Valid][24][UTT][ MSE = 0.1932 | LCC = 0.9522 | SRCC = 0.9548 ] +INFO:root:Average loss=0.05670236019790172 +INFO:root: +[Valid][25][UTT][ MSE = 0.2313 | LCC = 0.9554 | SRCC = 0.9577 ] +INFO:root:Average loss=0.05475941252708435 +INFO:root: +[Valid][26][UTT][ MSE = 0.2000 | LCC = 0.9565 | SRCC = 0.9574 ] +INFO:root:Average loss=0.05212402307987213 +INFO:root: +[Valid][27][UTT][ MSE = 0.1912 | LCC = 0.9562 | SRCC = 0.9572 ] +INFO:root:Average loss=0.053873953938484195 +INFO:root: +[Valid][28][UTT][ MSE = 0.2011 | LCC = 0.9529 | SRCC = 0.9543 ] +INFO:root:Average loss=0.052291763067245485 +INFO:root: +[Valid][29][UTT][ MSE = 0.1910 | LCC = 0.9551 | SRCC = 0.9563 ] +INFO:root:Average loss=0.04960975813865662 +INFO:root: +[Valid][30][UTT][ MSE = 0.2269 | LCC = 0.9569 | SRCC = 0.9579 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2452 | LCC = 0.9577 | SRCC = 0.9591 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4ecf6ecde9e22b8674bf217010d2b234ade02a8d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:71d3361b94613315881ca783edb2feec6d46500a7fa0e7b3e65f3c941a945502 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c6cc0b81ea353bec24744c89bd7318ac8fec56b3 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6baaaeb1bb12066ddd20aeeb73756e933a1dd7b353ae1061b0c4891d1537c10a +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5e7c78b1067978de55d472e21dbe31b890329d84 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3102157621383667 +INFO:root: +[Valid][0][UTT][ MSE = 0.3594 | LCC = 0.9436 | SRCC = 0.9444 ] +INFO:root:Average loss=0.15229062759876252 +INFO:root: +[Valid][1][UTT][ MSE = 0.2652 | LCC = 0.9496 | SRCC = 0.9507 ] +INFO:root:Average loss=0.13676434534788132 +INFO:root: +[Valid][2][UTT][ MSE = 0.3008 | LCC = 0.9525 | SRCC = 0.9539 ] +INFO:root:Average loss=0.13011730599403382 +INFO:root: +[Valid][3][UTT][ MSE = 0.2410 | LCC = 0.9498 | SRCC = 0.9533 ] +INFO:root:Average loss=0.12417296999692917 +INFO:root: +[Valid][4][UTT][ MSE = 0.2466 | LCC = 0.9471 | SRCC = 0.9519 ] +INFO:root:Average loss=0.11737407863140106 +INFO:root: +[Valid][5][UTT][ MSE = 0.2729 | LCC = 0.9538 | SRCC = 0.9562 ] +INFO:root:Average loss=0.11001595562696456 +INFO:root: +[Valid][6][UTT][ MSE = 0.2239 | LCC = 0.9546 | SRCC = 0.9559 ] +INFO:root:Average loss=0.10658535945415497 +INFO:root: +[Valid][7][UTT][ MSE = 0.2403 | LCC = 0.9564 | SRCC = 0.9577 ] +INFO:root:Average loss=0.10113639768958092 +INFO:root: +[Valid][8][UTT][ MSE = 0.2153 | LCC = 0.9555 | SRCC = 0.9570 ] +INFO:root:Average loss=0.09616619282960892 +INFO:root: +[Valid][9][UTT][ MSE = 0.2345 | LCC = 0.9544 | SRCC = 0.9574 ] +INFO:root:Average loss=0.09603134328126907 +INFO:root: +[Valid][10][UTT][ MSE = 0.2245 | LCC = 0.9535 | SRCC = 0.9575 ] +INFO:root:Average loss=0.09248003116250038 +INFO:root: +[Valid][11][UTT][ MSE = 0.2479 | LCC = 0.9567 | SRCC = 0.9580 ] +INFO:root:Average loss=0.08779804429411889 +INFO:root: +[Valid][12][UTT][ MSE = 0.2391 | LCC = 0.9549 | SRCC = 0.9574 ] +INFO:root:Average loss=0.08530081307888031 +INFO:root: +[Valid][13][UTT][ MSE = 0.2376 | LCC = 0.9534 | SRCC = 0.9544 ] +INFO:root:Average loss=0.08373410752415657 +INFO:root: +[Valid][14][UTT][ MSE = 0.2255 | LCC = 0.9561 | SRCC = 0.9577 ] +INFO:root:Average loss=0.07858064219355583 +INFO:root: +[Valid][15][UTT][ MSE = 0.2766 | LCC = 0.9567 | SRCC = 0.9580 ] +INFO:root:Average loss=0.07720868414640426 +INFO:root: +[Valid][16][UTT][ MSE = 0.2275 | LCC = 0.9554 | SRCC = 0.9556 ] +INFO:root:Average loss=0.07461480775475501 +INFO:root: +[Valid][17][UTT][ MSE = 0.2171 | LCC = 0.9559 | SRCC = 0.9577 ] +INFO:root:Average loss=0.06947526663541793 +INFO:root: +[Valid][18][UTT][ MSE = 0.2215 | LCC = 0.9562 | SRCC = 0.9579 ] +INFO:root:Average loss=0.06873399299383164 +INFO:root: +[Valid][19][UTT][ MSE = 0.2596 | LCC = 0.9568 | SRCC = 0.9581 ] +INFO:root:Average loss=0.06436493280529976 +INFO:root: +[Valid][20][UTT][ MSE = 0.2411 | LCC = 0.9555 | SRCC = 0.9561 ] +INFO:root:Average loss=0.06248075750470161 +INFO:root: +[Valid][21][UTT][ MSE = 0.2451 | LCC = 0.9555 | SRCC = 0.9567 ] +INFO:root:Average loss=0.06324950110912322 +INFO:root: +[Valid][22][UTT][ MSE = 0.2411 | LCC = 0.9559 | SRCC = 0.9570 ] +INFO:root:Average loss=0.06002864575386047 +INFO:root: +[Valid][23][UTT][ MSE = 0.2253 | LCC = 0.9555 | SRCC = 0.9562 ] +INFO:root:Average loss=0.05682684803009033 +INFO:root: +[Valid][24][UTT][ MSE = 0.2370 | LCC = 0.9556 | SRCC = 0.9565 ] +INFO:root:Average loss=0.05765553116798401 +INFO:root: +[Valid][25][UTT][ MSE = 0.2147 | LCC = 0.9556 | SRCC = 0.9560 ] +INFO:root:Average loss=0.055255173951387404 +INFO:root: +[Valid][26][UTT][ MSE = 0.2123 | LCC = 0.9553 | SRCC = 0.9569 ] +INFO:root:Average loss=0.05359240609407425 +INFO:root: +[Valid][27][UTT][ MSE = 0.2133 | LCC = 0.9548 | SRCC = 0.9561 ] +INFO:root:Average loss=0.053808876648545265 +INFO:root: +[Valid][28][UTT][ MSE = 0.2297 | LCC = 0.9562 | SRCC = 0.9569 ] +INFO:root:Average loss=0.050293172031641006 +INFO:root: +[Valid][29][UTT][ MSE = 0.2250 | LCC = 0.9554 | SRCC = 0.9561 ] +INFO:root:Average loss=0.04964933297038078 +INFO:root: +[Valid][30][UTT][ MSE = 0.2395 | LCC = 0.9556 | SRCC = 0.9561 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2693 | LCC = 0.9577 | SRCC = 0.9592 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4edb91e9ebb71a388450982feaadbec7591314d2 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b6861671d38bf3e92f6ae22cd07c1601344b66603277971baae51849cb361c9 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..1258ec04b72b9f228d57fa410fa45b4753427dbb --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1c5167a5bd6b68a04e51195827f7ebdd27e9844ec9e3a5737edaaae2ce83834b +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9b8ff7caf959efb7fe566c59e64d57fb26446064 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer8/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.26389545255899427 +INFO:root: +[Valid][0][UTT][ MSE = 0.3102 | LCC = 0.9415 | SRCC = 0.9422 ] +INFO:root:Average loss=0.1525710398554802 +INFO:root: +[Valid][1][UTT][ MSE = 0.2662 | LCC = 0.9469 | SRCC = 0.9481 ] +INFO:root:Average loss=0.13801996499300004 +INFO:root: +[Valid][2][UTT][ MSE = 0.3143 | LCC = 0.9499 | SRCC = 0.9515 ] +INFO:root:Average loss=0.13088900208473206 +INFO:root: +[Valid][3][UTT][ MSE = 0.2629 | LCC = 0.9519 | SRCC = 0.9528 ] +INFO:root:Average loss=0.11954002505540848 +INFO:root: +[Valid][4][UTT][ MSE = 0.2743 | LCC = 0.9536 | SRCC = 0.9545 ] +INFO:root:Average loss=0.11437464216351509 +INFO:root: +[Valid][5][UTT][ MSE = 0.2649 | LCC = 0.9532 | SRCC = 0.9556 ] +INFO:root:Average loss=0.10740545868873597 +INFO:root: +[Valid][6][UTT][ MSE = 0.2508 | LCC = 0.9551 | SRCC = 0.9564 ] +INFO:root:Average loss=0.10359200370311737 +INFO:root: +[Valid][7][UTT][ MSE = 0.2366 | LCC = 0.9543 | SRCC = 0.9548 ] +INFO:root:Average loss=0.1033398607969284 +INFO:root: +[Valid][8][UTT][ MSE = 0.2545 | LCC = 0.9552 | SRCC = 0.9563 ] +INFO:root:Average loss=0.09794021686911583 +INFO:root: +[Valid][9][UTT][ MSE = 0.2287 | LCC = 0.9564 | SRCC = 0.9570 ] +INFO:root:Average loss=0.09071828013658524 +INFO:root: +[Valid][10][UTT][ MSE = 0.2451 | LCC = 0.9553 | SRCC = 0.9573 ] +INFO:root:Average loss=0.0884531799852848 +INFO:root: +[Valid][11][UTT][ MSE = 0.2497 | LCC = 0.9563 | SRCC = 0.9573 ] +INFO:root:Average loss=0.08436953341960907 +INFO:root: +[Valid][12][UTT][ MSE = 0.2511 | LCC = 0.9560 | SRCC = 0.9572 ] +INFO:root:Average loss=0.08464269718527793 +INFO:root: +[Valid][13][UTT][ MSE = 0.2316 | LCC = 0.9574 | SRCC = 0.9590 ] +INFO:root:Average loss=0.07798559102416039 +INFO:root: +[Valid][14][UTT][ MSE = 0.2305 | LCC = 0.9575 | SRCC = 0.9585 ] +INFO:root:Average loss=0.07841711819171905 +INFO:root: +[Valid][15][UTT][ MSE = 0.2026 | LCC = 0.9559 | SRCC = 0.9572 ] +INFO:root:Average loss=0.07317124611139297 +INFO:root: +[Valid][16][UTT][ MSE = 0.2512 | LCC = 0.9565 | SRCC = 0.9579 ] +INFO:root:Average loss=0.07041464948654175 +INFO:root: +[Valid][17][UTT][ MSE = 0.2289 | LCC = 0.9557 | SRCC = 0.9572 ] +INFO:root:Average loss=0.07073857089877128 +INFO:root: +[Valid][18][UTT][ MSE = 0.2069 | LCC = 0.9563 | SRCC = 0.9570 ] +INFO:root:Average loss=0.06634580513834953 +INFO:root: +[Valid][19][UTT][ MSE = 0.2476 | LCC = 0.9557 | SRCC = 0.9575 ] +INFO:root:Average loss=0.06489538097381592 +INFO:root: +[Valid][20][UTT][ MSE = 0.2122 | LCC = 0.9542 | SRCC = 0.9551 ] +INFO:root:Average loss=0.0640416862666607 +INFO:root: +[Valid][21][UTT][ MSE = 0.2060 | LCC = 0.9550 | SRCC = 0.9557 ] +INFO:root:Average loss=0.062276922523975375 +INFO:root: +[Valid][22][UTT][ MSE = 0.2402 | LCC = 0.9556 | SRCC = 0.9585 ] +INFO:root:Average loss=0.06012867604196072 +INFO:root: +[Valid][23][UTT][ MSE = 0.2078 | LCC = 0.9559 | SRCC = 0.9571 ] +INFO:root:Average loss=0.06048339974880219 +INFO:root: +[Valid][24][UTT][ MSE = 0.2074 | LCC = 0.9572 | SRCC = 0.9579 ] +INFO:root:Average loss=0.058254021897912026 +INFO:root: +[Valid][25][UTT][ MSE = 0.2321 | LCC = 0.9570 | SRCC = 0.9587 ] +INFO:root:Average loss=0.05594653928279877 +INFO:root: +[Valid][26][UTT][ MSE = 0.2306 | LCC = 0.9562 | SRCC = 0.9566 ] +INFO:root:Average loss=0.05571062672138214 +INFO:root: +[Valid][27][UTT][ MSE = 0.2077 | LCC = 0.9560 | SRCC = 0.9560 ] +INFO:root:Average loss=0.05257274095714092 +INFO:root: +[Valid][28][UTT][ MSE = 0.2378 | LCC = 0.9552 | SRCC = 0.9555 ] +INFO:root:Average loss=0.05149648380279541 +INFO:root: +[Valid][29][UTT][ MSE = 0.2087 | LCC = 0.9567 | SRCC = 0.9569 ] +INFO:root:Average loss=0.04850392358005047 +INFO:root: +[Valid][30][UTT][ MSE = 0.2139 | LCC = 0.9576 | SRCC = 0.9581 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2193 | LCC = 0.9580 | SRCC = 0.9588 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..dec0a433ae7bece1f92a95853803f2198d340275 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9d793f3714b943d173bd594c1264b215ed08b5d0982b1225582a53e57affedf6 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..bbcfae08da44f31b436f2b6cbf8a5a999a2b9fd1 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:23add5f1bf70e8131366ea0dac9c727f901509007ab366159db13d2ef1b8f2be +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..98bc37660361ec8de273067c29d9ad5dd980a78d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.24120024049282074 +INFO:root: +[Valid][0][UTT][ MSE = 0.3270 | LCC = 0.9439 | SRCC = 0.9439 ] +INFO:root:Average loss=0.1521771274805069 +INFO:root: +[Valid][1][UTT][ MSE = 0.2714 | LCC = 0.9486 | SRCC = 0.9497 ] +INFO:root:Average loss=0.13914799058437347 +INFO:root: +[Valid][2][UTT][ MSE = 0.2607 | LCC = 0.9482 | SRCC = 0.9513 ] +INFO:root:Average loss=0.13221525800228118 +INFO:root: +[Valid][3][UTT][ MSE = 0.2744 | LCC = 0.9502 | SRCC = 0.9512 ] +INFO:root:Average loss=0.12864045444130898 +INFO:root: +[Valid][4][UTT][ MSE = 0.2592 | LCC = 0.9519 | SRCC = 0.9531 ] +INFO:root:Average loss=0.12085840484499931 +INFO:root: +[Valid][5][UTT][ MSE = 0.2818 | LCC = 0.9507 | SRCC = 0.9534 ] +INFO:root:Average loss=0.11834758305549621 +INFO:root: +[Valid][6][UTT][ MSE = 0.2649 | LCC = 0.9531 | SRCC = 0.9548 ] +INFO:root:Average loss=0.11553855293989182 +INFO:root: +[Valid][7][UTT][ MSE = 0.2412 | LCC = 0.9537 | SRCC = 0.9550 ] +INFO:root:Average loss=0.11055506157875061 +INFO:root: +[Valid][8][UTT][ MSE = 0.2769 | LCC = 0.9542 | SRCC = 0.9558 ] +INFO:root:Average loss=0.10921399760246277 +INFO:root: +[Valid][9][UTT][ MSE = 0.2511 | LCC = 0.9547 | SRCC = 0.9559 ] +INFO:root:Average loss=0.10074901556968689 +INFO:root: +[Valid][10][UTT][ MSE = 0.2489 | LCC = 0.9544 | SRCC = 0.9557 ] +INFO:root:Average loss=0.10219382816553116 +INFO:root: +[Valid][11][UTT][ MSE = 0.2925 | LCC = 0.9534 | SRCC = 0.9549 ] +INFO:root:Average loss=0.09769243109226226 +INFO:root: +[Valid][12][UTT][ MSE = 0.2329 | LCC = 0.9529 | SRCC = 0.9563 ] +INFO:root:Average loss=0.09599550026655197 +INFO:root: +[Valid][13][UTT][ MSE = 0.2471 | LCC = 0.9530 | SRCC = 0.9541 ] +INFO:root:Average loss=0.09205226457118988 +INFO:root: +[Valid][14][UTT][ MSE = 0.2385 | LCC = 0.9556 | SRCC = 0.9564 ] +INFO:root:Average loss=0.08939030039310455 +INFO:root: +[Valid][15][UTT][ MSE = 0.2580 | LCC = 0.9547 | SRCC = 0.9556 ] +INFO:root:Average loss=0.08824439963698387 +INFO:root: +[Valid][16][UTT][ MSE = 0.2529 | LCC = 0.9541 | SRCC = 0.9554 ] +INFO:root:Average loss=0.08684580737352371 +INFO:root: +[Valid][17][UTT][ MSE = 0.2536 | LCC = 0.9545 | SRCC = 0.9561 ] +INFO:root:Average loss=0.08334383139014244 +INFO:root: +[Valid][18][UTT][ MSE = 0.2317 | LCC = 0.9518 | SRCC = 0.9544 ] +INFO:root:Average loss=0.0820989582836628 +INFO:root: +[Valid][19][UTT][ MSE = 0.2311 | LCC = 0.9530 | SRCC = 0.9534 ] +INFO:root:Average loss=0.0794095609486103 +INFO:root: +[Valid][20][UTT][ MSE = 0.2645 | LCC = 0.9547 | SRCC = 0.9559 ] +INFO:root:Average loss=0.07964213535189629 +INFO:root: +[Valid][21][UTT][ MSE = 0.2005 | LCC = 0.9521 | SRCC = 0.9541 ] +INFO:root:Average loss=0.0767576954960823 +INFO:root: +[Valid][22][UTT][ MSE = 0.2419 | LCC = 0.9536 | SRCC = 0.9554 ] +INFO:root:Average loss=0.07521128517389297 +INFO:root: +[Valid][23][UTT][ MSE = 0.2335 | LCC = 0.9543 | SRCC = 0.9555 ] +INFO:root:Average loss=0.0724043212234974 +INFO:root: +[Valid][24][UTT][ MSE = 0.2618 | LCC = 0.9535 | SRCC = 0.9538 ] +INFO:root:Average loss=0.0708741279244423 +INFO:root: +[Valid][25][UTT][ MSE = 0.2177 | LCC = 0.9548 | SRCC = 0.9563 ] +INFO:root:Average loss=0.07306920313835144 +INFO:root: +[Valid][26][UTT][ MSE = 0.2261 | LCC = 0.9522 | SRCC = 0.9545 ] +INFO:root:Average loss=0.0706908166706562 +INFO:root: +[Valid][27][UTT][ MSE = 0.2454 | LCC = 0.9523 | SRCC = 0.9541 ] +INFO:root:Average loss=0.06727608945965767 +INFO:root: +[Valid][28][UTT][ MSE = 0.2317 | LCC = 0.9565 | SRCC = 0.9570 ] +INFO:root:Average loss=0.06564120271801949 +INFO:root: +[Valid][29][UTT][ MSE = 0.2151 | LCC = 0.9539 | SRCC = 0.9552 ] +INFO:root:Average loss=0.06451748365163804 +INFO:root: +[Valid][30][UTT][ MSE = 0.2347 | LCC = 0.9542 | SRCC = 0.9555 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2372 | LCC = 0.9580 | SRCC = 0.9590 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..905c9db5fa440f0be71c212365d94871f9579ce3 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b8ae8ba77d66ff6597fac8bdb4e534d28e55f056a436c32380696858a14f593 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..204822c0fe1a729ea6d2dac3ca360849b4a14086 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:73d78fd7ba7dc048ffce123332d7e3640f7e09d4d8f604fca45be8153fcde13a +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c01a020a350befec318ec44d2d6c699fbca69bec --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.2843470188379288 +INFO:root: +[Valid][0][UTT][ MSE = 0.3437 | LCC = 0.9423 | SRCC = 0.9417 ] +INFO:root:Average loss=0.1535177664756775 +INFO:root: +[Valid][1][UTT][ MSE = 0.3364 | LCC = 0.9480 | SRCC = 0.9490 ] +INFO:root:Average loss=0.1405117742419243 +INFO:root: +[Valid][2][UTT][ MSE = 0.2944 | LCC = 0.9501 | SRCC = 0.9504 ] +INFO:root:Average loss=0.1330141837000847 +INFO:root: +[Valid][3][UTT][ MSE = 0.2934 | LCC = 0.9510 | SRCC = 0.9518 ] +INFO:root:Average loss=0.128079969227314 +INFO:root: +[Valid][4][UTT][ MSE = 0.2494 | LCC = 0.9483 | SRCC = 0.9523 ] +INFO:root:Average loss=0.12369242405891419 +INFO:root: +[Valid][5][UTT][ MSE = 0.3256 | LCC = 0.9525 | SRCC = 0.9539 ] +INFO:root:Average loss=0.11804235768318176 +INFO:root: +[Valid][6][UTT][ MSE = 0.2533 | LCC = 0.9535 | SRCC = 0.9546 ] +INFO:root:Average loss=0.11327557590603829 +INFO:root: +[Valid][7][UTT][ MSE = 0.2704 | LCC = 0.9531 | SRCC = 0.9548 ] +INFO:root:Average loss=0.11345205289125443 +INFO:root: +[Valid][8][UTT][ MSE = 0.2635 | LCC = 0.9534 | SRCC = 0.9556 ] +INFO:root:Average loss=0.1115402106642723 +INFO:root: +[Valid][9][UTT][ MSE = 0.2666 | LCC = 0.9537 | SRCC = 0.9549 ] +INFO:root:Average loss=0.10670518034696579 +INFO:root: +[Valid][10][UTT][ MSE = 0.2424 | LCC = 0.9532 | SRCC = 0.9545 ] +INFO:root:Average loss=0.10159704023599625 +INFO:root: +[Valid][11][UTT][ MSE = 0.2448 | LCC = 0.9544 | SRCC = 0.9560 ] +INFO:root:Average loss=0.1002401256263256 +INFO:root: +[Valid][12][UTT][ MSE = 0.2899 | LCC = 0.9535 | SRCC = 0.9554 ] +INFO:root:Average loss=0.09571648749709129 +INFO:root: +[Valid][13][UTT][ MSE = 0.2462 | LCC = 0.9547 | SRCC = 0.9560 ] +INFO:root:Average loss=0.09858473533391952 +INFO:root: +[Valid][14][UTT][ MSE = 0.2563 | LCC = 0.9542 | SRCC = 0.9559 ] +INFO:root:Average loss=0.08991972768306732 +INFO:root: +[Valid][15][UTT][ MSE = 0.2336 | LCC = 0.9541 | SRCC = 0.9548 ] +INFO:root:Average loss=0.08975285145640373 +INFO:root: +[Valid][16][UTT][ MSE = 0.2777 | LCC = 0.9548 | SRCC = 0.9560 ] +INFO:root:Average loss=0.0867073692381382 +INFO:root: +[Valid][17][UTT][ MSE = 0.2358 | LCC = 0.9550 | SRCC = 0.9567 ] +INFO:root:Average loss=0.0831231089234352 +INFO:root: +[Valid][18][UTT][ MSE = 0.2488 | LCC = 0.9545 | SRCC = 0.9559 ] +INFO:root:Average loss=0.08091745758056641 +INFO:root: +[Valid][19][UTT][ MSE = 0.2608 | LCC = 0.9545 | SRCC = 0.9556 ] +INFO:root:Average loss=0.08075040748715401 +INFO:root: +[Valid][20][UTT][ MSE = 0.2558 | LCC = 0.9554 | SRCC = 0.9565 ] +INFO:root:Average loss=0.07714841231703758 +INFO:root: +[Valid][21][UTT][ MSE = 0.2752 | LCC = 0.9551 | SRCC = 0.9564 ] +INFO:root:Average loss=0.07632816442847253 +INFO:root: +[Valid][22][UTT][ MSE = 0.2524 | LCC = 0.9530 | SRCC = 0.9556 ] +INFO:root:Average loss=0.07308623301982879 +INFO:root: +[Valid][23][UTT][ MSE = 0.2628 | LCC = 0.9539 | SRCC = 0.9555 ] +INFO:root:Average loss=0.07379822918772698 +INFO:root: +[Valid][24][UTT][ MSE = 0.2833 | LCC = 0.9512 | SRCC = 0.9548 ] +INFO:root:Average loss=0.07250735187530517 +INFO:root: +[Valid][25][UTT][ MSE = 0.2445 | LCC = 0.9533 | SRCC = 0.9540 ] +INFO:root:Average loss=0.07053626410663127 +INFO:root: +[Valid][26][UTT][ MSE = 0.2480 | LCC = 0.9521 | SRCC = 0.9533 ] +INFO:root:Average loss=0.06835156154632568 +INFO:root: +[Valid][27][UTT][ MSE = 0.2544 | LCC = 0.9542 | SRCC = 0.9557 ] +INFO:root:Average loss=0.06687511324882507 +INFO:root: +[Valid][28][UTT][ MSE = 0.2417 | LCC = 0.9551 | SRCC = 0.9562 ] +INFO:root:Average loss=0.06918006202578544 +INFO:root: +[Valid][29][UTT][ MSE = 0.2448 | LCC = 0.9546 | SRCC = 0.9556 ] +INFO:root:Average loss=0.0651599968969822 +INFO:root: +[Valid][30][UTT][ MSE = 0.2364 | LCC = 0.9542 | SRCC = 0.9552 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2621 | LCC = 0.9587 | SRCC = 0.9604 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..307d1a7954ae73bf1647d47781287aca7e309d0c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:001992ea341cb19cdd61dbe036c1237e3a00ad0dc58b259653c5d6462bbbbefd +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a744209e0fa8eb2e97b1b049f63f82aae9a585ed --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:716c767b7a7e7eb0f74adbed892ac19a074acba48aef5b528b1b1b606451b5e4 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3193e792db57dd1a7355d8959f28e092c1ba00ad --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.31915111941099167 +INFO:root: +[Valid][0][UTT][ MSE = 0.2782 | LCC = 0.9391 | SRCC = 0.9432 ] +INFO:root:Average loss=0.1508343113064766 +INFO:root: +[Valid][1][UTT][ MSE = 0.2505 | LCC = 0.9445 | SRCC = 0.9491 ] +INFO:root:Average loss=0.13889614182710647 +INFO:root: +[Valid][2][UTT][ MSE = 0.2652 | LCC = 0.9441 | SRCC = 0.9516 ] +INFO:root:Average loss=0.13646643543243409 +INFO:root: +[Valid][3][UTT][ MSE = 0.2596 | LCC = 0.9514 | SRCC = 0.9530 ] +INFO:root:Average loss=0.1258525252342224 +INFO:root: +[Valid][4][UTT][ MSE = 0.2604 | LCC = 0.9519 | SRCC = 0.9540 ] +INFO:root:Average loss=0.12460966509580612 +INFO:root: +[Valid][5][UTT][ MSE = 0.2444 | LCC = 0.9510 | SRCC = 0.9547 ] +INFO:root:Average loss=0.11820398437976837 +INFO:root: +[Valid][6][UTT][ MSE = 0.2345 | LCC = 0.9531 | SRCC = 0.9543 ] +INFO:root:Average loss=0.11316458588838577 +INFO:root: +[Valid][7][UTT][ MSE = 0.2240 | LCC = 0.9527 | SRCC = 0.9558 ] +INFO:root:Average loss=0.11253042635321617 +INFO:root: +[Valid][8][UTT][ MSE = 0.2460 | LCC = 0.9525 | SRCC = 0.9555 ] +INFO:root:Average loss=0.10853406694531441 +INFO:root: +[Valid][9][UTT][ MSE = 0.2486 | LCC = 0.9531 | SRCC = 0.9558 ] +INFO:root:Average loss=0.10431531608104706 +INFO:root: +[Valid][10][UTT][ MSE = 0.2111 | LCC = 0.9538 | SRCC = 0.9564 ] +INFO:root:Average loss=0.10404213789105415 +INFO:root: +[Valid][11][UTT][ MSE = 0.2407 | LCC = 0.9562 | SRCC = 0.9571 ] +INFO:root:Average loss=0.0967855291068554 +INFO:root: +[Valid][12][UTT][ MSE = 0.2194 | LCC = 0.9552 | SRCC = 0.9566 ] +INFO:root:Average loss=0.09523350673913955 +INFO:root: +[Valid][13][UTT][ MSE = 0.2158 | LCC = 0.9548 | SRCC = 0.9554 ] +INFO:root:Average loss=0.09500109195709229 +INFO:root: +[Valid][14][UTT][ MSE = 0.2644 | LCC = 0.9493 | SRCC = 0.9552 ] +INFO:root:Average loss=0.09188585954904556 +INFO:root: +[Valid][15][UTT][ MSE = 0.2138 | LCC = 0.9530 | SRCC = 0.9552 ] +INFO:root:Average loss=0.0876659876704216 +INFO:root: +[Valid][16][UTT][ MSE = 0.2392 | LCC = 0.9536 | SRCC = 0.9547 ] +INFO:root:Average loss=0.08459763604402543 +INFO:root: +[Valid][17][UTT][ MSE = 0.2322 | LCC = 0.9548 | SRCC = 0.9561 ] +INFO:root:Average loss=0.0857811500132084 +INFO:root: +[Valid][18][UTT][ MSE = 0.2152 | LCC = 0.9553 | SRCC = 0.9567 ] +INFO:root:Average loss=0.08296056166291237 +INFO:root: +[Valid][19][UTT][ MSE = 0.2270 | LCC = 0.9542 | SRCC = 0.9554 ] +INFO:root:Average loss=0.07866793876886367 +INFO:root: +[Valid][20][UTT][ MSE = 0.1920 | LCC = 0.9541 | SRCC = 0.9562 ] +INFO:root:Average loss=0.08170650607347488 +INFO:root: +[Valid][21][UTT][ MSE = 0.2420 | LCC = 0.9528 | SRCC = 0.9553 ] +INFO:root:Average loss=0.07616163608431815 +INFO:root: +[Valid][22][UTT][ MSE = 0.2135 | LCC = 0.9539 | SRCC = 0.9553 ] +INFO:root:Average loss=0.07750520044565201 +INFO:root: +[Valid][23][UTT][ MSE = 0.2304 | LCC = 0.9531 | SRCC = 0.9549 ] +INFO:root:Average loss=0.07317914080619813 +INFO:root: +[Valid][24][UTT][ MSE = 0.2128 | LCC = 0.9550 | SRCC = 0.9559 ] +INFO:root:Average loss=0.07136685261130334 +INFO:root: +[Valid][25][UTT][ MSE = 0.2368 | LCC = 0.9531 | SRCC = 0.9543 ] +INFO:root:Average loss=0.06965050220489502 +INFO:root: +[Valid][26][UTT][ MSE = 0.2221 | LCC = 0.9497 | SRCC = 0.9546 ] +INFO:root:Average loss=0.06959742060303688 +INFO:root: +[Valid][27][UTT][ MSE = 0.2062 | LCC = 0.9543 | SRCC = 0.9561 ] +INFO:root:Average loss=0.06899940186738968 +INFO:root: +[Valid][28][UTT][ MSE = 0.2509 | LCC = 0.9523 | SRCC = 0.9544 ] +INFO:root:Average loss=0.06576290065050125 +INFO:root: +[Valid][29][UTT][ MSE = 0.2250 | LCC = 0.9545 | SRCC = 0.9561 ] +INFO:root:Average loss=0.06472568264603615 +INFO:root: +[Valid][30][UTT][ MSE = 0.1900 | LCC = 0.9530 | SRCC = 0.9553 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2427 | LCC = 0.9585 | SRCC = 0.9599 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1490626b49b540a3c312232858c20c12a05339cb --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7d082b644c3e45768aec980863f726daf01b6fb87fa136e5ee6b58fe99114977 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a0c7df0f60b36906e7ebe651d1190f8958ad8dfc --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:95832a04fad14e3f0c3baf179fafac460f5bdb12a0ffe3493b8cedea4a34e245 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..bbed4cfc41e956cc1b29601237e813f412be78a3 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.2706651109457016 +INFO:root: +[Valid][0][UTT][ MSE = 0.3514 | LCC = 0.9425 | SRCC = 0.9428 ] +INFO:root:Average loss=0.15532790780067443 +INFO:root: +[Valid][1][UTT][ MSE = 0.2985 | LCC = 0.9426 | SRCC = 0.9463 ] +INFO:root:Average loss=0.1447292937040329 +INFO:root: +[Valid][2][UTT][ MSE = 0.2729 | LCC = 0.9469 | SRCC = 0.9506 ] +INFO:root:Average loss=0.13397059792280197 +INFO:root: +[Valid][3][UTT][ MSE = 0.2795 | LCC = 0.9502 | SRCC = 0.9521 ] +INFO:root:Average loss=0.12837685310840607 +INFO:root: +[Valid][4][UTT][ MSE = 0.2892 | LCC = 0.9494 | SRCC = 0.9518 ] +INFO:root:Average loss=0.12232298010587692 +INFO:root: +[Valid][5][UTT][ MSE = 0.2798 | LCC = 0.9510 | SRCC = 0.9526 ] +INFO:root:Average loss=0.12113465049862862 +INFO:root: +[Valid][6][UTT][ MSE = 0.2753 | LCC = 0.9517 | SRCC = 0.9531 ] +INFO:root:Average loss=0.1155844691991806 +INFO:root: +[Valid][7][UTT][ MSE = 0.2778 | LCC = 0.9507 | SRCC = 0.9523 ] +INFO:root:Average loss=0.11662431526184082 +INFO:root: +[Valid][8][UTT][ MSE = 0.2863 | LCC = 0.9499 | SRCC = 0.9540 ] +INFO:root:Average loss=0.10904685169458389 +INFO:root: +[Valid][9][UTT][ MSE = 0.2583 | LCC = 0.9486 | SRCC = 0.9531 ] +INFO:root:Average loss=0.10567890411615372 +INFO:root: +[Valid][10][UTT][ MSE = 0.2479 | LCC = 0.9501 | SRCC = 0.9533 ] +INFO:root:Average loss=0.10444077175855637 +INFO:root: +[Valid][11][UTT][ MSE = 0.2634 | LCC = 0.9517 | SRCC = 0.9547 ] +INFO:root:Average loss=0.09995058622956277 +INFO:root: +[Valid][12][UTT][ MSE = 0.2863 | LCC = 0.9508 | SRCC = 0.9545 ] +INFO:root:Average loss=0.09804367500543594 +INFO:root: +[Valid][13][UTT][ MSE = 0.2403 | LCC = 0.9528 | SRCC = 0.9550 ] +INFO:root:Average loss=0.0993726597726345 +INFO:root: +[Valid][14][UTT][ MSE = 0.2401 | LCC = 0.9524 | SRCC = 0.9544 ] +INFO:root:Average loss=0.09472164112329483 +INFO:root: +[Valid][15][UTT][ MSE = 0.2808 | LCC = 0.9526 | SRCC = 0.9547 ] +INFO:root:Average loss=0.08799813967943192 +INFO:root: +[Valid][16][UTT][ MSE = 0.2242 | LCC = 0.9520 | SRCC = 0.9542 ] +INFO:root:Average loss=0.08666343128681182 +INFO:root: +[Valid][17][UTT][ MSE = 0.2937 | LCC = 0.9527 | SRCC = 0.9540 ] +INFO:root:Average loss=0.08685955074429512 +INFO:root: +[Valid][18][UTT][ MSE = 0.2721 | LCC = 0.9523 | SRCC = 0.9535 ] +INFO:root:Average loss=0.08450204578042031 +INFO:root: +[Valid][19][UTT][ MSE = 0.2348 | LCC = 0.9527 | SRCC = 0.9551 ] +INFO:root:Average loss=0.08157676449418068 +INFO:root: +[Valid][20][UTT][ MSE = 0.2459 | LCC = 0.9534 | SRCC = 0.9541 ] +INFO:root:Average loss=0.08030714693665504 +INFO:root: +[Valid][21][UTT][ MSE = 0.2598 | LCC = 0.9521 | SRCC = 0.9536 ] +INFO:root:Average loss=0.07775619778037071 +INFO:root: +[Valid][22][UTT][ MSE = 0.2338 | LCC = 0.9534 | SRCC = 0.9546 ] +INFO:root:Average loss=0.0752270772755146 +INFO:root: +[Valid][23][UTT][ MSE = 0.2398 | LCC = 0.9515 | SRCC = 0.9536 ] +INFO:root:Average loss=0.07193093219399452 +INFO:root: +[Valid][24][UTT][ MSE = 0.2389 | LCC = 0.9519 | SRCC = 0.9528 ] +INFO:root:Average loss=0.07622614440321922 +INFO:root: +[Valid][25][UTT][ MSE = 0.2352 | LCC = 0.9467 | SRCC = 0.9532 ] +INFO:root:Average loss=0.07262670981884002 +INFO:root: +[Valid][26][UTT][ MSE = 0.2286 | LCC = 0.9511 | SRCC = 0.9544 ] +INFO:root:Average loss=0.06932133316993713 +INFO:root: +[Valid][27][UTT][ MSE = 0.2441 | LCC = 0.9528 | SRCC = 0.9538 ] +INFO:root:Average loss=0.06862735772132873 +INFO:root: +[Valid][28][UTT][ MSE = 0.2653 | LCC = 0.9521 | SRCC = 0.9527 ] +INFO:root:Average loss=0.06641999724507332 +INFO:root: +[Valid][29][UTT][ MSE = 0.2868 | LCC = 0.9519 | SRCC = 0.9532 ] +INFO:root:Average loss=0.06654781770706177 +INFO:root: +[Valid][30][UTT][ MSE = 0.2242 | LCC = 0.9516 | SRCC = 0.9530 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2475 | LCC = 0.9581 | SRCC = 0.9593 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d525bb19c72c93f44b0959c225d6fc5c2d0ff648 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fa0c92f16e449bf302b0362d68cd56cff9a8a8dc5daabacad9aaa9066622b999 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b740cfa9a39fc3180535c36b21168ad1c0b4c26f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6bf0b699e580cff035a8b0d545a95b6cb6ac71c96755f2f0916b31d28292218f +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9854c9fad5c822c0000db62daabbcf8cad3c20a7 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_10epoch/layer9/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.26453088903427124 +INFO:root: +[Valid][0][UTT][ MSE = 0.2948 | LCC = 0.9450 | SRCC = 0.9453 ] +INFO:root:Average loss=0.15134877663850785 +INFO:root: +[Valid][1][UTT][ MSE = 0.3176 | LCC = 0.9471 | SRCC = 0.9476 ] +INFO:root:Average loss=0.14720772653818132 +INFO:root: +[Valid][2][UTT][ MSE = 0.2789 | LCC = 0.9473 | SRCC = 0.9506 ] +INFO:root:Average loss=0.13081148636341094 +INFO:root: +[Valid][3][UTT][ MSE = 0.2707 | LCC = 0.9507 | SRCC = 0.9524 ] +INFO:root:Average loss=0.1254156795144081 +INFO:root: +[Valid][4][UTT][ MSE = 0.2510 | LCC = 0.9513 | SRCC = 0.9540 ] +INFO:root:Average loss=0.12379711943864823 +INFO:root: +[Valid][5][UTT][ MSE = 0.2502 | LCC = 0.9529 | SRCC = 0.9550 ] +INFO:root:Average loss=0.11699004429578781 +INFO:root: +[Valid][6][UTT][ MSE = 0.2405 | LCC = 0.9534 | SRCC = 0.9550 ] +INFO:root:Average loss=0.11342701196670532 +INFO:root: +[Valid][7][UTT][ MSE = 0.2974 | LCC = 0.9524 | SRCC = 0.9549 ] +INFO:root:Average loss=0.11203417068719863 +INFO:root: +[Valid][8][UTT][ MSE = 0.2610 | LCC = 0.9514 | SRCC = 0.9534 ] +INFO:root:Average loss=0.10871828997135162 +INFO:root: +[Valid][9][UTT][ MSE = 0.2652 | LCC = 0.9537 | SRCC = 0.9552 ] +INFO:root:Average loss=0.10623682054877281 +INFO:root: +[Valid][10][UTT][ MSE = 0.2466 | LCC = 0.9542 | SRCC = 0.9557 ] +INFO:root:Average loss=0.10235799542069435 +INFO:root: +[Valid][11][UTT][ MSE = 0.2347 | LCC = 0.9526 | SRCC = 0.9554 ] +INFO:root:Average loss=0.09603022691607475 +INFO:root: +[Valid][12][UTT][ MSE = 0.2423 | LCC = 0.9550 | SRCC = 0.9560 ] +INFO:root:Average loss=0.09845858982205391 +INFO:root: +[Valid][13][UTT][ MSE = 0.2347 | LCC = 0.9551 | SRCC = 0.9558 ] +INFO:root:Average loss=0.09461126554012299 +INFO:root: +[Valid][14][UTT][ MSE = 0.2201 | LCC = 0.9546 | SRCC = 0.9560 ] +INFO:root:Average loss=0.09043939086794853 +INFO:root: +[Valid][15][UTT][ MSE = 0.2126 | LCC = 0.9546 | SRCC = 0.9555 ] +INFO:root:Average loss=0.08892817333340645 +INFO:root: +[Valid][16][UTT][ MSE = 0.2148 | LCC = 0.9559 | SRCC = 0.9570 ] +INFO:root:Average loss=0.08517366334795952 +INFO:root: +[Valid][17][UTT][ MSE = 0.2324 | LCC = 0.9551 | SRCC = 0.9562 ] +INFO:root:Average loss=0.08534333297610283 +INFO:root: +[Valid][18][UTT][ MSE = 0.2185 | LCC = 0.9554 | SRCC = 0.9556 ] +INFO:root:Average loss=0.08250140458345413 +INFO:root: +[Valid][19][UTT][ MSE = 0.2281 | LCC = 0.9541 | SRCC = 0.9563 ] +INFO:root:Average loss=0.08342384445667267 +INFO:root: +[Valid][20][UTT][ MSE = 0.2177 | LCC = 0.9543 | SRCC = 0.9552 ] +INFO:root:Average loss=0.08226179400086403 +INFO:root: +[Valid][21][UTT][ MSE = 0.2357 | LCC = 0.9542 | SRCC = 0.9559 ] +INFO:root:Average loss=0.07728950923681259 +INFO:root: +[Valid][22][UTT][ MSE = 0.2242 | LCC = 0.9534 | SRCC = 0.9554 ] +INFO:root:Average loss=0.07330785426497459 +INFO:root: +[Valid][23][UTT][ MSE = 0.2324 | LCC = 0.9542 | SRCC = 0.9549 ] +INFO:root:Average loss=0.0735078575015068 +INFO:root: +[Valid][24][UTT][ MSE = 0.2080 | LCC = 0.9545 | SRCC = 0.9562 ] +INFO:root:Average loss=0.07243457865715026 +INFO:root: +[Valid][25][UTT][ MSE = 0.2375 | LCC = 0.9536 | SRCC = 0.9551 ] +INFO:root:Average loss=0.0703130792081356 +INFO:root: +[Valid][26][UTT][ MSE = 0.2458 | LCC = 0.9545 | SRCC = 0.9555 ] +INFO:root:Average loss=0.0658481949865818 +INFO:root: +[Valid][27][UTT][ MSE = 0.2157 | LCC = 0.9538 | SRCC = 0.9551 ] +INFO:root:Average loss=0.06988300895690917 +INFO:root: +[Valid][28][UTT][ MSE = 0.2491 | LCC = 0.9539 | SRCC = 0.9556 ] +INFO:root:Average loss=0.06867827516794205 +INFO:root: +[Valid][29][UTT][ MSE = 0.2208 | LCC = 0.9540 | SRCC = 0.9551 ] +INFO:root:Average loss=0.06402230873703957 +INFO:root: +[Valid][30][UTT][ MSE = 0.2442 | LCC = 0.9545 | SRCC = 0.9561 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2151 | LCC = 0.9590 | SRCC = 0.9598 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0cc0a75d9ebb206f9d93167232b50cc8accabf40 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ee7db30d9f9e8c9751cf6f5a91fcac38863e804553c754b0a3cf4b3f4244be88 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e6e5aa890641feb84393a0457c50c6c488b9d368 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8d1abba0e191b30ec351c2a68826bf20a7f74a7110bcb5324a63369877b51754 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a8c4530cf9cbfe7fd5e29f347617d6c880c4e44f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5823194668292999 +INFO:root: +[Valid][0][UTT][ MSE = 0.4426 | LCC = 0.8883 | SRCC = 0.8873 ] +INFO:root:Average loss=0.26784711599349975 +INFO:root: +[Valid][1][UTT][ MSE = 0.3208 | LCC = 0.9188 | SRCC = 0.9202 ] +INFO:root:Average loss=0.20783009624481202 +INFO:root: +[Valid][2][UTT][ MSE = 0.2700 | LCC = 0.9232 | SRCC = 0.9237 ] +INFO:root:Average loss=0.19004937255382537 +INFO:root: +[Valid][3][UTT][ MSE = 0.2942 | LCC = 0.9306 | SRCC = 0.9330 ] +INFO:root:Average loss=0.16520969915390016 +INFO:root: +[Valid][4][UTT][ MSE = 0.2734 | LCC = 0.9357 | SRCC = 0.9363 ] +INFO:root:Average loss=0.1535733545422554 +INFO:root: +[Valid][5][UTT][ MSE = 0.2607 | LCC = 0.9407 | SRCC = 0.9422 ] +INFO:root:Average loss=0.13603111702203752 +INFO:root: +[Valid][6][UTT][ MSE = 0.2445 | LCC = 0.9394 | SRCC = 0.9411 ] +INFO:root:Average loss=0.12928756177425385 +INFO:root: +[Valid][7][UTT][ MSE = 0.2432 | LCC = 0.9405 | SRCC = 0.9418 ] +INFO:root:Average loss=0.1234587515592575 +INFO:root: +[Valid][8][UTT][ MSE = 0.2350 | LCC = 0.9475 | SRCC = 0.9488 ] +INFO:root:Average loss=0.11372279405593871 +INFO:root: +[Valid][9][UTT][ MSE = 0.2095 | LCC = 0.9486 | SRCC = 0.9498 ] +INFO:root:Average loss=0.10762161809206008 +INFO:root: +[Valid][10][UTT][ MSE = 0.2225 | LCC = 0.9483 | SRCC = 0.9498 ] +INFO:root:Average loss=0.10598676961660385 +INFO:root: +[Valid][11][UTT][ MSE = 0.2545 | LCC = 0.9435 | SRCC = 0.9451 ] +INFO:root:Average loss=0.10083114260435104 +INFO:root: +[Valid][12][UTT][ MSE = 0.1983 | LCC = 0.9497 | SRCC = 0.9509 ] +INFO:root:Average loss=0.09154768785834312 +INFO:root: +[Valid][13][UTT][ MSE = 0.2050 | LCC = 0.9500 | SRCC = 0.9511 ] +INFO:root:Average loss=0.08883169561624527 +INFO:root: +[Valid][14][UTT][ MSE = 0.2049 | LCC = 0.9520 | SRCC = 0.9529 ] +INFO:root:Average loss=0.0876413779258728 +INFO:root: +[Valid][15][UTT][ MSE = 0.2353 | LCC = 0.9525 | SRCC = 0.9536 ] +INFO:root:Average loss=0.08354928717017174 +INFO:root: +[Valid][16][UTT][ MSE = 0.2125 | LCC = 0.9526 | SRCC = 0.9537 ] +INFO:root:Average loss=0.07813201588392257 +INFO:root: +[Valid][17][UTT][ MSE = 0.2053 | LCC = 0.9513 | SRCC = 0.9524 ] +INFO:root:Average loss=0.0771672083735466 +INFO:root: +[Valid][18][UTT][ MSE = 0.2041 | LCC = 0.9523 | SRCC = 0.9535 ] +INFO:root:Average loss=0.07520710563659667 +INFO:root: +[Valid][19][UTT][ MSE = 0.1865 | LCC = 0.9505 | SRCC = 0.9523 ] +INFO:root:Average loss=0.07275090584158897 +INFO:root: +[Valid][20][UTT][ MSE = 0.2292 | LCC = 0.9534 | SRCC = 0.9546 ] +INFO:root:Average loss=0.0694021272957325 +INFO:root: +[Valid][21][UTT][ MSE = 0.2210 | LCC = 0.9526 | SRCC = 0.9542 ] +INFO:root:Average loss=0.0669014522433281 +INFO:root: +[Valid][22][UTT][ MSE = 0.2149 | LCC = 0.9549 | SRCC = 0.9562 ] +INFO:root:Average loss=0.06518024915456772 +INFO:root: +[Valid][23][UTT][ MSE = 0.1984 | LCC = 0.9517 | SRCC = 0.9541 ] +INFO:root:Average loss=0.06568436616659165 +INFO:root: +[Valid][24][UTT][ MSE = 0.2069 | LCC = 0.9549 | SRCC = 0.9562 ] +INFO:root:Average loss=0.05995625594258308 +INFO:root: +[Valid][25][UTT][ MSE = 0.1890 | LCC = 0.9536 | SRCC = 0.9560 ] +INFO:root:Average loss=0.061530302807688717 +INFO:root: +[Valid][26][UTT][ MSE = 0.1978 | LCC = 0.9566 | SRCC = 0.9583 ] +INFO:root:Average loss=0.05811383029818535 +INFO:root: +[Valid][27][UTT][ MSE = 0.1949 | LCC = 0.9545 | SRCC = 0.9565 ] +INFO:root:Average loss=0.05884148758649826 +INFO:root: +[Valid][28][UTT][ MSE = 0.1947 | LCC = 0.9552 | SRCC = 0.9563 ] +INFO:root:Average loss=0.05378641359508038 +INFO:root: +[Valid][29][UTT][ MSE = 0.1935 | LCC = 0.9551 | SRCC = 0.9566 ] +INFO:root:Average loss=0.05417945420742035 +INFO:root: +[Valid][30][UTT][ MSE = 0.1960 | LCC = 0.9554 | SRCC = 0.9573 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2055 | LCC = 0.9566 | SRCC = 0.9572 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d299dea677339a69d55bb1da780c43569336ada3 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a01afd1cf48075c40b95931038374dbedf88c20d92fb4a1e16d441a9616ed95 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d40507a1d3b5f1da0467b28ba64cd48f0338da1d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4a146edccfb03a2cb7720f60a41527bd0c41230284d51896e160ca2c06b6a7e2 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6e22c1a1ae1eb0663c485f00b874aabf2d9617a8 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5543786433935165 +INFO:root: +[Valid][0][UTT][ MSE = 0.3695 | LCC = 0.8895 | SRCC = 0.8906 ] +INFO:root:Average loss=0.2489934574365616 +INFO:root: +[Valid][1][UTT][ MSE = 0.2919 | LCC = 0.9216 | SRCC = 0.9246 ] +INFO:root:Average loss=0.19947820281982423 +INFO:root: +[Valid][2][UTT][ MSE = 0.2446 | LCC = 0.9276 | SRCC = 0.9330 ] +INFO:root:Average loss=0.17555677086114885 +INFO:root: +[Valid][3][UTT][ MSE = 0.2376 | LCC = 0.9265 | SRCC = 0.9314 ] +INFO:root:Average loss=0.15670201766490938 +INFO:root: +[Valid][4][UTT][ MSE = 0.2295 | LCC = 0.9356 | SRCC = 0.9387 ] +INFO:root:Average loss=0.1446509209871292 +INFO:root: +[Valid][5][UTT][ MSE = 0.2511 | LCC = 0.9420 | SRCC = 0.9449 ] +INFO:root:Average loss=0.13136403024196625 +INFO:root: +[Valid][6][UTT][ MSE = 0.2574 | LCC = 0.9456 | SRCC = 0.9480 ] +INFO:root:Average loss=0.12397407341003418 +INFO:root: +[Valid][7][UTT][ MSE = 0.2263 | LCC = 0.9449 | SRCC = 0.9467 ] +INFO:root:Average loss=0.1166672521829605 +INFO:root: +[Valid][8][UTT][ MSE = 0.2142 | LCC = 0.9471 | SRCC = 0.9489 ] +INFO:root:Average loss=0.11163745951652528 +INFO:root: +[Valid][9][UTT][ MSE = 0.2204 | LCC = 0.9496 | SRCC = 0.9510 ] +INFO:root:Average loss=0.10773501181602478 +INFO:root: +[Valid][10][UTT][ MSE = 0.2149 | LCC = 0.9478 | SRCC = 0.9489 ] +INFO:root:Average loss=0.0969874062538147 +INFO:root: +[Valid][11][UTT][ MSE = 0.2462 | LCC = 0.9503 | SRCC = 0.9526 ] +INFO:root:Average loss=0.09797781401872635 +INFO:root: +[Valid][12][UTT][ MSE = 0.2228 | LCC = 0.9500 | SRCC = 0.9514 ] +INFO:root:Average loss=0.08919251716136932 +INFO:root: +[Valid][13][UTT][ MSE = 0.1935 | LCC = 0.9467 | SRCC = 0.9492 ] +INFO:root:Average loss=0.08200861611962318 +INFO:root: +[Valid][14][UTT][ MSE = 0.2160 | LCC = 0.9510 | SRCC = 0.9527 ] +INFO:root:Average loss=0.08025262799859047 +INFO:root: +[Valid][15][UTT][ MSE = 0.2019 | LCC = 0.9481 | SRCC = 0.9494 ] +INFO:root:Average loss=0.07888882291316986 +INFO:root: +[Valid][16][UTT][ MSE = 0.1914 | LCC = 0.9521 | SRCC = 0.9536 ] +INFO:root:Average loss=0.08142725083231926 +INFO:root: +[Valid][17][UTT][ MSE = 0.2163 | LCC = 0.9488 | SRCC = 0.9502 ] +INFO:root:Average loss=0.07525500518083572 +INFO:root: +[Valid][18][UTT][ MSE = 0.2134 | LCC = 0.9521 | SRCC = 0.9533 ] +INFO:root:Average loss=0.07186941328644753 +INFO:root: +[Valid][19][UTT][ MSE = 0.2144 | LCC = 0.9516 | SRCC = 0.9535 ] +INFO:root:Average loss=0.0667783541381359 +INFO:root: +[Valid][20][UTT][ MSE = 0.2256 | LCC = 0.9489 | SRCC = 0.9504 ] +INFO:root:Average loss=0.06509316846728325 +INFO:root: +[Valid][21][UTT][ MSE = 0.1943 | LCC = 0.9513 | SRCC = 0.9526 ] +INFO:root:Average loss=0.06429456168413163 +INFO:root: +[Valid][22][UTT][ MSE = 0.1938 | LCC = 0.9529 | SRCC = 0.9542 ] +INFO:root:Average loss=0.061296018183231354 +INFO:root: +[Valid][23][UTT][ MSE = 0.1948 | LCC = 0.9519 | SRCC = 0.9530 ] +INFO:root:Average loss=0.060742075979709625 +INFO:root: +[Valid][24][UTT][ MSE = 0.2040 | LCC = 0.9523 | SRCC = 0.9543 ] +INFO:root:Average loss=0.05860231375694275 +INFO:root: +[Valid][25][UTT][ MSE = 0.2235 | LCC = 0.9533 | SRCC = 0.9553 ] +INFO:root:Average loss=0.05827286215126514 +INFO:root: +[Valid][26][UTT][ MSE = 0.1913 | LCC = 0.9535 | SRCC = 0.9556 ] +INFO:root:Average loss=0.05546455723047256 +INFO:root: +[Valid][27][UTT][ MSE = 0.1901 | LCC = 0.9497 | SRCC = 0.9510 ] +INFO:root:Average loss=0.05375006610155106 +INFO:root: +[Valid][28][UTT][ MSE = 0.1949 | LCC = 0.9544 | SRCC = 0.9560 ] +INFO:root:Average loss=0.05316811202466488 +INFO:root: +[Valid][29][UTT][ MSE = 0.2038 | LCC = 0.9541 | SRCC = 0.9553 ] +INFO:root:Average loss=0.05141941243410111 +INFO:root: +[Valid][30][UTT][ MSE = 0.1922 | LCC = 0.9530 | SRCC = 0.9549 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1893 | LCC = 0.9584 | SRCC = 0.9588 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..61a45e18d180d2931c1ab608bb4244f24de49c96 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:31dda9e292bbfe7f1651f1fdf379678e74203dccb302f52a7295db1ca188d791 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..663325ad9d12f29be6ab3af8388e5d378fc7ccf3 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6815bc98d5ab9660e52f81e630b2bc45eb201196afa994a0d09f0e48600629c0 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0a3c82e4399b924a4d665424952cac67878c59a7 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.527700179219246 +INFO:root: +[Valid][0][UTT][ MSE = 0.3254 | LCC = 0.9023 | SRCC = 0.9058 ] +INFO:root:Average loss=0.24617263507843018 +INFO:root: +[Valid][1][UTT][ MSE = 0.2951 | LCC = 0.9220 | SRCC = 0.9237 ] +INFO:root:Average loss=0.19726095491647722 +INFO:root: +[Valid][2][UTT][ MSE = 0.2583 | LCC = 0.9247 | SRCC = 0.9287 ] +INFO:root:Average loss=0.1701278772354126 +INFO:root: +[Valid][3][UTT][ MSE = 0.2450 | LCC = 0.9362 | SRCC = 0.9378 ] +INFO:root:Average loss=0.15749092304706575 +INFO:root: +[Valid][4][UTT][ MSE = 0.2247 | LCC = 0.9418 | SRCC = 0.9427 ] +INFO:root:Average loss=0.1420180374979973 +INFO:root: +[Valid][5][UTT][ MSE = 0.2033 | LCC = 0.9409 | SRCC = 0.9438 ] +INFO:root:Average loss=0.1370375542640686 +INFO:root: +[Valid][6][UTT][ MSE = 0.2054 | LCC = 0.9453 | SRCC = 0.9462 ] +INFO:root:Average loss=0.12705052667856218 +INFO:root: +[Valid][7][UTT][ MSE = 0.2239 | LCC = 0.9432 | SRCC = 0.9448 ] +INFO:root:Average loss=0.11777202153205872 +INFO:root: +[Valid][8][UTT][ MSE = 0.2074 | LCC = 0.9464 | SRCC = 0.9484 ] +INFO:root:Average loss=0.11012624233961105 +INFO:root: +[Valid][9][UTT][ MSE = 0.2212 | LCC = 0.9478 | SRCC = 0.9489 ] +INFO:root:Average loss=0.1041307349205017 +INFO:root: +[Valid][10][UTT][ MSE = 0.2234 | LCC = 0.9505 | SRCC = 0.9511 ] +INFO:root:Average loss=0.10118119379878045 +INFO:root: +[Valid][11][UTT][ MSE = 0.1941 | LCC = 0.9515 | SRCC = 0.9530 ] +INFO:root:Average loss=0.09690193429589272 +INFO:root: +[Valid][12][UTT][ MSE = 0.2071 | LCC = 0.9517 | SRCC = 0.9527 ] +INFO:root:Average loss=0.09453065568208695 +INFO:root: +[Valid][13][UTT][ MSE = 0.1986 | LCC = 0.9534 | SRCC = 0.9544 ] +INFO:root:Average loss=0.08537901586294175 +INFO:root: +[Valid][14][UTT][ MSE = 0.1962 | LCC = 0.9523 | SRCC = 0.9536 ] +INFO:root:Average loss=0.08522356876730919 +INFO:root: +[Valid][15][UTT][ MSE = 0.1789 | LCC = 0.9534 | SRCC = 0.9544 ] +INFO:root:Average loss=0.07893284004926682 +INFO:root: +[Valid][16][UTT][ MSE = 0.2114 | LCC = 0.9517 | SRCC = 0.9526 ] +INFO:root:Average loss=0.07915207555890083 +INFO:root: +[Valid][17][UTT][ MSE = 0.1675 | LCC = 0.9533 | SRCC = 0.9544 ] +INFO:root:Average loss=0.07264740830659866 +INFO:root: +[Valid][18][UTT][ MSE = 0.1926 | LCC = 0.9533 | SRCC = 0.9547 ] +INFO:root:Average loss=0.0726307370364666 +INFO:root: +[Valid][19][UTT][ MSE = 0.1749 | LCC = 0.9550 | SRCC = 0.9560 ] +INFO:root:Average loss=0.07040416541695595 +INFO:root: +[Valid][20][UTT][ MSE = 0.2061 | LCC = 0.9570 | SRCC = 0.9577 ] +INFO:root:Average loss=0.06809036514163017 +INFO:root: +[Valid][21][UTT][ MSE = 0.1785 | LCC = 0.9564 | SRCC = 0.9576 ] +INFO:root:Average loss=0.06581580376625061 +INFO:root: +[Valid][22][UTT][ MSE = 0.1851 | LCC = 0.9549 | SRCC = 0.9554 ] +INFO:root:Average loss=0.06399893274903297 +INFO:root: +[Valid][23][UTT][ MSE = 0.1806 | LCC = 0.9565 | SRCC = 0.9577 ] +INFO:root:Average loss=0.06519667470455169 +INFO:root: +[Valid][24][UTT][ MSE = 0.2049 | LCC = 0.9566 | SRCC = 0.9574 ] +INFO:root:Average loss=0.06024394407868385 +INFO:root: +[Valid][25][UTT][ MSE = 0.1927 | LCC = 0.9568 | SRCC = 0.9582 ] +INFO:root:Average loss=0.055139185309410095 +INFO:root: +[Valid][26][UTT][ MSE = 0.2085 | LCC = 0.9561 | SRCC = 0.9564 ] +INFO:root:Average loss=0.05579034033417702 +INFO:root: +[Valid][27][UTT][ MSE = 0.1714 | LCC = 0.9579 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05616809040307999 +INFO:root: +[Valid][28][UTT][ MSE = 0.1890 | LCC = 0.9564 | SRCC = 0.9572 ] +INFO:root:Average loss=0.05250369779765606 +INFO:root: +[Valid][29][UTT][ MSE = 0.1812 | LCC = 0.9561 | SRCC = 0.9571 ] +INFO:root:Average loss=0.05333210316300392 +INFO:root: +[Valid][30][UTT][ MSE = 0.1869 | LCC = 0.9580 | SRCC = 0.9584 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1801 | LCC = 0.9598 | SRCC = 0.9604 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3084a5ad24376ba7cc00f5a2c082a460e2b26dab --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:05e6c20c2edebb92652374687489c8ec357b578e34923b3cdee5da8bb22f01f9 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..13e579cced42971df88f5ff3a6769df32893808a --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0eb30e61dff0e59af0f8614eca425456df7bd046d64232822b16ff1c0064a833 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3a98bf06f89b643b8d55d56ebbac659962617f40 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5409844946861267 +INFO:root: +[Valid][0][UTT][ MSE = 0.3315 | LCC = 0.8997 | SRCC = 0.9041 ] +INFO:root:Average loss=0.23768842792510986 +INFO:root: +[Valid][1][UTT][ MSE = 0.3484 | LCC = 0.9257 | SRCC = 0.9273 ] +INFO:root:Average loss=0.19648792874813079 +INFO:root: +[Valid][2][UTT][ MSE = 0.2757 | LCC = 0.9302 | SRCC = 0.9315 ] +INFO:root:Average loss=0.1851567855477333 +INFO:root: +[Valid][3][UTT][ MSE = 0.2764 | LCC = 0.9353 | SRCC = 0.9371 ] +INFO:root:Average loss=0.16410772812366486 +INFO:root: +[Valid][4][UTT][ MSE = 0.2405 | LCC = 0.9383 | SRCC = 0.9405 ] +INFO:root:Average loss=0.14571646082401277 +INFO:root: +[Valid][5][UTT][ MSE = 0.2455 | LCC = 0.9413 | SRCC = 0.9435 ] +INFO:root:Average loss=0.13535246527194977 +INFO:root: +[Valid][6][UTT][ MSE = 0.2180 | LCC = 0.9454 | SRCC = 0.9467 ] +INFO:root:Average loss=0.12529711616039277 +INFO:root: +[Valid][7][UTT][ MSE = 0.2291 | LCC = 0.9469 | SRCC = 0.9475 ] +INFO:root:Average loss=0.12336722046136855 +INFO:root: +[Valid][8][UTT][ MSE = 0.2050 | LCC = 0.9441 | SRCC = 0.9489 ] +INFO:root:Average loss=0.11390282195806503 +INFO:root: +[Valid][9][UTT][ MSE = 0.1973 | LCC = 0.9470 | SRCC = 0.9506 ] +INFO:root:Average loss=0.10596581956744194 +INFO:root: +[Valid][10][UTT][ MSE = 0.2485 | LCC = 0.9475 | SRCC = 0.9485 ] +INFO:root:Average loss=0.10216250681877136 +INFO:root: +[Valid][11][UTT][ MSE = 0.2248 | LCC = 0.9492 | SRCC = 0.9509 ] +INFO:root:Average loss=0.09293656671047211 +INFO:root: +[Valid][12][UTT][ MSE = 0.2407 | LCC = 0.9492 | SRCC = 0.9505 ] +INFO:root:Average loss=0.09160766413807869 +INFO:root: +[Valid][13][UTT][ MSE = 0.2196 | LCC = 0.9526 | SRCC = 0.9532 ] +INFO:root:Average loss=0.08982816761732101 +INFO:root: +[Valid][14][UTT][ MSE = 0.2430 | LCC = 0.9505 | SRCC = 0.9515 ] +INFO:root:Average loss=0.08542284914851189 +INFO:root: +[Valid][15][UTT][ MSE = 0.2495 | LCC = 0.9502 | SRCC = 0.9517 ] +INFO:root:Average loss=0.07976681712269783 +INFO:root: +[Valid][16][UTT][ MSE = 0.2046 | LCC = 0.9524 | SRCC = 0.9531 ] +INFO:root:Average loss=0.07748889604210854 +INFO:root: +[Valid][17][UTT][ MSE = 0.2185 | LCC = 0.9526 | SRCC = 0.9535 ] +INFO:root:Average loss=0.0773507530093193 +INFO:root: +[Valid][18][UTT][ MSE = 0.2084 | LCC = 0.9511 | SRCC = 0.9528 ] +INFO:root:Average loss=0.07420722183585167 +INFO:root: +[Valid][19][UTT][ MSE = 0.2172 | LCC = 0.9524 | SRCC = 0.9535 ] +INFO:root:Average loss=0.07277760306000709 +INFO:root: +[Valid][20][UTT][ MSE = 0.1972 | LCC = 0.9539 | SRCC = 0.9551 ] +INFO:root:Average loss=0.0664137302339077 +INFO:root: +[Valid][21][UTT][ MSE = 0.1981 | LCC = 0.9548 | SRCC = 0.9558 ] +INFO:root:Average loss=0.06823394784331321 +INFO:root: +[Valid][22][UTT][ MSE = 0.1946 | LCC = 0.9549 | SRCC = 0.9559 ] +INFO:root:Average loss=0.06166636618971825 +INFO:root: +[Valid][23][UTT][ MSE = 0.2246 | LCC = 0.9544 | SRCC = 0.9551 ] +INFO:root:Average loss=0.06094460636377334 +INFO:root: +[Valid][24][UTT][ MSE = 0.2140 | LCC = 0.9535 | SRCC = 0.9545 ] +INFO:root:Average loss=0.06215003603696823 +INFO:root: +[Valid][25][UTT][ MSE = 0.2059 | LCC = 0.9543 | SRCC = 0.9548 ] +INFO:root:Average loss=0.059016861230134963 +INFO:root: +[Valid][26][UTT][ MSE = 0.1804 | LCC = 0.9528 | SRCC = 0.9546 ] +INFO:root:Average loss=0.053498825281858446 +INFO:root: +[Valid][27][UTT][ MSE = 0.2178 | LCC = 0.9540 | SRCC = 0.9549 ] +INFO:root:Average loss=0.0564657197445631 +INFO:root: +[Valid][28][UTT][ MSE = 0.1961 | LCC = 0.9548 | SRCC = 0.9555 ] +INFO:root:Average loss=0.05502486841380596 +INFO:root: +[Valid][29][UTT][ MSE = 0.2300 | LCC = 0.9531 | SRCC = 0.9543 ] +INFO:root:Average loss=0.05146248510479927 +INFO:root: +[Valid][30][UTT][ MSE = 0.1953 | LCC = 0.9539 | SRCC = 0.9545 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1947 | LCC = 0.9577 | SRCC = 0.9585 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f7bb86880cee5972d2449f4503f20a715ffbf06c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af972c7eafc4f75cf4706eeda979059a4f2cccb5182088ec948b3fd0d6084dab +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..147d83ca3a2ad663c40707bd30325a3c1159294c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9370fd4f2f85833f5ded1c921dccaa09c490848f7f92beb9f42309d099daf876 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0c4550dcd3a1f146800abec7296cab89a12f1602 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer0/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.557819393157959 +INFO:root: +[Valid][0][UTT][ MSE = 0.3764 | LCC = 0.9002 | SRCC = 0.9007 ] +INFO:root:Average loss=0.24769192636013032 +INFO:root: +[Valid][1][UTT][ MSE = 0.3317 | LCC = 0.9196 | SRCC = 0.9203 ] +INFO:root:Average loss=0.19951852703094483 +INFO:root: +[Valid][2][UTT][ MSE = 0.3126 | LCC = 0.9273 | SRCC = 0.9307 ] +INFO:root:Average loss=0.18255970305204391 +INFO:root: +[Valid][3][UTT][ MSE = 0.2255 | LCC = 0.9357 | SRCC = 0.9376 ] +INFO:root:Average loss=0.15641753923892976 +INFO:root: +[Valid][4][UTT][ MSE = 0.2494 | LCC = 0.9415 | SRCC = 0.9434 ] +INFO:root:Average loss=0.1490214609503746 +INFO:root: +[Valid][5][UTT][ MSE = 0.2400 | LCC = 0.9400 | SRCC = 0.9434 ] +INFO:root:Average loss=0.1290153543949127 +INFO:root: +[Valid][6][UTT][ MSE = 0.2181 | LCC = 0.9440 | SRCC = 0.9455 ] +INFO:root:Average loss=0.12453412747383118 +INFO:root: +[Valid][7][UTT][ MSE = 0.2381 | LCC = 0.9457 | SRCC = 0.9467 ] +INFO:root:Average loss=0.11779754942655564 +INFO:root: +[Valid][8][UTT][ MSE = 0.2705 | LCC = 0.9429 | SRCC = 0.9460 ] +INFO:root:Average loss=0.11571571865677834 +INFO:root: +[Valid][9][UTT][ MSE = 0.2179 | LCC = 0.9495 | SRCC = 0.9511 ] +INFO:root:Average loss=0.10437114870548249 +INFO:root: +[Valid][10][UTT][ MSE = 0.2178 | LCC = 0.9462 | SRCC = 0.9470 ] +INFO:root:Average loss=0.10062645182013512 +INFO:root: +[Valid][11][UTT][ MSE = 0.2331 | LCC = 0.9495 | SRCC = 0.9503 ] +INFO:root:Average loss=0.09521615481376648 +INFO:root: +[Valid][12][UTT][ MSE = 0.2207 | LCC = 0.9518 | SRCC = 0.9524 ] +INFO:root:Average loss=0.08994716686010361 +INFO:root: +[Valid][13][UTT][ MSE = 0.2112 | LCC = 0.9497 | SRCC = 0.9505 ] +INFO:root:Average loss=0.08778494489192963 +INFO:root: +[Valid][14][UTT][ MSE = 0.2025 | LCC = 0.9524 | SRCC = 0.9536 ] +INFO:root:Average loss=0.0847416012585163 +INFO:root: +[Valid][15][UTT][ MSE = 0.2283 | LCC = 0.9514 | SRCC = 0.9523 ] +INFO:root:Average loss=0.07859733238816262 +INFO:root: +[Valid][16][UTT][ MSE = 0.2007 | LCC = 0.9519 | SRCC = 0.9535 ] +INFO:root:Average loss=0.07627680110931397 +INFO:root: +[Valid][17][UTT][ MSE = 0.2298 | LCC = 0.9517 | SRCC = 0.9526 ] +INFO:root:Average loss=0.07353291621804238 +INFO:root: +[Valid][18][UTT][ MSE = 0.1904 | LCC = 0.9524 | SRCC = 0.9542 ] +INFO:root:Average loss=0.06945389366149902 +INFO:root: +[Valid][19][UTT][ MSE = 0.1938 | LCC = 0.9526 | SRCC = 0.9556 ] +INFO:root:Average loss=0.06735689690709114 +INFO:root: +[Valid][20][UTT][ MSE = 0.2309 | LCC = 0.9518 | SRCC = 0.9532 ] +INFO:root:Average loss=0.06902114582061768 +INFO:root: +[Valid][21][UTT][ MSE = 0.1990 | LCC = 0.9533 | SRCC = 0.9547 ] +INFO:root:Average loss=0.06487697052955628 +INFO:root: +[Valid][22][UTT][ MSE = 0.1823 | LCC = 0.9541 | SRCC = 0.9562 ] +INFO:root:Average loss=0.06129820293188095 +INFO:root: +[Valid][23][UTT][ MSE = 0.1726 | LCC = 0.9532 | SRCC = 0.9544 ] +INFO:root:Average loss=0.05733481751382351 +INFO:root: +[Valid][24][UTT][ MSE = 0.2039 | LCC = 0.9549 | SRCC = 0.9560 ] +INFO:root:Average loss=0.057057929426431654 +INFO:root: +[Valid][25][UTT][ MSE = 0.1810 | LCC = 0.9563 | SRCC = 0.9575 ] +INFO:root:Average loss=0.06035015614330769 +INFO:root: +[Valid][26][UTT][ MSE = 0.1936 | LCC = 0.9549 | SRCC = 0.9570 ] +INFO:root:Average loss=0.054653173923492435 +INFO:root: +[Valid][27][UTT][ MSE = 0.2086 | LCC = 0.9561 | SRCC = 0.9571 ] +INFO:root:Average loss=0.05370331659913063 +INFO:root: +[Valid][28][UTT][ MSE = 0.1954 | LCC = 0.9563 | SRCC = 0.9574 ] +INFO:root:Average loss=0.05308150035142899 +INFO:root: +[Valid][29][UTT][ MSE = 0.1822 | LCC = 0.9544 | SRCC = 0.9555 ] +INFO:root:Average loss=0.051317504212260244 +INFO:root: +[Valid][30][UTT][ MSE = 0.1903 | LCC = 0.9543 | SRCC = 0.9559 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1759 | LCC = 0.9590 | SRCC = 0.9599 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3feabfba28d789c9544b68f49e2610963e4f0993 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:55344c07c4a6af1440bc42613a745b27416dbc74826a95ccb6f066060a9e7aed +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4b93cab718f728249da3dafe4c1ab3e4dcf839ef --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3313543eb0c38214d81e2c5569c75108454528005826d1885e54f85de88ddd4e +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..513f1192ca9340860d5b408ffc681ea4f4f88306 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44070279228687287 +INFO:root: +[Valid][0][UTT][ MSE = 0.3375 | LCC = 0.9074 | SRCC = 0.9113 ] +INFO:root:Average loss=0.2166972661614418 +INFO:root: +[Valid][1][UTT][ MSE = 0.2592 | LCC = 0.9279 | SRCC = 0.9310 ] +INFO:root:Average loss=0.1724157184958458 +INFO:root: +[Valid][2][UTT][ MSE = 0.2369 | LCC = 0.9359 | SRCC = 0.9360 ] +INFO:root:Average loss=0.15242008763551712 +INFO:root: +[Valid][3][UTT][ MSE = 0.2228 | LCC = 0.9440 | SRCC = 0.9458 ] +INFO:root:Average loss=0.12979104924201965 +INFO:root: +[Valid][4][UTT][ MSE = 0.2202 | LCC = 0.9466 | SRCC = 0.9466 ] +INFO:root:Average loss=0.12244685506820678 +INFO:root: +[Valid][5][UTT][ MSE = 0.2216 | LCC = 0.9428 | SRCC = 0.9428 ] +INFO:root:Average loss=0.11143797865509987 +INFO:root: +[Valid][6][UTT][ MSE = 0.1937 | LCC = 0.9507 | SRCC = 0.9508 ] +INFO:root:Average loss=0.10403558102250099 +INFO:root: +[Valid][7][UTT][ MSE = 0.1939 | LCC = 0.9509 | SRCC = 0.9517 ] +INFO:root:Average loss=0.09661781495809554 +INFO:root: +[Valid][8][UTT][ MSE = 0.1983 | LCC = 0.9517 | SRCC = 0.9519 ] +INFO:root:Average loss=0.08961035332083703 +INFO:root: +[Valid][9][UTT][ MSE = 0.2068 | LCC = 0.9520 | SRCC = 0.9517 ] +INFO:root:Average loss=0.08531681603193284 +INFO:root: +[Valid][10][UTT][ MSE = 0.1992 | LCC = 0.9519 | SRCC = 0.9520 ] +INFO:root:Average loss=0.08222639694809913 +INFO:root: +[Valid][11][UTT][ MSE = 0.1711 | LCC = 0.9526 | SRCC = 0.9539 ] +INFO:root:Average loss=0.07725544610619545 +INFO:root: +[Valid][12][UTT][ MSE = 0.1839 | LCC = 0.9542 | SRCC = 0.9540 ] +INFO:root:Average loss=0.07504747766256332 +INFO:root: +[Valid][13][UTT][ MSE = 0.1891 | LCC = 0.9538 | SRCC = 0.9540 ] +INFO:root:Average loss=0.07020125806331634 +INFO:root: +[Valid][14][UTT][ MSE = 0.1835 | LCC = 0.9542 | SRCC = 0.9536 ] +INFO:root:Average loss=0.06519092240929604 +INFO:root: +[Valid][15][UTT][ MSE = 0.2262 | LCC = 0.9530 | SRCC = 0.9537 ] +INFO:root:Average loss=0.06927060148119926 +INFO:root: +[Valid][16][UTT][ MSE = 0.1799 | LCC = 0.9542 | SRCC = 0.9547 ] +INFO:root:Average loss=0.06250716534256935 +INFO:root: +[Valid][17][UTT][ MSE = 0.1752 | LCC = 0.9534 | SRCC = 0.9542 ] +INFO:root:Average loss=0.05797010588645935 +INFO:root: +[Valid][18][UTT][ MSE = 0.1911 | LCC = 0.9540 | SRCC = 0.9533 ] +INFO:root:Average loss=0.05950957250595093 +INFO:root: +[Valid][19][UTT][ MSE = 0.1760 | LCC = 0.9542 | SRCC = 0.9542 ] +INFO:root:Average loss=0.05849827912449837 +INFO:root: +[Valid][20][UTT][ MSE = 0.2014 | LCC = 0.9551 | SRCC = 0.9550 ] +INFO:root:Average loss=0.053640374541282655 +INFO:root: +[Valid][21][UTT][ MSE = 0.1828 | LCC = 0.9556 | SRCC = 0.9554 ] +INFO:root:Average loss=0.052775131195783614 +INFO:root: +[Valid][22][UTT][ MSE = 0.1799 | LCC = 0.9562 | SRCC = 0.9559 ] +INFO:root:Average loss=0.04807243171334267 +INFO:root: +[Valid][23][UTT][ MSE = 0.1865 | LCC = 0.9557 | SRCC = 0.9557 ] +INFO:root:Average loss=0.04591294838488102 +INFO:root: +[Valid][24][UTT][ MSE = 0.1858 | LCC = 0.9537 | SRCC = 0.9530 ] +INFO:root:Average loss=0.046369081780314444 +INFO:root: +[Valid][25][UTT][ MSE = 0.1763 | LCC = 0.9554 | SRCC = 0.9547 ] +INFO:root:Average loss=0.04436561270058155 +INFO:root: +[Valid][26][UTT][ MSE = 0.1902 | LCC = 0.9527 | SRCC = 0.9527 ] +INFO:root:Average loss=0.04346632993221283 +INFO:root: +[Valid][27][UTT][ MSE = 0.1975 | LCC = 0.9560 | SRCC = 0.9558 ] +INFO:root:Average loss=0.04101228515803814 +INFO:root: +[Valid][28][UTT][ MSE = 0.1790 | LCC = 0.9547 | SRCC = 0.9543 ] +INFO:root:Average loss=0.03868407414853573 +INFO:root: +[Valid][29][UTT][ MSE = 0.1625 | LCC = 0.9555 | SRCC = 0.9564 ] +INFO:root:Average loss=0.03799951495230198 +INFO:root: +[Valid][30][UTT][ MSE = 0.1948 | LCC = 0.9563 | SRCC = 0.9564 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1926 | LCC = 0.9609 | SRCC = 0.9620 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0e31d35f9d6c0bfef44502ea31055f83b73ddd53 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e3a645f67ac28a718f9336a5b0cb4c9e6be52cf6243b77a4e60f791ff6ed818 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e93eeb0c27a3e77369a306905c0326b14c264880 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cb73fe91e8da04ce40de3ad744ae1a38f02800f0fcc843b7095a6a3e81ae4e76 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..859a95dfbeb198b7dc562f8300e0ddf94c7a466b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.43629271066188813 +INFO:root: +[Valid][0][UTT][ MSE = 0.2792 | LCC = 0.9076 | SRCC = 0.9136 ] +INFO:root:Average loss=0.20519364947080612 +INFO:root: +[Valid][1][UTT][ MSE = 0.2577 | LCC = 0.9291 | SRCC = 0.9300 ] +INFO:root:Average loss=0.1681509991288185 +INFO:root: +[Valid][2][UTT][ MSE = 0.2187 | LCC = 0.9406 | SRCC = 0.9425 ] +INFO:root:Average loss=0.1447545763850212 +INFO:root: +[Valid][3][UTT][ MSE = 0.2150 | LCC = 0.9457 | SRCC = 0.9468 ] +INFO:root:Average loss=0.12945595556497574 +INFO:root: +[Valid][4][UTT][ MSE = 0.2116 | LCC = 0.9486 | SRCC = 0.9490 ] +INFO:root:Average loss=0.11969594383239746 +INFO:root: +[Valid][5][UTT][ MSE = 0.2074 | LCC = 0.9516 | SRCC = 0.9520 ] +INFO:root:Average loss=0.1139531654715538 +INFO:root: +[Valid][6][UTT][ MSE = 0.1850 | LCC = 0.9511 | SRCC = 0.9519 ] +INFO:root:Average loss=0.10026192033290864 +INFO:root: +[Valid][7][UTT][ MSE = 0.1912 | LCC = 0.9527 | SRCC = 0.9532 ] +INFO:root:Average loss=0.09658238673210144 +INFO:root: +[Valid][8][UTT][ MSE = 0.2019 | LCC = 0.9534 | SRCC = 0.9538 ] +INFO:root:Average loss=0.08903549337387084 +INFO:root: +[Valid][9][UTT][ MSE = 0.1946 | LCC = 0.9542 | SRCC = 0.9545 ] +INFO:root:Average loss=0.08454969584941864 +INFO:root: +[Valid][10][UTT][ MSE = 0.1913 | LCC = 0.9510 | SRCC = 0.9517 ] +INFO:root:Average loss=0.08508250379562378 +INFO:root: +[Valid][11][UTT][ MSE = 0.1921 | LCC = 0.9543 | SRCC = 0.9547 ] +INFO:root:Average loss=0.07859320867061614 +INFO:root: +[Valid][12][UTT][ MSE = 0.1903 | LCC = 0.9541 | SRCC = 0.9552 ] +INFO:root:Average loss=0.07423370921611785 +INFO:root: +[Valid][13][UTT][ MSE = 0.1845 | LCC = 0.9547 | SRCC = 0.9550 ] +INFO:root:Average loss=0.06916590133309364 +INFO:root: +[Valid][14][UTT][ MSE = 0.1705 | LCC = 0.9556 | SRCC = 0.9567 ] +INFO:root:Average loss=0.06520826435089111 +INFO:root: +[Valid][15][UTT][ MSE = 0.1939 | LCC = 0.9550 | SRCC = 0.9557 ] +INFO:root:Average loss=0.06410733503103257 +INFO:root: +[Valid][16][UTT][ MSE = 0.1709 | LCC = 0.9561 | SRCC = 0.9571 ] +INFO:root:Average loss=0.06326877528429031 +INFO:root: +[Valid][17][UTT][ MSE = 0.1690 | LCC = 0.9571 | SRCC = 0.9579 ] +INFO:root:Average loss=0.05805692587792873 +INFO:root: +[Valid][18][UTT][ MSE = 0.1638 | LCC = 0.9574 | SRCC = 0.9584 ] +INFO:root:Average loss=0.05434042382240296 +INFO:root: +[Valid][19][UTT][ MSE = 0.1758 | LCC = 0.9568 | SRCC = 0.9572 ] +INFO:root:Average loss=0.05455326281487942 +INFO:root: +[Valid][20][UTT][ MSE = 0.1902 | LCC = 0.9580 | SRCC = 0.9582 ] +INFO:root:Average loss=0.05308151154220104 +INFO:root: +[Valid][21][UTT][ MSE = 0.1805 | LCC = 0.9584 | SRCC = 0.9585 ] +INFO:root:Average loss=0.04945381647348404 +INFO:root: +[Valid][22][UTT][ MSE = 0.1774 | LCC = 0.9572 | SRCC = 0.9569 ] +INFO:root:Average loss=0.04651333028078079 +INFO:root: +[Valid][23][UTT][ MSE = 0.1694 | LCC = 0.9584 | SRCC = 0.9582 ] +INFO:root:Average loss=0.04679550333321095 +INFO:root: +[Valid][24][UTT][ MSE = 0.1614 | LCC = 0.9579 | SRCC = 0.9580 ] +INFO:root:Average loss=0.04488469363749027 +INFO:root: +[Valid][25][UTT][ MSE = 0.1879 | LCC = 0.9592 | SRCC = 0.9592 ] +INFO:root:Average loss=0.042277544841170314 +INFO:root: +[Valid][26][UTT][ MSE = 0.1616 | LCC = 0.9583 | SRCC = 0.9583 ] +INFO:root:Average loss=0.04158027118444443 +INFO:root: +[Valid][27][UTT][ MSE = 0.1716 | LCC = 0.9566 | SRCC = 0.9564 ] +INFO:root:Average loss=0.04228869187831879 +INFO:root: +[Valid][28][UTT][ MSE = 0.1736 | LCC = 0.9583 | SRCC = 0.9582 ] +INFO:root:Average loss=0.04051175490021706 +INFO:root: +[Valid][29][UTT][ MSE = 0.1683 | LCC = 0.9589 | SRCC = 0.9587 ] +INFO:root:Average loss=0.03833329246938229 +INFO:root: +[Valid][30][UTT][ MSE = 0.1620 | LCC = 0.9561 | SRCC = 0.9563 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1923 | LCC = 0.9599 | SRCC = 0.9610 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..492fec37ba353dbc125e93c39f0f98cb211a3716 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ba05b89c5df2f42e51eb1c7a20597c909cc20e20291d34c8841980afde8f0a1b +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..8f059bbd97855a07e2acbbd9d70c02591e153f5f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0021bf2c76dafee260a8ae470184d6c2e4a0f58a46e4870765da80a1308b5d6 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d7ac80020df4a211a6ad5666fbcda3a34c9af369 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5381969361305237 +INFO:root: +[Valid][0][UTT][ MSE = 0.3115 | LCC = 0.9119 | SRCC = 0.9148 ] +INFO:root:Average loss=0.2061135401725769 +INFO:root: +[Valid][1][UTT][ MSE = 0.2455 | LCC = 0.9289 | SRCC = 0.9311 ] +INFO:root:Average loss=0.16439072757959367 +INFO:root: +[Valid][2][UTT][ MSE = 0.2366 | LCC = 0.9351 | SRCC = 0.9368 ] +INFO:root:Average loss=0.14950133460760118 +INFO:root: +[Valid][3][UTT][ MSE = 0.2360 | LCC = 0.9421 | SRCC = 0.9439 ] +INFO:root:Average loss=0.13071510922908783 +INFO:root: +[Valid][4][UTT][ MSE = 0.2225 | LCC = 0.9475 | SRCC = 0.9489 ] +INFO:root:Average loss=0.12062310481071473 +INFO:root: +[Valid][5][UTT][ MSE = 0.1999 | LCC = 0.9501 | SRCC = 0.9514 ] +INFO:root:Average loss=0.1107893925011158 +INFO:root: +[Valid][6][UTT][ MSE = 0.1989 | LCC = 0.9488 | SRCC = 0.9503 ] +INFO:root:Average loss=0.10314005494117737 +INFO:root: +[Valid][7][UTT][ MSE = 0.2111 | LCC = 0.9515 | SRCC = 0.9525 ] +INFO:root:Average loss=0.09925200715661049 +INFO:root: +[Valid][8][UTT][ MSE = 0.2177 | LCC = 0.9516 | SRCC = 0.9526 ] +INFO:root:Average loss=0.09103800374269486 +INFO:root: +[Valid][9][UTT][ MSE = 0.2085 | LCC = 0.9543 | SRCC = 0.9550 ] +INFO:root:Average loss=0.08593437021970748 +INFO:root: +[Valid][10][UTT][ MSE = 0.2036 | LCC = 0.9534 | SRCC = 0.9540 ] +INFO:root:Average loss=0.0794707798063755 +INFO:root: +[Valid][11][UTT][ MSE = 0.2168 | LCC = 0.9535 | SRCC = 0.9545 ] +INFO:root:Average loss=0.07757580733299255 +INFO:root: +[Valid][12][UTT][ MSE = 0.2120 | LCC = 0.9522 | SRCC = 0.9525 ] +INFO:root:Average loss=0.07022411891818046 +INFO:root: +[Valid][13][UTT][ MSE = 0.2163 | LCC = 0.9542 | SRCC = 0.9542 ] +INFO:root:Average loss=0.06974675339460373 +INFO:root: +[Valid][14][UTT][ MSE = 0.1758 | LCC = 0.9553 | SRCC = 0.9556 ] +INFO:root:Average loss=0.06673781561851501 +INFO:root: +[Valid][15][UTT][ MSE = 0.1879 | LCC = 0.9557 | SRCC = 0.9561 ] +INFO:root:Average loss=0.0640679602921009 +INFO:root: +[Valid][16][UTT][ MSE = 0.1907 | LCC = 0.9559 | SRCC = 0.9567 ] +INFO:root:Average loss=0.06226389974355698 +INFO:root: +[Valid][17][UTT][ MSE = 0.1773 | LCC = 0.9570 | SRCC = 0.9571 ] +INFO:root:Average loss=0.05668555247783661 +INFO:root: +[Valid][18][UTT][ MSE = 0.1824 | LCC = 0.9534 | SRCC = 0.9534 ] +INFO:root:Average loss=0.05576067623496055 +INFO:root: +[Valid][19][UTT][ MSE = 0.1847 | LCC = 0.9565 | SRCC = 0.9565 ] +INFO:root:Average loss=0.05302490405738354 +INFO:root: +[Valid][20][UTT][ MSE = 0.2088 | LCC = 0.9535 | SRCC = 0.9531 ] +INFO:root:Average loss=0.05258875820040703 +INFO:root: +[Valid][21][UTT][ MSE = 0.1875 | LCC = 0.9566 | SRCC = 0.9568 ] +INFO:root:Average loss=0.05262869882583618 +INFO:root: +[Valid][22][UTT][ MSE = 0.1671 | LCC = 0.9569 | SRCC = 0.9568 ] +INFO:root:Average loss=0.046940534308552745 +INFO:root: +[Valid][23][UTT][ MSE = 0.1889 | LCC = 0.9568 | SRCC = 0.9572 ] +INFO:root:Average loss=0.046296240255236626 +INFO:root: +[Valid][24][UTT][ MSE = 0.1838 | LCC = 0.9575 | SRCC = 0.9578 ] +INFO:root:Average loss=0.043472914069890976 +INFO:root: +[Valid][25][UTT][ MSE = 0.1714 | LCC = 0.9564 | SRCC = 0.9563 ] +INFO:root:Average loss=0.04337498107552528 +INFO:root: +[Valid][26][UTT][ MSE = 0.1736 | LCC = 0.9572 | SRCC = 0.9574 ] +INFO:root:Average loss=0.04075286519527435 +INFO:root: +[Valid][27][UTT][ MSE = 0.1837 | LCC = 0.9579 | SRCC = 0.9581 ] +INFO:root:Average loss=0.041175436854362486 +INFO:root: +[Valid][28][UTT][ MSE = 0.1878 | LCC = 0.9571 | SRCC = 0.9578 ] +INFO:root:Average loss=0.0378303580135107 +INFO:root: +[Valid][29][UTT][ MSE = 0.1886 | LCC = 0.9557 | SRCC = 0.9557 ] +INFO:root:Average loss=0.039149532318115236 +INFO:root: +[Valid][30][UTT][ MSE = 0.1742 | LCC = 0.9572 | SRCC = 0.9574 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1872 | LCC = 0.9594 | SRCC = 0.9600 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2bf833a9d64e863bd1085de807f9b65ad552e43c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7f31ec6350e68110cd9dde25ad271c7bfcab942ec118ffd345760ed6cd830341 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..fd51e829dfde735ed7385a31d559c69bb99a6e2e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6fafded68daa0af26a9977c6cbdb503f0b58195fcf736be5afcd2a40a04e6dbd +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6292645094206b5aec06bd2d9a73a8738c9b4b19 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4602924358844757 +INFO:root: +[Valid][0][UTT][ MSE = 0.3095 | LCC = 0.9125 | SRCC = 0.9161 ] +INFO:root:Average loss=0.19946460258960724 +INFO:root: +[Valid][1][UTT][ MSE = 0.2374 | LCC = 0.9328 | SRCC = 0.9344 ] +INFO:root:Average loss=0.16337242370843888 +INFO:root: +[Valid][2][UTT][ MSE = 0.2625 | LCC = 0.9384 | SRCC = 0.9385 ] +INFO:root:Average loss=0.1569132800102234 +INFO:root: +[Valid][3][UTT][ MSE = 0.2338 | LCC = 0.9438 | SRCC = 0.9452 ] +INFO:root:Average loss=0.13289984112977982 +INFO:root: +[Valid][4][UTT][ MSE = 0.2363 | LCC = 0.9455 | SRCC = 0.9469 ] +INFO:root:Average loss=0.12062103408575058 +INFO:root: +[Valid][5][UTT][ MSE = 0.2028 | LCC = 0.9477 | SRCC = 0.9499 ] +INFO:root:Average loss=0.11122004386782647 +INFO:root: +[Valid][6][UTT][ MSE = 0.1993 | LCC = 0.9526 | SRCC = 0.9531 ] +INFO:root:Average loss=0.10767459654808044 +INFO:root: +[Valid][7][UTT][ MSE = 0.2118 | LCC = 0.9445 | SRCC = 0.9485 ] +INFO:root:Average loss=0.0981822763979435 +INFO:root: +[Valid][8][UTT][ MSE = 0.2326 | LCC = 0.9538 | SRCC = 0.9546 ] +INFO:root:Average loss=0.09316218554973603 +INFO:root: +[Valid][9][UTT][ MSE = 0.1859 | LCC = 0.9546 | SRCC = 0.9552 ] +INFO:root:Average loss=0.08526782813668252 +INFO:root: +[Valid][10][UTT][ MSE = 0.2107 | LCC = 0.9547 | SRCC = 0.9551 ] +INFO:root:Average loss=0.0799938046336174 +INFO:root: +[Valid][11][UTT][ MSE = 0.1911 | LCC = 0.9555 | SRCC = 0.9555 ] +INFO:root:Average loss=0.0765812036395073 +INFO:root: +[Valid][12][UTT][ MSE = 0.1871 | LCC = 0.9563 | SRCC = 0.9567 ] +INFO:root:Average loss=0.07426928383111954 +INFO:root: +[Valid][13][UTT][ MSE = 0.1954 | LCC = 0.9572 | SRCC = 0.9571 ] +INFO:root:Average loss=0.07017257484793663 +INFO:root: +[Valid][14][UTT][ MSE = 0.1916 | LCC = 0.9533 | SRCC = 0.9530 ] +INFO:root:Average loss=0.06691239297389984 +INFO:root: +[Valid][15][UTT][ MSE = 0.1674 | LCC = 0.9578 | SRCC = 0.9581 ] +INFO:root:Average loss=0.06540151506662369 +INFO:root: +[Valid][16][UTT][ MSE = 0.1947 | LCC = 0.9566 | SRCC = 0.9567 ] +INFO:root:Average loss=0.061096594154834746 +INFO:root: +[Valid][17][UTT][ MSE = 0.1950 | LCC = 0.9575 | SRCC = 0.9573 ] +INFO:root:Average loss=0.055762057036161425 +INFO:root: +[Valid][18][UTT][ MSE = 0.1980 | LCC = 0.9565 | SRCC = 0.9567 ] +INFO:root:Average loss=0.056930550664663314 +INFO:root: +[Valid][19][UTT][ MSE = 0.1781 | LCC = 0.9569 | SRCC = 0.9570 ] +INFO:root:Average loss=0.05392321696877479 +INFO:root: +[Valid][20][UTT][ MSE = 0.1833 | LCC = 0.9576 | SRCC = 0.9578 ] +INFO:root:Average loss=0.05740423461794853 +INFO:root: +[Valid][21][UTT][ MSE = 0.1957 | LCC = 0.9544 | SRCC = 0.9559 ] +INFO:root:Average loss=0.05070932833850384 +INFO:root: +[Valid][22][UTT][ MSE = 0.1846 | LCC = 0.9573 | SRCC = 0.9572 ] +INFO:root:Average loss=0.048580299153923985 +INFO:root: +[Valid][23][UTT][ MSE = 0.1967 | LCC = 0.9568 | SRCC = 0.9564 ] +INFO:root:Average loss=0.045896751523017885 +INFO:root: +[Valid][24][UTT][ MSE = 0.2093 | LCC = 0.9545 | SRCC = 0.9540 ] +INFO:root:Average loss=0.046270659178495405 +INFO:root: +[Valid][25][UTT][ MSE = 0.1721 | LCC = 0.9576 | SRCC = 0.9569 ] +INFO:root:Average loss=0.04335870341956615 +INFO:root: +[Valid][26][UTT][ MSE = 0.1802 | LCC = 0.9571 | SRCC = 0.9569 ] +INFO:root:Average loss=0.041812303826212885 +INFO:root: +[Valid][27][UTT][ MSE = 0.2024 | LCC = 0.9568 | SRCC = 0.9565 ] +INFO:root:Average loss=0.04086300508677959 +INFO:root: +[Valid][28][UTT][ MSE = 0.1787 | LCC = 0.9562 | SRCC = 0.9568 ] +INFO:root:Average loss=0.03926291438937187 +INFO:root: +[Valid][29][UTT][ MSE = 0.2030 | LCC = 0.9572 | SRCC = 0.9571 ] +INFO:root:Average loss=0.038076966121792795 +INFO:root: +[Valid][30][UTT][ MSE = 0.1768 | LCC = 0.9570 | SRCC = 0.9568 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1677 | LCC = 0.9595 | SRCC = 0.9601 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c9b6e9225f96270eb7ea1494cc6924fa510ef877 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a993968f32eb860f1a436ea64799ff3fc4b8c6b002bd25b14956c0deca30e3b0 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..3277bd674bc551be8e0d2fdcb0fec90a68f819e1 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aeae6d0a71a14223acd626bb766986de2548ee92c0a8aecf8b30ec0b5a2c8e5d +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..37f4c9ae09b27aa9bc39299540e8bf54ae109fea --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer1/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4644504265785217 +INFO:root: +[Valid][0][UTT][ MSE = 0.3671 | LCC = 0.9008 | SRCC = 0.9107 ] +INFO:root:Average loss=0.22051585710048677 +INFO:root: +[Valid][1][UTT][ MSE = 0.3086 | LCC = 0.9277 | SRCC = 0.9297 ] +INFO:root:Average loss=0.16833397465944291 +INFO:root: +[Valid][2][UTT][ MSE = 0.2615 | LCC = 0.9401 | SRCC = 0.9420 ] +INFO:root:Average loss=0.14518246391415596 +INFO:root: +[Valid][3][UTT][ MSE = 0.2298 | LCC = 0.9468 | SRCC = 0.9482 ] +INFO:root:Average loss=0.13454131627082824 +INFO:root: +[Valid][4][UTT][ MSE = 0.1901 | LCC = 0.9470 | SRCC = 0.9491 ] +INFO:root:Average loss=0.12192892521619797 +INFO:root: +[Valid][5][UTT][ MSE = 0.2149 | LCC = 0.9511 | SRCC = 0.9515 ] +INFO:root:Average loss=0.11015139889717102 +INFO:root: +[Valid][6][UTT][ MSE = 0.1726 | LCC = 0.9495 | SRCC = 0.9521 ] +INFO:root:Average loss=0.10599474227428436 +INFO:root: +[Valid][7][UTT][ MSE = 0.2018 | LCC = 0.9526 | SRCC = 0.9529 ] +INFO:root:Average loss=0.09681357657909394 +INFO:root: +[Valid][8][UTT][ MSE = 0.2222 | LCC = 0.9516 | SRCC = 0.9518 ] +INFO:root:Average loss=0.09145489487051964 +INFO:root: +[Valid][9][UTT][ MSE = 0.2009 | LCC = 0.9499 | SRCC = 0.9500 ] +INFO:root:Average loss=0.08682288402318955 +INFO:root: +[Valid][10][UTT][ MSE = 0.1867 | LCC = 0.9524 | SRCC = 0.9535 ] +INFO:root:Average loss=0.08371237781643867 +INFO:root: +[Valid][11][UTT][ MSE = 0.1954 | LCC = 0.9547 | SRCC = 0.9549 ] +INFO:root:Average loss=0.0775154250562191 +INFO:root: +[Valid][12][UTT][ MSE = 0.1765 | LCC = 0.9562 | SRCC = 0.9559 ] +INFO:root:Average loss=0.0734540363252163 +INFO:root: +[Valid][13][UTT][ MSE = 0.2097 | LCC = 0.9563 | SRCC = 0.9563 ] +INFO:root:Average loss=0.07052721494436263 +INFO:root: +[Valid][14][UTT][ MSE = 0.1737 | LCC = 0.9537 | SRCC = 0.9541 ] +INFO:root:Average loss=0.07101986396312714 +INFO:root: +[Valid][15][UTT][ MSE = 0.1920 | LCC = 0.9547 | SRCC = 0.9547 ] +INFO:root:Average loss=0.06258304744958877 +INFO:root: +[Valid][16][UTT][ MSE = 0.1898 | LCC = 0.9560 | SRCC = 0.9560 ] +INFO:root:Average loss=0.06289290469884873 +INFO:root: +[Valid][17][UTT][ MSE = 0.1791 | LCC = 0.9576 | SRCC = 0.9577 ] +INFO:root:Average loss=0.058662702172994616 +INFO:root: +[Valid][18][UTT][ MSE = 0.1937 | LCC = 0.9571 | SRCC = 0.9573 ] +INFO:root:Average loss=0.05687642423808575 +INFO:root: +[Valid][19][UTT][ MSE = 0.1847 | LCC = 0.9571 | SRCC = 0.9572 ] +INFO:root:Average loss=0.05618154036998749 +INFO:root: +[Valid][20][UTT][ MSE = 0.1794 | LCC = 0.9553 | SRCC = 0.9557 ] +INFO:root:Average loss=0.05858135569095611 +INFO:root: +[Valid][21][UTT][ MSE = 0.1754 | LCC = 0.9560 | SRCC = 0.9572 ] +INFO:root:Average loss=0.04966901242733002 +INFO:root: +[Valid][22][UTT][ MSE = 0.1816 | LCC = 0.9566 | SRCC = 0.9574 ] +INFO:root:Average loss=0.04738688090443611 +INFO:root: +[Valid][23][UTT][ MSE = 0.2093 | LCC = 0.9571 | SRCC = 0.9580 ] +INFO:root:Average loss=0.048839219659566876 +INFO:root: +[Valid][24][UTT][ MSE = 0.1727 | LCC = 0.9591 | SRCC = 0.9587 ] +INFO:root:Average loss=0.04405650401115418 +INFO:root: +[Valid][25][UTT][ MSE = 0.1772 | LCC = 0.9587 | SRCC = 0.9589 ] +INFO:root:Average loss=0.042384812042117116 +INFO:root: +[Valid][26][UTT][ MSE = 0.1809 | LCC = 0.9569 | SRCC = 0.9580 ] +INFO:root:Average loss=0.04257926522195339 +INFO:root: +[Valid][27][UTT][ MSE = 0.1988 | LCC = 0.9594 | SRCC = 0.9598 ] +INFO:root:Average loss=0.04079847992956638 +INFO:root: +[Valid][28][UTT][ MSE = 0.1885 | LCC = 0.9583 | SRCC = 0.9584 ] +INFO:root:Average loss=0.040474271342158315 +INFO:root: +[Valid][29][UTT][ MSE = 0.1875 | LCC = 0.9547 | SRCC = 0.9547 ] +INFO:root:Average loss=0.04333827371895313 +INFO:root: +[Valid][30][UTT][ MSE = 0.1949 | LCC = 0.9551 | SRCC = 0.9555 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2040 | LCC = 0.9604 | SRCC = 0.9615 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b17203b9673f5a9f32ce272fc682e254d099bf3a --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3cf58cdf826d3b65f2fcede42cfa7d0d314f0ff42086add65afebc531a3752e0 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..5ef093662a9ef86d13182cfeeb26ee8c6f6514c4 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ccc33db8b305e8cc0c49b5d8a60e021cea4c5c0c8a0fa70a7d3602eaa68190d6 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1e8e5eb8ff883e63f791e4a399836f88925a08f8 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.40299374067783356 +INFO:root: +[Valid][0][UTT][ MSE = 0.2970 | LCC = 0.9124 | SRCC = 0.9172 ] +INFO:root:Average loss=0.21385713148117066 +INFO:root: +[Valid][1][UTT][ MSE = 0.3347 | LCC = 0.9290 | SRCC = 0.9311 ] +INFO:root:Average loss=0.18865110021829606 +INFO:root: +[Valid][2][UTT][ MSE = 0.2730 | LCC = 0.9354 | SRCC = 0.9395 ] +INFO:root:Average loss=0.17018378460407257 +INFO:root: +[Valid][3][UTT][ MSE = 0.2955 | LCC = 0.9328 | SRCC = 0.9354 ] +INFO:root:Average loss=0.1624770833849907 +INFO:root: +[Valid][4][UTT][ MSE = 0.3337 | LCC = 0.9371 | SRCC = 0.9396 ] +INFO:root:Average loss=0.15776405906677246 +INFO:root: +[Valid][5][UTT][ MSE = 0.2683 | LCC = 0.9414 | SRCC = 0.9451 ] +INFO:root:Average loss=0.14686900508403777 +INFO:root: +[Valid][6][UTT][ MSE = 0.3120 | LCC = 0.9342 | SRCC = 0.9365 ] +INFO:root:Average loss=0.144317160487175 +INFO:root: +[Valid][7][UTT][ MSE = 0.2915 | LCC = 0.9433 | SRCC = 0.9444 ] +INFO:root:Average loss=0.13887519335746765 +INFO:root: +[Valid][8][UTT][ MSE = 0.2554 | LCC = 0.9394 | SRCC = 0.9423 ] +INFO:root:Average loss=0.1323683388531208 +INFO:root: +[Valid][9][UTT][ MSE = 0.2330 | LCC = 0.9445 | SRCC = 0.9481 ] +INFO:root:Average loss=0.1307642513513565 +INFO:root: +[Valid][10][UTT][ MSE = 0.2338 | LCC = 0.9466 | SRCC = 0.9483 ] +INFO:root:Average loss=0.13197209537029267 +INFO:root: +[Valid][11][UTT][ MSE = 0.2629 | LCC = 0.9442 | SRCC = 0.9473 ] +INFO:root:Average loss=0.12656863313913344 +INFO:root: +[Valid][12][UTT][ MSE = 0.2960 | LCC = 0.9435 | SRCC = 0.9452 ] +INFO:root:Average loss=0.11576347184181214 +INFO:root: +[Valid][13][UTT][ MSE = 0.2774 | LCC = 0.9471 | SRCC = 0.9498 ] +INFO:root:Average loss=0.11422830760478973 +INFO:root: +[Valid][14][UTT][ MSE = 0.2520 | LCC = 0.9490 | SRCC = 0.9502 ] +INFO:root:Average loss=0.10911568337678909 +INFO:root: +[Valid][15][UTT][ MSE = 0.2403 | LCC = 0.9496 | SRCC = 0.9508 ] +INFO:root:Average loss=0.10759615501761437 +INFO:root: +[Valid][16][UTT][ MSE = 0.2433 | LCC = 0.9481 | SRCC = 0.9505 ] +INFO:root:Average loss=0.10817556017637253 +INFO:root: +[Valid][17][UTT][ MSE = 0.3081 | LCC = 0.9399 | SRCC = 0.9405 ] +INFO:root:Average loss=0.10384493571519851 +INFO:root: +[Valid][18][UTT][ MSE = 0.2368 | LCC = 0.9450 | SRCC = 0.9473 ] +INFO:root:Average loss=0.09786012309789657 +INFO:root: +[Valid][19][UTT][ MSE = 0.2375 | LCC = 0.9470 | SRCC = 0.9493 ] +INFO:root:Average loss=0.10065971094369888 +INFO:root: +[Valid][20][UTT][ MSE = 0.2686 | LCC = 0.9492 | SRCC = 0.9503 ] +INFO:root:Average loss=0.09928636592626572 +INFO:root: +[Valid][21][UTT][ MSE = 0.2461 | LCC = 0.9504 | SRCC = 0.9515 ] +INFO:root:Average loss=0.09462209117412568 +INFO:root: +[Valid][22][UTT][ MSE = 0.2329 | LCC = 0.9494 | SRCC = 0.9512 ] +INFO:root:Average loss=0.0937059360742569 +INFO:root: +[Valid][23][UTT][ MSE = 0.2451 | LCC = 0.9473 | SRCC = 0.9477 ] +INFO:root:Average loss=0.0887704781293869 +INFO:root: +[Valid][24][UTT][ MSE = 0.2276 | LCC = 0.9488 | SRCC = 0.9489 ] +INFO:root:Average loss=0.08915168902277947 +INFO:root: +[Valid][25][UTT][ MSE = 0.2391 | LCC = 0.9498 | SRCC = 0.9516 ] +INFO:root:Average loss=0.08744154232740402 +INFO:root: +[Valid][26][UTT][ MSE = 0.2288 | LCC = 0.9499 | SRCC = 0.9505 ] +INFO:root:Average loss=0.08352775833010674 +INFO:root: +[Valid][27][UTT][ MSE = 0.2191 | LCC = 0.9494 | SRCC = 0.9516 ] +INFO:root:Average loss=0.08484130749106407 +INFO:root: +[Valid][28][UTT][ MSE = 0.2419 | LCC = 0.9395 | SRCC = 0.9437 ] +INFO:root:Average loss=0.08382597616314888 +INFO:root: +[Valid][29][UTT][ MSE = 0.2360 | LCC = 0.9459 | SRCC = 0.9472 ] +INFO:root:Average loss=0.0786683678328991 +INFO:root: +[Valid][30][UTT][ MSE = 0.2399 | LCC = 0.9495 | SRCC = 0.9495 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2501 | LCC = 0.9531 | SRCC = 0.9547 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1cfb50190c9ad2bc8668af02ed46da76ef36fef5 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5cd5f3d5d61ccef1b51926a460571f7588cf2d39fe8e6a65bc6e6fea301fbeb0 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2df6720c3a18005b0ac80d88d629bb7a0678781b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:03c510f630c67d659cf4ade9388fc5f014957fed17d03311a34ea8aefd867a37 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..51fa94f2811b39ac9d76de929c40ba6a1c4205ab --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.37994052469730377 +INFO:root: +[Valid][0][UTT][ MSE = 0.3164 | LCC = 0.9179 | SRCC = 0.9223 ] +INFO:root:Average loss=0.20864988338947296 +INFO:root: +[Valid][1][UTT][ MSE = 0.2843 | LCC = 0.9285 | SRCC = 0.9328 ] +INFO:root:Average loss=0.18352319806814194 +INFO:root: +[Valid][2][UTT][ MSE = 0.3015 | LCC = 0.9363 | SRCC = 0.9387 ] +INFO:root:Average loss=0.1761537043452263 +INFO:root: +[Valid][3][UTT][ MSE = 0.3521 | LCC = 0.9372 | SRCC = 0.9393 ] +INFO:root:Average loss=0.1694780637025833 +INFO:root: +[Valid][4][UTT][ MSE = 0.2702 | LCC = 0.9397 | SRCC = 0.9433 ] +INFO:root:Average loss=0.1516256942152977 +INFO:root: +[Valid][5][UTT][ MSE = 0.3022 | LCC = 0.9413 | SRCC = 0.9436 ] +INFO:root:Average loss=0.14944559401273727 +INFO:root: +[Valid][6][UTT][ MSE = 0.2950 | LCC = 0.9431 | SRCC = 0.9457 ] +INFO:root:Average loss=0.14473241555690766 +INFO:root: +[Valid][7][UTT][ MSE = 0.2890 | LCC = 0.9422 | SRCC = 0.9439 ] +INFO:root:Average loss=0.13582397156953813 +INFO:root: +[Valid][8][UTT][ MSE = 0.2511 | LCC = 0.9446 | SRCC = 0.9471 ] +INFO:root:Average loss=0.13170379191637038 +INFO:root: +[Valid][9][UTT][ MSE = 0.2604 | LCC = 0.9470 | SRCC = 0.9492 ] +INFO:root:Average loss=0.13004101687669753 +INFO:root: +[Valid][10][UTT][ MSE = 0.2428 | LCC = 0.9448 | SRCC = 0.9484 ] +INFO:root:Average loss=0.12259475958347321 +INFO:root: +[Valid][11][UTT][ MSE = 0.2722 | LCC = 0.9478 | SRCC = 0.9496 ] +INFO:root:Average loss=0.12715710759162901 +INFO:root: +[Valid][12][UTT][ MSE = 0.2501 | LCC = 0.9444 | SRCC = 0.9451 ] +INFO:root:Average loss=0.11807680815458298 +INFO:root: +[Valid][13][UTT][ MSE = 0.2743 | LCC = 0.9462 | SRCC = 0.9480 ] +INFO:root:Average loss=0.11998533076047897 +INFO:root: +[Valid][14][UTT][ MSE = 0.2273 | LCC = 0.9484 | SRCC = 0.9492 ] +INFO:root:Average loss=0.10940578112006187 +INFO:root: +[Valid][15][UTT][ MSE = 0.2437 | LCC = 0.9479 | SRCC = 0.9484 ] +INFO:root:Average loss=0.11471582758426667 +INFO:root: +[Valid][16][UTT][ MSE = 0.2342 | LCC = 0.9498 | SRCC = 0.9520 ] +INFO:root:Average loss=0.10732698222994805 +INFO:root: +[Valid][17][UTT][ MSE = 0.2494 | LCC = 0.9498 | SRCC = 0.9519 ] +INFO:root:Average loss=0.10284119752049446 +INFO:root: +[Valid][18][UTT][ MSE = 0.2257 | LCC = 0.9474 | SRCC = 0.9495 ] +INFO:root:Average loss=0.10014478981494904 +INFO:root: +[Valid][19][UTT][ MSE = 0.2536 | LCC = 0.9489 | SRCC = 0.9500 ] +INFO:root:Average loss=0.0984404926598072 +INFO:root: +[Valid][20][UTT][ MSE = 0.2243 | LCC = 0.9505 | SRCC = 0.9515 ] +INFO:root:Average loss=0.09856238448619843 +INFO:root: +[Valid][21][UTT][ MSE = 0.2693 | LCC = 0.9485 | SRCC = 0.9505 ] +INFO:root:Average loss=0.09244059398770332 +INFO:root: +[Valid][22][UTT][ MSE = 0.2327 | LCC = 0.9516 | SRCC = 0.9525 ] +INFO:root:Average loss=0.09195635074377059 +INFO:root: +[Valid][23][UTT][ MSE = 0.2463 | LCC = 0.9490 | SRCC = 0.9491 ] +INFO:root:Average loss=0.08759557059407234 +INFO:root: +[Valid][24][UTT][ MSE = 0.2225 | LCC = 0.9481 | SRCC = 0.9483 ] +INFO:root:Average loss=0.08883537465333939 +INFO:root: +[Valid][25][UTT][ MSE = 0.2409 | LCC = 0.9508 | SRCC = 0.9515 ] +INFO:root:Average loss=0.08690361934900284 +INFO:root: +[Valid][26][UTT][ MSE = 0.2213 | LCC = 0.9496 | SRCC = 0.9510 ] +INFO:root:Average loss=0.08501261329650879 +INFO:root: +[Valid][27][UTT][ MSE = 0.2358 | LCC = 0.9510 | SRCC = 0.9522 ] +INFO:root:Average loss=0.07889066174626351 +INFO:root: +[Valid][28][UTT][ MSE = 0.2453 | LCC = 0.9490 | SRCC = 0.9492 ] +INFO:root:Average loss=0.08019016748666763 +INFO:root: +[Valid][29][UTT][ MSE = 0.2484 | LCC = 0.9511 | SRCC = 0.9517 ] +INFO:root:Average loss=0.07827933102846146 +INFO:root: +[Valid][30][UTT][ MSE = 0.2396 | LCC = 0.9517 | SRCC = 0.9521 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2455 | LCC = 0.9521 | SRCC = 0.9536 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..abb00d7ba9e1cc35795a89cfd0262408668ed395 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6488aa670f8811c27b5e5f8d3b2370712e11600faf7e9095a3d598f81ddca8d4 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a1b59b848eae390efe2662bd74ce5b3a9d07c58f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:99e857dc68daaa7ea11ff2d7fb65f950c300b8a45067c3c00f3b02afe8792ba3 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d8a44b26c26da74955292fe33c63575a756a01ed --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.40743475139141083 +INFO:root: +[Valid][0][UTT][ MSE = 0.3463 | LCC = 0.9139 | SRCC = 0.9174 ] +INFO:root:Average loss=0.2175307474732399 +INFO:root: +[Valid][1][UTT][ MSE = 0.2900 | LCC = 0.9289 | SRCC = 0.9338 ] +INFO:root:Average loss=0.1947902617454529 +INFO:root: +[Valid][2][UTT][ MSE = 0.2498 | LCC = 0.9321 | SRCC = 0.9336 ] +INFO:root:Average loss=0.17307522463798522 +INFO:root: +[Valid][3][UTT][ MSE = 0.2773 | LCC = 0.9390 | SRCC = 0.9420 ] +INFO:root:Average loss=0.16670358169078828 +INFO:root: +[Valid][4][UTT][ MSE = 0.2799 | LCC = 0.9351 | SRCC = 0.9360 ] +INFO:root:Average loss=0.15476633197069167 +INFO:root: +[Valid][5][UTT][ MSE = 0.2334 | LCC = 0.9401 | SRCC = 0.9446 ] +INFO:root:Average loss=0.1474744874238968 +INFO:root: +[Valid][6][UTT][ MSE = 0.2780 | LCC = 0.9391 | SRCC = 0.9427 ] +INFO:root:Average loss=0.14120288920402527 +INFO:root: +[Valid][7][UTT][ MSE = 0.2668 | LCC = 0.9417 | SRCC = 0.9436 ] +INFO:root:Average loss=0.1382540831565857 +INFO:root: +[Valid][8][UTT][ MSE = 0.2362 | LCC = 0.9439 | SRCC = 0.9461 ] +INFO:root:Average loss=0.1308819016814232 +INFO:root: +[Valid][9][UTT][ MSE = 0.2513 | LCC = 0.9413 | SRCC = 0.9446 ] +INFO:root:Average loss=0.130722771525383 +INFO:root: +[Valid][10][UTT][ MSE = 0.2574 | LCC = 0.9463 | SRCC = 0.9486 ] +INFO:root:Average loss=0.12423805981874465 +INFO:root: +[Valid][11][UTT][ MSE = 0.2459 | LCC = 0.9488 | SRCC = 0.9508 ] +INFO:root:Average loss=0.11776781976222991 +INFO:root: +[Valid][12][UTT][ MSE = 0.2281 | LCC = 0.9489 | SRCC = 0.9500 ] +INFO:root:Average loss=0.11524169430136681 +INFO:root: +[Valid][13][UTT][ MSE = 0.2473 | LCC = 0.9498 | SRCC = 0.9516 ] +INFO:root:Average loss=0.11179597172141074 +INFO:root: +[Valid][14][UTT][ MSE = 0.2442 | LCC = 0.9493 | SRCC = 0.9518 ] +INFO:root:Average loss=0.11153377372026443 +INFO:root: +[Valid][15][UTT][ MSE = 0.2656 | LCC = 0.9453 | SRCC = 0.9466 ] +INFO:root:Average loss=0.10804919701814651 +INFO:root: +[Valid][16][UTT][ MSE = 0.2795 | LCC = 0.9466 | SRCC = 0.9502 ] +INFO:root:Average loss=0.10984214904904366 +INFO:root: +[Valid][17][UTT][ MSE = 0.2357 | LCC = 0.9457 | SRCC = 0.9489 ] +INFO:root:Average loss=0.10392764961719513 +INFO:root: +[Valid][18][UTT][ MSE = 0.2325 | LCC = 0.9505 | SRCC = 0.9522 ] +INFO:root:Average loss=0.09890953853726386 +INFO:root: +[Valid][19][UTT][ MSE = 0.2343 | LCC = 0.9475 | SRCC = 0.9490 ] +INFO:root:Average loss=0.09911041247844696 +INFO:root: +[Valid][20][UTT][ MSE = 0.2108 | LCC = 0.9506 | SRCC = 0.9527 ] +INFO:root:Average loss=0.0962057380080223 +INFO:root: +[Valid][21][UTT][ MSE = 0.2266 | LCC = 0.9493 | SRCC = 0.9509 ] +INFO:root:Average loss=0.09133183574676514 +INFO:root: +[Valid][22][UTT][ MSE = 0.2215 | LCC = 0.9461 | SRCC = 0.9481 ] +INFO:root:Average loss=0.08905693230032921 +INFO:root: +[Valid][23][UTT][ MSE = 0.2292 | LCC = 0.9519 | SRCC = 0.9526 ] +INFO:root:Average loss=0.08945249515771866 +INFO:root: +[Valid][24][UTT][ MSE = 0.2255 | LCC = 0.9508 | SRCC = 0.9523 ] +INFO:root:Average loss=0.08638399639725686 +INFO:root: +[Valid][25][UTT][ MSE = 0.2353 | LCC = 0.9506 | SRCC = 0.9518 ] +INFO:root:Average loss=0.09285736575722695 +INFO:root: +[Valid][26][UTT][ MSE = 0.2081 | LCC = 0.9502 | SRCC = 0.9525 ] +INFO:root:Average loss=0.0831476933658123 +INFO:root: +[Valid][27][UTT][ MSE = 0.2116 | LCC = 0.9475 | SRCC = 0.9506 ] +INFO:root:Average loss=0.08340623891353607 +INFO:root: +[Valid][28][UTT][ MSE = 0.2399 | LCC = 0.9457 | SRCC = 0.9465 ] +INFO:root:Average loss=0.08166679987311364 +INFO:root: +[Valid][29][UTT][ MSE = 0.2216 | LCC = 0.9507 | SRCC = 0.9523 ] +INFO:root:Average loss=0.07648542049527168 +INFO:root: +[Valid][30][UTT][ MSE = 0.2294 | LCC = 0.9497 | SRCC = 0.9511 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2381 | LCC = 0.9514 | SRCC = 0.9526 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..003936e286ce6a2a269f260cb22e27ddeebf2e88 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f65c8ae1ce1daf6cb8d36d49e182ad108b55982a49d475df886fda8f2526ac85 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c7c101c6862e2afd9e221840a2f9c92b3dfe352c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:abe3027d7e46c30b058d0a611411d776dbbd59a409c903b84af9e6bf9a0b66d9 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e0ac58d53d74426b40c96b8e8ccbc8e653597964 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42384013164043427 +INFO:root: +[Valid][0][UTT][ MSE = 0.3230 | LCC = 0.9173 | SRCC = 0.9224 ] +INFO:root:Average loss=0.2111757550239563 +INFO:root: +[Valid][1][UTT][ MSE = 0.2973 | LCC = 0.9262 | SRCC = 0.9328 ] +INFO:root:Average loss=0.18274634718894958 +INFO:root: +[Valid][2][UTT][ MSE = 0.3300 | LCC = 0.9294 | SRCC = 0.9309 ] +INFO:root:Average loss=0.17066530627012252 +INFO:root: +[Valid][3][UTT][ MSE = 0.2671 | LCC = 0.9384 | SRCC = 0.9412 ] +INFO:root:Average loss=0.16617247754335404 +INFO:root: +[Valid][4][UTT][ MSE = 0.3040 | LCC = 0.9404 | SRCC = 0.9427 ] +INFO:root:Average loss=0.15221384179592132 +INFO:root: +[Valid][5][UTT][ MSE = 0.2739 | LCC = 0.9460 | SRCC = 0.9485 ] +INFO:root:Average loss=0.14647549849748612 +INFO:root: +[Valid][6][UTT][ MSE = 0.2584 | LCC = 0.9387 | SRCC = 0.9402 ] +INFO:root:Average loss=0.13817827373743058 +INFO:root: +[Valid][7][UTT][ MSE = 0.2742 | LCC = 0.9438 | SRCC = 0.9469 ] +INFO:root:Average loss=0.1345999194383621 +INFO:root: +[Valid][8][UTT][ MSE = 0.2499 | LCC = 0.9438 | SRCC = 0.9454 ] +INFO:root:Average loss=0.13138851910829544 +INFO:root: +[Valid][9][UTT][ MSE = 0.2561 | LCC = 0.9467 | SRCC = 0.9493 ] +INFO:root:Average loss=0.12846589678525924 +INFO:root: +[Valid][10][UTT][ MSE = 0.2457 | LCC = 0.9475 | SRCC = 0.9488 ] +INFO:root:Average loss=0.12280567482113838 +INFO:root: +[Valid][11][UTT][ MSE = 0.2536 | LCC = 0.9453 | SRCC = 0.9475 ] +INFO:root:Average loss=0.11954163271188736 +INFO:root: +[Valid][12][UTT][ MSE = 0.2257 | LCC = 0.9493 | SRCC = 0.9506 ] +INFO:root:Average loss=0.1153273150920868 +INFO:root: +[Valid][13][UTT][ MSE = 0.2298 | LCC = 0.9509 | SRCC = 0.9523 ] +INFO:root:Average loss=0.11428883284330368 +INFO:root: +[Valid][14][UTT][ MSE = 0.2691 | LCC = 0.9491 | SRCC = 0.9511 ] +INFO:root:Average loss=0.10902082261443138 +INFO:root: +[Valid][15][UTT][ MSE = 0.2405 | LCC = 0.9491 | SRCC = 0.9496 ] +INFO:root:Average loss=0.10611328047513961 +INFO:root: +[Valid][16][UTT][ MSE = 0.2379 | LCC = 0.9469 | SRCC = 0.9490 ] +INFO:root:Average loss=0.10156619477272033 +INFO:root: +[Valid][17][UTT][ MSE = 0.2617 | LCC = 0.9461 | SRCC = 0.9479 ] +INFO:root:Average loss=0.10235964620113373 +INFO:root: +[Valid][18][UTT][ MSE = 0.2583 | LCC = 0.9491 | SRCC = 0.9510 ] +INFO:root:Average loss=0.10289288729429245 +INFO:root: +[Valid][19][UTT][ MSE = 0.2768 | LCC = 0.9487 | SRCC = 0.9506 ] +INFO:root:Average loss=0.09728418692946433 +INFO:root: +[Valid][20][UTT][ MSE = 0.2512 | LCC = 0.9499 | SRCC = 0.9519 ] +INFO:root:Average loss=0.10255775305628777 +INFO:root: +[Valid][21][UTT][ MSE = 0.2476 | LCC = 0.9504 | SRCC = 0.9516 ] +INFO:root:Average loss=0.09276688545942306 +INFO:root: +[Valid][22][UTT][ MSE = 0.2537 | LCC = 0.9488 | SRCC = 0.9495 ] +INFO:root:Average loss=0.08890741404891014 +INFO:root: +[Valid][23][UTT][ MSE = 0.2564 | LCC = 0.9494 | SRCC = 0.9499 ] +INFO:root:Average loss=0.08691971588134766 +INFO:root: +[Valid][24][UTT][ MSE = 0.2216 | LCC = 0.9509 | SRCC = 0.9526 ] +INFO:root:Average loss=0.08651251921057701 +INFO:root: +[Valid][25][UTT][ MSE = 0.2643 | LCC = 0.9497 | SRCC = 0.9508 ] +INFO:root:Average loss=0.08914873442053795 +INFO:root: +[Valid][26][UTT][ MSE = 0.2543 | LCC = 0.9481 | SRCC = 0.9497 ] +INFO:root:Average loss=0.08197639241814614 +INFO:root: +[Valid][27][UTT][ MSE = 0.2321 | LCC = 0.9505 | SRCC = 0.9507 ] +INFO:root:Average loss=0.08310658821463585 +INFO:root: +[Valid][28][UTT][ MSE = 0.2500 | LCC = 0.9464 | SRCC = 0.9464 ] +INFO:root:Average loss=0.08130264022946358 +INFO:root: +[Valid][29][UTT][ MSE = 0.2246 | LCC = 0.9509 | SRCC = 0.9518 ] +INFO:root:Average loss=0.07801731744408608 +INFO:root: +[Valid][30][UTT][ MSE = 0.2309 | LCC = 0.9523 | SRCC = 0.9527 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2376 | LCC = 0.9533 | SRCC = 0.9550 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..25fcfecf9fe45046b11cf00d86ce247b66775e6d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6ad3e3a9130de11b2707531c7d9173edf5ed1d947af330ae825620c0182bd6f8 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..36f27b834aa643ce2ec1807fb9dd9a5b9176842c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d974f7a81039b5c016a2980ea2dff41f7efac60b166e84ad51d78198732908a0 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5cdaf1e43482d3c4c078713584a8ab990157f726 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer10/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3601028636693954 +INFO:root: +[Valid][0][UTT][ MSE = 0.3700 | LCC = 0.9196 | SRCC = 0.9234 ] +INFO:root:Average loss=0.20719003224372864 +INFO:root: +[Valid][1][UTT][ MSE = 0.2633 | LCC = 0.9310 | SRCC = 0.9338 ] +INFO:root:Average loss=0.18701741701364516 +INFO:root: +[Valid][2][UTT][ MSE = 0.2545 | LCC = 0.9376 | SRCC = 0.9397 ] +INFO:root:Average loss=0.16908651602268218 +INFO:root: +[Valid][3][UTT][ MSE = 0.2894 | LCC = 0.9365 | SRCC = 0.9402 ] +INFO:root:Average loss=0.16108353090286254 +INFO:root: +[Valid][4][UTT][ MSE = 0.2898 | LCC = 0.9404 | SRCC = 0.9417 ] +INFO:root:Average loss=0.15046485722064973 +INFO:root: +[Valid][5][UTT][ MSE = 0.2462 | LCC = 0.9429 | SRCC = 0.9460 ] +INFO:root:Average loss=0.14038673156499862 +INFO:root: +[Valid][6][UTT][ MSE = 0.2857 | LCC = 0.9445 | SRCC = 0.9472 ] +INFO:root:Average loss=0.14159893822669983 +INFO:root: +[Valid][7][UTT][ MSE = 0.2251 | LCC = 0.9415 | SRCC = 0.9459 ] +INFO:root:Average loss=0.13502092760801315 +INFO:root: +[Valid][8][UTT][ MSE = 0.2851 | LCC = 0.9452 | SRCC = 0.9472 ] +INFO:root:Average loss=0.12660063362121582 +INFO:root: +[Valid][9][UTT][ MSE = 0.2726 | LCC = 0.9440 | SRCC = 0.9469 ] +INFO:root:Average loss=0.12546887975931167 +INFO:root: +[Valid][10][UTT][ MSE = 0.2554 | LCC = 0.9438 | SRCC = 0.9451 ] +INFO:root:Average loss=0.12114377933740615 +INFO:root: +[Valid][11][UTT][ MSE = 0.2651 | LCC = 0.9454 | SRCC = 0.9487 ] +INFO:root:Average loss=0.11805935227870941 +INFO:root: +[Valid][12][UTT][ MSE = 0.2871 | LCC = 0.9462 | SRCC = 0.9496 ] +INFO:root:Average loss=0.11712227880954743 +INFO:root: +[Valid][13][UTT][ MSE = 0.2207 | LCC = 0.9491 | SRCC = 0.9504 ] +INFO:root:Average loss=0.10774191501736641 +INFO:root: +[Valid][14][UTT][ MSE = 0.2337 | LCC = 0.9487 | SRCC = 0.9503 ] +INFO:root:Average loss=0.11197157639265061 +INFO:root: +[Valid][15][UTT][ MSE = 0.2304 | LCC = 0.9510 | SRCC = 0.9522 ] +INFO:root:Average loss=0.10908032888174057 +INFO:root: +[Valid][16][UTT][ MSE = 0.2271 | LCC = 0.9496 | SRCC = 0.9503 ] +INFO:root:Average loss=0.10247273117303848 +INFO:root: +[Valid][17][UTT][ MSE = 0.2717 | LCC = 0.9487 | SRCC = 0.9505 ] +INFO:root:Average loss=0.10406913498044014 +INFO:root: +[Valid][18][UTT][ MSE = 0.2544 | LCC = 0.9483 | SRCC = 0.9502 ] +INFO:root:Average loss=0.09657077863812447 +INFO:root: +[Valid][19][UTT][ MSE = 0.2328 | LCC = 0.9494 | SRCC = 0.9507 ] +INFO:root:Average loss=0.09916670548915862 +INFO:root: +[Valid][20][UTT][ MSE = 0.2306 | LCC = 0.9491 | SRCC = 0.9499 ] +INFO:root:Average loss=0.0933251587152481 +INFO:root: +[Valid][21][UTT][ MSE = 0.2273 | LCC = 0.9457 | SRCC = 0.9464 ] +INFO:root:Average loss=0.09335344982147217 +INFO:root: +[Valid][22][UTT][ MSE = 0.2197 | LCC = 0.9489 | SRCC = 0.9498 ] +INFO:root:Average loss=0.088473398655653 +INFO:root: +[Valid][23][UTT][ MSE = 0.2510 | LCC = 0.9474 | SRCC = 0.9485 ] +INFO:root:Average loss=0.08622092866897584 +INFO:root: +[Valid][24][UTT][ MSE = 0.2455 | LCC = 0.9490 | SRCC = 0.9503 ] +INFO:root:Average loss=0.08561786609888077 +INFO:root: +[Valid][25][UTT][ MSE = 0.2545 | LCC = 0.9488 | SRCC = 0.9505 ] +INFO:root:Average loss=0.086263601064682 +INFO:root: +[Valid][26][UTT][ MSE = 0.2266 | LCC = 0.9496 | SRCC = 0.9504 ] +INFO:root:Average loss=0.08285729178786277 +INFO:root: +[Valid][27][UTT][ MSE = 0.2477 | LCC = 0.9480 | SRCC = 0.9491 ] +INFO:root:Average loss=0.08354954022169113 +INFO:root: +[Valid][28][UTT][ MSE = 0.2437 | LCC = 0.9487 | SRCC = 0.9494 ] +INFO:root:Average loss=0.07656415206193924 +INFO:root: +[Valid][29][UTT][ MSE = 0.2426 | LCC = 0.9458 | SRCC = 0.9466 ] +INFO:root:Average loss=0.08045700454711914 +INFO:root: +[Valid][30][UTT][ MSE = 0.2601 | LCC = 0.9464 | SRCC = 0.9474 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2308 | LCC = 0.9550 | SRCC = 0.9568 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b490b1dcb42de3d4d3f1d19a6d3fdc661fa6951b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e3be639d35854b766f6fea9b56fbf6c3de4aac162847c37658beb14f50a6355d +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..86f46fa21d07333fd8d0bb2e18da71e7fd060a2c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:350b9a2ed6aa345afa2620e20c92ba48bcdd7069bdb55998daad691e31f4de18 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c55745b2d0224bc11d5a73f9b1effaf895bfca11 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.41522878241539 +INFO:root: +[Valid][0][UTT][ MSE = 0.4301 | LCC = 0.9122 | SRCC = 0.9149 ] +INFO:root:Average loss=0.2421001250743866 +INFO:root: +[Valid][1][UTT][ MSE = 0.3240 | LCC = 0.9254 | SRCC = 0.9294 ] +INFO:root:Average loss=0.20548478776216508 +INFO:root: +[Valid][2][UTT][ MSE = 0.2879 | LCC = 0.9293 | SRCC = 0.9354 ] +INFO:root:Average loss=0.19374740624427794 +INFO:root: +[Valid][3][UTT][ MSE = 0.3068 | LCC = 0.9339 | SRCC = 0.9368 ] +INFO:root:Average loss=0.18645928245782853 +INFO:root: +[Valid][4][UTT][ MSE = 0.2759 | LCC = 0.9351 | SRCC = 0.9379 ] +INFO:root:Average loss=0.16930195671319961 +INFO:root: +[Valid][5][UTT][ MSE = 0.3108 | LCC = 0.9382 | SRCC = 0.9404 ] +INFO:root:Average loss=0.1615273047685623 +INFO:root: +[Valid][6][UTT][ MSE = 0.2768 | LCC = 0.9364 | SRCC = 0.9391 ] +INFO:root:Average loss=0.16388378608226775 +INFO:root: +[Valid][7][UTT][ MSE = 0.3302 | LCC = 0.9376 | SRCC = 0.9383 ] +INFO:root:Average loss=0.16093012231588363 +INFO:root: +[Valid][8][UTT][ MSE = 0.2673 | LCC = 0.9411 | SRCC = 0.9422 ] +INFO:root:Average loss=0.15463815063238143 +INFO:root: +[Valid][9][UTT][ MSE = 0.2868 | LCC = 0.9345 | SRCC = 0.9375 ] +INFO:root:Average loss=0.14618439769744873 +INFO:root: +[Valid][10][UTT][ MSE = 0.2689 | LCC = 0.9394 | SRCC = 0.9413 ] +INFO:root:Average loss=0.15118766176700593 +INFO:root: +[Valid][11][UTT][ MSE = 0.2669 | LCC = 0.9426 | SRCC = 0.9447 ] +INFO:root:Average loss=0.14646750873327255 +INFO:root: +[Valid][12][UTT][ MSE = 0.2585 | LCC = 0.9397 | SRCC = 0.9414 ] +INFO:root:Average loss=0.138441412627697 +INFO:root: +[Valid][13][UTT][ MSE = 0.2477 | LCC = 0.9401 | SRCC = 0.9439 ] +INFO:root:Average loss=0.13394436755776407 +INFO:root: +[Valid][14][UTT][ MSE = 0.2457 | LCC = 0.9394 | SRCC = 0.9415 ] +INFO:root:Average loss=0.13256582289934157 +INFO:root: +[Valid][15][UTT][ MSE = 0.2500 | LCC = 0.9430 | SRCC = 0.9449 ] +INFO:root:Average loss=0.12742667776346206 +INFO:root: +[Valid][16][UTT][ MSE = 0.2529 | LCC = 0.9451 | SRCC = 0.9463 ] +INFO:root:Average loss=0.13035705387592317 +INFO:root: +[Valid][17][UTT][ MSE = 0.2470 | LCC = 0.9451 | SRCC = 0.9465 ] +INFO:root:Average loss=0.12680176216363906 +INFO:root: +[Valid][18][UTT][ MSE = 0.2645 | LCC = 0.9450 | SRCC = 0.9465 ] +INFO:root:Average loss=0.13382211816310882 +INFO:root: +[Valid][19][UTT][ MSE = 0.2608 | LCC = 0.9456 | SRCC = 0.9475 ] +INFO:root:Average loss=0.12334785217046737 +INFO:root: +[Valid][20][UTT][ MSE = 0.2243 | LCC = 0.9458 | SRCC = 0.9481 ] +INFO:root:Average loss=0.11631283086538315 +INFO:root: +[Valid][21][UTT][ MSE = 0.2520 | LCC = 0.9465 | SRCC = 0.9483 ] +INFO:root:Average loss=0.1154336560368538 +INFO:root: +[Valid][22][UTT][ MSE = 0.2537 | LCC = 0.9457 | SRCC = 0.9466 ] +INFO:root:Average loss=0.11044210559129715 +INFO:root: +[Valid][23][UTT][ MSE = 0.2672 | LCC = 0.9434 | SRCC = 0.9447 ] +INFO:root:Average loss=0.11324213546514511 +INFO:root: +[Valid][24][UTT][ MSE = 0.2984 | LCC = 0.9420 | SRCC = 0.9443 ] +INFO:root:Average loss=0.11302363926172257 +INFO:root: +[Valid][25][UTT][ MSE = 0.2755 | LCC = 0.9476 | SRCC = 0.9503 ] +INFO:root:Average loss=0.11051336151361466 +INFO:root: +[Valid][26][UTT][ MSE = 0.2703 | LCC = 0.9470 | SRCC = 0.9481 ] +INFO:root:Average loss=0.10669129636883735 +INFO:root: +[Valid][27][UTT][ MSE = 0.2505 | LCC = 0.9457 | SRCC = 0.9467 ] +INFO:root:Average loss=0.10273310586810112 +INFO:root: +[Valid][28][UTT][ MSE = 0.2445 | LCC = 0.9480 | SRCC = 0.9499 ] +INFO:root:Average loss=0.1045612205862999 +INFO:root: +[Valid][29][UTT][ MSE = 0.2680 | LCC = 0.9459 | SRCC = 0.9474 ] +INFO:root:Average loss=0.10041570165753365 +INFO:root: +[Valid][30][UTT][ MSE = 0.2551 | LCC = 0.9461 | SRCC = 0.9481 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2460 | LCC = 0.9497 | SRCC = 0.9511 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f4288495285989c9be49b0915c467f33b76cf67a --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:131d6985e8da97b459aa7da7a5d9250de5c3f8e693eb937ff7c43041a63536ce +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b2000cf130feae5d5a346eb6c2f2a99b063fd4bf --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ca590afa577a73831b110d44935afba405eb161e6ba332fce3915f1e78ad5bb8 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..871322dd64d7b951c33974a7180cd77fc8850c3f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3368654507398605 +INFO:root: +[Valid][0][UTT][ MSE = 0.3559 | LCC = 0.9179 | SRCC = 0.9209 ] +INFO:root:Average loss=0.224651160299778 +INFO:root: +[Valid][1][UTT][ MSE = 0.3488 | LCC = 0.9259 | SRCC = 0.9284 ] +INFO:root:Average loss=0.2071571618914604 +INFO:root: +[Valid][2][UTT][ MSE = 0.3274 | LCC = 0.9304 | SRCC = 0.9320 ] +INFO:root:Average loss=0.19641174334287642 +INFO:root: +[Valid][3][UTT][ MSE = 0.3236 | LCC = 0.9287 | SRCC = 0.9328 ] +INFO:root:Average loss=0.18482674193382262 +INFO:root: +[Valid][4][UTT][ MSE = 0.2726 | LCC = 0.9364 | SRCC = 0.9384 ] +INFO:root:Average loss=0.16882791876792907 +INFO:root: +[Valid][5][UTT][ MSE = 0.3328 | LCC = 0.9333 | SRCC = 0.9349 ] +INFO:root:Average loss=0.17143876552581788 +INFO:root: +[Valid][6][UTT][ MSE = 0.2655 | LCC = 0.9393 | SRCC = 0.9421 ] +INFO:root:Average loss=0.16020436030626298 +INFO:root: +[Valid][7][UTT][ MSE = 0.3042 | LCC = 0.9403 | SRCC = 0.9423 ] +INFO:root:Average loss=0.16081881511211396 +INFO:root: +[Valid][8][UTT][ MSE = 0.3098 | LCC = 0.9419 | SRCC = 0.9449 ] +INFO:root:Average loss=0.15237155824899673 +INFO:root: +[Valid][9][UTT][ MSE = 0.3117 | LCC = 0.9411 | SRCC = 0.9436 ] +INFO:root:Average loss=0.15261457258462907 +INFO:root: +[Valid][10][UTT][ MSE = 0.2685 | LCC = 0.9417 | SRCC = 0.9437 ] +INFO:root:Average loss=0.14573836332559587 +INFO:root: +[Valid][11][UTT][ MSE = 0.2766 | LCC = 0.9440 | SRCC = 0.9461 ] +INFO:root:Average loss=0.14044533789157868 +INFO:root: +[Valid][12][UTT][ MSE = 0.2968 | LCC = 0.9392 | SRCC = 0.9413 ] +INFO:root:Average loss=0.14030967485904694 +INFO:root: +[Valid][13][UTT][ MSE = 0.2506 | LCC = 0.9440 | SRCC = 0.9463 ] +INFO:root:Average loss=0.13056066703796387 +INFO:root: +[Valid][14][UTT][ MSE = 0.2846 | LCC = 0.9454 | SRCC = 0.9471 ] +INFO:root:Average loss=0.1331564946770668 +INFO:root: +[Valid][15][UTT][ MSE = 0.2844 | LCC = 0.9409 | SRCC = 0.9439 ] +INFO:root:Average loss=0.12926165115833282 +INFO:root: +[Valid][16][UTT][ MSE = 0.2636 | LCC = 0.9459 | SRCC = 0.9474 ] +INFO:root:Average loss=0.12280038416385651 +INFO:root: +[Valid][17][UTT][ MSE = 0.2469 | LCC = 0.9449 | SRCC = 0.9459 ] +INFO:root:Average loss=0.1228160738348961 +INFO:root: +[Valid][18][UTT][ MSE = 0.2425 | LCC = 0.9457 | SRCC = 0.9468 ] +INFO:root:Average loss=0.1172400376200676 +INFO:root: +[Valid][19][UTT][ MSE = 0.2747 | LCC = 0.9475 | SRCC = 0.9492 ] +INFO:root:Average loss=0.11417147773504258 +INFO:root: +[Valid][20][UTT][ MSE = 0.2430 | LCC = 0.9471 | SRCC = 0.9480 ] +INFO:root:Average loss=0.1185100370645523 +INFO:root: +[Valid][21][UTT][ MSE = 0.2756 | LCC = 0.9477 | SRCC = 0.9490 ] +INFO:root:Average loss=0.11466803535819053 +INFO:root: +[Valid][22][UTT][ MSE = 0.2591 | LCC = 0.9444 | SRCC = 0.9463 ] +INFO:root:Average loss=0.10833827111124993 +INFO:root: +[Valid][23][UTT][ MSE = 0.2479 | LCC = 0.9480 | SRCC = 0.9491 ] +INFO:root:Average loss=0.10996413165330887 +INFO:root: +[Valid][24][UTT][ MSE = 0.2429 | LCC = 0.9475 | SRCC = 0.9492 ] +INFO:root:Average loss=0.10542584979534149 +INFO:root: +[Valid][25][UTT][ MSE = 0.2441 | LCC = 0.9456 | SRCC = 0.9468 ] +INFO:root:Average loss=0.10347605589032173 +INFO:root: +[Valid][26][UTT][ MSE = 0.2629 | LCC = 0.9409 | SRCC = 0.9428 ] +INFO:root:Average loss=0.10097892567515374 +INFO:root: +[Valid][27][UTT][ MSE = 0.2695 | LCC = 0.9455 | SRCC = 0.9473 ] +INFO:root:Average loss=0.09776717606186867 +INFO:root: +[Valid][28][UTT][ MSE = 0.2600 | LCC = 0.9421 | SRCC = 0.9444 ] +INFO:root:Average loss=0.09804394778609275 +INFO:root: +[Valid][29][UTT][ MSE = 0.2365 | LCC = 0.9412 | SRCC = 0.9448 ] +INFO:root:Average loss=0.0975721144080162 +INFO:root: +[Valid][30][UTT][ MSE = 0.2874 | LCC = 0.9444 | SRCC = 0.9456 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2522 | LCC = 0.9485 | SRCC = 0.9506 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0b17e8fb9d163aa35e28d8403825f03c39c0669b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e055b28bbf39b1b7cebcafc829e4e36c2fb45455b745025cbe349cc070aaa638 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ce09406f834b2dd661846554c0189513839b82bc --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:95b3a26128309d8fdc9d9504f9748f72a16ec4d3d97c87945ce21434965d0ecb +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c647f84c78e041d9ff0ff38e07c6d16131f7ea39 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42007274127006533 +INFO:root: +[Valid][0][UTT][ MSE = 0.3310 | LCC = 0.9104 | SRCC = 0.9101 ] +INFO:root:Average loss=0.23298434698581696 +INFO:root: +[Valid][1][UTT][ MSE = 0.2862 | LCC = 0.9257 | SRCC = 0.9284 ] +INFO:root:Average loss=0.20482568079233168 +INFO:root: +[Valid][2][UTT][ MSE = 0.3678 | LCC = 0.9306 | SRCC = 0.9323 ] +INFO:root:Average loss=0.19232718032598495 +INFO:root: +[Valid][3][UTT][ MSE = 0.2860 | LCC = 0.9298 | SRCC = 0.9346 ] +INFO:root:Average loss=0.17974474358558654 +INFO:root: +[Valid][4][UTT][ MSE = 0.2752 | LCC = 0.9370 | SRCC = 0.9389 ] +INFO:root:Average loss=0.17594153428077697 +INFO:root: +[Valid][5][UTT][ MSE = 0.2964 | LCC = 0.9360 | SRCC = 0.9381 ] +INFO:root:Average loss=0.17025170928239822 +INFO:root: +[Valid][6][UTT][ MSE = 0.2625 | LCC = 0.9387 | SRCC = 0.9408 ] +INFO:root:Average loss=0.16214322113990784 +INFO:root: +[Valid][7][UTT][ MSE = 0.2723 | LCC = 0.9411 | SRCC = 0.9423 ] +INFO:root:Average loss=0.1577310982942581 +INFO:root: +[Valid][8][UTT][ MSE = 0.2628 | LCC = 0.9414 | SRCC = 0.9443 ] +INFO:root:Average loss=0.15049179905653 +INFO:root: +[Valid][9][UTT][ MSE = 0.2483 | LCC = 0.9429 | SRCC = 0.9457 ] +INFO:root:Average loss=0.1496007301211357 +INFO:root: +[Valid][10][UTT][ MSE = 0.2641 | LCC = 0.9403 | SRCC = 0.9412 ] +INFO:root:Average loss=0.14547106564044954 +INFO:root: +[Valid][11][UTT][ MSE = 0.2465 | LCC = 0.9444 | SRCC = 0.9467 ] +INFO:root:Average loss=0.13890598827600478 +INFO:root: +[Valid][12][UTT][ MSE = 0.2661 | LCC = 0.9436 | SRCC = 0.9452 ] +INFO:root:Average loss=0.1395161492228508 +INFO:root: +[Valid][13][UTT][ MSE = 0.2682 | LCC = 0.9436 | SRCC = 0.9445 ] +INFO:root:Average loss=0.1384703884422779 +INFO:root: +[Valid][14][UTT][ MSE = 0.2638 | LCC = 0.9468 | SRCC = 0.9483 ] +INFO:root:Average loss=0.13021820449829102 +INFO:root: +[Valid][15][UTT][ MSE = 0.2946 | LCC = 0.9429 | SRCC = 0.9443 ] +INFO:root:Average loss=0.12981693452596665 +INFO:root: +[Valid][16][UTT][ MSE = 0.2712 | LCC = 0.9377 | SRCC = 0.9396 ] +INFO:root:Average loss=0.12749102586507796 +INFO:root: +[Valid][17][UTT][ MSE = 0.2567 | LCC = 0.9473 | SRCC = 0.9490 ] +INFO:root:Average loss=0.12129549211263657 +INFO:root: +[Valid][18][UTT][ MSE = 0.2456 | LCC = 0.9437 | SRCC = 0.9457 ] +INFO:root:Average loss=0.12450907871127129 +INFO:root: +[Valid][19][UTT][ MSE = 0.2394 | LCC = 0.9445 | SRCC = 0.9466 ] +INFO:root:Average loss=0.11923452842235566 +INFO:root: +[Valid][20][UTT][ MSE = 0.2771 | LCC = 0.9443 | SRCC = 0.9480 ] +INFO:root:Average loss=0.11494418650865555 +INFO:root: +[Valid][21][UTT][ MSE = 0.2119 | LCC = 0.9441 | SRCC = 0.9464 ] +INFO:root:Average loss=0.11604009234905242 +INFO:root: +[Valid][22][UTT][ MSE = 0.2223 | LCC = 0.9474 | SRCC = 0.9489 ] +INFO:root:Average loss=0.11141461083292961 +INFO:root: +[Valid][23][UTT][ MSE = 0.2109 | LCC = 0.9483 | SRCC = 0.9495 ] +INFO:root:Average loss=0.11599620443582535 +INFO:root: +[Valid][24][UTT][ MSE = 0.2874 | LCC = 0.9425 | SRCC = 0.9429 ] +INFO:root:Average loss=0.11263208001852036 +INFO:root: +[Valid][25][UTT][ MSE = 0.2251 | LCC = 0.9464 | SRCC = 0.9488 ] +INFO:root:Average loss=0.10467201289534568 +INFO:root: +[Valid][26][UTT][ MSE = 0.2797 | LCC = 0.9450 | SRCC = 0.9478 ] +INFO:root:Average loss=0.10391859462857246 +INFO:root: +[Valid][27][UTT][ MSE = 0.2596 | LCC = 0.9467 | SRCC = 0.9477 ] +INFO:root:Average loss=0.09849146217107772 +INFO:root: +[Valid][28][UTT][ MSE = 0.2420 | LCC = 0.9453 | SRCC = 0.9468 ] +INFO:root:Average loss=0.09823001778125763 +INFO:root: +[Valid][29][UTT][ MSE = 0.2269 | LCC = 0.9469 | SRCC = 0.9491 ] +INFO:root:Average loss=0.09940572443604469 +INFO:root: +[Valid][30][UTT][ MSE = 0.2359 | LCC = 0.9487 | SRCC = 0.9505 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2385 | LCC = 0.9482 | SRCC = 0.9502 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..80b9f59a90fca38de09e5e17143f1e163d6bd3cc --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:71fb3ea865e30abb161f32cad6bed808c27f774c2e30f944fc5e3794d62d477c +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..3d4de1ec68c85dba722fa8a1146ad2c86a1c50fe --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ec0631ba7b4ec714c8d475db30a93d7e18c46b58e67a40131a76ea53693179f5 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a37a56c53238853d783a4ccef942567a87bef0ff --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3702618168592453 +INFO:root: +[Valid][0][UTT][ MSE = 0.3850 | LCC = 0.9079 | SRCC = 0.9110 ] +INFO:root:Average loss=0.23379302901029586 +INFO:root: +[Valid][1][UTT][ MSE = 0.3783 | LCC = 0.9246 | SRCC = 0.9290 ] +INFO:root:Average loss=0.21051879203319548 +INFO:root: +[Valid][2][UTT][ MSE = 0.3804 | LCC = 0.9205 | SRCC = 0.9223 ] +INFO:root:Average loss=0.19099934709072114 +INFO:root: +[Valid][3][UTT][ MSE = 0.3276 | LCC = 0.9327 | SRCC = 0.9361 ] +INFO:root:Average loss=0.18025127136707306 +INFO:root: +[Valid][4][UTT][ MSE = 0.2928 | LCC = 0.9350 | SRCC = 0.9389 ] +INFO:root:Average loss=0.17647414326667785 +INFO:root: +[Valid][5][UTT][ MSE = 0.2935 | LCC = 0.9320 | SRCC = 0.9380 ] +INFO:root:Average loss=0.1634721393585205 +INFO:root: +[Valid][6][UTT][ MSE = 0.3456 | LCC = 0.9351 | SRCC = 0.9368 ] +INFO:root:Average loss=0.16132601326704024 +INFO:root: +[Valid][7][UTT][ MSE = 0.3085 | LCC = 0.9334 | SRCC = 0.9368 ] +INFO:root:Average loss=0.15842292958498 +INFO:root: +[Valid][8][UTT][ MSE = 0.2946 | LCC = 0.9372 | SRCC = 0.9410 ] +INFO:root:Average loss=0.15480930387973785 +INFO:root: +[Valid][9][UTT][ MSE = 0.2631 | LCC = 0.9307 | SRCC = 0.9333 ] +INFO:root:Average loss=0.14631808668375015 +INFO:root: +[Valid][10][UTT][ MSE = 0.2950 | LCC = 0.9367 | SRCC = 0.9380 ] +INFO:root:Average loss=0.14869032675027846 +INFO:root: +[Valid][11][UTT][ MSE = 0.3086 | LCC = 0.9415 | SRCC = 0.9438 ] +INFO:root:Average loss=0.14410859405994414 +INFO:root: +[Valid][12][UTT][ MSE = 0.2974 | LCC = 0.9422 | SRCC = 0.9446 ] +INFO:root:Average loss=0.13514404082298279 +INFO:root: +[Valid][13][UTT][ MSE = 0.2833 | LCC = 0.9377 | SRCC = 0.9410 ] +INFO:root:Average loss=0.13268422001600266 +INFO:root: +[Valid][14][UTT][ MSE = 0.2262 | LCC = 0.9444 | SRCC = 0.9465 ] +INFO:root:Average loss=0.13343682610988616 +INFO:root: +[Valid][15][UTT][ MSE = 0.2724 | LCC = 0.9457 | SRCC = 0.9474 ] +INFO:root:Average loss=0.12816194850206375 +INFO:root: +[Valid][16][UTT][ MSE = 0.2875 | LCC = 0.9418 | SRCC = 0.9456 ] +INFO:root:Average loss=0.12735943180322648 +INFO:root: +[Valid][17][UTT][ MSE = 0.2938 | LCC = 0.9435 | SRCC = 0.9468 ] +INFO:root:Average loss=0.1221565267443657 +INFO:root: +[Valid][18][UTT][ MSE = 0.2602 | LCC = 0.9441 | SRCC = 0.9451 ] +INFO:root:Average loss=0.12166238275170327 +INFO:root: +[Valid][19][UTT][ MSE = 0.2829 | LCC = 0.9418 | SRCC = 0.9444 ] +INFO:root:Average loss=0.12124770736694336 +INFO:root: +[Valid][20][UTT][ MSE = 0.2899 | LCC = 0.9446 | SRCC = 0.9468 ] +INFO:root:Average loss=0.11799427098035813 +INFO:root: +[Valid][21][UTT][ MSE = 0.2498 | LCC = 0.9392 | SRCC = 0.9418 ] +INFO:root:Average loss=0.11529795849323272 +INFO:root: +[Valid][22][UTT][ MSE = 0.2461 | LCC = 0.9447 | SRCC = 0.9470 ] +INFO:root:Average loss=0.11176589131355286 +INFO:root: +[Valid][23][UTT][ MSE = 0.2599 | LCC = 0.9457 | SRCC = 0.9470 ] +INFO:root:Average loss=0.11336428275704384 +INFO:root: +[Valid][24][UTT][ MSE = 0.2827 | LCC = 0.9458 | SRCC = 0.9473 ] +INFO:root:Average loss=0.10582331508398056 +INFO:root: +[Valid][25][UTT][ MSE = 0.2524 | LCC = 0.9428 | SRCC = 0.9443 ] +INFO:root:Average loss=0.10699823862314224 +INFO:root: +[Valid][26][UTT][ MSE = 0.2499 | LCC = 0.9454 | SRCC = 0.9463 ] +INFO:root:Average loss=0.10738137209415435 +INFO:root: +[Valid][27][UTT][ MSE = 0.2576 | LCC = 0.9423 | SRCC = 0.9431 ] +INFO:root:Average loss=0.10311560827493668 +INFO:root: +[Valid][28][UTT][ MSE = 0.2435 | LCC = 0.9453 | SRCC = 0.9479 ] +INFO:root:Average loss=0.10089719301462173 +INFO:root: +[Valid][29][UTT][ MSE = 0.2551 | LCC = 0.9451 | SRCC = 0.9466 ] +INFO:root:Average loss=0.10163188624382019 +INFO:root: +[Valid][30][UTT][ MSE = 0.2809 | LCC = 0.9370 | SRCC = 0.9381 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2968 | LCC = 0.9449 | SRCC = 0.9475 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..fc865f5b685c7775ee8e2eb23c75e1146f97460a --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c11dc24851d45593d073c4b7dac9607bd16ed644c03cf45d2939175038f554f2 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7596f3769c92aef1182cc8fb7229c72f347b7731 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0f014919d9050cda1650a2e3997a63cceadf7938a4fba4d353564537e6e8bd1a +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..10058e4b5ee484a12c32cc7ccb73656013952eda --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer11/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.395048437833786 +INFO:root: +[Valid][0][UTT][ MSE = 0.3313 | LCC = 0.9124 | SRCC = 0.9159 ] +INFO:root:Average loss=0.22984091508388518 +INFO:root: +[Valid][1][UTT][ MSE = 0.3978 | LCC = 0.9215 | SRCC = 0.9230 ] +INFO:root:Average loss=0.21489734971523286 +INFO:root: +[Valid][2][UTT][ MSE = 0.3018 | LCC = 0.9265 | SRCC = 0.9284 ] +INFO:root:Average loss=0.1927092844247818 +INFO:root: +[Valid][3][UTT][ MSE = 0.2821 | LCC = 0.9301 | SRCC = 0.9321 ] +INFO:root:Average loss=0.18766978085041047 +INFO:root: +[Valid][4][UTT][ MSE = 0.3088 | LCC = 0.9301 | SRCC = 0.9339 ] +INFO:root:Average loss=0.17477403551340104 +INFO:root: +[Valid][5][UTT][ MSE = 0.2641 | LCC = 0.9384 | SRCC = 0.9399 ] +INFO:root:Average loss=0.1680288386940956 +INFO:root: +[Valid][6][UTT][ MSE = 0.3191 | LCC = 0.9279 | SRCC = 0.9317 ] +INFO:root:Average loss=0.1583864072561264 +INFO:root: +[Valid][7][UTT][ MSE = 0.3294 | LCC = 0.9394 | SRCC = 0.9426 ] +INFO:root:Average loss=0.16309096163511277 +INFO:root: +[Valid][8][UTT][ MSE = 0.2894 | LCC = 0.9419 | SRCC = 0.9443 ] +INFO:root:Average loss=0.1505994525551796 +INFO:root: +[Valid][9][UTT][ MSE = 0.2514 | LCC = 0.9405 | SRCC = 0.9435 ] +INFO:root:Average loss=0.14686058163642884 +INFO:root: +[Valid][10][UTT][ MSE = 0.2745 | LCC = 0.9388 | SRCC = 0.9415 ] +INFO:root:Average loss=0.14826195979118348 +INFO:root: +[Valid][11][UTT][ MSE = 0.2669 | LCC = 0.9438 | SRCC = 0.9453 ] +INFO:root:Average loss=0.14302513480186463 +INFO:root: +[Valid][12][UTT][ MSE = 0.2462 | LCC = 0.9369 | SRCC = 0.9405 ] +INFO:root:Average loss=0.13684074980020522 +INFO:root: +[Valid][13][UTT][ MSE = 0.2405 | LCC = 0.9384 | SRCC = 0.9427 ] +INFO:root:Average loss=0.1383340058326721 +INFO:root: +[Valid][14][UTT][ MSE = 0.2751 | LCC = 0.9416 | SRCC = 0.9435 ] +INFO:root:Average loss=0.132425388276577 +INFO:root: +[Valid][15][UTT][ MSE = 0.2622 | LCC = 0.9423 | SRCC = 0.9444 ] +INFO:root:Average loss=0.12647800272703172 +INFO:root: +[Valid][16][UTT][ MSE = 0.2508 | LCC = 0.9353 | SRCC = 0.9370 ] +INFO:root:Average loss=0.12616669607162476 +INFO:root: +[Valid][17][UTT][ MSE = 0.2808 | LCC = 0.9439 | SRCC = 0.9454 ] +INFO:root:Average loss=0.12395866233110428 +INFO:root: +[Valid][18][UTT][ MSE = 0.2720 | LCC = 0.9398 | SRCC = 0.9416 ] +INFO:root:Average loss=0.12350140225887299 +INFO:root: +[Valid][19][UTT][ MSE = 0.2400 | LCC = 0.9444 | SRCC = 0.9461 ] +INFO:root:Average loss=0.11871919387578964 +INFO:root: +[Valid][20][UTT][ MSE = 0.2649 | LCC = 0.9452 | SRCC = 0.9472 ] +INFO:root:Average loss=0.11596958827972412 +INFO:root: +[Valid][21][UTT][ MSE = 0.2381 | LCC = 0.9467 | SRCC = 0.9478 ] +INFO:root:Average loss=0.11660139524936676 +INFO:root: +[Valid][22][UTT][ MSE = 0.2769 | LCC = 0.9454 | SRCC = 0.9476 ] +INFO:root:Average loss=0.11067340284585953 +INFO:root: +[Valid][23][UTT][ MSE = 0.2455 | LCC = 0.9470 | SRCC = 0.9483 ] +INFO:root:Average loss=0.11053654956817627 +INFO:root: +[Valid][24][UTT][ MSE = 0.2440 | LCC = 0.9462 | SRCC = 0.9483 ] +INFO:root:Average loss=0.10660282683372498 +INFO:root: +[Valid][25][UTT][ MSE = 0.2384 | LCC = 0.9448 | SRCC = 0.9460 ] +INFO:root:Average loss=0.10610139161348343 +INFO:root: +[Valid][26][UTT][ MSE = 0.2305 | LCC = 0.9470 | SRCC = 0.9486 ] +INFO:root:Average loss=0.1041984159052372 +INFO:root: +[Valid][27][UTT][ MSE = 0.2799 | LCC = 0.9428 | SRCC = 0.9438 ] +INFO:root:Average loss=0.10062821382284165 +INFO:root: +[Valid][28][UTT][ MSE = 0.2573 | LCC = 0.9415 | SRCC = 0.9428 ] +INFO:root:Average loss=0.10087851247191429 +INFO:root: +[Valid][29][UTT][ MSE = 0.2608 | LCC = 0.9444 | SRCC = 0.9461 ] +INFO:root:Average loss=0.1014231983423233 +INFO:root: +[Valid][30][UTT][ MSE = 0.2577 | LCC = 0.9463 | SRCC = 0.9479 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2447 | LCC = 0.9508 | SRCC = 0.9518 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e35cf004f99447df134901f21fad1eccc964bb23 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c9967be5a8de9b46c0ef7bcc56d9e61ef1e2183db0851271171a26ad2ff53f2 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..bce88326bcf4b1c57cb70b45007923ade0ffde77 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:27d8e829eeaa5d92f10c1830d7c3a46d46db7f5bee5b25adb4c526c07e37261d +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..af5b52fafd8fa2afe9088a4366e9291a5a09880c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4117129452228546 +INFO:root: +[Valid][0][UTT][ MSE = 0.2727 | LCC = 0.9156 | SRCC = 0.9180 ] +INFO:root:Average loss=0.20048323178291322 +INFO:root: +[Valid][1][UTT][ MSE = 0.2497 | LCC = 0.9356 | SRCC = 0.9364 ] +INFO:root:Average loss=0.1574930672645569 +INFO:root: +[Valid][2][UTT][ MSE = 0.2195 | LCC = 0.9461 | SRCC = 0.9467 ] +INFO:root:Average loss=0.1356112723350525 +INFO:root: +[Valid][3][UTT][ MSE = 0.2109 | LCC = 0.9483 | SRCC = 0.9488 ] +INFO:root:Average loss=0.11931850332021714 +INFO:root: +[Valid][4][UTT][ MSE = 0.1995 | LCC = 0.9514 | SRCC = 0.9515 ] +INFO:root:Average loss=0.10683352142572403 +INFO:root: +[Valid][5][UTT][ MSE = 0.1823 | LCC = 0.9518 | SRCC = 0.9526 ] +INFO:root:Average loss=0.10276756420731545 +INFO:root: +[Valid][6][UTT][ MSE = 0.1825 | LCC = 0.9537 | SRCC = 0.9542 ] +INFO:root:Average loss=0.09474822154641152 +INFO:root: +[Valid][7][UTT][ MSE = 0.1874 | LCC = 0.9551 | SRCC = 0.9547 ] +INFO:root:Average loss=0.08594685822725295 +INFO:root: +[Valid][8][UTT][ MSE = 0.2034 | LCC = 0.9530 | SRCC = 0.9547 ] +INFO:root:Average loss=0.08285098138451576 +INFO:root: +[Valid][9][UTT][ MSE = 0.1820 | LCC = 0.9559 | SRCC = 0.9561 ] +INFO:root:Average loss=0.0788572643995285 +INFO:root: +[Valid][10][UTT][ MSE = 0.1731 | LCC = 0.9565 | SRCC = 0.9574 ] +INFO:root:Average loss=0.07493534168601036 +INFO:root: +[Valid][11][UTT][ MSE = 0.1816 | LCC = 0.9572 | SRCC = 0.9570 ] +INFO:root:Average loss=0.07432757851481438 +INFO:root: +[Valid][12][UTT][ MSE = 0.1936 | LCC = 0.9564 | SRCC = 0.9559 ] +INFO:root:Average loss=0.06503676447272301 +INFO:root: +[Valid][13][UTT][ MSE = 0.1726 | LCC = 0.9573 | SRCC = 0.9581 ] +INFO:root:Average loss=0.06309953266382218 +INFO:root: +[Valid][14][UTT][ MSE = 0.1694 | LCC = 0.9589 | SRCC = 0.9592 ] +INFO:root:Average loss=0.06035502225160599 +INFO:root: +[Valid][15][UTT][ MSE = 0.1716 | LCC = 0.9585 | SRCC = 0.9589 ] +INFO:root:Average loss=0.05665142534673214 +INFO:root: +[Valid][16][UTT][ MSE = 0.1817 | LCC = 0.9590 | SRCC = 0.9592 ] +INFO:root:Average loss=0.053808479130268094 +INFO:root: +[Valid][17][UTT][ MSE = 0.1559 | LCC = 0.9587 | SRCC = 0.9589 ] +INFO:root:Average loss=0.05224324733018875 +INFO:root: +[Valid][18][UTT][ MSE = 0.1842 | LCC = 0.9573 | SRCC = 0.9567 ] +INFO:root:Average loss=0.04894383059442043 +INFO:root: +[Valid][19][UTT][ MSE = 0.2003 | LCC = 0.9569 | SRCC = 0.9567 ] +INFO:root:Average loss=0.05148208171129227 +INFO:root: +[Valid][20][UTT][ MSE = 0.1856 | LCC = 0.9585 | SRCC = 0.9584 ] +INFO:root:Average loss=0.04610644632577896 +INFO:root: +[Valid][21][UTT][ MSE = 0.1782 | LCC = 0.9575 | SRCC = 0.9569 ] +INFO:root:Average loss=0.04448753994703293 +INFO:root: +[Valid][22][UTT][ MSE = 0.1601 | LCC = 0.9584 | SRCC = 0.9582 ] +INFO:root:Average loss=0.044050680309534074 +INFO:root: +[Valid][23][UTT][ MSE = 0.1657 | LCC = 0.9595 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04020033034682274 +INFO:root: +[Valid][24][UTT][ MSE = 0.1644 | LCC = 0.9583 | SRCC = 0.9584 ] +INFO:root:Average loss=0.0409704018086195 +INFO:root: +[Valid][25][UTT][ MSE = 0.1784 | LCC = 0.9585 | SRCC = 0.9580 ] +INFO:root:Average loss=0.03806517544388771 +INFO:root: +[Valid][26][UTT][ MSE = 0.2120 | LCC = 0.9581 | SRCC = 0.9584 ] +INFO:root:Average loss=0.03654275169968605 +INFO:root: +[Valid][27][UTT][ MSE = 0.1788 | LCC = 0.9599 | SRCC = 0.9595 ] +INFO:root:Average loss=0.03631840474903583 +INFO:root: +[Valid][28][UTT][ MSE = 0.1541 | LCC = 0.9580 | SRCC = 0.9590 ] +INFO:root:Average loss=0.03675487771630287 +INFO:root: +[Valid][29][UTT][ MSE = 0.1672 | LCC = 0.9590 | SRCC = 0.9590 ] +INFO:root:Average loss=0.03393508368730545 +INFO:root: +[Valid][30][UTT][ MSE = 0.1682 | LCC = 0.9595 | SRCC = 0.9596 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1766 | LCC = 0.9610 | SRCC = 0.9615 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e454cbf4e272620eb4080496b071c7694cca2a03 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2c17021bd2cdf27e48c4d31fef4e116070de344cfab0c36dc3f83ab7dab335e3 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..27fcc2a03e5337f55e5b45fbff2d9b9463551d98 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:604471073ad9d008f687f30dd062ef0524cd074c8f0024654315dee0fbbc2063 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3942eeba30afb9a06041da55b543a153035c39c6 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.43172702729701995 +INFO:root: +[Valid][0][UTT][ MSE = 0.3166 | LCC = 0.9203 | SRCC = 0.9217 ] +INFO:root:Average loss=0.1942877213358879 +INFO:root: +[Valid][1][UTT][ MSE = 0.2420 | LCC = 0.9401 | SRCC = 0.9409 ] +INFO:root:Average loss=0.15566967177391053 +INFO:root: +[Valid][2][UTT][ MSE = 0.2822 | LCC = 0.9425 | SRCC = 0.9417 ] +INFO:root:Average loss=0.1327896137237549 +INFO:root: +[Valid][3][UTT][ MSE = 0.2216 | LCC = 0.9482 | SRCC = 0.9484 ] +INFO:root:Average loss=0.12443702560663224 +INFO:root: +[Valid][4][UTT][ MSE = 0.2112 | LCC = 0.9510 | SRCC = 0.9525 ] +INFO:root:Average loss=0.11129049372673035 +INFO:root: +[Valid][5][UTT][ MSE = 0.1877 | LCC = 0.9550 | SRCC = 0.9558 ] +INFO:root:Average loss=0.09962961921095848 +INFO:root: +[Valid][6][UTT][ MSE = 0.2234 | LCC = 0.9548 | SRCC = 0.9568 ] +INFO:root:Average loss=0.09425131872296333 +INFO:root: +[Valid][7][UTT][ MSE = 0.2006 | LCC = 0.9567 | SRCC = 0.9578 ] +INFO:root:Average loss=0.08661315283179283 +INFO:root: +[Valid][8][UTT][ MSE = 0.1850 | LCC = 0.9572 | SRCC = 0.9577 ] +INFO:root:Average loss=0.08411098718643188 +INFO:root: +[Valid][9][UTT][ MSE = 0.2002 | LCC = 0.9577 | SRCC = 0.9588 ] +INFO:root:Average loss=0.07630190020799636 +INFO:root: +[Valid][10][UTT][ MSE = 0.2059 | LCC = 0.9559 | SRCC = 0.9577 ] +INFO:root:Average loss=0.07141833546757698 +INFO:root: +[Valid][11][UTT][ MSE = 0.1802 | LCC = 0.9581 | SRCC = 0.9586 ] +INFO:root:Average loss=0.06875888791680336 +INFO:root: +[Valid][12][UTT][ MSE = 0.2029 | LCC = 0.9580 | SRCC = 0.9584 ] +INFO:root:Average loss=0.06348221409320831 +INFO:root: +[Valid][13][UTT][ MSE = 0.2037 | LCC = 0.9568 | SRCC = 0.9572 ] +INFO:root:Average loss=0.060443800911307334 +INFO:root: +[Valid][14][UTT][ MSE = 0.2098 | LCC = 0.9587 | SRCC = 0.9596 ] +INFO:root:Average loss=0.05834884795546531 +INFO:root: +[Valid][15][UTT][ MSE = 0.2125 | LCC = 0.9583 | SRCC = 0.9595 ] +INFO:root:Average loss=0.05393544816970825 +INFO:root: +[Valid][16][UTT][ MSE = 0.2046 | LCC = 0.9582 | SRCC = 0.9587 ] +INFO:root:Average loss=0.05081468990445137 +INFO:root: +[Valid][17][UTT][ MSE = 0.1885 | LCC = 0.9592 | SRCC = 0.9593 ] +INFO:root:Average loss=0.049287012949585915 +INFO:root: +[Valid][18][UTT][ MSE = 0.1878 | LCC = 0.9601 | SRCC = 0.9603 ] +INFO:root:Average loss=0.04660970824956894 +INFO:root: +[Valid][19][UTT][ MSE = 0.1940 | LCC = 0.9585 | SRCC = 0.9590 ] +INFO:root:Average loss=0.0464377276301384 +INFO:root: +[Valid][20][UTT][ MSE = 0.1964 | LCC = 0.9588 | SRCC = 0.9593 ] +INFO:root:Average loss=0.04722593580186367 +INFO:root: +[Valid][21][UTT][ MSE = 0.1923 | LCC = 0.9592 | SRCC = 0.9599 ] +INFO:root:Average loss=0.042096694722771645 +INFO:root: +[Valid][22][UTT][ MSE = 0.2048 | LCC = 0.9598 | SRCC = 0.9603 ] +INFO:root:Average loss=0.041717765420675276 +INFO:root: +[Valid][23][UTT][ MSE = 0.1948 | LCC = 0.9604 | SRCC = 0.9609 ] +INFO:root:Average loss=0.03901925763487816 +INFO:root: +[Valid][24][UTT][ MSE = 0.2151 | LCC = 0.9597 | SRCC = 0.9606 ] +INFO:root:Average loss=0.03891501075029373 +INFO:root: +[Valid][25][UTT][ MSE = 0.1782 | LCC = 0.9603 | SRCC = 0.9605 ] +INFO:root:Average loss=0.03684741531312466 +INFO:root: +[Valid][26][UTT][ MSE = 0.2038 | LCC = 0.9604 | SRCC = 0.9609 ] +INFO:root:Average loss=0.03836204634606838 +INFO:root: +[Valid][27][UTT][ MSE = 0.1993 | LCC = 0.9605 | SRCC = 0.9607 ] +INFO:root:Average loss=0.03528076972067356 +INFO:root: +[Valid][28][UTT][ MSE = 0.1850 | LCC = 0.9597 | SRCC = 0.9597 ] +INFO:root:Average loss=0.03306329780817032 +INFO:root: +[Valid][29][UTT][ MSE = 0.1870 | LCC = 0.9603 | SRCC = 0.9605 ] +INFO:root:Average loss=0.032874192461371424 +INFO:root: +[Valid][30][UTT][ MSE = 0.1835 | LCC = 0.9605 | SRCC = 0.9609 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1983 | LCC = 0.9622 | SRCC = 0.9629 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1189c6c1c17918c1963e1772fb77661b544b2303 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2423a15058c71872fc205fc282b676b40bb1b7192509b9d9acdec46635ad9086 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..23c73d61233ed30b2f9bb0ea0ff5a67852f860b7 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:99454b17103a2714f42e710443c82541ae3a9ec2475f093156e080da50335dfc +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..495ef512cf6c14394aee37bfdf3292c90430abb5 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4474734629392624 +INFO:root: +[Valid][0][UTT][ MSE = 0.2865 | LCC = 0.9177 | SRCC = 0.9190 ] +INFO:root:Average loss=0.19678541588783263 +INFO:root: +[Valid][1][UTT][ MSE = 0.2691 | LCC = 0.9352 | SRCC = 0.9360 ] +INFO:root:Average loss=0.1570467372536659 +INFO:root: +[Valid][2][UTT][ MSE = 0.2374 | LCC = 0.9456 | SRCC = 0.9465 ] +INFO:root:Average loss=0.13418486726284026 +INFO:root: +[Valid][3][UTT][ MSE = 0.2131 | LCC = 0.9485 | SRCC = 0.9496 ] +INFO:root:Average loss=0.12360858422517776 +INFO:root: +[Valid][4][UTT][ MSE = 0.2000 | LCC = 0.9504 | SRCC = 0.9506 ] +INFO:root:Average loss=0.11165792700648308 +INFO:root: +[Valid][5][UTT][ MSE = 0.2051 | LCC = 0.9495 | SRCC = 0.9515 ] +INFO:root:Average loss=0.10332634258270264 +INFO:root: +[Valid][6][UTT][ MSE = 0.2261 | LCC = 0.9548 | SRCC = 0.9555 ] +INFO:root:Average loss=0.09489333108067513 +INFO:root: +[Valid][7][UTT][ MSE = 0.1881 | LCC = 0.9554 | SRCC = 0.9558 ] +INFO:root:Average loss=0.08730387529730797 +INFO:root: +[Valid][8][UTT][ MSE = 0.2039 | LCC = 0.9567 | SRCC = 0.9570 ] +INFO:root:Average loss=0.08931015515327453 +INFO:root: +[Valid][9][UTT][ MSE = 0.2216 | LCC = 0.9567 | SRCC = 0.9577 ] +INFO:root:Average loss=0.08055379527807235 +INFO:root: +[Valid][10][UTT][ MSE = 0.1860 | LCC = 0.9571 | SRCC = 0.9572 ] +INFO:root:Average loss=0.07643610182404519 +INFO:root: +[Valid][11][UTT][ MSE = 0.1937 | LCC = 0.9578 | SRCC = 0.9579 ] +INFO:root:Average loss=0.07192817172408104 +INFO:root: +[Valid][12][UTT][ MSE = 0.1833 | LCC = 0.9574 | SRCC = 0.9584 ] +INFO:root:Average loss=0.06547468695044517 +INFO:root: +[Valid][13][UTT][ MSE = 0.1996 | LCC = 0.9537 | SRCC = 0.9574 ] +INFO:root:Average loss=0.06426449990272522 +INFO:root: +[Valid][14][UTT][ MSE = 0.1744 | LCC = 0.9591 | SRCC = 0.9589 ] +INFO:root:Average loss=0.05994209235906601 +INFO:root: +[Valid][15][UTT][ MSE = 0.1975 | LCC = 0.9572 | SRCC = 0.9573 ] +INFO:root:Average loss=0.05763139125704765 +INFO:root: +[Valid][16][UTT][ MSE = 0.1794 | LCC = 0.9593 | SRCC = 0.9594 ] +INFO:root:Average loss=0.05506098560988903 +INFO:root: +[Valid][17][UTT][ MSE = 0.1885 | LCC = 0.9588 | SRCC = 0.9588 ] +INFO:root:Average loss=0.05283013814687729 +INFO:root: +[Valid][18][UTT][ MSE = 0.1735 | LCC = 0.9591 | SRCC = 0.9590 ] +INFO:root:Average loss=0.05121131193637848 +INFO:root: +[Valid][19][UTT][ MSE = 0.1852 | LCC = 0.9572 | SRCC = 0.9571 ] +INFO:root:Average loss=0.04759121043980122 +INFO:root: +[Valid][20][UTT][ MSE = 0.1783 | LCC = 0.9585 | SRCC = 0.9588 ] +INFO:root:Average loss=0.04781092554330826 +INFO:root: +[Valid][21][UTT][ MSE = 0.1997 | LCC = 0.9597 | SRCC = 0.9600 ] +INFO:root:Average loss=0.046990641757845876 +INFO:root: +[Valid][22][UTT][ MSE = 0.1857 | LCC = 0.9586 | SRCC = 0.9586 ] +INFO:root:Average loss=0.04555111196637154 +INFO:root: +[Valid][23][UTT][ MSE = 0.2142 | LCC = 0.9577 | SRCC = 0.9589 ] +INFO:root:Average loss=0.043549300283193586 +INFO:root: +[Valid][24][UTT][ MSE = 0.1731 | LCC = 0.9598 | SRCC = 0.9601 ] +INFO:root:Average loss=0.04106211857497692 +INFO:root: +[Valid][25][UTT][ MSE = 0.1899 | LCC = 0.9574 | SRCC = 0.9569 ] +INFO:root:Average loss=0.03694812278449535 +INFO:root: +[Valid][26][UTT][ MSE = 0.1852 | LCC = 0.9598 | SRCC = 0.9597 ] +INFO:root:Average loss=0.03802586558461189 +INFO:root: +[Valid][27][UTT][ MSE = 0.1909 | LCC = 0.9577 | SRCC = 0.9577 ] +INFO:root:Average loss=0.03834469810128212 +INFO:root: +[Valid][28][UTT][ MSE = 0.1903 | LCC = 0.9599 | SRCC = 0.9598 ] +INFO:root:Average loss=0.034009083300828934 +INFO:root: +[Valid][29][UTT][ MSE = 0.2035 | LCC = 0.9578 | SRCC = 0.9573 ] +INFO:root:Average loss=0.033868610128760336 +INFO:root: +[Valid][30][UTT][ MSE = 0.1713 | LCC = 0.9597 | SRCC = 0.9595 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1870 | LCC = 0.9616 | SRCC = 0.9625 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..841ae0ddb7d868557ecbb456ec9162335f9d68e3 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ec35ca397089300b261e0a3786e3eb686911fbb4b8b8fd95e273afaad019aee5 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..33b693688df3bae5a17608f70e74e36cbb6135b2 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:086977d079182d4d537889ca59803b42ba1c9f855c968e8549d8211faf55aefa +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..aded5b6beba1a8b0a87674448859177e6d49c7bb --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.47199773240089415 +INFO:root: +[Valid][0][UTT][ MSE = 0.3127 | LCC = 0.9106 | SRCC = 0.9148 ] +INFO:root:Average loss=0.20348889362812042 +INFO:root: +[Valid][1][UTT][ MSE = 0.2713 | LCC = 0.9384 | SRCC = 0.9392 ] +INFO:root:Average loss=0.16628122609853743 +INFO:root: +[Valid][2][UTT][ MSE = 0.2707 | LCC = 0.9429 | SRCC = 0.9447 ] +INFO:root:Average loss=0.13521206659078597 +INFO:root: +[Valid][3][UTT][ MSE = 0.2306 | LCC = 0.9469 | SRCC = 0.9478 ] +INFO:root:Average loss=0.12485281068086625 +INFO:root: +[Valid][4][UTT][ MSE = 0.2287 | LCC = 0.9515 | SRCC = 0.9524 ] +INFO:root:Average loss=0.11834340578317643 +INFO:root: +[Valid][5][UTT][ MSE = 0.2237 | LCC = 0.9516 | SRCC = 0.9521 ] +INFO:root:Average loss=0.10575118365883827 +INFO:root: +[Valid][6][UTT][ MSE = 0.2363 | LCC = 0.9525 | SRCC = 0.9528 ] +INFO:root:Average loss=0.09624920606613159 +INFO:root: +[Valid][7][UTT][ MSE = 0.2214 | LCC = 0.9527 | SRCC = 0.9535 ] +INFO:root:Average loss=0.08618006911873817 +INFO:root: +[Valid][8][UTT][ MSE = 0.2118 | LCC = 0.9557 | SRCC = 0.9560 ] +INFO:root:Average loss=0.08697067239880561 +INFO:root: +[Valid][9][UTT][ MSE = 0.2340 | LCC = 0.9553 | SRCC = 0.9555 ] +INFO:root:Average loss=0.07707777705788613 +INFO:root: +[Valid][10][UTT][ MSE = 0.1967 | LCC = 0.9544 | SRCC = 0.9561 ] +INFO:root:Average loss=0.07279629427194595 +INFO:root: +[Valid][11][UTT][ MSE = 0.1903 | LCC = 0.9551 | SRCC = 0.9552 ] +INFO:root:Average loss=0.06860028979182244 +INFO:root: +[Valid][12][UTT][ MSE = 0.1852 | LCC = 0.9535 | SRCC = 0.9558 ] +INFO:root:Average loss=0.06937534430623055 +INFO:root: +[Valid][13][UTT][ MSE = 0.2053 | LCC = 0.9590 | SRCC = 0.9594 ] +INFO:root:Average loss=0.06116833928227425 +INFO:root: +[Valid][14][UTT][ MSE = 0.2078 | LCC = 0.9580 | SRCC = 0.9583 ] +INFO:root:Average loss=0.059419028654694554 +INFO:root: +[Valid][15][UTT][ MSE = 0.1731 | LCC = 0.9566 | SRCC = 0.9575 ] +INFO:root:Average loss=0.05989677712321281 +INFO:root: +[Valid][16][UTT][ MSE = 0.1845 | LCC = 0.9575 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05299357303977013 +INFO:root: +[Valid][17][UTT][ MSE = 0.1848 | LCC = 0.9583 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05180080917477608 +INFO:root: +[Valid][18][UTT][ MSE = 0.1858 | LCC = 0.9591 | SRCC = 0.9594 ] +INFO:root:Average loss=0.050198452323675154 +INFO:root: +[Valid][19][UTT][ MSE = 0.1775 | LCC = 0.9587 | SRCC = 0.9587 ] +INFO:root:Average loss=0.046691305696964266 +INFO:root: +[Valid][20][UTT][ MSE = 0.2017 | LCC = 0.9580 | SRCC = 0.9584 ] +INFO:root:Average loss=0.04675274163484573 +INFO:root: +[Valid][21][UTT][ MSE = 0.1975 | LCC = 0.9596 | SRCC = 0.9593 ] +INFO:root:Average loss=0.045056015476584436 +INFO:root: +[Valid][22][UTT][ MSE = 0.1789 | LCC = 0.9575 | SRCC = 0.9587 ] +INFO:root:Average loss=0.04270039498806 +INFO:root: +[Valid][23][UTT][ MSE = 0.1641 | LCC = 0.9570 | SRCC = 0.9582 ] +INFO:root:Average loss=0.040477574333548545 +INFO:root: +[Valid][24][UTT][ MSE = 0.1910 | LCC = 0.9598 | SRCC = 0.9598 ] +INFO:root:Average loss=0.03818038235604763 +INFO:root: +[Valid][25][UTT][ MSE = 0.1961 | LCC = 0.9592 | SRCC = 0.9592 ] +INFO:root:Average loss=0.04083500951528549 +INFO:root: +[Valid][26][UTT][ MSE = 0.1956 | LCC = 0.9598 | SRCC = 0.9604 ] +INFO:root:Average loss=0.03840601409971714 +INFO:root: +[Valid][27][UTT][ MSE = 0.1791 | LCC = 0.9597 | SRCC = 0.9598 ] +INFO:root:Average loss=0.036476661443710326 +INFO:root: +[Valid][28][UTT][ MSE = 0.1836 | LCC = 0.9597 | SRCC = 0.9598 ] +INFO:root:Average loss=0.03406227308511734 +INFO:root: +[Valid][29][UTT][ MSE = 0.1733 | LCC = 0.9587 | SRCC = 0.9594 ] +INFO:root:Average loss=0.035088737100362775 +INFO:root: +[Valid][30][UTT][ MSE = 0.1754 | LCC = 0.9591 | SRCC = 0.9595 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1972 | LCC = 0.9617 | SRCC = 0.9627 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1de5dd2bfd3e7e6b12388d96a18df7c59bbcbb1c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1f75ed54c926a5098038862f838de7c0b45c35ee65522adc74669f7447e85c4d +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2f26e75a484ef793d2191b3a0486a1c43819359a --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9c16fab2cc253849c5f61119845af1836ba486a95baddccaad1f737c8fcf742c +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..cf93cee674c7d51b0cc85e34f81d360b4e695232 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer2/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4472167545557022 +INFO:root: +[Valid][0][UTT][ MSE = 0.2742 | LCC = 0.9148 | SRCC = 0.9180 ] +INFO:root:Average loss=0.19092176979780198 +INFO:root: +[Valid][1][UTT][ MSE = 0.2171 | LCC = 0.9372 | SRCC = 0.9387 ] +INFO:root:Average loss=0.1570450792312622 +INFO:root: +[Valid][2][UTT][ MSE = 0.2154 | LCC = 0.9438 | SRCC = 0.9450 ] +INFO:root:Average loss=0.13333258008956908 +INFO:root: +[Valid][3][UTT][ MSE = 0.2389 | LCC = 0.9467 | SRCC = 0.9476 ] +INFO:root:Average loss=0.12132747161388398 +INFO:root: +[Valid][4][UTT][ MSE = 0.2210 | LCC = 0.9511 | SRCC = 0.9521 ] +INFO:root:Average loss=0.11029983171820641 +INFO:root: +[Valid][5][UTT][ MSE = 0.2037 | LCC = 0.9518 | SRCC = 0.9528 ] +INFO:root:Average loss=0.10589992791414261 +INFO:root: +[Valid][6][UTT][ MSE = 0.2161 | LCC = 0.9546 | SRCC = 0.9550 ] +INFO:root:Average loss=0.09244923642277718 +INFO:root: +[Valid][7][UTT][ MSE = 0.1882 | LCC = 0.9554 | SRCC = 0.9558 ] +INFO:root:Average loss=0.09113950052857399 +INFO:root: +[Valid][8][UTT][ MSE = 0.2161 | LCC = 0.9554 | SRCC = 0.9567 ] +INFO:root:Average loss=0.08092880076169967 +INFO:root: +[Valid][9][UTT][ MSE = 0.2445 | LCC = 0.9573 | SRCC = 0.9584 ] +INFO:root:Average loss=0.07713496890664101 +INFO:root: +[Valid][10][UTT][ MSE = 0.1810 | LCC = 0.9528 | SRCC = 0.9550 ] +INFO:root:Average loss=0.07327425694465638 +INFO:root: +[Valid][11][UTT][ MSE = 0.1830 | LCC = 0.9560 | SRCC = 0.9567 ] +INFO:root:Average loss=0.07031730863451957 +INFO:root: +[Valid][12][UTT][ MSE = 0.1858 | LCC = 0.9579 | SRCC = 0.9582 ] +INFO:root:Average loss=0.0673485418856144 +INFO:root: +[Valid][13][UTT][ MSE = 0.1701 | LCC = 0.9568 | SRCC = 0.9571 ] +INFO:root:Average loss=0.06132769629359245 +INFO:root: +[Valid][14][UTT][ MSE = 0.1932 | LCC = 0.9541 | SRCC = 0.9541 ] +INFO:root:Average loss=0.060218247920274734 +INFO:root: +[Valid][15][UTT][ MSE = 0.1995 | LCC = 0.9571 | SRCC = 0.9574 ] +INFO:root:Average loss=0.05626587273180485 +INFO:root: +[Valid][16][UTT][ MSE = 0.2182 | LCC = 0.9577 | SRCC = 0.9580 ] +INFO:root:Average loss=0.05276948484778404 +INFO:root: +[Valid][17][UTT][ MSE = 0.1743 | LCC = 0.9570 | SRCC = 0.9579 ] +INFO:root:Average loss=0.053474965065717694 +INFO:root: +[Valid][18][UTT][ MSE = 0.1635 | LCC = 0.9592 | SRCC = 0.9593 ] +INFO:root:Average loss=0.04727851079404354 +INFO:root: +[Valid][19][UTT][ MSE = 0.2029 | LCC = 0.9584 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04793760153651237 +INFO:root: +[Valid][20][UTT][ MSE = 0.2101 | LCC = 0.9591 | SRCC = 0.9593 ] +INFO:root:Average loss=0.04593830120563507 +INFO:root: +[Valid][21][UTT][ MSE = 0.1938 | LCC = 0.9574 | SRCC = 0.9576 ] +INFO:root:Average loss=0.04455663627386093 +INFO:root: +[Valid][22][UTT][ MSE = 0.2237 | LCC = 0.9572 | SRCC = 0.9573 ] +INFO:root:Average loss=0.042885970026254654 +INFO:root: +[Valid][23][UTT][ MSE = 0.1981 | LCC = 0.9585 | SRCC = 0.9589 ] +INFO:root:Average loss=0.041641562834382057 +INFO:root: +[Valid][24][UTT][ MSE = 0.1786 | LCC = 0.9606 | SRCC = 0.9608 ] +INFO:root:Average loss=0.03898173813521862 +INFO:root: +[Valid][25][UTT][ MSE = 0.1959 | LCC = 0.9591 | SRCC = 0.9594 ] +INFO:root:Average loss=0.03717898596823215 +INFO:root: +[Valid][26][UTT][ MSE = 0.1852 | LCC = 0.9597 | SRCC = 0.9599 ] +INFO:root:Average loss=0.03806995168328285 +INFO:root: +[Valid][27][UTT][ MSE = 0.1927 | LCC = 0.9596 | SRCC = 0.9598 ] +INFO:root:Average loss=0.03589150670170784 +INFO:root: +[Valid][28][UTT][ MSE = 0.1759 | LCC = 0.9599 | SRCC = 0.9600 ] +INFO:root:Average loss=0.035229853630065915 +INFO:root: +[Valid][29][UTT][ MSE = 0.1667 | LCC = 0.9588 | SRCC = 0.9591 ] +INFO:root:Average loss=0.03473836690187454 +INFO:root: +[Valid][30][UTT][ MSE = 0.1834 | LCC = 0.9595 | SRCC = 0.9593 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1762 | LCC = 0.9614 | SRCC = 0.9618 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..28062b84654178e0617547974933cf48c433e9ac --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19209962250c7b8f49c3007e772cdd3967dfbb09438eaacff2c39feb1e8df938 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..bbd859a4a9b3ec57d35159b76b739a18b1127487 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f1d5abb494b67ee02258a64e5583bb16b3d1735b9ae07c57b77eb16c7e0edeba +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3d04b70d105c5ed44941048c39b5b0845b473aaa --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3930329396724701 +INFO:root: +[Valid][0][UTT][ MSE = 0.2769 | LCC = 0.9187 | SRCC = 0.9198 ] +INFO:root:Average loss=0.1899458568096161 +INFO:root: +[Valid][1][UTT][ MSE = 0.2806 | LCC = 0.9321 | SRCC = 0.9317 ] +INFO:root:Average loss=0.15509059500694275 +INFO:root: +[Valid][2][UTT][ MSE = 0.2149 | LCC = 0.9443 | SRCC = 0.9458 ] +INFO:root:Average loss=0.13407164758443832 +INFO:root: +[Valid][3][UTT][ MSE = 0.2383 | LCC = 0.9506 | SRCC = 0.9519 ] +INFO:root:Average loss=0.11553000915050507 +INFO:root: +[Valid][4][UTT][ MSE = 0.2127 | LCC = 0.9529 | SRCC = 0.9537 ] +INFO:root:Average loss=0.10759120380878448 +INFO:root: +[Valid][5][UTT][ MSE = 0.2660 | LCC = 0.9521 | SRCC = 0.9538 ] +INFO:root:Average loss=0.10283369356393814 +INFO:root: +[Valid][6][UTT][ MSE = 0.2452 | LCC = 0.9547 | SRCC = 0.9552 ] +INFO:root:Average loss=0.09053250113129616 +INFO:root: +[Valid][7][UTT][ MSE = 0.2134 | LCC = 0.9549 | SRCC = 0.9556 ] +INFO:root:Average loss=0.08831464219093323 +INFO:root: +[Valid][8][UTT][ MSE = 0.2252 | LCC = 0.9548 | SRCC = 0.9563 ] +INFO:root:Average loss=0.08126643329858779 +INFO:root: +[Valid][9][UTT][ MSE = 0.2049 | LCC = 0.9561 | SRCC = 0.9563 ] +INFO:root:Average loss=0.0768296613395214 +INFO:root: +[Valid][10][UTT][ MSE = 0.2044 | LCC = 0.9560 | SRCC = 0.9554 ] +INFO:root:Average loss=0.07114951148629188 +INFO:root: +[Valid][11][UTT][ MSE = 0.2215 | LCC = 0.9552 | SRCC = 0.9552 ] +INFO:root:Average loss=0.06846161243319511 +INFO:root: +[Valid][12][UTT][ MSE = 0.2118 | LCC = 0.9557 | SRCC = 0.9561 ] +INFO:root:Average loss=0.0646586728990078 +INFO:root: +[Valid][13][UTT][ MSE = 0.1989 | LCC = 0.9553 | SRCC = 0.9555 ] +INFO:root:Average loss=0.06153799882531166 +INFO:root: +[Valid][14][UTT][ MSE = 0.2053 | LCC = 0.9562 | SRCC = 0.9561 ] +INFO:root:Average loss=0.05867836157977581 +INFO:root: +[Valid][15][UTT][ MSE = 0.2233 | LCC = 0.9563 | SRCC = 0.9564 ] +INFO:root:Average loss=0.05578739856183529 +INFO:root: +[Valid][16][UTT][ MSE = 0.2096 | LCC = 0.9560 | SRCC = 0.9559 ] +INFO:root:Average loss=0.05536657392978668 +INFO:root: +[Valid][17][UTT][ MSE = 0.2139 | LCC = 0.9552 | SRCC = 0.9546 ] +INFO:root:Average loss=0.05099894224107265 +INFO:root: +[Valid][18][UTT][ MSE = 0.2253 | LCC = 0.9572 | SRCC = 0.9575 ] +INFO:root:Average loss=0.052090692579746245 +INFO:root: +[Valid][19][UTT][ MSE = 0.2107 | LCC = 0.9553 | SRCC = 0.9552 ] +INFO:root:Average loss=0.04815145289897919 +INFO:root: +[Valid][20][UTT][ MSE = 0.2043 | LCC = 0.9570 | SRCC = 0.9566 ] +INFO:root:Average loss=0.04427082623541355 +INFO:root: +[Valid][21][UTT][ MSE = 0.1962 | LCC = 0.9563 | SRCC = 0.9560 ] +INFO:root:Average loss=0.04648575548827648 +INFO:root: +[Valid][22][UTT][ MSE = 0.2031 | LCC = 0.9573 | SRCC = 0.9574 ] +INFO:root:Average loss=0.04279519799351692 +INFO:root: +[Valid][23][UTT][ MSE = 0.1953 | LCC = 0.9574 | SRCC = 0.9578 ] +INFO:root:Average loss=0.04174157966673374 +INFO:root: +[Valid][24][UTT][ MSE = 0.1991 | LCC = 0.9563 | SRCC = 0.9562 ] +INFO:root:Average loss=0.03774881088733673 +INFO:root: +[Valid][25][UTT][ MSE = 0.2053 | LCC = 0.9559 | SRCC = 0.9556 ] +INFO:root:Average loss=0.03799510328471661 +INFO:root: +[Valid][26][UTT][ MSE = 0.2278 | LCC = 0.9565 | SRCC = 0.9567 ] +INFO:root:Average loss=0.03595484346151352 +INFO:root: +[Valid][27][UTT][ MSE = 0.1992 | LCC = 0.9542 | SRCC = 0.9537 ] +INFO:root:Average loss=0.03439992955327034 +INFO:root: +[Valid][28][UTT][ MSE = 0.2037 | LCC = 0.9559 | SRCC = 0.9560 ] +INFO:root:Average loss=0.03713197238743305 +INFO:root: +[Valid][29][UTT][ MSE = 0.2024 | LCC = 0.9554 | SRCC = 0.9550 ] +INFO:root:Average loss=0.03402327986061573 +INFO:root: +[Valid][30][UTT][ MSE = 0.2048 | LCC = 0.9565 | SRCC = 0.9565 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2006 | LCC = 0.9603 | SRCC = 0.9613 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f559c361b3531e3a836fbf42f298b25902c68338 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aa2992a996f968eae37380f8269a47397ba63d3325f24a032d965a12dc8b3954 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f87cc12f259ea47f33e1b39c07e42e560d7a8767 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0294dbe448e8435660d26a9a0ecfd071ebdda11a94fd0c277e45eebf981e7007 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..dbe6cc93c7eeab219694ccb5ddecabbd323700fd --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.402008615732193 +INFO:root: +[Valid][0][UTT][ MSE = 0.2968 | LCC = 0.9288 | SRCC = 0.9288 ] +INFO:root:Average loss=0.17971222549676896 +INFO:root: +[Valid][1][UTT][ MSE = 0.2901 | LCC = 0.9385 | SRCC = 0.9417 ] +INFO:root:Average loss=0.14927011543512345 +INFO:root: +[Valid][2][UTT][ MSE = 0.2405 | LCC = 0.9461 | SRCC = 0.9464 ] +INFO:root:Average loss=0.1261805494427681 +INFO:root: +[Valid][3][UTT][ MSE = 0.2344 | LCC = 0.9507 | SRCC = 0.9517 ] +INFO:root:Average loss=0.1120335454940796 +INFO:root: +[Valid][4][UTT][ MSE = 0.2521 | LCC = 0.9525 | SRCC = 0.9532 ] +INFO:root:Average loss=0.1060435101389885 +INFO:root: +[Valid][5][UTT][ MSE = 0.2564 | LCC = 0.9498 | SRCC = 0.9505 ] +INFO:root:Average loss=0.0976867446899414 +INFO:root: +[Valid][6][UTT][ MSE = 0.2230 | LCC = 0.9545 | SRCC = 0.9554 ] +INFO:root:Average loss=0.08769423824548721 +INFO:root: +[Valid][7][UTT][ MSE = 0.2093 | LCC = 0.9553 | SRCC = 0.9561 ] +INFO:root:Average loss=0.0842697349190712 +INFO:root: +[Valid][8][UTT][ MSE = 0.2028 | LCC = 0.9550 | SRCC = 0.9557 ] +INFO:root:Average loss=0.08011507079005241 +INFO:root: +[Valid][9][UTT][ MSE = 0.2243 | LCC = 0.9554 | SRCC = 0.9559 ] +INFO:root:Average loss=0.07279838252067566 +INFO:root: +[Valid][10][UTT][ MSE = 0.2323 | LCC = 0.9504 | SRCC = 0.9515 ] +INFO:root:Average loss=0.06900475028157234 +INFO:root: +[Valid][11][UTT][ MSE = 0.2160 | LCC = 0.9571 | SRCC = 0.9573 ] +INFO:root:Average loss=0.06629369595646858 +INFO:root: +[Valid][12][UTT][ MSE = 0.2031 | LCC = 0.9568 | SRCC = 0.9569 ] +INFO:root:Average loss=0.06431689524650573 +INFO:root: +[Valid][13][UTT][ MSE = 0.2235 | LCC = 0.9536 | SRCC = 0.9537 ] +INFO:root:Average loss=0.0593640204668045 +INFO:root: +[Valid][14][UTT][ MSE = 0.2528 | LCC = 0.9547 | SRCC = 0.9562 ] +INFO:root:Average loss=0.05636858773231506 +INFO:root: +[Valid][15][UTT][ MSE = 0.2495 | LCC = 0.9562 | SRCC = 0.9572 ] +INFO:root:Average loss=0.053998994141817094 +INFO:root: +[Valid][16][UTT][ MSE = 0.2387 | LCC = 0.9554 | SRCC = 0.9570 ] +INFO:root:Average loss=0.05264511679112911 +INFO:root: +[Valid][17][UTT][ MSE = 0.2681 | LCC = 0.9547 | SRCC = 0.9567 ] +INFO:root:Average loss=0.050765877395868304 +INFO:root: +[Valid][18][UTT][ MSE = 0.2460 | LCC = 0.9562 | SRCC = 0.9575 ] +INFO:root:Average loss=0.0469324634373188 +INFO:root: +[Valid][19][UTT][ MSE = 0.1927 | LCC = 0.9570 | SRCC = 0.9572 ] +INFO:root:Average loss=0.04540518367290497 +INFO:root: +[Valid][20][UTT][ MSE = 0.2219 | LCC = 0.9557 | SRCC = 0.9556 ] +INFO:root:Average loss=0.04435427774488926 +INFO:root: +[Valid][21][UTT][ MSE = 0.2156 | LCC = 0.9583 | SRCC = 0.9584 ] +INFO:root:Average loss=0.04095488578081131 +INFO:root: +[Valid][22][UTT][ MSE = 0.2094 | LCC = 0.9584 | SRCC = 0.9585 ] +INFO:root:Average loss=0.04052720901370049 +INFO:root: +[Valid][23][UTT][ MSE = 0.2035 | LCC = 0.9568 | SRCC = 0.9569 ] +INFO:root:Average loss=0.039856266841292384 +INFO:root: +[Valid][24][UTT][ MSE = 0.2289 | LCC = 0.9568 | SRCC = 0.9582 ] +INFO:root:Average loss=0.038795601889491084 +INFO:root: +[Valid][25][UTT][ MSE = 0.2203 | LCC = 0.9575 | SRCC = 0.9579 ] +INFO:root:Average loss=0.03698466604948044 +INFO:root: +[Valid][26][UTT][ MSE = 0.2123 | LCC = 0.9561 | SRCC = 0.9562 ] +INFO:root:Average loss=0.034691643252968786 +INFO:root: +[Valid][27][UTT][ MSE = 0.2088 | LCC = 0.9564 | SRCC = 0.9565 ] +INFO:root:Average loss=0.03457070502638817 +INFO:root: +[Valid][28][UTT][ MSE = 0.2186 | LCC = 0.9587 | SRCC = 0.9589 ] +INFO:root:Average loss=0.032460535511374476 +INFO:root: +[Valid][29][UTT][ MSE = 0.2279 | LCC = 0.9579 | SRCC = 0.9584 ] +INFO:root:Average loss=0.03144793945550919 +INFO:root: +[Valid][30][UTT][ MSE = 0.2146 | LCC = 0.9561 | SRCC = 0.9566 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2316 | LCC = 0.9582 | SRCC = 0.9594 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..21e35f0c7bb91cf7de2618d86f7c9d2852a5168f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:49c389052c03bf89e3bf7759fdcafdc805ec842beb24868c0366abc7a47a0b2a +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b30e03b7bf12cb96ca8fdd164d81ae3adee5a936 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:379df8119368574d75392f7651184c5c9522fdd576509d797a3dcd95b1d5d013 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..cf2b917f6f6d55829f5391832b905650a292708b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4177987816333771 +INFO:root: +[Valid][0][UTT][ MSE = 0.2962 | LCC = 0.9134 | SRCC = 0.9170 ] +INFO:root:Average loss=0.18974414312839508 +INFO:root: +[Valid][1][UTT][ MSE = 0.2442 | LCC = 0.9360 | SRCC = 0.9407 ] +INFO:root:Average loss=0.15717399376630783 +INFO:root: +[Valid][2][UTT][ MSE = 0.2476 | LCC = 0.9473 | SRCC = 0.9487 ] +INFO:root:Average loss=0.13231262677907943 +INFO:root: +[Valid][3][UTT][ MSE = 0.2144 | LCC = 0.9489 | SRCC = 0.9502 ] +INFO:root:Average loss=0.1184765099287033 +INFO:root: +[Valid][4][UTT][ MSE = 0.2274 | LCC = 0.9532 | SRCC = 0.9544 ] +INFO:root:Average loss=0.11513209092617036 +INFO:root: +[Valid][5][UTT][ MSE = 0.2075 | LCC = 0.9542 | SRCC = 0.9558 ] +INFO:root:Average loss=0.09689042577147484 +INFO:root: +[Valid][6][UTT][ MSE = 0.2203 | LCC = 0.9545 | SRCC = 0.9550 ] +INFO:root:Average loss=0.09302342569828033 +INFO:root: +[Valid][7][UTT][ MSE = 0.1978 | LCC = 0.9557 | SRCC = 0.9561 ] +INFO:root:Average loss=0.08606567642092705 +INFO:root: +[Valid][8][UTT][ MSE = 0.2047 | LCC = 0.9551 | SRCC = 0.9561 ] +INFO:root:Average loss=0.08414586344361305 +INFO:root: +[Valid][9][UTT][ MSE = 0.2042 | LCC = 0.9581 | SRCC = 0.9581 ] +INFO:root:Average loss=0.07680516266822815 +INFO:root: +[Valid][10][UTT][ MSE = 0.2278 | LCC = 0.9561 | SRCC = 0.9581 ] +INFO:root:Average loss=0.0741670727133751 +INFO:root: +[Valid][11][UTT][ MSE = 0.2060 | LCC = 0.9572 | SRCC = 0.9575 ] +INFO:root:Average loss=0.06803818362951279 +INFO:root: +[Valid][12][UTT][ MSE = 0.2052 | LCC = 0.9585 | SRCC = 0.9593 ] +INFO:root:Average loss=0.06388348871469497 +INFO:root: +[Valid][13][UTT][ MSE = 0.1881 | LCC = 0.9559 | SRCC = 0.9570 ] +INFO:root:Average loss=0.059675974041223524 +INFO:root: +[Valid][14][UTT][ MSE = 0.2095 | LCC = 0.9570 | SRCC = 0.9583 ] +INFO:root:Average loss=0.05864660441875458 +INFO:root: +[Valid][15][UTT][ MSE = 0.1834 | LCC = 0.9584 | SRCC = 0.9589 ] +INFO:root:Average loss=0.052858634546399115 +INFO:root: +[Valid][16][UTT][ MSE = 0.1872 | LCC = 0.9567 | SRCC = 0.9572 ] +INFO:root:Average loss=0.052335058972239494 +INFO:root: +[Valid][17][UTT][ MSE = 0.1910 | LCC = 0.9579 | SRCC = 0.9578 ] +INFO:root:Average loss=0.049616965889930725 +INFO:root: +[Valid][18][UTT][ MSE = 0.1974 | LCC = 0.9568 | SRCC = 0.9571 ] +INFO:root:Average loss=0.047705186501145365 +INFO:root: +[Valid][19][UTT][ MSE = 0.1705 | LCC = 0.9583 | SRCC = 0.9587 ] +INFO:root:Average loss=0.044880654469132426 +INFO:root: +[Valid][20][UTT][ MSE = 0.1948 | LCC = 0.9579 | SRCC = 0.9583 ] +INFO:root:Average loss=0.04431298080086708 +INFO:root: +[Valid][21][UTT][ MSE = 0.1694 | LCC = 0.9588 | SRCC = 0.9589 ] +INFO:root:Average loss=0.041061096400022506 +INFO:root: +[Valid][22][UTT][ MSE = 0.1831 | LCC = 0.9585 | SRCC = 0.9585 ] +INFO:root:Average loss=0.03912197281420231 +INFO:root: +[Valid][23][UTT][ MSE = 0.1934 | LCC = 0.9584 | SRCC = 0.9585 ] +INFO:root:Average loss=0.03859921738505363 +INFO:root: +[Valid][24][UTT][ MSE = 0.1826 | LCC = 0.9586 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04033810685575008 +INFO:root: +[Valid][25][UTT][ MSE = 0.1767 | LCC = 0.9584 | SRCC = 0.9584 ] +INFO:root:Average loss=0.038527246162295345 +INFO:root: +[Valid][26][UTT][ MSE = 0.2171 | LCC = 0.9582 | SRCC = 0.9586 ] +INFO:root:Average loss=0.033550711780786516 +INFO:root: +[Valid][27][UTT][ MSE = 0.1957 | LCC = 0.9573 | SRCC = 0.9580 ] +INFO:root:Average loss=0.03297511634230614 +INFO:root: +[Valid][28][UTT][ MSE = 0.1793 | LCC = 0.9585 | SRCC = 0.9584 ] +INFO:root:Average loss=0.033830926418304444 +INFO:root: +[Valid][29][UTT][ MSE = 0.1870 | LCC = 0.9594 | SRCC = 0.9593 ] +INFO:root:Average loss=0.03233312672376633 +INFO:root: +[Valid][30][UTT][ MSE = 0.1726 | LCC = 0.9587 | SRCC = 0.9586 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1893 | LCC = 0.9613 | SRCC = 0.9617 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..66eb2eaf21c4de0ebef7b3b664f44a1a8fcd7eb4 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eb8704cf953948ccb81398efbceb03da849c1509a45f2a9dd860e3630be4bd4f +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..774317e1416e778a7370d2b0e8c5e8961a6dbcfd --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:da9d1dffdd1196d9c84b1d1819348219a1343da13b85b5d2b8f2d088bc8a569d +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..4054904492f0bf0dda4d4dc1a4d282e094a88434 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.47871339750289915 +INFO:root: +[Valid][0][UTT][ MSE = 0.3240 | LCC = 0.9199 | SRCC = 0.9227 ] +INFO:root:Average loss=0.1858013836145401 +INFO:root: +[Valid][1][UTT][ MSE = 0.2572 | LCC = 0.9432 | SRCC = 0.9436 ] +INFO:root:Average loss=0.14667196303606034 +INFO:root: +[Valid][2][UTT][ MSE = 0.2307 | LCC = 0.9495 | SRCC = 0.9504 ] +INFO:root:Average loss=0.12689044496417046 +INFO:root: +[Valid][3][UTT][ MSE = 0.2309 | LCC = 0.9526 | SRCC = 0.9529 ] +INFO:root:Average loss=0.11426289781928062 +INFO:root: +[Valid][4][UTT][ MSE = 0.1967 | LCC = 0.9546 | SRCC = 0.9548 ] +INFO:root:Average loss=0.1019235149025917 +INFO:root: +[Valid][5][UTT][ MSE = 0.2019 | LCC = 0.9559 | SRCC = 0.9567 ] +INFO:root:Average loss=0.09562926340103149 +INFO:root: +[Valid][6][UTT][ MSE = 0.2215 | LCC = 0.9560 | SRCC = 0.9566 ] +INFO:root:Average loss=0.08844079083204269 +INFO:root: +[Valid][7][UTT][ MSE = 0.2019 | LCC = 0.9548 | SRCC = 0.9558 ] +INFO:root:Average loss=0.08284713357686996 +INFO:root: +[Valid][8][UTT][ MSE = 0.2411 | LCC = 0.9577 | SRCC = 0.9583 ] +INFO:root:Average loss=0.07718702971935272 +INFO:root: +[Valid][9][UTT][ MSE = 0.2310 | LCC = 0.9578 | SRCC = 0.9591 ] +INFO:root:Average loss=0.07190829849243165 +INFO:root: +[Valid][10][UTT][ MSE = 0.2208 | LCC = 0.9547 | SRCC = 0.9569 ] +INFO:root:Average loss=0.06837224557995797 +INFO:root: +[Valid][11][UTT][ MSE = 0.2031 | LCC = 0.9531 | SRCC = 0.9542 ] +INFO:root:Average loss=0.06546760043501854 +INFO:root: +[Valid][12][UTT][ MSE = 0.1777 | LCC = 0.9573 | SRCC = 0.9577 ] +INFO:root:Average loss=0.061222145199775695 +INFO:root: +[Valid][13][UTT][ MSE = 0.1894 | LCC = 0.9572 | SRCC = 0.9571 ] +INFO:root:Average loss=0.05877900265157223 +INFO:root: +[Valid][14][UTT][ MSE = 0.2177 | LCC = 0.9581 | SRCC = 0.9587 ] +INFO:root:Average loss=0.05417710843682289 +INFO:root: +[Valid][15][UTT][ MSE = 0.1891 | LCC = 0.9555 | SRCC = 0.9554 ] +INFO:root:Average loss=0.05383208544552326 +INFO:root: +[Valid][16][UTT][ MSE = 0.1896 | LCC = 0.9569 | SRCC = 0.9582 ] +INFO:root:Average loss=0.051987668558955195 +INFO:root: +[Valid][17][UTT][ MSE = 0.2001 | LCC = 0.9582 | SRCC = 0.9585 ] +INFO:root:Average loss=0.048752890244126316 +INFO:root: +[Valid][18][UTT][ MSE = 0.1911 | LCC = 0.9582 | SRCC = 0.9587 ] +INFO:root:Average loss=0.04695938222110271 +INFO:root: +[Valid][19][UTT][ MSE = 0.1921 | LCC = 0.9580 | SRCC = 0.9586 ] +INFO:root:Average loss=0.04478889189660549 +INFO:root: +[Valid][20][UTT][ MSE = 0.2016 | LCC = 0.9584 | SRCC = 0.9589 ] +INFO:root:Average loss=0.04248959177732468 +INFO:root: +[Valid][21][UTT][ MSE = 0.1985 | LCC = 0.9571 | SRCC = 0.9571 ] +INFO:root:Average loss=0.040486804947257045 +INFO:root: +[Valid][22][UTT][ MSE = 0.1778 | LCC = 0.9575 | SRCC = 0.9580 ] +INFO:root:Average loss=0.040887934878468514 +INFO:root: +[Valid][23][UTT][ MSE = 0.2130 | LCC = 0.9579 | SRCC = 0.9583 ] +INFO:root:Average loss=0.03714893032610416 +INFO:root: +[Valid][24][UTT][ MSE = 0.1861 | LCC = 0.9592 | SRCC = 0.9591 ] +INFO:root:Average loss=0.03519544404745102 +INFO:root: +[Valid][25][UTT][ MSE = 0.1824 | LCC = 0.9556 | SRCC = 0.9561 ] +INFO:root:Average loss=0.036345255240798 +INFO:root: +[Valid][26][UTT][ MSE = 0.1848 | LCC = 0.9589 | SRCC = 0.9584 ] +INFO:root:Average loss=0.034178961016237734 +INFO:root: +[Valid][27][UTT][ MSE = 0.1813 | LCC = 0.9587 | SRCC = 0.9581 ] +INFO:root:Average loss=0.032129583328962326 +INFO:root: +[Valid][28][UTT][ MSE = 0.1896 | LCC = 0.9585 | SRCC = 0.9581 ] +INFO:root:Average loss=0.03170903220772743 +INFO:root: +[Valid][29][UTT][ MSE = 0.1945 | LCC = 0.9584 | SRCC = 0.9577 ] +INFO:root:Average loss=0.030456169210374355 +INFO:root: +[Valid][30][UTT][ MSE = 0.1952 | LCC = 0.9574 | SRCC = 0.9574 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1871 | LCC = 0.9613 | SRCC = 0.9620 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f6147301bd4e7987896153213b9083b84f13f5f0 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d61f4b0f06f6b327d7174e4f2c213e338cded110c2a7899dc54c0cae8a2221f3 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0ca201eac0f5901fe74e34f7ba803b7de34b42a8 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fdb6cde5c1e932e19caa3c04b96d53d7e4e435a7cc3fb1588083b6188139f3cb +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..cc70d09414ffd646975f97ec20669f4caec7a30f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer3/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4431458704471588 +INFO:root: +[Valid][0][UTT][ MSE = 0.3019 | LCC = 0.9232 | SRCC = 0.9233 ] +INFO:root:Average loss=0.18162953007221222 +INFO:root: +[Valid][1][UTT][ MSE = 0.2536 | LCC = 0.9412 | SRCC = 0.9414 ] +INFO:root:Average loss=0.14241428196430206 +INFO:root: +[Valid][2][UTT][ MSE = 0.2437 | LCC = 0.9469 | SRCC = 0.9490 ] +INFO:root:Average loss=0.12600676345825196 +INFO:root: +[Valid][3][UTT][ MSE = 0.2399 | LCC = 0.9496 | SRCC = 0.9498 ] +INFO:root:Average loss=0.11284900271892548 +INFO:root: +[Valid][4][UTT][ MSE = 0.1894 | LCC = 0.9530 | SRCC = 0.9545 ] +INFO:root:Average loss=0.10402562287449836 +INFO:root: +[Valid][5][UTT][ MSE = 0.2065 | LCC = 0.9541 | SRCC = 0.9555 ] +INFO:root:Average loss=0.09966082522273063 +INFO:root: +[Valid][6][UTT][ MSE = 0.2088 | LCC = 0.9551 | SRCC = 0.9561 ] +INFO:root:Average loss=0.08970675954222679 +INFO:root: +[Valid][7][UTT][ MSE = 0.2151 | LCC = 0.9554 | SRCC = 0.9560 ] +INFO:root:Average loss=0.08239482000470161 +INFO:root: +[Valid][8][UTT][ MSE = 0.2288 | LCC = 0.9577 | SRCC = 0.9583 ] +INFO:root:Average loss=0.08035438138246537 +INFO:root: +[Valid][9][UTT][ MSE = 0.1869 | LCC = 0.9570 | SRCC = 0.9580 ] +INFO:root:Average loss=0.07461319568753243 +INFO:root: +[Valid][10][UTT][ MSE = 0.2187 | LCC = 0.9538 | SRCC = 0.9545 ] +INFO:root:Average loss=0.06766999658942223 +INFO:root: +[Valid][11][UTT][ MSE = 0.1965 | LCC = 0.9576 | SRCC = 0.9583 ] +INFO:root:Average loss=0.065140420794487 +INFO:root: +[Valid][12][UTT][ MSE = 0.2106 | LCC = 0.9563 | SRCC = 0.9564 ] +INFO:root:Average loss=0.06419787445664406 +INFO:root: +[Valid][13][UTT][ MSE = 0.2039 | LCC = 0.9583 | SRCC = 0.9584 ] +INFO:root:Average loss=0.05705048510432243 +INFO:root: +[Valid][14][UTT][ MSE = 0.2029 | LCC = 0.9580 | SRCC = 0.9581 ] +INFO:root:Average loss=0.05516422954201698 +INFO:root: +[Valid][15][UTT][ MSE = 0.1891 | LCC = 0.9583 | SRCC = 0.9584 ] +INFO:root:Average loss=0.05417808301746845 +INFO:root: +[Valid][16][UTT][ MSE = 0.2130 | LCC = 0.9569 | SRCC = 0.9575 ] +INFO:root:Average loss=0.050631227999925615 +INFO:root: +[Valid][17][UTT][ MSE = 0.2173 | LCC = 0.9583 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04930158424377441 +INFO:root: +[Valid][18][UTT][ MSE = 0.1832 | LCC = 0.9576 | SRCC = 0.9583 ] +INFO:root:Average loss=0.04749850019812584 +INFO:root: +[Valid][19][UTT][ MSE = 0.1737 | LCC = 0.9568 | SRCC = 0.9581 ] +INFO:root:Average loss=0.04402059504389763 +INFO:root: +[Valid][20][UTT][ MSE = 0.2217 | LCC = 0.9570 | SRCC = 0.9569 ] +INFO:root:Average loss=0.04608544833958149 +INFO:root: +[Valid][21][UTT][ MSE = 0.2022 | LCC = 0.9569 | SRCC = 0.9574 ] +INFO:root:Average loss=0.04142291843891144 +INFO:root: +[Valid][22][UTT][ MSE = 0.1971 | LCC = 0.9545 | SRCC = 0.9549 ] +INFO:root:Average loss=0.04045705609023571 +INFO:root: +[Valid][23][UTT][ MSE = 0.1914 | LCC = 0.9598 | SRCC = 0.9601 ] +INFO:root:Average loss=0.03907561133801937 +INFO:root: +[Valid][24][UTT][ MSE = 0.2095 | LCC = 0.9581 | SRCC = 0.9579 ] +INFO:root:Average loss=0.03690817373991013 +INFO:root: +[Valid][25][UTT][ MSE = 0.2049 | LCC = 0.9583 | SRCC = 0.9589 ] +INFO:root:Average loss=0.03461463070660829 +INFO:root: +[Valid][26][UTT][ MSE = 0.1854 | LCC = 0.9578 | SRCC = 0.9579 ] +INFO:root:Average loss=0.03573756130039692 +INFO:root: +[Valid][27][UTT][ MSE = 0.2037 | LCC = 0.9567 | SRCC = 0.9569 ] +INFO:root:Average loss=0.03343098133802414 +INFO:root: +[Valid][28][UTT][ MSE = 0.2013 | LCC = 0.9575 | SRCC = 0.9574 ] +INFO:root:Average loss=0.03197427412867546 +INFO:root: +[Valid][29][UTT][ MSE = 0.1881 | LCC = 0.9583 | SRCC = 0.9585 ] +INFO:root:Average loss=0.03063482528179884 +INFO:root: +[Valid][30][UTT][ MSE = 0.1907 | LCC = 0.9602 | SRCC = 0.9606 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1892 | LCC = 0.9611 | SRCC = 0.9620 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ace5b3b12b0374708e437ecb1616e200787c56dd --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2ddd5effad8e76b95ec0c7123376113aac13d3e7bb7d5b6cb85680f5c00c9895 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..1855b3faf0f5ba93810118bf7a29dc6f48a94f9c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ebb7283ae4186bc3ed709672882bbe244db9fe6cce3dc8429d2313240d447233 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e82c5cfb886d631f7145b6934b064b7c2c589276 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4109841995239258 +INFO:root: +[Valid][0][UTT][ MSE = 0.3160 | LCC = 0.9252 | SRCC = 0.9263 ] +INFO:root:Average loss=0.17779469668865203 +INFO:root: +[Valid][1][UTT][ MSE = 0.2484 | LCC = 0.9433 | SRCC = 0.9456 ] +INFO:root:Average loss=0.15172755777835845 +INFO:root: +[Valid][2][UTT][ MSE = 0.3021 | LCC = 0.9449 | SRCC = 0.9469 ] +INFO:root:Average loss=0.13237388944625855 +INFO:root: +[Valid][3][UTT][ MSE = 0.2329 | LCC = 0.9519 | SRCC = 0.9529 ] +INFO:root:Average loss=0.11887434595823287 +INFO:root: +[Valid][4][UTT][ MSE = 0.2206 | LCC = 0.9519 | SRCC = 0.9524 ] +INFO:root:Average loss=0.10738262841105461 +INFO:root: +[Valid][5][UTT][ MSE = 0.2661 | LCC = 0.9539 | SRCC = 0.9553 ] +INFO:root:Average loss=0.1024795241355896 +INFO:root: +[Valid][6][UTT][ MSE = 0.2144 | LCC = 0.9562 | SRCC = 0.9569 ] +INFO:root:Average loss=0.09276399785280227 +INFO:root: +[Valid][7][UTT][ MSE = 0.2369 | LCC = 0.9553 | SRCC = 0.9563 ] +INFO:root:Average loss=0.08698916152119636 +INFO:root: +[Valid][8][UTT][ MSE = 0.1926 | LCC = 0.9539 | SRCC = 0.9539 ] +INFO:root:Average loss=0.081055908203125 +INFO:root: +[Valid][9][UTT][ MSE = 0.2090 | LCC = 0.9574 | SRCC = 0.9583 ] +INFO:root:Average loss=0.07544783654808998 +INFO:root: +[Valid][10][UTT][ MSE = 0.2142 | LCC = 0.9557 | SRCC = 0.9556 ] +INFO:root:Average loss=0.07202144941687584 +INFO:root: +[Valid][11][UTT][ MSE = 0.1955 | LCC = 0.9554 | SRCC = 0.9559 ] +INFO:root:Average loss=0.06772127485275269 +INFO:root: +[Valid][12][UTT][ MSE = 0.2150 | LCC = 0.9578 | SRCC = 0.9584 ] +INFO:root:Average loss=0.06314583882689476 +INFO:root: +[Valid][13][UTT][ MSE = 0.2117 | LCC = 0.9582 | SRCC = 0.9588 ] +INFO:root:Average loss=0.059902838438749315 +INFO:root: +[Valid][14][UTT][ MSE = 0.2081 | LCC = 0.9573 | SRCC = 0.9577 ] +INFO:root:Average loss=0.059154370903968814 +INFO:root: +[Valid][15][UTT][ MSE = 0.2153 | LCC = 0.9570 | SRCC = 0.9572 ] +INFO:root:Average loss=0.05262029515206814 +INFO:root: +[Valid][16][UTT][ MSE = 0.2119 | LCC = 0.9577 | SRCC = 0.9579 ] +INFO:root:Average loss=0.05238126994669438 +INFO:root: +[Valid][17][UTT][ MSE = 0.2198 | LCC = 0.9581 | SRCC = 0.9581 ] +INFO:root:Average loss=0.049792143017053606 +INFO:root: +[Valid][18][UTT][ MSE = 0.2182 | LCC = 0.9548 | SRCC = 0.9553 ] +INFO:root:Average loss=0.046118711709976196 +INFO:root: +[Valid][19][UTT][ MSE = 0.2144 | LCC = 0.9576 | SRCC = 0.9576 ] +INFO:root:Average loss=0.04478552033007145 +INFO:root: +[Valid][20][UTT][ MSE = 0.2263 | LCC = 0.9569 | SRCC = 0.9574 ] +INFO:root:Average loss=0.04703011672198772 +INFO:root: +[Valid][21][UTT][ MSE = 0.1972 | LCC = 0.9557 | SRCC = 0.9559 ] +INFO:root:Average loss=0.04095473338663578 +INFO:root: +[Valid][22][UTT][ MSE = 0.1832 | LCC = 0.9547 | SRCC = 0.9553 ] +INFO:root:Average loss=0.040740761056542396 +INFO:root: +[Valid][23][UTT][ MSE = 0.2279 | LCC = 0.9527 | SRCC = 0.9529 ] +INFO:root:Average loss=0.04030142103135586 +INFO:root: +[Valid][24][UTT][ MSE = 0.1923 | LCC = 0.9573 | SRCC = 0.9577 ] +INFO:root:Average loss=0.03793483678996563 +INFO:root: +[Valid][25][UTT][ MSE = 0.1852 | LCC = 0.9568 | SRCC = 0.9569 ] +INFO:root:Average loss=0.038748045042157175 +INFO:root: +[Valid][26][UTT][ MSE = 0.2080 | LCC = 0.9573 | SRCC = 0.9573 ] +INFO:root:Average loss=0.03363405810296535 +INFO:root: +[Valid][27][UTT][ MSE = 0.1854 | LCC = 0.9567 | SRCC = 0.9567 ] +INFO:root:Average loss=0.03378842556476593 +INFO:root: +[Valid][28][UTT][ MSE = 0.1971 | LCC = 0.9568 | SRCC = 0.9568 ] +INFO:root:Average loss=0.03347498458623886 +INFO:root: +[Valid][29][UTT][ MSE = 0.2006 | LCC = 0.9566 | SRCC = 0.9569 ] +INFO:root:Average loss=0.033655927270650865 +INFO:root: +[Valid][30][UTT][ MSE = 0.2036 | LCC = 0.9573 | SRCC = 0.9573 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2248 | LCC = 0.9569 | SRCC = 0.9582 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..191afd5b8e9ddabd0df58686fbc6d3c04df77cac --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:84c6ab497c93f52c84cec0d32d9abb0971734f85623fabaa4c86715fb08dc7f8 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..bc17f8646182c1ff9c5d2ed83be4852bb32c58d0 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eec7b449551f5436924c66f67e91d3216ebbde313886b8001eb82b6150b61e6a +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..8d1f748b829eaf29e8d8a68ef8bdd2c9a897069a --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.34505884754657745 +INFO:root: +[Valid][0][UTT][ MSE = 0.2785 | LCC = 0.9299 | SRCC = 0.9313 ] +INFO:root:Average loss=0.1872590844631195 +INFO:root: +[Valid][1][UTT][ MSE = 0.2456 | LCC = 0.9431 | SRCC = 0.9434 ] +INFO:root:Average loss=0.14968288546800612 +INFO:root: +[Valid][2][UTT][ MSE = 0.2527 | LCC = 0.9468 | SRCC = 0.9465 ] +INFO:root:Average loss=0.1306014972925186 +INFO:root: +[Valid][3][UTT][ MSE = 0.2493 | LCC = 0.9458 | SRCC = 0.9493 ] +INFO:root:Average loss=0.12069900751113892 +INFO:root: +[Valid][4][UTT][ MSE = 0.2496 | LCC = 0.9512 | SRCC = 0.9521 ] +INFO:root:Average loss=0.11285368132591247 +INFO:root: +[Valid][5][UTT][ MSE = 0.2877 | LCC = 0.9504 | SRCC = 0.9547 ] +INFO:root:Average loss=0.10220553296804429 +INFO:root: +[Valid][6][UTT][ MSE = 0.2002 | LCC = 0.9540 | SRCC = 0.9543 ] +INFO:root:Average loss=0.09520770838856697 +INFO:root: +[Valid][7][UTT][ MSE = 0.2403 | LCC = 0.9540 | SRCC = 0.9556 ] +INFO:root:Average loss=0.08972676658630371 +INFO:root: +[Valid][8][UTT][ MSE = 0.2064 | LCC = 0.9543 | SRCC = 0.9547 ] +INFO:root:Average loss=0.08690524291992187 +INFO:root: +[Valid][9][UTT][ MSE = 0.1841 | LCC = 0.9559 | SRCC = 0.9568 ] +INFO:root:Average loss=0.07491436007618904 +INFO:root: +[Valid][10][UTT][ MSE = 0.2130 | LCC = 0.9554 | SRCC = 0.9553 ] +INFO:root:Average loss=0.07200428828597069 +INFO:root: +[Valid][11][UTT][ MSE = 0.2008 | LCC = 0.9560 | SRCC = 0.9560 ] +INFO:root:Average loss=0.06764858248829841 +INFO:root: +[Valid][12][UTT][ MSE = 0.1977 | LCC = 0.9555 | SRCC = 0.9559 ] +INFO:root:Average loss=0.06416667696833611 +INFO:root: +[Valid][13][UTT][ MSE = 0.2176 | LCC = 0.9565 | SRCC = 0.9577 ] +INFO:root:Average loss=0.0625202457010746 +INFO:root: +[Valid][14][UTT][ MSE = 0.2122 | LCC = 0.9564 | SRCC = 0.9571 ] +INFO:root:Average loss=0.06353186416625976 +INFO:root: +[Valid][15][UTT][ MSE = 0.2209 | LCC = 0.9558 | SRCC = 0.9557 ] +INFO:root:Average loss=0.05920123109221458 +INFO:root: +[Valid][16][UTT][ MSE = 0.1900 | LCC = 0.9588 | SRCC = 0.9593 ] +INFO:root:Average loss=0.05377444788813591 +INFO:root: +[Valid][17][UTT][ MSE = 0.2206 | LCC = 0.9560 | SRCC = 0.9566 ] +INFO:root:Average loss=0.05111109472811222 +INFO:root: +[Valid][18][UTT][ MSE = 0.2132 | LCC = 0.9561 | SRCC = 0.9566 ] +INFO:root:Average loss=0.04894554570317269 +INFO:root: +[Valid][19][UTT][ MSE = 0.2104 | LCC = 0.9557 | SRCC = 0.9566 ] +INFO:root:Average loss=0.04521441906690597 +INFO:root: +[Valid][20][UTT][ MSE = 0.1881 | LCC = 0.9565 | SRCC = 0.9570 ] +INFO:root:Average loss=0.04246924017369747 +INFO:root: +[Valid][21][UTT][ MSE = 0.1974 | LCC = 0.9574 | SRCC = 0.9572 ] +INFO:root:Average loss=0.042791969686746596 +INFO:root: +[Valid][22][UTT][ MSE = 0.2122 | LCC = 0.9568 | SRCC = 0.9572 ] +INFO:root:Average loss=0.040818573996424676 +INFO:root: +[Valid][23][UTT][ MSE = 0.2242 | LCC = 0.9567 | SRCC = 0.9572 ] +INFO:root:Average loss=0.03883113442361355 +INFO:root: +[Valid][24][UTT][ MSE = 0.2144 | LCC = 0.9561 | SRCC = 0.9562 ] +INFO:root:Average loss=0.03765919172763824 +INFO:root: +[Valid][25][UTT][ MSE = 0.2049 | LCC = 0.9576 | SRCC = 0.9575 ] +INFO:root:Average loss=0.0371649002879858 +INFO:root: +[Valid][26][UTT][ MSE = 0.2050 | LCC = 0.9561 | SRCC = 0.9559 ] +INFO:root:Average loss=0.03462309722602368 +INFO:root: +[Valid][27][UTT][ MSE = 0.1914 | LCC = 0.9570 | SRCC = 0.9569 ] +INFO:root:Average loss=0.033511832654476166 +INFO:root: +[Valid][28][UTT][ MSE = 0.1940 | LCC = 0.9568 | SRCC = 0.9564 ] +INFO:root:Average loss=0.033512157112360004 +INFO:root: +[Valid][29][UTT][ MSE = 0.2100 | LCC = 0.9573 | SRCC = 0.9571 ] +INFO:root:Average loss=0.03335204717516899 +INFO:root: +[Valid][30][UTT][ MSE = 0.2032 | LCC = 0.9583 | SRCC = 0.9585 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1997 | LCC = 0.9584 | SRCC = 0.9596 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a93102f33e2f5607efe8e9c8b8b7e89e60421a9d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d10abec64646ace6e3b9fa21472c4fb6d76b34e0602cef5b2afb9d9f094c9bc8 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f02ac4a7c337ccd2f2b17b2ab0fd23cb031b1326 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:464ec1d7c95a98967e40de5fc5bf0f63af77c46dae2872ee0d0bcbbe3d3ba5a3 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..27e59558695b405ec811bf942d9c32584ec2d195 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.40266659975051877 +INFO:root: +[Valid][0][UTT][ MSE = 0.3133 | LCC = 0.9239 | SRCC = 0.9248 ] +INFO:root:Average loss=0.18190583086013795 +INFO:root: +[Valid][1][UTT][ MSE = 0.2401 | LCC = 0.9426 | SRCC = 0.9433 ] +INFO:root:Average loss=0.15138049095869063 +INFO:root: +[Valid][2][UTT][ MSE = 0.2301 | LCC = 0.9473 | SRCC = 0.9486 ] +INFO:root:Average loss=0.13196607699990273 +INFO:root: +[Valid][3][UTT][ MSE = 0.2393 | LCC = 0.9502 | SRCC = 0.9518 ] +INFO:root:Average loss=0.1147264450788498 +INFO:root: +[Valid][4][UTT][ MSE = 0.2158 | LCC = 0.9524 | SRCC = 0.9521 ] +INFO:root:Average loss=0.10628675836324691 +INFO:root: +[Valid][5][UTT][ MSE = 0.2072 | LCC = 0.9542 | SRCC = 0.9550 ] +INFO:root:Average loss=0.10306210750341416 +INFO:root: +[Valid][6][UTT][ MSE = 0.1976 | LCC = 0.9557 | SRCC = 0.9567 ] +INFO:root:Average loss=0.0902158836722374 +INFO:root: +[Valid][7][UTT][ MSE = 0.1880 | LCC = 0.9555 | SRCC = 0.9557 ] +INFO:root:Average loss=0.0865193048119545 +INFO:root: +[Valid][8][UTT][ MSE = 0.1936 | LCC = 0.9579 | SRCC = 0.9575 ] +INFO:root:Average loss=0.07758637472987175 +INFO:root: +[Valid][9][UTT][ MSE = 0.1929 | LCC = 0.9574 | SRCC = 0.9575 ] +INFO:root:Average loss=0.07374835756421089 +INFO:root: +[Valid][10][UTT][ MSE = 0.1903 | LCC = 0.9566 | SRCC = 0.9563 ] +INFO:root:Average loss=0.07463263189792634 +INFO:root: +[Valid][11][UTT][ MSE = 0.1989 | LCC = 0.9568 | SRCC = 0.9562 ] +INFO:root:Average loss=0.06480506545305252 +INFO:root: +[Valid][12][UTT][ MSE = 0.1690 | LCC = 0.9574 | SRCC = 0.9572 ] +INFO:root:Average loss=0.06546191042661667 +INFO:root: +[Valid][13][UTT][ MSE = 0.1869 | LCC = 0.9586 | SRCC = 0.9581 ] +INFO:root:Average loss=0.06140268057584763 +INFO:root: +[Valid][14][UTT][ MSE = 0.1614 | LCC = 0.9570 | SRCC = 0.9569 ] +INFO:root:Average loss=0.05555249300599098 +INFO:root: +[Valid][15][UTT][ MSE = 0.1908 | LCC = 0.9558 | SRCC = 0.9562 ] +INFO:root:Average loss=0.0537428984940052 +INFO:root: +[Valid][16][UTT][ MSE = 0.1946 | LCC = 0.9548 | SRCC = 0.9540 ] +INFO:root:Average loss=0.05093618522584438 +INFO:root: +[Valid][17][UTT][ MSE = 0.1844 | LCC = 0.9591 | SRCC = 0.9585 ] +INFO:root:Average loss=0.048768073067069054 +INFO:root: +[Valid][18][UTT][ MSE = 0.2006 | LCC = 0.9572 | SRCC = 0.9553 ] +INFO:root:Average loss=0.04921975582838058 +INFO:root: +[Valid][19][UTT][ MSE = 0.1875 | LCC = 0.9579 | SRCC = 0.9572 ] +INFO:root:Average loss=0.048951120942831036 +INFO:root: +[Valid][20][UTT][ MSE = 0.1793 | LCC = 0.9560 | SRCC = 0.9550 ] +INFO:root:Average loss=0.0443808760792017 +INFO:root: +[Valid][21][UTT][ MSE = 0.1862 | LCC = 0.9580 | SRCC = 0.9579 ] +INFO:root:Average loss=0.040367054030299186 +INFO:root: +[Valid][22][UTT][ MSE = 0.1830 | LCC = 0.9573 | SRCC = 0.9566 ] +INFO:root:Average loss=0.039141142189502716 +INFO:root: +[Valid][23][UTT][ MSE = 0.1819 | LCC = 0.9573 | SRCC = 0.9571 ] +INFO:root:Average loss=0.03813562296330929 +INFO:root: +[Valid][24][UTT][ MSE = 0.1726 | LCC = 0.9559 | SRCC = 0.9554 ] +INFO:root:Average loss=0.03932301090657711 +INFO:root: +[Valid][25][UTT][ MSE = 0.1704 | LCC = 0.9586 | SRCC = 0.9580 ] +INFO:root:Average loss=0.03470719636976719 +INFO:root: +[Valid][26][UTT][ MSE = 0.1843 | LCC = 0.9590 | SRCC = 0.9584 ] +INFO:root:Average loss=0.03198687352240086 +INFO:root: +[Valid][27][UTT][ MSE = 0.1855 | LCC = 0.9576 | SRCC = 0.9566 ] +INFO:root:Average loss=0.03292430330812931 +INFO:root: +[Valid][28][UTT][ MSE = 0.1749 | LCC = 0.9589 | SRCC = 0.9584 ] +INFO:root:Average loss=0.0321594677567482 +INFO:root: +[Valid][29][UTT][ MSE = 0.1915 | LCC = 0.9570 | SRCC = 0.9562 ] +INFO:root:Average loss=0.03245609647035599 +INFO:root: +[Valid][30][UTT][ MSE = 0.1691 | LCC = 0.9578 | SRCC = 0.9573 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1960 | LCC = 0.9582 | SRCC = 0.9593 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3c2da9bac802a24465d3c6b1b48cc89abdf9ac81 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:955b6e3453cd2c5e82e4f353cdd977d72b1283c70f362a6ab22da7f14e7e236e +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..57fa94be5e8c08bd3cfda2086ff0e58a180c3571 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c5f8ad279eb160c8878a35d9883508148c4f62230b1d318cd5e5381c0b62a4fd +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..61c831234abc72e6c316cba8f2a3a8981b02ac04 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3803337653875351 +INFO:root: +[Valid][0][UTT][ MSE = 0.2503 | LCC = 0.9321 | SRCC = 0.9340 ] +INFO:root:Average loss=0.1817507405281067 +INFO:root: +[Valid][1][UTT][ MSE = 0.2069 | LCC = 0.9423 | SRCC = 0.9453 ] +INFO:root:Average loss=0.14614957910776138 +INFO:root: +[Valid][2][UTT][ MSE = 0.2009 | LCC = 0.9482 | SRCC = 0.9478 ] +INFO:root:Average loss=0.12671906638145447 +INFO:root: +[Valid][3][UTT][ MSE = 0.2747 | LCC = 0.9446 | SRCC = 0.9469 ] +INFO:root:Average loss=0.1170536438524723 +INFO:root: +[Valid][4][UTT][ MSE = 0.1934 | LCC = 0.9537 | SRCC = 0.9538 ] +INFO:root:Average loss=0.10651950204372405 +INFO:root: +[Valid][5][UTT][ MSE = 0.2010 | LCC = 0.9544 | SRCC = 0.9544 ] +INFO:root:Average loss=0.09801290959119796 +INFO:root: +[Valid][6][UTT][ MSE = 0.2087 | LCC = 0.9558 | SRCC = 0.9555 ] +INFO:root:Average loss=0.09224679416418076 +INFO:root: +[Valid][7][UTT][ MSE = 0.2336 | LCC = 0.9494 | SRCC = 0.9491 ] +INFO:root:Average loss=0.09083675476908684 +INFO:root: +[Valid][8][UTT][ MSE = 0.1872 | LCC = 0.9548 | SRCC = 0.9547 ] +INFO:root:Average loss=0.08033430486917496 +INFO:root: +[Valid][9][UTT][ MSE = 0.2122 | LCC = 0.9558 | SRCC = 0.9561 ] +INFO:root:Average loss=0.07396079182624817 +INFO:root: +[Valid][10][UTT][ MSE = 0.2097 | LCC = 0.9564 | SRCC = 0.9576 ] +INFO:root:Average loss=0.07452303084731102 +INFO:root: +[Valid][11][UTT][ MSE = 0.2084 | LCC = 0.9570 | SRCC = 0.9588 ] +INFO:root:Average loss=0.06727582633495331 +INFO:root: +[Valid][12][UTT][ MSE = 0.1859 | LCC = 0.9527 | SRCC = 0.9518 ] +INFO:root:Average loss=0.061339378774166106 +INFO:root: +[Valid][13][UTT][ MSE = 0.1765 | LCC = 0.9571 | SRCC = 0.9570 ] +INFO:root:Average loss=0.06085424986481667 +INFO:root: +[Valid][14][UTT][ MSE = 0.1889 | LCC = 0.9568 | SRCC = 0.9565 ] +INFO:root:Average loss=0.05601621246337891 +INFO:root: +[Valid][15][UTT][ MSE = 0.1869 | LCC = 0.9591 | SRCC = 0.9595 ] +INFO:root:Average loss=0.05311819180846214 +INFO:root: +[Valid][16][UTT][ MSE = 0.1673 | LCC = 0.9575 | SRCC = 0.9574 ] +INFO:root:Average loss=0.05106305706501007 +INFO:root: +[Valid][17][UTT][ MSE = 0.1743 | LCC = 0.9581 | SRCC = 0.9581 ] +INFO:root:Average loss=0.05036891664564609 +INFO:root: +[Valid][18][UTT][ MSE = 0.2048 | LCC = 0.9579 | SRCC = 0.9591 ] +INFO:root:Average loss=0.04678826655447483 +INFO:root: +[Valid][19][UTT][ MSE = 0.1899 | LCC = 0.9584 | SRCC = 0.9586 ] +INFO:root:Average loss=0.043676041468977925 +INFO:root: +[Valid][20][UTT][ MSE = 0.1904 | LCC = 0.9561 | SRCC = 0.9555 ] +INFO:root:Average loss=0.045529259145259855 +INFO:root: +[Valid][21][UTT][ MSE = 0.2020 | LCC = 0.9588 | SRCC = 0.9586 ] +INFO:root:Average loss=0.041104242399334905 +INFO:root: +[Valid][22][UTT][ MSE = 0.1816 | LCC = 0.9591 | SRCC = 0.9589 ] +INFO:root:Average loss=0.03906361697614193 +INFO:root: +[Valid][23][UTT][ MSE = 0.1931 | LCC = 0.9571 | SRCC = 0.9572 ] +INFO:root:Average loss=0.036754277974367144 +INFO:root: +[Valid][24][UTT][ MSE = 0.1798 | LCC = 0.9588 | SRCC = 0.9585 ] +INFO:root:Average loss=0.03562306070327759 +INFO:root: +[Valid][25][UTT][ MSE = 0.1761 | LCC = 0.9590 | SRCC = 0.9594 ] +INFO:root:Average loss=0.034613435983657834 +INFO:root: +[Valid][26][UTT][ MSE = 0.1690 | LCC = 0.9599 | SRCC = 0.9599 ] +INFO:root:Average loss=0.035258956104516986 +INFO:root: +[Valid][27][UTT][ MSE = 0.2055 | LCC = 0.9561 | SRCC = 0.9574 ] +INFO:root:Average loss=0.03526039572060108 +INFO:root: +[Valid][28][UTT][ MSE = 0.1795 | LCC = 0.9572 | SRCC = 0.9578 ] +INFO:root:Average loss=0.03387531600892544 +INFO:root: +[Valid][29][UTT][ MSE = 0.1782 | LCC = 0.9599 | SRCC = 0.9604 ] +INFO:root:Average loss=0.0309514689296484 +INFO:root: +[Valid][30][UTT][ MSE = 0.1991 | LCC = 0.9598 | SRCC = 0.9602 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1860 | LCC = 0.9588 | SRCC = 0.9605 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5af338143aec1e6471265cac704e07b71f352adb --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:faf272308f095b8c8ea6ea6a12edcea71e360775e122dfdd46ec3240fc89b9c5 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e7a4ee4bf33a9b0575492dac8b39ee6b489caae5 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bdf08d237fb9d1c51198028e8a986bcfca36dbbe6ac6b641bafa2d4d6aa3d5e7 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..490392ba3ae032a42ccca4c9f846a35832e3274e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer4/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.43342790865898134 +INFO:root: +[Valid][0][UTT][ MSE = 0.2782 | LCC = 0.9254 | SRCC = 0.9293 ] +INFO:root:Average loss=0.18762906479835512 +INFO:root: +[Valid][1][UTT][ MSE = 0.2673 | LCC = 0.9421 | SRCC = 0.9443 ] +INFO:root:Average loss=0.15010770463943482 +INFO:root: +[Valid][2][UTT][ MSE = 0.2481 | LCC = 0.9489 | SRCC = 0.9502 ] +INFO:root:Average loss=0.1360975494980812 +INFO:root: +[Valid][3][UTT][ MSE = 0.2315 | LCC = 0.9504 | SRCC = 0.9515 ] +INFO:root:Average loss=0.11954224890470505 +INFO:root: +[Valid][4][UTT][ MSE = 0.2104 | LCC = 0.9536 | SRCC = 0.9549 ] +INFO:root:Average loss=0.11151766270399094 +INFO:root: +[Valid][5][UTT][ MSE = 0.2427 | LCC = 0.9536 | SRCC = 0.9549 ] +INFO:root:Average loss=0.1006438801586628 +INFO:root: +[Valid][6][UTT][ MSE = 0.1989 | LCC = 0.9555 | SRCC = 0.9563 ] +INFO:root:Average loss=0.09224212762713432 +INFO:root: +[Valid][7][UTT][ MSE = 0.2006 | LCC = 0.9555 | SRCC = 0.9557 ] +INFO:root:Average loss=0.08522941946983337 +INFO:root: +[Valid][8][UTT][ MSE = 0.2105 | LCC = 0.9562 | SRCC = 0.9561 ] +INFO:root:Average loss=0.07948589435219765 +INFO:root: +[Valid][9][UTT][ MSE = 0.2200 | LCC = 0.9571 | SRCC = 0.9575 ] +INFO:root:Average loss=0.07645492872595787 +INFO:root: +[Valid][10][UTT][ MSE = 0.1929 | LCC = 0.9575 | SRCC = 0.9576 ] +INFO:root:Average loss=0.06990088587999344 +INFO:root: +[Valid][11][UTT][ MSE = 0.2204 | LCC = 0.9567 | SRCC = 0.9572 ] +INFO:root:Average loss=0.06610057485103607 +INFO:root: +[Valid][12][UTT][ MSE = 0.1899 | LCC = 0.9573 | SRCC = 0.9576 ] +INFO:root:Average loss=0.06398276275396347 +INFO:root: +[Valid][13][UTT][ MSE = 0.1983 | LCC = 0.9557 | SRCC = 0.9548 ] +INFO:root:Average loss=0.061416313290596006 +INFO:root: +[Valid][14][UTT][ MSE = 0.1946 | LCC = 0.9574 | SRCC = 0.9569 ] +INFO:root:Average loss=0.05603737132251262 +INFO:root: +[Valid][15][UTT][ MSE = 0.2080 | LCC = 0.9578 | SRCC = 0.9578 ] +INFO:root:Average loss=0.05438374108076095 +INFO:root: +[Valid][16][UTT][ MSE = 0.2378 | LCC = 0.9568 | SRCC = 0.9566 ] +INFO:root:Average loss=0.049625846400856974 +INFO:root: +[Valid][17][UTT][ MSE = 0.2105 | LCC = 0.9580 | SRCC = 0.9576 ] +INFO:root:Average loss=0.04661442185938358 +INFO:root: +[Valid][18][UTT][ MSE = 0.2151 | LCC = 0.9577 | SRCC = 0.9577 ] +INFO:root:Average loss=0.04702937941253185 +INFO:root: +[Valid][19][UTT][ MSE = 0.2411 | LCC = 0.9573 | SRCC = 0.9579 ] +INFO:root:Average loss=0.04458038803935051 +INFO:root: +[Valid][20][UTT][ MSE = 0.2278 | LCC = 0.9590 | SRCC = 0.9588 ] +INFO:root:Average loss=0.04394047489762306 +INFO:root: +[Valid][21][UTT][ MSE = 0.2066 | LCC = 0.9590 | SRCC = 0.9585 ] +INFO:root:Average loss=0.04049551185965538 +INFO:root: +[Valid][22][UTT][ MSE = 0.1847 | LCC = 0.9601 | SRCC = 0.9593 ] +INFO:root:Average loss=0.041118221431970595 +INFO:root: +[Valid][23][UTT][ MSE = 0.1854 | LCC = 0.9592 | SRCC = 0.9588 ] +INFO:root:Average loss=0.04041275672614574 +INFO:root: +[Valid][24][UTT][ MSE = 0.2034 | LCC = 0.9595 | SRCC = 0.9589 ] +INFO:root:Average loss=0.037004569336771966 +INFO:root: +[Valid][25][UTT][ MSE = 0.2084 | LCC = 0.9577 | SRCC = 0.9570 ] +INFO:root:Average loss=0.03523528423905373 +INFO:root: +[Valid][26][UTT][ MSE = 0.1825 | LCC = 0.9581 | SRCC = 0.9579 ] +INFO:root:Average loss=0.0344405555576086 +INFO:root: +[Valid][27][UTT][ MSE = 0.2060 | LCC = 0.9591 | SRCC = 0.9587 ] +INFO:root:Average loss=0.03500353348255157 +INFO:root: +[Valid][28][UTT][ MSE = 0.1939 | LCC = 0.9600 | SRCC = 0.9599 ] +INFO:root:Average loss=0.03300565625727177 +INFO:root: +[Valid][29][UTT][ MSE = 0.1818 | LCC = 0.9601 | SRCC = 0.9592 ] +INFO:root:Average loss=0.03173108524084091 +INFO:root: +[Valid][30][UTT][ MSE = 0.2044 | LCC = 0.9595 | SRCC = 0.9595 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1964 | LCC = 0.9594 | SRCC = 0.9598 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..06d233a2853e03a9af7d63a1247a209fc2b3a148 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1878b1af5fd5ed96d39b06c31aa2b9a10d8e39f2288126c5f5d152b121e552ec +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ae54c37cd8f354903f3993a9a4d3f4b94a34f808 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dee341ea1660d60b4fa04dc4cc62fa934dafa44fadac0f6f27ec79aaa16d80b8 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..dc29e20894a476ad67a4c1d7bf216b25ef67cf48 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.45631615352630617 +INFO:root: +[Valid][0][UTT][ MSE = 0.3036 | LCC = 0.9265 | SRCC = 0.9277 ] +INFO:root:Average loss=0.1756517783999443 +INFO:root: +[Valid][1][UTT][ MSE = 0.3112 | LCC = 0.9424 | SRCC = 0.9442 ] +INFO:root:Average loss=0.14883575481176375 +INFO:root: +[Valid][2][UTT][ MSE = 0.2691 | LCC = 0.9475 | SRCC = 0.9476 ] +INFO:root:Average loss=0.13077098000049592 +INFO:root: +[Valid][3][UTT][ MSE = 0.2381 | LCC = 0.9488 | SRCC = 0.9494 ] +INFO:root:Average loss=0.11819477647542953 +INFO:root: +[Valid][4][UTT][ MSE = 0.2620 | LCC = 0.9508 | SRCC = 0.9520 ] +INFO:root:Average loss=0.10615454882383346 +INFO:root: +[Valid][5][UTT][ MSE = 0.2307 | LCC = 0.9512 | SRCC = 0.9518 ] +INFO:root:Average loss=0.09928559678792953 +INFO:root: +[Valid][6][UTT][ MSE = 0.2465 | LCC = 0.9525 | SRCC = 0.9534 ] +INFO:root:Average loss=0.08830497020483016 +INFO:root: +[Valid][7][UTT][ MSE = 0.2715 | LCC = 0.9513 | SRCC = 0.9533 ] +INFO:root:Average loss=0.08649312034249305 +INFO:root: +[Valid][8][UTT][ MSE = 0.2222 | LCC = 0.9519 | SRCC = 0.9529 ] +INFO:root:Average loss=0.07936748561263085 +INFO:root: +[Valid][9][UTT][ MSE = 0.2698 | LCC = 0.9532 | SRCC = 0.9549 ] +INFO:root:Average loss=0.07632498866319656 +INFO:root: +[Valid][10][UTT][ MSE = 0.2348 | LCC = 0.9526 | SRCC = 0.9534 ] +INFO:root:Average loss=0.06963409769535064 +INFO:root: +[Valid][11][UTT][ MSE = 0.2442 | LCC = 0.9540 | SRCC = 0.9550 ] +INFO:root:Average loss=0.06680078506469726 +INFO:root: +[Valid][12][UTT][ MSE = 0.2255 | LCC = 0.9528 | SRCC = 0.9536 ] +INFO:root:Average loss=0.06187844005227089 +INFO:root: +[Valid][13][UTT][ MSE = 0.2599 | LCC = 0.9525 | SRCC = 0.9540 ] +INFO:root:Average loss=0.0598592137247324 +INFO:root: +[Valid][14][UTT][ MSE = 0.2535 | LCC = 0.9524 | SRCC = 0.9542 ] +INFO:root:Average loss=0.057256523251533506 +INFO:root: +[Valid][15][UTT][ MSE = 0.2582 | LCC = 0.9527 | SRCC = 0.9540 ] +INFO:root:Average loss=0.05472505435347557 +INFO:root: +[Valid][16][UTT][ MSE = 0.2649 | LCC = 0.9531 | SRCC = 0.9543 ] +INFO:root:Average loss=0.051671618938446046 +INFO:root: +[Valid][17][UTT][ MSE = 0.2269 | LCC = 0.9533 | SRCC = 0.9543 ] +INFO:root:Average loss=0.04711918130517006 +INFO:root: +[Valid][18][UTT][ MSE = 0.2385 | LCC = 0.9541 | SRCC = 0.9547 ] +INFO:root:Average loss=0.04472645181417465 +INFO:root: +[Valid][19][UTT][ MSE = 0.2575 | LCC = 0.9528 | SRCC = 0.9542 ] +INFO:root:Average loss=0.044523023918271065 +INFO:root: +[Valid][20][UTT][ MSE = 0.2481 | LCC = 0.9534 | SRCC = 0.9542 ] +INFO:root:Average loss=0.042760106667876246 +INFO:root: +[Valid][21][UTT][ MSE = 0.2346 | LCC = 0.9509 | SRCC = 0.9516 ] +INFO:root:Average loss=0.04323615960776806 +INFO:root: +[Valid][22][UTT][ MSE = 0.2557 | LCC = 0.9519 | SRCC = 0.9537 ] +INFO:root:Average loss=0.03934671029448509 +INFO:root: +[Valid][23][UTT][ MSE = 0.2347 | LCC = 0.9539 | SRCC = 0.9547 ] +INFO:root:Average loss=0.037562045320868494 +INFO:root: +[Valid][24][UTT][ MSE = 0.2395 | LCC = 0.9524 | SRCC = 0.9535 ] +INFO:root:Average loss=0.03727534034848213 +INFO:root: +[Valid][25][UTT][ MSE = 0.2484 | LCC = 0.9543 | SRCC = 0.9551 ] +INFO:root:Average loss=0.03384411442279816 +INFO:root: +[Valid][26][UTT][ MSE = 0.2369 | LCC = 0.9540 | SRCC = 0.9548 ] +INFO:root:Average loss=0.03279544846713543 +INFO:root: +[Valid][27][UTT][ MSE = 0.2578 | LCC = 0.9534 | SRCC = 0.9551 ] +INFO:root:Average loss=0.03329531566798687 +INFO:root: +[Valid][28][UTT][ MSE = 0.2332 | LCC = 0.9547 | SRCC = 0.9556 ] +INFO:root:Average loss=0.03138474352657795 +INFO:root: +[Valid][29][UTT][ MSE = 0.2406 | LCC = 0.9548 | SRCC = 0.9555 ] +INFO:root:Average loss=0.029529418885707856 +INFO:root: +[Valid][30][UTT][ MSE = 0.2573 | LCC = 0.9539 | SRCC = 0.9555 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2473 | LCC = 0.9577 | SRCC = 0.9591 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5bc87abcf1ed563641222fe0a687720174e62939 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f3972b9a3d1d6c95aff058dafe1b3a196b514482bc5d2c701b81ef2cbbf97240 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c1cfa76adeee90d0ded8785024828c7bd0d96f91 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9721d552bce1f6414d5f706f943035ce88f6466b47be13f7002e3e657ec3b494 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5f148680a8870e5b49e79f36d4311df5943d3df3 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.369248903632164 +INFO:root: +[Valid][0][UTT][ MSE = 0.2810 | LCC = 0.9269 | SRCC = 0.9287 ] +INFO:root:Average loss=0.18397222328186036 +INFO:root: +[Valid][1][UTT][ MSE = 0.2240 | LCC = 0.9365 | SRCC = 0.9384 ] +INFO:root:Average loss=0.152589022397995 +INFO:root: +[Valid][2][UTT][ MSE = 0.2100 | LCC = 0.9458 | SRCC = 0.9474 ] +INFO:root:Average loss=0.1398293048143387 +INFO:root: +[Valid][3][UTT][ MSE = 0.2268 | LCC = 0.9465 | SRCC = 0.9487 ] +INFO:root:Average loss=0.12051390135288238 +INFO:root: +[Valid][4][UTT][ MSE = 0.2327 | LCC = 0.9475 | SRCC = 0.9489 ] +INFO:root:Average loss=0.11405166977643967 +INFO:root: +[Valid][5][UTT][ MSE = 0.2128 | LCC = 0.9472 | SRCC = 0.9490 ] +INFO:root:Average loss=0.1040172860622406 +INFO:root: +[Valid][6][UTT][ MSE = 0.2227 | LCC = 0.9529 | SRCC = 0.9538 ] +INFO:root:Average loss=0.0944249067902565 +INFO:root: +[Valid][7][UTT][ MSE = 0.2047 | LCC = 0.9509 | SRCC = 0.9520 ] +INFO:root:Average loss=0.08935329806804657 +INFO:root: +[Valid][8][UTT][ MSE = 0.2050 | LCC = 0.9525 | SRCC = 0.9541 ] +INFO:root:Average loss=0.08262434396147728 +INFO:root: +[Valid][9][UTT][ MSE = 0.1904 | LCC = 0.9513 | SRCC = 0.9531 ] +INFO:root:Average loss=0.0757561129629612 +INFO:root: +[Valid][10][UTT][ MSE = 0.2067 | LCC = 0.9509 | SRCC = 0.9522 ] +INFO:root:Average loss=0.07185161802172661 +INFO:root: +[Valid][11][UTT][ MSE = 0.1936 | LCC = 0.9493 | SRCC = 0.9506 ] +INFO:root:Average loss=0.07015671774744987 +INFO:root: +[Valid][12][UTT][ MSE = 0.1884 | LCC = 0.9536 | SRCC = 0.9549 ] +INFO:root:Average loss=0.06732827401161194 +INFO:root: +[Valid][13][UTT][ MSE = 0.2384 | LCC = 0.9506 | SRCC = 0.9515 ] +INFO:root:Average loss=0.06046520522236824 +INFO:root: +[Valid][14][UTT][ MSE = 0.1886 | LCC = 0.9518 | SRCC = 0.9524 ] +INFO:root:Average loss=0.059767757803201677 +INFO:root: +[Valid][15][UTT][ MSE = 0.1961 | LCC = 0.9540 | SRCC = 0.9552 ] +INFO:root:Average loss=0.058068855732679364 +INFO:root: +[Valid][16][UTT][ MSE = 0.1999 | LCC = 0.9495 | SRCC = 0.9513 ] +INFO:root:Average loss=0.054335319578647616 +INFO:root: +[Valid][17][UTT][ MSE = 0.1834 | LCC = 0.9525 | SRCC = 0.9540 ] +INFO:root:Average loss=0.048712417170405385 +INFO:root: +[Valid][18][UTT][ MSE = 0.2072 | LCC = 0.9511 | SRCC = 0.9522 ] +INFO:root:Average loss=0.04720996604859829 +INFO:root: +[Valid][19][UTT][ MSE = 0.1928 | LCC = 0.9510 | SRCC = 0.9523 ] +INFO:root:Average loss=0.045316089004278186 +INFO:root: +[Valid][20][UTT][ MSE = 0.1942 | LCC = 0.9508 | SRCC = 0.9521 ] +INFO:root:Average loss=0.04241244073212147 +INFO:root: +[Valid][21][UTT][ MSE = 0.1898 | LCC = 0.9514 | SRCC = 0.9529 ] +INFO:root:Average loss=0.04081263878941536 +INFO:root: +[Valid][22][UTT][ MSE = 0.1922 | LCC = 0.9520 | SRCC = 0.9532 ] +INFO:root:Average loss=0.04046694707870484 +INFO:root: +[Valid][23][UTT][ MSE = 0.2089 | LCC = 0.9530 | SRCC = 0.9542 ] +INFO:root:Average loss=0.03880376835167408 +INFO:root: +[Valid][24][UTT][ MSE = 0.2061 | LCC = 0.9540 | SRCC = 0.9549 ] +INFO:root:Average loss=0.03696328681707382 +INFO:root: +[Valid][25][UTT][ MSE = 0.2046 | LCC = 0.9535 | SRCC = 0.9549 ] +INFO:root:Average loss=0.03547971074283123 +INFO:root: +[Valid][26][UTT][ MSE = 0.2105 | LCC = 0.9510 | SRCC = 0.9527 ] +INFO:root:Average loss=0.035475707665085794 +INFO:root: +[Valid][27][UTT][ MSE = 0.2092 | LCC = 0.9525 | SRCC = 0.9537 ] +INFO:root:Average loss=0.034734173595905304 +INFO:root: +[Valid][28][UTT][ MSE = 0.2169 | LCC = 0.9533 | SRCC = 0.9544 ] +INFO:root:Average loss=0.03301887421309948 +INFO:root: +[Valid][29][UTT][ MSE = 0.2032 | LCC = 0.9507 | SRCC = 0.9519 ] +INFO:root:Average loss=0.030127456203103066 +INFO:root: +[Valid][30][UTT][ MSE = 0.2041 | LCC = 0.9535 | SRCC = 0.9547 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1959 | LCC = 0.9583 | SRCC = 0.9591 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e7c3e01299a7c4b7129efa15e0430a3f397379c5 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:26bbe3bd4a95c7750be03cc044eef173b2f4855115ba12fd215b93f1dc849d4f +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..315a5d556e3c2ef157b79aab59ef39b75eeee336 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5c6a9ac0021e4b148df3050e3fac333e335a6ac3f6b4427988784885bf653ee7 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6fd4a78ce9720007aa273749dafc7e2d26236c35 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.48805931985378265 +INFO:root: +[Valid][0][UTT][ MSE = 0.3097 | LCC = 0.9229 | SRCC = 0.9224 ] +INFO:root:Average loss=0.18133438169956206 +INFO:root: +[Valid][1][UTT][ MSE = 0.2798 | LCC = 0.9406 | SRCC = 0.9416 ] +INFO:root:Average loss=0.1472295544743538 +INFO:root: +[Valid][2][UTT][ MSE = 0.2510 | LCC = 0.9473 | SRCC = 0.9485 ] +INFO:root:Average loss=0.13458774268627166 +INFO:root: +[Valid][3][UTT][ MSE = 0.2297 | LCC = 0.9485 | SRCC = 0.9508 ] +INFO:root:Average loss=0.12333846706151962 +INFO:root: +[Valid][4][UTT][ MSE = 0.2056 | LCC = 0.9474 | SRCC = 0.9489 ] +INFO:root:Average loss=0.10922713258862496 +INFO:root: +[Valid][5][UTT][ MSE = 0.2144 | LCC = 0.9520 | SRCC = 0.9530 ] +INFO:root:Average loss=0.10505904507637023 +INFO:root: +[Valid][6][UTT][ MSE = 0.2555 | LCC = 0.9489 | SRCC = 0.9502 ] +INFO:root:Average loss=0.09170190846920014 +INFO:root: +[Valid][7][UTT][ MSE = 0.2122 | LCC = 0.9529 | SRCC = 0.9534 ] +INFO:root:Average loss=0.08708799782395363 +INFO:root: +[Valid][8][UTT][ MSE = 0.2240 | LCC = 0.9520 | SRCC = 0.9535 ] +INFO:root:Average loss=0.0835012000799179 +INFO:root: +[Valid][9][UTT][ MSE = 0.1956 | LCC = 0.9532 | SRCC = 0.9539 ] +INFO:root:Average loss=0.07261229673027993 +INFO:root: +[Valid][10][UTT][ MSE = 0.2083 | LCC = 0.9539 | SRCC = 0.9541 ] +INFO:root:Average loss=0.06937389555573463 +INFO:root: +[Valid][11][UTT][ MSE = 0.2237 | LCC = 0.9535 | SRCC = 0.9543 ] +INFO:root:Average loss=0.06645282125473022 +INFO:root: +[Valid][12][UTT][ MSE = 0.2086 | LCC = 0.9520 | SRCC = 0.9528 ] +INFO:root:Average loss=0.06118219283223152 +INFO:root: +[Valid][13][UTT][ MSE = 0.2117 | LCC = 0.9501 | SRCC = 0.9514 ] +INFO:root:Average loss=0.05830035239458084 +INFO:root: +[Valid][14][UTT][ MSE = 0.2076 | LCC = 0.9522 | SRCC = 0.9528 ] +INFO:root:Average loss=0.05649072992801666 +INFO:root: +[Valid][15][UTT][ MSE = 0.1916 | LCC = 0.9536 | SRCC = 0.9538 ] +INFO:root:Average loss=0.05413280439376831 +INFO:root: +[Valid][16][UTT][ MSE = 0.2092 | LCC = 0.9532 | SRCC = 0.9540 ] +INFO:root:Average loss=0.051900198876857755 +INFO:root: +[Valid][17][UTT][ MSE = 0.2112 | LCC = 0.9520 | SRCC = 0.9533 ] +INFO:root:Average loss=0.04774604474008083 +INFO:root: +[Valid][18][UTT][ MSE = 0.2144 | LCC = 0.9537 | SRCC = 0.9542 ] +INFO:root:Average loss=0.04609258005023003 +INFO:root: +[Valid][19][UTT][ MSE = 0.2085 | LCC = 0.9511 | SRCC = 0.9516 ] +INFO:root:Average loss=0.04358355037868023 +INFO:root: +[Valid][20][UTT][ MSE = 0.1905 | LCC = 0.9531 | SRCC = 0.9534 ] +INFO:root:Average loss=0.04134274852275848 +INFO:root: +[Valid][21][UTT][ MSE = 0.2021 | LCC = 0.9532 | SRCC = 0.9536 ] +INFO:root:Average loss=0.03902082338929176 +INFO:root: +[Valid][22][UTT][ MSE = 0.2181 | LCC = 0.9524 | SRCC = 0.9530 ] +INFO:root:Average loss=0.03824545119702816 +INFO:root: +[Valid][23][UTT][ MSE = 0.1801 | LCC = 0.9524 | SRCC = 0.9537 ] +INFO:root:Average loss=0.03761099833250046 +INFO:root: +[Valid][24][UTT][ MSE = 0.2116 | LCC = 0.9529 | SRCC = 0.9536 ] +INFO:root:Average loss=0.03532977657020092 +INFO:root: +[Valid][25][UTT][ MSE = 0.1930 | LCC = 0.9549 | SRCC = 0.9551 ] +INFO:root:Average loss=0.03290769663453102 +INFO:root: +[Valid][26][UTT][ MSE = 0.2040 | LCC = 0.9542 | SRCC = 0.9548 ] +INFO:root:Average loss=0.03327473025023937 +INFO:root: +[Valid][27][UTT][ MSE = 0.1923 | LCC = 0.9527 | SRCC = 0.9535 ] +INFO:root:Average loss=0.03246244192123413 +INFO:root: +[Valid][28][UTT][ MSE = 0.2100 | LCC = 0.9536 | SRCC = 0.9543 ] +INFO:root:Average loss=0.030406089201569557 +INFO:root: +[Valid][29][UTT][ MSE = 0.1907 | LCC = 0.9520 | SRCC = 0.9528 ] +INFO:root:Average loss=0.028475490659475328 +INFO:root: +[Valid][30][UTT][ MSE = 0.1915 | LCC = 0.9536 | SRCC = 0.9538 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1938 | LCC = 0.9554 | SRCC = 0.9565 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9a6a25c7dbc68ac03962acd1ff4b0f55ff35608a --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e7fb9df62b489ccce7cf119f4e3515ff87420d42c9f15a3bf94e146e72da558d +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..57037695d696ffdad38505af0b80dfcf73e8db3d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:88b752475dd0a230984b35d2fb2cf8a07c2d791bceaeb0e619a46335caa4cc91 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..19329225382a82a87515503c2e150361d27eb698 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46284975373744963 +INFO:root: +[Valid][0][UTT][ MSE = 0.2791 | LCC = 0.9178 | SRCC = 0.9200 ] +INFO:root:Average loss=0.1828877449631691 +INFO:root: +[Valid][1][UTT][ MSE = 0.2227 | LCC = 0.9395 | SRCC = 0.9423 ] +INFO:root:Average loss=0.15283601993322374 +INFO:root: +[Valid][2][UTT][ MSE = 0.2231 | LCC = 0.9441 | SRCC = 0.9458 ] +INFO:root:Average loss=0.13687439718842506 +INFO:root: +[Valid][3][UTT][ MSE = 0.2084 | LCC = 0.9465 | SRCC = 0.9490 ] +INFO:root:Average loss=0.11953235167264939 +INFO:root: +[Valid][4][UTT][ MSE = 0.2301 | LCC = 0.9492 | SRCC = 0.9509 ] +INFO:root:Average loss=0.10996338725090027 +INFO:root: +[Valid][5][UTT][ MSE = 0.2082 | LCC = 0.9477 | SRCC = 0.9496 ] +INFO:root:Average loss=0.10044216448068619 +INFO:root: +[Valid][6][UTT][ MSE = 0.2150 | LCC = 0.9488 | SRCC = 0.9508 ] +INFO:root:Average loss=0.09626852771639824 +INFO:root: +[Valid][7][UTT][ MSE = 0.2256 | LCC = 0.9518 | SRCC = 0.9533 ] +INFO:root:Average loss=0.08527649706602096 +INFO:root: +[Valid][8][UTT][ MSE = 0.2175 | LCC = 0.9512 | SRCC = 0.9526 ] +INFO:root:Average loss=0.08124754863977432 +INFO:root: +[Valid][9][UTT][ MSE = 0.2183 | LCC = 0.9486 | SRCC = 0.9503 ] +INFO:root:Average loss=0.07634615305066109 +INFO:root: +[Valid][10][UTT][ MSE = 0.2031 | LCC = 0.9506 | SRCC = 0.9519 ] +INFO:root:Average loss=0.07074149543046951 +INFO:root: +[Valid][11][UTT][ MSE = 0.2011 | LCC = 0.9501 | SRCC = 0.9516 ] +INFO:root:Average loss=0.0670567201077938 +INFO:root: +[Valid][12][UTT][ MSE = 0.1934 | LCC = 0.9490 | SRCC = 0.9500 ] +INFO:root:Average loss=0.06379567980766296 +INFO:root: +[Valid][13][UTT][ MSE = 0.2246 | LCC = 0.9478 | SRCC = 0.9491 ] +INFO:root:Average loss=0.06191250002384186 +INFO:root: +[Valid][14][UTT][ MSE = 0.2030 | LCC = 0.9508 | SRCC = 0.9516 ] +INFO:root:Average loss=0.05946513599157333 +INFO:root: +[Valid][15][UTT][ MSE = 0.2013 | LCC = 0.9508 | SRCC = 0.9510 ] +INFO:root:Average loss=0.05527076295018196 +INFO:root: +[Valid][16][UTT][ MSE = 0.2209 | LCC = 0.9454 | SRCC = 0.9467 ] +INFO:root:Average loss=0.053272455602884294 +INFO:root: +[Valid][17][UTT][ MSE = 0.2162 | LCC = 0.9527 | SRCC = 0.9535 ] +INFO:root:Average loss=0.04818961195647717 +INFO:root: +[Valid][18][UTT][ MSE = 0.2080 | LCC = 0.9467 | SRCC = 0.9472 ] +INFO:root:Average loss=0.04619884313642979 +INFO:root: +[Valid][19][UTT][ MSE = 0.1865 | LCC = 0.9468 | SRCC = 0.9482 ] +INFO:root:Average loss=0.042913806319236754 +INFO:root: +[Valid][20][UTT][ MSE = 0.2357 | LCC = 0.9493 | SRCC = 0.9505 ] +INFO:root:Average loss=0.046159168884158136 +INFO:root: +[Valid][21][UTT][ MSE = 0.2028 | LCC = 0.9510 | SRCC = 0.9517 ] +INFO:root:Average loss=0.03895376299321651 +INFO:root: +[Valid][22][UTT][ MSE = 0.1917 | LCC = 0.9519 | SRCC = 0.9527 ] +INFO:root:Average loss=0.03894115829467774 +INFO:root: +[Valid][23][UTT][ MSE = 0.2044 | LCC = 0.9510 | SRCC = 0.9516 ] +INFO:root:Average loss=0.0383162659406662 +INFO:root: +[Valid][24][UTT][ MSE = 0.2016 | LCC = 0.9524 | SRCC = 0.9532 ] +INFO:root:Average loss=0.03624134257435799 +INFO:root: +[Valid][25][UTT][ MSE = 0.1857 | LCC = 0.9490 | SRCC = 0.9503 ] +INFO:root:Average loss=0.0355315427929163 +INFO:root: +[Valid][26][UTT][ MSE = 0.2084 | LCC = 0.9493 | SRCC = 0.9503 ] +INFO:root:Average loss=0.03530453124642372 +INFO:root: +[Valid][27][UTT][ MSE = 0.1958 | LCC = 0.9496 | SRCC = 0.9505 ] +INFO:root:Average loss=0.031497067958116534 +INFO:root: +[Valid][28][UTT][ MSE = 0.2008 | LCC = 0.9507 | SRCC = 0.9515 ] +INFO:root:Average loss=0.031173504129052163 +INFO:root: +[Valid][29][UTT][ MSE = 0.2024 | LCC = 0.9524 | SRCC = 0.9538 ] +INFO:root:Average loss=0.031260938912630085 +INFO:root: +[Valid][30][UTT][ MSE = 0.2073 | LCC = 0.9525 | SRCC = 0.9529 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2187 | LCC = 0.9579 | SRCC = 0.9589 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c17ab0430aa98694e92a5972842800fa0ec34afd --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f1a06789422ea38584bbef60fbf90d6550cc4c0d861307126c660efdf333d1b7 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..51954be5ca2a9ab4e5ad8036ebc0b9bf4181674e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eea561f2080c8b45635b346f12d5bb8b70af2a3438a295c577d4a0451647f547 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f7ae6c7d5d712d55b78fa7a972a2e206afdf4c7a --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer5/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3616385979652405 +INFO:root: +[Valid][0][UTT][ MSE = 0.2633 | LCC = 0.9279 | SRCC = 0.9314 ] +INFO:root:Average loss=0.17157283937931062 +INFO:root: +[Valid][1][UTT][ MSE = 0.2547 | LCC = 0.9429 | SRCC = 0.9449 ] +INFO:root:Average loss=0.14582651150226594 +INFO:root: +[Valid][2][UTT][ MSE = 0.2324 | LCC = 0.9472 | SRCC = 0.9486 ] +INFO:root:Average loss=0.1323048641681671 +INFO:root: +[Valid][3][UTT][ MSE = 0.2381 | LCC = 0.9426 | SRCC = 0.9429 ] +INFO:root:Average loss=0.11814508455991744 +INFO:root: +[Valid][4][UTT][ MSE = 0.2008 | LCC = 0.9513 | SRCC = 0.9529 ] +INFO:root:Average loss=0.10650392565131188 +INFO:root: +[Valid][5][UTT][ MSE = 0.2018 | LCC = 0.9507 | SRCC = 0.9515 ] +INFO:root:Average loss=0.09977437129616737 +INFO:root: +[Valid][6][UTT][ MSE = 0.2225 | LCC = 0.9507 | SRCC = 0.9521 ] +INFO:root:Average loss=0.09212336778640746 +INFO:root: +[Valid][7][UTT][ MSE = 0.2194 | LCC = 0.9518 | SRCC = 0.9524 ] +INFO:root:Average loss=0.08902364620566368 +INFO:root: +[Valid][8][UTT][ MSE = 0.1932 | LCC = 0.9534 | SRCC = 0.9545 ] +INFO:root:Average loss=0.08263257643580436 +INFO:root: +[Valid][9][UTT][ MSE = 0.2433 | LCC = 0.9521 | SRCC = 0.9527 ] +INFO:root:Average loss=0.07425819164514541 +INFO:root: +[Valid][10][UTT][ MSE = 0.2056 | LCC = 0.9539 | SRCC = 0.9548 ] +INFO:root:Average loss=0.06894444349408149 +INFO:root: +[Valid][11][UTT][ MSE = 0.1767 | LCC = 0.9538 | SRCC = 0.9551 ] +INFO:root:Average loss=0.06382294717431068 +INFO:root: +[Valid][12][UTT][ MSE = 0.1894 | LCC = 0.9527 | SRCC = 0.9532 ] +INFO:root:Average loss=0.06422997677326202 +INFO:root: +[Valid][13][UTT][ MSE = 0.1819 | LCC = 0.9541 | SRCC = 0.9555 ] +INFO:root:Average loss=0.057980885088443755 +INFO:root: +[Valid][14][UTT][ MSE = 0.1798 | LCC = 0.9536 | SRCC = 0.9542 ] +INFO:root:Average loss=0.05523533537983894 +INFO:root: +[Valid][15][UTT][ MSE = 0.2163 | LCC = 0.9500 | SRCC = 0.9507 ] +INFO:root:Average loss=0.05486509470641613 +INFO:root: +[Valid][16][UTT][ MSE = 0.1929 | LCC = 0.9530 | SRCC = 0.9539 ] +INFO:root:Average loss=0.049399322777986525 +INFO:root: +[Valid][17][UTT][ MSE = 0.1833 | LCC = 0.9544 | SRCC = 0.9549 ] +INFO:root:Average loss=0.048783723205327985 +INFO:root: +[Valid][18][UTT][ MSE = 0.1986 | LCC = 0.9534 | SRCC = 0.9541 ] +INFO:root:Average loss=0.04582230693101883 +INFO:root: +[Valid][19][UTT][ MSE = 0.1922 | LCC = 0.9529 | SRCC = 0.9536 ] +INFO:root:Average loss=0.042277437672019004 +INFO:root: +[Valid][20][UTT][ MSE = 0.1869 | LCC = 0.9535 | SRCC = 0.9541 ] +INFO:root:Average loss=0.041624211013317106 +INFO:root: +[Valid][21][UTT][ MSE = 0.1926 | LCC = 0.9540 | SRCC = 0.9548 ] +INFO:root:Average loss=0.04017722791433334 +INFO:root: +[Valid][22][UTT][ MSE = 0.1844 | LCC = 0.9537 | SRCC = 0.9538 ] +INFO:root:Average loss=0.037837415516376496 +INFO:root: +[Valid][23][UTT][ MSE = 0.1757 | LCC = 0.9541 | SRCC = 0.9548 ] +INFO:root:Average loss=0.038683407217264175 +INFO:root: +[Valid][24][UTT][ MSE = 0.1932 | LCC = 0.9541 | SRCC = 0.9546 ] +INFO:root:Average loss=0.035947467654943464 +INFO:root: +[Valid][25][UTT][ MSE = 0.2007 | LCC = 0.9525 | SRCC = 0.9529 ] +INFO:root:Average loss=0.03526121042668819 +INFO:root: +[Valid][26][UTT][ MSE = 0.1849 | LCC = 0.9532 | SRCC = 0.9537 ] +INFO:root:Average loss=0.03247630442678928 +INFO:root: +[Valid][27][UTT][ MSE = 0.1955 | LCC = 0.9514 | SRCC = 0.9519 ] +INFO:root:Average loss=0.03202079455554485 +INFO:root: +[Valid][28][UTT][ MSE = 0.1853 | LCC = 0.9529 | SRCC = 0.9534 ] +INFO:root:Average loss=0.0336299863755703 +INFO:root: +[Valid][29][UTT][ MSE = 0.2108 | LCC = 0.9517 | SRCC = 0.9518 ] +INFO:root:Average loss=0.03128068970143795 +INFO:root: +[Valid][30][UTT][ MSE = 0.2001 | LCC = 0.9533 | SRCC = 0.9537 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1821 | LCC = 0.9572 | SRCC = 0.9582 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..445d17f657a310999c02c6a0e02f23dc0431bf9d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ede49667d782e96328a45e2e69c3da78734a6193ba3c4529653dba81a37f07db +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..298238c46731059cd37eef7338bbb9727637e6dc --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c30cb28e35e1aad3d076d98cc4c4a8e2a459de99d94029de61e1f6fff9e97b05 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d34858b83dd9138f77616b0a8a05142c552d5107 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4456357583999634 +INFO:root: +[Valid][0][UTT][ MSE = 0.3366 | LCC = 0.9219 | SRCC = 0.9234 ] +INFO:root:Average loss=0.19182163363695146 +INFO:root: +[Valid][1][UTT][ MSE = 0.2543 | LCC = 0.9366 | SRCC = 0.9394 ] +INFO:root:Average loss=0.15322915363311768 +INFO:root: +[Valid][2][UTT][ MSE = 0.2509 | LCC = 0.9429 | SRCC = 0.9445 ] +INFO:root:Average loss=0.13887556171417237 +INFO:root: +[Valid][3][UTT][ MSE = 0.2583 | LCC = 0.9407 | SRCC = 0.9422 ] +INFO:root:Average loss=0.1253982278704643 +INFO:root: +[Valid][4][UTT][ MSE = 0.2471 | LCC = 0.9468 | SRCC = 0.9482 ] +INFO:root:Average loss=0.11375288835167885 +INFO:root: +[Valid][5][UTT][ MSE = 0.2335 | LCC = 0.9473 | SRCC = 0.9485 ] +INFO:root:Average loss=0.10303991869091987 +INFO:root: +[Valid][6][UTT][ MSE = 0.2486 | LCC = 0.9472 | SRCC = 0.9491 ] +INFO:root:Average loss=0.10141887751221657 +INFO:root: +[Valid][7][UTT][ MSE = 0.2165 | LCC = 0.9494 | SRCC = 0.9508 ] +INFO:root:Average loss=0.09323137930035591 +INFO:root: +[Valid][8][UTT][ MSE = 0.2426 | LCC = 0.9504 | SRCC = 0.9515 ] +INFO:root:Average loss=0.08241498455405236 +INFO:root: +[Valid][9][UTT][ MSE = 0.2451 | LCC = 0.9503 | SRCC = 0.9515 ] +INFO:root:Average loss=0.07972859928011894 +INFO:root: +[Valid][10][UTT][ MSE = 0.2505 | LCC = 0.9514 | SRCC = 0.9525 ] +INFO:root:Average loss=0.07444022226333619 +INFO:root: +[Valid][11][UTT][ MSE = 0.2300 | LCC = 0.9518 | SRCC = 0.9526 ] +INFO:root:Average loss=0.07145227181911469 +INFO:root: +[Valid][12][UTT][ MSE = 0.2456 | LCC = 0.9517 | SRCC = 0.9532 ] +INFO:root:Average loss=0.0693144624531269 +INFO:root: +[Valid][13][UTT][ MSE = 0.2305 | LCC = 0.9492 | SRCC = 0.9506 ] +INFO:root:Average loss=0.0628601356446743 +INFO:root: +[Valid][14][UTT][ MSE = 0.2486 | LCC = 0.9507 | SRCC = 0.9522 ] +INFO:root:Average loss=0.05854415595531463 +INFO:root: +[Valid][15][UTT][ MSE = 0.2087 | LCC = 0.9494 | SRCC = 0.9500 ] +INFO:root:Average loss=0.05345706906914711 +INFO:root: +[Valid][16][UTT][ MSE = 0.2309 | LCC = 0.9506 | SRCC = 0.9519 ] +INFO:root:Average loss=0.05431419748067856 +INFO:root: +[Valid][17][UTT][ MSE = 0.2135 | LCC = 0.9493 | SRCC = 0.9500 ] +INFO:root:Average loss=0.05180395565927029 +INFO:root: +[Valid][18][UTT][ MSE = 0.2295 | LCC = 0.9509 | SRCC = 0.9519 ] +INFO:root:Average loss=0.05229282483458519 +INFO:root: +[Valid][19][UTT][ MSE = 0.2172 | LCC = 0.9507 | SRCC = 0.9515 ] +INFO:root:Average loss=0.04534087853133678 +INFO:root: +[Valid][20][UTT][ MSE = 0.2192 | LCC = 0.9484 | SRCC = 0.9492 ] +INFO:root:Average loss=0.043073117151856424 +INFO:root: +[Valid][21][UTT][ MSE = 0.2197 | LCC = 0.9501 | SRCC = 0.9510 ] +INFO:root:Average loss=0.04301554125547409 +INFO:root: +[Valid][22][UTT][ MSE = 0.2234 | LCC = 0.9509 | SRCC = 0.9516 ] +INFO:root:Average loss=0.04051660239696503 +INFO:root: +[Valid][23][UTT][ MSE = 0.2506 | LCC = 0.9504 | SRCC = 0.9515 ] +INFO:root:Average loss=0.04221714209020138 +INFO:root: +[Valid][24][UTT][ MSE = 0.2416 | LCC = 0.9493 | SRCC = 0.9501 ] +INFO:root:Average loss=0.03693766714632511 +INFO:root: +[Valid][25][UTT][ MSE = 0.2267 | LCC = 0.9487 | SRCC = 0.9496 ] +INFO:root:Average loss=0.035482014685869215 +INFO:root: +[Valid][26][UTT][ MSE = 0.2415 | LCC = 0.9491 | SRCC = 0.9499 ] +INFO:root:Average loss=0.03434496283531189 +INFO:root: +[Valid][27][UTT][ MSE = 0.2173 | LCC = 0.9506 | SRCC = 0.9514 ] +INFO:root:Average loss=0.033162279784679415 +INFO:root: +[Valid][28][UTT][ MSE = 0.2165 | LCC = 0.9507 | SRCC = 0.9510 ] +INFO:root:Average loss=0.03141732768714428 +INFO:root: +[Valid][29][UTT][ MSE = 0.2264 | LCC = 0.9510 | SRCC = 0.9516 ] +INFO:root:Average loss=0.03087141777575016 +INFO:root: +[Valid][30][UTT][ MSE = 0.2319 | LCC = 0.9486 | SRCC = 0.9493 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2426 | LCC = 0.9519 | SRCC = 0.9534 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6bf0f53c3a67788f9cf953ceee9bee19386ab8b4 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5dd45bb334e419160e71ea6b86d05d25dd2e173829c306655f9001d46be71be1 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..075e006216e6acff397873dbd0617811b1eda808 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f71e36ad04e1e3a83e60a12a87f3a6fabd4c0cce74f9dff8bac8b1c56b53fe0f +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..78f121aea7bda167eaac85aa6a21ac6a29950193 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4186544963121414 +INFO:root: +[Valid][0][UTT][ MSE = 0.3182 | LCC = 0.9124 | SRCC = 0.9165 ] +INFO:root:Average loss=0.19035791635513305 +INFO:root: +[Valid][1][UTT][ MSE = 0.2982 | LCC = 0.9378 | SRCC = 0.9380 ] +INFO:root:Average loss=0.15223421376943588 +INFO:root: +[Valid][2][UTT][ MSE = 0.2628 | LCC = 0.9401 | SRCC = 0.9409 ] +INFO:root:Average loss=0.1356086004972458 +INFO:root: +[Valid][3][UTT][ MSE = 0.2519 | LCC = 0.9435 | SRCC = 0.9453 ] +INFO:root:Average loss=0.12433120292425155 +INFO:root: +[Valid][4][UTT][ MSE = 0.2220 | LCC = 0.9459 | SRCC = 0.9481 ] +INFO:root:Average loss=0.11714754921197891 +INFO:root: +[Valid][5][UTT][ MSE = 0.2411 | LCC = 0.9500 | SRCC = 0.9514 ] +INFO:root:Average loss=0.10445461076498032 +INFO:root: +[Valid][6][UTT][ MSE = 0.2214 | LCC = 0.9471 | SRCC = 0.9495 ] +INFO:root:Average loss=0.09652287983894348 +INFO:root: +[Valid][7][UTT][ MSE = 0.2050 | LCC = 0.9483 | SRCC = 0.9503 ] +INFO:root:Average loss=0.09238114169239998 +INFO:root: +[Valid][8][UTT][ MSE = 0.1924 | LCC = 0.9504 | SRCC = 0.9519 ] +INFO:root:Average loss=0.08335401302576065 +INFO:root: +[Valid][9][UTT][ MSE = 0.1947 | LCC = 0.9510 | SRCC = 0.9524 ] +INFO:root:Average loss=0.08021691787242889 +INFO:root: +[Valid][10][UTT][ MSE = 0.1918 | LCC = 0.9521 | SRCC = 0.9538 ] +INFO:root:Average loss=0.0773236910700798 +INFO:root: +[Valid][11][UTT][ MSE = 0.2353 | LCC = 0.9499 | SRCC = 0.9515 ] +INFO:root:Average loss=0.07108905449509621 +INFO:root: +[Valid][12][UTT][ MSE = 0.2184 | LCC = 0.9520 | SRCC = 0.9526 ] +INFO:root:Average loss=0.06602118653059005 +INFO:root: +[Valid][13][UTT][ MSE = 0.1935 | LCC = 0.9513 | SRCC = 0.9522 ] +INFO:root:Average loss=0.06355162400007248 +INFO:root: +[Valid][14][UTT][ MSE = 0.1941 | LCC = 0.9509 | SRCC = 0.9525 ] +INFO:root:Average loss=0.06257434576749801 +INFO:root: +[Valid][15][UTT][ MSE = 0.1959 | LCC = 0.9524 | SRCC = 0.9539 ] +INFO:root:Average loss=0.05597138431668282 +INFO:root: +[Valid][16][UTT][ MSE = 0.2010 | LCC = 0.9514 | SRCC = 0.9526 ] +INFO:root:Average loss=0.05496250462532044 +INFO:root: +[Valid][17][UTT][ MSE = 0.1932 | LCC = 0.9490 | SRCC = 0.9507 ] +INFO:root:Average loss=0.05083446238934994 +INFO:root: +[Valid][18][UTT][ MSE = 0.1949 | LCC = 0.9500 | SRCC = 0.9510 ] +INFO:root:Average loss=0.050518397748470306 +INFO:root: +[Valid][19][UTT][ MSE = 0.1921 | LCC = 0.9531 | SRCC = 0.9542 ] +INFO:root:Average loss=0.046143108233809474 +INFO:root: +[Valid][20][UTT][ MSE = 0.2044 | LCC = 0.9524 | SRCC = 0.9532 ] +INFO:root:Average loss=0.04454578506946564 +INFO:root: +[Valid][21][UTT][ MSE = 0.1870 | LCC = 0.9515 | SRCC = 0.9528 ] +INFO:root:Average loss=0.0427032832056284 +INFO:root: +[Valid][22][UTT][ MSE = 0.1929 | LCC = 0.9522 | SRCC = 0.9536 ] +INFO:root:Average loss=0.03944619990885258 +INFO:root: +[Valid][23][UTT][ MSE = 0.2021 | LCC = 0.9503 | SRCC = 0.9509 ] +INFO:root:Average loss=0.037923003137111666 +INFO:root: +[Valid][24][UTT][ MSE = 0.1926 | LCC = 0.9526 | SRCC = 0.9536 ] +INFO:root:Average loss=0.037460220918059346 +INFO:root: +[Valid][25][UTT][ MSE = 0.1974 | LCC = 0.9524 | SRCC = 0.9533 ] +INFO:root:Average loss=0.035568890035152435 +INFO:root: +[Valid][26][UTT][ MSE = 0.1902 | LCC = 0.9532 | SRCC = 0.9540 ] +INFO:root:Average loss=0.0344182386547327 +INFO:root: +[Valid][27][UTT][ MSE = 0.1986 | LCC = 0.9519 | SRCC = 0.9530 ] +INFO:root:Average loss=0.032967503041028974 +INFO:root: +[Valid][28][UTT][ MSE = 0.1905 | LCC = 0.9513 | SRCC = 0.9522 ] +INFO:root:Average loss=0.031432232692837715 +INFO:root: +[Valid][29][UTT][ MSE = 0.1981 | LCC = 0.9527 | SRCC = 0.9539 ] +INFO:root:Average loss=0.030703338488936423 +INFO:root: +[Valid][30][UTT][ MSE = 0.1821 | LCC = 0.9511 | SRCC = 0.9524 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1933 | LCC = 0.9539 | SRCC = 0.9543 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e111fa23d5da185327a9bc4ddff7fe01dd345130 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:26736ec9db902080af6a5a0b8787099a57910b2fcd92fb3f18af37887490d620 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..629027be0539c27150489f5b894e26a80de0fb96 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:18ff7b0fca02de1470d3f0718a4cfddfd64e4c811fd68c78dac8b9f873b57264 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..256e92a78cd2e7e3e24c01939cbda9bb64787025 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.43604246068000796 +INFO:root: +[Valid][0][UTT][ MSE = 0.2881 | LCC = 0.9265 | SRCC = 0.9284 ] +INFO:root:Average loss=0.19872375345230103 +INFO:root: +[Valid][1][UTT][ MSE = 0.2560 | LCC = 0.9380 | SRCC = 0.9390 ] +INFO:root:Average loss=0.1635867993235588 +INFO:root: +[Valid][2][UTT][ MSE = 0.2218 | LCC = 0.9438 | SRCC = 0.9455 ] +INFO:root:Average loss=0.14052477169036864 +INFO:root: +[Valid][3][UTT][ MSE = 0.1987 | LCC = 0.9446 | SRCC = 0.9458 ] +INFO:root:Average loss=0.12674545031785964 +INFO:root: +[Valid][4][UTT][ MSE = 0.2165 | LCC = 0.9464 | SRCC = 0.9483 ] +INFO:root:Average loss=0.11518640843033791 +INFO:root: +[Valid][5][UTT][ MSE = 0.2341 | LCC = 0.9458 | SRCC = 0.9472 ] +INFO:root:Average loss=0.10923742634058 +INFO:root: +[Valid][6][UTT][ MSE = 0.2268 | LCC = 0.9479 | SRCC = 0.9502 ] +INFO:root:Average loss=0.09969328278303147 +INFO:root: +[Valid][7][UTT][ MSE = 0.2302 | LCC = 0.9492 | SRCC = 0.9504 ] +INFO:root:Average loss=0.0918657295703888 +INFO:root: +[Valid][8][UTT][ MSE = 0.1957 | LCC = 0.9483 | SRCC = 0.9498 ] +INFO:root:Average loss=0.09101050764322281 +INFO:root: +[Valid][9][UTT][ MSE = 0.2140 | LCC = 0.9517 | SRCC = 0.9524 ] +INFO:root:Average loss=0.08219499760866165 +INFO:root: +[Valid][10][UTT][ MSE = 0.2177 | LCC = 0.9463 | SRCC = 0.9478 ] +INFO:root:Average loss=0.07848934495449066 +INFO:root: +[Valid][11][UTT][ MSE = 0.1978 | LCC = 0.9517 | SRCC = 0.9525 ] +INFO:root:Average loss=0.07186202335357667 +INFO:root: +[Valid][12][UTT][ MSE = 0.2165 | LCC = 0.9506 | SRCC = 0.9521 ] +INFO:root:Average loss=0.06632604935765267 +INFO:root: +[Valid][13][UTT][ MSE = 0.2159 | LCC = 0.9506 | SRCC = 0.9520 ] +INFO:root:Average loss=0.06605227550864219 +INFO:root: +[Valid][14][UTT][ MSE = 0.1896 | LCC = 0.9507 | SRCC = 0.9525 ] +INFO:root:Average loss=0.06386821231245994 +INFO:root: +[Valid][15][UTT][ MSE = 0.2258 | LCC = 0.9505 | SRCC = 0.9506 ] +INFO:root:Average loss=0.058172690063714984 +INFO:root: +[Valid][16][UTT][ MSE = 0.1928 | LCC = 0.9515 | SRCC = 0.9522 ] +INFO:root:Average loss=0.05428381896018982 +INFO:root: +[Valid][17][UTT][ MSE = 0.2488 | LCC = 0.9501 | SRCC = 0.9514 ] +INFO:root:Average loss=0.05444227029383183 +INFO:root: +[Valid][18][UTT][ MSE = 0.2170 | LCC = 0.9487 | SRCC = 0.9496 ] +INFO:root:Average loss=0.049666144490242 +INFO:root: +[Valid][19][UTT][ MSE = 0.2051 | LCC = 0.9500 | SRCC = 0.9502 ] +INFO:root:Average loss=0.04817500838637352 +INFO:root: +[Valid][20][UTT][ MSE = 0.1930 | LCC = 0.9519 | SRCC = 0.9530 ] +INFO:root:Average loss=0.044870812490582467 +INFO:root: +[Valid][21][UTT][ MSE = 0.1979 | LCC = 0.9489 | SRCC = 0.9506 ] +INFO:root:Average loss=0.04297021123766899 +INFO:root: +[Valid][22][UTT][ MSE = 0.2081 | LCC = 0.9488 | SRCC = 0.9501 ] +INFO:root:Average loss=0.04272268016636371 +INFO:root: +[Valid][23][UTT][ MSE = 0.2144 | LCC = 0.9512 | SRCC = 0.9520 ] +INFO:root:Average loss=0.04190243908762932 +INFO:root: +[Valid][24][UTT][ MSE = 0.2183 | LCC = 0.9520 | SRCC = 0.9522 ] +INFO:root:Average loss=0.03850781100988388 +INFO:root: +[Valid][25][UTT][ MSE = 0.1993 | LCC = 0.9514 | SRCC = 0.9521 ] +INFO:root:Average loss=0.03927642200887203 +INFO:root: +[Valid][26][UTT][ MSE = 0.2217 | LCC = 0.9514 | SRCC = 0.9524 ] +INFO:root:Average loss=0.035800802499055866 +INFO:root: +[Valid][27][UTT][ MSE = 0.2089 | LCC = 0.9480 | SRCC = 0.9489 ] +INFO:root:Average loss=0.03627059356868267 +INFO:root: +[Valid][28][UTT][ MSE = 0.2022 | LCC = 0.9524 | SRCC = 0.9529 ] +INFO:root:Average loss=0.03361176200211048 +INFO:root: +[Valid][29][UTT][ MSE = 0.2003 | LCC = 0.9534 | SRCC = 0.9537 ] +INFO:root:Average loss=0.033783186703920366 +INFO:root: +[Valid][30][UTT][ MSE = 0.2157 | LCC = 0.9518 | SRCC = 0.9523 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2061 | LCC = 0.9536 | SRCC = 0.9545 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2b57f463dbd0a1bec01c313b7715ba291d903038 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3a8bfadc7d416ecb192952dc218b17c533e7f696322e6ac2b6033c962267308a +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2d8e1edc00262a3b6d0c30df7372096221b31346 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fc54b7f7237d5db449f55884e5f2a5e8b94fcb095788b84b51ab59bf98c30e1c +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c30c1b55f1ed37b7336fa8cafe51975b13cb7207 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4516884636878967 +INFO:root: +[Valid][0][UTT][ MSE = 0.3208 | LCC = 0.9281 | SRCC = 0.9297 ] +INFO:root:Average loss=0.1790100499391556 +INFO:root: +[Valid][1][UTT][ MSE = 0.2560 | LCC = 0.9385 | SRCC = 0.9400 ] +INFO:root:Average loss=0.14707104277610777 +INFO:root: +[Valid][2][UTT][ MSE = 0.2183 | LCC = 0.9435 | SRCC = 0.9454 ] +INFO:root:Average loss=0.1302455193400383 +INFO:root: +[Valid][3][UTT][ MSE = 0.2245 | LCC = 0.9478 | SRCC = 0.9496 ] +INFO:root:Average loss=0.12394321233034133 +INFO:root: +[Valid][4][UTT][ MSE = 0.1904 | LCC = 0.9467 | SRCC = 0.9493 ] +INFO:root:Average loss=0.1106733640730381 +INFO:root: +[Valid][5][UTT][ MSE = 0.2373 | LCC = 0.9432 | SRCC = 0.9467 ] +INFO:root:Average loss=0.10051828318834305 +INFO:root: +[Valid][6][UTT][ MSE = 0.2222 | LCC = 0.9513 | SRCC = 0.9524 ] +INFO:root:Average loss=0.09348770266771317 +INFO:root: +[Valid][7][UTT][ MSE = 0.2110 | LCC = 0.9474 | SRCC = 0.9488 ] +INFO:root:Average loss=0.08401586240530014 +INFO:root: +[Valid][8][UTT][ MSE = 0.2469 | LCC = 0.9506 | SRCC = 0.9520 ] +INFO:root:Average loss=0.07991352865099907 +INFO:root: +[Valid][9][UTT][ MSE = 0.2440 | LCC = 0.9453 | SRCC = 0.9473 ] +INFO:root:Average loss=0.07295040953159332 +INFO:root: +[Valid][10][UTT][ MSE = 0.2096 | LCC = 0.9531 | SRCC = 0.9539 ] +INFO:root:Average loss=0.0698698762357235 +INFO:root: +[Valid][11][UTT][ MSE = 0.2114 | LCC = 0.9513 | SRCC = 0.9522 ] +INFO:root:Average loss=0.06538952666521072 +INFO:root: +[Valid][12][UTT][ MSE = 0.2118 | LCC = 0.9532 | SRCC = 0.9542 ] +INFO:root:Average loss=0.06197335204482079 +INFO:root: +[Valid][13][UTT][ MSE = 0.2147 | LCC = 0.9495 | SRCC = 0.9501 ] +INFO:root:Average loss=0.059723654374480246 +INFO:root: +[Valid][14][UTT][ MSE = 0.2080 | LCC = 0.9507 | SRCC = 0.9513 ] +INFO:root:Average loss=0.05415319401025772 +INFO:root: +[Valid][15][UTT][ MSE = 0.1997 | LCC = 0.9532 | SRCC = 0.9536 ] +INFO:root:Average loss=0.05090158009529114 +INFO:root: +[Valid][16][UTT][ MSE = 0.2099 | LCC = 0.9535 | SRCC = 0.9537 ] +INFO:root:Average loss=0.051486355543136596 +INFO:root: +[Valid][17][UTT][ MSE = 0.1945 | LCC = 0.9494 | SRCC = 0.9503 ] +INFO:root:Average loss=0.048547496870160106 +INFO:root: +[Valid][18][UTT][ MSE = 0.2294 | LCC = 0.9526 | SRCC = 0.9539 ] +INFO:root:Average loss=0.04814401641488075 +INFO:root: +[Valid][19][UTT][ MSE = 0.2142 | LCC = 0.9535 | SRCC = 0.9543 ] +INFO:root:Average loss=0.04228952641785145 +INFO:root: +[Valid][20][UTT][ MSE = 0.2131 | LCC = 0.9532 | SRCC = 0.9538 ] +INFO:root:Average loss=0.040211720421910284 +INFO:root: +[Valid][21][UTT][ MSE = 0.2176 | LCC = 0.9503 | SRCC = 0.9509 ] +INFO:root:Average loss=0.03999157133698463 +INFO:root: +[Valid][22][UTT][ MSE = 0.2129 | LCC = 0.9510 | SRCC = 0.9521 ] +INFO:root:Average loss=0.03936823189258575 +INFO:root: +[Valid][23][UTT][ MSE = 0.1935 | LCC = 0.9519 | SRCC = 0.9524 ] +INFO:root:Average loss=0.03915000906586647 +INFO:root: +[Valid][24][UTT][ MSE = 0.1906 | LCC = 0.9524 | SRCC = 0.9527 ] +INFO:root:Average loss=0.03743513847887516 +INFO:root: +[Valid][25][UTT][ MSE = 0.2156 | LCC = 0.9518 | SRCC = 0.9525 ] +INFO:root:Average loss=0.03440098913013935 +INFO:root: +[Valid][26][UTT][ MSE = 0.2115 | LCC = 0.9534 | SRCC = 0.9542 ] +INFO:root:Average loss=0.03289936943352222 +INFO:root: +[Valid][27][UTT][ MSE = 0.1851 | LCC = 0.9533 | SRCC = 0.9538 ] +INFO:root:Average loss=0.03309448350965977 +INFO:root: +[Valid][28][UTT][ MSE = 0.2097 | LCC = 0.9535 | SRCC = 0.9543 ] +INFO:root:Average loss=0.030794583924114705 +INFO:root: +[Valid][29][UTT][ MSE = 0.2055 | LCC = 0.9523 | SRCC = 0.9530 ] +INFO:root:Average loss=0.02958855830132961 +INFO:root: +[Valid][30][UTT][ MSE = 0.1983 | LCC = 0.9518 | SRCC = 0.9527 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2239 | LCC = 0.9540 | SRCC = 0.9550 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..418b14e5ebab88f7ec47add3d795408f064c8607 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7897bf0c06ef8ffd0cdf1c044b9f6e748d078e1aa70f0f556fd7c16e51f8ab57 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c6093bbcaf5a46bcd3af000974d6793661338633 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2055e1e6305434724e57369c3b207ea5528d23582991c2f25d44066d46255aff +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f32700e51ac635f7fe9e1c618f4f41f163c9d9f4 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer6/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.40889763820171354 +INFO:root: +[Valid][0][UTT][ MSE = 0.2813 | LCC = 0.9252 | SRCC = 0.9259 ] +INFO:root:Average loss=0.18182318532466887 +INFO:root: +[Valid][1][UTT][ MSE = 0.2337 | LCC = 0.9363 | SRCC = 0.9384 ] +INFO:root:Average loss=0.15410079234838486 +INFO:root: +[Valid][2][UTT][ MSE = 0.2178 | LCC = 0.9441 | SRCC = 0.9457 ] +INFO:root:Average loss=0.13352356904745102 +INFO:root: +[Valid][3][UTT][ MSE = 0.2480 | LCC = 0.9448 | SRCC = 0.9466 ] +INFO:root:Average loss=0.12351909840106964 +INFO:root: +[Valid][4][UTT][ MSE = 0.2103 | LCC = 0.9468 | SRCC = 0.9479 ] +INFO:root:Average loss=0.11363448613882064 +INFO:root: +[Valid][5][UTT][ MSE = 0.2242 | LCC = 0.9484 | SRCC = 0.9494 ] +INFO:root:Average loss=0.10469243422150612 +INFO:root: +[Valid][6][UTT][ MSE = 0.2135 | LCC = 0.9472 | SRCC = 0.9487 ] +INFO:root:Average loss=0.09436695045232772 +INFO:root: +[Valid][7][UTT][ MSE = 0.1913 | LCC = 0.9502 | SRCC = 0.9515 ] +INFO:root:Average loss=0.09425595360994339 +INFO:root: +[Valid][8][UTT][ MSE = 0.2034 | LCC = 0.9470 | SRCC = 0.9489 ] +INFO:root:Average loss=0.08328211745619774 +INFO:root: +[Valid][9][UTT][ MSE = 0.2514 | LCC = 0.9482 | SRCC = 0.9518 ] +INFO:root:Average loss=0.08176813051104545 +INFO:root: +[Valid][10][UTT][ MSE = 0.1996 | LCC = 0.9511 | SRCC = 0.9525 ] +INFO:root:Average loss=0.07654002526402473 +INFO:root: +[Valid][11][UTT][ MSE = 0.1875 | LCC = 0.9518 | SRCC = 0.9529 ] +INFO:root:Average loss=0.07319582250714302 +INFO:root: +[Valid][12][UTT][ MSE = 0.1953 | LCC = 0.9525 | SRCC = 0.9530 ] +INFO:root:Average loss=0.06734006237983704 +INFO:root: +[Valid][13][UTT][ MSE = 0.1943 | LCC = 0.9512 | SRCC = 0.9520 ] +INFO:root:Average loss=0.06398740473389626 +INFO:root: +[Valid][14][UTT][ MSE = 0.1960 | LCC = 0.9530 | SRCC = 0.9535 ] +INFO:root:Average loss=0.060372424453496935 +INFO:root: +[Valid][15][UTT][ MSE = 0.1933 | LCC = 0.9515 | SRCC = 0.9518 ] +INFO:root:Average loss=0.06020554268360138 +INFO:root: +[Valid][16][UTT][ MSE = 0.2112 | LCC = 0.9519 | SRCC = 0.9523 ] +INFO:root:Average loss=0.053811813920736314 +INFO:root: +[Valid][17][UTT][ MSE = 0.1973 | LCC = 0.9522 | SRCC = 0.9526 ] +INFO:root:Average loss=0.051851604864001276 +INFO:root: +[Valid][18][UTT][ MSE = 0.1794 | LCC = 0.9532 | SRCC = 0.9538 ] +INFO:root:Average loss=0.04973193246126175 +INFO:root: +[Valid][19][UTT][ MSE = 0.2048 | LCC = 0.9506 | SRCC = 0.9520 ] +INFO:root:Average loss=0.04772383844852447 +INFO:root: +[Valid][20][UTT][ MSE = 0.1772 | LCC = 0.9514 | SRCC = 0.9520 ] +INFO:root:Average loss=0.04376605641841889 +INFO:root: +[Valid][21][UTT][ MSE = 0.1955 | LCC = 0.9526 | SRCC = 0.9537 ] +INFO:root:Average loss=0.04407379359006882 +INFO:root: +[Valid][22][UTT][ MSE = 0.1808 | LCC = 0.9510 | SRCC = 0.9519 ] +INFO:root:Average loss=0.04480861052870751 +INFO:root: +[Valid][23][UTT][ MSE = 0.2039 | LCC = 0.9532 | SRCC = 0.9541 ] +INFO:root:Average loss=0.04044782271981239 +INFO:root: +[Valid][24][UTT][ MSE = 0.1862 | LCC = 0.9537 | SRCC = 0.9542 ] +INFO:root:Average loss=0.03646922397613525 +INFO:root: +[Valid][25][UTT][ MSE = 0.1912 | LCC = 0.9531 | SRCC = 0.9535 ] +INFO:root:Average loss=0.03647898119688034 +INFO:root: +[Valid][26][UTT][ MSE = 0.1998 | LCC = 0.9520 | SRCC = 0.9532 ] +INFO:root:Average loss=0.03584649051725865 +INFO:root: +[Valid][27][UTT][ MSE = 0.2176 | LCC = 0.9510 | SRCC = 0.9523 ] +INFO:root:Average loss=0.03425562085211277 +INFO:root: +[Valid][28][UTT][ MSE = 0.2121 | LCC = 0.9517 | SRCC = 0.9530 ] +INFO:root:Average loss=0.0332694047242403 +INFO:root: +[Valid][29][UTT][ MSE = 0.1999 | LCC = 0.9511 | SRCC = 0.9518 ] +INFO:root:Average loss=0.03262285892665386 +INFO:root: +[Valid][30][UTT][ MSE = 0.2229 | LCC = 0.9525 | SRCC = 0.9540 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1942 | LCC = 0.9534 | SRCC = 0.9545 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..afc5c0076fa62c123264a631a5ac771598e18961 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0216943dbcdddff58363ba80796ae687ffb4c87c87ea254b7dcca69e29342157 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..61fbb24ddfc5058ae56ca45b878b4b61c37bc91c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fd4de4ff88e66fe5716dd8f95556158854bfe84e3921466b2cc230439180f3e3 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0fb0192b9becb4a05912c50538166dd66adbc55e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3795255181789398 +INFO:root: +[Valid][0][UTT][ MSE = 0.2763 | LCC = 0.9298 | SRCC = 0.9302 ] +INFO:root:Average loss=0.17910365408658982 +INFO:root: +[Valid][1][UTT][ MSE = 0.2868 | LCC = 0.9382 | SRCC = 0.9391 ] +INFO:root:Average loss=0.14931206184625626 +INFO:root: +[Valid][2][UTT][ MSE = 0.2434 | LCC = 0.9454 | SRCC = 0.9468 ] +INFO:root:Average loss=0.13393901669979094 +INFO:root: +[Valid][3][UTT][ MSE = 0.2357 | LCC = 0.9441 | SRCC = 0.9453 ] +INFO:root:Average loss=0.12416048693656921 +INFO:root: +[Valid][4][UTT][ MSE = 0.2182 | LCC = 0.9475 | SRCC = 0.9493 ] +INFO:root:Average loss=0.11287460482120514 +INFO:root: +[Valid][5][UTT][ MSE = 0.2039 | LCC = 0.9503 | SRCC = 0.9511 ] +INFO:root:Average loss=0.10385601049661636 +INFO:root: +[Valid][6][UTT][ MSE = 0.2520 | LCC = 0.9472 | SRCC = 0.9499 ] +INFO:root:Average loss=0.09831344690918922 +INFO:root: +[Valid][7][UTT][ MSE = 0.2493 | LCC = 0.9512 | SRCC = 0.9526 ] +INFO:root:Average loss=0.09023834469914437 +INFO:root: +[Valid][8][UTT][ MSE = 0.2289 | LCC = 0.9460 | SRCC = 0.9475 ] +INFO:root:Average loss=0.09289654144644738 +INFO:root: +[Valid][9][UTT][ MSE = 0.2098 | LCC = 0.9516 | SRCC = 0.9523 ] +INFO:root:Average loss=0.08066764649748802 +INFO:root: +[Valid][10][UTT][ MSE = 0.2062 | LCC = 0.9506 | SRCC = 0.9507 ] +INFO:root:Average loss=0.07634541568160057 +INFO:root: +[Valid][11][UTT][ MSE = 0.1987 | LCC = 0.9497 | SRCC = 0.9507 ] +INFO:root:Average loss=0.07294633221626282 +INFO:root: +[Valid][12][UTT][ MSE = 0.2148 | LCC = 0.9470 | SRCC = 0.9476 ] +INFO:root:Average loss=0.06895221087336541 +INFO:root: +[Valid][13][UTT][ MSE = 0.1946 | LCC = 0.9522 | SRCC = 0.9528 ] +INFO:root:Average loss=0.06471780496835709 +INFO:root: +[Valid][14][UTT][ MSE = 0.2138 | LCC = 0.9521 | SRCC = 0.9527 ] +INFO:root:Average loss=0.06398489272594451 +INFO:root: +[Valid][15][UTT][ MSE = 0.2144 | LCC = 0.9513 | SRCC = 0.9516 ] +INFO:root:Average loss=0.06159736241400242 +INFO:root: +[Valid][16][UTT][ MSE = 0.2176 | LCC = 0.9470 | SRCC = 0.9479 ] +INFO:root:Average loss=0.06142305770516396 +INFO:root: +[Valid][17][UTT][ MSE = 0.2266 | LCC = 0.9487 | SRCC = 0.9503 ] +INFO:root:Average loss=0.05436309508979321 +INFO:root: +[Valid][18][UTT][ MSE = 0.2099 | LCC = 0.9504 | SRCC = 0.9511 ] +INFO:root:Average loss=0.05132994091510773 +INFO:root: +[Valid][19][UTT][ MSE = 0.2048 | LCC = 0.9512 | SRCC = 0.9519 ] +INFO:root:Average loss=0.05334633809328079 +INFO:root: +[Valid][20][UTT][ MSE = 0.2025 | LCC = 0.9519 | SRCC = 0.9524 ] +INFO:root:Average loss=0.04778014156222343 +INFO:root: +[Valid][21][UTT][ MSE = 0.1894 | LCC = 0.9493 | SRCC = 0.9498 ] +INFO:root:Average loss=0.04423848521709442 +INFO:root: +[Valid][22][UTT][ MSE = 0.2030 | LCC = 0.9523 | SRCC = 0.9528 ] +INFO:root:Average loss=0.047054816246032714 +INFO:root: +[Valid][23][UTT][ MSE = 0.2225 | LCC = 0.9510 | SRCC = 0.9518 ] +INFO:root:Average loss=0.04006380877643824 +INFO:root: +[Valid][24][UTT][ MSE = 0.1937 | LCC = 0.9526 | SRCC = 0.9527 ] +INFO:root:Average loss=0.041129816457629205 +INFO:root: +[Valid][25][UTT][ MSE = 0.2090 | LCC = 0.9517 | SRCC = 0.9518 ] +INFO:root:Average loss=0.03802778598666191 +INFO:root: +[Valid][26][UTT][ MSE = 0.2062 | LCC = 0.9506 | SRCC = 0.9508 ] +INFO:root:Average loss=0.03706763149797916 +INFO:root: +[Valid][27][UTT][ MSE = 0.2100 | LCC = 0.9498 | SRCC = 0.9505 ] +INFO:root:Average loss=0.03406005984544754 +INFO:root: +[Valid][28][UTT][ MSE = 0.2069 | LCC = 0.9515 | SRCC = 0.9511 ] +INFO:root:Average loss=0.03361578883230686 +INFO:root: +[Valid][29][UTT][ MSE = 0.2124 | LCC = 0.9502 | SRCC = 0.9507 ] +INFO:root:Average loss=0.03547283236682415 +INFO:root: +[Valid][30][UTT][ MSE = 0.2151 | LCC = 0.9499 | SRCC = 0.9510 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2037 | LCC = 0.9531 | SRCC = 0.9544 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1ac06f62e8c8b0d33e9894514b71b1910d26468e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:22106ed2e0f1c0af5fefa07d0bdf24c6eaeb7ffb63dda61fd09b438cab156c1c +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d33bea96fa4a85b1e128b79ce9a432a361cd9afe --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:32d89d0084cbeb0c520a98f8745e22b5038f87b9b929dd3b2a91198850a6ec38 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f5edec1621c3bd6bc82df441c33afca6830b6eac --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.45887871479988096 +INFO:root: +[Valid][0][UTT][ MSE = 0.2941 | LCC = 0.9186 | SRCC = 0.9220 ] +INFO:root:Average loss=0.1950674271583557 +INFO:root: +[Valid][1][UTT][ MSE = 0.2823 | LCC = 0.9386 | SRCC = 0.9404 ] +INFO:root:Average loss=0.16033083242177965 +INFO:root: +[Valid][2][UTT][ MSE = 0.2403 | LCC = 0.9434 | SRCC = 0.9440 ] +INFO:root:Average loss=0.13852513915300368 +INFO:root: +[Valid][3][UTT][ MSE = 0.2597 | LCC = 0.9452 | SRCC = 0.9460 ] +INFO:root:Average loss=0.12632779532670974 +INFO:root: +[Valid][4][UTT][ MSE = 0.2116 | LCC = 0.9458 | SRCC = 0.9464 ] +INFO:root:Average loss=0.1175848553776741 +INFO:root: +[Valid][5][UTT][ MSE = 0.2519 | LCC = 0.9481 | SRCC = 0.9488 ] +INFO:root:Average loss=0.10807828015089035 +INFO:root: +[Valid][6][UTT][ MSE = 0.2020 | LCC = 0.9485 | SRCC = 0.9504 ] +INFO:root:Average loss=0.10121943658590317 +INFO:root: +[Valid][7][UTT][ MSE = 0.2342 | LCC = 0.9465 | SRCC = 0.9476 ] +INFO:root:Average loss=0.09399103796482086 +INFO:root: +[Valid][8][UTT][ MSE = 0.2028 | LCC = 0.9501 | SRCC = 0.9513 ] +INFO:root:Average loss=0.08945922356843948 +INFO:root: +[Valid][9][UTT][ MSE = 0.2277 | LCC = 0.9511 | SRCC = 0.9521 ] +INFO:root:Average loss=0.08697024920582772 +INFO:root: +[Valid][10][UTT][ MSE = 0.2289 | LCC = 0.9505 | SRCC = 0.9510 ] +INFO:root:Average loss=0.07725988021492958 +INFO:root: +[Valid][11][UTT][ MSE = 0.2184 | LCC = 0.9492 | SRCC = 0.9495 ] +INFO:root:Average loss=0.0749389006793499 +INFO:root: +[Valid][12][UTT][ MSE = 0.2136 | LCC = 0.9491 | SRCC = 0.9495 ] +INFO:root:Average loss=0.06896125864982605 +INFO:root: +[Valid][13][UTT][ MSE = 0.1913 | LCC = 0.9494 | SRCC = 0.9501 ] +INFO:root:Average loss=0.06837906914949417 +INFO:root: +[Valid][14][UTT][ MSE = 0.1973 | LCC = 0.9494 | SRCC = 0.9498 ] +INFO:root:Average loss=0.06456496757268905 +INFO:root: +[Valid][15][UTT][ MSE = 0.2204 | LCC = 0.9497 | SRCC = 0.9498 ] +INFO:root:Average loss=0.06192379623651505 +INFO:root: +[Valid][16][UTT][ MSE = 0.1825 | LCC = 0.9483 | SRCC = 0.9491 ] +INFO:root:Average loss=0.05753123790025711 +INFO:root: +[Valid][17][UTT][ MSE = 0.2030 | LCC = 0.9502 | SRCC = 0.9505 ] +INFO:root:Average loss=0.055793226629495624 +INFO:root: +[Valid][18][UTT][ MSE = 0.2051 | LCC = 0.9503 | SRCC = 0.9507 ] +INFO:root:Average loss=0.05242875911295414 +INFO:root: +[Valid][19][UTT][ MSE = 0.2324 | LCC = 0.9470 | SRCC = 0.9469 ] +INFO:root:Average loss=0.04970262771844864 +INFO:root: +[Valid][20][UTT][ MSE = 0.2023 | LCC = 0.9488 | SRCC = 0.9491 ] +INFO:root:Average loss=0.05018509523570538 +INFO:root: +[Valid][21][UTT][ MSE = 0.2094 | LCC = 0.9501 | SRCC = 0.9500 ] +INFO:root:Average loss=0.04462411680817604 +INFO:root: +[Valid][22][UTT][ MSE = 0.2045 | LCC = 0.9482 | SRCC = 0.9479 ] +INFO:root:Average loss=0.04422852191329003 +INFO:root: +[Valid][23][UTT][ MSE = 0.2129 | LCC = 0.9500 | SRCC = 0.9505 ] +INFO:root:Average loss=0.042932215258479116 +INFO:root: +[Valid][24][UTT][ MSE = 0.2158 | LCC = 0.9489 | SRCC = 0.9487 ] +INFO:root:Average loss=0.040247301086783406 +INFO:root: +[Valid][25][UTT][ MSE = 0.2153 | LCC = 0.9499 | SRCC = 0.9501 ] +INFO:root:Average loss=0.03805620963871479 +INFO:root: +[Valid][26][UTT][ MSE = 0.2072 | LCC = 0.9492 | SRCC = 0.9504 ] +INFO:root:Average loss=0.037836986765265465 +INFO:root: +[Valid][27][UTT][ MSE = 0.2201 | LCC = 0.9513 | SRCC = 0.9515 ] +INFO:root:Average loss=0.03711707866191864 +INFO:root: +[Valid][28][UTT][ MSE = 0.2364 | LCC = 0.9493 | SRCC = 0.9504 ] +INFO:root:Average loss=0.03662942077219486 +INFO:root: +[Valid][29][UTT][ MSE = 0.2174 | LCC = 0.9493 | SRCC = 0.9493 ] +INFO:root:Average loss=0.03619437149167061 +INFO:root: +[Valid][30][UTT][ MSE = 0.2126 | LCC = 0.9502 | SRCC = 0.9503 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2275 | LCC = 0.9527 | SRCC = 0.9539 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d3178641f228c712805a516f26ae8574323e8d50 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fce6b5ec3d390a676ad7413efe30a2b8b2b4d1841cb4567462d8b614f99188f1 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f1faffb802b3eafa1d06535722797822d7107cbd --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:30aabd819e8300d45d04d2843cd0d784b5e44ef6b514dc001d7ac2f6b55f5083 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9e54b0189e272aefc1c4adfb81caa0c49ccddea4 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.43660208797454836 +INFO:root: +[Valid][0][UTT][ MSE = 0.3290 | LCC = 0.9236 | SRCC = 0.9241 ] +INFO:root:Average loss=0.1829475224018097 +INFO:root: +[Valid][1][UTT][ MSE = 0.2974 | LCC = 0.9402 | SRCC = 0.9422 ] +INFO:root:Average loss=0.1540798797607422 +INFO:root: +[Valid][2][UTT][ MSE = 0.2375 | LCC = 0.9448 | SRCC = 0.9451 ] +INFO:root:Average loss=0.13462268370389938 +INFO:root: +[Valid][3][UTT][ MSE = 0.2451 | LCC = 0.9475 | SRCC = 0.9484 ] +INFO:root:Average loss=0.12221343600749969 +INFO:root: +[Valid][4][UTT][ MSE = 0.2341 | LCC = 0.9490 | SRCC = 0.9504 ] +INFO:root:Average loss=0.1123434796333313 +INFO:root: +[Valid][5][UTT][ MSE = 0.2571 | LCC = 0.9487 | SRCC = 0.9500 ] +INFO:root:Average loss=0.10396945402026177 +INFO:root: +[Valid][6][UTT][ MSE = 0.2204 | LCC = 0.9501 | SRCC = 0.9512 ] +INFO:root:Average loss=0.09995429557561875 +INFO:root: +[Valid][7][UTT][ MSE = 0.2289 | LCC = 0.9454 | SRCC = 0.9486 ] +INFO:root:Average loss=0.09190564075112342 +INFO:root: +[Valid][8][UTT][ MSE = 0.2136 | LCC = 0.9509 | SRCC = 0.9516 ] +INFO:root:Average loss=0.08687495100498199 +INFO:root: +[Valid][9][UTT][ MSE = 0.2328 | LCC = 0.9491 | SRCC = 0.9509 ] +INFO:root:Average loss=0.08189962968230248 +INFO:root: +[Valid][10][UTT][ MSE = 0.2188 | LCC = 0.9501 | SRCC = 0.9511 ] +INFO:root:Average loss=0.07658350044488907 +INFO:root: +[Valid][11][UTT][ MSE = 0.2060 | LCC = 0.9513 | SRCC = 0.9514 ] +INFO:root:Average loss=0.07595232290029526 +INFO:root: +[Valid][12][UTT][ MSE = 0.2310 | LCC = 0.9499 | SRCC = 0.9518 ] +INFO:root:Average loss=0.07112147599458694 +INFO:root: +[Valid][13][UTT][ MSE = 0.2078 | LCC = 0.9505 | SRCC = 0.9514 ] +INFO:root:Average loss=0.06355426850914955 +INFO:root: +[Valid][14][UTT][ MSE = 0.2234 | LCC = 0.9507 | SRCC = 0.9516 ] +INFO:root:Average loss=0.06277950251102447 +INFO:root: +[Valid][15][UTT][ MSE = 0.2067 | LCC = 0.9507 | SRCC = 0.9514 ] +INFO:root:Average loss=0.059329343974590304 +INFO:root: +[Valid][16][UTT][ MSE = 0.2030 | LCC = 0.9478 | SRCC = 0.9482 ] +INFO:root:Average loss=0.055347108274698256 +INFO:root: +[Valid][17][UTT][ MSE = 0.2439 | LCC = 0.9476 | SRCC = 0.9505 ] +INFO:root:Average loss=0.05340147405862808 +INFO:root: +[Valid][18][UTT][ MSE = 0.2265 | LCC = 0.9504 | SRCC = 0.9509 ] +INFO:root:Average loss=0.04928492838144302 +INFO:root: +[Valid][19][UTT][ MSE = 0.2196 | LCC = 0.9502 | SRCC = 0.9502 ] +INFO:root:Average loss=0.0486734808087349 +INFO:root: +[Valid][20][UTT][ MSE = 0.1884 | LCC = 0.9514 | SRCC = 0.9518 ] +INFO:root:Average loss=0.04800995297729969 +INFO:root: +[Valid][21][UTT][ MSE = 0.2317 | LCC = 0.9492 | SRCC = 0.9496 ] +INFO:root:Average loss=0.0440369720607996 +INFO:root: +[Valid][22][UTT][ MSE = 0.1873 | LCC = 0.9514 | SRCC = 0.9513 ] +INFO:root:Average loss=0.043484204322099686 +INFO:root: +[Valid][23][UTT][ MSE = 0.2007 | LCC = 0.9517 | SRCC = 0.9517 ] +INFO:root:Average loss=0.04132666824758053 +INFO:root: +[Valid][24][UTT][ MSE = 0.2358 | LCC = 0.9485 | SRCC = 0.9498 ] +INFO:root:Average loss=0.03937691617012024 +INFO:root: +[Valid][25][UTT][ MSE = 0.2056 | LCC = 0.9487 | SRCC = 0.9486 ] +INFO:root:Average loss=0.037611574977636335 +INFO:root: +[Valid][26][UTT][ MSE = 0.2100 | LCC = 0.9490 | SRCC = 0.9494 ] +INFO:root:Average loss=0.0352730652987957 +INFO:root: +[Valid][27][UTT][ MSE = 0.2277 | LCC = 0.9491 | SRCC = 0.9501 ] +INFO:root:Average loss=0.03559019266068936 +INFO:root: +[Valid][28][UTT][ MSE = 0.2191 | LCC = 0.9509 | SRCC = 0.9519 ] +INFO:root:Average loss=0.034208733141422275 +INFO:root: +[Valid][29][UTT][ MSE = 0.2080 | LCC = 0.9515 | SRCC = 0.9521 ] +INFO:root:Average loss=0.034728865057229996 +INFO:root: +[Valid][30][UTT][ MSE = 0.2124 | LCC = 0.9511 | SRCC = 0.9512 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2041 | LCC = 0.9536 | SRCC = 0.9547 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3d57e122ad358d542c0667c6928cead877404203 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:231fb4dc1de079626f4171c5e2af570fad32fd311210d1322b26aefc0b312188 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..842eb04b0b67d872932eb0d5c72701cb059e2bca --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a1400821b17957d4be685fb15b94bda1adcb377aa5d6e5f0ade6286acb54565b +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..58c5c5e7b2e8c4da462a7a381db7d6b99992ad76 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.38220336055755616 +INFO:root: +[Valid][0][UTT][ MSE = 0.2899 | LCC = 0.9256 | SRCC = 0.9263 ] +INFO:root:Average loss=0.1884854957461357 +INFO:root: +[Valid][1][UTT][ MSE = 0.2869 | LCC = 0.9395 | SRCC = 0.9407 ] +INFO:root:Average loss=0.1587417197227478 +INFO:root: +[Valid][2][UTT][ MSE = 0.2217 | LCC = 0.9435 | SRCC = 0.9448 ] +INFO:root:Average loss=0.1373079694509506 +INFO:root: +[Valid][3][UTT][ MSE = 0.2153 | LCC = 0.9458 | SRCC = 0.9467 ] +INFO:root:Average loss=0.1226277107000351 +INFO:root: +[Valid][4][UTT][ MSE = 0.2064 | LCC = 0.9470 | SRCC = 0.9488 ] +INFO:root:Average loss=0.11391983732581139 +INFO:root: +[Valid][5][UTT][ MSE = 0.2108 | LCC = 0.9494 | SRCC = 0.9507 ] +INFO:root:Average loss=0.10584946650266647 +INFO:root: +[Valid][6][UTT][ MSE = 0.2246 | LCC = 0.9493 | SRCC = 0.9500 ] +INFO:root:Average loss=0.09900952008366584 +INFO:root: +[Valid][7][UTT][ MSE = 0.2346 | LCC = 0.9507 | SRCC = 0.9513 ] +INFO:root:Average loss=0.09382659104466438 +INFO:root: +[Valid][8][UTT][ MSE = 0.1901 | LCC = 0.9500 | SRCC = 0.9515 ] +INFO:root:Average loss=0.09143266308307647 +INFO:root: +[Valid][9][UTT][ MSE = 0.2314 | LCC = 0.9516 | SRCC = 0.9527 ] +INFO:root:Average loss=0.08498089396953583 +INFO:root: +[Valid][10][UTT][ MSE = 0.2112 | LCC = 0.9499 | SRCC = 0.9510 ] +INFO:root:Average loss=0.07888130983710288 +INFO:root: +[Valid][11][UTT][ MSE = 0.2015 | LCC = 0.9512 | SRCC = 0.9522 ] +INFO:root:Average loss=0.07249431186914444 +INFO:root: +[Valid][12][UTT][ MSE = 0.2201 | LCC = 0.9471 | SRCC = 0.9480 ] +INFO:root:Average loss=0.06838862332701683 +INFO:root: +[Valid][13][UTT][ MSE = 0.2045 | LCC = 0.9495 | SRCC = 0.9511 ] +INFO:root:Average loss=0.06524487778544426 +INFO:root: +[Valid][14][UTT][ MSE = 0.2154 | LCC = 0.9511 | SRCC = 0.9518 ] +INFO:root:Average loss=0.06281218707561492 +INFO:root: +[Valid][15][UTT][ MSE = 0.1934 | LCC = 0.9516 | SRCC = 0.9528 ] +INFO:root:Average loss=0.05848144352436066 +INFO:root: +[Valid][16][UTT][ MSE = 0.2040 | LCC = 0.9510 | SRCC = 0.9517 ] +INFO:root:Average loss=0.05694202318787575 +INFO:root: +[Valid][17][UTT][ MSE = 0.2109 | LCC = 0.9515 | SRCC = 0.9521 ] +INFO:root:Average loss=0.0541700418740511 +INFO:root: +[Valid][18][UTT][ MSE = 0.2094 | LCC = 0.9518 | SRCC = 0.9523 ] +INFO:root:Average loss=0.055184931606054304 +INFO:root: +[Valid][19][UTT][ MSE = 0.1995 | LCC = 0.9465 | SRCC = 0.9478 ] +INFO:root:Average loss=0.05106238490343094 +INFO:root: +[Valid][20][UTT][ MSE = 0.2443 | LCC = 0.9508 | SRCC = 0.9527 ] +INFO:root:Average loss=0.04964294075965881 +INFO:root: +[Valid][21][UTT][ MSE = 0.1957 | LCC = 0.9510 | SRCC = 0.9517 ] +INFO:root:Average loss=0.045653062716126444 +INFO:root: +[Valid][22][UTT][ MSE = 0.1965 | LCC = 0.9513 | SRCC = 0.9517 ] +INFO:root:Average loss=0.04386616553366184 +INFO:root: +[Valid][23][UTT][ MSE = 0.2150 | LCC = 0.9506 | SRCC = 0.9511 ] +INFO:root:Average loss=0.043527578085660934 +INFO:root: +[Valid][24][UTT][ MSE = 0.2060 | LCC = 0.9528 | SRCC = 0.9532 ] +INFO:root:Average loss=0.042334718465805055 +INFO:root: +[Valid][25][UTT][ MSE = 0.2028 | LCC = 0.9502 | SRCC = 0.9510 ] +INFO:root:Average loss=0.03829601709544659 +INFO:root: +[Valid][26][UTT][ MSE = 0.2434 | LCC = 0.9509 | SRCC = 0.9519 ] +INFO:root:Average loss=0.03759830643236637 +INFO:root: +[Valid][27][UTT][ MSE = 0.2042 | LCC = 0.9519 | SRCC = 0.9524 ] +INFO:root:Average loss=0.03517821328341961 +INFO:root: +[Valid][28][UTT][ MSE = 0.2082 | LCC = 0.9507 | SRCC = 0.9513 ] +INFO:root:Average loss=0.03499572090804577 +INFO:root: +[Valid][29][UTT][ MSE = 0.2106 | LCC = 0.9517 | SRCC = 0.9524 ] +INFO:root:Average loss=0.0329365508556366 +INFO:root: +[Valid][30][UTT][ MSE = 0.1906 | LCC = 0.9516 | SRCC = 0.9521 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2214 | LCC = 0.9528 | SRCC = 0.9538 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..96ffc2c0316356fe29615269c408d2a8b7979ddf --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bae736ae3e625579fb629081a55db0d5b1f652a6166418949af93870651af710 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f75df84dcff87c622963ef4ca359f779fbdb9ccf --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b8b296f73692fc19e22b71c425b21ac567860fb84e6f68d3a7ca9258a36c6b6 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..903038b806c9f0844397fa108cbc2dfa731f3ab2 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer7/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4542096072435379 +INFO:root: +[Valid][0][UTT][ MSE = 0.3031 | LCC = 0.9205 | SRCC = 0.9194 ] +INFO:root:Average loss=0.19370228737592698 +INFO:root: +[Valid][1][UTT][ MSE = 0.2260 | LCC = 0.9363 | SRCC = 0.9368 ] +INFO:root:Average loss=0.15248759555816652 +INFO:root: +[Valid][2][UTT][ MSE = 0.2180 | LCC = 0.9447 | SRCC = 0.9464 ] +INFO:root:Average loss=0.13588381844758987 +INFO:root: +[Valid][3][UTT][ MSE = 0.2120 | LCC = 0.9462 | SRCC = 0.9476 ] +INFO:root:Average loss=0.12204218655824661 +INFO:root: +[Valid][4][UTT][ MSE = 0.2171 | LCC = 0.9471 | SRCC = 0.9485 ] +INFO:root:Average loss=0.11430694288015365 +INFO:root: +[Valid][5][UTT][ MSE = 0.2320 | LCC = 0.9488 | SRCC = 0.9502 ] +INFO:root:Average loss=0.10613754856586456 +INFO:root: +[Valid][6][UTT][ MSE = 0.2241 | LCC = 0.9481 | SRCC = 0.9503 ] +INFO:root:Average loss=0.09835205808281898 +INFO:root: +[Valid][7][UTT][ MSE = 0.2025 | LCC = 0.9478 | SRCC = 0.9497 ] +INFO:root:Average loss=0.09308099725842475 +INFO:root: +[Valid][8][UTT][ MSE = 0.1932 | LCC = 0.9487 | SRCC = 0.9502 ] +INFO:root:Average loss=0.08856818062067032 +INFO:root: +[Valid][9][UTT][ MSE = 0.2039 | LCC = 0.9482 | SRCC = 0.9489 ] +INFO:root:Average loss=0.08019436922669411 +INFO:root: +[Valid][10][UTT][ MSE = 0.1843 | LCC = 0.9505 | SRCC = 0.9516 ] +INFO:root:Average loss=0.07588800442218781 +INFO:root: +[Valid][11][UTT][ MSE = 0.1867 | LCC = 0.9524 | SRCC = 0.9532 ] +INFO:root:Average loss=0.07546773076057434 +INFO:root: +[Valid][12][UTT][ MSE = 0.1865 | LCC = 0.9504 | SRCC = 0.9520 ] +INFO:root:Average loss=0.06727490738034249 +INFO:root: +[Valid][13][UTT][ MSE = 0.2080 | LCC = 0.9493 | SRCC = 0.9503 ] +INFO:root:Average loss=0.06571665495634078 +INFO:root: +[Valid][14][UTT][ MSE = 0.1925 | LCC = 0.9518 | SRCC = 0.9522 ] +INFO:root:Average loss=0.06297182032465935 +INFO:root: +[Valid][15][UTT][ MSE = 0.2073 | LCC = 0.9507 | SRCC = 0.9520 ] +INFO:root:Average loss=0.05752062195539474 +INFO:root: +[Valid][16][UTT][ MSE = 0.1792 | LCC = 0.9515 | SRCC = 0.9523 ] +INFO:root:Average loss=0.05494471031427384 +INFO:root: +[Valid][17][UTT][ MSE = 0.2085 | LCC = 0.9530 | SRCC = 0.9535 ] +INFO:root:Average loss=0.05245620128512383 +INFO:root: +[Valid][18][UTT][ MSE = 0.1886 | LCC = 0.9505 | SRCC = 0.9508 ] +INFO:root:Average loss=0.049716114044189456 +INFO:root: +[Valid][19][UTT][ MSE = 0.1792 | LCC = 0.9515 | SRCC = 0.9520 ] +INFO:root:Average loss=0.04794665163755417 +INFO:root: +[Valid][20][UTT][ MSE = 0.1918 | LCC = 0.9510 | SRCC = 0.9521 ] +INFO:root:Average loss=0.04461339715123176 +INFO:root: +[Valid][21][UTT][ MSE = 0.1783 | LCC = 0.9522 | SRCC = 0.9530 ] +INFO:root:Average loss=0.042451718270778656 +INFO:root: +[Valid][22][UTT][ MSE = 0.2049 | LCC = 0.9507 | SRCC = 0.9521 ] +INFO:root:Average loss=0.04237612058222294 +INFO:root: +[Valid][23][UTT][ MSE = 0.1865 | LCC = 0.9529 | SRCC = 0.9538 ] +INFO:root:Average loss=0.03916429553925991 +INFO:root: +[Valid][24][UTT][ MSE = 0.2059 | LCC = 0.9514 | SRCC = 0.9520 ] +INFO:root:Average loss=0.039668657913804056 +INFO:root: +[Valid][25][UTT][ MSE = 0.1906 | LCC = 0.9522 | SRCC = 0.9532 ] +INFO:root:Average loss=0.03599092073738575 +INFO:root: +[Valid][26][UTT][ MSE = 0.1978 | LCC = 0.9522 | SRCC = 0.9524 ] +INFO:root:Average loss=0.03786354283988476 +INFO:root: +[Valid][27][UTT][ MSE = 0.1760 | LCC = 0.9531 | SRCC = 0.9538 ] +INFO:root:Average loss=0.03466813752055168 +INFO:root: +[Valid][28][UTT][ MSE = 0.1929 | LCC = 0.9523 | SRCC = 0.9528 ] +INFO:root:Average loss=0.032485018074512484 +INFO:root: +[Valid][29][UTT][ MSE = 0.2005 | LCC = 0.9517 | SRCC = 0.9522 ] +INFO:root:Average loss=0.03234594416618347 +INFO:root: +[Valid][30][UTT][ MSE = 0.1949 | LCC = 0.9507 | SRCC = 0.9516 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1856 | LCC = 0.9522 | SRCC = 0.9536 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1532618fbc5f2537f7d52d2528ffe436ab2af4fc --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b48463e26822e61818e0ee18c51fdb8b491a084107e753202cf588b9be282410 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6462b21daf16d6a216a9c59b23c00868c3c6443e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6bd165f354514ac0fa0cf206e9923eafe5bc6cb3782015bae486d3833a90dc6a +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c57bfeaa494f5c5b2ee10f58a939dea830c608f8 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4535652527809143 +INFO:root: +[Valid][0][UTT][ MSE = 0.3390 | LCC = 0.9150 | SRCC = 0.9256 ] +INFO:root:Average loss=0.18639140504598617 +INFO:root: +[Valid][1][UTT][ MSE = 0.3103 | LCC = 0.9404 | SRCC = 0.9431 ] +INFO:root:Average loss=0.16002341037988663 +INFO:root: +[Valid][2][UTT][ MSE = 0.2892 | LCC = 0.9451 | SRCC = 0.9478 ] +INFO:root:Average loss=0.1376120421886444 +INFO:root: +[Valid][3][UTT][ MSE = 0.2753 | LCC = 0.9484 | SRCC = 0.9510 ] +INFO:root:Average loss=0.12881259500980377 +INFO:root: +[Valid][4][UTT][ MSE = 0.2629 | LCC = 0.9485 | SRCC = 0.9517 ] +INFO:root:Average loss=0.11776514947414399 +INFO:root: +[Valid][5][UTT][ MSE = 0.2357 | LCC = 0.9491 | SRCC = 0.9517 ] +INFO:root:Average loss=0.11249046158790589 +INFO:root: +[Valid][6][UTT][ MSE = 0.2228 | LCC = 0.9505 | SRCC = 0.9524 ] +INFO:root:Average loss=0.10393206351995468 +INFO:root: +[Valid][7][UTT][ MSE = 0.2248 | LCC = 0.9527 | SRCC = 0.9545 ] +INFO:root:Average loss=0.09752230760455131 +INFO:root: +[Valid][8][UTT][ MSE = 0.2474 | LCC = 0.9500 | SRCC = 0.9521 ] +INFO:root:Average loss=0.09673263019323348 +INFO:root: +[Valid][9][UTT][ MSE = 0.2310 | LCC = 0.9525 | SRCC = 0.9549 ] +INFO:root:Average loss=0.09023099383711815 +INFO:root: +[Valid][10][UTT][ MSE = 0.2190 | LCC = 0.9531 | SRCC = 0.9546 ] +INFO:root:Average loss=0.08414890491962433 +INFO:root: +[Valid][11][UTT][ MSE = 0.2085 | LCC = 0.9539 | SRCC = 0.9555 ] +INFO:root:Average loss=0.08095589643716812 +INFO:root: +[Valid][12][UTT][ MSE = 0.2384 | LCC = 0.9529 | SRCC = 0.9548 ] +INFO:root:Average loss=0.07919505688548088 +INFO:root: +[Valid][13][UTT][ MSE = 0.2394 | LCC = 0.9534 | SRCC = 0.9551 ] +INFO:root:Average loss=0.07406702515482902 +INFO:root: +[Valid][14][UTT][ MSE = 0.2383 | LCC = 0.9516 | SRCC = 0.9538 ] +INFO:root:Average loss=0.07214828231930732 +INFO:root: +[Valid][15][UTT][ MSE = 0.2331 | LCC = 0.9527 | SRCC = 0.9554 ] +INFO:root:Average loss=0.06623273220658302 +INFO:root: +[Valid][16][UTT][ MSE = 0.2229 | LCC = 0.9545 | SRCC = 0.9559 ] +INFO:root:Average loss=0.06335344323515892 +INFO:root: +[Valid][17][UTT][ MSE = 0.2415 | LCC = 0.9535 | SRCC = 0.9555 ] +INFO:root:Average loss=0.06342540580034256 +INFO:root: +[Valid][18][UTT][ MSE = 0.2146 | LCC = 0.9540 | SRCC = 0.9551 ] +INFO:root:Average loss=0.05764209574460983 +INFO:root: +[Valid][19][UTT][ MSE = 0.2212 | LCC = 0.9533 | SRCC = 0.9543 ] +INFO:root:Average loss=0.05656523650884628 +INFO:root: +[Valid][20][UTT][ MSE = 0.2034 | LCC = 0.9546 | SRCC = 0.9557 ] +INFO:root:Average loss=0.05524353355169296 +INFO:root: +[Valid][21][UTT][ MSE = 0.2053 | LCC = 0.9527 | SRCC = 0.9543 ] +INFO:root:Average loss=0.05540272024273872 +INFO:root: +[Valid][22][UTT][ MSE = 0.2289 | LCC = 0.9548 | SRCC = 0.9565 ] +INFO:root:Average loss=0.05362495401501655 +INFO:root: +[Valid][23][UTT][ MSE = 0.1979 | LCC = 0.9545 | SRCC = 0.9554 ] +INFO:root:Average loss=0.04919727647304535 +INFO:root: +[Valid][24][UTT][ MSE = 0.2287 | LCC = 0.9539 | SRCC = 0.9555 ] +INFO:root:Average loss=0.04845613619685173 +INFO:root: +[Valid][25][UTT][ MSE = 0.2071 | LCC = 0.9548 | SRCC = 0.9558 ] +INFO:root:Average loss=0.04578055272996426 +INFO:root: +[Valid][26][UTT][ MSE = 0.2088 | LCC = 0.9547 | SRCC = 0.9555 ] +INFO:root:Average loss=0.045270334735512736 +INFO:root: +[Valid][27][UTT][ MSE = 0.2055 | LCC = 0.9543 | SRCC = 0.9552 ] +INFO:root:Average loss=0.04410006956756115 +INFO:root: +[Valid][28][UTT][ MSE = 0.2161 | LCC = 0.9547 | SRCC = 0.9557 ] +INFO:root:Average loss=0.040104021161794665 +INFO:root: +[Valid][29][UTT][ MSE = 0.2232 | LCC = 0.9513 | SRCC = 0.9530 ] +INFO:root:Average loss=0.040417472437024114 +INFO:root: +[Valid][30][UTT][ MSE = 0.2108 | LCC = 0.9543 | SRCC = 0.9555 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2434 | LCC = 0.9536 | SRCC = 0.9559 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..889ac9d853064d2b04a9f8d5585c24da6b463569 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d16bbb58a5a88ebe8bbd2ed0e0f6a5815028bd556ce7b69562873fc159fc14b2 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..316951c4032b8178faa6a9a35d93adfb677030cc --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1ed457571f83bf07964583f42099f0469c4cb69a69c4ccaf91557a65edbc1405 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ad13fa84a0045901de8906c56117ba9967bfb01e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3366733160018921 +INFO:root: +[Valid][0][UTT][ MSE = 0.3810 | LCC = 0.9245 | SRCC = 0.9262 ] +INFO:root:Average loss=0.18449829423427583 +INFO:root: +[Valid][1][UTT][ MSE = 0.2802 | LCC = 0.9369 | SRCC = 0.9398 ] +INFO:root:Average loss=0.15696697825193406 +INFO:root: +[Valid][2][UTT][ MSE = 0.2498 | LCC = 0.9448 | SRCC = 0.9466 ] +INFO:root:Average loss=0.1393587148785591 +INFO:root: +[Valid][3][UTT][ MSE = 0.2550 | LCC = 0.9469 | SRCC = 0.9489 ] +INFO:root:Average loss=0.13038189947605133 +INFO:root: +[Valid][4][UTT][ MSE = 0.2453 | LCC = 0.9475 | SRCC = 0.9489 ] +INFO:root:Average loss=0.11975489574670792 +INFO:root: +[Valid][5][UTT][ MSE = 0.2250 | LCC = 0.9476 | SRCC = 0.9499 ] +INFO:root:Average loss=0.10991513666510581 +INFO:root: +[Valid][6][UTT][ MSE = 0.2558 | LCC = 0.9453 | SRCC = 0.9482 ] +INFO:root:Average loss=0.10592013278603554 +INFO:root: +[Valid][7][UTT][ MSE = 0.2339 | LCC = 0.9506 | SRCC = 0.9523 ] +INFO:root:Average loss=0.10180696192383766 +INFO:root: +[Valid][8][UTT][ MSE = 0.2355 | LCC = 0.9481 | SRCC = 0.9504 ] +INFO:root:Average loss=0.09751774615049362 +INFO:root: +[Valid][9][UTT][ MSE = 0.2551 | LCC = 0.9509 | SRCC = 0.9530 ] +INFO:root:Average loss=0.09420063948631287 +INFO:root: +[Valid][10][UTT][ MSE = 0.2438 | LCC = 0.9524 | SRCC = 0.9537 ] +INFO:root:Average loss=0.08452028626203537 +INFO:root: +[Valid][11][UTT][ MSE = 0.2211 | LCC = 0.9516 | SRCC = 0.9528 ] +INFO:root:Average loss=0.08708619731664657 +INFO:root: +[Valid][12][UTT][ MSE = 0.2551 | LCC = 0.9510 | SRCC = 0.9526 ] +INFO:root:Average loss=0.07739253088831902 +INFO:root: +[Valid][13][UTT][ MSE = 0.2312 | LCC = 0.9525 | SRCC = 0.9538 ] +INFO:root:Average loss=0.07420756849646569 +INFO:root: +[Valid][14][UTT][ MSE = 0.2604 | LCC = 0.9510 | SRCC = 0.9519 ] +INFO:root:Average loss=0.07543748834729194 +INFO:root: +[Valid][15][UTT][ MSE = 0.2074 | LCC = 0.9526 | SRCC = 0.9540 ] +INFO:root:Average loss=0.07014206576347351 +INFO:root: +[Valid][16][UTT][ MSE = 0.2065 | LCC = 0.9522 | SRCC = 0.9528 ] +INFO:root:Average loss=0.06711208987236023 +INFO:root: +[Valid][17][UTT][ MSE = 0.2741 | LCC = 0.9491 | SRCC = 0.9515 ] +INFO:root:Average loss=0.06439852470159531 +INFO:root: +[Valid][18][UTT][ MSE = 0.2180 | LCC = 0.9525 | SRCC = 0.9537 ] +INFO:root:Average loss=0.06246338814496994 +INFO:root: +[Valid][19][UTT][ MSE = 0.2397 | LCC = 0.9498 | SRCC = 0.9513 ] +INFO:root:Average loss=0.061339945435523985 +INFO:root: +[Valid][20][UTT][ MSE = 0.2191 | LCC = 0.9475 | SRCC = 0.9494 ] +INFO:root:Average loss=0.058814699336886406 +INFO:root: +[Valid][21][UTT][ MSE = 0.2262 | LCC = 0.9511 | SRCC = 0.9517 ] +INFO:root:Average loss=0.052870135843753815 +INFO:root: +[Valid][22][UTT][ MSE = 0.2279 | LCC = 0.9534 | SRCC = 0.9541 ] +INFO:root:Average loss=0.05249299532175064 +INFO:root: +[Valid][23][UTT][ MSE = 0.2498 | LCC = 0.9531 | SRCC = 0.9542 ] +INFO:root:Average loss=0.05361617854237556 +INFO:root: +[Valid][24][UTT][ MSE = 0.2156 | LCC = 0.9519 | SRCC = 0.9527 ] +INFO:root:Average loss=0.05333548824489116 +INFO:root: +[Valid][25][UTT][ MSE = 0.2781 | LCC = 0.9530 | SRCC = 0.9536 ] +INFO:root:Average loss=0.048378186851739884 +INFO:root: +[Valid][26][UTT][ MSE = 0.2229 | LCC = 0.9525 | SRCC = 0.9532 ] +INFO:root:Average loss=0.04693363696336746 +INFO:root: +[Valid][27][UTT][ MSE = 0.2257 | LCC = 0.9544 | SRCC = 0.9549 ] +INFO:root:Average loss=0.04660462480783462 +INFO:root: +[Valid][28][UTT][ MSE = 0.2237 | LCC = 0.9542 | SRCC = 0.9546 ] +INFO:root:Average loss=0.04400846502184868 +INFO:root: +[Valid][29][UTT][ MSE = 0.2292 | LCC = 0.9512 | SRCC = 0.9529 ] +INFO:root:Average loss=0.04312684372067452 +INFO:root: +[Valid][30][UTT][ MSE = 0.2124 | LCC = 0.9538 | SRCC = 0.9539 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2300 | LCC = 0.9541 | SRCC = 0.9552 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..27b3255ea56823d6ec1af74ce6480e2078257be5 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8ae90ddf0be43100a2679ecbea130cef5bd8cbe7bf31509b25f54b22b8511c5e +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..1f1feaf399b130bb7d47a7c4ca37bde1c997aa5f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aca2a7386e349dd7631ae126937051e64d800afa65f3af072829852994d66a83 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c055f718451d8a19b3210f358eb647542e3a9b7c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46459856307506564 +INFO:root: +[Valid][0][UTT][ MSE = 0.2869 | LCC = 0.9246 | SRCC = 0.9262 ] +INFO:root:Average loss=0.18894573777914048 +INFO:root: +[Valid][1][UTT][ MSE = 0.2701 | LCC = 0.9364 | SRCC = 0.9377 ] +INFO:root:Average loss=0.15589474385976793 +INFO:root: +[Valid][2][UTT][ MSE = 0.2874 | LCC = 0.9439 | SRCC = 0.9465 ] +INFO:root:Average loss=0.1382767550945282 +INFO:root: +[Valid][3][UTT][ MSE = 0.2397 | LCC = 0.9471 | SRCC = 0.9488 ] +INFO:root:Average loss=0.13020514500141145 +INFO:root: +[Valid][4][UTT][ MSE = 0.2427 | LCC = 0.9460 | SRCC = 0.9478 ] +INFO:root:Average loss=0.12349942547082901 +INFO:root: +[Valid][5][UTT][ MSE = 0.2439 | LCC = 0.9496 | SRCC = 0.9508 ] +INFO:root:Average loss=0.111974534034729 +INFO:root: +[Valid][6][UTT][ MSE = 0.2188 | LCC = 0.9521 | SRCC = 0.9533 ] +INFO:root:Average loss=0.10678492987155914 +INFO:root: +[Valid][7][UTT][ MSE = 0.2514 | LCC = 0.9499 | SRCC = 0.9522 ] +INFO:root:Average loss=0.10127689671516418 +INFO:root: +[Valid][8][UTT][ MSE = 0.2400 | LCC = 0.9524 | SRCC = 0.9530 ] +INFO:root:Average loss=0.09579245194792747 +INFO:root: +[Valid][9][UTT][ MSE = 0.2095 | LCC = 0.9524 | SRCC = 0.9538 ] +INFO:root:Average loss=0.09259192660450935 +INFO:root: +[Valid][10][UTT][ MSE = 0.2171 | LCC = 0.9515 | SRCC = 0.9524 ] +INFO:root:Average loss=0.09043845182657242 +INFO:root: +[Valid][11][UTT][ MSE = 0.2239 | LCC = 0.9535 | SRCC = 0.9543 ] +INFO:root:Average loss=0.08273971727490426 +INFO:root: +[Valid][12][UTT][ MSE = 0.2246 | LCC = 0.9544 | SRCC = 0.9549 ] +INFO:root:Average loss=0.07931618797779083 +INFO:root: +[Valid][13][UTT][ MSE = 0.2145 | LCC = 0.9539 | SRCC = 0.9549 ] +INFO:root:Average loss=0.07560688653588295 +INFO:root: +[Valid][14][UTT][ MSE = 0.2050 | LCC = 0.9536 | SRCC = 0.9539 ] +INFO:root:Average loss=0.07381805378198623 +INFO:root: +[Valid][15][UTT][ MSE = 0.2309 | LCC = 0.9541 | SRCC = 0.9547 ] +INFO:root:Average loss=0.06854941767454148 +INFO:root: +[Valid][16][UTT][ MSE = 0.2117 | LCC = 0.9535 | SRCC = 0.9548 ] +INFO:root:Average loss=0.06764546096324921 +INFO:root: +[Valid][17][UTT][ MSE = 0.2181 | LCC = 0.9516 | SRCC = 0.9529 ] +INFO:root:Average loss=0.06393024264276027 +INFO:root: +[Valid][18][UTT][ MSE = 0.2137 | LCC = 0.9550 | SRCC = 0.9556 ] +INFO:root:Average loss=0.060211755275726316 +INFO:root: +[Valid][19][UTT][ MSE = 0.2250 | LCC = 0.9523 | SRCC = 0.9543 ] +INFO:root:Average loss=0.060050109505653385 +INFO:root: +[Valid][20][UTT][ MSE = 0.2180 | LCC = 0.9549 | SRCC = 0.9556 ] +INFO:root:Average loss=0.05593120551109314 +INFO:root: +[Valid][21][UTT][ MSE = 0.2672 | LCC = 0.9517 | SRCC = 0.9528 ] +INFO:root:Average loss=0.05548521322011948 +INFO:root: +[Valid][22][UTT][ MSE = 0.2048 | LCC = 0.9536 | SRCC = 0.9534 ] +INFO:root:Average loss=0.05071151666343212 +INFO:root: +[Valid][23][UTT][ MSE = 0.2209 | LCC = 0.9540 | SRCC = 0.9543 ] +INFO:root:Average loss=0.051390826404094696 +INFO:root: +[Valid][24][UTT][ MSE = 0.2180 | LCC = 0.9540 | SRCC = 0.9545 ] +INFO:root:Average loss=0.048133090615272524 +INFO:root: +[Valid][25][UTT][ MSE = 0.2330 | LCC = 0.9530 | SRCC = 0.9535 ] +INFO:root:Average loss=0.048111342400312423 +INFO:root: +[Valid][26][UTT][ MSE = 0.2244 | LCC = 0.9491 | SRCC = 0.9506 ] +INFO:root:Average loss=0.04581630797684193 +INFO:root: +[Valid][27][UTT][ MSE = 0.2306 | LCC = 0.9534 | SRCC = 0.9539 ] +INFO:root:Average loss=0.0438236408084631 +INFO:root: +[Valid][28][UTT][ MSE = 0.2256 | LCC = 0.9544 | SRCC = 0.9552 ] +INFO:root:Average loss=0.043437341913580896 +INFO:root: +[Valid][29][UTT][ MSE = 0.2310 | LCC = 0.9509 | SRCC = 0.9513 ] +INFO:root:Average loss=0.041721360623836516 +INFO:root: +[Valid][30][UTT][ MSE = 0.2329 | LCC = 0.9539 | SRCC = 0.9555 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2192 | LCC = 0.9541 | SRCC = 0.9553 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2b4c120b5a99a57a0d74f35b98fcd78d76bb1d06 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a199ff31f298618b2db8ae0dfcae8735e8eaed1f67bbcf7d2c4f20b72d320dc8 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a54a96056cc47b481293bfe7fd496c4d08745546 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:48d797ce3f6e3548362163fe8ce94d08a24e24247e6bf5678290e3b4c7f36b42 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d3daa40f8ab4d01578aa6e39c7d65fb79f58aa2f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.34819504189491274 +INFO:root: +[Valid][0][UTT][ MSE = 0.3072 | LCC = 0.9240 | SRCC = 0.9261 ] +INFO:root:Average loss=0.19021975654363632 +INFO:root: +[Valid][1][UTT][ MSE = 0.2571 | LCC = 0.9373 | SRCC = 0.9396 ] +INFO:root:Average loss=0.16342935550212861 +INFO:root: +[Valid][2][UTT][ MSE = 0.3078 | LCC = 0.9427 | SRCC = 0.9474 ] +INFO:root:Average loss=0.1414556373357773 +INFO:root: +[Valid][3][UTT][ MSE = 0.2602 | LCC = 0.9461 | SRCC = 0.9484 ] +INFO:root:Average loss=0.13087890619039536 +INFO:root: +[Valid][4][UTT][ MSE = 0.2463 | LCC = 0.9482 | SRCC = 0.9512 ] +INFO:root:Average loss=0.12395320779085159 +INFO:root: +[Valid][5][UTT][ MSE = 0.2576 | LCC = 0.9482 | SRCC = 0.9502 ] +INFO:root:Average loss=0.11348430532217026 +INFO:root: +[Valid][6][UTT][ MSE = 0.2077 | LCC = 0.9474 | SRCC = 0.9483 ] +INFO:root:Average loss=0.10649880722165107 +INFO:root: +[Valid][7][UTT][ MSE = 0.2392 | LCC = 0.9484 | SRCC = 0.9505 ] +INFO:root:Average loss=0.09863660302758216 +INFO:root: +[Valid][8][UTT][ MSE = 0.2035 | LCC = 0.9506 | SRCC = 0.9516 ] +INFO:root:Average loss=0.09453171265125275 +INFO:root: +[Valid][9][UTT][ MSE = 0.2018 | LCC = 0.9517 | SRCC = 0.9523 ] +INFO:root:Average loss=0.09303064036369324 +INFO:root: +[Valid][10][UTT][ MSE = 0.2325 | LCC = 0.9503 | SRCC = 0.9510 ] +INFO:root:Average loss=0.08683363369107247 +INFO:root: +[Valid][11][UTT][ MSE = 0.2500 | LCC = 0.9484 | SRCC = 0.9496 ] +INFO:root:Average loss=0.0884852094054222 +INFO:root: +[Valid][12][UTT][ MSE = 0.1906 | LCC = 0.9517 | SRCC = 0.9527 ] +INFO:root:Average loss=0.07657169923186302 +INFO:root: +[Valid][13][UTT][ MSE = 0.2555 | LCC = 0.9458 | SRCC = 0.9490 ] +INFO:root:Average loss=0.07636092624068261 +INFO:root: +[Valid][14][UTT][ MSE = 0.2373 | LCC = 0.9533 | SRCC = 0.9543 ] +INFO:root:Average loss=0.07422048935294151 +INFO:root: +[Valid][15][UTT][ MSE = 0.2176 | LCC = 0.9499 | SRCC = 0.9510 ] +INFO:root:Average loss=0.06950309672951699 +INFO:root: +[Valid][16][UTT][ MSE = 0.2393 | LCC = 0.9542 | SRCC = 0.9549 ] +INFO:root:Average loss=0.06587564179301261 +INFO:root: +[Valid][17][UTT][ MSE = 0.2028 | LCC = 0.9526 | SRCC = 0.9530 ] +INFO:root:Average loss=0.06360327431559562 +INFO:root: +[Valid][18][UTT][ MSE = 0.1978 | LCC = 0.9535 | SRCC = 0.9543 ] +INFO:root:Average loss=0.06296480894088745 +INFO:root: +[Valid][19][UTT][ MSE = 0.1992 | LCC = 0.9536 | SRCC = 0.9540 ] +INFO:root:Average loss=0.060157086193561554 +INFO:root: +[Valid][20][UTT][ MSE = 0.2385 | LCC = 0.9533 | SRCC = 0.9549 ] +INFO:root:Average loss=0.05814656573534012 +INFO:root: +[Valid][21][UTT][ MSE = 0.2129 | LCC = 0.9540 | SRCC = 0.9549 ] +INFO:root:Average loss=0.05556665831804276 +INFO:root: +[Valid][22][UTT][ MSE = 0.2035 | LCC = 0.9511 | SRCC = 0.9517 ] +INFO:root:Average loss=0.055490629389882085 +INFO:root: +[Valid][23][UTT][ MSE = 0.2419 | LCC = 0.9525 | SRCC = 0.9543 ] +INFO:root:Average loss=0.05053889909386635 +INFO:root: +[Valid][24][UTT][ MSE = 0.2266 | LCC = 0.9524 | SRCC = 0.9529 ] +INFO:root:Average loss=0.04968774105608463 +INFO:root: +[Valid][25][UTT][ MSE = 0.2117 | LCC = 0.9540 | SRCC = 0.9547 ] +INFO:root:Average loss=0.04752098110318184 +INFO:root: +[Valid][26][UTT][ MSE = 0.1974 | LCC = 0.9540 | SRCC = 0.9548 ] +INFO:root:Average loss=0.04952722501754761 +INFO:root: +[Valid][27][UTT][ MSE = 0.2159 | LCC = 0.9534 | SRCC = 0.9536 ] +INFO:root:Average loss=0.045785805597901345 +INFO:root: +[Valid][28][UTT][ MSE = 0.1914 | LCC = 0.9511 | SRCC = 0.9523 ] +INFO:root:Average loss=0.04516926130652427 +INFO:root: +[Valid][29][UTT][ MSE = 0.2316 | LCC = 0.9521 | SRCC = 0.9532 ] +INFO:root:Average loss=0.04371372798085213 +INFO:root: +[Valid][30][UTT][ MSE = 0.2200 | LCC = 0.9543 | SRCC = 0.9549 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2293 | LCC = 0.9520 | SRCC = 0.9545 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ba851599e86ef64cd073365f6be5f234c7b572c6 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7019778bb4da798843abd0e5cee3aa7345903bf68af4fc00b499099657284adb +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..685ae610543b82ea67e01618e5eaf76b0aa53015 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4ca9de17336e6680a78d35ffa4aa8bfb9e7801c53bd0bf774d45857a3c2dc47a +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d7a91657d1d5118e2367ef0d05c871acc3484347 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer8/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3626426478624344 +INFO:root: +[Valid][0][UTT][ MSE = 0.2902 | LCC = 0.9262 | SRCC = 0.9289 ] +INFO:root:Average loss=0.18226851677894593 +INFO:root: +[Valid][1][UTT][ MSE = 0.2817 | LCC = 0.9390 | SRCC = 0.9422 ] +INFO:root:Average loss=0.15029009991884232 +INFO:root: +[Valid][2][UTT][ MSE = 0.2113 | LCC = 0.9438 | SRCC = 0.9460 ] +INFO:root:Average loss=0.13922085094451905 +INFO:root: +[Valid][3][UTT][ MSE = 0.2252 | LCC = 0.9469 | SRCC = 0.9483 ] +INFO:root:Average loss=0.12838545268774032 +INFO:root: +[Valid][4][UTT][ MSE = 0.2409 | LCC = 0.9418 | SRCC = 0.9455 ] +INFO:root:Average loss=0.11848302358388901 +INFO:root: +[Valid][5][UTT][ MSE = 0.2018 | LCC = 0.9474 | SRCC = 0.9498 ] +INFO:root:Average loss=0.11502853083610534 +INFO:root: +[Valid][6][UTT][ MSE = 0.2265 | LCC = 0.9456 | SRCC = 0.9467 ] +INFO:root:Average loss=0.10816139981150627 +INFO:root: +[Valid][7][UTT][ MSE = 0.2503 | LCC = 0.9469 | SRCC = 0.9500 ] +INFO:root:Average loss=0.09962704509496689 +INFO:root: +[Valid][8][UTT][ MSE = 0.2266 | LCC = 0.9503 | SRCC = 0.9522 ] +INFO:root:Average loss=0.09464568856358528 +INFO:root: +[Valid][9][UTT][ MSE = 0.2449 | LCC = 0.9506 | SRCC = 0.9524 ] +INFO:root:Average loss=0.08858411914110184 +INFO:root: +[Valid][10][UTT][ MSE = 0.2145 | LCC = 0.9509 | SRCC = 0.9525 ] +INFO:root:Average loss=0.08592538049817085 +INFO:root: +[Valid][11][UTT][ MSE = 0.2010 | LCC = 0.9498 | SRCC = 0.9522 ] +INFO:root:Average loss=0.08366714492440223 +INFO:root: +[Valid][12][UTT][ MSE = 0.2119 | LCC = 0.9514 | SRCC = 0.9534 ] +INFO:root:Average loss=0.07784932053089141 +INFO:root: +[Valid][13][UTT][ MSE = 0.1895 | LCC = 0.9518 | SRCC = 0.9529 ] +INFO:root:Average loss=0.07740395897626877 +INFO:root: +[Valid][14][UTT][ MSE = 0.2184 | LCC = 0.9487 | SRCC = 0.9507 ] +INFO:root:Average loss=0.06968799632787705 +INFO:root: +[Valid][15][UTT][ MSE = 0.2461 | LCC = 0.9518 | SRCC = 0.9532 ] +INFO:root:Average loss=0.07259088256955147 +INFO:root: +[Valid][16][UTT][ MSE = 0.2122 | LCC = 0.9496 | SRCC = 0.9504 ] +INFO:root:Average loss=0.065323023468256 +INFO:root: +[Valid][17][UTT][ MSE = 0.2143 | LCC = 0.9517 | SRCC = 0.9526 ] +INFO:root:Average loss=0.06099117937684059 +INFO:root: +[Valid][18][UTT][ MSE = 0.1910 | LCC = 0.9497 | SRCC = 0.9518 ] +INFO:root:Average loss=0.06264378181099892 +INFO:root: +[Valid][19][UTT][ MSE = 0.2186 | LCC = 0.9507 | SRCC = 0.9520 ] +INFO:root:Average loss=0.05586300620436668 +INFO:root: +[Valid][20][UTT][ MSE = 0.1937 | LCC = 0.9525 | SRCC = 0.9533 ] +INFO:root:Average loss=0.05385167649388313 +INFO:root: +[Valid][21][UTT][ MSE = 0.2303 | LCC = 0.9517 | SRCC = 0.9534 ] +INFO:root:Average loss=0.054095169991254806 +INFO:root: +[Valid][22][UTT][ MSE = 0.1906 | LCC = 0.9512 | SRCC = 0.9521 ] +INFO:root:Average loss=0.052861405476927754 +INFO:root: +[Valid][23][UTT][ MSE = 0.1812 | LCC = 0.9499 | SRCC = 0.9511 ] +INFO:root:Average loss=0.05074354399740696 +INFO:root: +[Valid][24][UTT][ MSE = 0.2157 | LCC = 0.9536 | SRCC = 0.9548 ] +INFO:root:Average loss=0.049030815184116366 +INFO:root: +[Valid][25][UTT][ MSE = 0.2059 | LCC = 0.9519 | SRCC = 0.9530 ] +INFO:root:Average loss=0.046913313195109364 +INFO:root: +[Valid][26][UTT][ MSE = 0.2092 | LCC = 0.9523 | SRCC = 0.9527 ] +INFO:root:Average loss=0.045731461003422735 +INFO:root: +[Valid][27][UTT][ MSE = 0.2086 | LCC = 0.9512 | SRCC = 0.9521 ] +INFO:root:Average loss=0.04456591689586639 +INFO:root: +[Valid][28][UTT][ MSE = 0.1940 | LCC = 0.9540 | SRCC = 0.9546 ] +INFO:root:Average loss=0.04103993345797062 +INFO:root: +[Valid][29][UTT][ MSE = 0.2335 | LCC = 0.9522 | SRCC = 0.9526 ] +INFO:root:Average loss=0.042427859872579576 +INFO:root: +[Valid][30][UTT][ MSE = 0.2280 | LCC = 0.9513 | SRCC = 0.9519 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1966 | LCC = 0.9554 | SRCC = 0.9568 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f9316fe04818c394b0fe1932f47ed4a4e2e2aab4 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8d39240aa66e7a61193ed116cff09273b9b8cfff11e72952f08de5ccfcee41a6 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..934d6cd5ada5921cfac8f729232349c935317205 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:100bb55a91bfaa6d1da20e61b6bab67a4db1681f00ab71555b617c196a948cf8 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..11f1d28eda63e2d1cd78f204e9af90a54bbaf4df --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.35699978935718535 +INFO:root: +[Valid][0][UTT][ MSE = 0.3203 | LCC = 0.9259 | SRCC = 0.9272 ] +INFO:root:Average loss=0.18752329295873643 +INFO:root: +[Valid][1][UTT][ MSE = 0.2620 | LCC = 0.9376 | SRCC = 0.9407 ] +INFO:root:Average loss=0.1673127804994583 +INFO:root: +[Valid][2][UTT][ MSE = 0.2262 | LCC = 0.9420 | SRCC = 0.9442 ] +INFO:root:Average loss=0.14537589782476426 +INFO:root: +[Valid][3][UTT][ MSE = 0.2586 | LCC = 0.9461 | SRCC = 0.9489 ] +INFO:root:Average loss=0.13527960878610612 +INFO:root: +[Valid][4][UTT][ MSE = 0.2322 | LCC = 0.9462 | SRCC = 0.9492 ] +INFO:root:Average loss=0.13053855830430985 +INFO:root: +[Valid][5][UTT][ MSE = 0.2662 | LCC = 0.9470 | SRCC = 0.9489 ] +INFO:root:Average loss=0.11974332171678544 +INFO:root: +[Valid][6][UTT][ MSE = 0.2592 | LCC = 0.9477 | SRCC = 0.9500 ] +INFO:root:Average loss=0.11571659398078918 +INFO:root: +[Valid][7][UTT][ MSE = 0.2180 | LCC = 0.9482 | SRCC = 0.9495 ] +INFO:root:Average loss=0.10720683252811432 +INFO:root: +[Valid][8][UTT][ MSE = 0.2005 | LCC = 0.9503 | SRCC = 0.9523 ] +INFO:root:Average loss=0.10243010523915291 +INFO:root: +[Valid][9][UTT][ MSE = 0.2206 | LCC = 0.9517 | SRCC = 0.9523 ] +INFO:root:Average loss=0.0991221079826355 +INFO:root: +[Valid][10][UTT][ MSE = 0.2108 | LCC = 0.9512 | SRCC = 0.9539 ] +INFO:root:Average loss=0.09804875740408897 +INFO:root: +[Valid][11][UTT][ MSE = 0.2160 | LCC = 0.9526 | SRCC = 0.9533 ] +INFO:root:Average loss=0.09294919484853745 +INFO:root: +[Valid][12][UTT][ MSE = 0.2202 | LCC = 0.9512 | SRCC = 0.9521 ] +INFO:root:Average loss=0.08752188688516617 +INFO:root: +[Valid][13][UTT][ MSE = 0.2122 | LCC = 0.9501 | SRCC = 0.9502 ] +INFO:root:Average loss=0.08453096410632134 +INFO:root: +[Valid][14][UTT][ MSE = 0.2093 | LCC = 0.9492 | SRCC = 0.9506 ] +INFO:root:Average loss=0.08271618375182152 +INFO:root: +[Valid][15][UTT][ MSE = 0.1995 | LCC = 0.9523 | SRCC = 0.9542 ] +INFO:root:Average loss=0.07775633171200752 +INFO:root: +[Valid][16][UTT][ MSE = 0.2045 | LCC = 0.9524 | SRCC = 0.9526 ] +INFO:root:Average loss=0.07815837088227272 +INFO:root: +[Valid][17][UTT][ MSE = 0.2145 | LCC = 0.9517 | SRCC = 0.9526 ] +INFO:root:Average loss=0.07439595392346382 +INFO:root: +[Valid][18][UTT][ MSE = 0.2080 | LCC = 0.9506 | SRCC = 0.9524 ] +INFO:root:Average loss=0.07409505811333657 +INFO:root: +[Valid][19][UTT][ MSE = 0.2048 | LCC = 0.9519 | SRCC = 0.9529 ] +INFO:root:Average loss=0.07154082551598549 +INFO:root: +[Valid][20][UTT][ MSE = 0.2390 | LCC = 0.9522 | SRCC = 0.9530 ] +INFO:root:Average loss=0.06701615595817566 +INFO:root: +[Valid][21][UTT][ MSE = 0.2218 | LCC = 0.9517 | SRCC = 0.9518 ] +INFO:root:Average loss=0.06327238556742668 +INFO:root: +[Valid][22][UTT][ MSE = 0.2092 | LCC = 0.9531 | SRCC = 0.9537 ] +INFO:root:Average loss=0.0626496179997921 +INFO:root: +[Valid][23][UTT][ MSE = 0.2082 | LCC = 0.9527 | SRCC = 0.9534 ] +INFO:root:Average loss=0.06546469354629517 +INFO:root: +[Valid][24][UTT][ MSE = 0.1951 | LCC = 0.9536 | SRCC = 0.9548 ] +INFO:root:Average loss=0.06091588142514229 +INFO:root: +[Valid][25][UTT][ MSE = 0.2004 | LCC = 0.9531 | SRCC = 0.9546 ] +INFO:root:Average loss=0.05880323511362076 +INFO:root: +[Valid][26][UTT][ MSE = 0.2236 | LCC = 0.9516 | SRCC = 0.9521 ] +INFO:root:Average loss=0.05706383264064789 +INFO:root: +[Valid][27][UTT][ MSE = 0.2209 | LCC = 0.9516 | SRCC = 0.9526 ] +INFO:root:Average loss=0.059761785954236986 +INFO:root: +[Valid][28][UTT][ MSE = 0.2118 | LCC = 0.9526 | SRCC = 0.9530 ] +INFO:root:Average loss=0.0542477745115757 +INFO:root: +[Valid][29][UTT][ MSE = 0.1988 | LCC = 0.9521 | SRCC = 0.9527 ] +INFO:root:Average loss=0.05621429693698883 +INFO:root: +[Valid][30][UTT][ MSE = 0.2341 | LCC = 0.9532 | SRCC = 0.9540 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1892 | LCC = 0.9587 | SRCC = 0.9602 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2ba9343dbd44bb885ddcdb6ed3a62c1fe96897a0 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:34d6cbad4519af50020c55b52e65cf27002d01a7b836dbe1a013f3ca78c09928 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..491b40ea517b4616dfbdddaf6e0997dd87a67cc5 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:13f91ca85607af662551cf3556761324c3b1ff5bcb10a39edffdb5ee1f2bedd1 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3b9f8c1f1e7e074b501fadd2c8e08c35d54ac1f9 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3531851681470871 +INFO:root: +[Valid][0][UTT][ MSE = 0.3053 | LCC = 0.9285 | SRCC = 0.9308 ] +INFO:root:Average loss=0.1916282371878624 +INFO:root: +[Valid][1][UTT][ MSE = 0.2514 | LCC = 0.9353 | SRCC = 0.9404 ] +INFO:root:Average loss=0.16302703177928923 +INFO:root: +[Valid][2][UTT][ MSE = 0.2429 | LCC = 0.9401 | SRCC = 0.9422 ] +INFO:root:Average loss=0.15268572092056273 +INFO:root: +[Valid][3][UTT][ MSE = 0.2487 | LCC = 0.9414 | SRCC = 0.9442 ] +INFO:root:Average loss=0.14087506210803985 +INFO:root: +[Valid][4][UTT][ MSE = 0.2595 | LCC = 0.9459 | SRCC = 0.9478 ] +INFO:root:Average loss=0.13006026947498323 +INFO:root: +[Valid][5][UTT][ MSE = 0.2735 | LCC = 0.9460 | SRCC = 0.9500 ] +INFO:root:Average loss=0.12199526220560074 +INFO:root: +[Valid][6][UTT][ MSE = 0.2375 | LCC = 0.9474 | SRCC = 0.9497 ] +INFO:root:Average loss=0.12222370862960816 +INFO:root: +[Valid][7][UTT][ MSE = 0.2173 | LCC = 0.9495 | SRCC = 0.9507 ] +INFO:root:Average loss=0.11398882257938385 +INFO:root: +[Valid][8][UTT][ MSE = 0.2372 | LCC = 0.9516 | SRCC = 0.9530 ] +INFO:root:Average loss=0.10909336537122727 +INFO:root: +[Valid][9][UTT][ MSE = 0.2183 | LCC = 0.9499 | SRCC = 0.9512 ] +INFO:root:Average loss=0.10245987397432327 +INFO:root: +[Valid][10][UTT][ MSE = 0.2274 | LCC = 0.9481 | SRCC = 0.9509 ] +INFO:root:Average loss=0.09985648894309998 +INFO:root: +[Valid][11][UTT][ MSE = 0.2253 | LCC = 0.9515 | SRCC = 0.9529 ] +INFO:root:Average loss=0.09296002697944641 +INFO:root: +[Valid][12][UTT][ MSE = 0.2272 | LCC = 0.9496 | SRCC = 0.9535 ] +INFO:root:Average loss=0.09075033074617386 +INFO:root: +[Valid][13][UTT][ MSE = 0.2367 | LCC = 0.9492 | SRCC = 0.9504 ] +INFO:root:Average loss=0.08890243056416512 +INFO:root: +[Valid][14][UTT][ MSE = 0.2126 | LCC = 0.9540 | SRCC = 0.9554 ] +INFO:root:Average loss=0.08496330899000168 +INFO:root: +[Valid][15][UTT][ MSE = 0.2261 | LCC = 0.9526 | SRCC = 0.9534 ] +INFO:root:Average loss=0.08629249960184097 +INFO:root: +[Valid][16][UTT][ MSE = 0.2362 | LCC = 0.9505 | SRCC = 0.9527 ] +INFO:root:Average loss=0.07979719951748848 +INFO:root: +[Valid][17][UTT][ MSE = 0.2189 | LCC = 0.9536 | SRCC = 0.9543 ] +INFO:root:Average loss=0.07741267478466034 +INFO:root: +[Valid][18][UTT][ MSE = 0.2547 | LCC = 0.9476 | SRCC = 0.9487 ] +INFO:root:Average loss=0.07601898995041848 +INFO:root: +[Valid][19][UTT][ MSE = 0.2285 | LCC = 0.9502 | SRCC = 0.9516 ] +INFO:root:Average loss=0.07114845699071884 +INFO:root: +[Valid][20][UTT][ MSE = 0.2136 | LCC = 0.9536 | SRCC = 0.9550 ] +INFO:root:Average loss=0.07197442474961281 +INFO:root: +[Valid][21][UTT][ MSE = 0.2434 | LCC = 0.9503 | SRCC = 0.9523 ] +INFO:root:Average loss=0.0703341164290905 +INFO:root: +[Valid][22][UTT][ MSE = 0.2243 | LCC = 0.9505 | SRCC = 0.9518 ] +INFO:root:Average loss=0.06939190918207169 +INFO:root: +[Valid][23][UTT][ MSE = 0.2003 | LCC = 0.9519 | SRCC = 0.9536 ] +INFO:root:Average loss=0.06324914500117301 +INFO:root: +[Valid][24][UTT][ MSE = 0.1965 | LCC = 0.9537 | SRCC = 0.9543 ] +INFO:root:Average loss=0.06536224120855331 +INFO:root: +[Valid][25][UTT][ MSE = 0.2002 | LCC = 0.9537 | SRCC = 0.9547 ] +INFO:root:Average loss=0.061659978419542315 +INFO:root: +[Valid][26][UTT][ MSE = 0.2027 | LCC = 0.9525 | SRCC = 0.9532 ] +INFO:root:Average loss=0.06281926813721657 +INFO:root: +[Valid][27][UTT][ MSE = 0.2267 | LCC = 0.9513 | SRCC = 0.9521 ] +INFO:root:Average loss=0.059936147719621656 +INFO:root: +[Valid][28][UTT][ MSE = 0.2323 | LCC = 0.9520 | SRCC = 0.9534 ] +INFO:root:Average loss=0.057917327344417575 +INFO:root: +[Valid][29][UTT][ MSE = 0.2091 | LCC = 0.9528 | SRCC = 0.9537 ] +INFO:root:Average loss=0.05461907097697258 +INFO:root: +[Valid][30][UTT][ MSE = 0.2158 | LCC = 0.9529 | SRCC = 0.9539 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2154 | LCC = 0.9573 | SRCC = 0.9589 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c2e5fef8b85a03e8ebca1a1bae0a7ede0b339027 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1c75787391102bfad1ce3d51122ad544283a2e2228b96a2908a33b3aa42a889b +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6c29c308832a6d55a45c158ad39d1c8b02efdd44 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:628a47974f1cf5f4fd4aad63292e1633fe08dfd007302c3ed96714360dfb6a02 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..16bb65398d5559127e410d106197b444fdabb79b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.33150491774082186 +INFO:root: +[Valid][0][UTT][ MSE = 0.2722 | LCC = 0.9194 | SRCC = 0.9243 ] +INFO:root:Average loss=0.18799919575452806 +INFO:root: +[Valid][1][UTT][ MSE = 0.2582 | LCC = 0.9320 | SRCC = 0.9371 ] +INFO:root:Average loss=0.16703034323453902 +INFO:root: +[Valid][2][UTT][ MSE = 0.2815 | LCC = 0.9409 | SRCC = 0.9433 ] +INFO:root:Average loss=0.14978729963302612 +INFO:root: +[Valid][3][UTT][ MSE = 0.2393 | LCC = 0.9421 | SRCC = 0.9460 ] +INFO:root:Average loss=0.14048668044805526 +INFO:root: +[Valid][4][UTT][ MSE = 0.2563 | LCC = 0.9426 | SRCC = 0.9466 ] +INFO:root:Average loss=0.1297797354757786 +INFO:root: +[Valid][5][UTT][ MSE = 0.2614 | LCC = 0.9483 | SRCC = 0.9508 ] +INFO:root:Average loss=0.1234441688656807 +INFO:root: +[Valid][6][UTT][ MSE = 0.2654 | LCC = 0.9463 | SRCC = 0.9489 ] +INFO:root:Average loss=0.1150764965415001 +INFO:root: +[Valid][7][UTT][ MSE = 0.2362 | LCC = 0.9493 | SRCC = 0.9511 ] +INFO:root:Average loss=0.10899504590034485 +INFO:root: +[Valid][8][UTT][ MSE = 0.2219 | LCC = 0.9500 | SRCC = 0.9516 ] +INFO:root:Average loss=0.11197624403238296 +INFO:root: +[Valid][9][UTT][ MSE = 0.2776 | LCC = 0.9463 | SRCC = 0.9481 ] +INFO:root:Average loss=0.10278315669298171 +INFO:root: +[Valid][10][UTT][ MSE = 0.2123 | LCC = 0.9480 | SRCC = 0.9517 ] +INFO:root:Average loss=0.09739012691378593 +INFO:root: +[Valid][11][UTT][ MSE = 0.2090 | LCC = 0.9517 | SRCC = 0.9531 ] +INFO:root:Average loss=0.09407022613286972 +INFO:root: +[Valid][12][UTT][ MSE = 0.2377 | LCC = 0.9461 | SRCC = 0.9497 ] +INFO:root:Average loss=0.09108651396632195 +INFO:root: +[Valid][13][UTT][ MSE = 0.2320 | LCC = 0.9473 | SRCC = 0.9495 ] +INFO:root:Average loss=0.08860067966580391 +INFO:root: +[Valid][14][UTT][ MSE = 0.2251 | LCC = 0.9517 | SRCC = 0.9540 ] +INFO:root:Average loss=0.0856282670199871 +INFO:root: +[Valid][15][UTT][ MSE = 0.2054 | LCC = 0.9488 | SRCC = 0.9510 ] +INFO:root:Average loss=0.0827737752199173 +INFO:root: +[Valid][16][UTT][ MSE = 0.2302 | LCC = 0.9470 | SRCC = 0.9492 ] +INFO:root:Average loss=0.08208924105763435 +INFO:root: +[Valid][17][UTT][ MSE = 0.1962 | LCC = 0.9507 | SRCC = 0.9541 ] +INFO:root:Average loss=0.07713314828276634 +INFO:root: +[Valid][18][UTT][ MSE = 0.2034 | LCC = 0.9507 | SRCC = 0.9522 ] +INFO:root:Average loss=0.07421728175878525 +INFO:root: +[Valid][19][UTT][ MSE = 0.2089 | LCC = 0.9480 | SRCC = 0.9490 ] +INFO:root:Average loss=0.07436045303940773 +INFO:root: +[Valid][20][UTT][ MSE = 0.2316 | LCC = 0.9505 | SRCC = 0.9524 ] +INFO:root:Average loss=0.06856267547607423 +INFO:root: +[Valid][21][UTT][ MSE = 0.2172 | LCC = 0.9531 | SRCC = 0.9542 ] +INFO:root:Average loss=0.06949113687872886 +INFO:root: +[Valid][22][UTT][ MSE = 0.2327 | LCC = 0.9516 | SRCC = 0.9523 ] +INFO:root:Average loss=0.06620137384533882 +INFO:root: +[Valid][23][UTT][ MSE = 0.2159 | LCC = 0.9525 | SRCC = 0.9531 ] +INFO:root:Average loss=0.06612128752470016 +INFO:root: +[Valid][24][UTT][ MSE = 0.2048 | LCC = 0.9530 | SRCC = 0.9541 ] +INFO:root:Average loss=0.06118071913719177 +INFO:root: +[Valid][25][UTT][ MSE = 0.2160 | LCC = 0.9505 | SRCC = 0.9525 ] +INFO:root:Average loss=0.06133281993865967 +INFO:root: +[Valid][26][UTT][ MSE = 0.2068 | LCC = 0.9501 | SRCC = 0.9516 ] +INFO:root:Average loss=0.0576343689262867 +INFO:root: +[Valid][27][UTT][ MSE = 0.2134 | LCC = 0.9500 | SRCC = 0.9510 ] +INFO:root:Average loss=0.05869892290234566 +INFO:root: +[Valid][28][UTT][ MSE = 0.2119 | LCC = 0.9504 | SRCC = 0.9506 ] +INFO:root:Average loss=0.05885292699933052 +INFO:root: +[Valid][29][UTT][ MSE = 0.2013 | LCC = 0.9527 | SRCC = 0.9535 ] +INFO:root:Average loss=0.051422727361321446 +INFO:root: +[Valid][30][UTT][ MSE = 0.2232 | LCC = 0.9506 | SRCC = 0.9507 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2213 | LCC = 0.9554 | SRCC = 0.9567 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7383e96328e67a7f1b8b4983ff2e38a0584bd82e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c3596c27122bc636106fd65921b3645ea35aa9986500882e0bbbe185fe844e45 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..31a28e1095c5d42450b35f422973174406cd3fc5 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:094664fc2adbc2c2d61eb4652534ebaa11759a7992d1afd5c3e8f95a2bdbd780 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1892b0e5bd8554015bfa354fe1a4739932b44be7 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3414685404896736 +INFO:root: +[Valid][0][UTT][ MSE = 0.3045 | LCC = 0.9211 | SRCC = 0.9280 ] +INFO:root:Average loss=0.1849859619140625 +INFO:root: +[Valid][1][UTT][ MSE = 0.2812 | LCC = 0.9389 | SRCC = 0.9412 ] +INFO:root:Average loss=0.17223060834407805 +INFO:root: +[Valid][2][UTT][ MSE = 0.2935 | LCC = 0.9388 | SRCC = 0.9428 ] +INFO:root:Average loss=0.14743355363607408 +INFO:root: +[Valid][3][UTT][ MSE = 0.2333 | LCC = 0.9394 | SRCC = 0.9435 ] +INFO:root:Average loss=0.1381827169060707 +INFO:root: +[Valid][4][UTT][ MSE = 0.2455 | LCC = 0.9435 | SRCC = 0.9459 ] +INFO:root:Average loss=0.12700683170557023 +INFO:root: +[Valid][5][UTT][ MSE = 0.2226 | LCC = 0.9467 | SRCC = 0.9496 ] +INFO:root:Average loss=0.12295842033624649 +INFO:root: +[Valid][6][UTT][ MSE = 0.2137 | LCC = 0.9481 | SRCC = 0.9502 ] +INFO:root:Average loss=0.11586453914642333 +INFO:root: +[Valid][7][UTT][ MSE = 0.2227 | LCC = 0.9481 | SRCC = 0.9519 ] +INFO:root:Average loss=0.10815437361598015 +INFO:root: +[Valid][8][UTT][ MSE = 0.2518 | LCC = 0.9478 | SRCC = 0.9505 ] +INFO:root:Average loss=0.10556991577148438 +INFO:root: +[Valid][9][UTT][ MSE = 0.2315 | LCC = 0.9489 | SRCC = 0.9515 ] +INFO:root:Average loss=0.10004620775580406 +INFO:root: +[Valid][10][UTT][ MSE = 0.2186 | LCC = 0.9511 | SRCC = 0.9522 ] +INFO:root:Average loss=0.09726959607005119 +INFO:root: +[Valid][11][UTT][ MSE = 0.2088 | LCC = 0.9515 | SRCC = 0.9528 ] +INFO:root:Average loss=0.09261023890972138 +INFO:root: +[Valid][12][UTT][ MSE = 0.2286 | LCC = 0.9492 | SRCC = 0.9522 ] +INFO:root:Average loss=0.08905469635128975 +INFO:root: +[Valid][13][UTT][ MSE = 0.1932 | LCC = 0.9527 | SRCC = 0.9545 ] +INFO:root:Average loss=0.08661129799485207 +INFO:root: +[Valid][14][UTT][ MSE = 0.2139 | LCC = 0.9534 | SRCC = 0.9542 ] +INFO:root:Average loss=0.08154126715660096 +INFO:root: +[Valid][15][UTT][ MSE = 0.2286 | LCC = 0.9502 | SRCC = 0.9523 ] +INFO:root:Average loss=0.07933334738016129 +INFO:root: +[Valid][16][UTT][ MSE = 0.2101 | LCC = 0.9515 | SRCC = 0.9530 ] +INFO:root:Average loss=0.07607685655355453 +INFO:root: +[Valid][17][UTT][ MSE = 0.2078 | LCC = 0.9503 | SRCC = 0.9516 ] +INFO:root:Average loss=0.07373991820216179 +INFO:root: +[Valid][18][UTT][ MSE = 0.2168 | LCC = 0.9523 | SRCC = 0.9545 ] +INFO:root:Average loss=0.07381205254793168 +INFO:root: +[Valid][19][UTT][ MSE = 0.2008 | LCC = 0.9479 | SRCC = 0.9512 ] +INFO:root:Average loss=0.07060438191890717 +INFO:root: +[Valid][20][UTT][ MSE = 0.1976 | LCC = 0.9537 | SRCC = 0.9542 ] +INFO:root:Average loss=0.06872892478108406 +INFO:root: +[Valid][21][UTT][ MSE = 0.2097 | LCC = 0.9534 | SRCC = 0.9540 ] +INFO:root:Average loss=0.06502419456839562 +INFO:root: +[Valid][22][UTT][ MSE = 0.2107 | LCC = 0.9534 | SRCC = 0.9542 ] +INFO:root:Average loss=0.06429591649770737 +INFO:root: +[Valid][23][UTT][ MSE = 0.1931 | LCC = 0.9532 | SRCC = 0.9539 ] +INFO:root:Average loss=0.06508260348439217 +INFO:root: +[Valid][24][UTT][ MSE = 0.2003 | LCC = 0.9507 | SRCC = 0.9517 ] +INFO:root:Average loss=0.06374369940161705 +INFO:root: +[Valid][25][UTT][ MSE = 0.1891 | LCC = 0.9526 | SRCC = 0.9532 ] +INFO:root:Average loss=0.05839028468728066 +INFO:root: +[Valid][26][UTT][ MSE = 0.1957 | LCC = 0.9536 | SRCC = 0.9552 ] +INFO:root:Average loss=0.05663471624255181 +INFO:root: +[Valid][27][UTT][ MSE = 0.1911 | LCC = 0.9535 | SRCC = 0.9544 ] +INFO:root:Average loss=0.0569037379026413 +INFO:root: +[Valid][28][UTT][ MSE = 0.2263 | LCC = 0.9535 | SRCC = 0.9540 ] +INFO:root:Average loss=0.05564570446312427 +INFO:root: +[Valid][29][UTT][ MSE = 0.1895 | LCC = 0.9545 | SRCC = 0.9548 ] +INFO:root:Average loss=0.053005217045545575 +INFO:root: +[Valid][30][UTT][ MSE = 0.2263 | LCC = 0.9495 | SRCC = 0.9505 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1887 | LCC = 0.9566 | SRCC = 0.9583 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..846d94e60617cc5adc1854b8935c02d644739e2b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:26e653fbe0bc2e280ebcad223c7ae9fffb698d52caee32d78627a524e850e40d +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a92bd8ffc793e0fe5bd999a363e335738be664d0 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3287114daaa7d62e8ebc72813c7a15d5bf851bb1b5680ffcfb8145dd3d49c1a0 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..fa07ee4bed8784fb620cd7f4dd779b674f8bbf45 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_1epoch/layer9/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.31600039803981783 +INFO:root: +[Valid][0][UTT][ MSE = 0.2898 | LCC = 0.9281 | SRCC = 0.9298 ] +INFO:root:Average loss=0.18810705220699311 +INFO:root: +[Valid][1][UTT][ MSE = 0.2404 | LCC = 0.9389 | SRCC = 0.9404 ] +INFO:root:Average loss=0.16404403299093245 +INFO:root: +[Valid][2][UTT][ MSE = 0.2441 | LCC = 0.9382 | SRCC = 0.9428 ] +INFO:root:Average loss=0.14609589391946792 +INFO:root: +[Valid][3][UTT][ MSE = 0.2247 | LCC = 0.9441 | SRCC = 0.9472 ] +INFO:root:Average loss=0.13476836061477662 +INFO:root: +[Valid][4][UTT][ MSE = 0.2405 | LCC = 0.9441 | SRCC = 0.9473 ] +INFO:root:Average loss=0.1303654638528824 +INFO:root: +[Valid][5][UTT][ MSE = 0.2493 | LCC = 0.9493 | SRCC = 0.9510 ] +INFO:root:Average loss=0.11797622787952423 +INFO:root: +[Valid][6][UTT][ MSE = 0.2117 | LCC = 0.9482 | SRCC = 0.9511 ] +INFO:root:Average loss=0.11488485091924668 +INFO:root: +[Valid][7][UTT][ MSE = 0.2274 | LCC = 0.9496 | SRCC = 0.9518 ] +INFO:root:Average loss=0.111002107411623 +INFO:root: +[Valid][8][UTT][ MSE = 0.2388 | LCC = 0.9496 | SRCC = 0.9511 ] +INFO:root:Average loss=0.10666505843400956 +INFO:root: +[Valid][9][UTT][ MSE = 0.1925 | LCC = 0.9495 | SRCC = 0.9528 ] +INFO:root:Average loss=0.10084576290845872 +INFO:root: +[Valid][10][UTT][ MSE = 0.2272 | LCC = 0.9493 | SRCC = 0.9517 ] +INFO:root:Average loss=0.09528447246551514 +INFO:root: +[Valid][11][UTT][ MSE = 0.2205 | LCC = 0.9526 | SRCC = 0.9539 ] +INFO:root:Average loss=0.09200021609663964 +INFO:root: +[Valid][12][UTT][ MSE = 0.1865 | LCC = 0.9518 | SRCC = 0.9533 ] +INFO:root:Average loss=0.0883315002322197 +INFO:root: +[Valid][13][UTT][ MSE = 0.2024 | LCC = 0.9514 | SRCC = 0.9539 ] +INFO:root:Average loss=0.08649074944853782 +INFO:root: +[Valid][14][UTT][ MSE = 0.2192 | LCC = 0.9527 | SRCC = 0.9539 ] +INFO:root:Average loss=0.08353441017866135 +INFO:root: +[Valid][15][UTT][ MSE = 0.1975 | LCC = 0.9523 | SRCC = 0.9534 ] +INFO:root:Average loss=0.07753622516989708 +INFO:root: +[Valid][16][UTT][ MSE = 0.1991 | LCC = 0.9521 | SRCC = 0.9543 ] +INFO:root:Average loss=0.0749312601685524 +INFO:root: +[Valid][17][UTT][ MSE = 0.2069 | LCC = 0.9522 | SRCC = 0.9533 ] +INFO:root:Average loss=0.07449931013584137 +INFO:root: +[Valid][18][UTT][ MSE = 0.2050 | LCC = 0.9490 | SRCC = 0.9501 ] +INFO:root:Average loss=0.0714966915845871 +INFO:root: +[Valid][19][UTT][ MSE = 0.1975 | LCC = 0.9517 | SRCC = 0.9531 ] +INFO:root:Average loss=0.0701470872759819 +INFO:root: +[Valid][20][UTT][ MSE = 0.2059 | LCC = 0.9520 | SRCC = 0.9528 ] +INFO:root:Average loss=0.06707534644007683 +INFO:root: +[Valid][21][UTT][ MSE = 0.1978 | LCC = 0.9529 | SRCC = 0.9542 ] +INFO:root:Average loss=0.06813260379433632 +INFO:root: +[Valid][22][UTT][ MSE = 0.2211 | LCC = 0.9513 | SRCC = 0.9522 ] +INFO:root:Average loss=0.0652239807844162 +INFO:root: +[Valid][23][UTT][ MSE = 0.1983 | LCC = 0.9523 | SRCC = 0.9534 ] +INFO:root:Average loss=0.06112687730789185 +INFO:root: +[Valid][24][UTT][ MSE = 0.1995 | LCC = 0.9521 | SRCC = 0.9531 ] +INFO:root:Average loss=0.05955331316590309 +INFO:root: +[Valid][25][UTT][ MSE = 0.2155 | LCC = 0.9516 | SRCC = 0.9531 ] +INFO:root:Average loss=0.05908935305476189 +INFO:root: +[Valid][26][UTT][ MSE = 0.2026 | LCC = 0.9513 | SRCC = 0.9526 ] +INFO:root:Average loss=0.05396010422706604 +INFO:root: +[Valid][27][UTT][ MSE = 0.2083 | LCC = 0.9523 | SRCC = 0.9533 ] +INFO:root:Average loss=0.05634060567617417 +INFO:root: +[Valid][28][UTT][ MSE = 0.2000 | LCC = 0.9519 | SRCC = 0.9525 ] +INFO:root:Average loss=0.05577262619137764 +INFO:root: +[Valid][29][UTT][ MSE = 0.1901 | LCC = 0.9505 | SRCC = 0.9524 ] +INFO:root:Average loss=0.053091096252202985 +INFO:root: +[Valid][30][UTT][ MSE = 0.1934 | LCC = 0.9513 | SRCC = 0.9520 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1984 | LCC = 0.9564 | SRCC = 0.9579 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ef3b269ec6f1ef35d07357c5b08085b5fc258508 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:93fdd4094909a2389d30482079029ebb76a34f4d0b0829453e0f7b38b94e81e7 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9310988153ec74cfd7ddd2ab1856a3fc7fa7db23 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:59cda2b564f18695531f200f40df689b9a3f7e4eb675ba73a6777e078fb34e42 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..19d4304c0bb0d6f758ef1516eba2966d97f89bfd --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.494384445309639 +INFO:root: +[Valid][0][UTT][ MSE = 0.3154 | LCC = 0.9050 | SRCC = 0.9070 ] +INFO:root:Average loss=0.24269290018081666 +INFO:root: +[Valid][1][UTT][ MSE = 0.3018 | LCC = 0.9222 | SRCC = 0.9247 ] +INFO:root:Average loss=0.20740264666080474 +INFO:root: +[Valid][2][UTT][ MSE = 0.2691 | LCC = 0.9295 | SRCC = 0.9320 ] +INFO:root:Average loss=0.17591745883226395 +INFO:root: +[Valid][3][UTT][ MSE = 0.2643 | LCC = 0.9378 | SRCC = 0.9389 ] +INFO:root:Average loss=0.1636647173166275 +INFO:root: +[Valid][4][UTT][ MSE = 0.2495 | LCC = 0.9385 | SRCC = 0.9401 ] +INFO:root:Average loss=0.14473668313026428 +INFO:root: +[Valid][5][UTT][ MSE = 0.2305 | LCC = 0.9420 | SRCC = 0.9437 ] +INFO:root:Average loss=0.13416014623641967 +INFO:root: +[Valid][6][UTT][ MSE = 0.2175 | LCC = 0.9421 | SRCC = 0.9435 ] +INFO:root:Average loss=0.1262117406129837 +INFO:root: +[Valid][7][UTT][ MSE = 0.2259 | LCC = 0.9446 | SRCC = 0.9461 ] +INFO:root:Average loss=0.12037474289536476 +INFO:root: +[Valid][8][UTT][ MSE = 0.2238 | LCC = 0.9466 | SRCC = 0.9477 ] +INFO:root:Average loss=0.1136349352300167 +INFO:root: +[Valid][9][UTT][ MSE = 0.2122 | LCC = 0.9475 | SRCC = 0.9487 ] +INFO:root:Average loss=0.11127658706903458 +INFO:root: +[Valid][10][UTT][ MSE = 0.2234 | LCC = 0.9469 | SRCC = 0.9489 ] +INFO:root:Average loss=0.10187796780467033 +INFO:root: +[Valid][11][UTT][ MSE = 0.2285 | LCC = 0.9479 | SRCC = 0.9491 ] +INFO:root:Average loss=0.09381412160396575 +INFO:root: +[Valid][12][UTT][ MSE = 0.2044 | LCC = 0.9484 | SRCC = 0.9499 ] +INFO:root:Average loss=0.09388641789555549 +INFO:root: +[Valid][13][UTT][ MSE = 0.1867 | LCC = 0.9502 | SRCC = 0.9517 ] +INFO:root:Average loss=0.08785476049780845 +INFO:root: +[Valid][14][UTT][ MSE = 0.2203 | LCC = 0.9489 | SRCC = 0.9501 ] +INFO:root:Average loss=0.08789532679319381 +INFO:root: +[Valid][15][UTT][ MSE = 0.1926 | LCC = 0.9505 | SRCC = 0.9524 ] +INFO:root:Average loss=0.08529668682813644 +INFO:root: +[Valid][16][UTT][ MSE = 0.2226 | LCC = 0.9517 | SRCC = 0.9527 ] +INFO:root:Average loss=0.07874932682514191 +INFO:root: +[Valid][17][UTT][ MSE = 0.2011 | LCC = 0.9523 | SRCC = 0.9531 ] +INFO:root:Average loss=0.0757429983317852 +INFO:root: +[Valid][18][UTT][ MSE = 0.2063 | LCC = 0.9513 | SRCC = 0.9527 ] +INFO:root:Average loss=0.07004423031210899 +INFO:root: +[Valid][19][UTT][ MSE = 0.2116 | LCC = 0.9495 | SRCC = 0.9504 ] +INFO:root:Average loss=0.07070749270915985 +INFO:root: +[Valid][20][UTT][ MSE = 0.2154 | LCC = 0.9526 | SRCC = 0.9536 ] +INFO:root:Average loss=0.06596742272377014 +INFO:root: +[Valid][21][UTT][ MSE = 0.2380 | LCC = 0.9529 | SRCC = 0.9543 ] +INFO:root:Average loss=0.06631094267964363 +INFO:root: +[Valid][22][UTT][ MSE = 0.2088 | LCC = 0.9525 | SRCC = 0.9539 ] +INFO:root:Average loss=0.06524488024413586 +INFO:root: +[Valid][23][UTT][ MSE = 0.2036 | LCC = 0.9534 | SRCC = 0.9545 ] +INFO:root:Average loss=0.05941160935163498 +INFO:root: +[Valid][24][UTT][ MSE = 0.1930 | LCC = 0.9530 | SRCC = 0.9537 ] +INFO:root:Average loss=0.056928825750947 +INFO:root: +[Valid][25][UTT][ MSE = 0.1963 | LCC = 0.9522 | SRCC = 0.9533 ] +INFO:root:Average loss=0.05748575001955032 +INFO:root: +[Valid][26][UTT][ MSE = 0.1988 | LCC = 0.9538 | SRCC = 0.9550 ] +INFO:root:Average loss=0.056298536151647566 +INFO:root: +[Valid][27][UTT][ MSE = 0.2073 | LCC = 0.9533 | SRCC = 0.9544 ] +INFO:root:Average loss=0.05608313456177712 +INFO:root: +[Valid][28][UTT][ MSE = 0.2119 | LCC = 0.9547 | SRCC = 0.9559 ] +INFO:root:Average loss=0.05581942591071129 +INFO:root: +[Valid][29][UTT][ MSE = 0.1945 | LCC = 0.9548 | SRCC = 0.9562 ] +INFO:root:Average loss=0.05224558882415295 +INFO:root: +[Valid][30][UTT][ MSE = 0.1883 | LCC = 0.9536 | SRCC = 0.9548 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1940 | LCC = 0.9570 | SRCC = 0.9579 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7a161eacc0d27bededa1dc8463cd306c3ff01c53 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9c010c90cfc287ec539a452e1c3c0753ef591c09029524ec37a0a7c2c363a813 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..084b14cddcce7393aa1992f2bf2c9f178bde91b9 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bfaa6edbd882d88c7fc831bc613038f92fa8d3778e1e3b5bcea0670b7bdb8917 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3b435ac3c7f3dd440a50571c7f5d60cb96c29507 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6408195750713348 +INFO:root: +[Valid][0][UTT][ MSE = 0.3528 | LCC = 0.8970 | SRCC = 0.8976 ] +INFO:root:Average loss=0.2519067157506943 +INFO:root: +[Valid][1][UTT][ MSE = 0.2630 | LCC = 0.9238 | SRCC = 0.9265 ] +INFO:root:Average loss=0.2028078436255455 +INFO:root: +[Valid][2][UTT][ MSE = 0.2599 | LCC = 0.9297 | SRCC = 0.9311 ] +INFO:root:Average loss=0.1768379318714142 +INFO:root: +[Valid][3][UTT][ MSE = 0.2532 | LCC = 0.9341 | SRCC = 0.9363 ] +INFO:root:Average loss=0.16006737446784974 +INFO:root: +[Valid][4][UTT][ MSE = 0.2119 | LCC = 0.9393 | SRCC = 0.9418 ] +INFO:root:Average loss=0.1441127724647522 +INFO:root: +[Valid][5][UTT][ MSE = 0.2620 | LCC = 0.9430 | SRCC = 0.9445 ] +INFO:root:Average loss=0.1400240575671196 +INFO:root: +[Valid][6][UTT][ MSE = 0.1943 | LCC = 0.9417 | SRCC = 0.9436 ] +INFO:root:Average loss=0.12514630898833276 +INFO:root: +[Valid][7][UTT][ MSE = 0.2249 | LCC = 0.9461 | SRCC = 0.9475 ] +INFO:root:Average loss=0.12020890474319458 +INFO:root: +[Valid][8][UTT][ MSE = 0.2074 | LCC = 0.9473 | SRCC = 0.9487 ] +INFO:root:Average loss=0.11671304798126221 +INFO:root: +[Valid][9][UTT][ MSE = 0.2120 | LCC = 0.9492 | SRCC = 0.9511 ] +INFO:root:Average loss=0.1088497828245163 +INFO:root: +[Valid][10][UTT][ MSE = 0.1789 | LCC = 0.9502 | SRCC = 0.9520 ] +INFO:root:Average loss=0.10209456741809846 +INFO:root: +[Valid][11][UTT][ MSE = 0.1818 | LCC = 0.9485 | SRCC = 0.9501 ] +INFO:root:Average loss=0.09573487284779549 +INFO:root: +[Valid][12][UTT][ MSE = 0.1981 | LCC = 0.9507 | SRCC = 0.9520 ] +INFO:root:Average loss=0.09517885279655457 +INFO:root: +[Valid][13][UTT][ MSE = 0.1947 | LCC = 0.9513 | SRCC = 0.9523 ] +INFO:root:Average loss=0.08771763563156128 +INFO:root: +[Valid][14][UTT][ MSE = 0.1988 | LCC = 0.9515 | SRCC = 0.9527 ] +INFO:root:Average loss=0.08488858395814895 +INFO:root: +[Valid][15][UTT][ MSE = 0.1672 | LCC = 0.9497 | SRCC = 0.9526 ] +INFO:root:Average loss=0.082753875374794 +INFO:root: +[Valid][16][UTT][ MSE = 0.1914 | LCC = 0.9512 | SRCC = 0.9520 ] +INFO:root:Average loss=0.07867987242341042 +INFO:root: +[Valid][17][UTT][ MSE = 0.1711 | LCC = 0.9529 | SRCC = 0.9544 ] +INFO:root:Average loss=0.07438462290167809 +INFO:root: +[Valid][18][UTT][ MSE = 0.1782 | LCC = 0.9523 | SRCC = 0.9531 ] +INFO:root:Average loss=0.07131885114312173 +INFO:root: +[Valid][19][UTT][ MSE = 0.1996 | LCC = 0.9539 | SRCC = 0.9552 ] +INFO:root:Average loss=0.07154967305064201 +INFO:root: +[Valid][20][UTT][ MSE = 0.1835 | LCC = 0.9540 | SRCC = 0.9548 ] +INFO:root:Average loss=0.06622391703724861 +INFO:root: +[Valid][21][UTT][ MSE = 0.1687 | LCC = 0.9517 | SRCC = 0.9528 ] +INFO:root:Average loss=0.06649587252736092 +INFO:root: +[Valid][22][UTT][ MSE = 0.1679 | LCC = 0.9516 | SRCC = 0.9550 ] +INFO:root:Average loss=0.06555596113204956 +INFO:root: +[Valid][23][UTT][ MSE = 0.1710 | LCC = 0.9511 | SRCC = 0.9531 ] +INFO:root:Average loss=0.06325250604748726 +INFO:root: +[Valid][24][UTT][ MSE = 0.1688 | LCC = 0.9522 | SRCC = 0.9546 ] +INFO:root:Average loss=0.059674049288034436 +INFO:root: +[Valid][25][UTT][ MSE = 0.1650 | LCC = 0.9533 | SRCC = 0.9556 ] +INFO:root:Average loss=0.05920306473970413 +INFO:root: +[Valid][26][UTT][ MSE = 0.1704 | LCC = 0.9545 | SRCC = 0.9551 ] +INFO:root:Average loss=0.059165849208831785 +INFO:root: +[Valid][27][UTT][ MSE = 0.1722 | LCC = 0.9545 | SRCC = 0.9561 ] +INFO:root:Average loss=0.05357003879547119 +INFO:root: +[Valid][28][UTT][ MSE = 0.1836 | LCC = 0.9546 | SRCC = 0.9552 ] +INFO:root:Average loss=0.05435469478368759 +INFO:root: +[Valid][29][UTT][ MSE = 0.1761 | LCC = 0.9539 | SRCC = 0.9558 ] +INFO:root:Average loss=0.05234866440296173 +INFO:root: +[Valid][30][UTT][ MSE = 0.1701 | LCC = 0.9568 | SRCC = 0.9575 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1752 | LCC = 0.9556 | SRCC = 0.9562 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bd228fea72cf05622527d3d19e8dcf497a03b960 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:30464168038cab32a548dd08220ed9fd140d1f3c0bb8a10124f724d035e69176 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..5d84f1d7ddd910642006f8db9367cded55271be4 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ce30197233d15be34212ca1b769d66650c88d2f69476c95dda054eadabc84b11 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..47bba83d69853e79388eee3bcf69455fba909d51 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5604454100131988 +INFO:root: +[Valid][0][UTT][ MSE = 0.3761 | LCC = 0.8946 | SRCC = 0.8960 ] +INFO:root:Average loss=0.2513816796541214 +INFO:root: +[Valid][1][UTT][ MSE = 0.2898 | LCC = 0.9216 | SRCC = 0.9243 ] +INFO:root:Average loss=0.21093139535188674 +INFO:root: +[Valid][2][UTT][ MSE = 0.2911 | LCC = 0.9288 | SRCC = 0.9306 ] +INFO:root:Average loss=0.17651208728551865 +INFO:root: +[Valid][3][UTT][ MSE = 0.3015 | LCC = 0.9337 | SRCC = 0.9360 ] +INFO:root:Average loss=0.16643233054876327 +INFO:root: +[Valid][4][UTT][ MSE = 0.2318 | LCC = 0.9386 | SRCC = 0.9414 ] +INFO:root:Average loss=0.15527827495336533 +INFO:root: +[Valid][5][UTT][ MSE = 0.2565 | LCC = 0.9425 | SRCC = 0.9440 ] +INFO:root:Average loss=0.1444030237197876 +INFO:root: +[Valid][6][UTT][ MSE = 0.2559 | LCC = 0.9418 | SRCC = 0.9444 ] +INFO:root:Average loss=0.1296231433749199 +INFO:root: +[Valid][7][UTT][ MSE = 0.2678 | LCC = 0.9429 | SRCC = 0.9444 ] +INFO:root:Average loss=0.13175793552398682 +INFO:root: +[Valid][8][UTT][ MSE = 0.2267 | LCC = 0.9461 | SRCC = 0.9479 ] +INFO:root:Average loss=0.12297363317012787 +INFO:root: +[Valid][9][UTT][ MSE = 0.2437 | LCC = 0.9478 | SRCC = 0.9494 ] +INFO:root:Average loss=0.11268791103363038 +INFO:root: +[Valid][10][UTT][ MSE = 0.2130 | LCC = 0.9482 | SRCC = 0.9502 ] +INFO:root:Average loss=0.10581791210174561 +INFO:root: +[Valid][11][UTT][ MSE = 0.2699 | LCC = 0.9468 | SRCC = 0.9485 ] +INFO:root:Average loss=0.09911538875102997 +INFO:root: +[Valid][12][UTT][ MSE = 0.2262 | LCC = 0.9499 | SRCC = 0.9511 ] +INFO:root:Average loss=0.0977169460952282 +INFO:root: +[Valid][13][UTT][ MSE = 0.2128 | LCC = 0.9504 | SRCC = 0.9522 ] +INFO:root:Average loss=0.09377087637782097 +INFO:root: +[Valid][14][UTT][ MSE = 0.2104 | LCC = 0.9472 | SRCC = 0.9490 ] +INFO:root:Average loss=0.08655714938044548 +INFO:root: +[Valid][15][UTT][ MSE = 0.2408 | LCC = 0.9506 | SRCC = 0.9513 ] +INFO:root:Average loss=0.08502242505550385 +INFO:root: +[Valid][16][UTT][ MSE = 0.2225 | LCC = 0.9491 | SRCC = 0.9501 ] +INFO:root:Average loss=0.08198450964689255 +INFO:root: +[Valid][17][UTT][ MSE = 0.2083 | LCC = 0.9515 | SRCC = 0.9529 ] +INFO:root:Average loss=0.07964264640212058 +INFO:root: +[Valid][18][UTT][ MSE = 0.2189 | LCC = 0.9513 | SRCC = 0.9526 ] +INFO:root:Average loss=0.07566532737016678 +INFO:root: +[Valid][19][UTT][ MSE = 0.2516 | LCC = 0.9487 | SRCC = 0.9502 ] +INFO:root:Average loss=0.07683645728230476 +INFO:root: +[Valid][20][UTT][ MSE = 0.2051 | LCC = 0.9507 | SRCC = 0.9518 ] +INFO:root:Average loss=0.06682324907183647 +INFO:root: +[Valid][21][UTT][ MSE = 0.1752 | LCC = 0.9514 | SRCC = 0.9543 ] +INFO:root:Average loss=0.06758657273650169 +INFO:root: +[Valid][22][UTT][ MSE = 0.2010 | LCC = 0.9528 | SRCC = 0.9543 ] +INFO:root:Average loss=0.06459805431962014 +INFO:root: +[Valid][23][UTT][ MSE = 0.1854 | LCC = 0.9537 | SRCC = 0.9551 ] +INFO:root:Average loss=0.06106165960431099 +INFO:root: +[Valid][24][UTT][ MSE = 0.2102 | LCC = 0.9540 | SRCC = 0.9547 ] +INFO:root:Average loss=0.06050471252202988 +INFO:root: +[Valid][25][UTT][ MSE = 0.2146 | LCC = 0.9531 | SRCC = 0.9538 ] +INFO:root:Average loss=0.06270685169100762 +INFO:root: +[Valid][26][UTT][ MSE = 0.2046 | LCC = 0.9529 | SRCC = 0.9534 ] +INFO:root:Average loss=0.05670363238453865 +INFO:root: +[Valid][27][UTT][ MSE = 0.1991 | LCC = 0.9537 | SRCC = 0.9550 ] +INFO:root:Average loss=0.05540425725281239 +INFO:root: +[Valid][28][UTT][ MSE = 0.1980 | LCC = 0.9537 | SRCC = 0.9548 ] +INFO:root:Average loss=0.056522105395793915 +INFO:root: +[Valid][29][UTT][ MSE = 0.2504 | LCC = 0.9540 | SRCC = 0.9552 ] +INFO:root:Average loss=0.05440563103556633 +INFO:root: +[Valid][30][UTT][ MSE = 0.1842 | LCC = 0.9559 | SRCC = 0.9572 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1826 | LCC = 0.9588 | SRCC = 0.9592 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3fe301479e6b50a2361dba8b9820a510e53e248b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e629ac3b798856052c232a27f17fb3ee85bded02d755e481de8643e24cacbc68 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..5a76227d412e28c9f7fe689046c12d132a3205f2 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6915c89f543f178580ec0f4fb41178755f488f9a37e2f5d2deeac736983d844e +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..510777dd9d561d6bd1bd12188241b9b28904ab62 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5909262309074402 +INFO:root: +[Valid][0][UTT][ MSE = 0.3412 | LCC = 0.8977 | SRCC = 0.9022 ] +INFO:root:Average loss=0.24451898682117462 +INFO:root: +[Valid][1][UTT][ MSE = 0.3172 | LCC = 0.9243 | SRCC = 0.9255 ] +INFO:root:Average loss=0.20416161841154098 +INFO:root: +[Valid][2][UTT][ MSE = 0.2937 | LCC = 0.9298 | SRCC = 0.9318 ] +INFO:root:Average loss=0.171036923289299 +INFO:root: +[Valid][3][UTT][ MSE = 0.2382 | LCC = 0.9376 | SRCC = 0.9394 ] +INFO:root:Average loss=0.1594510021805763 +INFO:root: +[Valid][4][UTT][ MSE = 0.2301 | LCC = 0.9381 | SRCC = 0.9398 ] +INFO:root:Average loss=0.14303830999135972 +INFO:root: +[Valid][5][UTT][ MSE = 0.2190 | LCC = 0.9395 | SRCC = 0.9407 ] +INFO:root:Average loss=0.1336745474934578 +INFO:root: +[Valid][6][UTT][ MSE = 0.2365 | LCC = 0.9444 | SRCC = 0.9460 ] +INFO:root:Average loss=0.13520059943199159 +INFO:root: +[Valid][7][UTT][ MSE = 0.2624 | LCC = 0.9452 | SRCC = 0.9471 ] +INFO:root:Average loss=0.1185574049949646 +INFO:root: +[Valid][8][UTT][ MSE = 0.2240 | LCC = 0.9468 | SRCC = 0.9482 ] +INFO:root:Average loss=0.11691705459356308 +INFO:root: +[Valid][9][UTT][ MSE = 0.2138 | LCC = 0.9466 | SRCC = 0.9478 ] +INFO:root:Average loss=0.10906050810217857 +INFO:root: +[Valid][10][UTT][ MSE = 0.2212 | LCC = 0.9452 | SRCC = 0.9460 ] +INFO:root:Average loss=0.10450141218304634 +INFO:root: +[Valid][11][UTT][ MSE = 0.1941 | LCC = 0.9466 | SRCC = 0.9484 ] +INFO:root:Average loss=0.09950383445620536 +INFO:root: +[Valid][12][UTT][ MSE = 0.1915 | LCC = 0.9494 | SRCC = 0.9509 ] +INFO:root:Average loss=0.09525548204779626 +INFO:root: +[Valid][13][UTT][ MSE = 0.2214 | LCC = 0.9507 | SRCC = 0.9520 ] +INFO:root:Average loss=0.08795954191684723 +INFO:root: +[Valid][14][UTT][ MSE = 0.2391 | LCC = 0.9475 | SRCC = 0.9483 ] +INFO:root:Average loss=0.087048394292593 +INFO:root: +[Valid][15][UTT][ MSE = 0.2300 | LCC = 0.9497 | SRCC = 0.9510 ] +INFO:root:Average loss=0.08472292593121529 +INFO:root: +[Valid][16][UTT][ MSE = 0.2010 | LCC = 0.9504 | SRCC = 0.9511 ] +INFO:root:Average loss=0.07873646456003189 +INFO:root: +[Valid][17][UTT][ MSE = 0.2258 | LCC = 0.9518 | SRCC = 0.9527 ] +INFO:root:Average loss=0.07703364086151122 +INFO:root: +[Valid][18][UTT][ MSE = 0.2144 | LCC = 0.9525 | SRCC = 0.9532 ] +INFO:root:Average loss=0.0719699901342392 +INFO:root: +[Valid][19][UTT][ MSE = 0.2073 | LCC = 0.9534 | SRCC = 0.9544 ] +INFO:root:Average loss=0.0709581505060196 +INFO:root: +[Valid][20][UTT][ MSE = 0.2176 | LCC = 0.9521 | SRCC = 0.9532 ] +INFO:root:Average loss=0.06598455291986466 +INFO:root: +[Valid][21][UTT][ MSE = 0.2172 | LCC = 0.9543 | SRCC = 0.9553 ] +INFO:root:Average loss=0.06751970830559731 +INFO:root: +[Valid][22][UTT][ MSE = 0.1956 | LCC = 0.9534 | SRCC = 0.9542 ] +INFO:root:Average loss=0.07032349568605423 +INFO:root: +[Valid][23][UTT][ MSE = 0.2281 | LCC = 0.9491 | SRCC = 0.9508 ] +INFO:root:Average loss=0.06084752401709557 +INFO:root: +[Valid][24][UTT][ MSE = 0.2198 | LCC = 0.9532 | SRCC = 0.9545 ] +INFO:root:Average loss=0.05783099380135536 +INFO:root: +[Valid][25][UTT][ MSE = 0.2134 | LCC = 0.9531 | SRCC = 0.9539 ] +INFO:root:Average loss=0.0581059385240078 +INFO:root: +[Valid][26][UTT][ MSE = 0.1925 | LCC = 0.9540 | SRCC = 0.9555 ] +INFO:root:Average loss=0.05693699821829796 +INFO:root: +[Valid][27][UTT][ MSE = 0.2049 | LCC = 0.9544 | SRCC = 0.9552 ] +INFO:root:Average loss=0.055612679570913315 +INFO:root: +[Valid][28][UTT][ MSE = 0.2004 | LCC = 0.9540 | SRCC = 0.9550 ] +INFO:root:Average loss=0.05467387172579765 +INFO:root: +[Valid][29][UTT][ MSE = 0.2007 | LCC = 0.9546 | SRCC = 0.9555 ] +INFO:root:Average loss=0.05374634401500225 +INFO:root: +[Valid][30][UTT][ MSE = 0.1960 | LCC = 0.9534 | SRCC = 0.9545 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2044 | LCC = 0.9556 | SRCC = 0.9563 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d6d19502fcdb36c29a338660fad01ae445f3b854 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4712d038053d311fbca6ba1763f157b65845ed8922eb420822568e7e094d9678 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..3bda3e81150c2d286d025793a82d74fa99995f5f --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e90c50833650005f25bed2faa105b07e0cd2aeb6bc5d5a811b7b32dad2f5e44e +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f93cfcb7fcc566542d82279e92da0fa087fc94bb --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer0/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.666659742474556 +INFO:root: +[Valid][0][UTT][ MSE = 0.4610 | LCC = 0.8856 | SRCC = 0.8913 ] +INFO:root:Average loss=0.26194700825214384 +INFO:root: +[Valid][1][UTT][ MSE = 0.3607 | LCC = 0.9154 | SRCC = 0.9215 ] +INFO:root:Average loss=0.20499116390943528 +INFO:root: +[Valid][2][UTT][ MSE = 0.3347 | LCC = 0.9254 | SRCC = 0.9287 ] +INFO:root:Average loss=0.1762750727534294 +INFO:root: +[Valid][3][UTT][ MSE = 0.2796 | LCC = 0.9345 | SRCC = 0.9360 ] +INFO:root:Average loss=0.15920547038316726 +INFO:root: +[Valid][4][UTT][ MSE = 0.2771 | LCC = 0.9387 | SRCC = 0.9401 ] +INFO:root:Average loss=0.14534761321544648 +INFO:root: +[Valid][5][UTT][ MSE = 0.2264 | LCC = 0.9369 | SRCC = 0.9390 ] +INFO:root:Average loss=0.13880151438713073 +INFO:root: +[Valid][6][UTT][ MSE = 0.3138 | LCC = 0.9410 | SRCC = 0.9433 ] +INFO:root:Average loss=0.12777202808856963 +INFO:root: +[Valid][7][UTT][ MSE = 0.2775 | LCC = 0.9426 | SRCC = 0.9451 ] +INFO:root:Average loss=0.12057858008146286 +INFO:root: +[Valid][8][UTT][ MSE = 0.2389 | LCC = 0.9456 | SRCC = 0.9467 ] +INFO:root:Average loss=0.11735507184267044 +INFO:root: +[Valid][9][UTT][ MSE = 0.2306 | LCC = 0.9469 | SRCC = 0.9479 ] +INFO:root:Average loss=0.1061612587273121 +INFO:root: +[Valid][10][UTT][ MSE = 0.2762 | LCC = 0.9462 | SRCC = 0.9474 ] +INFO:root:Average loss=0.10348297670483589 +INFO:root: +[Valid][11][UTT][ MSE = 0.2739 | LCC = 0.9470 | SRCC = 0.9486 ] +INFO:root:Average loss=0.1010675144791603 +INFO:root: +[Valid][12][UTT][ MSE = 0.2199 | LCC = 0.9503 | SRCC = 0.9506 ] +INFO:root:Average loss=0.09547651985287667 +INFO:root: +[Valid][13][UTT][ MSE = 0.2418 | LCC = 0.9503 | SRCC = 0.9507 ] +INFO:root:Average loss=0.0886446884572506 +INFO:root: +[Valid][14][UTT][ MSE = 0.2474 | LCC = 0.9505 | SRCC = 0.9514 ] +INFO:root:Average loss=0.08410595589876176 +INFO:root: +[Valid][15][UTT][ MSE = 0.2039 | LCC = 0.9516 | SRCC = 0.9520 ] +INFO:root:Average loss=0.08146147710084915 +INFO:root: +[Valid][16][UTT][ MSE = 0.2211 | LCC = 0.9534 | SRCC = 0.9540 ] +INFO:root:Average loss=0.07598590078949928 +INFO:root: +[Valid][17][UTT][ MSE = 0.2386 | LCC = 0.9521 | SRCC = 0.9530 ] +INFO:root:Average loss=0.0756042227447033 +INFO:root: +[Valid][18][UTT][ MSE = 0.2163 | LCC = 0.9532 | SRCC = 0.9542 ] +INFO:root:Average loss=0.07405531910061837 +INFO:root: +[Valid][19][UTT][ MSE = 0.2215 | LCC = 0.9530 | SRCC = 0.9542 ] +INFO:root:Average loss=0.06900795030593872 +INFO:root: +[Valid][20][UTT][ MSE = 0.2561 | LCC = 0.9517 | SRCC = 0.9531 ] +INFO:root:Average loss=0.06679015001654624 +INFO:root: +[Valid][21][UTT][ MSE = 0.2423 | LCC = 0.9531 | SRCC = 0.9542 ] +INFO:root:Average loss=0.06613256803154946 +INFO:root: +[Valid][22][UTT][ MSE = 0.1948 | LCC = 0.9543 | SRCC = 0.9548 ] +INFO:root:Average loss=0.062377174824476245 +INFO:root: +[Valid][23][UTT][ MSE = 0.2054 | LCC = 0.9541 | SRCC = 0.9547 ] +INFO:root:Average loss=0.06169612547755241 +INFO:root: +[Valid][24][UTT][ MSE = 0.2346 | LCC = 0.9509 | SRCC = 0.9518 ] +INFO:root:Average loss=0.057737328261137005 +INFO:root: +[Valid][25][UTT][ MSE = 0.2259 | LCC = 0.9554 | SRCC = 0.9557 ] +INFO:root:Average loss=0.0594765405356884 +INFO:root: +[Valid][26][UTT][ MSE = 0.2348 | LCC = 0.9546 | SRCC = 0.9550 ] +INFO:root:Average loss=0.05775196287035942 +INFO:root: +[Valid][27][UTT][ MSE = 0.2906 | LCC = 0.9507 | SRCC = 0.9537 ] +INFO:root:Average loss=0.05883470359444618 +INFO:root: +[Valid][28][UTT][ MSE = 0.2331 | LCC = 0.9534 | SRCC = 0.9539 ] +INFO:root:Average loss=0.05417452000081539 +INFO:root: +[Valid][29][UTT][ MSE = 0.2095 | LCC = 0.9552 | SRCC = 0.9554 ] +INFO:root:Average loss=0.0532923119366169 +INFO:root: +[Valid][30][UTT][ MSE = 0.2052 | LCC = 0.9571 | SRCC = 0.9572 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2111 | LCC = 0.9583 | SRCC = 0.9586 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4032df51a2f6c7e096fb23ee63cfec45e83f2603 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e927d0366046e639b3b23dd1e7e4b90bfb50ecd9b36d134067a532929edaa295 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7357432e3432985629586f145d387e6fab9240b1 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b3090ff3dadcc0c838c74c0df809cf930efe07bfa32bea65233198fd7854700 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c1c37a4b68e58134b1232e83181a62174eca46d0 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44481729531288144 +INFO:root: +[Valid][0][UTT][ MSE = 0.3037 | LCC = 0.9143 | SRCC = 0.9174 ] +INFO:root:Average loss=0.21866742223501207 +INFO:root: +[Valid][1][UTT][ MSE = 0.2555 | LCC = 0.9316 | SRCC = 0.9337 ] +INFO:root:Average loss=0.1877202265262604 +INFO:root: +[Valid][2][UTT][ MSE = 0.2718 | LCC = 0.9353 | SRCC = 0.9409 ] +INFO:root:Average loss=0.15488941663503647 +INFO:root: +[Valid][3][UTT][ MSE = 0.2373 | LCC = 0.9427 | SRCC = 0.9436 ] +INFO:root:Average loss=0.1400339560508728 +INFO:root: +[Valid][4][UTT][ MSE = 0.2067 | LCC = 0.9472 | SRCC = 0.9492 ] +INFO:root:Average loss=0.1279034076333046 +INFO:root: +[Valid][5][UTT][ MSE = 0.2355 | LCC = 0.9467 | SRCC = 0.9507 ] +INFO:root:Average loss=0.11876237291097641 +INFO:root: +[Valid][6][UTT][ MSE = 0.2102 | LCC = 0.9488 | SRCC = 0.9503 ] +INFO:root:Average loss=0.10763017162680626 +INFO:root: +[Valid][7][UTT][ MSE = 0.2048 | LCC = 0.9525 | SRCC = 0.9544 ] +INFO:root:Average loss=0.10419221299886704 +INFO:root: +[Valid][8][UTT][ MSE = 0.1930 | LCC = 0.9516 | SRCC = 0.9539 ] +INFO:root:Average loss=0.09508992463350296 +INFO:root: +[Valid][9][UTT][ MSE = 0.1875 | LCC = 0.9525 | SRCC = 0.9540 ] +INFO:root:Average loss=0.09173485547304154 +INFO:root: +[Valid][10][UTT][ MSE = 0.2144 | LCC = 0.9522 | SRCC = 0.9538 ] +INFO:root:Average loss=0.08209247097373008 +INFO:root: +[Valid][11][UTT][ MSE = 0.2154 | LCC = 0.9513 | SRCC = 0.9522 ] +INFO:root:Average loss=0.08643882092833519 +INFO:root: +[Valid][12][UTT][ MSE = 0.1907 | LCC = 0.9530 | SRCC = 0.9534 ] +INFO:root:Average loss=0.07754873767495156 +INFO:root: +[Valid][13][UTT][ MSE = 0.1867 | LCC = 0.9536 | SRCC = 0.9556 ] +INFO:root:Average loss=0.07894439974427223 +INFO:root: +[Valid][14][UTT][ MSE = 0.2394 | LCC = 0.9512 | SRCC = 0.9537 ] +INFO:root:Average loss=0.07250741821527482 +INFO:root: +[Valid][15][UTT][ MSE = 0.2050 | LCC = 0.9559 | SRCC = 0.9569 ] +INFO:root:Average loss=0.06750080749392509 +INFO:root: +[Valid][16][UTT][ MSE = 0.1901 | LCC = 0.9517 | SRCC = 0.9548 ] +INFO:root:Average loss=0.06687535184621811 +INFO:root: +[Valid][17][UTT][ MSE = 0.1723 | LCC = 0.9559 | SRCC = 0.9576 ] +INFO:root:Average loss=0.06305904948711395 +INFO:root: +[Valid][18][UTT][ MSE = 0.2107 | LCC = 0.9552 | SRCC = 0.9567 ] +INFO:root:Average loss=0.06079751744866371 +INFO:root: +[Valid][19][UTT][ MSE = 0.1835 | LCC = 0.9561 | SRCC = 0.9567 ] +INFO:root:Average loss=0.062153345197439196 +INFO:root: +[Valid][20][UTT][ MSE = 0.1808 | LCC = 0.9567 | SRCC = 0.9571 ] +INFO:root:Average loss=0.055455834120512006 +INFO:root: +[Valid][21][UTT][ MSE = 0.2146 | LCC = 0.9526 | SRCC = 0.9560 ] +INFO:root:Average loss=0.054196180552244184 +INFO:root: +[Valid][22][UTT][ MSE = 0.1872 | LCC = 0.9577 | SRCC = 0.9584 ] +INFO:root:Average loss=0.049102734953165056 +INFO:root: +[Valid][23][UTT][ MSE = 0.1771 | LCC = 0.9564 | SRCC = 0.9580 ] +INFO:root:Average loss=0.05248659712076187 +INFO:root: +[Valid][24][UTT][ MSE = 0.2007 | LCC = 0.9547 | SRCC = 0.9556 ] +INFO:root:Average loss=0.04637669813632965 +INFO:root: +[Valid][25][UTT][ MSE = 0.2087 | LCC = 0.9546 | SRCC = 0.9566 ] +INFO:root:Average loss=0.04542199146747589 +INFO:root: +[Valid][26][UTT][ MSE = 0.1981 | LCC = 0.9560 | SRCC = 0.9570 ] +INFO:root:Average loss=0.04768392352759838 +INFO:root: +[Valid][27][UTT][ MSE = 0.2038 | LCC = 0.9587 | SRCC = 0.9590 ] +INFO:root:Average loss=0.045563000336289405 +INFO:root: +[Valid][28][UTT][ MSE = 0.2056 | LCC = 0.9564 | SRCC = 0.9584 ] +INFO:root:Average loss=0.04325862763822079 +INFO:root: +[Valid][29][UTT][ MSE = 0.1684 | LCC = 0.9564 | SRCC = 0.9580 ] +INFO:root:Average loss=0.04052740615606308 +INFO:root: +[Valid][30][UTT][ MSE = 0.2162 | LCC = 0.9547 | SRCC = 0.9549 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2016 | LCC = 0.9600 | SRCC = 0.9608 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..dae657b60f1d1a82c378ddd03f558d3a21ff6800 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:30550b138e50d49adafd7b38ddd8baf14a503764c4736283ae82661f9f1820ae +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9aed855906a9a761b6930e22d5f2d7b03c8ef0c4 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b7586c030e49082e97c44c47986300a6596fc18c9e4395fdcd61684cb9e58302 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..12aed4ef60a6c9f60a5187acb4a4d4b9b71ee6d9 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5150217670202255 +INFO:root: +[Valid][0][UTT][ MSE = 0.3479 | LCC = 0.9088 | SRCC = 0.9120 ] +INFO:root:Average loss=0.21187816995382308 +INFO:root: +[Valid][1][UTT][ MSE = 0.3485 | LCC = 0.9341 | SRCC = 0.9355 ] +INFO:root:Average loss=0.17448822236061096 +INFO:root: +[Valid][2][UTT][ MSE = 0.2840 | LCC = 0.9399 | SRCC = 0.9415 ] +INFO:root:Average loss=0.14547593384981156 +INFO:root: +[Valid][3][UTT][ MSE = 0.2568 | LCC = 0.9456 | SRCC = 0.9478 ] +INFO:root:Average loss=0.1332418442964554 +INFO:root: +[Valid][4][UTT][ MSE = 0.2342 | LCC = 0.9499 | SRCC = 0.9514 ] +INFO:root:Average loss=0.12247860640287399 +INFO:root: +[Valid][5][UTT][ MSE = 0.2468 | LCC = 0.9491 | SRCC = 0.9507 ] +INFO:root:Average loss=0.11078035843372346 +INFO:root: +[Valid][6][UTT][ MSE = 0.2191 | LCC = 0.9516 | SRCC = 0.9524 ] +INFO:root:Average loss=0.10859825176000595 +INFO:root: +[Valid][7][UTT][ MSE = 0.2512 | LCC = 0.9529 | SRCC = 0.9545 ] +INFO:root:Average loss=0.10298047983646393 +INFO:root: +[Valid][8][UTT][ MSE = 0.2294 | LCC = 0.9528 | SRCC = 0.9535 ] +INFO:root:Average loss=0.09299596235156059 +INFO:root: +[Valid][9][UTT][ MSE = 0.2239 | LCC = 0.9538 | SRCC = 0.9546 ] +INFO:root:Average loss=0.08550921306014062 +INFO:root: +[Valid][10][UTT][ MSE = 0.2252 | LCC = 0.9558 | SRCC = 0.9564 ] +INFO:root:Average loss=0.08302932679653167 +INFO:root: +[Valid][11][UTT][ MSE = 0.2012 | LCC = 0.9560 | SRCC = 0.9570 ] +INFO:root:Average loss=0.08223216927051544 +INFO:root: +[Valid][12][UTT][ MSE = 0.1957 | LCC = 0.9546 | SRCC = 0.9550 ] +INFO:root:Average loss=0.07633789095282555 +INFO:root: +[Valid][13][UTT][ MSE = 0.2100 | LCC = 0.9529 | SRCC = 0.9547 ] +INFO:root:Average loss=0.076051448315382 +INFO:root: +[Valid][14][UTT][ MSE = 0.2115 | LCC = 0.9543 | SRCC = 0.9555 ] +INFO:root:Average loss=0.0672678400874138 +INFO:root: +[Valid][15][UTT][ MSE = 0.2055 | LCC = 0.9559 | SRCC = 0.9566 ] +INFO:root:Average loss=0.07396036764979362 +INFO:root: +[Valid][16][UTT][ MSE = 0.2168 | LCC = 0.9569 | SRCC = 0.9578 ] +INFO:root:Average loss=0.06303637748956681 +INFO:root: +[Valid][17][UTT][ MSE = 0.2133 | LCC = 0.9546 | SRCC = 0.9550 ] +INFO:root:Average loss=0.0607840980887413 +INFO:root: +[Valid][18][UTT][ MSE = 0.2224 | LCC = 0.9564 | SRCC = 0.9568 ] +INFO:root:Average loss=0.06005714723467827 +INFO:root: +[Valid][19][UTT][ MSE = 0.2076 | LCC = 0.9560 | SRCC = 0.9561 ] +INFO:root:Average loss=0.05448469734191894 +INFO:root: +[Valid][20][UTT][ MSE = 0.1896 | LCC = 0.9569 | SRCC = 0.9577 ] +INFO:root:Average loss=0.056158369481563566 +INFO:root: +[Valid][21][UTT][ MSE = 0.2052 | LCC = 0.9589 | SRCC = 0.9597 ] +INFO:root:Average loss=0.053050378158688544 +INFO:root: +[Valid][22][UTT][ MSE = 0.2205 | LCC = 0.9565 | SRCC = 0.9577 ] +INFO:root:Average loss=0.052451262265443804 +INFO:root: +[Valid][23][UTT][ MSE = 0.1907 | LCC = 0.9575 | SRCC = 0.9578 ] +INFO:root:Average loss=0.04768886469304562 +INFO:root: +[Valid][24][UTT][ MSE = 0.2221 | LCC = 0.9573 | SRCC = 0.9579 ] +INFO:root:Average loss=0.05055049948394299 +INFO:root: +[Valid][25][UTT][ MSE = 0.1992 | LCC = 0.9578 | SRCC = 0.9583 ] +INFO:root:Average loss=0.04675011250376701 +INFO:root: +[Valid][26][UTT][ MSE = 0.2209 | LCC = 0.9585 | SRCC = 0.9588 ] +INFO:root:Average loss=0.04292815791070461 +INFO:root: +[Valid][27][UTT][ MSE = 0.2040 | LCC = 0.9583 | SRCC = 0.9591 ] +INFO:root:Average loss=0.04157303349673748 +INFO:root: +[Valid][28][UTT][ MSE = 0.2020 | LCC = 0.9560 | SRCC = 0.9559 ] +INFO:root:Average loss=0.04174956314265728 +INFO:root: +[Valid][29][UTT][ MSE = 0.2071 | LCC = 0.9574 | SRCC = 0.9584 ] +INFO:root:Average loss=0.03968598692119121 +INFO:root: +[Valid][30][UTT][ MSE = 0.1912 | LCC = 0.9570 | SRCC = 0.9581 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1984 | LCC = 0.9599 | SRCC = 0.9607 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..cae541edf4cc1cd06b14bc59593bb29b9107a1ee --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:641a47957a249f1813e7fed5ed204607b6f8bedf600434d7e14537ea7d3e8e22 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..52788130aff8ca766d3fe2fc981659c444b7cf81 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:93ec3ea1e0bcd641ed51f2d1d4e31dbaba2b2eb589a9a24304423280c07e4727 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..de32136482e3cf9e6ced6fc6501a2d0c7d0c1654 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6509030554294586 +INFO:root: +[Valid][0][UTT][ MSE = 0.3144 | LCC = 0.9061 | SRCC = 0.9089 ] +INFO:root:Average loss=0.2196166058778763 +INFO:root: +[Valid][1][UTT][ MSE = 0.2770 | LCC = 0.9318 | SRCC = 0.9341 ] +INFO:root:Average loss=0.184027741253376 +INFO:root: +[Valid][2][UTT][ MSE = 0.2720 | LCC = 0.9368 | SRCC = 0.9392 ] +INFO:root:Average loss=0.15107028275728226 +INFO:root: +[Valid][3][UTT][ MSE = 0.2392 | LCC = 0.9400 | SRCC = 0.9416 ] +INFO:root:Average loss=0.13752109748125077 +INFO:root: +[Valid][4][UTT][ MSE = 0.2305 | LCC = 0.9458 | SRCC = 0.9479 ] +INFO:root:Average loss=0.12978473004698754 +INFO:root: +[Valid][5][UTT][ MSE = 0.2680 | LCC = 0.9500 | SRCC = 0.9518 ] +INFO:root:Average loss=0.11947654485702515 +INFO:root: +[Valid][6][UTT][ MSE = 0.2247 | LCC = 0.9517 | SRCC = 0.9526 ] +INFO:root:Average loss=0.10761991074681282 +INFO:root: +[Valid][7][UTT][ MSE = 0.2275 | LCC = 0.9501 | SRCC = 0.9516 ] +INFO:root:Average loss=0.10158700439333916 +INFO:root: +[Valid][8][UTT][ MSE = 0.2298 | LCC = 0.9484 | SRCC = 0.9493 ] +INFO:root:Average loss=0.09620315608382225 +INFO:root: +[Valid][9][UTT][ MSE = 0.2133 | LCC = 0.9520 | SRCC = 0.9526 ] +INFO:root:Average loss=0.09117890396714211 +INFO:root: +[Valid][10][UTT][ MSE = 0.2214 | LCC = 0.9528 | SRCC = 0.9543 ] +INFO:root:Average loss=0.0855025874376297 +INFO:root: +[Valid][11][UTT][ MSE = 0.2015 | LCC = 0.9518 | SRCC = 0.9530 ] +INFO:root:Average loss=0.08177813014388084 +INFO:root: +[Valid][12][UTT][ MSE = 0.1886 | LCC = 0.9519 | SRCC = 0.9537 ] +INFO:root:Average loss=0.07571210381388664 +INFO:root: +[Valid][13][UTT][ MSE = 0.2052 | LCC = 0.9554 | SRCC = 0.9562 ] +INFO:root:Average loss=0.07383665415644645 +INFO:root: +[Valid][14][UTT][ MSE = 0.2090 | LCC = 0.9574 | SRCC = 0.9583 ] +INFO:root:Average loss=0.07070428261160851 +INFO:root: +[Valid][15][UTT][ MSE = 0.2236 | LCC = 0.9510 | SRCC = 0.9518 ] +INFO:root:Average loss=0.0692132218182087 +INFO:root: +[Valid][16][UTT][ MSE = 0.2068 | LCC = 0.9565 | SRCC = 0.9568 ] +INFO:root:Average loss=0.0625254147350788 +INFO:root: +[Valid][17][UTT][ MSE = 0.2133 | LCC = 0.9566 | SRCC = 0.9573 ] +INFO:root:Average loss=0.059911476463079455 +INFO:root: +[Valid][18][UTT][ MSE = 0.2236 | LCC = 0.9574 | SRCC = 0.9580 ] +INFO:root:Average loss=0.05838600564002991 +INFO:root: +[Valid][19][UTT][ MSE = 0.2332 | LCC = 0.9539 | SRCC = 0.9552 ] +INFO:root:Average loss=0.057701872438192366 +INFO:root: +[Valid][20][UTT][ MSE = 0.1871 | LCC = 0.9555 | SRCC = 0.9564 ] +INFO:root:Average loss=0.05403106135129929 +INFO:root: +[Valid][21][UTT][ MSE = 0.2000 | LCC = 0.9571 | SRCC = 0.9572 ] +INFO:root:Average loss=0.052114736288785934 +INFO:root: +[Valid][22][UTT][ MSE = 0.1917 | LCC = 0.9562 | SRCC = 0.9568 ] +INFO:root:Average loss=0.05113162779808044 +INFO:root: +[Valid][23][UTT][ MSE = 0.2030 | LCC = 0.9555 | SRCC = 0.9561 ] +INFO:root:Average loss=0.04942956402897835 +INFO:root: +[Valid][24][UTT][ MSE = 0.2077 | LCC = 0.9566 | SRCC = 0.9573 ] +INFO:root:Average loss=0.05031870165467262 +INFO:root: +[Valid][25][UTT][ MSE = 0.1996 | LCC = 0.9563 | SRCC = 0.9566 ] +INFO:root:Average loss=0.04535461121797562 +INFO:root: +[Valid][26][UTT][ MSE = 0.2024 | LCC = 0.9585 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04811444312334061 +INFO:root: +[Valid][27][UTT][ MSE = 0.1890 | LCC = 0.9563 | SRCC = 0.9569 ] +INFO:root:Average loss=0.04145051583647728 +INFO:root: +[Valid][28][UTT][ MSE = 0.1977 | LCC = 0.9569 | SRCC = 0.9568 ] +INFO:root:Average loss=0.04091439640522003 +INFO:root: +[Valid][29][UTT][ MSE = 0.2001 | LCC = 0.9577 | SRCC = 0.9582 ] +INFO:root:Average loss=0.03975096778571606 +INFO:root: +[Valid][30][UTT][ MSE = 0.1908 | LCC = 0.9574 | SRCC = 0.9575 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2122 | LCC = 0.9586 | SRCC = 0.9601 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5d83229d220f74bf17511064b828dc1d640fcf81 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4be3e2704aed8e94bfba75f4d37fa8ac63c2f7cc932f27596b66d1b7945046e +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..29103d660cc27e893e0126ef8585d39b5c345856 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6a446423dc965a20163a389a07ca41c41ecb7c46dca51d649e11bad8a612d7f6 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a6cafa47ec1c3189653aa68e1c8bb89ceaf273c0 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4801592128276825 +INFO:root: +[Valid][0][UTT][ MSE = 0.3521 | LCC = 0.9143 | SRCC = 0.9167 ] +INFO:root:Average loss=0.21139605975151063 +INFO:root: +[Valid][1][UTT][ MSE = 0.3475 | LCC = 0.9313 | SRCC = 0.9347 ] +INFO:root:Average loss=0.175518878698349 +INFO:root: +[Valid][2][UTT][ MSE = 0.2492 | LCC = 0.9383 | SRCC = 0.9396 ] +INFO:root:Average loss=0.15592331421375275 +INFO:root: +[Valid][3][UTT][ MSE = 0.2733 | LCC = 0.9448 | SRCC = 0.9459 ] +INFO:root:Average loss=0.13868592250347136 +INFO:root: +[Valid][4][UTT][ MSE = 0.2428 | LCC = 0.9443 | SRCC = 0.9460 ] +INFO:root:Average loss=0.13076802706718446 +INFO:root: +[Valid][5][UTT][ MSE = 0.2591 | LCC = 0.9475 | SRCC = 0.9485 ] +INFO:root:Average loss=0.11506363397836686 +INFO:root: +[Valid][6][UTT][ MSE = 0.2083 | LCC = 0.9483 | SRCC = 0.9492 ] +INFO:root:Average loss=0.1096720617711544 +INFO:root: +[Valid][7][UTT][ MSE = 0.2423 | LCC = 0.9485 | SRCC = 0.9500 ] +INFO:root:Average loss=0.10704722860455514 +INFO:root: +[Valid][8][UTT][ MSE = 0.2085 | LCC = 0.9524 | SRCC = 0.9533 ] +INFO:root:Average loss=0.09666902127861976 +INFO:root: +[Valid][9][UTT][ MSE = 0.2660 | LCC = 0.9508 | SRCC = 0.9525 ] +INFO:root:Average loss=0.09244434490799903 +INFO:root: +[Valid][10][UTT][ MSE = 0.2191 | LCC = 0.9500 | SRCC = 0.9507 ] +INFO:root:Average loss=0.0917695144712925 +INFO:root: +[Valid][11][UTT][ MSE = 0.2223 | LCC = 0.9515 | SRCC = 0.9514 ] +INFO:root:Average loss=0.0833442396223545 +INFO:root: +[Valid][12][UTT][ MSE = 0.2292 | LCC = 0.9516 | SRCC = 0.9528 ] +INFO:root:Average loss=0.08487044627964496 +INFO:root: +[Valid][13][UTT][ MSE = 0.2306 | LCC = 0.9513 | SRCC = 0.9519 ] +INFO:root:Average loss=0.07574659231305123 +INFO:root: +[Valid][14][UTT][ MSE = 0.1879 | LCC = 0.9523 | SRCC = 0.9529 ] +INFO:root:Average loss=0.07141567960381508 +INFO:root: +[Valid][15][UTT][ MSE = 0.2005 | LCC = 0.9540 | SRCC = 0.9543 ] +INFO:root:Average loss=0.0681918685734272 +INFO:root: +[Valid][16][UTT][ MSE = 0.1851 | LCC = 0.9541 | SRCC = 0.9544 ] +INFO:root:Average loss=0.06579469987750053 +INFO:root: +[Valid][17][UTT][ MSE = 0.2066 | LCC = 0.9533 | SRCC = 0.9539 ] +INFO:root:Average loss=0.06638138657808304 +INFO:root: +[Valid][18][UTT][ MSE = 0.2101 | LCC = 0.9544 | SRCC = 0.9551 ] +INFO:root:Average loss=0.06080639371275902 +INFO:root: +[Valid][19][UTT][ MSE = 0.2108 | LCC = 0.9522 | SRCC = 0.9532 ] +INFO:root:Average loss=0.05721925491094589 +INFO:root: +[Valid][20][UTT][ MSE = 0.2057 | LCC = 0.9542 | SRCC = 0.9544 ] +INFO:root:Average loss=0.05646757915616035 +INFO:root: +[Valid][21][UTT][ MSE = 0.1961 | LCC = 0.9533 | SRCC = 0.9534 ] +INFO:root:Average loss=0.05461681348085404 +INFO:root: +[Valid][22][UTT][ MSE = 0.2217 | LCC = 0.9539 | SRCC = 0.9544 ] +INFO:root:Average loss=0.05627953711152077 +INFO:root: +[Valid][23][UTT][ MSE = 0.1913 | LCC = 0.9557 | SRCC = 0.9561 ] +INFO:root:Average loss=0.05011064854264259 +INFO:root: +[Valid][24][UTT][ MSE = 0.2065 | LCC = 0.9554 | SRCC = 0.9561 ] +INFO:root:Average loss=0.05051308655738831 +INFO:root: +[Valid][25][UTT][ MSE = 0.2032 | LCC = 0.9549 | SRCC = 0.9555 ] +INFO:root:Average loss=0.04700484454631806 +INFO:root: +[Valid][26][UTT][ MSE = 0.1762 | LCC = 0.9546 | SRCC = 0.9554 ] +INFO:root:Average loss=0.045230320632457736 +INFO:root: +[Valid][27][UTT][ MSE = 0.2141 | LCC = 0.9540 | SRCC = 0.9546 ] +INFO:root:Average loss=0.045513241678476334 +INFO:root: +[Valid][28][UTT][ MSE = 0.1858 | LCC = 0.9533 | SRCC = 0.9540 ] +INFO:root:Average loss=0.041326691970229146 +INFO:root: +[Valid][29][UTT][ MSE = 0.1942 | LCC = 0.9546 | SRCC = 0.9553 ] +INFO:root:Average loss=0.04537670554220676 +INFO:root: +[Valid][30][UTT][ MSE = 0.1946 | LCC = 0.9546 | SRCC = 0.9550 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1971 | LCC = 0.9556 | SRCC = 0.9564 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b0e0906b8d55a5e8c89fb8c2ac2ffa74db63738e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e6b1c3621b6210bdc449dea5c04dd0bcf3c8da388adc5f2efb4c6187f11ff16 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..29e8047412e926b3158c5de65bd79697806d8915 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:455d82db471b32cf9070f7dd3afe25f093e7d8ecd5cbb23f3d7917db2e347655 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..567b5acc9e2c480d3fe0cac912a7085f7a4edb8d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer1/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4878673813343048 +INFO:root: +[Valid][0][UTT][ MSE = 0.3108 | LCC = 0.9107 | SRCC = 0.9145 ] +INFO:root:Average loss=0.20861506992578507 +INFO:root: +[Valid][1][UTT][ MSE = 0.2634 | LCC = 0.9324 | SRCC = 0.9358 ] +INFO:root:Average loss=0.1681030638217926 +INFO:root: +[Valid][2][UTT][ MSE = 0.2508 | LCC = 0.9395 | SRCC = 0.9420 ] +INFO:root:Average loss=0.1472177372574806 +INFO:root: +[Valid][3][UTT][ MSE = 0.2588 | LCC = 0.9469 | SRCC = 0.9485 ] +INFO:root:Average loss=0.13371166253089906 +INFO:root: +[Valid][4][UTT][ MSE = 0.2328 | LCC = 0.9489 | SRCC = 0.9503 ] +INFO:root:Average loss=0.12144188195466996 +INFO:root: +[Valid][5][UTT][ MSE = 0.2080 | LCC = 0.9502 | SRCC = 0.9509 ] +INFO:root:Average loss=0.11185720795392991 +INFO:root: +[Valid][6][UTT][ MSE = 0.2006 | LCC = 0.9530 | SRCC = 0.9539 ] +INFO:root:Average loss=0.1060076994895935 +INFO:root: +[Valid][7][UTT][ MSE = 0.1985 | LCC = 0.9532 | SRCC = 0.9545 ] +INFO:root:Average loss=0.10169418492913246 +INFO:root: +[Valid][8][UTT][ MSE = 0.2115 | LCC = 0.9535 | SRCC = 0.9538 ] +INFO:root:Average loss=0.09133067446947098 +INFO:root: +[Valid][9][UTT][ MSE = 0.1990 | LCC = 0.9547 | SRCC = 0.9565 ] +INFO:root:Average loss=0.08662062603235245 +INFO:root: +[Valid][10][UTT][ MSE = 0.2122 | LCC = 0.9560 | SRCC = 0.9567 ] +INFO:root:Average loss=0.08328334367275238 +INFO:root: +[Valid][11][UTT][ MSE = 0.2134 | LCC = 0.9549 | SRCC = 0.9565 ] +INFO:root:Average loss=0.07847896921634674 +INFO:root: +[Valid][12][UTT][ MSE = 0.2075 | LCC = 0.9567 | SRCC = 0.9573 ] +INFO:root:Average loss=0.07226548463106156 +INFO:root: +[Valid][13][UTT][ MSE = 0.1643 | LCC = 0.9541 | SRCC = 0.9563 ] +INFO:root:Average loss=0.07053855687379837 +INFO:root: +[Valid][14][UTT][ MSE = 0.2069 | LCC = 0.9562 | SRCC = 0.9569 ] +INFO:root:Average loss=0.06564089077711105 +INFO:root: +[Valid][15][UTT][ MSE = 0.1857 | LCC = 0.9574 | SRCC = 0.9581 ] +INFO:root:Average loss=0.06441515734791756 +INFO:root: +[Valid][16][UTT][ MSE = 0.2158 | LCC = 0.9560 | SRCC = 0.9567 ] +INFO:root:Average loss=0.060239887177944186 +INFO:root: +[Valid][17][UTT][ MSE = 0.1915 | LCC = 0.9585 | SRCC = 0.9591 ] +INFO:root:Average loss=0.05873961353302002 +INFO:root: +[Valid][18][UTT][ MSE = 0.1810 | LCC = 0.9579 | SRCC = 0.9584 ] +INFO:root:Average loss=0.056284116357564924 +INFO:root: +[Valid][19][UTT][ MSE = 0.1828 | LCC = 0.9574 | SRCC = 0.9583 ] +INFO:root:Average loss=0.052535724207758906 +INFO:root: +[Valid][20][UTT][ MSE = 0.1872 | LCC = 0.9584 | SRCC = 0.9591 ] +INFO:root:Average loss=0.0491207074970007 +INFO:root: +[Valid][21][UTT][ MSE = 0.1922 | LCC = 0.9568 | SRCC = 0.9575 ] +INFO:root:Average loss=0.04701900026202202 +INFO:root: +[Valid][22][UTT][ MSE = 0.1781 | LCC = 0.9574 | SRCC = 0.9583 ] +INFO:root:Average loss=0.04750551576912403 +INFO:root: +[Valid][23][UTT][ MSE = 0.1918 | LCC = 0.9567 | SRCC = 0.9575 ] +INFO:root:Average loss=0.047822620421648025 +INFO:root: +[Valid][24][UTT][ MSE = 0.1770 | LCC = 0.9577 | SRCC = 0.9577 ] +INFO:root:Average loss=0.04637842412292957 +INFO:root: +[Valid][25][UTT][ MSE = 0.1797 | LCC = 0.9586 | SRCC = 0.9585 ] +INFO:root:Average loss=0.045138125732541085 +INFO:root: +[Valid][26][UTT][ MSE = 0.1981 | LCC = 0.9567 | SRCC = 0.9570 ] +INFO:root:Average loss=0.040138355031609535 +INFO:root: +[Valid][27][UTT][ MSE = 0.1809 | LCC = 0.9582 | SRCC = 0.9584 ] +INFO:root:Average loss=0.038421537816524506 +INFO:root: +[Valid][28][UTT][ MSE = 0.1833 | LCC = 0.9564 | SRCC = 0.9572 ] +INFO:root:Average loss=0.04264209255576134 +INFO:root: +[Valid][29][UTT][ MSE = 0.1757 | LCC = 0.9567 | SRCC = 0.9567 ] +INFO:root:Average loss=0.0374107653349638 +INFO:root: +[Valid][30][UTT][ MSE = 0.1893 | LCC = 0.9572 | SRCC = 0.9568 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1803 | LCC = 0.9581 | SRCC = 0.9592 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0e00e03ecfff7243587992b009a4e2a27b4d0f9d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:004eb4522e77fbc4d92c5c8474e5afc3d4406c047fca4ce84c9cf403d343cc5f +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0be8165218e6dbd8f8390f3769d0404d6e0358e2 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9f712adbb1f33ea3a336f837c41494d32d751bbded7cad6d7dbc06bab8b1c1bd +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3e0ce857bebc9b11c16c9ccefdca490de9e4ad5e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.30902277714014054 +INFO:root: +[Valid][0][UTT][ MSE = 0.3342 | LCC = 0.9213 | SRCC = 0.9224 ] +INFO:root:Average loss=0.20452969163656234 +INFO:root: +[Valid][1][UTT][ MSE = 0.3025 | LCC = 0.9325 | SRCC = 0.9352 ] +INFO:root:Average loss=0.1853622596859932 +INFO:root: +[Valid][2][UTT][ MSE = 0.2749 | LCC = 0.9354 | SRCC = 0.9401 ] +INFO:root:Average loss=0.16694013464450835 +INFO:root: +[Valid][3][UTT][ MSE = 0.2878 | LCC = 0.9379 | SRCC = 0.9434 ] +INFO:root:Average loss=0.1594066536426544 +INFO:root: +[Valid][4][UTT][ MSE = 0.2582 | LCC = 0.9424 | SRCC = 0.9435 ] +INFO:root:Average loss=0.1482121803164482 +INFO:root: +[Valid][5][UTT][ MSE = 0.2396 | LCC = 0.9380 | SRCC = 0.9458 ] +INFO:root:Average loss=0.14500034099817277 +INFO:root: +[Valid][6][UTT][ MSE = 0.2265 | LCC = 0.9461 | SRCC = 0.9468 ] +INFO:root:Average loss=0.1375709791779518 +INFO:root: +[Valid][7][UTT][ MSE = 0.2512 | LCC = 0.9477 | SRCC = 0.9499 ] +INFO:root:Average loss=0.13082195854187012 +INFO:root: +[Valid][8][UTT][ MSE = 0.2987 | LCC = 0.9474 | SRCC = 0.9498 ] +INFO:root:Average loss=0.1280507496893406 +INFO:root: +[Valid][9][UTT][ MSE = 0.2846 | LCC = 0.9480 | SRCC = 0.9496 ] +INFO:root:Average loss=0.1222425658106804 +INFO:root: +[Valid][10][UTT][ MSE = 0.2725 | LCC = 0.9496 | SRCC = 0.9515 ] +INFO:root:Average loss=0.12109947550296783 +INFO:root: +[Valid][11][UTT][ MSE = 0.2887 | LCC = 0.9496 | SRCC = 0.9508 ] +INFO:root:Average loss=0.11630650275945663 +INFO:root: +[Valid][12][UTT][ MSE = 0.2916 | LCC = 0.9484 | SRCC = 0.9508 ] +INFO:root:Average loss=0.11803143483400345 +INFO:root: +[Valid][13][UTT][ MSE = 0.2609 | LCC = 0.9486 | SRCC = 0.9502 ] +INFO:root:Average loss=0.11145109289884567 +INFO:root: +[Valid][14][UTT][ MSE = 0.2315 | LCC = 0.9462 | SRCC = 0.9468 ] +INFO:root:Average loss=0.11050088834762574 +INFO:root: +[Valid][15][UTT][ MSE = 0.2509 | LCC = 0.9488 | SRCC = 0.9492 ] +INFO:root:Average loss=0.1083276771903038 +INFO:root: +[Valid][16][UTT][ MSE = 0.2530 | LCC = 0.9476 | SRCC = 0.9492 ] +INFO:root:Average loss=0.10524966710805893 +INFO:root: +[Valid][17][UTT][ MSE = 0.2817 | LCC = 0.9496 | SRCC = 0.9505 ] +INFO:root:Average loss=0.105453584253788 +INFO:root: +[Valid][18][UTT][ MSE = 0.2529 | LCC = 0.9480 | SRCC = 0.9501 ] +INFO:root:Average loss=0.10577457869052886 +INFO:root: +[Valid][19][UTT][ MSE = 0.2253 | LCC = 0.9490 | SRCC = 0.9522 ] +INFO:root:Average loss=0.09772588109970093 +INFO:root: +[Valid][20][UTT][ MSE = 0.2441 | LCC = 0.9502 | SRCC = 0.9514 ] +INFO:root:Average loss=0.09594934260845184 +INFO:root: +[Valid][21][UTT][ MSE = 0.2235 | LCC = 0.9502 | SRCC = 0.9521 ] +INFO:root:Average loss=0.09405272835493088 +INFO:root: +[Valid][22][UTT][ MSE = 0.2147 | LCC = 0.9488 | SRCC = 0.9499 ] +INFO:root:Average loss=0.09284023290872574 +INFO:root: +[Valid][23][UTT][ MSE = 0.2495 | LCC = 0.9479 | SRCC = 0.9498 ] +INFO:root:Average loss=0.09107664868235588 +INFO:root: +[Valid][24][UTT][ MSE = 0.2296 | LCC = 0.9480 | SRCC = 0.9515 ] +INFO:root:Average loss=0.08911662608385086 +INFO:root: +[Valid][25][UTT][ MSE = 0.2343 | LCC = 0.9510 | SRCC = 0.9518 ] +INFO:root:Average loss=0.0872859560251236 +INFO:root: +[Valid][26][UTT][ MSE = 0.2198 | LCC = 0.9503 | SRCC = 0.9526 ] +INFO:root:Average loss=0.0886726096868515 +INFO:root: +[Valid][27][UTT][ MSE = 0.2181 | LCC = 0.9483 | SRCC = 0.9494 ] +INFO:root:Average loss=0.08210958717763424 +INFO:root: +[Valid][28][UTT][ MSE = 0.2443 | LCC = 0.9512 | SRCC = 0.9522 ] +INFO:root:Average loss=0.08263995656371116 +INFO:root: +[Valid][29][UTT][ MSE = 0.2260 | LCC = 0.9508 | SRCC = 0.9514 ] +INFO:root:Average loss=0.08141993701457978 +INFO:root: +[Valid][30][UTT][ MSE = 0.2320 | LCC = 0.9504 | SRCC = 0.9522 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2439 | LCC = 0.9537 | SRCC = 0.9550 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6ad78a65663b76f1d4b08e88b748ce1e59019c04 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d37581e4559c65fdb9d07d5c5164f29d6a7c4ab10030d481498d910976ebb088 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..8894a219a9df19cbbaa2c21ec604226af4ef81b2 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e104cf2fe8aa1b8024515e7ccbc0f325a567b9f749370e120c8550d5270e51b4 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..7d314f0e609cb73b7b8a02dd878f14a64bd1eb91 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3520560179948807 +INFO:root: +[Valid][0][UTT][ MSE = 0.4050 | LCC = 0.9202 | SRCC = 0.9233 ] +INFO:root:Average loss=0.20852748835086823 +INFO:root: +[Valid][1][UTT][ MSE = 0.3264 | LCC = 0.9295 | SRCC = 0.9324 ] +INFO:root:Average loss=0.17937929981946946 +INFO:root: +[Valid][2][UTT][ MSE = 0.2975 | LCC = 0.9388 | SRCC = 0.9414 ] +INFO:root:Average loss=0.17485316222906114 +INFO:root: +[Valid][3][UTT][ MSE = 0.2503 | LCC = 0.9366 | SRCC = 0.9405 ] +INFO:root:Average loss=0.16155420517921448 +INFO:root: +[Valid][4][UTT][ MSE = 0.3162 | LCC = 0.9406 | SRCC = 0.9434 ] +INFO:root:Average loss=0.15285056245326994 +INFO:root: +[Valid][5][UTT][ MSE = 0.2914 | LCC = 0.9409 | SRCC = 0.9453 ] +INFO:root:Average loss=0.15011487358808517 +INFO:root: +[Valid][6][UTT][ MSE = 0.2553 | LCC = 0.9424 | SRCC = 0.9466 ] +INFO:root:Average loss=0.14741260021924973 +INFO:root: +[Valid][7][UTT][ MSE = 0.2539 | LCC = 0.9432 | SRCC = 0.9464 ] +INFO:root:Average loss=0.13381069004535676 +INFO:root: +[Valid][8][UTT][ MSE = 0.2469 | LCC = 0.9426 | SRCC = 0.9450 ] +INFO:root:Average loss=0.13425644034147263 +INFO:root: +[Valid][9][UTT][ MSE = 0.2405 | LCC = 0.9424 | SRCC = 0.9477 ] +INFO:root:Average loss=0.12746263313293457 +INFO:root: +[Valid][10][UTT][ MSE = 0.2667 | LCC = 0.9443 | SRCC = 0.9486 ] +INFO:root:Average loss=0.12322832119464874 +INFO:root: +[Valid][11][UTT][ MSE = 0.2743 | LCC = 0.9481 | SRCC = 0.9500 ] +INFO:root:Average loss=0.1234422602057457 +INFO:root: +[Valid][12][UTT][ MSE = 0.2522 | LCC = 0.9483 | SRCC = 0.9498 ] +INFO:root:Average loss=0.11882060784101486 +INFO:root: +[Valid][13][UTT][ MSE = 0.2605 | LCC = 0.9452 | SRCC = 0.9505 ] +INFO:root:Average loss=0.11949072790145875 +INFO:root: +[Valid][14][UTT][ MSE = 0.2239 | LCC = 0.9469 | SRCC = 0.9492 ] +INFO:root:Average loss=0.11683094239234924 +INFO:root: +[Valid][15][UTT][ MSE = 0.2476 | LCC = 0.9485 | SRCC = 0.9503 ] +INFO:root:Average loss=0.11320522367954254 +INFO:root: +[Valid][16][UTT][ MSE = 0.2673 | LCC = 0.9454 | SRCC = 0.9484 ] +INFO:root:Average loss=0.10847462695837021 +INFO:root: +[Valid][17][UTT][ MSE = 0.2449 | LCC = 0.9485 | SRCC = 0.9503 ] +INFO:root:Average loss=0.10558279493451118 +INFO:root: +[Valid][18][UTT][ MSE = 0.2455 | LCC = 0.9475 | SRCC = 0.9501 ] +INFO:root:Average loss=0.10519059750437737 +INFO:root: +[Valid][19][UTT][ MSE = 0.2305 | LCC = 0.9472 | SRCC = 0.9501 ] +INFO:root:Average loss=0.10295894053578376 +INFO:root: +[Valid][20][UTT][ MSE = 0.2568 | LCC = 0.9475 | SRCC = 0.9489 ] +INFO:root:Average loss=0.09978912377357482 +INFO:root: +[Valid][21][UTT][ MSE = 0.2848 | LCC = 0.9451 | SRCC = 0.9476 ] +INFO:root:Average loss=0.09689242160320281 +INFO:root: +[Valid][22][UTT][ MSE = 0.2296 | LCC = 0.9498 | SRCC = 0.9508 ] +INFO:root:Average loss=0.09752023214101792 +INFO:root: +[Valid][23][UTT][ MSE = 0.2389 | LCC = 0.9494 | SRCC = 0.9500 ] +INFO:root:Average loss=0.09832394447922707 +INFO:root: +[Valid][24][UTT][ MSE = 0.2893 | LCC = 0.9466 | SRCC = 0.9490 ] +INFO:root:Average loss=0.0949598371386528 +INFO:root: +[Valid][25][UTT][ MSE = 0.2284 | LCC = 0.9493 | SRCC = 0.9497 ] +INFO:root:Average loss=0.09335098704695702 +INFO:root: +[Valid][26][UTT][ MSE = 0.2319 | LCC = 0.9490 | SRCC = 0.9514 ] +INFO:root:Average loss=0.09167593407630921 +INFO:root: +[Valid][27][UTT][ MSE = 0.2383 | LCC = 0.9485 | SRCC = 0.9508 ] +INFO:root:Average loss=0.08645264172554017 +INFO:root: +[Valid][28][UTT][ MSE = 0.2557 | LCC = 0.9499 | SRCC = 0.9513 ] +INFO:root:Average loss=0.08679896438121795 +INFO:root: +[Valid][29][UTT][ MSE = 0.2412 | LCC = 0.9514 | SRCC = 0.9525 ] +INFO:root:Average loss=0.08410239008069038 +INFO:root: +[Valid][30][UTT][ MSE = 0.2594 | LCC = 0.9475 | SRCC = 0.9495 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2506 | LCC = 0.9525 | SRCC = 0.9538 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..425a47eb92d2a5722ec24cfadd9b9d777fd0fe6b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:34115c02a574d964655c15c704660b3326d4998b29fcde804014e7bae32695d8 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e9eb98392aa21b4ecf29d11696397c13d31a98bf --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0526fc5c278ae21c032d24f8a5473d3b13ec4d0c8933b5b971ed3a6674f11866 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3a50bd3b947d449f7ed9fc68f68f2d7bee3213ef --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3239538811445236 +INFO:root: +[Valid][0][UTT][ MSE = 0.3408 | LCC = 0.9228 | SRCC = 0.9248 ] +INFO:root:Average loss=0.21295914483070374 +INFO:root: +[Valid][1][UTT][ MSE = 0.2996 | LCC = 0.9334 | SRCC = 0.9351 ] +INFO:root:Average loss=0.17897224807739257 +INFO:root: +[Valid][2][UTT][ MSE = 0.3141 | LCC = 0.9388 | SRCC = 0.9396 ] +INFO:root:Average loss=0.1672461394071579 +INFO:root: +[Valid][3][UTT][ MSE = 0.3080 | LCC = 0.9420 | SRCC = 0.9441 ] +INFO:root:Average loss=0.15521497821807861 +INFO:root: +[Valid][4][UTT][ MSE = 0.2797 | LCC = 0.9452 | SRCC = 0.9466 ] +INFO:root:Average loss=0.1446857762336731 +INFO:root: +[Valid][5][UTT][ MSE = 0.2857 | LCC = 0.9394 | SRCC = 0.9418 ] +INFO:root:Average loss=0.1421384015083313 +INFO:root: +[Valid][6][UTT][ MSE = 0.2797 | LCC = 0.9438 | SRCC = 0.9455 ] +INFO:root:Average loss=0.1338391322493553 +INFO:root: +[Valid][7][UTT][ MSE = 0.2648 | LCC = 0.9475 | SRCC = 0.9493 ] +INFO:root:Average loss=0.13420029240846634 +INFO:root: +[Valid][8][UTT][ MSE = 0.3004 | LCC = 0.9440 | SRCC = 0.9463 ] +INFO:root:Average loss=0.1326946273446083 +INFO:root: +[Valid][9][UTT][ MSE = 0.2768 | LCC = 0.9493 | SRCC = 0.9506 ] +INFO:root:Average loss=0.12767786294221878 +INFO:root: +[Valid][10][UTT][ MSE = 0.2684 | LCC = 0.9492 | SRCC = 0.9509 ] +INFO:root:Average loss=0.12093556350469589 +INFO:root: +[Valid][11][UTT][ MSE = 0.2279 | LCC = 0.9486 | SRCC = 0.9503 ] +INFO:root:Average loss=0.12013656282424927 +INFO:root: +[Valid][12][UTT][ MSE = 0.2392 | LCC = 0.9493 | SRCC = 0.9506 ] +INFO:root:Average loss=0.11481758326292038 +INFO:root: +[Valid][13][UTT][ MSE = 0.3099 | LCC = 0.9439 | SRCC = 0.9464 ] +INFO:root:Average loss=0.11424020394682884 +INFO:root: +[Valid][14][UTT][ MSE = 0.2614 | LCC = 0.9511 | SRCC = 0.9527 ] +INFO:root:Average loss=0.11124946820735931 +INFO:root: +[Valid][15][UTT][ MSE = 0.2634 | LCC = 0.9502 | SRCC = 0.9515 ] +INFO:root:Average loss=0.11019123324751853 +INFO:root: +[Valid][16][UTT][ MSE = 0.2479 | LCC = 0.9506 | SRCC = 0.9518 ] +INFO:root:Average loss=0.10920317450165748 +INFO:root: +[Valid][17][UTT][ MSE = 0.2149 | LCC = 0.9502 | SRCC = 0.9518 ] +INFO:root:Average loss=0.10415299960970878 +INFO:root: +[Valid][18][UTT][ MSE = 0.2600 | LCC = 0.9520 | SRCC = 0.9526 ] +INFO:root:Average loss=0.1039080962240696 +INFO:root: +[Valid][19][UTT][ MSE = 0.2578 | LCC = 0.9501 | SRCC = 0.9506 ] +INFO:root:Average loss=0.101131854981184 +INFO:root: +[Valid][20][UTT][ MSE = 0.2169 | LCC = 0.9484 | SRCC = 0.9527 ] +INFO:root:Average loss=0.10124831146001816 +INFO:root: +[Valid][21][UTT][ MSE = 0.2034 | LCC = 0.9504 | SRCC = 0.9541 ] +INFO:root:Average loss=0.09813296037912368 +INFO:root: +[Valid][22][UTT][ MSE = 0.2455 | LCC = 0.9519 | SRCC = 0.9528 ] +INFO:root:Average loss=0.09422659742832183 +INFO:root: +[Valid][23][UTT][ MSE = 0.2610 | LCC = 0.9501 | SRCC = 0.9507 ] +INFO:root:Average loss=0.09316587817668914 +INFO:root: +[Valid][24][UTT][ MSE = 0.2610 | LCC = 0.9495 | SRCC = 0.9505 ] +INFO:root:Average loss=0.09194156813621521 +INFO:root: +[Valid][25][UTT][ MSE = 0.2694 | LCC = 0.9465 | SRCC = 0.9474 ] +INFO:root:Average loss=0.08882942795753479 +INFO:root: +[Valid][26][UTT][ MSE = 0.2281 | LCC = 0.9508 | SRCC = 0.9529 ] +INFO:root:Average loss=0.08897866144776344 +INFO:root: +[Valid][27][UTT][ MSE = 0.3040 | LCC = 0.9442 | SRCC = 0.9444 ] +INFO:root:Average loss=0.0880424055159092 +INFO:root: +[Valid][28][UTT][ MSE = 0.2618 | LCC = 0.9453 | SRCC = 0.9470 ] +INFO:root:Average loss=0.08696771407127381 +INFO:root: +[Valid][29][UTT][ MSE = 0.2379 | LCC = 0.9513 | SRCC = 0.9523 ] +INFO:root:Average loss=0.08140960940718651 +INFO:root: +[Valid][30][UTT][ MSE = 0.2419 | LCC = 0.9518 | SRCC = 0.9533 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2628 | LCC = 0.9507 | SRCC = 0.9520 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ee0d24cf9140094c5a08330b8e9cb5b029a962bc --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20c04abbae2418033934c602754b1c6f1e34612c100b2f79ddf1220f3cd0bffb +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b4a27412a969fb83b3cd335b22b903dfa7fd2800 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bdb715b5cfc97dc1a27372e709c863577d3b48ae937f1d7b93eb44e0aafefe54 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..8253a9d7778797243c07e045827e1a22659df7e2 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.34505713200569155 +INFO:root: +[Valid][0][UTT][ MSE = 0.3919 | LCC = 0.9269 | SRCC = 0.9295 ] +INFO:root:Average loss=0.20443237555027008 +INFO:root: +[Valid][1][UTT][ MSE = 0.3035 | LCC = 0.9288 | SRCC = 0.9353 ] +INFO:root:Average loss=0.1792861914038658 +INFO:root: +[Valid][2][UTT][ MSE = 0.3522 | LCC = 0.9379 | SRCC = 0.9421 ] +INFO:root:Average loss=0.16785356432199477 +INFO:root: +[Valid][3][UTT][ MSE = 0.2930 | LCC = 0.9404 | SRCC = 0.9442 ] +INFO:root:Average loss=0.15630795007944107 +INFO:root: +[Valid][4][UTT][ MSE = 0.2740 | LCC = 0.9417 | SRCC = 0.9455 ] +INFO:root:Average loss=0.15082917326688766 +INFO:root: +[Valid][5][UTT][ MSE = 0.2493 | LCC = 0.9443 | SRCC = 0.9480 ] +INFO:root:Average loss=0.14577870416641237 +INFO:root: +[Valid][6][UTT][ MSE = 0.2514 | LCC = 0.9430 | SRCC = 0.9455 ] +INFO:root:Average loss=0.1335325345993042 +INFO:root: +[Valid][7][UTT][ MSE = 0.2701 | LCC = 0.9474 | SRCC = 0.9493 ] +INFO:root:Average loss=0.12985334554314612 +INFO:root: +[Valid][8][UTT][ MSE = 0.2844 | LCC = 0.9470 | SRCC = 0.9502 ] +INFO:root:Average loss=0.12921459752321243 +INFO:root: +[Valid][9][UTT][ MSE = 0.2392 | LCC = 0.9489 | SRCC = 0.9504 ] +INFO:root:Average loss=0.12376989763975144 +INFO:root: +[Valid][10][UTT][ MSE = 0.2369 | LCC = 0.9478 | SRCC = 0.9499 ] +INFO:root:Average loss=0.118032390832901 +INFO:root: +[Valid][11][UTT][ MSE = 0.2442 | LCC = 0.9492 | SRCC = 0.9515 ] +INFO:root:Average loss=0.12058747243881225 +INFO:root: +[Valid][12][UTT][ MSE = 0.2568 | LCC = 0.9496 | SRCC = 0.9521 ] +INFO:root:Average loss=0.11906627410650253 +INFO:root: +[Valid][13][UTT][ MSE = 0.2754 | LCC = 0.9494 | SRCC = 0.9515 ] +INFO:root:Average loss=0.11321356201171875 +INFO:root: +[Valid][14][UTT][ MSE = 0.2548 | LCC = 0.9470 | SRCC = 0.9493 ] +INFO:root:Average loss=0.10820669746398925 +INFO:root: +[Valid][15][UTT][ MSE = 0.2250 | LCC = 0.9509 | SRCC = 0.9518 ] +INFO:root:Average loss=0.10821562829613686 +INFO:root: +[Valid][16][UTT][ MSE = 0.2618 | LCC = 0.9458 | SRCC = 0.9474 ] +INFO:root:Average loss=0.10736116683483124 +INFO:root: +[Valid][17][UTT][ MSE = 0.2386 | LCC = 0.9491 | SRCC = 0.9527 ] +INFO:root:Average loss=0.10241256624460221 +INFO:root: +[Valid][18][UTT][ MSE = 0.2751 | LCC = 0.9500 | SRCC = 0.9525 ] +INFO:root:Average loss=0.10097370314598084 +INFO:root: +[Valid][19][UTT][ MSE = 0.2509 | LCC = 0.9511 | SRCC = 0.9523 ] +INFO:root:Average loss=0.09947004851698875 +INFO:root: +[Valid][20][UTT][ MSE = 0.2444 | LCC = 0.9477 | SRCC = 0.9494 ] +INFO:root:Average loss=0.10048803162574768 +INFO:root: +[Valid][21][UTT][ MSE = 0.2454 | LCC = 0.9519 | SRCC = 0.9535 ] +INFO:root:Average loss=0.09399746459722519 +INFO:root: +[Valid][22][UTT][ MSE = 0.2477 | LCC = 0.9499 | SRCC = 0.9506 ] +INFO:root:Average loss=0.09441613483428955 +INFO:root: +[Valid][23][UTT][ MSE = 0.2541 | LCC = 0.9495 | SRCC = 0.9511 ] +INFO:root:Average loss=0.0915441006720066 +INFO:root: +[Valid][24][UTT][ MSE = 0.2517 | LCC = 0.9506 | SRCC = 0.9522 ] +INFO:root:Average loss=0.0913265515267849 +INFO:root: +[Valid][25][UTT][ MSE = 0.2429 | LCC = 0.9508 | SRCC = 0.9521 ] +INFO:root:Average loss=0.08710516315698624 +INFO:root: +[Valid][26][UTT][ MSE = 0.2647 | LCC = 0.9460 | SRCC = 0.9492 ] +INFO:root:Average loss=0.08552612343430518 +INFO:root: +[Valid][27][UTT][ MSE = 0.2790 | LCC = 0.9472 | SRCC = 0.9516 ] +INFO:root:Average loss=0.08892702051997185 +INFO:root: +[Valid][28][UTT][ MSE = 0.2315 | LCC = 0.9488 | SRCC = 0.9507 ] +INFO:root:Average loss=0.08337264394760131 +INFO:root: +[Valid][29][UTT][ MSE = 0.2353 | LCC = 0.9478 | SRCC = 0.9512 ] +INFO:root:Average loss=0.08499873566627503 +INFO:root: +[Valid][30][UTT][ MSE = 0.2764 | LCC = 0.9494 | SRCC = 0.9520 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2527 | LCC = 0.9518 | SRCC = 0.9537 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..02f150f6423334385f5563af2276be56916a2214 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bfd2f21da07d24bc0112a0f51290c031c00fa2770aee784430a1af18c906c673 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..836bed2b0210777ef4a8a5faa6106c50310612dd --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d8bb20730d3e3e4476957e4ec93148982c6ca78bb092d28b2e7c427677b37047 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ffa0dacec0da8e6e0ee6ece54bbcbebf059184ef --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer10/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3098862881064415 +INFO:root: +[Valid][0][UTT][ MSE = 0.3380 | LCC = 0.9225 | SRCC = 0.9234 ] +INFO:root:Average loss=0.21435169839859008 +INFO:root: +[Valid][1][UTT][ MSE = 0.3321 | LCC = 0.9318 | SRCC = 0.9353 ] +INFO:root:Average loss=0.18230284720659257 +INFO:root: +[Valid][2][UTT][ MSE = 0.2820 | LCC = 0.9377 | SRCC = 0.9395 ] +INFO:root:Average loss=0.1678971411585808 +INFO:root: +[Valid][3][UTT][ MSE = 0.3141 | LCC = 0.9341 | SRCC = 0.9356 ] +INFO:root:Average loss=0.16098832148313522 +INFO:root: +[Valid][4][UTT][ MSE = 0.2771 | LCC = 0.9435 | SRCC = 0.9452 ] +INFO:root:Average loss=0.15073406690359115 +INFO:root: +[Valid][5][UTT][ MSE = 0.2782 | LCC = 0.9429 | SRCC = 0.9448 ] +INFO:root:Average loss=0.14143346992135047 +INFO:root: +[Valid][6][UTT][ MSE = 0.2936 | LCC = 0.9424 | SRCC = 0.9450 ] +INFO:root:Average loss=0.14239372557401658 +INFO:root: +[Valid][7][UTT][ MSE = 0.2606 | LCC = 0.9470 | SRCC = 0.9488 ] +INFO:root:Average loss=0.13356979167461394 +INFO:root: +[Valid][8][UTT][ MSE = 0.2612 | LCC = 0.9468 | SRCC = 0.9477 ] +INFO:root:Average loss=0.1288651922941208 +INFO:root: +[Valid][9][UTT][ MSE = 0.2853 | LCC = 0.9456 | SRCC = 0.9465 ] +INFO:root:Average loss=0.12918928456306458 +INFO:root: +[Valid][10][UTT][ MSE = 0.2950 | LCC = 0.9451 | SRCC = 0.9478 ] +INFO:root:Average loss=0.12447459369897843 +INFO:root: +[Valid][11][UTT][ MSE = 0.2984 | LCC = 0.9437 | SRCC = 0.9474 ] +INFO:root:Average loss=0.12554457306861877 +INFO:root: +[Valid][12][UTT][ MSE = 0.2950 | LCC = 0.9409 | SRCC = 0.9449 ] +INFO:root:Average loss=0.11742309698462486 +INFO:root: +[Valid][13][UTT][ MSE = 0.2581 | LCC = 0.9483 | SRCC = 0.9501 ] +INFO:root:Average loss=0.11469560873508454 +INFO:root: +[Valid][14][UTT][ MSE = 0.2471 | LCC = 0.9483 | SRCC = 0.9500 ] +INFO:root:Average loss=0.11256054875254631 +INFO:root: +[Valid][15][UTT][ MSE = 0.2511 | LCC = 0.9472 | SRCC = 0.9502 ] +INFO:root:Average loss=0.11139565956592559 +INFO:root: +[Valid][16][UTT][ MSE = 0.2697 | LCC = 0.9474 | SRCC = 0.9498 ] +INFO:root:Average loss=0.11042464578151703 +INFO:root: +[Valid][17][UTT][ MSE = 0.2356 | LCC = 0.9497 | SRCC = 0.9506 ] +INFO:root:Average loss=0.10790228861570358 +INFO:root: +[Valid][18][UTT][ MSE = 0.2640 | LCC = 0.9485 | SRCC = 0.9506 ] +INFO:root:Average loss=0.10736958420276642 +INFO:root: +[Valid][19][UTT][ MSE = 0.2486 | LCC = 0.9472 | SRCC = 0.9499 ] +INFO:root:Average loss=0.10209155946969986 +INFO:root: +[Valid][20][UTT][ MSE = 0.2468 | LCC = 0.9480 | SRCC = 0.9495 ] +INFO:root:Average loss=0.09999256417155265 +INFO:root: +[Valid][21][UTT][ MSE = 0.2658 | LCC = 0.9476 | SRCC = 0.9487 ] +INFO:root:Average loss=0.09904227960109711 +INFO:root: +[Valid][22][UTT][ MSE = 0.2606 | LCC = 0.9466 | SRCC = 0.9481 ] +INFO:root:Average loss=0.09497362154722214 +INFO:root: +[Valid][23][UTT][ MSE = 0.2351 | LCC = 0.9490 | SRCC = 0.9505 ] +INFO:root:Average loss=0.0918066918849945 +INFO:root: +[Valid][24][UTT][ MSE = 0.2804 | LCC = 0.9464 | SRCC = 0.9486 ] +INFO:root:Average loss=0.09236401310563087 +INFO:root: +[Valid][25][UTT][ MSE = 0.2637 | LCC = 0.9443 | SRCC = 0.9475 ] +INFO:root:Average loss=0.09372627863287926 +INFO:root: +[Valid][26][UTT][ MSE = 0.2429 | LCC = 0.9463 | SRCC = 0.9495 ] +INFO:root:Average loss=0.08789754602313042 +INFO:root: +[Valid][27][UTT][ MSE = 0.2262 | LCC = 0.9472 | SRCC = 0.9479 ] +INFO:root:Average loss=0.08836045098304748 +INFO:root: +[Valid][28][UTT][ MSE = 0.2784 | LCC = 0.9483 | SRCC = 0.9503 ] +INFO:root:Average loss=0.08777776750922203 +INFO:root: +[Valid][29][UTT][ MSE = 0.2704 | LCC = 0.9441 | SRCC = 0.9453 ] +INFO:root:Average loss=0.08487269672751427 +INFO:root: +[Valid][30][UTT][ MSE = 0.2549 | LCC = 0.9495 | SRCC = 0.9508 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2401 | LCC = 0.9506 | SRCC = 0.9519 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..162263f868a340e95ec718e4704a382ccf4ce6a5 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8f6d497447c5390a0a2953a6ba8a06772b475eed7d1c44affc8da28c2c66f953 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6f836dbce570bf3d4203a43e37206bc8747d144d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1937493cbd83983216398a57609019ce30ea8d2d11bb846d299bc99bc3715e4a +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5a352d8c7c97b1f3c64dfcab783a0e95dd9242ac --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.36533413457870484 +INFO:root: +[Valid][0][UTT][ MSE = 0.3763 | LCC = 0.9204 | SRCC = 0.9252 ] +INFO:root:Average loss=0.21246522039175034 +INFO:root: +[Valid][1][UTT][ MSE = 0.3116 | LCC = 0.9320 | SRCC = 0.9327 ] +INFO:root:Average loss=0.1982731122970581 +INFO:root: +[Valid][2][UTT][ MSE = 0.3249 | LCC = 0.9318 | SRCC = 0.9354 ] +INFO:root:Average loss=0.18281660711765288 +INFO:root: +[Valid][3][UTT][ MSE = 0.3323 | LCC = 0.9349 | SRCC = 0.9376 ] +INFO:root:Average loss=0.1776858444213867 +INFO:root: +[Valid][4][UTT][ MSE = 0.3212 | LCC = 0.9368 | SRCC = 0.9391 ] +INFO:root:Average loss=0.1731673179268837 +INFO:root: +[Valid][5][UTT][ MSE = 0.2933 | LCC = 0.9393 | SRCC = 0.9403 ] +INFO:root:Average loss=0.1643575987815857 +INFO:root: +[Valid][6][UTT][ MSE = 0.3153 | LCC = 0.9340 | SRCC = 0.9403 ] +INFO:root:Average loss=0.1610022369623184 +INFO:root: +[Valid][7][UTT][ MSE = 0.3038 | LCC = 0.9350 | SRCC = 0.9371 ] +INFO:root:Average loss=0.15642806470394136 +INFO:root: +[Valid][8][UTT][ MSE = 0.3231 | LCC = 0.9405 | SRCC = 0.9426 ] +INFO:root:Average loss=0.14884266424179077 +INFO:root: +[Valid][9][UTT][ MSE = 0.2709 | LCC = 0.9393 | SRCC = 0.9433 ] +INFO:root:Average loss=0.1454704994559288 +INFO:root: +[Valid][10][UTT][ MSE = 0.3221 | LCC = 0.9386 | SRCC = 0.9411 ] +INFO:root:Average loss=0.14402996438741683 +INFO:root: +[Valid][11][UTT][ MSE = 0.2664 | LCC = 0.9419 | SRCC = 0.9441 ] +INFO:root:Average loss=0.13815940016508102 +INFO:root: +[Valid][12][UTT][ MSE = 0.2807 | LCC = 0.9431 | SRCC = 0.9443 ] +INFO:root:Average loss=0.13685102063417434 +INFO:root: +[Valid][13][UTT][ MSE = 0.2781 | LCC = 0.9443 | SRCC = 0.9457 ] +INFO:root:Average loss=0.1360374614596367 +INFO:root: +[Valid][14][UTT][ MSE = 0.2892 | LCC = 0.9444 | SRCC = 0.9460 ] +INFO:root:Average loss=0.1343663500547409 +INFO:root: +[Valid][15][UTT][ MSE = 0.2815 | LCC = 0.9430 | SRCC = 0.9448 ] +INFO:root:Average loss=0.126043055921793 +INFO:root: +[Valid][16][UTT][ MSE = 0.2689 | LCC = 0.9443 | SRCC = 0.9452 ] +INFO:root:Average loss=0.12891441866755485 +INFO:root: +[Valid][17][UTT][ MSE = 0.2723 | LCC = 0.9434 | SRCC = 0.9451 ] +INFO:root:Average loss=0.12347266346216201 +INFO:root: +[Valid][18][UTT][ MSE = 0.2587 | LCC = 0.9454 | SRCC = 0.9458 ] +INFO:root:Average loss=0.12103471845388412 +INFO:root: +[Valid][19][UTT][ MSE = 0.2456 | LCC = 0.9421 | SRCC = 0.9453 ] +INFO:root:Average loss=0.12800812321901323 +INFO:root: +[Valid][20][UTT][ MSE = 0.2625 | LCC = 0.9455 | SRCC = 0.9465 ] +INFO:root:Average loss=0.11933435356616974 +INFO:root: +[Valid][21][UTT][ MSE = 0.2638 | LCC = 0.9446 | SRCC = 0.9456 ] +INFO:root:Average loss=0.11673335629701614 +INFO:root: +[Valid][22][UTT][ MSE = 0.2489 | LCC = 0.9463 | SRCC = 0.9473 ] +INFO:root:Average loss=0.11631698843836784 +INFO:root: +[Valid][23][UTT][ MSE = 0.2747 | LCC = 0.9452 | SRCC = 0.9465 ] +INFO:root:Average loss=0.11289132741093635 +INFO:root: +[Valid][24][UTT][ MSE = 0.2616 | LCC = 0.9450 | SRCC = 0.9459 ] +INFO:root:Average loss=0.11055779218673706 +INFO:root: +[Valid][25][UTT][ MSE = 0.2535 | LCC = 0.9456 | SRCC = 0.9463 ] +INFO:root:Average loss=0.10770225557684898 +INFO:root: +[Valid][26][UTT][ MSE = 0.2617 | LCC = 0.9462 | SRCC = 0.9464 ] +INFO:root:Average loss=0.10756360438466073 +INFO:root: +[Valid][27][UTT][ MSE = 0.2620 | LCC = 0.9438 | SRCC = 0.9457 ] +INFO:root:Average loss=0.10740320038795471 +INFO:root: +[Valid][28][UTT][ MSE = 0.2798 | LCC = 0.9415 | SRCC = 0.9423 ] +INFO:root:Average loss=0.10587838616967202 +INFO:root: +[Valid][29][UTT][ MSE = 0.2491 | LCC = 0.9473 | SRCC = 0.9481 ] +INFO:root:Average loss=0.10104695534706115 +INFO:root: +[Valid][30][UTT][ MSE = 0.2823 | LCC = 0.9426 | SRCC = 0.9436 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2577 | LCC = 0.9492 | SRCC = 0.9508 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ebea298889b8ad103d837c26d914d61cec1075d7 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:12f383bc878e6c60f2b599edfd6ad488fcd55010944c16a4bcb714acc565641c +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..eeaabeb0cf13fa78af89b6f4bababc18bdbf0f76 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8e376eec6922b5767743f28df42e3c1809bf69fafbff17db2346535923d17065 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b21ceed8c10ce37da8e39b10561a794e4c7a0681 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.33730274510383607 +INFO:root: +[Valid][0][UTT][ MSE = 0.4037 | LCC = 0.9215 | SRCC = 0.9261 ] +INFO:root:Average loss=0.2142590662240982 +INFO:root: +[Valid][1][UTT][ MSE = 0.3578 | LCC = 0.9280 | SRCC = 0.9322 ] +INFO:root:Average loss=0.19527875965833663 +INFO:root: +[Valid][2][UTT][ MSE = 0.3405 | LCC = 0.9313 | SRCC = 0.9360 ] +INFO:root:Average loss=0.18440199851989747 +INFO:root: +[Valid][3][UTT][ MSE = 0.3459 | LCC = 0.9345 | SRCC = 0.9376 ] +INFO:root:Average loss=0.17580604308843611 +INFO:root: +[Valid][4][UTT][ MSE = 0.3366 | LCC = 0.9392 | SRCC = 0.9408 ] +INFO:root:Average loss=0.16639700645208358 +INFO:root: +[Valid][5][UTT][ MSE = 0.2909 | LCC = 0.9334 | SRCC = 0.9407 ] +INFO:root:Average loss=0.16454034465551376 +INFO:root: +[Valid][6][UTT][ MSE = 0.2827 | LCC = 0.9383 | SRCC = 0.9433 ] +INFO:root:Average loss=0.1614515143632889 +INFO:root: +[Valid][7][UTT][ MSE = 0.2917 | LCC = 0.9368 | SRCC = 0.9411 ] +INFO:root:Average loss=0.15439063566923142 +INFO:root: +[Valid][8][UTT][ MSE = 0.2807 | LCC = 0.9351 | SRCC = 0.9420 ] +INFO:root:Average loss=0.14838156765699387 +INFO:root: +[Valid][9][UTT][ MSE = 0.3161 | LCC = 0.9438 | SRCC = 0.9452 ] +INFO:root:Average loss=0.14588492786884308 +INFO:root: +[Valid][10][UTT][ MSE = 0.2839 | LCC = 0.9429 | SRCC = 0.9443 ] +INFO:root:Average loss=0.14158741343021392 +INFO:root: +[Valid][11][UTT][ MSE = 0.3041 | LCC = 0.9429 | SRCC = 0.9454 ] +INFO:root:Average loss=0.14156825423240663 +INFO:root: +[Valid][12][UTT][ MSE = 0.3016 | LCC = 0.9431 | SRCC = 0.9464 ] +INFO:root:Average loss=0.1340863714814186 +INFO:root: +[Valid][13][UTT][ MSE = 0.2778 | LCC = 0.9362 | SRCC = 0.9423 ] +INFO:root:Average loss=0.13787866175174712 +INFO:root: +[Valid][14][UTT][ MSE = 0.2814 | LCC = 0.9443 | SRCC = 0.9452 ] +INFO:root:Average loss=0.13044489359855652 +INFO:root: +[Valid][15][UTT][ MSE = 0.2989 | LCC = 0.9424 | SRCC = 0.9445 ] +INFO:root:Average loss=0.13094159084558488 +INFO:root: +[Valid][16][UTT][ MSE = 0.2979 | LCC = 0.9461 | SRCC = 0.9466 ] +INFO:root:Average loss=0.12586733746528625 +INFO:root: +[Valid][17][UTT][ MSE = 0.3039 | LCC = 0.9447 | SRCC = 0.9463 ] +INFO:root:Average loss=0.12470481485128403 +INFO:root: +[Valid][18][UTT][ MSE = 0.2882 | LCC = 0.9443 | SRCC = 0.9458 ] +INFO:root:Average loss=0.12136745718121529 +INFO:root: +[Valid][19][UTT][ MSE = 0.2832 | LCC = 0.9461 | SRCC = 0.9475 ] +INFO:root:Average loss=0.12103758743405342 +INFO:root: +[Valid][20][UTT][ MSE = 0.2822 | LCC = 0.9451 | SRCC = 0.9456 ] +INFO:root:Average loss=0.11863759723305702 +INFO:root: +[Valid][21][UTT][ MSE = 0.2488 | LCC = 0.9423 | SRCC = 0.9445 ] +INFO:root:Average loss=0.11581328395009041 +INFO:root: +[Valid][22][UTT][ MSE = 0.2869 | LCC = 0.9466 | SRCC = 0.9475 ] +INFO:root:Average loss=0.1166506884098053 +INFO:root: +[Valid][23][UTT][ MSE = 0.2625 | LCC = 0.9462 | SRCC = 0.9468 ] +INFO:root:Average loss=0.11215945771336555 +INFO:root: +[Valid][24][UTT][ MSE = 0.2584 | LCC = 0.9459 | SRCC = 0.9472 ] +INFO:root:Average loss=0.10747527801990509 +INFO:root: +[Valid][25][UTT][ MSE = 0.2631 | LCC = 0.9471 | SRCC = 0.9478 ] +INFO:root:Average loss=0.1106845289170742 +INFO:root: +[Valid][26][UTT][ MSE = 0.2892 | LCC = 0.9436 | SRCC = 0.9466 ] +INFO:root:Average loss=0.1072735739350319 +INFO:root: +[Valid][27][UTT][ MSE = 0.3059 | LCC = 0.9464 | SRCC = 0.9475 ] +INFO:root:Average loss=0.10253626757860183 +INFO:root: +[Valid][28][UTT][ MSE = 0.2816 | LCC = 0.9434 | SRCC = 0.9451 ] +INFO:root:Average loss=0.10386406436562538 +INFO:root: +[Valid][29][UTT][ MSE = 0.2850 | LCC = 0.9454 | SRCC = 0.9465 ] +INFO:root:Average loss=0.09896584179997445 +INFO:root: +[Valid][30][UTT][ MSE = 0.2795 | LCC = 0.9468 | SRCC = 0.9479 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2694 | LCC = 0.9504 | SRCC = 0.9513 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6466b81d481be8ba9cc130f97d41592698a3a195 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3dbc2b17cd1f0e14dd03d951a1ff88f65a2d59d4a0bef81c1de56e505141df2e +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..92e460aafea9228d66dd660f4c680bce14473a83 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fa911bc890f704734cd6c769977638161f4608b7cf503936478010f2b785959b +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..581a04054e63b3438cf821893219b57767b814f8 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3004972951412201 +INFO:root: +[Valid][0][UTT][ MSE = 0.3638 | LCC = 0.9244 | SRCC = 0.9262 ] +INFO:root:Average loss=0.21979464864730835 +INFO:root: +[Valid][1][UTT][ MSE = 0.3378 | LCC = 0.9291 | SRCC = 0.9310 ] +INFO:root:Average loss=0.1987296858429909 +INFO:root: +[Valid][2][UTT][ MSE = 0.3091 | LCC = 0.9330 | SRCC = 0.9355 ] +INFO:root:Average loss=0.18633621227741243 +INFO:root: +[Valid][3][UTT][ MSE = 0.3015 | LCC = 0.9374 | SRCC = 0.9384 ] +INFO:root:Average loss=0.18024483144283296 +INFO:root: +[Valid][4][UTT][ MSE = 0.3293 | LCC = 0.9319 | SRCC = 0.9354 ] +INFO:root:Average loss=0.17140125864744185 +INFO:root: +[Valid][5][UTT][ MSE = 0.2747 | LCC = 0.9383 | SRCC = 0.9428 ] +INFO:root:Average loss=0.16464992797374725 +INFO:root: +[Valid][6][UTT][ MSE = 0.2732 | LCC = 0.9389 | SRCC = 0.9423 ] +INFO:root:Average loss=0.15995975017547606 +INFO:root: +[Valid][7][UTT][ MSE = 0.3652 | LCC = 0.9338 | SRCC = 0.9383 ] +INFO:root:Average loss=0.15429225158691406 +INFO:root: +[Valid][8][UTT][ MSE = 0.3230 | LCC = 0.9401 | SRCC = 0.9432 ] +INFO:root:Average loss=0.15190215200185775 +INFO:root: +[Valid][9][UTT][ MSE = 0.3056 | LCC = 0.9381 | SRCC = 0.9419 ] +INFO:root:Average loss=0.14335860919952392 +INFO:root: +[Valid][10][UTT][ MSE = 0.2926 | LCC = 0.9421 | SRCC = 0.9441 ] +INFO:root:Average loss=0.14201903700828553 +INFO:root: +[Valid][11][UTT][ MSE = 0.2702 | LCC = 0.9397 | SRCC = 0.9428 ] +INFO:root:Average loss=0.14028863483667373 +INFO:root: +[Valid][12][UTT][ MSE = 0.2822 | LCC = 0.9428 | SRCC = 0.9444 ] +INFO:root:Average loss=0.1416088863015175 +INFO:root: +[Valid][13][UTT][ MSE = 0.2641 | LCC = 0.9420 | SRCC = 0.9460 ] +INFO:root:Average loss=0.13469027855992316 +INFO:root: +[Valid][14][UTT][ MSE = 0.2417 | LCC = 0.9440 | SRCC = 0.9458 ] +INFO:root:Average loss=0.1347194178700447 +INFO:root: +[Valid][15][UTT][ MSE = 0.2412 | LCC = 0.9450 | SRCC = 0.9465 ] +INFO:root:Average loss=0.13070469346642494 +INFO:root: +[Valid][16][UTT][ MSE = 0.2479 | LCC = 0.9397 | SRCC = 0.9452 ] +INFO:root:Average loss=0.1294267457127571 +INFO:root: +[Valid][17][UTT][ MSE = 0.2358 | LCC = 0.9380 | SRCC = 0.9425 ] +INFO:root:Average loss=0.12901247501373292 +INFO:root: +[Valid][18][UTT][ MSE = 0.2862 | LCC = 0.9444 | SRCC = 0.9457 ] +INFO:root:Average loss=0.12441481542587281 +INFO:root: +[Valid][19][UTT][ MSE = 0.2733 | LCC = 0.9438 | SRCC = 0.9448 ] +INFO:root:Average loss=0.12074654588103294 +INFO:root: +[Valid][20][UTT][ MSE = 0.2770 | LCC = 0.9448 | SRCC = 0.9456 ] +INFO:root:Average loss=0.1194706355035305 +INFO:root: +[Valid][21][UTT][ MSE = 0.2579 | LCC = 0.9413 | SRCC = 0.9418 ] +INFO:root:Average loss=0.11663777416944504 +INFO:root: +[Valid][22][UTT][ MSE = 0.2848 | LCC = 0.9431 | SRCC = 0.9447 ] +INFO:root:Average loss=0.11896583434939384 +INFO:root: +[Valid][23][UTT][ MSE = 0.2503 | LCC = 0.9455 | SRCC = 0.9467 ] +INFO:root:Average loss=0.11396031278371811 +INFO:root: +[Valid][24][UTT][ MSE = 0.2380 | LCC = 0.9459 | SRCC = 0.9473 ] +INFO:root:Average loss=0.11150911122560501 +INFO:root: +[Valid][25][UTT][ MSE = 0.2645 | LCC = 0.9436 | SRCC = 0.9449 ] +INFO:root:Average loss=0.1078699756860733 +INFO:root: +[Valid][26][UTT][ MSE = 0.2584 | LCC = 0.9418 | SRCC = 0.9448 ] +INFO:root:Average loss=0.10812105688452721 +INFO:root: +[Valid][27][UTT][ MSE = 0.2315 | LCC = 0.9442 | SRCC = 0.9460 ] +INFO:root:Average loss=0.10911165183782577 +INFO:root: +[Valid][28][UTT][ MSE = 0.2536 | LCC = 0.9442 | SRCC = 0.9442 ] +INFO:root:Average loss=0.1040264932513237 +INFO:root: +[Valid][29][UTT][ MSE = 0.2650 | LCC = 0.9432 | SRCC = 0.9443 ] +INFO:root:Average loss=0.10362424784898758 +INFO:root: +[Valid][30][UTT][ MSE = 0.2840 | LCC = 0.9422 | SRCC = 0.9430 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2376 | LCC = 0.9489 | SRCC = 0.9504 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0ab9326cf71e81720a1626a37c0e3ed060a6103a --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:95cfef43959838a879a7f71b691edba9fd0ed74e97f85be5b0d048ce630b9224 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..921a07ca0c612080a3f498e8ba7d32010a95dfb3 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a941262ed477e2e9c461ff1b4f75769341541bddc79213b7541deac96e86066f +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..40a71af3635dcac44677f1974ae8370b3eb73d51 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3057872973680496 +INFO:root: +[Valid][0][UTT][ MSE = 0.3050 | LCC = 0.9190 | SRCC = 0.9213 ] +INFO:root:Average loss=0.21334175491333007 +INFO:root: +[Valid][1][UTT][ MSE = 0.2713 | LCC = 0.9257 | SRCC = 0.9268 ] +INFO:root:Average loss=0.20008187568187713 +INFO:root: +[Valid][2][UTT][ MSE = 0.2431 | LCC = 0.9313 | SRCC = 0.9325 ] +INFO:root:Average loss=0.18726534825563432 +INFO:root: +[Valid][3][UTT][ MSE = 0.2949 | LCC = 0.9382 | SRCC = 0.9393 ] +INFO:root:Average loss=0.1743638710975647 +INFO:root: +[Valid][4][UTT][ MSE = 0.2761 | LCC = 0.9395 | SRCC = 0.9404 ] +INFO:root:Average loss=0.16944974756240844 +INFO:root: +[Valid][5][UTT][ MSE = 0.2842 | LCC = 0.9374 | SRCC = 0.9386 ] +INFO:root:Average loss=0.16276761072874069 +INFO:root: +[Valid][6][UTT][ MSE = 0.2849 | LCC = 0.9413 | SRCC = 0.9432 ] +INFO:root:Average loss=0.15679599469900132 +INFO:root: +[Valid][7][UTT][ MSE = 0.2781 | LCC = 0.9415 | SRCC = 0.9425 ] +INFO:root:Average loss=0.15583100211620332 +INFO:root: +[Valid][8][UTT][ MSE = 0.2840 | LCC = 0.9410 | SRCC = 0.9426 ] +INFO:root:Average loss=0.14734458112716675 +INFO:root: +[Valid][9][UTT][ MSE = 0.2594 | LCC = 0.9396 | SRCC = 0.9421 ] +INFO:root:Average loss=0.1456520709991455 +INFO:root: +[Valid][10][UTT][ MSE = 0.2610 | LCC = 0.9401 | SRCC = 0.9419 ] +INFO:root:Average loss=0.1437021780014038 +INFO:root: +[Valid][11][UTT][ MSE = 0.2667 | LCC = 0.9407 | SRCC = 0.9429 ] +INFO:root:Average loss=0.14000432831048965 +INFO:root: +[Valid][12][UTT][ MSE = 0.2972 | LCC = 0.9401 | SRCC = 0.9437 ] +INFO:root:Average loss=0.13583118903636932 +INFO:root: +[Valid][13][UTT][ MSE = 0.2379 | LCC = 0.9420 | SRCC = 0.9445 ] +INFO:root:Average loss=0.13488049647212028 +INFO:root: +[Valid][14][UTT][ MSE = 0.2426 | LCC = 0.9427 | SRCC = 0.9445 ] +INFO:root:Average loss=0.13368328791856765 +INFO:root: +[Valid][15][UTT][ MSE = 0.2568 | LCC = 0.9427 | SRCC = 0.9431 ] +INFO:root:Average loss=0.12734372431039812 +INFO:root: +[Valid][16][UTT][ MSE = 0.2233 | LCC = 0.9461 | SRCC = 0.9464 ] +INFO:root:Average loss=0.12610205405950548 +INFO:root: +[Valid][17][UTT][ MSE = 0.2434 | LCC = 0.9404 | SRCC = 0.9440 ] +INFO:root:Average loss=0.12158321300148964 +INFO:root: +[Valid][18][UTT][ MSE = 0.2450 | LCC = 0.9431 | SRCC = 0.9444 ] +INFO:root:Average loss=0.12101853209733963 +INFO:root: +[Valid][19][UTT][ MSE = 0.2782 | LCC = 0.9457 | SRCC = 0.9464 ] +INFO:root:Average loss=0.11857024317979813 +INFO:root: +[Valid][20][UTT][ MSE = 0.2957 | LCC = 0.9406 | SRCC = 0.9408 ] +INFO:root:Average loss=0.1189413076043129 +INFO:root: +[Valid][21][UTT][ MSE = 0.2425 | LCC = 0.9434 | SRCC = 0.9453 ] +INFO:root:Average loss=0.11645249211788178 +INFO:root: +[Valid][22][UTT][ MSE = 0.2458 | LCC = 0.9426 | SRCC = 0.9454 ] +INFO:root:Average loss=0.1136017809510231 +INFO:root: +[Valid][23][UTT][ MSE = 0.2403 | LCC = 0.9445 | SRCC = 0.9453 ] +INFO:root:Average loss=0.1114174634218216 +INFO:root: +[Valid][24][UTT][ MSE = 0.2273 | LCC = 0.9439 | SRCC = 0.9460 ] +INFO:root:Average loss=0.10805136185884476 +INFO:root: +[Valid][25][UTT][ MSE = 0.2503 | LCC = 0.9425 | SRCC = 0.9439 ] +INFO:root:Average loss=0.10658091709017753 +INFO:root: +[Valid][26][UTT][ MSE = 0.2650 | LCC = 0.9462 | SRCC = 0.9467 ] +INFO:root:Average loss=0.10718781271576881 +INFO:root: +[Valid][27][UTT][ MSE = 0.2267 | LCC = 0.9431 | SRCC = 0.9434 ] +INFO:root:Average loss=0.10434659197926521 +INFO:root: +[Valid][28][UTT][ MSE = 0.2502 | LCC = 0.9446 | SRCC = 0.9451 ] +INFO:root:Average loss=0.1049755728840828 +INFO:root: +[Valid][29][UTT][ MSE = 0.2315 | LCC = 0.9434 | SRCC = 0.9439 ] +INFO:root:Average loss=0.09871793574094773 +INFO:root: +[Valid][30][UTT][ MSE = 0.2187 | LCC = 0.9431 | SRCC = 0.9444 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2567 | LCC = 0.9511 | SRCC = 0.9514 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b020d913badf3c6a0bd7ea2594723e7defcd4087 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:96a1b27f67be415e1fffc3f5c00a82731eac3ea3d9db48a05979aac13c4d0780 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..bf113369c97b0e153bcb5f68baeb8c85cbb35ce2 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af64206f458d8be33ebf0e1b8f4e45f7f69facd0fb1dd3585d1e5f0c455d75fe +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c57546dda519734ece7d7f7b25316e8afe46deb2 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer11/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.32962817084789275 +INFO:root: +[Valid][0][UTT][ MSE = 0.3433 | LCC = 0.9225 | SRCC = 0.9256 ] +INFO:root:Average loss=0.21765461230278016 +INFO:root: +[Valid][1][UTT][ MSE = 0.3471 | LCC = 0.9282 | SRCC = 0.9316 ] +INFO:root:Average loss=0.20246832925081254 +INFO:root: +[Valid][2][UTT][ MSE = 0.3637 | LCC = 0.9316 | SRCC = 0.9355 ] +INFO:root:Average loss=0.18324586504697798 +INFO:root: +[Valid][3][UTT][ MSE = 0.3155 | LCC = 0.9345 | SRCC = 0.9392 ] +INFO:root:Average loss=0.17448062747716903 +INFO:root: +[Valid][4][UTT][ MSE = 0.3333 | LCC = 0.9388 | SRCC = 0.9409 ] +INFO:root:Average loss=0.16527196753025056 +INFO:root: +[Valid][5][UTT][ MSE = 0.3100 | LCC = 0.9369 | SRCC = 0.9393 ] +INFO:root:Average loss=0.1614203817844391 +INFO:root: +[Valid][6][UTT][ MSE = 0.3144 | LCC = 0.9416 | SRCC = 0.9431 ] +INFO:root:Average loss=0.1583761940598488 +INFO:root: +[Valid][7][UTT][ MSE = 0.2832 | LCC = 0.9411 | SRCC = 0.9425 ] +INFO:root:Average loss=0.1506015293598175 +INFO:root: +[Valid][8][UTT][ MSE = 0.2963 | LCC = 0.9402 | SRCC = 0.9413 ] +INFO:root:Average loss=0.1511352277994156 +INFO:root: +[Valid][9][UTT][ MSE = 0.2416 | LCC = 0.9411 | SRCC = 0.9444 ] +INFO:root:Average loss=0.14804847854375838 +INFO:root: +[Valid][10][UTT][ MSE = 0.2818 | LCC = 0.9451 | SRCC = 0.9462 ] +INFO:root:Average loss=0.1395268673300743 +INFO:root: +[Valid][11][UTT][ MSE = 0.3385 | LCC = 0.9406 | SRCC = 0.9438 ] +INFO:root:Average loss=0.13978132003545762 +INFO:root: +[Valid][12][UTT][ MSE = 0.2890 | LCC = 0.9457 | SRCC = 0.9471 ] +INFO:root:Average loss=0.13119865554571153 +INFO:root: +[Valid][13][UTT][ MSE = 0.2815 | LCC = 0.9403 | SRCC = 0.9433 ] +INFO:root:Average loss=0.13012206494808198 +INFO:root: +[Valid][14][UTT][ MSE = 0.3449 | LCC = 0.9418 | SRCC = 0.9437 ] +INFO:root:Average loss=0.12820315054059028 +INFO:root: +[Valid][15][UTT][ MSE = 0.2716 | LCC = 0.9453 | SRCC = 0.9468 ] +INFO:root:Average loss=0.12708798968791962 +INFO:root: +[Valid][16][UTT][ MSE = 0.2880 | LCC = 0.9451 | SRCC = 0.9460 ] +INFO:root:Average loss=0.12375780215859414 +INFO:root: +[Valid][17][UTT][ MSE = 0.2827 | LCC = 0.9411 | SRCC = 0.9424 ] +INFO:root:Average loss=0.12200377660989761 +INFO:root: +[Valid][18][UTT][ MSE = 0.2495 | LCC = 0.9452 | SRCC = 0.9470 ] +INFO:root:Average loss=0.118022929251194 +INFO:root: +[Valid][19][UTT][ MSE = 0.2824 | LCC = 0.9463 | SRCC = 0.9474 ] +INFO:root:Average loss=0.1181783444583416 +INFO:root: +[Valid][20][UTT][ MSE = 0.2665 | LCC = 0.9466 | SRCC = 0.9471 ] +INFO:root:Average loss=0.1227158949971199 +INFO:root: +[Valid][21][UTT][ MSE = 0.2497 | LCC = 0.9407 | SRCC = 0.9419 ] +INFO:root:Average loss=0.1172878388762474 +INFO:root: +[Valid][22][UTT][ MSE = 0.3052 | LCC = 0.9431 | SRCC = 0.9436 ] +INFO:root:Average loss=0.11388633459806442 +INFO:root: +[Valid][23][UTT][ MSE = 0.3004 | LCC = 0.9446 | SRCC = 0.9462 ] +INFO:root:Average loss=0.11103305742144584 +INFO:root: +[Valid][24][UTT][ MSE = 0.2675 | LCC = 0.9442 | SRCC = 0.9465 ] +INFO:root:Average loss=0.10894894480705261 +INFO:root: +[Valid][25][UTT][ MSE = 0.2562 | LCC = 0.9473 | SRCC = 0.9484 ] +INFO:root:Average loss=0.10587113630771637 +INFO:root: +[Valid][26][UTT][ MSE = 0.2570 | LCC = 0.9447 | SRCC = 0.9463 ] +INFO:root:Average loss=0.10534653025865555 +INFO:root: +[Valid][27][UTT][ MSE = 0.2614 | LCC = 0.9466 | SRCC = 0.9473 ] +INFO:root:Average loss=0.10389642488956452 +INFO:root: +[Valid][28][UTT][ MSE = 0.2583 | LCC = 0.9442 | SRCC = 0.9460 ] +INFO:root:Average loss=0.10427488285303116 +INFO:root: +[Valid][29][UTT][ MSE = 0.3117 | LCC = 0.9450 | SRCC = 0.9457 ] +INFO:root:Average loss=0.10079590022563935 +INFO:root: +[Valid][30][UTT][ MSE = 0.2910 | LCC = 0.9448 | SRCC = 0.9457 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2532 | LCC = 0.9505 | SRCC = 0.9520 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0d80b3947e3c16928071c09aec29c2e14cf72143 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3226b063610e4a0d849ad015960927d47495482567a1bcf6f0f4624c87e4a104 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..5d0cddcb3cdd1c58f98be293e8f3ed5be7b1b8fb --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a2570358d56e4e12020830b649d169c922b0ae02fde72ea133c8fefe3c48567 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5eec9c150dfd64672f067bbd79b2d034bfbdbdf2 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.45611376655101776 +INFO:root: +[Valid][0][UTT][ MSE = 0.4411 | LCC = 0.9046 | SRCC = 0.9103 ] +INFO:root:Average loss=0.1970333775281906 +INFO:root: +[Valid][1][UTT][ MSE = 0.3090 | LCC = 0.9351 | SRCC = 0.9384 ] +INFO:root:Average loss=0.15867075049877166 +INFO:root: +[Valid][2][UTT][ MSE = 0.2527 | LCC = 0.9442 | SRCC = 0.9462 ] +INFO:root:Average loss=0.13815816563367844 +INFO:root: +[Valid][3][UTT][ MSE = 0.2416 | LCC = 0.9501 | SRCC = 0.9508 ] +INFO:root:Average loss=0.12584821099042892 +INFO:root: +[Valid][4][UTT][ MSE = 0.2329 | LCC = 0.9500 | SRCC = 0.9516 ] +INFO:root:Average loss=0.11227120947837829 +INFO:root: +[Valid][5][UTT][ MSE = 0.2145 | LCC = 0.9553 | SRCC = 0.9563 ] +INFO:root:Average loss=0.10358537271618844 +INFO:root: +[Valid][6][UTT][ MSE = 0.2012 | LCC = 0.9549 | SRCC = 0.9549 ] +INFO:root:Average loss=0.09809633529186249 +INFO:root: +[Valid][7][UTT][ MSE = 0.2440 | LCC = 0.9556 | SRCC = 0.9568 ] +INFO:root:Average loss=0.0884268479347229 +INFO:root: +[Valid][8][UTT][ MSE = 0.2210 | LCC = 0.9567 | SRCC = 0.9584 ] +INFO:root:Average loss=0.085064682751894 +INFO:root: +[Valid][9][UTT][ MSE = 0.2626 | LCC = 0.9545 | SRCC = 0.9549 ] +INFO:root:Average loss=0.08162278625369072 +INFO:root: +[Valid][10][UTT][ MSE = 0.2128 | LCC = 0.9585 | SRCC = 0.9597 ] +INFO:root:Average loss=0.07270104986429214 +INFO:root: +[Valid][11][UTT][ MSE = 0.2143 | LCC = 0.9561 | SRCC = 0.9562 ] +INFO:root:Average loss=0.07030773308873177 +INFO:root: +[Valid][12][UTT][ MSE = 0.2288 | LCC = 0.9558 | SRCC = 0.9564 ] +INFO:root:Average loss=0.0704692345559597 +INFO:root: +[Valid][13][UTT][ MSE = 0.2559 | LCC = 0.9494 | SRCC = 0.9499 ] +INFO:root:Average loss=0.06699200981855392 +INFO:root: +[Valid][14][UTT][ MSE = 0.2313 | LCC = 0.9535 | SRCC = 0.9566 ] +INFO:root:Average loss=0.06270150139927864 +INFO:root: +[Valid][15][UTT][ MSE = 0.2133 | LCC = 0.9583 | SRCC = 0.9590 ] +INFO:root:Average loss=0.059093268603086474 +INFO:root: +[Valid][16][UTT][ MSE = 0.2161 | LCC = 0.9576 | SRCC = 0.9583 ] +INFO:root:Average loss=0.055670050889253614 +INFO:root: +[Valid][17][UTT][ MSE = 0.1896 | LCC = 0.9582 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05458234244585037 +INFO:root: +[Valid][18][UTT][ MSE = 0.2131 | LCC = 0.9603 | SRCC = 0.9607 ] +INFO:root:Average loss=0.052345961824059484 +INFO:root: +[Valid][19][UTT][ MSE = 0.2022 | LCC = 0.9598 | SRCC = 0.9601 ] +INFO:root:Average loss=0.05100585119426251 +INFO:root: +[Valid][20][UTT][ MSE = 0.1827 | LCC = 0.9602 | SRCC = 0.9606 ] +INFO:root:Average loss=0.050096614211797716 +INFO:root: +[Valid][21][UTT][ MSE = 0.2093 | LCC = 0.9573 | SRCC = 0.9577 ] +INFO:root:Average loss=0.04660657887160778 +INFO:root: +[Valid][22][UTT][ MSE = 0.2023 | LCC = 0.9591 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04721663217246532 +INFO:root: +[Valid][23][UTT][ MSE = 0.2201 | LCC = 0.9563 | SRCC = 0.9577 ] +INFO:root:Average loss=0.044515154719352724 +INFO:root: +[Valid][24][UTT][ MSE = 0.2219 | LCC = 0.9563 | SRCC = 0.9573 ] +INFO:root:Average loss=0.04295226721465588 +INFO:root: +[Valid][25][UTT][ MSE = 0.2038 | LCC = 0.9570 | SRCC = 0.9583 ] +INFO:root:Average loss=0.041227276965975765 +INFO:root: +[Valid][26][UTT][ MSE = 0.1879 | LCC = 0.9586 | SRCC = 0.9598 ] +INFO:root:Average loss=0.03922981142997742 +INFO:root: +[Valid][27][UTT][ MSE = 0.2122 | LCC = 0.9598 | SRCC = 0.9598 ] +INFO:root:Average loss=0.03745564274489879 +INFO:root: +[Valid][28][UTT][ MSE = 0.1990 | LCC = 0.9586 | SRCC = 0.9592 ] +INFO:root:Average loss=0.04005966849625111 +INFO:root: +[Valid][29][UTT][ MSE = 0.2088 | LCC = 0.9591 | SRCC = 0.9596 ] +INFO:root:Average loss=0.037854685992002486 +INFO:root: +[Valid][30][UTT][ MSE = 0.2114 | LCC = 0.9584 | SRCC = 0.9593 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2051 | LCC = 0.9618 | SRCC = 0.9624 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..88eaedddfe60ef726771f627f613e28c8613a090 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5c40343a292668b89083906823b1ee03a6e17e40ebcff4417966aad47d03dcd1 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..63b7435dbe972453c96348421dc4cfbed603e754 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:94c5ee8703237911d6009d0676d617e375f2557e4fa2115ceb19c9c4fbb175cd +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e59df82be78a4514ff824df1d5112e34d6becad1 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42488954079151153 +INFO:root: +[Valid][0][UTT][ MSE = 0.3505 | LCC = 0.9172 | SRCC = 0.9173 ] +INFO:root:Average loss=0.20006745278835297 +INFO:root: +[Valid][1][UTT][ MSE = 0.2526 | LCC = 0.9361 | SRCC = 0.9384 ] +INFO:root:Average loss=0.15733577126264572 +INFO:root: +[Valid][2][UTT][ MSE = 0.2196 | LCC = 0.9450 | SRCC = 0.9455 ] +INFO:root:Average loss=0.138455528318882 +INFO:root: +[Valid][3][UTT][ MSE = 0.2021 | LCC = 0.9484 | SRCC = 0.9502 ] +INFO:root:Average loss=0.12403804382681846 +INFO:root: +[Valid][4][UTT][ MSE = 0.2052 | LCC = 0.9512 | SRCC = 0.9535 ] +INFO:root:Average loss=0.11070508277416229 +INFO:root: +[Valid][5][UTT][ MSE = 0.1955 | LCC = 0.9542 | SRCC = 0.9551 ] +INFO:root:Average loss=0.10454723256826401 +INFO:root: +[Valid][6][UTT][ MSE = 0.2003 | LCC = 0.9551 | SRCC = 0.9555 ] +INFO:root:Average loss=0.09653816547989845 +INFO:root: +[Valid][7][UTT][ MSE = 0.2400 | LCC = 0.9554 | SRCC = 0.9560 ] +INFO:root:Average loss=0.09184713298082352 +INFO:root: +[Valid][8][UTT][ MSE = 0.1877 | LCC = 0.9566 | SRCC = 0.9574 ] +INFO:root:Average loss=0.08482714533805848 +INFO:root: +[Valid][9][UTT][ MSE = 0.1827 | LCC = 0.9573 | SRCC = 0.9583 ] +INFO:root:Average loss=0.08117881748080254 +INFO:root: +[Valid][10][UTT][ MSE = 0.1810 | LCC = 0.9575 | SRCC = 0.9582 ] +INFO:root:Average loss=0.07682055327296257 +INFO:root: +[Valid][11][UTT][ MSE = 0.1819 | LCC = 0.9576 | SRCC = 0.9584 ] +INFO:root:Average loss=0.06902608522772789 +INFO:root: +[Valid][12][UTT][ MSE = 0.1942 | LCC = 0.9584 | SRCC = 0.9589 ] +INFO:root:Average loss=0.07143232786655426 +INFO:root: +[Valid][13][UTT][ MSE = 0.2067 | LCC = 0.9585 | SRCC = 0.9592 ] +INFO:root:Average loss=0.06736758887767792 +INFO:root: +[Valid][14][UTT][ MSE = 0.2002 | LCC = 0.9575 | SRCC = 0.9581 ] +INFO:root:Average loss=0.06219977399706841 +INFO:root: +[Valid][15][UTT][ MSE = 0.1966 | LCC = 0.9581 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05911544415354729 +INFO:root: +[Valid][16][UTT][ MSE = 0.2271 | LCC = 0.9587 | SRCC = 0.9593 ] +INFO:root:Average loss=0.056372369930148124 +INFO:root: +[Valid][17][UTT][ MSE = 0.1907 | LCC = 0.9586 | SRCC = 0.9590 ] +INFO:root:Average loss=0.0551698026061058 +INFO:root: +[Valid][18][UTT][ MSE = 0.2035 | LCC = 0.9590 | SRCC = 0.9597 ] +INFO:root:Average loss=0.0515468453168869 +INFO:root: +[Valid][19][UTT][ MSE = 0.1820 | LCC = 0.9588 | SRCC = 0.9594 ] +INFO:root:Average loss=0.05290971378982067 +INFO:root: +[Valid][20][UTT][ MSE = 0.1669 | LCC = 0.9592 | SRCC = 0.9599 ] +INFO:root:Average loss=0.049373692080378534 +INFO:root: +[Valid][21][UTT][ MSE = 0.2069 | LCC = 0.9591 | SRCC = 0.9606 ] +INFO:root:Average loss=0.045949924901127814 +INFO:root: +[Valid][22][UTT][ MSE = 0.1816 | LCC = 0.9596 | SRCC = 0.9604 ] +INFO:root:Average loss=0.04442909847199917 +INFO:root: +[Valid][23][UTT][ MSE = 0.1850 | LCC = 0.9588 | SRCC = 0.9599 ] +INFO:root:Average loss=0.045120372638106344 +INFO:root: +[Valid][24][UTT][ MSE = 0.1765 | LCC = 0.9589 | SRCC = 0.9601 ] +INFO:root:Average loss=0.04316778801381588 +INFO:root: +[Valid][25][UTT][ MSE = 0.1993 | LCC = 0.9597 | SRCC = 0.9606 ] +INFO:root:Average loss=0.04083941039443016 +INFO:root: +[Valid][26][UTT][ MSE = 0.1885 | LCC = 0.9597 | SRCC = 0.9610 ] +INFO:root:Average loss=0.0397479695379734 +INFO:root: +[Valid][27][UTT][ MSE = 0.1840 | LCC = 0.9598 | SRCC = 0.9608 ] +INFO:root:Average loss=0.03887163041532039 +INFO:root: +[Valid][28][UTT][ MSE = 0.1804 | LCC = 0.9595 | SRCC = 0.9604 ] +INFO:root:Average loss=0.040229948356747626 +INFO:root: +[Valid][29][UTT][ MSE = 0.1716 | LCC = 0.9602 | SRCC = 0.9611 ] +INFO:root:Average loss=0.0361282053142786 +INFO:root: +[Valid][30][UTT][ MSE = 0.1811 | LCC = 0.9573 | SRCC = 0.9579 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1671 | LCC = 0.9615 | SRCC = 0.9620 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..fe1876a58b138cf32088d61a9ebca051d180c1bd --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:102c30b92a49d689b9d189db53c93c88d11d199e8d7f02146c0380bf328c1e4d +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d6955bd21f77f5681fe666f0908ef48694350f39 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43a969594ce3f0e9a93d8fb6682c3c17a7467ecb9784596f70d7b96f69f63eb5 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..60459f3f0c8856baa19627a65a207b044cf3a2dd --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46851400566101076 +INFO:root: +[Valid][0][UTT][ MSE = 0.3740 | LCC = 0.9140 | SRCC = 0.9166 ] +INFO:root:Average loss=0.195388265311718 +INFO:root: +[Valid][1][UTT][ MSE = 0.2793 | LCC = 0.9371 | SRCC = 0.9381 ] +INFO:root:Average loss=0.15347489327192307 +INFO:root: +[Valid][2][UTT][ MSE = 0.2897 | LCC = 0.9446 | SRCC = 0.9464 ] +INFO:root:Average loss=0.13501997423171996 +INFO:root: +[Valid][3][UTT][ MSE = 0.2127 | LCC = 0.9501 | SRCC = 0.9518 ] +INFO:root:Average loss=0.1204314700961113 +INFO:root: +[Valid][4][UTT][ MSE = 0.2447 | LCC = 0.9481 | SRCC = 0.9502 ] +INFO:root:Average loss=0.11057447209954262 +INFO:root: +[Valid][5][UTT][ MSE = 0.2092 | LCC = 0.9510 | SRCC = 0.9539 ] +INFO:root:Average loss=0.1017061361670494 +INFO:root: +[Valid][6][UTT][ MSE = 0.2338 | LCC = 0.9562 | SRCC = 0.9575 ] +INFO:root:Average loss=0.09917033812403679 +INFO:root: +[Valid][7][UTT][ MSE = 0.2346 | LCC = 0.9568 | SRCC = 0.9582 ] +INFO:root:Average loss=0.08987247183918953 +INFO:root: +[Valid][8][UTT][ MSE = 0.2026 | LCC = 0.9569 | SRCC = 0.9579 ] +INFO:root:Average loss=0.08588396880030633 +INFO:root: +[Valid][9][UTT][ MSE = 0.2266 | LCC = 0.9569 | SRCC = 0.9584 ] +INFO:root:Average loss=0.07691318072378635 +INFO:root: +[Valid][10][UTT][ MSE = 0.1939 | LCC = 0.9563 | SRCC = 0.9572 ] +INFO:root:Average loss=0.07215703326463699 +INFO:root: +[Valid][11][UTT][ MSE = 0.1993 | LCC = 0.9583 | SRCC = 0.9591 ] +INFO:root:Average loss=0.06778483417630196 +INFO:root: +[Valid][12][UTT][ MSE = 0.1830 | LCC = 0.9567 | SRCC = 0.9577 ] +INFO:root:Average loss=0.06788343521952629 +INFO:root: +[Valid][13][UTT][ MSE = 0.1996 | LCC = 0.9575 | SRCC = 0.9586 ] +INFO:root:Average loss=0.06273102441430092 +INFO:root: +[Valid][14][UTT][ MSE = 0.2106 | LCC = 0.9582 | SRCC = 0.9592 ] +INFO:root:Average loss=0.06351986202597618 +INFO:root: +[Valid][15][UTT][ MSE = 0.2248 | LCC = 0.9591 | SRCC = 0.9600 ] +INFO:root:Average loss=0.056358633995056154 +INFO:root: +[Valid][16][UTT][ MSE = 0.2004 | LCC = 0.9576 | SRCC = 0.9591 ] +INFO:root:Average loss=0.055550382822752 +INFO:root: +[Valid][17][UTT][ MSE = 0.2256 | LCC = 0.9581 | SRCC = 0.9590 ] +INFO:root:Average loss=0.05426924280822277 +INFO:root: +[Valid][18][UTT][ MSE = 0.2253 | LCC = 0.9591 | SRCC = 0.9599 ] +INFO:root:Average loss=0.05265174898505211 +INFO:root: +[Valid][19][UTT][ MSE = 0.1766 | LCC = 0.9591 | SRCC = 0.9594 ] +INFO:root:Average loss=0.04848474657535553 +INFO:root: +[Valid][20][UTT][ MSE = 0.2147 | LCC = 0.9589 | SRCC = 0.9593 ] +INFO:root:Average loss=0.0477328524440527 +INFO:root: +[Valid][21][UTT][ MSE = 0.1865 | LCC = 0.9584 | SRCC = 0.9585 ] +INFO:root:Average loss=0.04642121535539627 +INFO:root: +[Valid][22][UTT][ MSE = 0.2056 | LCC = 0.9589 | SRCC = 0.9594 ] +INFO:root:Average loss=0.04394263881444931 +INFO:root: +[Valid][23][UTT][ MSE = 0.2107 | LCC = 0.9587 | SRCC = 0.9592 ] +INFO:root:Average loss=0.04300291639566421 +INFO:root: +[Valid][24][UTT][ MSE = 0.2146 | LCC = 0.9588 | SRCC = 0.9594 ] +INFO:root:Average loss=0.04033444191515446 +INFO:root: +[Valid][25][UTT][ MSE = 0.1994 | LCC = 0.9585 | SRCC = 0.9587 ] +INFO:root:Average loss=0.04066796438395977 +INFO:root: +[Valid][26][UTT][ MSE = 0.1940 | LCC = 0.9606 | SRCC = 0.9607 ] +INFO:root:Average loss=0.03731056345999241 +INFO:root: +[Valid][27][UTT][ MSE = 0.2106 | LCC = 0.9595 | SRCC = 0.9601 ] +INFO:root:Average loss=0.03662172901630402 +INFO:root: +[Valid][28][UTT][ MSE = 0.2024 | LCC = 0.9598 | SRCC = 0.9600 ] +INFO:root:Average loss=0.036063070580363274 +INFO:root: +[Valid][29][UTT][ MSE = 0.2073 | LCC = 0.9592 | SRCC = 0.9596 ] +INFO:root:Average loss=0.03453338925540447 +INFO:root: +[Valid][30][UTT][ MSE = 0.2005 | LCC = 0.9597 | SRCC = 0.9598 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1937 | LCC = 0.9629 | SRCC = 0.9634 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a5de39038290f5cb171e0ec0952898cd258b4fd2 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6c525d9325339c949acfc22fa1404ffdeda11ef920aa45e64b7238080cd2a41 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d37d35affb0e73960d540e99e583f945e79e1db6 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:da02b4619d6127311d046e63dd1bae2998f023ffa20cc66c9ba66ee6eda44f1a +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d0a741807ff166708c8334862dc1c51be3ac4dba --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.41366678941249846 +INFO:root: +[Valid][0][UTT][ MSE = 0.2844 | LCC = 0.9192 | SRCC = 0.9223 ] +INFO:root:Average loss=0.19561233353614807 +INFO:root: +[Valid][1][UTT][ MSE = 0.2751 | LCC = 0.9401 | SRCC = 0.9417 ] +INFO:root:Average loss=0.1588566505908966 +INFO:root: +[Valid][2][UTT][ MSE = 0.2437 | LCC = 0.9450 | SRCC = 0.9453 ] +INFO:root:Average loss=0.13683633756637573 +INFO:root: +[Valid][3][UTT][ MSE = 0.2121 | LCC = 0.9517 | SRCC = 0.9522 ] +INFO:root:Average loss=0.12254817968606949 +INFO:root: +[Valid][4][UTT][ MSE = 0.2319 | LCC = 0.9532 | SRCC = 0.9541 ] +INFO:root:Average loss=0.11772914338111877 +INFO:root: +[Valid][5][UTT][ MSE = 0.2891 | LCC = 0.9497 | SRCC = 0.9519 ] +INFO:root:Average loss=0.11496923047304153 +INFO:root: +[Valid][6][UTT][ MSE = 0.2031 | LCC = 0.9539 | SRCC = 0.9541 ] +INFO:root:Average loss=0.0977235116660595 +INFO:root: +[Valid][7][UTT][ MSE = 0.2300 | LCC = 0.9569 | SRCC = 0.9573 ] +INFO:root:Average loss=0.09060642817616463 +INFO:root: +[Valid][8][UTT][ MSE = 0.1945 | LCC = 0.9553 | SRCC = 0.9555 ] +INFO:root:Average loss=0.08360666427016258 +INFO:root: +[Valid][9][UTT][ MSE = 0.2182 | LCC = 0.9565 | SRCC = 0.9575 ] +INFO:root:Average loss=0.07746437558531762 +INFO:root: +[Valid][10][UTT][ MSE = 0.1893 | LCC = 0.9572 | SRCC = 0.9579 ] +INFO:root:Average loss=0.07443345302343368 +INFO:root: +[Valid][11][UTT][ MSE = 0.2088 | LCC = 0.9581 | SRCC = 0.9585 ] +INFO:root:Average loss=0.06899174571037292 +INFO:root: +[Valid][12][UTT][ MSE = 0.2176 | LCC = 0.9581 | SRCC = 0.9583 ] +INFO:root:Average loss=0.0694892684519291 +INFO:root: +[Valid][13][UTT][ MSE = 0.1932 | LCC = 0.9575 | SRCC = 0.9584 ] +INFO:root:Average loss=0.06445646664500236 +INFO:root: +[Valid][14][UTT][ MSE = 0.2227 | LCC = 0.9581 | SRCC = 0.9587 ] +INFO:root:Average loss=0.06099912759661674 +INFO:root: +[Valid][15][UTT][ MSE = 0.2103 | LCC = 0.9568 | SRCC = 0.9574 ] +INFO:root:Average loss=0.05981821852922439 +INFO:root: +[Valid][16][UTT][ MSE = 0.2006 | LCC = 0.9595 | SRCC = 0.9594 ] +INFO:root:Average loss=0.060251038551330564 +INFO:root: +[Valid][17][UTT][ MSE = 0.2002 | LCC = 0.9579 | SRCC = 0.9588 ] +INFO:root:Average loss=0.05458149260282517 +INFO:root: +[Valid][18][UTT][ MSE = 0.2157 | LCC = 0.9584 | SRCC = 0.9589 ] +INFO:root:Average loss=0.05358489659428597 +INFO:root: +[Valid][19][UTT][ MSE = 0.2077 | LCC = 0.9588 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05024181950092316 +INFO:root: +[Valid][20][UTT][ MSE = 0.2004 | LCC = 0.9581 | SRCC = 0.9583 ] +INFO:root:Average loss=0.04724397857487202 +INFO:root: +[Valid][21][UTT][ MSE = 0.1835 | LCC = 0.9580 | SRCC = 0.9582 ] +INFO:root:Average loss=0.04645976796746254 +INFO:root: +[Valid][22][UTT][ MSE = 0.1903 | LCC = 0.9601 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04689102141559124 +INFO:root: +[Valid][23][UTT][ MSE = 0.1846 | LCC = 0.9582 | SRCC = 0.9587 ] +INFO:root:Average loss=0.04220510110259056 +INFO:root: +[Valid][24][UTT][ MSE = 0.1801 | LCC = 0.9595 | SRCC = 0.9594 ] +INFO:root:Average loss=0.039575854912400246 +INFO:root: +[Valid][25][UTT][ MSE = 0.2105 | LCC = 0.9586 | SRCC = 0.9592 ] +INFO:root:Average loss=0.03882134936749935 +INFO:root: +[Valid][26][UTT][ MSE = 0.2088 | LCC = 0.9591 | SRCC = 0.9593 ] +INFO:root:Average loss=0.03819887863099575 +INFO:root: +[Valid][27][UTT][ MSE = 0.1946 | LCC = 0.9583 | SRCC = 0.9586 ] +INFO:root:Average loss=0.03723120164871216 +INFO:root: +[Valid][28][UTT][ MSE = 0.1682 | LCC = 0.9584 | SRCC = 0.9591 ] +INFO:root:Average loss=0.03749848793447018 +INFO:root: +[Valid][29][UTT][ MSE = 0.1919 | LCC = 0.9589 | SRCC = 0.9589 ] +INFO:root:Average loss=0.03649663718044758 +INFO:root: +[Valid][30][UTT][ MSE = 0.1795 | LCC = 0.9590 | SRCC = 0.9590 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1844 | LCC = 0.9630 | SRCC = 0.9633 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c7c2a8ed14898579b14fe4a24c746a93bff0a656 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:09464bca08d6fc760131b28ef246378b96b50c778f5ef0a1d97c136c7d25dcb9 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..bdf3b3059e3d43468e91883321ee868c0957ba50 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0b47c6b01e1b58d32184acdfa007219293c0a98818e9887e89c649c0924cf85 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..8ef6a2857fb7f0ce4fa7164c33f67b1ce3c5e252 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer2/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42170892310142516 +INFO:root: +[Valid][0][UTT][ MSE = 0.2963 | LCC = 0.9235 | SRCC = 0.9243 ] +INFO:root:Average loss=0.19843198621273042 +INFO:root: +[Valid][1][UTT][ MSE = 0.2827 | LCC = 0.9371 | SRCC = 0.9390 ] +INFO:root:Average loss=0.15768149876594542 +INFO:root: +[Valid][2][UTT][ MSE = 0.2821 | LCC = 0.9416 | SRCC = 0.9428 ] +INFO:root:Average loss=0.1365991051197052 +INFO:root: +[Valid][3][UTT][ MSE = 0.2470 | LCC = 0.9440 | SRCC = 0.9462 ] +INFO:root:Average loss=0.12262591224908828 +INFO:root: +[Valid][4][UTT][ MSE = 0.2156 | LCC = 0.9525 | SRCC = 0.9535 ] +INFO:root:Average loss=0.11029033470153808 +INFO:root: +[Valid][5][UTT][ MSE = 0.2257 | LCC = 0.9520 | SRCC = 0.9531 ] +INFO:root:Average loss=0.10283486890792846 +INFO:root: +[Valid][6][UTT][ MSE = 0.1908 | LCC = 0.9537 | SRCC = 0.9548 ] +INFO:root:Average loss=0.09501331162452697 +INFO:root: +[Valid][7][UTT][ MSE = 0.2152 | LCC = 0.9546 | SRCC = 0.9555 ] +INFO:root:Average loss=0.0904442910850048 +INFO:root: +[Valid][8][UTT][ MSE = 0.2381 | LCC = 0.9559 | SRCC = 0.9569 ] +INFO:root:Average loss=0.08610500150918961 +INFO:root: +[Valid][9][UTT][ MSE = 0.1805 | LCC = 0.9572 | SRCC = 0.9577 ] +INFO:root:Average loss=0.07697503244876862 +INFO:root: +[Valid][10][UTT][ MSE = 0.2332 | LCC = 0.9551 | SRCC = 0.9553 ] +INFO:root:Average loss=0.07571949037909508 +INFO:root: +[Valid][11][UTT][ MSE = 0.1848 | LCC = 0.9574 | SRCC = 0.9584 ] +INFO:root:Average loss=0.0724712927043438 +INFO:root: +[Valid][12][UTT][ MSE = 0.1927 | LCC = 0.9556 | SRCC = 0.9559 ] +INFO:root:Average loss=0.0714154523909092 +INFO:root: +[Valid][13][UTT][ MSE = 0.1829 | LCC = 0.9580 | SRCC = 0.9589 ] +INFO:root:Average loss=0.06547898790240288 +INFO:root: +[Valid][14][UTT][ MSE = 0.2139 | LCC = 0.9565 | SRCC = 0.9568 ] +INFO:root:Average loss=0.06457890301942826 +INFO:root: +[Valid][15][UTT][ MSE = 0.2064 | LCC = 0.9567 | SRCC = 0.9578 ] +INFO:root:Average loss=0.0593600692152977 +INFO:root: +[Valid][16][UTT][ MSE = 0.2073 | LCC = 0.9588 | SRCC = 0.9600 ] +INFO:root:Average loss=0.05411391481757164 +INFO:root: +[Valid][17][UTT][ MSE = 0.1998 | LCC = 0.9577 | SRCC = 0.9579 ] +INFO:root:Average loss=0.055621402025222776 +INFO:root: +[Valid][18][UTT][ MSE = 0.2260 | LCC = 0.9583 | SRCC = 0.9584 ] +INFO:root:Average loss=0.05208730928599834 +INFO:root: +[Valid][19][UTT][ MSE = 0.1932 | LCC = 0.9562 | SRCC = 0.9563 ] +INFO:root:Average loss=0.04931080640852451 +INFO:root: +[Valid][20][UTT][ MSE = 0.1964 | LCC = 0.9581 | SRCC = 0.9584 ] +INFO:root:Average loss=0.04590767423808575 +INFO:root: +[Valid][21][UTT][ MSE = 0.2030 | LCC = 0.9596 | SRCC = 0.9600 ] +INFO:root:Average loss=0.04908711029589176 +INFO:root: +[Valid][22][UTT][ MSE = 0.1831 | LCC = 0.9588 | SRCC = 0.9587 ] +INFO:root:Average loss=0.04513048368692398 +INFO:root: +[Valid][23][UTT][ MSE = 0.1941 | LCC = 0.9584 | SRCC = 0.9583 ] +INFO:root:Average loss=0.046488383501768114 +INFO:root: +[Valid][24][UTT][ MSE = 0.1955 | LCC = 0.9568 | SRCC = 0.9572 ] +INFO:root:Average loss=0.043727996706962584 +INFO:root: +[Valid][25][UTT][ MSE = 0.2099 | LCC = 0.9575 | SRCC = 0.9576 ] +INFO:root:Average loss=0.04244305971264839 +INFO:root: +[Valid][26][UTT][ MSE = 0.1950 | LCC = 0.9587 | SRCC = 0.9589 ] +INFO:root:Average loss=0.03753298135101795 +INFO:root: +[Valid][27][UTT][ MSE = 0.1857 | LCC = 0.9591 | SRCC = 0.9589 ] +INFO:root:Average loss=0.03837004165351391 +INFO:root: +[Valid][28][UTT][ MSE = 0.1952 | LCC = 0.9583 | SRCC = 0.9581 ] +INFO:root:Average loss=0.03761759276688099 +INFO:root: +[Valid][29][UTT][ MSE = 0.2015 | LCC = 0.9602 | SRCC = 0.9605 ] +INFO:root:Average loss=0.03537790620326996 +INFO:root: +[Valid][30][UTT][ MSE = 0.1960 | LCC = 0.9604 | SRCC = 0.9604 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1954 | LCC = 0.9606 | SRCC = 0.9613 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..39058a7f5b54680cb8d89c4681903138e24e98c1 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1bde1783743253d2d829d9106892c960f74eaf864dbe03923c0ec10049d32431 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9d09427ad999d727aff44853f1903f3923ab5956 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4b9498a3ca117fb119e68156d99113ae2ff9e8c9e33e9daa217095c3e371393 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c7ace12e05c4b1574d73280c171891c011c38cef --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4567294031381607 +INFO:root: +[Valid][0][UTT][ MSE = 0.3395 | LCC = 0.9186 | SRCC = 0.9226 ] +INFO:root:Average loss=0.18890932148694992 +INFO:root: +[Valid][1][UTT][ MSE = 0.2424 | LCC = 0.9417 | SRCC = 0.9429 ] +INFO:root:Average loss=0.15482299733161925 +INFO:root: +[Valid][2][UTT][ MSE = 0.2454 | LCC = 0.9486 | SRCC = 0.9508 ] +INFO:root:Average loss=0.13257532852888107 +INFO:root: +[Valid][3][UTT][ MSE = 0.2212 | LCC = 0.9521 | SRCC = 0.9530 ] +INFO:root:Average loss=0.1172676129937172 +INFO:root: +[Valid][4][UTT][ MSE = 0.2273 | LCC = 0.9550 | SRCC = 0.9561 ] +INFO:root:Average loss=0.1104501966536045 +INFO:root: +[Valid][5][UTT][ MSE = 0.1998 | LCC = 0.9537 | SRCC = 0.9543 ] +INFO:root:Average loss=0.10344121924042701 +INFO:root: +[Valid][6][UTT][ MSE = 0.2096 | LCC = 0.9570 | SRCC = 0.9575 ] +INFO:root:Average loss=0.09207313016057014 +INFO:root: +[Valid][7][UTT][ MSE = 0.1972 | LCC = 0.9543 | SRCC = 0.9546 ] +INFO:root:Average loss=0.08660174569487572 +INFO:root: +[Valid][8][UTT][ MSE = 0.1671 | LCC = 0.9542 | SRCC = 0.9540 ] +INFO:root:Average loss=0.08061042904853821 +INFO:root: +[Valid][9][UTT][ MSE = 0.1884 | LCC = 0.9565 | SRCC = 0.9572 ] +INFO:root:Average loss=0.07655969521403312 +INFO:root: +[Valid][10][UTT][ MSE = 0.1926 | LCC = 0.9564 | SRCC = 0.9564 ] +INFO:root:Average loss=0.07076771837472916 +INFO:root: +[Valid][11][UTT][ MSE = 0.2071 | LCC = 0.9591 | SRCC = 0.9601 ] +INFO:root:Average loss=0.07024510109424591 +INFO:root: +[Valid][12][UTT][ MSE = 0.1824 | LCC = 0.9568 | SRCC = 0.9575 ] +INFO:root:Average loss=0.06612272569537163 +INFO:root: +[Valid][13][UTT][ MSE = 0.1774 | LCC = 0.9577 | SRCC = 0.9580 ] +INFO:root:Average loss=0.06249737536907196 +INFO:root: +[Valid][14][UTT][ MSE = 0.1885 | LCC = 0.9595 | SRCC = 0.9596 ] +INFO:root:Average loss=0.05970617899298668 +INFO:root: +[Valid][15][UTT][ MSE = 0.1865 | LCC = 0.9576 | SRCC = 0.9579 ] +INFO:root:Average loss=0.05517768579721451 +INFO:root: +[Valid][16][UTT][ MSE = 0.1843 | LCC = 0.9592 | SRCC = 0.9595 ] +INFO:root:Average loss=0.05150590486824513 +INFO:root: +[Valid][17][UTT][ MSE = 0.1775 | LCC = 0.9574 | SRCC = 0.9576 ] +INFO:root:Average loss=0.04705406846106052 +INFO:root: +[Valid][18][UTT][ MSE = 0.1944 | LCC = 0.9567 | SRCC = 0.9568 ] +INFO:root:Average loss=0.04740243792533874 +INFO:root: +[Valid][19][UTT][ MSE = 0.2133 | LCC = 0.9595 | SRCC = 0.9603 ] +INFO:root:Average loss=0.04741250276565552 +INFO:root: +[Valid][20][UTT][ MSE = 0.1918 | LCC = 0.9600 | SRCC = 0.9604 ] +INFO:root:Average loss=0.04429354657232761 +INFO:root: +[Valid][21][UTT][ MSE = 0.1751 | LCC = 0.9587 | SRCC = 0.9592 ] +INFO:root:Average loss=0.04300968590378761 +INFO:root: +[Valid][22][UTT][ MSE = 0.1876 | LCC = 0.9601 | SRCC = 0.9605 ] +INFO:root:Average loss=0.040639152392745015 +INFO:root: +[Valid][23][UTT][ MSE = 0.1819 | LCC = 0.9600 | SRCC = 0.9601 ] +INFO:root:Average loss=0.039260679915547374 +INFO:root: +[Valid][24][UTT][ MSE = 0.1913 | LCC = 0.9590 | SRCC = 0.9588 ] +INFO:root:Average loss=0.03741108752787113 +INFO:root: +[Valid][25][UTT][ MSE = 0.1798 | LCC = 0.9586 | SRCC = 0.9593 ] +INFO:root:Average loss=0.03775206217169762 +INFO:root: +[Valid][26][UTT][ MSE = 0.1772 | LCC = 0.9600 | SRCC = 0.9602 ] +INFO:root:Average loss=0.033640537291765214 +INFO:root: +[Valid][27][UTT][ MSE = 0.1779 | LCC = 0.9580 | SRCC = 0.9584 ] +INFO:root:Average loss=0.03237975639104843 +INFO:root: +[Valid][28][UTT][ MSE = 0.1904 | LCC = 0.9589 | SRCC = 0.9594 ] +INFO:root:Average loss=0.0339259278178215 +INFO:root: +[Valid][29][UTT][ MSE = 0.1823 | LCC = 0.9593 | SRCC = 0.9596 ] +INFO:root:Average loss=0.03235561287403107 +INFO:root: +[Valid][30][UTT][ MSE = 0.1883 | LCC = 0.9583 | SRCC = 0.9584 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1941 | LCC = 0.9605 | SRCC = 0.9616 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b83714875318929fe5bbb1ac56a9506a385c935d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e5e815f940b747513c5dc41f15f66f65fccde27ea6600b5c24a08e4c359cac4 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6f76c0ab6150a2963ff03734d879fb4ab92a3bca --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e448bdea05c6af9d0f32fb28afb3219a825304cb4c8ad641e1c8fc74ad424222 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5d342c94c04af57381410dcf618a3f54ebb10cd2 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42165069830417634 +INFO:root: +[Valid][0][UTT][ MSE = 0.2903 | LCC = 0.9281 | SRCC = 0.9293 ] +INFO:root:Average loss=0.18713951462507247 +INFO:root: +[Valid][1][UTT][ MSE = 0.2510 | LCC = 0.9420 | SRCC = 0.9437 ] +INFO:root:Average loss=0.14878211653232573 +INFO:root: +[Valid][2][UTT][ MSE = 0.2027 | LCC = 0.9507 | SRCC = 0.9524 ] +INFO:root:Average loss=0.13291565579175948 +INFO:root: +[Valid][3][UTT][ MSE = 0.2330 | LCC = 0.9521 | SRCC = 0.9529 ] +INFO:root:Average loss=0.11972375673055649 +INFO:root: +[Valid][4][UTT][ MSE = 0.2038 | LCC = 0.9556 | SRCC = 0.9561 ] +INFO:root:Average loss=0.11093790709972381 +INFO:root: +[Valid][5][UTT][ MSE = 0.2216 | LCC = 0.9577 | SRCC = 0.9580 ] +INFO:root:Average loss=0.10145853838324546 +INFO:root: +[Valid][6][UTT][ MSE = 0.1706 | LCC = 0.9583 | SRCC = 0.9588 ] +INFO:root:Average loss=0.09913208493590354 +INFO:root: +[Valid][7][UTT][ MSE = 0.1730 | LCC = 0.9583 | SRCC = 0.9591 ] +INFO:root:Average loss=0.08803795024752617 +INFO:root: +[Valid][8][UTT][ MSE = 0.2010 | LCC = 0.9593 | SRCC = 0.9595 ] +INFO:root:Average loss=0.08523006823658943 +INFO:root: +[Valid][9][UTT][ MSE = 0.1937 | LCC = 0.9602 | SRCC = 0.9603 ] +INFO:root:Average loss=0.08161310106515884 +INFO:root: +[Valid][10][UTT][ MSE = 0.1734 | LCC = 0.9595 | SRCC = 0.9599 ] +INFO:root:Average loss=0.07286287543177605 +INFO:root: +[Valid][11][UTT][ MSE = 0.1890 | LCC = 0.9606 | SRCC = 0.9605 ] +INFO:root:Average loss=0.07127228826284408 +INFO:root: +[Valid][12][UTT][ MSE = 0.2308 | LCC = 0.9556 | SRCC = 0.9557 ] +INFO:root:Average loss=0.07035966935753822 +INFO:root: +[Valid][13][UTT][ MSE = 0.2022 | LCC = 0.9611 | SRCC = 0.9613 ] +INFO:root:Average loss=0.06207516705989838 +INFO:root: +[Valid][14][UTT][ MSE = 0.1782 | LCC = 0.9614 | SRCC = 0.9615 ] +INFO:root:Average loss=0.06238397970795631 +INFO:root: +[Valid][15][UTT][ MSE = 0.1730 | LCC = 0.9612 | SRCC = 0.9612 ] +INFO:root:Average loss=0.05655361571907997 +INFO:root: +[Valid][16][UTT][ MSE = 0.1853 | LCC = 0.9618 | SRCC = 0.9616 ] +INFO:root:Average loss=0.05623822724819184 +INFO:root: +[Valid][17][UTT][ MSE = 0.1867 | LCC = 0.9587 | SRCC = 0.9590 ] +INFO:root:Average loss=0.053727394729852676 +INFO:root: +[Valid][18][UTT][ MSE = 0.1744 | LCC = 0.9606 | SRCC = 0.9603 ] +INFO:root:Average loss=0.05011829248070717 +INFO:root: +[Valid][19][UTT][ MSE = 0.1850 | LCC = 0.9591 | SRCC = 0.9593 ] +INFO:root:Average loss=0.05148288983106613 +INFO:root: +[Valid][20][UTT][ MSE = 0.2049 | LCC = 0.9610 | SRCC = 0.9608 ] +INFO:root:Average loss=0.04906926855444908 +INFO:root: +[Valid][21][UTT][ MSE = 0.1979 | LCC = 0.9604 | SRCC = 0.9602 ] +INFO:root:Average loss=0.04482899557054043 +INFO:root: +[Valid][22][UTT][ MSE = 0.1982 | LCC = 0.9601 | SRCC = 0.9604 ] +INFO:root:Average loss=0.04373997528851032 +INFO:root: +[Valid][23][UTT][ MSE = 0.1919 | LCC = 0.9607 | SRCC = 0.9608 ] +INFO:root:Average loss=0.04250522962212563 +INFO:root: +[Valid][24][UTT][ MSE = 0.2170 | LCC = 0.9592 | SRCC = 0.9595 ] +INFO:root:Average loss=0.04095675568282604 +INFO:root: +[Valid][25][UTT][ MSE = 0.1862 | LCC = 0.9608 | SRCC = 0.9616 ] +INFO:root:Average loss=0.03868122889101505 +INFO:root: +[Valid][26][UTT][ MSE = 0.1842 | LCC = 0.9602 | SRCC = 0.9603 ] +INFO:root:Average loss=0.03780360403656959 +INFO:root: +[Valid][27][UTT][ MSE = 0.2029 | LCC = 0.9577 | SRCC = 0.9578 ] +INFO:root:Average loss=0.037369033992290496 +INFO:root: +[Valid][28][UTT][ MSE = 0.1851 | LCC = 0.9613 | SRCC = 0.9617 ] +INFO:root:Average loss=0.03526256054639816 +INFO:root: +[Valid][29][UTT][ MSE = 0.1856 | LCC = 0.9606 | SRCC = 0.9604 ] +INFO:root:Average loss=0.0344747554063797 +INFO:root: +[Valid][30][UTT][ MSE = 0.1746 | LCC = 0.9606 | SRCC = 0.9604 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1921 | LCC = 0.9593 | SRCC = 0.9602 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..05f8b5b222e264b742e821ba70f7c96009fd7eb5 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9cb09efc9e38fc5f37b5c990d230a53407367b99b74fb5739f3cd76ba4b4c614 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..3fd17e96e57c04f9f5c3935c18236197950db394 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cf2a74bb66a6a24ad721a263041dae2d01352c62ebfa11b0d6a09e731c19be00 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e50b316ac71834cd9fb1437e17d72300df3695d8 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42793388760089873 +INFO:root: +[Valid][0][UTT][ MSE = 0.2993 | LCC = 0.9297 | SRCC = 0.9305 ] +INFO:root:Average loss=0.18139684474468232 +INFO:root: +[Valid][1][UTT][ MSE = 0.2689 | LCC = 0.9432 | SRCC = 0.9460 ] +INFO:root:Average loss=0.1494204699397087 +INFO:root: +[Valid][2][UTT][ MSE = 0.2261 | LCC = 0.9501 | SRCC = 0.9513 ] +INFO:root:Average loss=0.1264487292766571 +INFO:root: +[Valid][3][UTT][ MSE = 0.2060 | LCC = 0.9512 | SRCC = 0.9519 ] +INFO:root:Average loss=0.1158354377746582 +INFO:root: +[Valid][4][UTT][ MSE = 0.2598 | LCC = 0.9512 | SRCC = 0.9534 ] +INFO:root:Average loss=0.10586250916123391 +INFO:root: +[Valid][5][UTT][ MSE = 0.2093 | LCC = 0.9549 | SRCC = 0.9563 ] +INFO:root:Average loss=0.09725140681862832 +INFO:root: +[Valid][6][UTT][ MSE = 0.2469 | LCC = 0.9534 | SRCC = 0.9575 ] +INFO:root:Average loss=0.090783736795187 +INFO:root: +[Valid][7][UTT][ MSE = 0.1939 | LCC = 0.9578 | SRCC = 0.9588 ] +INFO:root:Average loss=0.08535895875096321 +INFO:root: +[Valid][8][UTT][ MSE = 0.1951 | LCC = 0.9573 | SRCC = 0.9585 ] +INFO:root:Average loss=0.08261223623156548 +INFO:root: +[Valid][9][UTT][ MSE = 0.2233 | LCC = 0.9561 | SRCC = 0.9584 ] +INFO:root:Average loss=0.07457572776079179 +INFO:root: +[Valid][10][UTT][ MSE = 0.2184 | LCC = 0.9578 | SRCC = 0.9594 ] +INFO:root:Average loss=0.07011301410198212 +INFO:root: +[Valid][11][UTT][ MSE = 0.2378 | LCC = 0.9567 | SRCC = 0.9597 ] +INFO:root:Average loss=0.06476230278611184 +INFO:root: +[Valid][12][UTT][ MSE = 0.2396 | LCC = 0.9575 | SRCC = 0.9603 ] +INFO:root:Average loss=0.0660365988612175 +INFO:root: +[Valid][13][UTT][ MSE = 0.2086 | LCC = 0.9583 | SRCC = 0.9597 ] +INFO:root:Average loss=0.06020735067129135 +INFO:root: +[Valid][14][UTT][ MSE = 0.2136 | LCC = 0.9572 | SRCC = 0.9582 ] +INFO:root:Average loss=0.0554944072663784 +INFO:root: +[Valid][15][UTT][ MSE = 0.1948 | LCC = 0.9579 | SRCC = 0.9587 ] +INFO:root:Average loss=0.05353314882516861 +INFO:root: +[Valid][16][UTT][ MSE = 0.1971 | LCC = 0.9595 | SRCC = 0.9603 ] +INFO:root:Average loss=0.052266601160168645 +INFO:root: +[Valid][17][UTT][ MSE = 0.1996 | LCC = 0.9592 | SRCC = 0.9598 ] +INFO:root:Average loss=0.04753543566167354 +INFO:root: +[Valid][18][UTT][ MSE = 0.2230 | LCC = 0.9565 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04834233483672142 +INFO:root: +[Valid][19][UTT][ MSE = 0.1991 | LCC = 0.9592 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04403045231103897 +INFO:root: +[Valid][20][UTT][ MSE = 0.2096 | LCC = 0.9574 | SRCC = 0.9586 ] +INFO:root:Average loss=0.047193175315856935 +INFO:root: +[Valid][21][UTT][ MSE = 0.2013 | LCC = 0.9589 | SRCC = 0.9595 ] +INFO:root:Average loss=0.041832013100385666 +INFO:root: +[Valid][22][UTT][ MSE = 0.2013 | LCC = 0.9590 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04350387440621853 +INFO:root: +[Valid][23][UTT][ MSE = 0.1878 | LCC = 0.9587 | SRCC = 0.9594 ] +INFO:root:Average loss=0.03891048803925514 +INFO:root: +[Valid][24][UTT][ MSE = 0.2207 | LCC = 0.9591 | SRCC = 0.9606 ] +INFO:root:Average loss=0.035988285169005395 +INFO:root: +[Valid][25][UTT][ MSE = 0.2042 | LCC = 0.9587 | SRCC = 0.9596 ] +INFO:root:Average loss=0.035604889452457426 +INFO:root: +[Valid][26][UTT][ MSE = 0.2033 | LCC = 0.9592 | SRCC = 0.9602 ] +INFO:root:Average loss=0.03612201915681362 +INFO:root: +[Valid][27][UTT][ MSE = 0.1922 | LCC = 0.9602 | SRCC = 0.9603 ] +INFO:root:Average loss=0.03471210265159607 +INFO:root: +[Valid][28][UTT][ MSE = 0.1995 | LCC = 0.9589 | SRCC = 0.9599 ] +INFO:root:Average loss=0.032526455074548724 +INFO:root: +[Valid][29][UTT][ MSE = 0.2115 | LCC = 0.9602 | SRCC = 0.9608 ] +INFO:root:Average loss=0.0312832103818655 +INFO:root: +[Valid][30][UTT][ MSE = 0.2071 | LCC = 0.9585 | SRCC = 0.9594 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1977 | LCC = 0.9602 | SRCC = 0.9615 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5f136f1637db54eb4a6fdb73b968546c77b3eb10 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:756ceea9b7161fe585e598b89b801d7020c8f97b8d7df3d7e8abb22ce20597dd +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..053e4a4fdb5aa2fcf9cf408c1f550b4ff112627c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:17fbd9fe6e86f80fc30de8cbb0adeadefa7e80a03e43ff87598db750b091ec63 +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0f8827465ddcc4831a6856ec1088ccf2b3170144 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42605183088779447 +INFO:root: +[Valid][0][UTT][ MSE = 0.2835 | LCC = 0.9259 | SRCC = 0.9290 ] +INFO:root:Average loss=0.18787039011716844 +INFO:root: +[Valid][1][UTT][ MSE = 0.2204 | LCC = 0.9427 | SRCC = 0.9440 ] +INFO:root:Average loss=0.1560110582113266 +INFO:root: +[Valid][2][UTT][ MSE = 0.2444 | LCC = 0.9471 | SRCC = 0.9489 ] +INFO:root:Average loss=0.1343741362094879 +INFO:root: +[Valid][3][UTT][ MSE = 0.2094 | LCC = 0.9520 | SRCC = 0.9532 ] +INFO:root:Average loss=0.12348667800426483 +INFO:root: +[Valid][4][UTT][ MSE = 0.2364 | LCC = 0.9537 | SRCC = 0.9550 ] +INFO:root:Average loss=0.10810300546884537 +INFO:root: +[Valid][5][UTT][ MSE = 0.2307 | LCC = 0.9546 | SRCC = 0.9553 ] +INFO:root:Average loss=0.10097509890794754 +INFO:root: +[Valid][6][UTT][ MSE = 0.1882 | LCC = 0.9557 | SRCC = 0.9568 ] +INFO:root:Average loss=0.0950009543299675 +INFO:root: +[Valid][7][UTT][ MSE = 0.2008 | LCC = 0.9564 | SRCC = 0.9571 ] +INFO:root:Average loss=0.08807534757256508 +INFO:root: +[Valid][8][UTT][ MSE = 0.1973 | LCC = 0.9577 | SRCC = 0.9581 ] +INFO:root:Average loss=0.08210239100456238 +INFO:root: +[Valid][9][UTT][ MSE = 0.1901 | LCC = 0.9561 | SRCC = 0.9562 ] +INFO:root:Average loss=0.0804307973086834 +INFO:root: +[Valid][10][UTT][ MSE = 0.2101 | LCC = 0.9571 | SRCC = 0.9583 ] +INFO:root:Average loss=0.07420174366235734 +INFO:root: +[Valid][11][UTT][ MSE = 0.1914 | LCC = 0.9535 | SRCC = 0.9534 ] +INFO:root:Average loss=0.06736176419258118 +INFO:root: +[Valid][12][UTT][ MSE = 0.1665 | LCC = 0.9578 | SRCC = 0.9590 ] +INFO:root:Average loss=0.06517408722639084 +INFO:root: +[Valid][13][UTT][ MSE = 0.1803 | LCC = 0.9576 | SRCC = 0.9581 ] +INFO:root:Average loss=0.06543282820284367 +INFO:root: +[Valid][14][UTT][ MSE = 0.1994 | LCC = 0.9596 | SRCC = 0.9602 ] +INFO:root:Average loss=0.05989030769467354 +INFO:root: +[Valid][15][UTT][ MSE = 0.1934 | LCC = 0.9583 | SRCC = 0.9594 ] +INFO:root:Average loss=0.05661010912060738 +INFO:root: +[Valid][16][UTT][ MSE = 0.1744 | LCC = 0.9577 | SRCC = 0.9579 ] +INFO:root:Average loss=0.053971335887908936 +INFO:root: +[Valid][17][UTT][ MSE = 0.1860 | LCC = 0.9595 | SRCC = 0.9603 ] +INFO:root:Average loss=0.0512073715031147 +INFO:root: +[Valid][18][UTT][ MSE = 0.1935 | LCC = 0.9595 | SRCC = 0.9597 ] +INFO:root:Average loss=0.049134950786828994 +INFO:root: +[Valid][19][UTT][ MSE = 0.1850 | LCC = 0.9601 | SRCC = 0.9610 ] +INFO:root:Average loss=0.04908493752777576 +INFO:root: +[Valid][20][UTT][ MSE = 0.1996 | LCC = 0.9588 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04656567294895649 +INFO:root: +[Valid][21][UTT][ MSE = 0.1886 | LCC = 0.9600 | SRCC = 0.9605 ] +INFO:root:Average loss=0.04336367526650429 +INFO:root: +[Valid][22][UTT][ MSE = 0.2025 | LCC = 0.9586 | SRCC = 0.9585 ] +INFO:root:Average loss=0.04221031524240971 +INFO:root: +[Valid][23][UTT][ MSE = 0.1759 | LCC = 0.9601 | SRCC = 0.9603 ] +INFO:root:Average loss=0.04248048865795136 +INFO:root: +[Valid][24][UTT][ MSE = 0.2065 | LCC = 0.9598 | SRCC = 0.9598 ] +INFO:root:Average loss=0.04210212701559067 +INFO:root: +[Valid][25][UTT][ MSE = 0.1793 | LCC = 0.9598 | SRCC = 0.9602 ] +INFO:root:Average loss=0.03719288502633571 +INFO:root: +[Valid][26][UTT][ MSE = 0.1716 | LCC = 0.9594 | SRCC = 0.9601 ] +INFO:root:Average loss=0.03764729607105255 +INFO:root: +[Valid][27][UTT][ MSE = 0.2076 | LCC = 0.9569 | SRCC = 0.9576 ] +INFO:root:Average loss=0.036376521423459054 +INFO:root: +[Valid][28][UTT][ MSE = 0.1671 | LCC = 0.9588 | SRCC = 0.9597 ] +INFO:root:Average loss=0.035047107651829716 +INFO:root: +[Valid][29][UTT][ MSE = 0.1824 | LCC = 0.9583 | SRCC = 0.9586 ] +INFO:root:Average loss=0.03213062910735607 +INFO:root: +[Valid][30][UTT][ MSE = 0.2087 | LCC = 0.9587 | SRCC = 0.9589 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1868 | LCC = 0.9603 | SRCC = 0.9611 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..56c6c86041f864ec9bdceba31cb0a22a50d2a2a6 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2bf4c42796ce00954a5cde3e7cf68a6d217f9032e2180e90535f67f5f20e8fc2 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6df2e467d9260baa38318f053a46baf8b21abec2 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1021703b20905afe1d6905d39ff75a2946409bcbb18abebaf091aa35dc6e25cd +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..190ab7aa187929286d6941e331918d365d05550a --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer3/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.40672116029262545 +INFO:root: +[Valid][0][UTT][ MSE = 0.3718 | LCC = 0.9262 | SRCC = 0.9291 ] +INFO:root:Average loss=0.18374787986278535 +INFO:root: +[Valid][1][UTT][ MSE = 0.2594 | LCC = 0.9414 | SRCC = 0.9436 ] +INFO:root:Average loss=0.15089002603292465 +INFO:root: +[Valid][2][UTT][ MSE = 0.2609 | LCC = 0.9492 | SRCC = 0.9504 ] +INFO:root:Average loss=0.12958956599235535 +INFO:root: +[Valid][3][UTT][ MSE = 0.2667 | LCC = 0.9497 | SRCC = 0.9514 ] +INFO:root:Average loss=0.12106955033540726 +INFO:root: +[Valid][4][UTT][ MSE = 0.2457 | LCC = 0.9519 | SRCC = 0.9548 ] +INFO:root:Average loss=0.11277154576778411 +INFO:root: +[Valid][5][UTT][ MSE = 0.2536 | LCC = 0.9560 | SRCC = 0.9575 ] +INFO:root:Average loss=0.09903341907262803 +INFO:root: +[Valid][6][UTT][ MSE = 0.2169 | LCC = 0.9564 | SRCC = 0.9576 ] +INFO:root:Average loss=0.09372895836830139 +INFO:root: +[Valid][7][UTT][ MSE = 0.2222 | LCC = 0.9571 | SRCC = 0.9586 ] +INFO:root:Average loss=0.08402193564176559 +INFO:root: +[Valid][8][UTT][ MSE = 0.1953 | LCC = 0.9578 | SRCC = 0.9581 ] +INFO:root:Average loss=0.0811304191350937 +INFO:root: +[Valid][9][UTT][ MSE = 0.2122 | LCC = 0.9571 | SRCC = 0.9574 ] +INFO:root:Average loss=0.0802537427842617 +INFO:root: +[Valid][10][UTT][ MSE = 0.2361 | LCC = 0.9563 | SRCC = 0.9570 ] +INFO:root:Average loss=0.07048336577415466 +INFO:root: +[Valid][11][UTT][ MSE = 0.2483 | LCC = 0.9576 | SRCC = 0.9598 ] +INFO:root:Average loss=0.06751337069272995 +INFO:root: +[Valid][12][UTT][ MSE = 0.2428 | LCC = 0.9579 | SRCC = 0.9597 ] +INFO:root:Average loss=0.0661210739016533 +INFO:root: +[Valid][13][UTT][ MSE = 0.2446 | LCC = 0.9573 | SRCC = 0.9589 ] +INFO:root:Average loss=0.06411034089326859 +INFO:root: +[Valid][14][UTT][ MSE = 0.2593 | LCC = 0.9573 | SRCC = 0.9594 ] +INFO:root:Average loss=0.056224485635757444 +INFO:root: +[Valid][15][UTT][ MSE = 0.2221 | LCC = 0.9578 | SRCC = 0.9594 ] +INFO:root:Average loss=0.05651949329674244 +INFO:root: +[Valid][16][UTT][ MSE = 0.2352 | LCC = 0.9565 | SRCC = 0.9585 ] +INFO:root:Average loss=0.0560292499512434 +INFO:root: +[Valid][17][UTT][ MSE = 0.2210 | LCC = 0.9596 | SRCC = 0.9604 ] +INFO:root:Average loss=0.05033625008165837 +INFO:root: +[Valid][18][UTT][ MSE = 0.2071 | LCC = 0.9587 | SRCC = 0.9594 ] +INFO:root:Average loss=0.0500062614530325 +INFO:root: +[Valid][19][UTT][ MSE = 0.2257 | LCC = 0.9575 | SRCC = 0.9588 ] +INFO:root:Average loss=0.04642871583998203 +INFO:root: +[Valid][20][UTT][ MSE = 0.2192 | LCC = 0.9578 | SRCC = 0.9591 ] +INFO:root:Average loss=0.04559517307579517 +INFO:root: +[Valid][21][UTT][ MSE = 0.2237 | LCC = 0.9589 | SRCC = 0.9600 ] +INFO:root:Average loss=0.04261394739151001 +INFO:root: +[Valid][22][UTT][ MSE = 0.2025 | LCC = 0.9597 | SRCC = 0.9602 ] +INFO:root:Average loss=0.04195428490638733 +INFO:root: +[Valid][23][UTT][ MSE = 0.2559 | LCC = 0.9572 | SRCC = 0.9596 ] +INFO:root:Average loss=0.04276272839307785 +INFO:root: +[Valid][24][UTT][ MSE = 0.1985 | LCC = 0.9590 | SRCC = 0.9598 ] +INFO:root:Average loss=0.039798954471945765 +INFO:root: +[Valid][25][UTT][ MSE = 0.2310 | LCC = 0.9555 | SRCC = 0.9565 ] +INFO:root:Average loss=0.04110429221391678 +INFO:root: +[Valid][26][UTT][ MSE = 0.2273 | LCC = 0.9608 | SRCC = 0.9616 ] +INFO:root:Average loss=0.03710890752077103 +INFO:root: +[Valid][27][UTT][ MSE = 0.2208 | LCC = 0.9582 | SRCC = 0.9588 ] +INFO:root:Average loss=0.0339703518897295 +INFO:root: +[Valid][28][UTT][ MSE = 0.2180 | LCC = 0.9589 | SRCC = 0.9599 ] +INFO:root:Average loss=0.03451389193534851 +INFO:root: +[Valid][29][UTT][ MSE = 0.2104 | LCC = 0.9599 | SRCC = 0.9606 ] +INFO:root:Average loss=0.03204993318021297 +INFO:root: +[Valid][30][UTT][ MSE = 0.2228 | LCC = 0.9586 | SRCC = 0.9596 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2412 | LCC = 0.9599 | SRCC = 0.9613 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1b416727b3cd423d6bcaeaf01a8c87f4a7a97296 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6bb53cbef74387e039915fe46f7d827e63a099cd9eff695537d1e66721450818 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..859855503f13ee8e9cefd6ccd4701a63c4125800 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4d96496f5c3f8a01df546895cb1374e1d1ed1203332e76944d0b8a71defe125d +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..7ba72043eb819ef1c2c00e14a3fecb313fd5518c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.40976391315460203 +INFO:root: +[Valid][0][UTT][ MSE = 0.2786 | LCC = 0.9272 | SRCC = 0.9291 ] +INFO:root:Average loss=0.19563726055622102 +INFO:root: +[Valid][1][UTT][ MSE = 0.2701 | LCC = 0.9402 | SRCC = 0.9407 ] +INFO:root:Average loss=0.1568712238073349 +INFO:root: +[Valid][2][UTT][ MSE = 0.2493 | LCC = 0.9481 | SRCC = 0.9496 ] +INFO:root:Average loss=0.13730275040864945 +INFO:root: +[Valid][3][UTT][ MSE = 0.2272 | LCC = 0.9527 | SRCC = 0.9536 ] +INFO:root:Average loss=0.12468938422203064 +INFO:root: +[Valid][4][UTT][ MSE = 0.2335 | LCC = 0.9527 | SRCC = 0.9543 ] +INFO:root:Average loss=0.1111413520872593 +INFO:root: +[Valid][5][UTT][ MSE = 0.2233 | LCC = 0.9557 | SRCC = 0.9564 ] +INFO:root:Average loss=0.10374339339137077 +INFO:root: +[Valid][6][UTT][ MSE = 0.2152 | LCC = 0.9527 | SRCC = 0.9526 ] +INFO:root:Average loss=0.09980350521206856 +INFO:root: +[Valid][7][UTT][ MSE = 0.2063 | LCC = 0.9568 | SRCC = 0.9566 ] +INFO:root:Average loss=0.09289584240317345 +INFO:root: +[Valid][8][UTT][ MSE = 0.2101 | LCC = 0.9566 | SRCC = 0.9568 ] +INFO:root:Average loss=0.08598067528009415 +INFO:root: +[Valid][9][UTT][ MSE = 0.2086 | LCC = 0.9591 | SRCC = 0.9593 ] +INFO:root:Average loss=0.08236808416247368 +INFO:root: +[Valid][10][UTT][ MSE = 0.2079 | LCC = 0.9588 | SRCC = 0.9588 ] +INFO:root:Average loss=0.07429955765604973 +INFO:root: +[Valid][11][UTT][ MSE = 0.2211 | LCC = 0.9592 | SRCC = 0.9592 ] +INFO:root:Average loss=0.07402410253882408 +INFO:root: +[Valid][12][UTT][ MSE = 0.2255 | LCC = 0.9585 | SRCC = 0.9596 ] +INFO:root:Average loss=0.06767480972409248 +INFO:root: +[Valid][13][UTT][ MSE = 0.2220 | LCC = 0.9590 | SRCC = 0.9595 ] +INFO:root:Average loss=0.06449999190866947 +INFO:root: +[Valid][14][UTT][ MSE = 0.2084 | LCC = 0.9591 | SRCC = 0.9599 ] +INFO:root:Average loss=0.06057469426095486 +INFO:root: +[Valid][15][UTT][ MSE = 0.2223 | LCC = 0.9581 | SRCC = 0.9589 ] +INFO:root:Average loss=0.05665587759017944 +INFO:root: +[Valid][16][UTT][ MSE = 0.1957 | LCC = 0.9598 | SRCC = 0.9595 ] +INFO:root:Average loss=0.05477933821082115 +INFO:root: +[Valid][17][UTT][ MSE = 0.2076 | LCC = 0.9598 | SRCC = 0.9600 ] +INFO:root:Average loss=0.053514929041266444 +INFO:root: +[Valid][18][UTT][ MSE = 0.2349 | LCC = 0.9581 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05054462619125843 +INFO:root: +[Valid][19][UTT][ MSE = 0.2173 | LCC = 0.9595 | SRCC = 0.9598 ] +INFO:root:Average loss=0.04945348197221756 +INFO:root: +[Valid][20][UTT][ MSE = 0.1847 | LCC = 0.9598 | SRCC = 0.9589 ] +INFO:root:Average loss=0.046084348320961 +INFO:root: +[Valid][21][UTT][ MSE = 0.1827 | LCC = 0.9603 | SRCC = 0.9604 ] +INFO:root:Average loss=0.04651263071596622 +INFO:root: +[Valid][22][UTT][ MSE = 0.2106 | LCC = 0.9584 | SRCC = 0.9592 ] +INFO:root:Average loss=0.044324273273348806 +INFO:root: +[Valid][23][UTT][ MSE = 0.1918 | LCC = 0.9604 | SRCC = 0.9603 ] +INFO:root:Average loss=0.03949837630987167 +INFO:root: +[Valid][24][UTT][ MSE = 0.1850 | LCC = 0.9586 | SRCC = 0.9591 ] +INFO:root:Average loss=0.03916199314594269 +INFO:root: +[Valid][25][UTT][ MSE = 0.2045 | LCC = 0.9584 | SRCC = 0.9595 ] +INFO:root:Average loss=0.037430680349469184 +INFO:root: +[Valid][26][UTT][ MSE = 0.2013 | LCC = 0.9597 | SRCC = 0.9595 ] +INFO:root:Average loss=0.03709499007463455 +INFO:root: +[Valid][27][UTT][ MSE = 0.1987 | LCC = 0.9607 | SRCC = 0.9599 ] +INFO:root:Average loss=0.037674958020448686 +INFO:root: +[Valid][28][UTT][ MSE = 0.2080 | LCC = 0.9593 | SRCC = 0.9598 ] +INFO:root:Average loss=0.03314765465259552 +INFO:root: +[Valid][29][UTT][ MSE = 0.2313 | LCC = 0.9582 | SRCC = 0.9596 ] +INFO:root:Average loss=0.03306786145269871 +INFO:root: +[Valid][30][UTT][ MSE = 0.2257 | LCC = 0.9575 | SRCC = 0.9587 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2081 | LCC = 0.9579 | SRCC = 0.9596 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..121486db8d655612ba0fa0224dd7caeafc116af9 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:38cd2ed56f7abd2a071d3e37b7cd71cd95d906370cddde7ffd3c85ae6bb3c655 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..56cdf9ae15be3ee01fbe92ea82a849055b08d2be --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:79ae251e9f947b01ab3c0ea249c02c195fa4ab3b32999221755fb5ba44aa3813 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..08807a892fa51476122f6862d99dd5280ff3e3b5 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4197057344913483 +INFO:root: +[Valid][0][UTT][ MSE = 0.3098 | LCC = 0.9326 | SRCC = 0.9329 ] +INFO:root:Average loss=0.18309338849782944 +INFO:root: +[Valid][1][UTT][ MSE = 0.2447 | LCC = 0.9434 | SRCC = 0.9439 ] +INFO:root:Average loss=0.15081071561574935 +INFO:root: +[Valid][2][UTT][ MSE = 0.2279 | LCC = 0.9497 | SRCC = 0.9509 ] +INFO:root:Average loss=0.13151079428195953 +INFO:root: +[Valid][3][UTT][ MSE = 0.2112 | LCC = 0.9529 | SRCC = 0.9537 ] +INFO:root:Average loss=0.11891608482599259 +INFO:root: +[Valid][4][UTT][ MSE = 0.2501 | LCC = 0.9534 | SRCC = 0.9556 ] +INFO:root:Average loss=0.11334395962953568 +INFO:root: +[Valid][5][UTT][ MSE = 0.2376 | LCC = 0.9520 | SRCC = 0.9525 ] +INFO:root:Average loss=0.10710218712687493 +INFO:root: +[Valid][6][UTT][ MSE = 0.1997 | LCC = 0.9575 | SRCC = 0.9583 ] +INFO:root:Average loss=0.09637548556923867 +INFO:root: +[Valid][7][UTT][ MSE = 0.2262 | LCC = 0.9568 | SRCC = 0.9578 ] +INFO:root:Average loss=0.09212515684962273 +INFO:root: +[Valid][8][UTT][ MSE = 0.2189 | LCC = 0.9560 | SRCC = 0.9567 ] +INFO:root:Average loss=0.08677915301918983 +INFO:root: +[Valid][9][UTT][ MSE = 0.2179 | LCC = 0.9562 | SRCC = 0.9564 ] +INFO:root:Average loss=0.07832262563705444 +INFO:root: +[Valid][10][UTT][ MSE = 0.2026 | LCC = 0.9574 | SRCC = 0.9573 ] +INFO:root:Average loss=0.07343999251723289 +INFO:root: +[Valid][11][UTT][ MSE = 0.1815 | LCC = 0.9584 | SRCC = 0.9584 ] +INFO:root:Average loss=0.06838023012876511 +INFO:root: +[Valid][12][UTT][ MSE = 0.2064 | LCC = 0.9590 | SRCC = 0.9590 ] +INFO:root:Average loss=0.06429157224297523 +INFO:root: +[Valid][13][UTT][ MSE = 0.1776 | LCC = 0.9597 | SRCC = 0.9593 ] +INFO:root:Average loss=0.06371167540550232 +INFO:root: +[Valid][14][UTT][ MSE = 0.2194 | LCC = 0.9579 | SRCC = 0.9582 ] +INFO:root:Average loss=0.05894143351912499 +INFO:root: +[Valid][15][UTT][ MSE = 0.2238 | LCC = 0.9569 | SRCC = 0.9578 ] +INFO:root:Average loss=0.05548419666290283 +INFO:root: +[Valid][16][UTT][ MSE = 0.2339 | LCC = 0.9574 | SRCC = 0.9583 ] +INFO:root:Average loss=0.05434088909626007 +INFO:root: +[Valid][17][UTT][ MSE = 0.1793 | LCC = 0.9588 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05144071350991726 +INFO:root: +[Valid][18][UTT][ MSE = 0.1974 | LCC = 0.9573 | SRCC = 0.9573 ] +INFO:root:Average loss=0.04898289714753628 +INFO:root: +[Valid][19][UTT][ MSE = 0.1909 | LCC = 0.9588 | SRCC = 0.9579 ] +INFO:root:Average loss=0.04861725869774818 +INFO:root: +[Valid][20][UTT][ MSE = 0.2188 | LCC = 0.9590 | SRCC = 0.9587 ] +INFO:root:Average loss=0.04301817090809345 +INFO:root: +[Valid][21][UTT][ MSE = 0.2205 | LCC = 0.9590 | SRCC = 0.9589 ] +INFO:root:Average loss=0.04343731559813022 +INFO:root: +[Valid][22][UTT][ MSE = 0.2183 | LCC = 0.9597 | SRCC = 0.9592 ] +INFO:root:Average loss=0.044194620415568354 +INFO:root: +[Valid][23][UTT][ MSE = 0.1851 | LCC = 0.9601 | SRCC = 0.9593 ] +INFO:root:Average loss=0.03966616404056549 +INFO:root: +[Valid][24][UTT][ MSE = 0.2173 | LCC = 0.9583 | SRCC = 0.9588 ] +INFO:root:Average loss=0.038757167860865596 +INFO:root: +[Valid][25][UTT][ MSE = 0.2018 | LCC = 0.9589 | SRCC = 0.9588 ] +INFO:root:Average loss=0.03641550251841545 +INFO:root: +[Valid][26][UTT][ MSE = 0.1863 | LCC = 0.9602 | SRCC = 0.9594 ] +INFO:root:Average loss=0.0353197306394577 +INFO:root: +[Valid][27][UTT][ MSE = 0.2207 | LCC = 0.9599 | SRCC = 0.9593 ] +INFO:root:Average loss=0.0332793447971344 +INFO:root: +[Valid][28][UTT][ MSE = 0.1974 | LCC = 0.9595 | SRCC = 0.9588 ] +INFO:root:Average loss=0.03173627963662148 +INFO:root: +[Valid][29][UTT][ MSE = 0.2070 | LCC = 0.9586 | SRCC = 0.9587 ] +INFO:root:Average loss=0.031959924772381784 +INFO:root: +[Valid][30][UTT][ MSE = 0.1953 | LCC = 0.9605 | SRCC = 0.9597 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2026 | LCC = 0.9579 | SRCC = 0.9595 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..03d4477fc0b32a3c957a8e538f11bc5da596e2c4 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d95a50b02728c18c924063b4ce3b34f0835c9b82a75f90841775735bf08adb3b +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f745465634af4d452d1de1a3edbfca104fbca30d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:91c4c5f05b275196732a8dba3a20d334d18ebed839586c89d05ad0fe2b3cd809 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..511f73f21e844b59f767797238607ca852d4f9ba --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3961967119574547 +INFO:root: +[Valid][0][UTT][ MSE = 0.2862 | LCC = 0.9312 | SRCC = 0.9313 ] +INFO:root:Average loss=0.18133943802118302 +INFO:root: +[Valid][1][UTT][ MSE = 0.2785 | LCC = 0.9440 | SRCC = 0.9449 ] +INFO:root:Average loss=0.15179079782962798 +INFO:root: +[Valid][2][UTT][ MSE = 0.2273 | LCC = 0.9481 | SRCC = 0.9482 ] +INFO:root:Average loss=0.13727449199557304 +INFO:root: +[Valid][3][UTT][ MSE = 0.2267 | LCC = 0.9514 | SRCC = 0.9519 ] +INFO:root:Average loss=0.12001691228151322 +INFO:root: +[Valid][4][UTT][ MSE = 0.2193 | LCC = 0.9534 | SRCC = 0.9544 ] +INFO:root:Average loss=0.11664644759893418 +INFO:root: +[Valid][5][UTT][ MSE = 0.2396 | LCC = 0.9552 | SRCC = 0.9562 ] +INFO:root:Average loss=0.10758652719855309 +INFO:root: +[Valid][6][UTT][ MSE = 0.2055 | LCC = 0.9530 | SRCC = 0.9534 ] +INFO:root:Average loss=0.0977199000120163 +INFO:root: +[Valid][7][UTT][ MSE = 0.2217 | LCC = 0.9554 | SRCC = 0.9565 ] +INFO:root:Average loss=0.09323750710487366 +INFO:root: +[Valid][8][UTT][ MSE = 0.2227 | LCC = 0.9561 | SRCC = 0.9568 ] +INFO:root:Average loss=0.0870637620985508 +INFO:root: +[Valid][9][UTT][ MSE = 0.1780 | LCC = 0.9574 | SRCC = 0.9568 ] +INFO:root:Average loss=0.08309251695871353 +INFO:root: +[Valid][10][UTT][ MSE = 0.2196 | LCC = 0.9549 | SRCC = 0.9543 ] +INFO:root:Average loss=0.07762299010157585 +INFO:root: +[Valid][11][UTT][ MSE = 0.1903 | LCC = 0.9557 | SRCC = 0.9558 ] +INFO:root:Average loss=0.06932515984773636 +INFO:root: +[Valid][12][UTT][ MSE = 0.2043 | LCC = 0.9600 | SRCC = 0.9594 ] +INFO:root:Average loss=0.06639607399702072 +INFO:root: +[Valid][13][UTT][ MSE = 0.2130 | LCC = 0.9590 | SRCC = 0.9586 ] +INFO:root:Average loss=0.06320654782652856 +INFO:root: +[Valid][14][UTT][ MSE = 0.2128 | LCC = 0.9567 | SRCC = 0.9559 ] +INFO:root:Average loss=0.05954228800535202 +INFO:root: +[Valid][15][UTT][ MSE = 0.2102 | LCC = 0.9590 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05657498130202293 +INFO:root: +[Valid][16][UTT][ MSE = 0.1884 | LCC = 0.9598 | SRCC = 0.9589 ] +INFO:root:Average loss=0.05636400318145752 +INFO:root: +[Valid][17][UTT][ MSE = 0.2088 | LCC = 0.9593 | SRCC = 0.9587 ] +INFO:root:Average loss=0.053352175116539 +INFO:root: +[Valid][18][UTT][ MSE = 0.1988 | LCC = 0.9542 | SRCC = 0.9548 ] +INFO:root:Average loss=0.05158473858237266 +INFO:root: +[Valid][19][UTT][ MSE = 0.1851 | LCC = 0.9574 | SRCC = 0.9582 ] +INFO:root:Average loss=0.046585344552993777 +INFO:root: +[Valid][20][UTT][ MSE = 0.1889 | LCC = 0.9597 | SRCC = 0.9598 ] +INFO:root:Average loss=0.04430999553203583 +INFO:root: +[Valid][21][UTT][ MSE = 0.1978 | LCC = 0.9594 | SRCC = 0.9589 ] +INFO:root:Average loss=0.04596231846511364 +INFO:root: +[Valid][22][UTT][ MSE = 0.1761 | LCC = 0.9571 | SRCC = 0.9575 ] +INFO:root:Average loss=0.040791209906339644 +INFO:root: +[Valid][23][UTT][ MSE = 0.1912 | LCC = 0.9575 | SRCC = 0.9579 ] +INFO:root:Average loss=0.0415332874506712 +INFO:root: +[Valid][24][UTT][ MSE = 0.1788 | LCC = 0.9588 | SRCC = 0.9583 ] +INFO:root:Average loss=0.038877969473600386 +INFO:root: +[Valid][25][UTT][ MSE = 0.1859 | LCC = 0.9595 | SRCC = 0.9591 ] +INFO:root:Average loss=0.03844251717627049 +INFO:root: +[Valid][26][UTT][ MSE = 0.1956 | LCC = 0.9580 | SRCC = 0.9575 ] +INFO:root:Average loss=0.036193000480532644 +INFO:root: +[Valid][27][UTT][ MSE = 0.1985 | LCC = 0.9582 | SRCC = 0.9572 ] +INFO:root:Average loss=0.035830357208848 +INFO:root: +[Valid][28][UTT][ MSE = 0.1718 | LCC = 0.9590 | SRCC = 0.9587 ] +INFO:root:Average loss=0.03692905642092228 +INFO:root: +[Valid][29][UTT][ MSE = 0.1832 | LCC = 0.9596 | SRCC = 0.9593 ] +INFO:root:Average loss=0.03280617184937 +INFO:root: +[Valid][30][UTT][ MSE = 0.2194 | LCC = 0.9565 | SRCC = 0.9559 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2194 | LCC = 0.9575 | SRCC = 0.9587 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c27d8a2d9abef9effeddb84ec2588fc5bbbf1423 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e95724aaecdbaa3d28d7ad1f5cd97eb969e0275d620188572e5b0f7adb868cb3 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..fd4d9dd7397d415d456aca64934d543fe9e896ad --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1484e15d9d0f36c4189675f6febfac0fe7bcc7df50b1fd842368e6a64504deb6 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2fea7130f763f71476a42eb13865f433feb942b7 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.38186923730373384 +INFO:root: +[Valid][0][UTT][ MSE = 0.3019 | LCC = 0.9287 | SRCC = 0.9282 ] +INFO:root:Average loss=0.1836319875717163 +INFO:root: +[Valid][1][UTT][ MSE = 0.2818 | LCC = 0.9425 | SRCC = 0.9440 ] +INFO:root:Average loss=0.1577769368290901 +INFO:root: +[Valid][2][UTT][ MSE = 0.2568 | LCC = 0.9461 | SRCC = 0.9478 ] +INFO:root:Average loss=0.1347257359623909 +INFO:root: +[Valid][3][UTT][ MSE = 0.2285 | LCC = 0.9509 | SRCC = 0.9517 ] +INFO:root:Average loss=0.1245771821141243 +INFO:root: +[Valid][4][UTT][ MSE = 0.2397 | LCC = 0.9535 | SRCC = 0.9544 ] +INFO:root:Average loss=0.11474167525768281 +INFO:root: +[Valid][5][UTT][ MSE = 0.2395 | LCC = 0.9545 | SRCC = 0.9555 ] +INFO:root:Average loss=0.10471625837683678 +INFO:root: +[Valid][6][UTT][ MSE = 0.2032 | LCC = 0.9540 | SRCC = 0.9546 ] +INFO:root:Average loss=0.09726914557814598 +INFO:root: +[Valid][7][UTT][ MSE = 0.2091 | LCC = 0.9573 | SRCC = 0.9576 ] +INFO:root:Average loss=0.09280273455381394 +INFO:root: +[Valid][8][UTT][ MSE = 0.2134 | LCC = 0.9559 | SRCC = 0.9566 ] +INFO:root:Average loss=0.08436309742927552 +INFO:root: +[Valid][9][UTT][ MSE = 0.2370 | LCC = 0.9577 | SRCC = 0.9583 ] +INFO:root:Average loss=0.08172645425796508 +INFO:root: +[Valid][10][UTT][ MSE = 0.2055 | LCC = 0.9577 | SRCC = 0.9581 ] +INFO:root:Average loss=0.07526400217413902 +INFO:root: +[Valid][11][UTT][ MSE = 0.2156 | LCC = 0.9580 | SRCC = 0.9581 ] +INFO:root:Average loss=0.07210629764199257 +INFO:root: +[Valid][12][UTT][ MSE = 0.2042 | LCC = 0.9552 | SRCC = 0.9560 ] +INFO:root:Average loss=0.06623179668188095 +INFO:root: +[Valid][13][UTT][ MSE = 0.2022 | LCC = 0.9588 | SRCC = 0.9591 ] +INFO:root:Average loss=0.062233398020267484 +INFO:root: +[Valid][14][UTT][ MSE = 0.2200 | LCC = 0.9572 | SRCC = 0.9573 ] +INFO:root:Average loss=0.0586146282851696 +INFO:root: +[Valid][15][UTT][ MSE = 0.1977 | LCC = 0.9585 | SRCC = 0.9583 ] +INFO:root:Average loss=0.0583320207297802 +INFO:root: +[Valid][16][UTT][ MSE = 0.1876 | LCC = 0.9578 | SRCC = 0.9573 ] +INFO:root:Average loss=0.05586474722623825 +INFO:root: +[Valid][17][UTT][ MSE = 0.2321 | LCC = 0.9551 | SRCC = 0.9548 ] +INFO:root:Average loss=0.05145584818720818 +INFO:root: +[Valid][18][UTT][ MSE = 0.2053 | LCC = 0.9584 | SRCC = 0.9587 ] +INFO:root:Average loss=0.048877687618136406 +INFO:root: +[Valid][19][UTT][ MSE = 0.1915 | LCC = 0.9598 | SRCC = 0.9593 ] +INFO:root:Average loss=0.04658375705778599 +INFO:root: +[Valid][20][UTT][ MSE = 0.1768 | LCC = 0.9579 | SRCC = 0.9578 ] +INFO:root:Average loss=0.046444664478302 +INFO:root: +[Valid][21][UTT][ MSE = 0.1971 | LCC = 0.9596 | SRCC = 0.9592 ] +INFO:root:Average loss=0.04304086712002754 +INFO:root: +[Valid][22][UTT][ MSE = 0.1998 | LCC = 0.9598 | SRCC = 0.9594 ] +INFO:root:Average loss=0.04242153112590313 +INFO:root: +[Valid][23][UTT][ MSE = 0.1883 | LCC = 0.9591 | SRCC = 0.9587 ] +INFO:root:Average loss=0.04223848967254162 +INFO:root: +[Valid][24][UTT][ MSE = 0.1788 | LCC = 0.9593 | SRCC = 0.9589 ] +INFO:root:Average loss=0.04072936570644379 +INFO:root: +[Valid][25][UTT][ MSE = 0.2161 | LCC = 0.9578 | SRCC = 0.9568 ] +INFO:root:Average loss=0.039102175027132036 +INFO:root: +[Valid][26][UTT][ MSE = 0.1852 | LCC = 0.9606 | SRCC = 0.9599 ] +INFO:root:Average loss=0.03703291481733322 +INFO:root: +[Valid][27][UTT][ MSE = 0.1982 | LCC = 0.9599 | SRCC = 0.9594 ] +INFO:root:Average loss=0.03321303054690361 +INFO:root: +[Valid][28][UTT][ MSE = 0.2086 | LCC = 0.9591 | SRCC = 0.9588 ] +INFO:root:Average loss=0.03330492506921291 +INFO:root: +[Valid][29][UTT][ MSE = 0.1737 | LCC = 0.9601 | SRCC = 0.9595 ] +INFO:root:Average loss=0.03672857034206391 +INFO:root: +[Valid][30][UTT][ MSE = 0.1860 | LCC = 0.9577 | SRCC = 0.9573 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1979 | LCC = 0.9555 | SRCC = 0.9568 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c6f04b303693e32bf8a48a53b9a79d8da03f04a9 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43b59bc0cdde4a808713cc5fbe9514f8cb5f34d35c4a7e6dec1b3300c1576a75 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run4/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..8acd695edde6e8ef60c7dbc98288940c173cea3e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bfb9e2c7c4a893a035b5ab9087ba719699b239be09bf0931108f86720aaad58b +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run4/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..99ada83afdd5fe2ed3c57ce3f6408a6fc4c018a9 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer4/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3896739808320999 +INFO:root: +[Valid][0][UTT][ MSE = 0.2773 | LCC = 0.9323 | SRCC = 0.9352 ] +INFO:root:Average loss=0.1840283548235893 +INFO:root: +[Valid][1][UTT][ MSE = 0.2846 | LCC = 0.9451 | SRCC = 0.9466 ] +INFO:root:Average loss=0.14905599123239519 +INFO:root: +[Valid][2][UTT][ MSE = 0.2667 | LCC = 0.9500 | SRCC = 0.9517 ] +INFO:root:Average loss=0.13371629178524017 +INFO:root: +[Valid][3][UTT][ MSE = 0.2682 | LCC = 0.9518 | SRCC = 0.9531 ] +INFO:root:Average loss=0.1185966283082962 +INFO:root: +[Valid][4][UTT][ MSE = 0.2157 | LCC = 0.9554 | SRCC = 0.9562 ] +INFO:root:Average loss=0.10874298098683358 +INFO:root: +[Valid][5][UTT][ MSE = 0.2128 | LCC = 0.9570 | SRCC = 0.9578 ] +INFO:root:Average loss=0.1005050473511219 +INFO:root: +[Valid][6][UTT][ MSE = 0.2156 | LCC = 0.9568 | SRCC = 0.9572 ] +INFO:root:Average loss=0.09385505720973014 +INFO:root: +[Valid][7][UTT][ MSE = 0.2125 | LCC = 0.9578 | SRCC = 0.9587 ] +INFO:root:Average loss=0.09385308134555817 +INFO:root: +[Valid][8][UTT][ MSE = 0.1991 | LCC = 0.9587 | SRCC = 0.9597 ] +INFO:root:Average loss=0.08023590368032456 +INFO:root: +[Valid][9][UTT][ MSE = 0.1799 | LCC = 0.9603 | SRCC = 0.9602 ] +INFO:root:Average loss=0.07513035565614701 +INFO:root: +[Valid][10][UTT][ MSE = 0.2596 | LCC = 0.9596 | SRCC = 0.9607 ] +INFO:root:Average loss=0.07100962346792221 +INFO:root: +[Valid][11][UTT][ MSE = 0.2050 | LCC = 0.9596 | SRCC = 0.9607 ] +INFO:root:Average loss=0.06787458726763726 +INFO:root: +[Valid][12][UTT][ MSE = 0.1995 | LCC = 0.9607 | SRCC = 0.9608 ] +INFO:root:Average loss=0.06351774689555169 +INFO:root: +[Valid][13][UTT][ MSE = 0.1922 | LCC = 0.9596 | SRCC = 0.9596 ] +INFO:root:Average loss=0.058533595770597455 +INFO:root: +[Valid][14][UTT][ MSE = 0.2145 | LCC = 0.9604 | SRCC = 0.9610 ] +INFO:root:Average loss=0.05815828911960125 +INFO:root: +[Valid][15][UTT][ MSE = 0.2385 | LCC = 0.9590 | SRCC = 0.9593 ] +INFO:root:Average loss=0.057380455255508424 +INFO:root: +[Valid][16][UTT][ MSE = 0.1814 | LCC = 0.9603 | SRCC = 0.9601 ] +INFO:root:Average loss=0.05136481776833534 +INFO:root: +[Valid][17][UTT][ MSE = 0.2026 | LCC = 0.9599 | SRCC = 0.9600 ] +INFO:root:Average loss=0.049525676861405375 +INFO:root: +[Valid][18][UTT][ MSE = 0.1934 | LCC = 0.9601 | SRCC = 0.9601 ] +INFO:root:Average loss=0.046864748537540434 +INFO:root: +[Valid][19][UTT][ MSE = 0.2171 | LCC = 0.9586 | SRCC = 0.9596 ] +INFO:root:Average loss=0.04657359391450882 +INFO:root: +[Valid][20][UTT][ MSE = 0.2036 | LCC = 0.9596 | SRCC = 0.9598 ] +INFO:root:Average loss=0.04288273000717163 +INFO:root: +[Valid][21][UTT][ MSE = 0.1873 | LCC = 0.9592 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04085408987104893 +INFO:root: +[Valid][22][UTT][ MSE = 0.2188 | LCC = 0.9606 | SRCC = 0.9608 ] +INFO:root:Average loss=0.04206668275594711 +INFO:root: +[Valid][23][UTT][ MSE = 0.2038 | LCC = 0.9586 | SRCC = 0.9585 ] +INFO:root:Average loss=0.038968561366200444 +INFO:root: +[Valid][24][UTT][ MSE = 0.2463 | LCC = 0.9570 | SRCC = 0.9587 ] +INFO:root:Average loss=0.03748026525974274 +INFO:root: +[Valid][25][UTT][ MSE = 0.2028 | LCC = 0.9585 | SRCC = 0.9582 ] +INFO:root:Average loss=0.035954231277108194 +INFO:root: +[Valid][26][UTT][ MSE = 0.2002 | LCC = 0.9596 | SRCC = 0.9593 ] +INFO:root:Average loss=0.033476821616292 +INFO:root: +[Valid][27][UTT][ MSE = 0.1786 | LCC = 0.9604 | SRCC = 0.9606 ] +INFO:root:Average loss=0.032869873449206354 +INFO:root: +[Valid][28][UTT][ MSE = 0.1819 | LCC = 0.9616 | SRCC = 0.9613 ] +INFO:root:Average loss=0.03324310082197189 +INFO:root: +[Valid][29][UTT][ MSE = 0.2054 | LCC = 0.9614 | SRCC = 0.9611 ] +INFO:root:Average loss=0.03191984287649393 +INFO:root: +[Valid][30][UTT][ MSE = 0.1859 | LCC = 0.9602 | SRCC = 0.9599 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1908 | LCC = 0.9591 | SRCC = 0.9604 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run0/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run0/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..45abb1ac691106b62211e5d3f69e400a4e5f6373 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0356ad73e09789914651b1b5fe16cc9561b1513ec4b77b4185136c9c0fbdcb28 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run0/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ae8b67db501fc3be57a404870f71086b08736159 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:73fb580e721ca4ab9ca5aa463baa1f09e2c016c813bb24b2b3fa93acb5d6483d +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run0/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c3a9268ba8088faa2ab19eb4619a06c1447caea7 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3521515680551529 +INFO:root: +[Valid][0][UTT][ MSE = 0.3500 | LCC = 0.9283 | SRCC = 0.9296 ] +INFO:root:Average loss=0.18941102409362792 +INFO:root: +[Valid][1][UTT][ MSE = 0.2554 | LCC = 0.9411 | SRCC = 0.9419 ] +INFO:root:Average loss=0.1554175283908844 +INFO:root: +[Valid][2][UTT][ MSE = 0.2726 | LCC = 0.9470 | SRCC = 0.9474 ] +INFO:root:Average loss=0.13853217279911043 +INFO:root: +[Valid][3][UTT][ MSE = 0.2208 | LCC = 0.9479 | SRCC = 0.9494 ] +INFO:root:Average loss=0.1241320235133171 +INFO:root: +[Valid][4][UTT][ MSE = 0.2252 | LCC = 0.9510 | SRCC = 0.9537 ] +INFO:root:Average loss=0.11153641021251678 +INFO:root: +[Valid][5][UTT][ MSE = 0.2174 | LCC = 0.9539 | SRCC = 0.9543 ] +INFO:root:Average loss=0.10566975462436676 +INFO:root: +[Valid][6][UTT][ MSE = 0.2407 | LCC = 0.9542 | SRCC = 0.9545 ] +INFO:root:Average loss=0.09696273854374886 +INFO:root: +[Valid][7][UTT][ MSE = 0.1951 | LCC = 0.9549 | SRCC = 0.9550 ] +INFO:root:Average loss=0.09072465366125107 +INFO:root: +[Valid][8][UTT][ MSE = 0.1952 | LCC = 0.9522 | SRCC = 0.9527 ] +INFO:root:Average loss=0.08448094654083252 +INFO:root: +[Valid][9][UTT][ MSE = 0.1902 | LCC = 0.9539 | SRCC = 0.9549 ] +INFO:root:Average loss=0.07859994763135911 +INFO:root: +[Valid][10][UTT][ MSE = 0.1946 | LCC = 0.9561 | SRCC = 0.9560 ] +INFO:root:Average loss=0.07243141406774521 +INFO:root: +[Valid][11][UTT][ MSE = 0.2079 | LCC = 0.9554 | SRCC = 0.9559 ] +INFO:root:Average loss=0.06956052309274674 +INFO:root: +[Valid][12][UTT][ MSE = 0.2182 | LCC = 0.9549 | SRCC = 0.9552 ] +INFO:root:Average loss=0.06613886657357215 +INFO:root: +[Valid][13][UTT][ MSE = 0.2289 | LCC = 0.9557 | SRCC = 0.9569 ] +INFO:root:Average loss=0.06219570544362068 +INFO:root: +[Valid][14][UTT][ MSE = 0.2249 | LCC = 0.9537 | SRCC = 0.9556 ] +INFO:root:Average loss=0.058884975284338 +INFO:root: +[Valid][15][UTT][ MSE = 0.2049 | LCC = 0.9544 | SRCC = 0.9550 ] +INFO:root:Average loss=0.055446144297719005 +INFO:root: +[Valid][16][UTT][ MSE = 0.1878 | LCC = 0.9539 | SRCC = 0.9550 ] +INFO:root:Average loss=0.05088705229759216 +INFO:root: +[Valid][17][UTT][ MSE = 0.1986 | LCC = 0.9547 | SRCC = 0.9551 ] +INFO:root:Average loss=0.05053432022035122 +INFO:root: +[Valid][18][UTT][ MSE = 0.2210 | LCC = 0.9538 | SRCC = 0.9547 ] +INFO:root:Average loss=0.051710828974843026 +INFO:root: +[Valid][19][UTT][ MSE = 0.1827 | LCC = 0.9554 | SRCC = 0.9557 ] +INFO:root:Average loss=0.048445750415325166 +INFO:root: +[Valid][20][UTT][ MSE = 0.2059 | LCC = 0.9564 | SRCC = 0.9564 ] +INFO:root:Average loss=0.044172617346048354 +INFO:root: +[Valid][21][UTT][ MSE = 0.1860 | LCC = 0.9551 | SRCC = 0.9550 ] +INFO:root:Average loss=0.042987498968839646 +INFO:root: +[Valid][22][UTT][ MSE = 0.2179 | LCC = 0.9539 | SRCC = 0.9549 ] +INFO:root:Average loss=0.04357734721899033 +INFO:root: +[Valid][23][UTT][ MSE = 0.1907 | LCC = 0.9538 | SRCC = 0.9545 ] +INFO:root:Average loss=0.03958565130829811 +INFO:root: +[Valid][24][UTT][ MSE = 0.2111 | LCC = 0.9553 | SRCC = 0.9559 ] +INFO:root:Average loss=0.0384377239048481 +INFO:root: +[Valid][25][UTT][ MSE = 0.1990 | LCC = 0.9549 | SRCC = 0.9551 ] +INFO:root:Average loss=0.03674681879580021 +INFO:root: +[Valid][26][UTT][ MSE = 0.2012 | LCC = 0.9547 | SRCC = 0.9557 ] +INFO:root:Average loss=0.03502953916788101 +INFO:root: +[Valid][27][UTT][ MSE = 0.2084 | LCC = 0.9553 | SRCC = 0.9559 ] +INFO:root:Average loss=0.03348429052531719 +INFO:root: +[Valid][28][UTT][ MSE = 0.1965 | LCC = 0.9565 | SRCC = 0.9569 ] +INFO:root:Average loss=0.034138038143515584 +INFO:root: +[Valid][29][UTT][ MSE = 0.1802 | LCC = 0.9560 | SRCC = 0.9560 ] +INFO:root:Average loss=0.03209398387372494 +INFO:root: +[Valid][30][UTT][ MSE = 0.1987 | LCC = 0.9545 | SRCC = 0.9546 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1969 | LCC = 0.9576 | SRCC = 0.9586 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run1/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run1/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..687f11bd5cf1b8b834132a981b2ac55ed37cfe6e --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ee382c439293671c288eccc4d97ef070258cf9d16e6399697c65459ca54a267d +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run1/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b8c434d65d14b819439ee158875443c974ba1a77 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:23e7e1442ec2caed4dffb029604a875c4927900fef6980a71527a6e810b0b99f +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run1/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..de3c767723313519a3cc04e9893be51dbfa1cd71 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3891458629965782 +INFO:root: +[Valid][0][UTT][ MSE = 0.3550 | LCC = 0.9266 | SRCC = 0.9274 ] +INFO:root:Average loss=0.19456314653158188 +INFO:root: +[Valid][1][UTT][ MSE = 0.2208 | LCC = 0.9356 | SRCC = 0.9397 ] +INFO:root:Average loss=0.15910591447353364 +INFO:root: +[Valid][2][UTT][ MSE = 0.3203 | LCC = 0.9455 | SRCC = 0.9479 ] +INFO:root:Average loss=0.13797100490331649 +INFO:root: +[Valid][3][UTT][ MSE = 0.3005 | LCC = 0.9487 | SRCC = 0.9502 ] +INFO:root:Average loss=0.12398311501741409 +INFO:root: +[Valid][4][UTT][ MSE = 0.2511 | LCC = 0.9504 | SRCC = 0.9508 ] +INFO:root:Average loss=0.11648653811216354 +INFO:root: +[Valid][5][UTT][ MSE = 0.2535 | LCC = 0.9525 | SRCC = 0.9534 ] +INFO:root:Average loss=0.10736062636971473 +INFO:root: +[Valid][6][UTT][ MSE = 0.2494 | LCC = 0.9523 | SRCC = 0.9524 ] +INFO:root:Average loss=0.09847553911805153 +INFO:root: +[Valid][7][UTT][ MSE = 0.2329 | LCC = 0.9503 | SRCC = 0.9507 ] +INFO:root:Average loss=0.09022566020488738 +INFO:root: +[Valid][8][UTT][ MSE = 0.2213 | LCC = 0.9545 | SRCC = 0.9543 ] +INFO:root:Average loss=0.08403911915421486 +INFO:root: +[Valid][9][UTT][ MSE = 0.2273 | LCC = 0.9539 | SRCC = 0.9544 ] +INFO:root:Average loss=0.08007037076354027 +INFO:root: +[Valid][10][UTT][ MSE = 0.2699 | LCC = 0.9541 | SRCC = 0.9561 ] +INFO:root:Average loss=0.0745079609155655 +INFO:root: +[Valid][11][UTT][ MSE = 0.2154 | LCC = 0.9506 | SRCC = 0.9524 ] +INFO:root:Average loss=0.07193538227677346 +INFO:root: +[Valid][12][UTT][ MSE = 0.2640 | LCC = 0.9549 | SRCC = 0.9562 ] +INFO:root:Average loss=0.06636203473806382 +INFO:root: +[Valid][13][UTT][ MSE = 0.2180 | LCC = 0.9541 | SRCC = 0.9545 ] +INFO:root:Average loss=0.06596233819425106 +INFO:root: +[Valid][14][UTT][ MSE = 0.2338 | LCC = 0.9555 | SRCC = 0.9562 ] +INFO:root:Average loss=0.06072505688667297 +INFO:root: +[Valid][15][UTT][ MSE = 0.2000 | LCC = 0.9554 | SRCC = 0.9555 ] +INFO:root:Average loss=0.059289256244897844 +INFO:root: +[Valid][16][UTT][ MSE = 0.2200 | LCC = 0.9564 | SRCC = 0.9568 ] +INFO:root:Average loss=0.0545545459240675 +INFO:root: +[Valid][17][UTT][ MSE = 0.2272 | LCC = 0.9557 | SRCC = 0.9561 ] +INFO:root:Average loss=0.05039114433526993 +INFO:root: +[Valid][18][UTT][ MSE = 0.2093 | LCC = 0.9552 | SRCC = 0.9559 ] +INFO:root:Average loss=0.051560937032103536 +INFO:root: +[Valid][19][UTT][ MSE = 0.2447 | LCC = 0.9554 | SRCC = 0.9561 ] +INFO:root:Average loss=0.048017529010772705 +INFO:root: +[Valid][20][UTT][ MSE = 0.2360 | LCC = 0.9543 | SRCC = 0.9550 ] +INFO:root:Average loss=0.045288794338703156 +INFO:root: +[Valid][21][UTT][ MSE = 0.1860 | LCC = 0.9558 | SRCC = 0.9561 ] +INFO:root:Average loss=0.04373704770207405 +INFO:root: +[Valid][22][UTT][ MSE = 0.2084 | LCC = 0.9551 | SRCC = 0.9553 ] +INFO:root:Average loss=0.042929927825927736 +INFO:root: +[Valid][23][UTT][ MSE = 0.1987 | LCC = 0.9552 | SRCC = 0.9553 ] +INFO:root:Average loss=0.041255953326821325 +INFO:root: +[Valid][24][UTT][ MSE = 0.1968 | LCC = 0.9546 | SRCC = 0.9547 ] +INFO:root:Average loss=0.038155638635158536 +INFO:root: +[Valid][25][UTT][ MSE = 0.2262 | LCC = 0.9551 | SRCC = 0.9557 ] +INFO:root:Average loss=0.039377200305461886 +INFO:root: +[Valid][26][UTT][ MSE = 0.2340 | LCC = 0.9543 | SRCC = 0.9554 ] +INFO:root:Average loss=0.03599828954041004 +INFO:root: +[Valid][27][UTT][ MSE = 0.1986 | LCC = 0.9556 | SRCC = 0.9558 ] +INFO:root:Average loss=0.034996068730950354 +INFO:root: +[Valid][28][UTT][ MSE = 0.1925 | LCC = 0.9559 | SRCC = 0.9555 ] +INFO:root:Average loss=0.03464907328784466 +INFO:root: +[Valid][29][UTT][ MSE = 0.2188 | LCC = 0.9553 | SRCC = 0.9556 ] +INFO:root:Average loss=0.033943608850240704 +INFO:root: +[Valid][30][UTT][ MSE = 0.1982 | LCC = 0.9564 | SRCC = 0.9562 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2065 | LCC = 0.9577 | SRCC = 0.9587 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run2/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run2/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b7290a93922705264a99f58d599f65e2a0d03014 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0980680fd12bbd0099ad7cc66798e008941645a1fc59fcfea1d90035d1b20f0 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run2/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ccce80d38e79af09d879dabba9975299dfa1348d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fb5b2bf7e38c3ba321ffaa59b2fa9f93158a31376100c2ed02df7868cdfcdc7e +size 2749638 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run2/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d25a112259b0b22d38a5d992ca70399efa650a65 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.39394612568616866 +INFO:root: +[Valid][0][UTT][ MSE = 0.3444 | LCC = 0.9287 | SRCC = 0.9315 ] +INFO:root:Average loss=0.18959981256723404 +INFO:root: +[Valid][1][UTT][ MSE = 0.2767 | LCC = 0.9435 | SRCC = 0.9453 ] +INFO:root:Average loss=0.15768910813331605 +INFO:root: +[Valid][2][UTT][ MSE = 0.2889 | LCC = 0.9460 | SRCC = 0.9472 ] +INFO:root:Average loss=0.13533767998218538 +INFO:root: +[Valid][3][UTT][ MSE = 0.2324 | LCC = 0.9521 | SRCC = 0.9529 ] +INFO:root:Average loss=0.12125271648168563 +INFO:root: +[Valid][4][UTT][ MSE = 0.2494 | LCC = 0.9527 | SRCC = 0.9547 ] +INFO:root:Average loss=0.11371499171853065 +INFO:root: +[Valid][5][UTT][ MSE = 0.2265 | LCC = 0.9549 | SRCC = 0.9553 ] +INFO:root:Average loss=0.1046644850075245 +INFO:root: +[Valid][6][UTT][ MSE = 0.2402 | LCC = 0.9500 | SRCC = 0.9510 ] +INFO:root:Average loss=0.096652567923069 +INFO:root: +[Valid][7][UTT][ MSE = 0.2068 | LCC = 0.9561 | SRCC = 0.9574 ] +INFO:root:Average loss=0.08901439070701599 +INFO:root: +[Valid][8][UTT][ MSE = 0.2136 | LCC = 0.9555 | SRCC = 0.9563 ] +INFO:root:Average loss=0.08240639343857765 +INFO:root: +[Valid][9][UTT][ MSE = 0.2410 | LCC = 0.9538 | SRCC = 0.9546 ] +INFO:root:Average loss=0.07904003283381462 +INFO:root: +[Valid][10][UTT][ MSE = 0.2418 | LCC = 0.9549 | SRCC = 0.9561 ] +INFO:root:Average loss=0.07253500774502754 +INFO:root: +[Valid][11][UTT][ MSE = 0.2862 | LCC = 0.9541 | SRCC = 0.9552 ] +INFO:root:Average loss=0.06928543677926063 +INFO:root: +[Valid][12][UTT][ MSE = 0.2060 | LCC = 0.9537 | SRCC = 0.9561 ] +INFO:root:Average loss=0.0703354858160019 +INFO:root: +[Valid][13][UTT][ MSE = 0.2304 | LCC = 0.9530 | SRCC = 0.9540 ] +INFO:root:Average loss=0.06313909378647804 +INFO:root: +[Valid][14][UTT][ MSE = 0.2017 | LCC = 0.9566 | SRCC = 0.9574 ] +INFO:root:Average loss=0.057976797729730606 +INFO:root: +[Valid][15][UTT][ MSE = 0.2022 | LCC = 0.9562 | SRCC = 0.9567 ] +INFO:root:Average loss=0.053705230474472045 +INFO:root: +[Valid][16][UTT][ MSE = 0.2438 | LCC = 0.9550 | SRCC = 0.9551 ] +INFO:root:Average loss=0.05360560782253742 +INFO:root: +[Valid][17][UTT][ MSE = 0.2186 | LCC = 0.9555 | SRCC = 0.9560 ] +INFO:root:Average loss=0.05064309257268906 +INFO:root: +[Valid][18][UTT][ MSE = 0.2440 | LCC = 0.9521 | SRCC = 0.9529 ] +INFO:root:Average loss=0.04860201105475426 +INFO:root: +[Valid][19][UTT][ MSE = 0.2271 | LCC = 0.9558 | SRCC = 0.9568 ] +INFO:root:Average loss=0.047945031061768534 +INFO:root: +[Valid][20][UTT][ MSE = 0.2098 | LCC = 0.9529 | SRCC = 0.9537 ] +INFO:root:Average loss=0.043432119473814965 +INFO:root: +[Valid][21][UTT][ MSE = 0.2182 | LCC = 0.9544 | SRCC = 0.9552 ] +INFO:root:Average loss=0.04351049910485744 +INFO:root: +[Valid][22][UTT][ MSE = 0.2358 | LCC = 0.9538 | SRCC = 0.9548 ] +INFO:root:Average loss=0.04020920544862747 +INFO:root: +[Valid][23][UTT][ MSE = 0.2189 | LCC = 0.9527 | SRCC = 0.9530 ] +INFO:root:Average loss=0.03943390469253063 +INFO:root: +[Valid][24][UTT][ MSE = 0.2677 | LCC = 0.9519 | SRCC = 0.9526 ] +INFO:root:Average loss=0.03803786511719227 +INFO:root: +[Valid][25][UTT][ MSE = 0.2189 | LCC = 0.9539 | SRCC = 0.9546 ] +INFO:root:Average loss=0.037356244176626204 +INFO:root: +[Valid][26][UTT][ MSE = 0.2092 | LCC = 0.9542 | SRCC = 0.9549 ] +INFO:root:Average loss=0.03413469190895557 +INFO:root: +[Valid][27][UTT][ MSE = 0.2545 | LCC = 0.9546 | SRCC = 0.9548 ] +INFO:root:Average loss=0.0343871782720089 +INFO:root: +[Valid][28][UTT][ MSE = 0.2269 | LCC = 0.9542 | SRCC = 0.9549 ] +INFO:root:Average loss=0.03226800273358822 +INFO:root: +[Valid][29][UTT][ MSE = 0.2041 | LCC = 0.9552 | SRCC = 0.9556 ] +INFO:root:Average loss=0.03164504693448544 +INFO:root: +[Valid][30][UTT][ MSE = 0.2259 | LCC = 0.9537 | SRCC = 0.9542 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2005 | LCC = 0.9585 | SRCC = 0.9604 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run3/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run3/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5af950b84c6e4de862f9bccc92da445c34048c3b --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:277225a1fda6539d12c5dc1bc3b02e68faab0699e0833ce40ba527c8d5229166 +size 2747979 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run3/model_best.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..16dc33b25d5199b34b4bfb0f08def4443acb9927 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e53118559c49b661845ad352e5b005bd90ae831c3e20b796009469a5a7ea7645 +size 2749702 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run3/train.log b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..59bda0c3c144d125116be56d41af31d5d93d41a8 --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4045218495130539 +INFO:root: +[Valid][0][UTT][ MSE = 0.2940 | LCC = 0.9299 | SRCC = 0.9304 ] +INFO:root:Average loss=0.18607682800292968 +INFO:root: +[Valid][1][UTT][ MSE = 0.2818 | LCC = 0.9411 | SRCC = 0.9422 ] +INFO:root:Average loss=0.15562763583660125 +INFO:root: +[Valid][2][UTT][ MSE = 0.2670 | LCC = 0.9479 | SRCC = 0.9488 ] +INFO:root:Average loss=0.1380761194229126 +INFO:root: +[Valid][3][UTT][ MSE = 0.2506 | LCC = 0.9513 | SRCC = 0.9525 ] +INFO:root:Average loss=0.12045952808856965 +INFO:root: +[Valid][4][UTT][ MSE = 0.2380 | LCC = 0.9526 | SRCC = 0.9551 ] +INFO:root:Average loss=0.11198065149784088 +INFO:root: +[Valid][5][UTT][ MSE = 0.1978 | LCC = 0.9519 | SRCC = 0.9529 ] +INFO:root:Average loss=0.10356247815489769 +INFO:root: +[Valid][6][UTT][ MSE = 0.2023 | LCC = 0.9523 | SRCC = 0.9526 ] +INFO:root:Average loss=0.09373614689707756 +INFO:root: +[Valid][7][UTT][ MSE = 0.1799 | LCC = 0.9544 | SRCC = 0.9555 ] +INFO:root:Average loss=0.08650289258360863 +INFO:root: +[Valid][8][UTT][ MSE = 0.2082 | LCC = 0.9559 | SRCC = 0.9565 ] +INFO:root:Average loss=0.08403970861434937 +INFO:root: +[Valid][9][UTT][ MSE = 0.1753 | LCC = 0.9534 | SRCC = 0.9551 ] +INFO:root:Average loss=0.07531432613730431 +INFO:root: +[Valid][10][UTT][ MSE = 0.1893 | LCC = 0.9558 | SRCC = 0.9566 ] +INFO:root:Average loss=0.072122664809227 +INFO:root: +[Valid][11][UTT][ MSE = 0.2220 | LCC = 0.9571 | SRCC = 0.9583 ] +INFO:root:Average loss=0.06783606851100922 +INFO:root: +[Valid][12][UTT][ MSE = 0.2184 | LCC = 0.9569 | SRCC = 0.9578 ] +INFO:root:Average loss=0.06532798498868943 +INFO:root: +[Valid][13][UTT][ MSE = 0.1892 | LCC = 0.9559 | SRCC = 0.9568 ] +INFO:root:Average loss=0.062252119153738024 +INFO:root: +[Valid][14][UTT][ MSE = 0.1872 | LCC = 0.9573 | SRCC = 0.9579 ] +INFO:root:Average loss=0.05700781607627869 +INFO:root: +[Valid][15][UTT][ MSE = 0.2075 | LCC = 0.9555 | SRCC = 0.9560 ] +INFO:root:Average loss=0.05821480756998062 +INFO:root: +[Valid][16][UTT][ MSE = 0.2211 | LCC = 0.9530 | SRCC = 0.9547 ] +INFO:root:Average loss=0.05257145991921425 +INFO:root: +[Valid][17][UTT][ MSE = 0.2146 | LCC = 0.9565 | SRCC = 0.9576 ] +INFO:root:Average loss=0.05016828042268753 +INFO:root: +[Valid][18][UTT][ MSE = 0.1880 | LCC = 0.9545 | SRCC = 0.9559 ] +INFO:root:Average loss=0.04996839943528175 +INFO:root: +[Valid][19][UTT][ MSE = 0.2100 | LCC = 0.9560 | SRCC = 0.9570 ] +INFO:root:Average loss=0.04590244869887829 +INFO:root: +[Valid][20][UTT][ MSE = 0.1963 | LCC = 0.9550 | SRCC = 0.9557 ] +INFO:root:Average loss=0.043880213513970374 +INFO:root: +[Valid][21][UTT][ MSE = 0.1960 | LCC = 0.9565 | SRCC = 0.9575 ] +INFO:root:Average loss=0.044458669245243075 +INFO:root: +[Valid][22][UTT][ MSE = 0.1904 | LCC = 0.9549 | SRCC = 0.9550 ] +INFO:root:Average loss=0.040101041689515116 +INFO:root: +[Valid][23][UTT][ MSE = 0.1962 | LCC = 0.9564 | SRCC = 0.9570 ] +INFO:root:Average loss=0.039858437940478324 +INFO:root: +[Valid][24][UTT][ MSE = 0.1905 | LCC = 0.9565 | SRCC = 0.9570 ] +INFO:root:Average loss=0.03798139175772667 +INFO:root: +[Valid][25][UTT][ MSE = 0.1919 | LCC = 0.9559 | SRCC = 0.9569 ] +INFO:root:Average loss=0.03707741491496563 +INFO:root: +[Valid][26][UTT][ MSE = 0.2017 | LCC = 0.9553 | SRCC = 0.9562 ] +INFO:root:Average loss=0.03443771904706955 +INFO:root: +[Valid][27][UTT][ MSE = 0.2013 | LCC = 0.9551 | SRCC = 0.9560 ] +INFO:root:Average loss=0.033179854705929754 +INFO:root: +[Valid][28][UTT][ MSE = 0.2120 | LCC = 0.9550 | SRCC = 0.9563 ] +INFO:root:Average loss=0.03424662382900715 +INFO:root: +[Valid][29][UTT][ MSE = 0.2303 | LCC = 0.9536 | SRCC = 0.9548 ] +INFO:root:Average loss=0.030952404245734214 +INFO:root: +[Valid][30][UTT][ MSE = 0.2139 | LCC = 0.9551 | SRCC = 0.9561 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1979 | LCC = 0.9560 | SRCC = 0.9564 ] diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run4/config.gin b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run4/model.pt b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..17de44c827cc32ba67bfaeb934e2747f814e8d0c --- /dev/null +++ b/results_tencent/w2v2_base_tencent_finetuned_2epoch/layer5/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e237f01e9c3de560f1b6e48c092fff1240ba6f739324744f50ac85db97708a9 +size 2747979