Upload SSL Layer MOS checkpoints (part 12)

results_nisqa/hubert_base/layer9/run3/config.gin

@@ -0,0 +1,31 @@
+# Training config on NISQA dataset.
+
+# ------------------------------
+# Architecture
+# ------------------------------
+ProjectionHead:
+    conv_channels = (32, 32)
+    dense_neurons = (64, 32, 1)
+    use_poolings = (True, True)
+    kernel_size = 5
+    pool_size = 5
+    
+# ------------------------------
+# Dataset
+# ------------------------------
+Nisqa:
+    debug = False
+
+get_dataloader:
+    num_workers = 12
+    shuffle = True
+
+# ------------------------------
+# Training loop
+# ------------------------------
+TrainingLoop:
+    num_epochs = 30
+    learning_rate = 1e-4
+    dataset_cls = @Nisqa
+    weight_decay = 0.0
+    batch_size_train = 64

results_nisqa/hubert_base/layer9/run3/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:99946f3c781459a45dd213381c1fb92adedfdc2cde260ae37c6f52de6519bf87
+size 2747979

results_nisqa/hubert_base/layer9/run3/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2dc4cbc7d4e8e9e44f7973e26815035a40618d182c14c6db179e1721481a307a
+size 2749702

results_nisqa/hubert_base/layer9/run3/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 11020
+INFO:root:Num val speech clips: 2700
+INFO:root:Num test speech clips: 232
+INFO:root:Device=cuda
+INFO:root:Average loss=0.44666784700286183
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3856 | LCC = 0.8532 | SRCC = 0.8520 ]
+INFO:root:Average loss=0.30138865096031586
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3495 | LCC = 0.8645 | SRCC = 0.8629 ]
+INFO:root:Average loss=0.26457147825660043
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3389 | LCC = 0.8724 | SRCC = 0.8709 ]
+INFO:root:Average loss=0.24420551628391177
+INFO:root:
+[Valid][3][UTT][ MSE = 0.3357 | LCC = 0.8708 | SRCC = 0.8717 ]
+INFO:root:Average loss=0.2327808791157827
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3565 | LCC = 0.8749 | SRCC = 0.8752 ]
+INFO:root:Average loss=0.21240624518408252
+INFO:root:
+[Valid][5][UTT][ MSE = 0.3229 | LCC = 0.8763 | SRCC = 0.8744 ]
+INFO:root:Average loss=0.19559336824051907
+INFO:root:
+[Valid][6][UTT][ MSE = 0.3183 | LCC = 0.8748 | SRCC = 0.8737 ]
+INFO:root:Average loss=0.1823456175840659
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2970 | LCC = 0.8776 | SRCC = 0.8769 ]
+INFO:root:Average loss=0.1677803102600781
+INFO:root:
+[Valid][8][UTT][ MSE = 0.3239 | LCC = 0.8738 | SRCC = 0.8750 ]
+INFO:root:Average loss=0.15418062049012654
+INFO:root:
+[Valid][9][UTT][ MSE = 0.3531 | LCC = 0.8663 | SRCC = 0.8711 ]
+INFO:root:Average loss=0.14771959900511483
+INFO:root:
+[Valid][10][UTT][ MSE = 0.3080 | LCC = 0.8739 | SRCC = 0.8726 ]
+INFO:root:Average loss=0.14206625426435746
+INFO:root:
+[Valid][11][UTT][ MSE = 0.3271 | LCC = 0.8717 | SRCC = 0.8704 ]
+INFO:root:Average loss=0.13147686520925148
+INFO:root:
+[Valid][12][UTT][ MSE = 0.3326 | LCC = 0.8695 | SRCC = 0.8689 ]
+INFO:root:Average loss=0.12326985070040461
+INFO:root:
+[Valid][13][UTT][ MSE = 0.3117 | LCC = 0.8730 | SRCC = 0.8709 ]
+INFO:root:Average loss=0.11464816917573785
+INFO:root:
+[Valid][14][UTT][ MSE = 0.3987 | LCC = 0.8627 | SRCC = 0.8672 ]
+INFO:root:Average loss=0.10780032465740436
+INFO:root:
+[Valid][15][UTT][ MSE = 0.3106 | LCC = 0.8680 | SRCC = 0.8684 ]
+INFO:root:Average loss=0.10193173742363218
+INFO:root:
+[Valid][16][UTT][ MSE = 0.3404 | LCC = 0.8684 | SRCC = 0.8689 ]
+INFO:root:Average loss=0.09824519912373124
+INFO:root:
+[Valid][17][UTT][ MSE = 0.3119 | LCC = 0.8721 | SRCC = 0.8712 ]
+INFO:root:Average loss=0.0920214000335663
+INFO:root:
+[Valid][18][UTT][ MSE = 0.3173 | LCC = 0.8688 | SRCC = 0.8674 ]
+INFO:root:Average loss=0.08671017301220425
+INFO:root:
+[Valid][19][UTT][ MSE = 0.3322 | LCC = 0.8678 | SRCC = 0.8673 ]
+INFO:root:Average loss=0.08271780209561994
+INFO:root:
+[Valid][20][UTT][ MSE = 0.3579 | LCC = 0.8651 | SRCC = 0.8651 ]
+INFO:root:Average loss=0.08045609058500025
+INFO:root:
+[Valid][21][UTT][ MSE = 0.3408 | LCC = 0.8673 | SRCC = 0.8675 ]
+INFO:root:Average loss=0.07757445007045834
+INFO:root:
+[Valid][22][UTT][ MSE = 0.3567 | LCC = 0.8613 | SRCC = 0.8625 ]
+INFO:root:Average loss=0.07378173908057241
+INFO:root:
+[Valid][23][UTT][ MSE = 0.3234 | LCC = 0.8643 | SRCC = 0.8635 ]
+INFO:root:Average loss=0.07183158782046999
+INFO:root:
+[Valid][24][UTT][ MSE = 0.3241 | LCC = 0.8666 | SRCC = 0.8649 ]
+INFO:root:Average loss=0.0713757578218948
+INFO:root:
+[Valid][25][UTT][ MSE = 0.3464 | LCC = 0.8641 | SRCC = 0.8643 ]
+INFO:root:Average loss=0.06711374490247296
+INFO:root:
+[Valid][26][UTT][ MSE = 0.3321 | LCC = 0.8679 | SRCC = 0.8663 ]
+INFO:root:Average loss=0.06465638179138217
+INFO:root:
+[Valid][27][UTT][ MSE = 0.3475 | LCC = 0.8650 | SRCC = 0.8643 ]
+INFO:root:Average loss=0.06330957657770614
+INFO:root:
+[Valid][28][UTT][ MSE = 0.3680 | LCC = 0.8605 | SRCC = 0.8622 ]
+INFO:root:Average loss=0.05964362851715501
+INFO:root:
+[Valid][29][UTT][ MSE = 0.3249 | LCC = 0.8662 | SRCC = 0.8655 ]
+INFO:root:Average loss=0.05948429367359663
+INFO:root:
+[Valid][30][UTT][ MSE = 0.3555 | LCC = 0.8626 | SRCC = 0.8626 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.5098 | LCC = 0.7805 | SRCC = 0.7761 ]

results_nisqa/hubert_base/layer9/run4/config.gin

@@ -0,0 +1,31 @@
+# Training config on NISQA dataset.
+
+# ------------------------------
+# Architecture
+# ------------------------------
+ProjectionHead:
+    conv_channels = (32, 32)
+    dense_neurons = (64, 32, 1)
+    use_poolings = (True, True)
+    kernel_size = 5
+    pool_size = 5
+    
+# ------------------------------
+# Dataset
+# ------------------------------
+Nisqa:
+    debug = False
+
+get_dataloader:
+    num_workers = 12
+    shuffle = True
+
+# ------------------------------
+# Training loop
+# ------------------------------
+TrainingLoop:
+    num_epochs = 30
+    learning_rate = 1e-4
+    dataset_cls = @Nisqa
+    weight_decay = 0.0
+    batch_size_train = 64

results_nisqa/hubert_base/layer9/run4/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a23aac372c23ba27e7e3e7ff74eb1ed54003098a6f74908d1edf3bd2b7f3eb70
+size 2747979

results_nisqa/hubert_base/layer9/run4/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1af5e330ce1c397ad5a6df04f94a8619a34a9aa6d9c6794e12e7af0c18a8fb4
+size 2749702

results_nisqa/hubert_base/layer9/run4/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 11020
+INFO:root:Num val speech clips: 2700
+INFO:root:Num test speech clips: 232
+INFO:root:Device=cuda
+INFO:root:Average loss=0.4937498324868307
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3827 | LCC = 0.8519 | SRCC = 0.8503 ]
+INFO:root:Average loss=0.30047601375276645
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3292 | LCC = 0.8628 | SRCC = 0.8617 ]
+INFO:root:Average loss=0.2666946215436638
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3484 | LCC = 0.8684 | SRCC = 0.8683 ]
+INFO:root:Average loss=0.2429341184179907
+INFO:root:
+[Valid][3][UTT][ MSE = 0.3177 | LCC = 0.8739 | SRCC = 0.8721 ]
+INFO:root:Average loss=0.22515277977968226
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3360 | LCC = 0.8740 | SRCC = 0.8720 ]
+INFO:root:Average loss=0.20786380741982102
+INFO:root:
+[Valid][5][UTT][ MSE = 0.3156 | LCC = 0.8729 | SRCC = 0.8713 ]
+INFO:root:Average loss=0.1959835817617488
+INFO:root:
+[Valid][6][UTT][ MSE = 0.3137 | LCC = 0.8744 | SRCC = 0.8728 ]
+INFO:root:Average loss=0.18504263917145702
+INFO:root:
+[Valid][7][UTT][ MSE = 0.3269 | LCC = 0.8738 | SRCC = 0.8718 ]
+INFO:root:Average loss=0.167429916474516
+INFO:root:
+[Valid][8][UTT][ MSE = 0.3053 | LCC = 0.8739 | SRCC = 0.8722 ]
+INFO:root:Average loss=0.15698515610887825
+INFO:root:
+[Valid][9][UTT][ MSE = 0.3180 | LCC = 0.8728 | SRCC = 0.8704 ]
+INFO:root:Average loss=0.15077230657283971
+INFO:root:
+[Valid][10][UTT][ MSE = 0.3118 | LCC = 0.8709 | SRCC = 0.8689 ]
+INFO:root:Average loss=0.1383441848561943
+INFO:root:
+[Valid][11][UTT][ MSE = 0.3270 | LCC = 0.8726 | SRCC = 0.8717 ]
+INFO:root:Average loss=0.1290830308732959
+INFO:root:
+[Valid][12][UTT][ MSE = 0.3139 | LCC = 0.8713 | SRCC = 0.8696 ]
+INFO:root:Average loss=0.12285170363897534
+INFO:root:
+[Valid][13][UTT][ MSE = 0.3031 | LCC = 0.8697 | SRCC = 0.8683 ]
+INFO:root:Average loss=0.11503355820916292
+INFO:root:
+[Valid][14][UTT][ MSE = 0.3127 | LCC = 0.8699 | SRCC = 0.8676 ]
+INFO:root:Average loss=0.11209239009660103
+INFO:root:
+[Valid][15][UTT][ MSE = 0.3047 | LCC = 0.8709 | SRCC = 0.8689 ]
+INFO:root:Average loss=0.10471336519873212
+INFO:root:
+[Valid][16][UTT][ MSE = 0.3194 | LCC = 0.8702 | SRCC = 0.8670 ]
+INFO:root:Average loss=0.0993591742536236
+INFO:root:
+[Valid][17][UTT][ MSE = 0.3355 | LCC = 0.8644 | SRCC = 0.8635 ]
+INFO:root:Average loss=0.09705279147349341
+INFO:root:
+[Valid][18][UTT][ MSE = 0.3222 | LCC = 0.8670 | SRCC = 0.8641 ]
+INFO:root:Average loss=0.09146756687433044
+INFO:root:
+[Valid][19][UTT][ MSE = 0.3301 | LCC = 0.8649 | SRCC = 0.8631 ]
+INFO:root:Average loss=0.08754754488523296
+INFO:root:
+[Valid][20][UTT][ MSE = 0.3539 | LCC = 0.8606 | SRCC = 0.8599 ]
+INFO:root:Average loss=0.08255260767189064
+INFO:root:
+[Valid][21][UTT][ MSE = 0.3192 | LCC = 0.8654 | SRCC = 0.8637 ]
+INFO:root:Average loss=0.07871194203206569
+INFO:root:
+[Valid][22][UTT][ MSE = 0.3087 | LCC = 0.8687 | SRCC = 0.8663 ]
+INFO:root:Average loss=0.07684320067136274
+INFO:root:
+[Valid][23][UTT][ MSE = 0.3085 | LCC = 0.8667 | SRCC = 0.8650 ]
+INFO:root:Average loss=0.07221316186592758
+INFO:root:
+[Valid][24][UTT][ MSE = 0.3467 | LCC = 0.8621 | SRCC = 0.8613 ]
+INFO:root:Average loss=0.07282608355573147
+INFO:root:
+[Valid][25][UTT][ MSE = 0.3138 | LCC = 0.8663 | SRCC = 0.8638 ]
+INFO:root:Average loss=0.06812817032712733
+INFO:root:
+[Valid][26][UTT][ MSE = 0.3175 | LCC = 0.8656 | SRCC = 0.8630 ]
+INFO:root:Average loss=0.06732187889559421
+INFO:root:
+[Valid][27][UTT][ MSE = 0.3154 | LCC = 0.8667 | SRCC = 0.8639 ]
+INFO:root:Average loss=0.06591674758692008
+INFO:root:
+[Valid][28][UTT][ MSE = 0.3198 | LCC = 0.8652 | SRCC = 0.8644 ]
+INFO:root:Average loss=0.06197786611111867
+INFO:root:
+[Valid][29][UTT][ MSE = 0.3194 | LCC = 0.8660 | SRCC = 0.8634 ]
+INFO:root:Average loss=0.05978712813142752
+INFO:root:
+[Valid][30][UTT][ MSE = 0.3311 | LCC = 0.8634 | SRCC = 0.8610 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4424 | LCC = 0.7892 | SRCC = 0.7874 ]

results_nisqa/w2v2_base/layer0/run0/config.gin

@@ -0,0 +1,31 @@
+# Training config on NISQA dataset.
+
+# ------------------------------
+# Architecture
+# ------------------------------
+ProjectionHead:
+    conv_channels = (32, 32)
+    dense_neurons = (64, 32, 1)
+    use_poolings = (True, True)
+    kernel_size = 5
+    pool_size = 5
+    
+# ------------------------------
+# Dataset
+# ------------------------------
+Nisqa:
+    debug = False
+
+get_dataloader:
+    num_workers = 12
+    shuffle = True
+
+# ------------------------------
+# Training loop
+# ------------------------------
+TrainingLoop:
+    num_epochs = 30
+    learning_rate = 1e-4
+    dataset_cls = @Nisqa
+    weight_decay = 0.0
+    batch_size_train = 64

results_nisqa/w2v2_base/layer0/run0/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61d0e51fe7405ae014c33f344bf46308be40fde85c4570b7665356e2700daa54
+size 2747979

results_nisqa/w2v2_base/layer0/run0/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0855f198e5a69696cfab0693ca92b1be77b37807b4ada06e586767a0438c169
+size 2749702

results_nisqa/w2v2_base/layer0/run0/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 11020
+INFO:root:Num val speech clips: 2700
+INFO:root:Num test speech clips: 232
+INFO:root:Device=cuda
+INFO:root:Average loss=0.4656045253393967
+INFO:root:
+[Valid][0][UTT][ MSE = 0.4296 | LCC = 0.8454 | SRCC = 0.8440 ]
+INFO:root:Average loss=0.2881045989218475
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3587 | LCC = 0.8737 | SRCC = 0.8751 ]
+INFO:root:Average loss=0.24932437583890266
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3069 | LCC = 0.8864 | SRCC = 0.8859 ]
+INFO:root:Average loss=0.22377282561938885
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2915 | LCC = 0.8930 | SRCC = 0.8933 ]
+INFO:root:Average loss=0.20796405993445072
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2868 | LCC = 0.8979 | SRCC = 0.8974 ]
+INFO:root:Average loss=0.19377853071069442
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2831 | LCC = 0.9011 | SRCC = 0.9010 ]
+INFO:root:Average loss=0.1860185399909929
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2611 | LCC = 0.9041 | SRCC = 0.9044 ]
+INFO:root:Average loss=0.1798827164521107
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2870 | LCC = 0.9035 | SRCC = 0.9036 ]
+INFO:root:Average loss=0.17635803010766907
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2439 | LCC = 0.9045 | SRCC = 0.9053 ]
+INFO:root:Average loss=0.16112533646684163
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2626 | LCC = 0.9048 | SRCC = 0.9053 ]
+INFO:root:Average loss=0.15658956447433184
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2627 | LCC = 0.9064 | SRCC = 0.9068 ]
+INFO:root:Average loss=0.15010529110541923
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2559 | LCC = 0.9044 | SRCC = 0.9052 ]
+INFO:root:Average loss=0.14273553862737093
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2786 | LCC = 0.9052 | SRCC = 0.9054 ]
+INFO:root:Average loss=0.13428944495716535
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2431 | LCC = 0.9051 | SRCC = 0.9052 ]
+INFO:root:Average loss=0.13038814446829647
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2518 | LCC = 0.9060 | SRCC = 0.9062 ]
+INFO:root:Average loss=0.1292005862416215
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2544 | LCC = 0.9054 | SRCC = 0.9053 ]
+INFO:root:Average loss=0.12292390717247317
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2522 | LCC = 0.9048 | SRCC = 0.9050 ]
+INFO:root:Average loss=0.11747096125790149
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2617 | LCC = 0.9039 | SRCC = 0.9037 ]
+INFO:root:Average loss=0.1176645015447126
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2469 | LCC = 0.9032 | SRCC = 0.9036 ]
+INFO:root:Average loss=0.11091964119087065
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2576 | LCC = 0.9034 | SRCC = 0.9035 ]
+INFO:root:Average loss=0.10674060846684295
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2672 | LCC = 0.9027 | SRCC = 0.9027 ]
+INFO:root:Average loss=0.10354766060162142
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2502 | LCC = 0.9047 | SRCC = 0.9043 ]
+INFO:root:Average loss=0.10074021156898813
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2797 | LCC = 0.9037 | SRCC = 0.9040 ]
+INFO:root:Average loss=0.10130047016491779
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2531 | LCC = 0.9029 | SRCC = 0.9032 ]
+INFO:root:Average loss=0.09709900753260348
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2536 | LCC = 0.9035 | SRCC = 0.9034 ]
+INFO:root:Average loss=0.09214405493670805
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2509 | LCC = 0.9039 | SRCC = 0.9042 ]
+INFO:root:Average loss=0.08782778101074214
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2559 | LCC = 0.9041 | SRCC = 0.9040 ]
+INFO:root:Average loss=0.08744030854950062
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2558 | LCC = 0.9050 | SRCC = 0.9047 ]
+INFO:root:Average loss=0.08482128924372569
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2646 | LCC = 0.9017 | SRCC = 0.9022 ]
+INFO:root:Average loss=0.08370480800553554
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2535 | LCC = 0.9027 | SRCC = 0.9025 ]
+INFO:root:Average loss=0.07968193238784123
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2604 | LCC = 0.9020 | SRCC = 0.9021 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4201 | LCC = 0.7277 | SRCC = 0.6993 ]

results_nisqa/w2v2_base/layer0/run1/config.gin

@@ -0,0 +1,31 @@
+# Training config on NISQA dataset.
+
+# ------------------------------
+# Architecture
+# ------------------------------
+ProjectionHead:
+    conv_channels = (32, 32)
+    dense_neurons = (64, 32, 1)
+    use_poolings = (True, True)
+    kernel_size = 5
+    pool_size = 5
+    
+# ------------------------------
+# Dataset
+# ------------------------------
+Nisqa:
+    debug = False
+
+get_dataloader:
+    num_workers = 12
+    shuffle = True
+
+# ------------------------------
+# Training loop
+# ------------------------------
+TrainingLoop:
+    num_epochs = 30
+    learning_rate = 1e-4
+    dataset_cls = @Nisqa
+    weight_decay = 0.0
+    batch_size_train = 64

results_nisqa/w2v2_base/layer0/run1/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5774e1532747b2f9859fcab7ad3bcd65f0f489258882a7ccf3e213c09a39a92
+size 2747979

results_nisqa/w2v2_base/layer0/run1/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a63c4f5f2d9f76158709f6571d63f667023232d7f0cf4b5b0f5d1fd74fd3c4f4
+size 2749702

results_nisqa/w2v2_base/layer0/run1/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 11020
+INFO:root:Num val speech clips: 2700
+INFO:root:Num test speech clips: 232
+INFO:root:Device=cuda
+INFO:root:Average loss=0.4732976847818132
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3855 | LCC = 0.8543 | SRCC = 0.8531 ]
+INFO:root:Average loss=0.28063617250478334
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3462 | LCC = 0.8774 | SRCC = 0.8777 ]
+INFO:root:Average loss=0.2405943265023259
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3277 | LCC = 0.8921 | SRCC = 0.8921 ]
+INFO:root:Average loss=0.2183921815170718
+INFO:root:
+[Valid][3][UTT][ MSE = 0.3101 | LCC = 0.8943 | SRCC = 0.8946 ]
+INFO:root:Average loss=0.2088140014537497
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2758 | LCC = 0.8972 | SRCC = 0.8976 ]
+INFO:root:Average loss=0.19720859295887755
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2873 | LCC = 0.9004 | SRCC = 0.8999 ]
+INFO:root:Average loss=0.1835478444319929
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2769 | LCC = 0.9012 | SRCC = 0.9010 ]
+INFO:root:Average loss=0.17801272899256965
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2587 | LCC = 0.9032 | SRCC = 0.9031 ]
+INFO:root:Average loss=0.16980918285819147
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2702 | LCC = 0.9031 | SRCC = 0.9036 ]
+INFO:root:Average loss=0.1649634395376106
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2876 | LCC = 0.8990 | SRCC = 0.8990 ]
+INFO:root:Average loss=0.15810617274803923
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2770 | LCC = 0.9027 | SRCC = 0.9025 ]
+INFO:root:Average loss=0.1529789700501227
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2860 | LCC = 0.9016 | SRCC = 0.9014 ]
+INFO:root:Average loss=0.14302995684691247
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2550 | LCC = 0.9028 | SRCC = 0.9037 ]
+INFO:root:Average loss=0.1361115965023206
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2655 | LCC = 0.9036 | SRCC = 0.9035 ]
+INFO:root:Average loss=0.1316482866602826
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2738 | LCC = 0.9022 | SRCC = 0.9029 ]
+INFO:root:Average loss=0.12706508108473927
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2601 | LCC = 0.9027 | SRCC = 0.9028 ]
+INFO:root:Average loss=0.1263645576431572
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2417 | LCC = 0.9041 | SRCC = 0.9045 ]
+INFO:root:Average loss=0.11896640583442125
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2541 | LCC = 0.9014 | SRCC = 0.9017 ]
+INFO:root:Average loss=0.11688587771502533
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2544 | LCC = 0.9034 | SRCC = 0.9034 ]
+INFO:root:Average loss=0.1115981881975094
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2743 | LCC = 0.9014 | SRCC = 0.9012 ]
+INFO:root:Average loss=0.11078533978131465
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2618 | LCC = 0.9024 | SRCC = 0.9020 ]
+INFO:root:Average loss=0.10573375526535718
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2542 | LCC = 0.9046 | SRCC = 0.9054 ]
+INFO:root:Average loss=0.10360300489243744
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2642 | LCC = 0.9012 | SRCC = 0.9027 ]
+INFO:root:Average loss=0.09702793392487344
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2462 | LCC = 0.9049 | SRCC = 0.9052 ]
+INFO:root:Average loss=0.09721219438435025
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2551 | LCC = 0.9042 | SRCC = 0.9041 ]
+INFO:root:Average loss=0.09635214641087317
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2666 | LCC = 0.9003 | SRCC = 0.9004 ]
+INFO:root:Average loss=0.08935522841494208
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2550 | LCC = 0.9043 | SRCC = 0.9044 ]
+INFO:root:Average loss=0.08954595382503003
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2555 | LCC = 0.9029 | SRCC = 0.9027 ]
+INFO:root:Average loss=0.08561355422514712
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2596 | LCC = 0.9024 | SRCC = 0.9025 ]
+INFO:root:Average loss=0.08572480493801178
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2374 | LCC = 0.9036 | SRCC = 0.9040 ]
+INFO:root:Average loss=0.0844722409985658
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2662 | LCC = 0.9017 | SRCC = 0.9016 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4285 | LCC = 0.7448 | SRCC = 0.7164 ]

results_nisqa/w2v2_base/layer0/run2/config.gin

@@ -0,0 +1,31 @@
+# Training config on NISQA dataset.
+
+# ------------------------------
+# Architecture
+# ------------------------------
+ProjectionHead:
+    conv_channels = (32, 32)
+    dense_neurons = (64, 32, 1)
+    use_poolings = (True, True)
+    kernel_size = 5
+    pool_size = 5
+    
+# ------------------------------
+# Dataset
+# ------------------------------
+Nisqa:
+    debug = False
+
+get_dataloader:
+    num_workers = 12
+    shuffle = True
+
+# ------------------------------
+# Training loop
+# ------------------------------
+TrainingLoop:
+    num_epochs = 30
+    learning_rate = 1e-4
+    dataset_cls = @Nisqa
+    weight_decay = 0.0
+    batch_size_train = 64

results_nisqa/w2v2_base/layer0/run2/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:81b38ea2f7e8707f5ffe3034e2be74e285a6a3d9581027c712f406db0ed97beb
+size 2747979

results_nisqa/w2v2_base/layer0/run2/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9016a0583b9ddcd5a25c8a7c0519d25fc23556f87365ef6be0c29eb620c6410a
+size 2749702

results_nisqa/w2v2_base/layer0/run2/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 11020
+INFO:root:Num val speech clips: 2700
+INFO:root:Num test speech clips: 232
+INFO:root:Device=cuda
+INFO:root:Average loss=0.4887408753695516
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3927 | LCC = 0.8500 | SRCC = 0.8480 ]
+INFO:root:Average loss=0.28332467325505495
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3194 | LCC = 0.8769 | SRCC = 0.8772 ]
+INFO:root:Average loss=0.24259669906486667
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3202 | LCC = 0.8909 | SRCC = 0.8914 ]
+INFO:root:Average loss=0.21941497146738748
+INFO:root:
+[Valid][3][UTT][ MSE = 0.3445 | LCC = 0.8943 | SRCC = 0.8966 ]
+INFO:root:Average loss=0.20445749908685684
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2714 | LCC = 0.9012 | SRCC = 0.9011 ]
+INFO:root:Average loss=0.19050900078233265
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2769 | LCC = 0.9020 | SRCC = 0.9034 ]
+INFO:root:Average loss=0.1808594342945628
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2719 | LCC = 0.9028 | SRCC = 0.9043 ]
+INFO:root:Average loss=0.17513246985012396
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2617 | LCC = 0.9045 | SRCC = 0.9048 ]
+INFO:root:Average loss=0.1673779837313415
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2725 | LCC = 0.9050 | SRCC = 0.9054 ]
+INFO:root:Average loss=0.16050715847856048
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2628 | LCC = 0.9043 | SRCC = 0.9053 ]
+INFO:root:Average loss=0.15591477034236653
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2924 | LCC = 0.9047 | SRCC = 0.9068 ]
+INFO:root:Average loss=0.14948185873513967
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2852 | LCC = 0.9053 | SRCC = 0.9061 ]
+INFO:root:Average loss=0.14087455228746282
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2531 | LCC = 0.9055 | SRCC = 0.9058 ]
+INFO:root:Average loss=0.135630670956449
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2665 | LCC = 0.9040 | SRCC = 0.9041 ]
+INFO:root:Average loss=0.1318436085402621
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2614 | LCC = 0.9041 | SRCC = 0.9039 ]
+INFO:root:Average loss=0.1259858227534101
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2538 | LCC = 0.9042 | SRCC = 0.9043 ]
+INFO:root:Average loss=0.12247318236124997
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2388 | LCC = 0.9034 | SRCC = 0.9037 ]
+INFO:root:Average loss=0.11649351623016975
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2669 | LCC = 0.9029 | SRCC = 0.9043 ]
+INFO:root:Average loss=0.11803678797848652
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2443 | LCC = 0.9063 | SRCC = 0.9065 ]
+INFO:root:Average loss=0.11060947571703464
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2447 | LCC = 0.9049 | SRCC = 0.9053 ]
+INFO:root:Average loss=0.10761247545136193
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2664 | LCC = 0.9024 | SRCC = 0.9029 ]
+INFO:root:Average loss=0.10479283677360227
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2520 | LCC = 0.9041 | SRCC = 0.9044 ]
+INFO:root:Average loss=0.10309677463390923
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2890 | LCC = 0.9011 | SRCC = 0.9019 ]
+INFO:root:Average loss=0.09764214361764792
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2806 | LCC = 0.9034 | SRCC = 0.9035 ]
+INFO:root:Average loss=0.09634519002341121
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2606 | LCC = 0.9007 | SRCC = 0.9006 ]
+INFO:root:Average loss=0.09268231267873951
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2639 | LCC = 0.9012 | SRCC = 0.9023 ]
+INFO:root:Average loss=0.08857534429586003
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2776 | LCC = 0.8987 | SRCC = 0.9000 ]
+INFO:root:Average loss=0.0879936558275209
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2583 | LCC = 0.9019 | SRCC = 0.9025 ]
+INFO:root:Average loss=0.08507861446783033
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2648 | LCC = 0.8995 | SRCC = 0.9000 ]
+INFO:root:Average loss=0.08290507116844889
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2548 | LCC = 0.9017 | SRCC = 0.9033 ]
+INFO:root:Average loss=0.08242985438232477
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2711 | LCC = 0.8992 | SRCC = 0.9010 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4162 | LCC = 0.7503 | SRCC = 0.7217 ]

results_nisqa/w2v2_base/layer0/run3/config.gin

@@ -0,0 +1,31 @@
+# Training config on NISQA dataset.
+
+# ------------------------------
+# Architecture
+# ------------------------------
+ProjectionHead:
+    conv_channels = (32, 32)
+    dense_neurons = (64, 32, 1)
+    use_poolings = (True, True)
+    kernel_size = 5
+    pool_size = 5
+    
+# ------------------------------
+# Dataset
+# ------------------------------
+Nisqa:
+    debug = False
+
+get_dataloader:
+    num_workers = 12
+    shuffle = True
+
+# ------------------------------
+# Training loop
+# ------------------------------
+TrainingLoop:
+    num_epochs = 30
+    learning_rate = 1e-4
+    dataset_cls = @Nisqa
+    weight_decay = 0.0
+    batch_size_train = 64

results_nisqa/w2v2_base/layer0/run3/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e814eb9a3191a178d427139e62537e35c048426f956ebc57bf0fff038f5208a
+size 2747979

results_nisqa/w2v2_base/layer0/run3/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:43bfbb1a7e57171f024754192acde8a3bf23ae52546b0ec78fefc252cec7f87b
+size 2749702

results_nisqa/w2v2_base/layer0/run3/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 11020
+INFO:root:Num val speech clips: 2700
+INFO:root:Num test speech clips: 232
+INFO:root:Device=cuda
+INFO:root:Average loss=0.5220217747839889
+INFO:root:
+[Valid][0][UTT][ MSE = 0.4032 | LCC = 0.8419 | SRCC = 0.8411 ]
+INFO:root:Average loss=0.28966531148880204
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3432 | LCC = 0.8742 | SRCC = 0.8737 ]
+INFO:root:Average loss=0.250793971939583
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3026 | LCC = 0.8901 | SRCC = 0.8903 ]
+INFO:root:Average loss=0.2198167531132009
+INFO:root:
+[Valid][3][UTT][ MSE = 0.3049 | LCC = 0.8950 | SRCC = 0.8955 ]
+INFO:root:Average loss=0.2081440923186396
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3326 | LCC = 0.8971 | SRCC = 0.8968 ]
+INFO:root:Average loss=0.2001598441187357
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2656 | LCC = 0.9010 | SRCC = 0.9015 ]
+INFO:root:Average loss=0.1877970003840551
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2857 | LCC = 0.9042 | SRCC = 0.9042 ]
+INFO:root:Average loss=0.1772624268869444
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2674 | LCC = 0.9045 | SRCC = 0.9046 ]
+INFO:root:Average loss=0.16934027436668472
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2725 | LCC = 0.9063 | SRCC = 0.9064 ]
+INFO:root:Average loss=0.15953071996828036
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2621 | LCC = 0.9062 | SRCC = 0.9066 ]
+INFO:root:Average loss=0.15449726000169797
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2561 | LCC = 0.9066 | SRCC = 0.9069 ]
+INFO:root:Average loss=0.14500464847838948
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2649 | LCC = 0.9080 | SRCC = 0.9083 ]
+INFO:root:Average loss=0.14202140465464896
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2754 | LCC = 0.9039 | SRCC = 0.9048 ]
+INFO:root:Average loss=0.13688732076414747
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2677 | LCC = 0.9058 | SRCC = 0.9068 ]
+INFO:root:Average loss=0.13234757808591588
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2486 | LCC = 0.9042 | SRCC = 0.9048 ]
+INFO:root:Average loss=0.12894457259963701
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2687 | LCC = 0.9057 | SRCC = 0.9064 ]
+INFO:root:Average loss=0.12374647375131619
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2413 | LCC = 0.9051 | SRCC = 0.9056 ]
+INFO:root:Average loss=0.11876645692855636
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2600 | LCC = 0.9045 | SRCC = 0.9046 ]
+INFO:root:Average loss=0.11688838709641054
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2567 | LCC = 0.9055 | SRCC = 0.9059 ]
+INFO:root:Average loss=0.11554037413500637
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2588 | LCC = 0.9049 | SRCC = 0.9050 ]
+INFO:root:Average loss=0.10992274030698518
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2398 | LCC = 0.9062 | SRCC = 0.9063 ]
+INFO:root:Average loss=0.10452693054786308
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2716 | LCC = 0.9054 | SRCC = 0.9054 ]
+INFO:root:Average loss=0.10080644931924136
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2677 | LCC = 0.9031 | SRCC = 0.9034 ]
+INFO:root:Average loss=0.09887692747863731
+INFO:root:
+[Valid][23][UTT][ MSE = 0.3013 | LCC = 0.9033 | SRCC = 0.9052 ]
+INFO:root:Average loss=0.09655310990924092
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2610 | LCC = 0.9043 | SRCC = 0.9049 ]
+INFO:root:Average loss=0.09397958337485446
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2477 | LCC = 0.9031 | SRCC = 0.9045 ]
+INFO:root:Average loss=0.09055395103517296
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2443 | LCC = 0.9050 | SRCC = 0.9054 ]
+INFO:root:Average loss=0.08929413292190932
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2532 | LCC = 0.9037 | SRCC = 0.9050 ]
+INFO:root:Average loss=0.08656429417560556
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2522 | LCC = 0.9039 | SRCC = 0.9040 ]
+INFO:root:Average loss=0.08417933042338818
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2467 | LCC = 0.9048 | SRCC = 0.9047 ]
+INFO:root:Average loss=0.08278639220519562
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2524 | LCC = 0.9023 | SRCC = 0.9033 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4541 | LCC = 0.7437 | SRCC = 0.7166 ]

results_nisqa/w2v2_base/layer0/run4/config.gin

@@ -0,0 +1,31 @@
+# Training config on NISQA dataset.
+
+# ------------------------------
+# Architecture
+# ------------------------------
+ProjectionHead:
+    conv_channels = (32, 32)
+    dense_neurons = (64, 32, 1)
+    use_poolings = (True, True)
+    kernel_size = 5
+    pool_size = 5
+    
+# ------------------------------
+# Dataset
+# ------------------------------
+Nisqa:
+    debug = False
+
+get_dataloader:
+    num_workers = 12
+    shuffle = True
+
+# ------------------------------
+# Training loop
+# ------------------------------
+TrainingLoop:
+    num_epochs = 30
+    learning_rate = 1e-4
+    dataset_cls = @Nisqa
+    weight_decay = 0.0
+    batch_size_train = 64

results_nisqa/w2v2_base/layer0/run4/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a763bc68b50bffcaf52f08e34c84bb7492ad2d797b0e90d79b6042701d53387b
+size 2747979

results_nisqa/w2v2_base/layer0/run4/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fdb2a49fae8ab66beedb69826f383b22aba9c354a6a60eb352e410d9f86b718
+size 2749702

results_nisqa/w2v2_base/layer0/run4/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 11020
+INFO:root:Num val speech clips: 2700
+INFO:root:Num test speech clips: 232
+INFO:root:Device=cuda
+INFO:root:Average loss=0.513914836796722
+INFO:root:
+[Valid][0][UTT][ MSE = 0.4477 | LCC = 0.8500 | SRCC = 0.8489 ]
+INFO:root:Average loss=0.29364840453759783
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3454 | LCC = 0.8797 | SRCC = 0.8799 ]
+INFO:root:Average loss=0.24287864600302855
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3200 | LCC = 0.8918 | SRCC = 0.8915 ]
+INFO:root:Average loss=0.22607805127697872
+INFO:root:
+[Valid][3][UTT][ MSE = 0.3000 | LCC = 0.8966 | SRCC = 0.8964 ]
+INFO:root:Average loss=0.20207453243477497
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3050 | LCC = 0.8992 | SRCC = 0.8990 ]
+INFO:root:Average loss=0.1935510774051523
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2584 | LCC = 0.9026 | SRCC = 0.9026 ]
+INFO:root:Average loss=0.1800973029925644
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2706 | LCC = 0.9055 | SRCC = 0.9056 ]
+INFO:root:Average loss=0.17491466491725405
+INFO:root:
+[Valid][7][UTT][ MSE = 0.3130 | LCC = 0.9052 | SRCC = 0.9054 ]
+INFO:root:Average loss=0.1673860390148411
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2746 | LCC = 0.9055 | SRCC = 0.9060 ]
+INFO:root:Average loss=0.15654053289249453
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2566 | LCC = 0.9056 | SRCC = 0.9064 ]
+INFO:root:Average loss=0.14868969513329466
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2714 | LCC = 0.9038 | SRCC = 0.9046 ]
+INFO:root:Average loss=0.14407660549431178
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2805 | LCC = 0.9065 | SRCC = 0.9066 ]
+INFO:root:Average loss=0.14161831123291413
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2564 | LCC = 0.9047 | SRCC = 0.9065 ]
+INFO:root:Average loss=0.13910104803784046
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2765 | LCC = 0.9054 | SRCC = 0.9054 ]
+INFO:root:Average loss=0.13281678834747027
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2669 | LCC = 0.9059 | SRCC = 0.9064 ]
+INFO:root:Average loss=0.12765254011388458
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2558 | LCC = 0.9047 | SRCC = 0.9058 ]
+INFO:root:Average loss=0.12245727308912774
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2426 | LCC = 0.9051 | SRCC = 0.9061 ]
+INFO:root:Average loss=0.11970482650347528
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2766 | LCC = 0.9060 | SRCC = 0.9063 ]
+INFO:root:Average loss=0.11510303894305504
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2615 | LCC = 0.9059 | SRCC = 0.9059 ]
+INFO:root:Average loss=0.10870681621263482
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2746 | LCC = 0.8996 | SRCC = 0.9022 ]
+INFO:root:Average loss=0.10789603044445804
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2677 | LCC = 0.9040 | SRCC = 0.9041 ]
+INFO:root:Average loss=0.10374947996928513
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2868 | LCC = 0.9054 | SRCC = 0.9053 ]
+INFO:root:Average loss=0.09831848436955772
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2843 | LCC = 0.9054 | SRCC = 0.9057 ]
+INFO:root:Average loss=0.09606771358434175
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2845 | LCC = 0.9017 | SRCC = 0.9021 ]
+INFO:root:Average loss=0.09312217320837726
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2564 | LCC = 0.9028 | SRCC = 0.9042 ]
+INFO:root:Average loss=0.09122585636601283
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2746 | LCC = 0.9039 | SRCC = 0.9042 ]
+INFO:root:Average loss=0.08769452315620604
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2721 | LCC = 0.9046 | SRCC = 0.9046 ]
+INFO:root:Average loss=0.08359550333694915
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2698 | LCC = 0.9014 | SRCC = 0.9018 ]
+INFO:root:Average loss=0.08518337271030928
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2908 | LCC = 0.9028 | SRCC = 0.9025 ]
+INFO:root:Average loss=0.08358871803290582
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2914 | LCC = 0.9020 | SRCC = 0.9024 ]
+INFO:root:Average loss=0.08230808368652542
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2872 | LCC = 0.8990 | SRCC = 0.8993 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.3905 | LCC = 0.7612 | SRCC = 0.7375 ]

results_nisqa/w2v2_base/layer1/run0/config.gin

@@ -0,0 +1,31 @@
+# Training config on NISQA dataset.
+
+# ------------------------------
+# Architecture
+# ------------------------------
+ProjectionHead:
+    conv_channels = (32, 32)
+    dense_neurons = (64, 32, 1)
+    use_poolings = (True, True)
+    kernel_size = 5
+    pool_size = 5
+    
+# ------------------------------
+# Dataset
+# ------------------------------
+Nisqa:
+    debug = False
+
+get_dataloader:
+    num_workers = 12
+    shuffle = True
+
+# ------------------------------
+# Training loop
+# ------------------------------
+TrainingLoop:
+    num_epochs = 30
+    learning_rate = 1e-4
+    dataset_cls = @Nisqa
+    weight_decay = 0.0
+    batch_size_train = 64

results_nisqa/w2v2_base/layer1/run0/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b15991197b2e30bf79784b99670ca8795c98e8de5716ccbc374e6db20567e374
+size 2747979

results_nisqa/w2v2_base/layer1/run0/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65605e5b0f3dc77683e5e358674fd49ede74426f9dfc1eb214ea4406ac094973
+size 2749702

results_nisqa/w2v2_base/layer1/run0/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 11020
+INFO:root:Num val speech clips: 2700
+INFO:root:Num test speech clips: 232
+INFO:root:Device=cuda
+INFO:root:Average loss=0.4140820809010136
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3245 | LCC = 0.8797 | SRCC = 0.8805 ]
+INFO:root:Average loss=0.23183250117164128
+INFO:root:
+[Valid][1][UTT][ MSE = 0.2930 | LCC = 0.9033 | SRCC = 0.9035 ]
+INFO:root:Average loss=0.20205044651651657
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2731 | LCC = 0.9090 | SRCC = 0.9091 ]
+INFO:root:Average loss=0.1890778933043425
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2869 | LCC = 0.9117 | SRCC = 0.9118 ]
+INFO:root:Average loss=0.17378311050851222
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2665 | LCC = 0.9130 | SRCC = 0.9137 ]
+INFO:root:Average loss=0.16254348633003374
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2355 | LCC = 0.9140 | SRCC = 0.9137 ]
+INFO:root:Average loss=0.15618486904856788
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2485 | LCC = 0.9145 | SRCC = 0.9147 ]
+INFO:root:Average loss=0.14985793971085135
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2850 | LCC = 0.9124 | SRCC = 0.9140 ]
+INFO:root:Average loss=0.14395758961838795
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2684 | LCC = 0.9137 | SRCC = 0.9135 ]
+INFO:root:Average loss=0.13762679747763396
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2550 | LCC = 0.9127 | SRCC = 0.9121 ]
+INFO:root:Average loss=0.12752486054780166
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2527 | LCC = 0.9139 | SRCC = 0.9136 ]
+INFO:root:Average loss=0.12279291593098227
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2595 | LCC = 0.9131 | SRCC = 0.9127 ]
+INFO:root:Average loss=0.11543510203933441
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2744 | LCC = 0.9107 | SRCC = 0.9106 ]
+INFO:root:Average loss=0.11302425195543753
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2633 | LCC = 0.9101 | SRCC = 0.9103 ]
+INFO:root:Average loss=0.10984737400202393
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2571 | LCC = 0.9102 | SRCC = 0.9111 ]
+INFO:root:Average loss=0.1022053825631307
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2549 | LCC = 0.9121 | SRCC = 0.9121 ]
+INFO:root:Average loss=0.09984387272182917
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2662 | LCC = 0.9108 | SRCC = 0.9106 ]
+INFO:root:Average loss=0.0966098906117023
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2813 | LCC = 0.9077 | SRCC = 0.9086 ]
+INFO:root:Average loss=0.0947185921660393
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2572 | LCC = 0.9081 | SRCC = 0.9090 ]
+INFO:root:Average loss=0.08736853285375014
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2481 | LCC = 0.9086 | SRCC = 0.9082 ]
+INFO:root:Average loss=0.08675893103283953
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2653 | LCC = 0.9066 | SRCC = 0.9065 ]
+INFO:root:Average loss=0.0818977086142653
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2435 | LCC = 0.9094 | SRCC = 0.9090 ]
+INFO:root:Average loss=0.07939190652242975
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2463 | LCC = 0.9083 | SRCC = 0.9080 ]
+INFO:root:Average loss=0.07457428244669313
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2480 | LCC = 0.9090 | SRCC = 0.9086 ]
+INFO:root:Average loss=0.0751065140902307
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2504 | LCC = 0.9060 | SRCC = 0.9078 ]
+INFO:root:Average loss=0.07503754975048103
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2637 | LCC = 0.9079 | SRCC = 0.9081 ]
+INFO:root:Average loss=0.07139340473737331
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2577 | LCC = 0.9091 | SRCC = 0.9088 ]
+INFO:root:Average loss=0.06834535201677697
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2505 | LCC = 0.9094 | SRCC = 0.9093 ]
+INFO:root:Average loss=0.06673628929464114
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2459 | LCC = 0.9099 | SRCC = 0.9098 ]
+INFO:root:Average loss=0.06375883312928195
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2763 | LCC = 0.9072 | SRCC = 0.9072 ]
+INFO:root:Average loss=0.06435214954218424
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2502 | LCC = 0.9088 | SRCC = 0.9091 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4162 | LCC = 0.7712 | SRCC = 0.7517 ]

results_nisqa/w2v2_base/layer1/run1/config.gin

@@ -0,0 +1,31 @@
+# Training config on NISQA dataset.
+
+# ------------------------------
+# Architecture
+# ------------------------------
+ProjectionHead:
+    conv_channels = (32, 32)
+    dense_neurons = (64, 32, 1)
+    use_poolings = (True, True)
+    kernel_size = 5
+    pool_size = 5
+    
+# ------------------------------
+# Dataset
+# ------------------------------
+Nisqa:
+    debug = False
+
+get_dataloader:
+    num_workers = 12
+    shuffle = True
+
+# ------------------------------
+# Training loop
+# ------------------------------
+TrainingLoop:
+    num_epochs = 30
+    learning_rate = 1e-4
+    dataset_cls = @Nisqa
+    weight_decay = 0.0
+    batch_size_train = 64

results_nisqa/w2v2_base/layer1/run1/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13c5d924c08911398aa2cc8d0fe2e60de7a453182439a0dd70172cbfa8428a0f
+size 2747979

results_nisqa/w2v2_base/layer1/run1/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1e2a26517f12610325f1083b8ed29b924f7aacb260855c9c8aecb7f207006b9
+size 2749702

results_nisqa/w2v2_base/layer1/run1/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 11020
+INFO:root:Num val speech clips: 2700
+INFO:root:Num test speech clips: 232
+INFO:root:Device=cuda
+INFO:root:Average loss=0.42079639727669643
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3297 | LCC = 0.8823 | SRCC = 0.8824 ]
+INFO:root:Average loss=0.2348495078396935
+INFO:root:
+[Valid][1][UTT][ MSE = 0.2934 | LCC = 0.9044 | SRCC = 0.9044 ]
+INFO:root:Average loss=0.20348848049351245
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2665 | LCC = 0.9097 | SRCC = 0.9098 ]
+INFO:root:Average loss=0.18695151289074408
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2520 | LCC = 0.9130 | SRCC = 0.9131 ]
+INFO:root:Average loss=0.17688856235129297
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2796 | LCC = 0.9131 | SRCC = 0.9131 ]
+INFO:root:Average loss=0.1672476589076781
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2535 | LCC = 0.9146 | SRCC = 0.9150 ]
+INFO:root:Average loss=0.15619305310221765
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2674 | LCC = 0.9139 | SRCC = 0.9141 ]
+INFO:root:Average loss=0.14986810475760112
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2593 | LCC = 0.9126 | SRCC = 0.9137 ]
+INFO:root:Average loss=0.1423182886287656
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2339 | LCC = 0.9152 | SRCC = 0.9153 ]
+INFO:root:Average loss=0.13629879875679238
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2504 | LCC = 0.9133 | SRCC = 0.9135 ]
+INFO:root:Average loss=0.12806375173991816
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2726 | LCC = 0.9121 | SRCC = 0.9132 ]
+INFO:root:Average loss=0.12159248785993268
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2435 | LCC = 0.9112 | SRCC = 0.9128 ]
+INFO:root:Average loss=0.11674804373972678
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2457 | LCC = 0.9092 | SRCC = 0.9108 ]
+INFO:root:Average loss=0.11240020493379219
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2488 | LCC = 0.9101 | SRCC = 0.9113 ]
+INFO:root:Average loss=0.11014750679534983
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2314 | LCC = 0.9116 | SRCC = 0.9129 ]
+INFO:root:Average loss=0.10269484040192786
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2731 | LCC = 0.9117 | SRCC = 0.9114 ]
+INFO:root:Average loss=0.09724281840241714
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2653 | LCC = 0.9095 | SRCC = 0.9092 ]
+INFO:root:Average loss=0.094509947571279
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2753 | LCC = 0.9090 | SRCC = 0.9091 ]
+INFO:root:Average loss=0.09011944254181978
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2590 | LCC = 0.9099 | SRCC = 0.9096 ]
+INFO:root:Average loss=0.08746569644595158
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2466 | LCC = 0.9093 | SRCC = 0.9093 ]
+INFO:root:Average loss=0.08325344138924097
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2385 | LCC = 0.9095 | SRCC = 0.9097 ]
+INFO:root:Average loss=0.08433052712265467
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2619 | LCC = 0.9082 | SRCC = 0.9092 ]
+INFO:root:Average loss=0.07685908785014484
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2554 | LCC = 0.9088 | SRCC = 0.9087 ]
+INFO:root:Average loss=0.07621902591317375
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2360 | LCC = 0.9090 | SRCC = 0.9096 ]
+INFO:root:Average loss=0.07313514382422315
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2707 | LCC = 0.9079 | SRCC = 0.9077 ]
+INFO:root:Average loss=0.07038647879582609
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2534 | LCC = 0.9056 | SRCC = 0.9063 ]
+INFO:root:Average loss=0.0685674838850953
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2696 | LCC = 0.9077 | SRCC = 0.9071 ]
+INFO:root:Average loss=0.06732527011564013
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2636 | LCC = 0.9073 | SRCC = 0.9067 ]
+INFO:root:Average loss=0.06531374765872267
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2495 | LCC = 0.9065 | SRCC = 0.9062 ]
+INFO:root:Average loss=0.06180603806510826
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2459 | LCC = 0.9065 | SRCC = 0.9066 ]
+INFO:root:Average loss=0.0629519805139889
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2620 | LCC = 0.9079 | SRCC = 0.9074 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.3836 | LCC = 0.7560 | SRCC = 0.7420 ]

results_nisqa/w2v2_base/layer1/run2/config.gin

@@ -0,0 +1,31 @@
+# Training config on NISQA dataset.
+
+# ------------------------------
+# Architecture
+# ------------------------------
+ProjectionHead:
+    conv_channels = (32, 32)
+    dense_neurons = (64, 32, 1)
+    use_poolings = (True, True)
+    kernel_size = 5
+    pool_size = 5
+    
+# ------------------------------
+# Dataset
+# ------------------------------
+Nisqa:
+    debug = False
+
+get_dataloader:
+    num_workers = 12
+    shuffle = True
+
+# ------------------------------
+# Training loop
+# ------------------------------
+TrainingLoop:
+    num_epochs = 30
+    learning_rate = 1e-4
+    dataset_cls = @Nisqa
+    weight_decay = 0.0
+    batch_size_train = 64

results_nisqa/w2v2_base/layer1/run2/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:665a4c7f2c71e2b7561dbc31c77840cc83373cc9ff4d971157326ae2433304d7
+size 2747979

results_nisqa/w2v2_base/layer1/run2/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5453bf1d41260f8c7f7e1306aa4a3b89a5c4482dea1be51c3e6dbd15dfbc5f4
+size 2749702

results_nisqa/w2v2_base/layer1/run2/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 11020
+INFO:root:Num val speech clips: 2700
+INFO:root:Num test speech clips: 232
+INFO:root:Device=cuda
+INFO:root:Average loss=0.47191362100184997
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3142 | LCC = 0.8775 | SRCC = 0.8791 ]
+INFO:root:Average loss=0.23304167678418186
+INFO:root:
+[Valid][1][UTT][ MSE = 0.2943 | LCC = 0.8997 | SRCC = 0.8995 ]
+INFO:root:Average loss=0.2027546578662933
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2740 | LCC = 0.9066 | SRCC = 0.9069 ]
+INFO:root:Average loss=0.18250450720159994
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2883 | LCC = 0.9106 | SRCC = 0.9114 ]
+INFO:root:Average loss=0.1740097487506839
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2504 | LCC = 0.9117 | SRCC = 0.9117 ]
+INFO:root:Average loss=0.1616970681386187
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2687 | LCC = 0.9135 | SRCC = 0.9136 ]
+INFO:root:Average loss=0.1547936099457603
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2519 | LCC = 0.9140 | SRCC = 0.9138 ]
+INFO:root:Average loss=0.14812706169709994
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2374 | LCC = 0.9138 | SRCC = 0.9138 ]
+INFO:root:Average loss=0.14149325504640622
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2478 | LCC = 0.9131 | SRCC = 0.9134 ]
+INFO:root:Average loss=0.1326942708278667
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2475 | LCC = 0.9105 | SRCC = 0.9117 ]
+INFO:root:Average loss=0.12701404581352466
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2556 | LCC = 0.9105 | SRCC = 0.9106 ]
+INFO:root:Average loss=0.1213962070514701
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2622 | LCC = 0.9105 | SRCC = 0.9116 ]
+INFO:root:Average loss=0.11688236336659834
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2399 | LCC = 0.9114 | SRCC = 0.9120 ]
+INFO:root:Average loss=0.11268372301852082
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2370 | LCC = 0.9115 | SRCC = 0.9117 ]
+INFO:root:Average loss=0.10770984213648505
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2331 | LCC = 0.9125 | SRCC = 0.9127 ]
+INFO:root:Average loss=0.10341569124532572
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2618 | LCC = 0.9103 | SRCC = 0.9105 ]
+INFO:root:Average loss=0.0990697529403805
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2236 | LCC = 0.9103 | SRCC = 0.9110 ]
+INFO:root:Average loss=0.09471679946160041
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2380 | LCC = 0.9078 | SRCC = 0.9080 ]
+INFO:root:Average loss=0.09024654563537912
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2533 | LCC = 0.9074 | SRCC = 0.9077 ]
+INFO:root:Average loss=0.087887292566327
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2457 | LCC = 0.9088 | SRCC = 0.9088 ]
+INFO:root:Average loss=0.08342296802411879
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2420 | LCC = 0.9077 | SRCC = 0.9082 ]
+INFO:root:Average loss=0.08034802107624925
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2400 | LCC = 0.9074 | SRCC = 0.9084 ]
+INFO:root:Average loss=0.07772412213545314
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2467 | LCC = 0.9070 | SRCC = 0.9072 ]
+INFO:root:Average loss=0.0757277177248387
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2451 | LCC = 0.9078 | SRCC = 0.9079 ]
+INFO:root:Average loss=0.07218075621162535
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2346 | LCC = 0.9083 | SRCC = 0.9089 ]
+INFO:root:Average loss=0.07157445427654796
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2473 | LCC = 0.9042 | SRCC = 0.9052 ]
+INFO:root:Average loss=0.07045848148664988
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2350 | LCC = 0.9080 | SRCC = 0.9081 ]
+INFO:root:Average loss=0.06662582547936825
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2464 | LCC = 0.9059 | SRCC = 0.9066 ]
+INFO:root:Average loss=0.06482528206068656
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2310 | LCC = 0.9075 | SRCC = 0.9079 ]
+INFO:root:Average loss=0.06322503957703624
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2557 | LCC = 0.9057 | SRCC = 0.9060 ]
+INFO:root:Average loss=0.059369667675453805
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2503 | LCC = 0.9062 | SRCC = 0.9063 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.3757 | LCC = 0.7794 | SRCC = 0.7656 ]

results_nisqa/w2v2_base/layer1/run3/config.gin

@@ -0,0 +1,31 @@
+# Training config on NISQA dataset.
+
+# ------------------------------
+# Architecture
+# ------------------------------
+ProjectionHead:
+    conv_channels = (32, 32)
+    dense_neurons = (64, 32, 1)
+    use_poolings = (True, True)
+    kernel_size = 5
+    pool_size = 5
+    
+# ------------------------------
+# Dataset
+# ------------------------------
+Nisqa:
+    debug = False
+
+get_dataloader:
+    num_workers = 12
+    shuffle = True
+
+# ------------------------------
+# Training loop
+# ------------------------------
+TrainingLoop:
+    num_epochs = 30
+    learning_rate = 1e-4
+    dataset_cls = @Nisqa
+    weight_decay = 0.0
+    batch_size_train = 64

results_nisqa/w2v2_base/layer1/run3/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b79a32a58ed01e0805397b6b56b3a336d287292b4c2a016f9572f6f37c43a7c8
+size 2747979

results_nisqa/w2v2_base/layer1/run3/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ead62719480caaeb0f107548155acbf58a0ef932d00dcb099a5224c17c9a4835
+size 2749702

results_nisqa/w2v2_base/layer1/run3/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 11020
+INFO:root:Num val speech clips: 2700
+INFO:root:Num test speech clips: 232
+INFO:root:Device=cuda
+INFO:root:Average loss=0.4645358320605548
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3285 | LCC = 0.8762 | SRCC = 0.8765 ]
+INFO:root:Average loss=0.24353808681399836
+INFO:root:
+[Valid][1][UTT][ MSE = 0.2822 | LCC = 0.9008 | SRCC = 0.9008 ]
+INFO:root:Average loss=0.20912874412502166
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2796 | LCC = 0.9057 | SRCC = 0.9058 ]
+INFO:root:Average loss=0.18992455678351353
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2659 | LCC = 0.9112 | SRCC = 0.9109 ]
+INFO:root:Average loss=0.17423045574929674
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2782 | LCC = 0.9113 | SRCC = 0.9116 ]
+INFO:root:Average loss=0.16558513846356054
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2479 | LCC = 0.9127 | SRCC = 0.9122 ]
+INFO:root:Average loss=0.15647276951743
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2634 | LCC = 0.9137 | SRCC = 0.9136 ]
+INFO:root:Average loss=0.14936454038571761
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2645 | LCC = 0.9119 | SRCC = 0.9122 ]
+INFO:root:Average loss=0.142234451853471
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2366 | LCC = 0.9108 | SRCC = 0.9118 ]
+INFO:root:Average loss=0.13475031289406594
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2474 | LCC = 0.9118 | SRCC = 0.9120 ]
+INFO:root:Average loss=0.13033452130466527
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2517 | LCC = 0.9113 | SRCC = 0.9119 ]
+INFO:root:Average loss=0.12352152508979587
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2616 | LCC = 0.9123 | SRCC = 0.9118 ]
+INFO:root:Average loss=0.12031384841727384
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2661 | LCC = 0.9118 | SRCC = 0.9115 ]
+INFO:root:Average loss=0.11395652344547255
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2609 | LCC = 0.9115 | SRCC = 0.9115 ]
+INFO:root:Average loss=0.10535275264282447
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2402 | LCC = 0.9113 | SRCC = 0.9113 ]
+INFO:root:Average loss=0.10479879112257434
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2355 | LCC = 0.9099 | SRCC = 0.9110 ]
+INFO:root:Average loss=0.10019654591131762
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2480 | LCC = 0.9091 | SRCC = 0.9094 ]
+INFO:root:Average loss=0.0961237923172168
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2462 | LCC = 0.9095 | SRCC = 0.9096 ]
+INFO:root:Average loss=0.09073398544350801
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2364 | LCC = 0.9079 | SRCC = 0.9086 ]
+INFO:root:Average loss=0.08823640007159614
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2556 | LCC = 0.9091 | SRCC = 0.9090 ]
+INFO:root:Average loss=0.08493320788176073
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2417 | LCC = 0.9091 | SRCC = 0.9094 ]
+INFO:root:Average loss=0.0802434212555086
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2619 | LCC = 0.9080 | SRCC = 0.9084 ]
+INFO:root:Average loss=0.0777926423689189
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2570 | LCC = 0.9093 | SRCC = 0.9090 ]
+INFO:root:Average loss=0.07505168800236862
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2439 | LCC = 0.9075 | SRCC = 0.9086 ]
+INFO:root:Average loss=0.07597914521146372
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2540 | LCC = 0.9042 | SRCC = 0.9052 ]
+INFO:root:Average loss=0.071441778234836
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2487 | LCC = 0.9078 | SRCC = 0.9085 ]
+INFO:root:Average loss=0.06841913184334088
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2482 | LCC = 0.9074 | SRCC = 0.9083 ]
+INFO:root:Average loss=0.06771057969056113
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2570 | LCC = 0.9069 | SRCC = 0.9073 ]
+INFO:root:Average loss=0.06449337178744333
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2523 | LCC = 0.9080 | SRCC = 0.9079 ]
+INFO:root:Average loss=0.06529931504430109
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2413 | LCC = 0.9076 | SRCC = 0.9080 ]
+INFO:root:Average loss=0.06153249826734466
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2588 | LCC = 0.9084 | SRCC = 0.9085 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.3843 | LCC = 0.7690 | SRCC = 0.7518 ]

results_nisqa/w2v2_base/layer1/run4/config.gin

@@ -0,0 +1,31 @@
+# Training config on NISQA dataset.
+
+# ------------------------------
+# Architecture
+# ------------------------------
+ProjectionHead:
+    conv_channels = (32, 32)
+    dense_neurons = (64, 32, 1)
+    use_poolings = (True, True)
+    kernel_size = 5
+    pool_size = 5
+    
+# ------------------------------
+# Dataset
+# ------------------------------
+Nisqa:
+    debug = False
+
+get_dataloader:
+    num_workers = 12
+    shuffle = True
+
+# ------------------------------
+# Training loop
+# ------------------------------
+TrainingLoop:
+    num_epochs = 30
+    learning_rate = 1e-4
+    dataset_cls = @Nisqa
+    weight_decay = 0.0
+    batch_size_train = 64

results_nisqa/w2v2_base/layer1/run4/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e81b05bc71d35d78aea0c35382b86c0e91930d6484cb52fa826290c014efb8e
+size 2747979

results_nisqa/w2v2_base/layer1/run4/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c73f95b0b1218500cdfa0fde754002f64e428afc30d129949aeb42d12939f41
+size 2749702

results_nisqa/w2v2_base/layer1/run4/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 11020
+INFO:root:Num val speech clips: 2700
+INFO:root:Num test speech clips: 232
+INFO:root:Device=cuda
+INFO:root:Average loss=0.43465099910091115
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3250 | LCC = 0.8756 | SRCC = 0.8760 ]
+INFO:root:Average loss=0.24063599747040368
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3248 | LCC = 0.8976 | SRCC = 0.8991 ]
+INFO:root:Average loss=0.20355439870860537
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2591 | LCC = 0.9062 | SRCC = 0.9061 ]
+INFO:root:Average loss=0.19028889206964847
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2934 | LCC = 0.9066 | SRCC = 0.9085 ]
+INFO:root:Average loss=0.17772545140085882
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2912 | LCC = 0.9087 | SRCC = 0.9091 ]
+INFO:root:Average loss=0.1654128308241078
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2675 | LCC = 0.9088 | SRCC = 0.9121 ]
+INFO:root:Average loss=0.15829817528669546
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2867 | LCC = 0.9099 | SRCC = 0.9119 ]
+INFO:root:Average loss=0.1490263454400735
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2547 | LCC = 0.9087 | SRCC = 0.9100 ]
+INFO:root:Average loss=0.1446479712263008
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2491 | LCC = 0.9101 | SRCC = 0.9112 ]
+INFO:root:Average loss=0.13472810180890077
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2642 | LCC = 0.9093 | SRCC = 0.9098 ]
+INFO:root:Average loss=0.1325780803585328
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2803 | LCC = 0.9095 | SRCC = 0.9100 ]
+INFO:root:Average loss=0.12557296911416027
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2484 | LCC = 0.9087 | SRCC = 0.9088 ]
+INFO:root:Average loss=0.12142900554086432
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2803 | LCC = 0.9072 | SRCC = 0.9094 ]
+INFO:root:Average loss=0.11248545544598833
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2651 | LCC = 0.9074 | SRCC = 0.9079 ]
+INFO:root:Average loss=0.11051415583129563
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2563 | LCC = 0.9079 | SRCC = 0.9096 ]
+INFO:root:Average loss=0.10501352105440431
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2533 | LCC = 0.9061 | SRCC = 0.9089 ]
+INFO:root:Average loss=0.10199560436037923
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2655 | LCC = 0.9077 | SRCC = 0.9083 ]
+INFO:root:Average loss=0.0957174086243431
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2507 | LCC = 0.9049 | SRCC = 0.9049 ]
+INFO:root:Average loss=0.09607243387168543
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2661 | LCC = 0.9074 | SRCC = 0.9076 ]
+INFO:root:Average loss=0.08995198434746335
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2497 | LCC = 0.9060 | SRCC = 0.9061 ]
+INFO:root:Average loss=0.0868710470789602
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2717 | LCC = 0.9036 | SRCC = 0.9058 ]
+INFO:root:Average loss=0.0841633923825501
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2621 | LCC = 0.9034 | SRCC = 0.9037 ]
+INFO:root:Average loss=0.08126699435659226
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2489 | LCC = 0.9053 | SRCC = 0.9054 ]
+INFO:root:Average loss=0.0778068660440817
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2542 | LCC = 0.9038 | SRCC = 0.9038 ]
+INFO:root:Average loss=0.0765922138671999
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2731 | LCC = 0.9016 | SRCC = 0.9017 ]
+INFO:root:Average loss=0.07271993795485165
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2821 | LCC = 0.9009 | SRCC = 0.9032 ]
+INFO:root:Average loss=0.0709245972809075
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2509 | LCC = 0.9055 | SRCC = 0.9053 ]
+INFO:root:Average loss=0.06831825020685361
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2640 | LCC = 0.9037 | SRCC = 0.9041 ]
+INFO:root:Average loss=0.06718278379109553
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2802 | LCC = 0.9019 | SRCC = 0.9027 ]
+INFO:root:Average loss=0.06434275987261982
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2558 | LCC = 0.9043 | SRCC = 0.9047 ]
+INFO:root:Average loss=0.06542826297483004
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2626 | LCC = 0.9022 | SRCC = 0.9026 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4340 | LCC = 0.7715 | SRCC = 0.7549 ]

results_nisqa/w2v2_base/layer10/run0/config.gin

@@ -0,0 +1,31 @@
+# Training config on NISQA dataset.
+
+# ------------------------------
+# Architecture
+# ------------------------------
+ProjectionHead:
+    conv_channels = (32, 32)
+    dense_neurons = (64, 32, 1)
+    use_poolings = (True, True)
+    kernel_size = 5
+    pool_size = 5
+    
+# ------------------------------
+# Dataset
+# ------------------------------
+Nisqa:
+    debug = False
+
+get_dataloader:
+    num_workers = 12
+    shuffle = True
+
+# ------------------------------
+# Training loop
+# ------------------------------
+TrainingLoop:
+    num_epochs = 30
+    learning_rate = 1e-4
+    dataset_cls = @Nisqa
+    weight_decay = 0.0
+    batch_size_train = 64

results_nisqa/w2v2_base/layer10/run0/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df75d9f57b929dda8c6031d7d37d0105806499842e89499b771a55921fdc9cfa
+size 2747979