Upload SSL Layer MOS checkpoints (part 13)

results_nisqa/w2v2_large_xlsr53/layer10/run3/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06eb72ee97db195363627ee2dda98b0fbd4a7c54aee06d68fd6784549c8458b0
+size 3403339

results_nisqa/w2v2_large_xlsr53/layer10/run3/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29dc9033e4d0af89733844cf9abd606179751e7ea535c75bcbb3cd3f7a788175
+size 3405062

results_nisqa/w2v2_large_xlsr53/layer10/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=0.40070874558363345
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3338 | LCC = 0.8954 | SRCC = 0.8979 ]
+INFO:root:Average loss=0.22120705255538742
+INFO:root:
+[Valid][1][UTT][ MSE = 0.2907 | LCC = 0.9055 | SRCC = 0.9053 ]
+INFO:root:Average loss=0.20961992365087387
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3271 | LCC = 0.9083 | SRCC = 0.9082 ]
+INFO:root:Average loss=0.20643512603175435
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2972 | LCC = 0.9091 | SRCC = 0.9098 ]
+INFO:root:Average loss=0.18861697497912225
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2813 | LCC = 0.9121 | SRCC = 0.9121 ]
+INFO:root:Average loss=0.17471699705185917
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2638 | LCC = 0.9138 | SRCC = 0.9137 ]
+INFO:root:Average loss=0.16829351762126635
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2784 | LCC = 0.9133 | SRCC = 0.9132 ]
+INFO:root:Average loss=0.16861853068102303
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2589 | LCC = 0.9139 | SRCC = 0.9140 ]
+INFO:root:Average loss=0.15918979935907904
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2787 | LCC = 0.9152 | SRCC = 0.9149 ]
+INFO:root:Average loss=0.15523253450159394
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2677 | LCC = 0.9144 | SRCC = 0.9146 ]
+INFO:root:Average loss=0.1480444418212582
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2846 | LCC = 0.9148 | SRCC = 0.9148 ]
+INFO:root:Average loss=0.1450222345187485
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2451 | LCC = 0.9160 | SRCC = 0.9158 ]
+INFO:root:Average loss=0.13845875208949768
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2540 | LCC = 0.9120 | SRCC = 0.9139 ]
+INFO:root:Average loss=0.13945366205335352
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2595 | LCC = 0.9086 | SRCC = 0.9125 ]
+INFO:root:Average loss=0.13058247658386396
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2345 | LCC = 0.9125 | SRCC = 0.9133 ]
+INFO:root:Average loss=0.12289333119557772
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2555 | LCC = 0.9135 | SRCC = 0.9132 ]
+INFO:root:Average loss=0.12034882226086765
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2411 | LCC = 0.9129 | SRCC = 0.9132 ]
+INFO:root:Average loss=0.11495334762713813
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2472 | LCC = 0.9134 | SRCC = 0.9131 ]
+INFO:root:Average loss=0.11150705232957885
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2420 | LCC = 0.9136 | SRCC = 0.9135 ]
+INFO:root:Average loss=0.1064118500906608
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2478 | LCC = 0.9097 | SRCC = 0.9108 ]
+INFO:root:Average loss=0.10259875534586824
+INFO:root:
+[Valid][20][UTT][ MSE = 0.3160 | LCC = 0.9104 | SRCC = 0.9104 ]
+INFO:root:Average loss=0.1036702170106717
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2903 | LCC = 0.9117 | SRCC = 0.9113 ]
+INFO:root:Average loss=0.09263259364087457
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2560 | LCC = 0.9080 | SRCC = 0.9089 ]
+INFO:root:Average loss=0.08877095842809346
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2596 | LCC = 0.9059 | SRCC = 0.9071 ]
+INFO:root:Average loss=0.08955498134469711
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2444 | LCC = 0.9092 | SRCC = 0.9093 ]
+INFO:root:Average loss=0.08312550187110901
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2627 | LCC = 0.9085 | SRCC = 0.9082 ]
+INFO:root:Average loss=0.0846172599821766
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2546 | LCC = 0.9098 | SRCC = 0.9098 ]
+INFO:root:Average loss=0.07862708006980103
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2429 | LCC = 0.9085 | SRCC = 0.9087 ]
+INFO:root:Average loss=0.07714248097614745
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2494 | LCC = 0.9081 | SRCC = 0.9078 ]
+INFO:root:Average loss=0.07187956666326248
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2475 | LCC = 0.9077 | SRCC = 0.9082 ]
+INFO:root:Average loss=0.07286350409677952
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2487 | LCC = 0.9095 | SRCC = 0.9091 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4165 | LCC = 0.7304 | SRCC = 0.7183 ]

results_nisqa/w2v2_large_xlsr53/layer10/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_large_xlsr53/layer10/run4/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf40050d4582b9ef36794027c6eee2122b0b3f700b89a920d2ff984a7340cd5f
+size 3403339

results_nisqa/w2v2_large_xlsr53/layer10/run4/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93332dcc3aa7ad959cd4d67657117303219b05d5f77f1e25006cc65695d21748
+size 3405062

results_nisqa/w2v2_large_xlsr53/layer10/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=0.43597569036690487
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3082 | LCC = 0.8989 | SRCC = 0.8987 ]
+INFO:root:Average loss=0.22928062322511839
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3814 | LCC = 0.9051 | SRCC = 0.9060 ]
+INFO:root:Average loss=0.2044524481100154
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3219 | LCC = 0.9098 | SRCC = 0.9096 ]
+INFO:root:Average loss=0.20417099773367017
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2859 | LCC = 0.9118 | SRCC = 0.9115 ]
+INFO:root:Average loss=0.18740450952611218
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3453 | LCC = 0.9093 | SRCC = 0.9116 ]
+INFO:root:Average loss=0.1819288198313961
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2820 | LCC = 0.9135 | SRCC = 0.9136 ]
+INFO:root:Average loss=0.17206568486256407
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2609 | LCC = 0.9137 | SRCC = 0.9139 ]
+INFO:root:Average loss=0.17222920921496573
+INFO:root:
+[Valid][7][UTT][ MSE = 0.3216 | LCC = 0.9136 | SRCC = 0.9135 ]
+INFO:root:Average loss=0.15886955893453145
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2519 | LCC = 0.9139 | SRCC = 0.9151 ]
+INFO:root:Average loss=0.15477564457179493
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2501 | LCC = 0.9156 | SRCC = 0.9150 ]
+INFO:root:Average loss=0.15001520217326336
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2448 | LCC = 0.9167 | SRCC = 0.9164 ]
+INFO:root:Average loss=0.14798956691701978
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2210 | LCC = 0.9159 | SRCC = 0.9160 ]
+INFO:root:Average loss=0.14249783698361734
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2354 | LCC = 0.9155 | SRCC = 0.9156 ]
+INFO:root:Average loss=0.13580096577633324
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2481 | LCC = 0.9125 | SRCC = 0.9135 ]
+INFO:root:Average loss=0.12954326512324327
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2235 | LCC = 0.9145 | SRCC = 0.9146 ]
+INFO:root:Average loss=0.12579415053818266
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2672 | LCC = 0.9148 | SRCC = 0.9155 ]
+INFO:root:Average loss=0.12007445704213457
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2690 | LCC = 0.9135 | SRCC = 0.9143 ]
+INFO:root:Average loss=0.11431484176330484
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2487 | LCC = 0.9151 | SRCC = 0.9145 ]
+INFO:root:Average loss=0.10842725679795177
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2273 | LCC = 0.9133 | SRCC = 0.9135 ]
+INFO:root:Average loss=0.10820932382706962
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2469 | LCC = 0.9132 | SRCC = 0.9132 ]
+INFO:root:Average loss=0.10493012397103227
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2918 | LCC = 0.9103 | SRCC = 0.9116 ]
+INFO:root:Average loss=0.09840076866780402
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2716 | LCC = 0.9104 | SRCC = 0.9117 ]
+INFO:root:Average loss=0.09348792978966167
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2382 | LCC = 0.9117 | SRCC = 0.9112 ]
+INFO:root:Average loss=0.09118205830024156
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2417 | LCC = 0.9126 | SRCC = 0.9130 ]
+INFO:root:Average loss=0.09100293617889371
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2520 | LCC = 0.9113 | SRCC = 0.9115 ]
+INFO:root:Average loss=0.08299180509985526
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2556 | LCC = 0.9102 | SRCC = 0.9101 ]
+INFO:root:Average loss=0.07898753856690978
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2541 | LCC = 0.9100 | SRCC = 0.9101 ]
+INFO:root:Average loss=0.07761970801935719
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2557 | LCC = 0.9096 | SRCC = 0.9113 ]
+INFO:root:Average loss=0.07477183304081074
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2986 | LCC = 0.9070 | SRCC = 0.9080 ]
+INFO:root:Average loss=0.07420015959553636
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2411 | LCC = 0.9105 | SRCC = 0.9104 ]
+INFO:root:Average loss=0.0741897067754944
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2525 | LCC = 0.9087 | SRCC = 0.9084 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4180 | LCC = 0.7486 | SRCC = 0.7352 ]

results_nisqa/w2v2_large_xlsr53/layer12/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_large_xlsr53/layer12/run0/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fad6f5f9f2a125119e553532969c14f30c8c2733b03869b9879bb71798464e13
+size 3403339

results_nisqa/w2v2_large_xlsr53/layer12/run0/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a093008afdfa0f72c03ed35dae181af4bdb15926325f1375876cbb5eb9faa947
+size 3405062

results_nisqa/w2v2_large_xlsr53/layer12/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=0.4029752202116685
+INFO:root:
+[Valid][0][UTT][ MSE = 0.2934 | LCC = 0.8938 | SRCC = 0.8939 ]
+INFO:root:Average loss=0.22426742504786895
+INFO:root:
+[Valid][1][UTT][ MSE = 0.2568 | LCC = 0.9040 | SRCC = 0.9043 ]
+INFO:root:Average loss=0.2093253300627532
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2545 | LCC = 0.9084 | SRCC = 0.9092 ]
+INFO:root:Average loss=0.19335106428648005
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2735 | LCC = 0.9103 | SRCC = 0.9098 ]
+INFO:root:Average loss=0.1836271121495032
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2470 | LCC = 0.9102 | SRCC = 0.9120 ]
+INFO:root:Average loss=0.1832838637722021
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2543 | LCC = 0.9129 | SRCC = 0.9129 ]
+INFO:root:Average loss=0.17199393348886788
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2802 | LCC = 0.9125 | SRCC = 0.9121 ]
+INFO:root:Average loss=0.1684250049508376
+INFO:root:
+[Valid][7][UTT][ MSE = 0.3855 | LCC = 0.9123 | SRCC = 0.9138 ]
+INFO:root:Average loss=0.1619690090331728
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2776 | LCC = 0.9114 | SRCC = 0.9132 ]
+INFO:root:Average loss=0.15225318922645095
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2625 | LCC = 0.9152 | SRCC = 0.9152 ]
+INFO:root:Average loss=0.14608368355070236
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2908 | LCC = 0.9145 | SRCC = 0.9151 ]
+INFO:root:Average loss=0.1417129935987423
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2662 | LCC = 0.9153 | SRCC = 0.9152 ]
+INFO:root:Average loss=0.13682200053351462
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2668 | LCC = 0.9142 | SRCC = 0.9144 ]
+INFO:root:Average loss=0.13240789811562942
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2516 | LCC = 0.9132 | SRCC = 0.9138 ]
+INFO:root:Average loss=0.12546036052221507
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2646 | LCC = 0.9107 | SRCC = 0.9127 ]
+INFO:root:Average loss=0.1218133266433815
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2536 | LCC = 0.9135 | SRCC = 0.9133 ]
+INFO:root:Average loss=0.11815535113473848
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2758 | LCC = 0.9134 | SRCC = 0.9133 ]
+INFO:root:Average loss=0.10961908704972681
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2230 | LCC = 0.9131 | SRCC = 0.9130 ]
+INFO:root:Average loss=0.10912312601687592
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2785 | LCC = 0.9120 | SRCC = 0.9129 ]
+INFO:root:Average loss=0.1040437143145269
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2348 | LCC = 0.9118 | SRCC = 0.9114 ]
+INFO:root:Average loss=0.1002599992977746
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2631 | LCC = 0.9109 | SRCC = 0.9114 ]
+INFO:root:Average loss=0.09711474399862952
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2461 | LCC = 0.9100 | SRCC = 0.9106 ]
+INFO:root:Average loss=0.09061994646928903
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2410 | LCC = 0.9110 | SRCC = 0.9105 ]
+INFO:root:Average loss=0.08500200279616896
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2645 | LCC = 0.9071 | SRCC = 0.9077 ]
+INFO:root:Average loss=0.08410124239549471
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2287 | LCC = 0.9088 | SRCC = 0.9086 ]
+INFO:root:Average loss=0.07820040463453773
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2428 | LCC = 0.9091 | SRCC = 0.9083 ]
+INFO:root:Average loss=0.07870422852780089
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2575 | LCC = 0.9074 | SRCC = 0.9076 ]
+INFO:root:Average loss=0.07328228355315379
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2539 | LCC = 0.9068 | SRCC = 0.9071 ]
+INFO:root:Average loss=0.07139649640531898
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2557 | LCC = 0.9087 | SRCC = 0.9085 ]
+INFO:root:Average loss=0.07043677792383757
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2567 | LCC = 0.9079 | SRCC = 0.9082 ]
+INFO:root:Average loss=0.0685034562655956
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2471 | LCC = 0.9087 | SRCC = 0.9090 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4575 | LCC = 0.7355 | SRCC = 0.7188 ]

results_nisqa/w2v2_large_xlsr53/layer12/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_large_xlsr53/layer12/run1/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:601eee5c0ecc4958720d4e7d0789e6d2c2a080f5eed25cbe0b7e30a530a5935c
+size 3403339

results_nisqa/w2v2_large_xlsr53/layer12/run1/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d6fcbee8786ca6b71f607b2c284c718b10f03f269a785fbab756574ba7968b0
+size 3405062

results_nisqa/w2v2_large_xlsr53/layer12/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=0.43012736935836043
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3385 | LCC = 0.8934 | SRCC = 0.8929 ]
+INFO:root:Average loss=0.237255014694495
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3177 | LCC = 0.9039 | SRCC = 0.9031 ]
+INFO:root:Average loss=0.2114692587900713
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3066 | LCC = 0.9041 | SRCC = 0.9052 ]
+INFO:root:Average loss=0.2018104854949637
+INFO:root:
+[Valid][3][UTT][ MSE = 0.3310 | LCC = 0.9071 | SRCC = 0.9075 ]
+INFO:root:Average loss=0.19089909853962805
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2785 | LCC = 0.9104 | SRCC = 0.9101 ]
+INFO:root:Average loss=0.18020728648225695
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2573 | LCC = 0.9112 | SRCC = 0.9118 ]
+INFO:root:Average loss=0.17722379067385127
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2794 | LCC = 0.9122 | SRCC = 0.9125 ]
+INFO:root:Average loss=0.17531734179554648
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2789 | LCC = 0.9146 | SRCC = 0.9137 ]
+INFO:root:Average loss=0.1645109484650496
+INFO:root:
+[Valid][8][UTT][ MSE = 0.3059 | LCC = 0.9133 | SRCC = 0.9125 ]
+INFO:root:Average loss=0.15708171972477367
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2326 | LCC = 0.9135 | SRCC = 0.9141 ]
+INFO:root:Average loss=0.14965870451031393
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2966 | LCC = 0.9148 | SRCC = 0.9145 ]
+INFO:root:Average loss=0.14506039385161648
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2890 | LCC = 0.9148 | SRCC = 0.9149 ]
+INFO:root:Average loss=0.13991672278656436
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2521 | LCC = 0.9132 | SRCC = 0.9138 ]
+INFO:root:Average loss=0.13374393282598154
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2973 | LCC = 0.9125 | SRCC = 0.9119 ]
+INFO:root:Average loss=0.1315865249216901
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2723 | LCC = 0.9131 | SRCC = 0.9127 ]
+INFO:root:Average loss=0.128966244529782
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2565 | LCC = 0.9132 | SRCC = 0.9125 ]
+INFO:root:Average loss=0.11806138833134161
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2618 | LCC = 0.9138 | SRCC = 0.9132 ]
+INFO:root:Average loss=0.11150327699549625
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2846 | LCC = 0.9131 | SRCC = 0.9127 ]
+INFO:root:Average loss=0.11161387522268847
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2603 | LCC = 0.9133 | SRCC = 0.9126 ]
+INFO:root:Average loss=0.10402258948525252
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2815 | LCC = 0.9127 | SRCC = 0.9118 ]
+INFO:root:Average loss=0.1012498461183785
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2657 | LCC = 0.9126 | SRCC = 0.9122 ]
+INFO:root:Average loss=0.09608005467436218
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2467 | LCC = 0.9112 | SRCC = 0.9106 ]
+INFO:root:Average loss=0.09316717363850918
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2470 | LCC = 0.9105 | SRCC = 0.9112 ]
+INFO:root:Average loss=0.08950366615990683
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2414 | LCC = 0.9107 | SRCC = 0.9109 ]
+INFO:root:Average loss=0.08536695174313005
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2613 | LCC = 0.9131 | SRCC = 0.9124 ]
+INFO:root:Average loss=0.08433568641284987
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2651 | LCC = 0.9114 | SRCC = 0.9113 ]
+INFO:root:Average loss=0.07802220569611285
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2462 | LCC = 0.9105 | SRCC = 0.9101 ]
+INFO:root:Average loss=0.07534260875572359
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2490 | LCC = 0.9097 | SRCC = 0.9093 ]
+INFO:root:Average loss=0.07296159833325128
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2532 | LCC = 0.9105 | SRCC = 0.9099 ]
+INFO:root:Average loss=0.07208628422779843
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2613 | LCC = 0.9100 | SRCC = 0.9090 ]
+INFO:root:Average loss=0.0690605456300209
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2584 | LCC = 0.9083 | SRCC = 0.9081 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4417 | LCC = 0.7255 | SRCC = 0.7146 ]

results_nisqa/w2v2_large_xlsr53/layer12/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_large_xlsr53/layer12/run2/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e341b8dee297b1782826f77b506ee9832b3dde5ba9cdf46527f9f4e7c1d240f3
+size 3403339

results_nisqa/w2v2_large_xlsr53/layer12/run2/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c8b167377f90ab39d70e3dbae906e90dfb1cf5a4e2982c018a435197c7dd3d7
+size 3405062

results_nisqa/w2v2_large_xlsr53/layer12/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=0.40218038822529634
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3424 | LCC = 0.8935 | SRCC = 0.8955 ]
+INFO:root:Average loss=0.2281859744921585
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3357 | LCC = 0.9021 | SRCC = 0.9032 ]
+INFO:root:Average loss=0.2033634959703925
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2382 | LCC = 0.9065 | SRCC = 0.9067 ]
+INFO:root:Average loss=0.2020685990594026
+INFO:root:
+[Valid][3][UTT][ MSE = 0.3178 | LCC = 0.9089 | SRCC = 0.9105 ]
+INFO:root:Average loss=0.1852702017808925
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3819 | LCC = 0.9077 | SRCC = 0.9115 ]
+INFO:root:Average loss=0.18207473716983905
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2600 | LCC = 0.9113 | SRCC = 0.9120 ]
+INFO:root:Average loss=0.17279740763192922
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2620 | LCC = 0.9135 | SRCC = 0.9134 ]
+INFO:root:Average loss=0.1622677012158267
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2746 | LCC = 0.9149 | SRCC = 0.9147 ]
+INFO:root:Average loss=0.15684048219912314
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2699 | LCC = 0.9153 | SRCC = 0.9150 ]
+INFO:root:Average loss=0.1532943186732386
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2619 | LCC = 0.9138 | SRCC = 0.9147 ]
+INFO:root:Average loss=0.14935055392325958
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2319 | LCC = 0.9144 | SRCC = 0.9138 ]
+INFO:root:Average loss=0.14296348203468873
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2455 | LCC = 0.9148 | SRCC = 0.9156 ]
+INFO:root:Average loss=0.14039647299258007
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2488 | LCC = 0.9151 | SRCC = 0.9157 ]
+INFO:root:Average loss=0.12894492811722563
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2267 | LCC = 0.9161 | SRCC = 0.9156 ]
+INFO:root:Average loss=0.12651109863395635
+INFO:root:
+[Valid][14][UTT][ MSE = 0.3236 | LCC = 0.9114 | SRCC = 0.9135 ]
+INFO:root:Average loss=0.12353469211758905
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2824 | LCC = 0.9114 | SRCC = 0.9140 ]
+INFO:root:Average loss=0.11322278710286741
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2509 | LCC = 0.9142 | SRCC = 0.9137 ]
+INFO:root:Average loss=0.10971474402040415
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2527 | LCC = 0.9121 | SRCC = 0.9130 ]
+INFO:root:Average loss=0.1063967769416426
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2592 | LCC = 0.9122 | SRCC = 0.9132 ]
+INFO:root:Average loss=0.09805945386690211
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2325 | LCC = 0.9099 | SRCC = 0.9103 ]
+INFO:root:Average loss=0.09751324945619341
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2413 | LCC = 0.9098 | SRCC = 0.9114 ]
+INFO:root:Average loss=0.09371230395973762
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2453 | LCC = 0.9106 | SRCC = 0.9117 ]
+INFO:root:Average loss=0.09089080898920236
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2720 | LCC = 0.9060 | SRCC = 0.9106 ]
+INFO:root:Average loss=0.08636566458237653
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2652 | LCC = 0.9087 | SRCC = 0.9102 ]
+INFO:root:Average loss=0.0818176319704235
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2748 | LCC = 0.9091 | SRCC = 0.9102 ]
+INFO:root:Average loss=0.0809261535073636
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2788 | LCC = 0.9060 | SRCC = 0.9088 ]
+INFO:root:Average loss=0.07530032630170012
+INFO:root:
+[Valid][26][UTT][ MSE = 0.3057 | LCC = 0.9016 | SRCC = 0.9067 ]
+INFO:root:Average loss=0.07357059199513727
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2640 | LCC = 0.9067 | SRCC = 0.9081 ]
+INFO:root:Average loss=0.072261465013544
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2952 | LCC = 0.9053 | SRCC = 0.9074 ]
+INFO:root:Average loss=0.06748448555610773
+INFO:root:
+[Valid][29][UTT][ MSE = 0.3155 | LCC = 0.9045 | SRCC = 0.9068 ]
+INFO:root:Average loss=0.06668400633111166
+INFO:root:
+[Valid][30][UTT][ MSE = 0.3416 | LCC = 0.9004 | SRCC = 0.9051 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4024 | LCC = 0.7381 | SRCC = 0.7229 ]

results_nisqa/w2v2_large_xlsr53/layer12/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_large_xlsr53/layer12/run3/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb5c921d2f0a27477de3d93b2586663f2ac1d6669d9918186d2337c63ac81e6b
+size 3403339

results_nisqa/w2v2_large_xlsr53/layer12/run3/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9bbb6e2ae18de1dd7521f8b1de824fd2d49c85609cf40f5dd8a0faa132582195
+size 3405062

results_nisqa/w2v2_large_xlsr53/layer12/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=0.40563768251782895
+INFO:root:
+[Valid][0][UTT][ MSE = 0.4083 | LCC = 0.8906 | SRCC = 0.8913 ]
+INFO:root:Average loss=0.23715700002419465
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3075 | LCC = 0.9027 | SRCC = 0.9018 ]
+INFO:root:Average loss=0.21821646672280537
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3354 | LCC = 0.9048 | SRCC = 0.9043 ]
+INFO:root:Average loss=0.20218434947074493
+INFO:root:
+[Valid][3][UTT][ MSE = 0.3262 | LCC = 0.9085 | SRCC = 0.9081 ]
+INFO:root:Average loss=0.19294118752024766
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3209 | LCC = 0.9104 | SRCC = 0.9099 ]
+INFO:root:Average loss=0.19193602538522267
+INFO:root:
+[Valid][5][UTT][ MSE = 0.3062 | LCC = 0.9109 | SRCC = 0.9115 ]
+INFO:root:Average loss=0.1701524776401203
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2721 | LCC = 0.9122 | SRCC = 0.9117 ]
+INFO:root:Average loss=0.1710044368670855
+INFO:root:
+[Valid][7][UTT][ MSE = 0.3103 | LCC = 0.9134 | SRCC = 0.9131 ]
+INFO:root:Average loss=0.15933955434946656
+INFO:root:
+[Valid][8][UTT][ MSE = 0.3089 | LCC = 0.9144 | SRCC = 0.9139 ]
+INFO:root:Average loss=0.15687527916679492
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2615 | LCC = 0.9145 | SRCC = 0.9144 ]
+INFO:root:Average loss=0.14676090799315128
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2821 | LCC = 0.9119 | SRCC = 0.9127 ]
+INFO:root:Average loss=0.14207780417162558
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2868 | LCC = 0.9132 | SRCC = 0.9131 ]
+INFO:root:Average loss=0.13825235830221563
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2519 | LCC = 0.9128 | SRCC = 0.9133 ]
+INFO:root:Average loss=0.13033690579192486
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2666 | LCC = 0.9122 | SRCC = 0.9119 ]
+INFO:root:Average loss=0.1324447486011279
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2810 | LCC = 0.9122 | SRCC = 0.9128 ]
+INFO:root:Average loss=0.12349162681426616
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2894 | LCC = 0.9133 | SRCC = 0.9129 ]
+INFO:root:Average loss=0.11536422479531669
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2960 | LCC = 0.9128 | SRCC = 0.9122 ]
+INFO:root:Average loss=0.11254478904897768
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2810 | LCC = 0.9111 | SRCC = 0.9107 ]
+INFO:root:Average loss=0.10933323549052884
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2731 | LCC = 0.9121 | SRCC = 0.9120 ]
+INFO:root:Average loss=0.10361595055615971
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2826 | LCC = 0.9114 | SRCC = 0.9108 ]
+INFO:root:Average loss=0.09769086828293828
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2708 | LCC = 0.9097 | SRCC = 0.9107 ]
+INFO:root:Average loss=0.09289148779359856
+INFO:root:
+[Valid][21][UTT][ MSE = 0.3200 | LCC = 0.9095 | SRCC = 0.9090 ]
+INFO:root:Average loss=0.09156402581602852
+INFO:root:
+[Valid][22][UTT][ MSE = 0.3064 | LCC = 0.9095 | SRCC = 0.9094 ]
+INFO:root:Average loss=0.08494042456580725
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2911 | LCC = 0.9109 | SRCC = 0.9103 ]
+INFO:root:Average loss=0.08076360739896753
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2740 | LCC = 0.9080 | SRCC = 0.9077 ]
+INFO:root:Average loss=0.0782911600491215
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2834 | LCC = 0.9072 | SRCC = 0.9066 ]
+INFO:root:Average loss=0.07541190161009055
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2676 | LCC = 0.9076 | SRCC = 0.9076 ]
+INFO:root:Average loss=0.07637436261583615
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2567 | LCC = 0.9082 | SRCC = 0.9081 ]
+INFO:root:Average loss=0.06907574416843452
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2533 | LCC = 0.9071 | SRCC = 0.9070 ]
+INFO:root:Average loss=0.06985286586632618
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2829 | LCC = 0.9056 | SRCC = 0.9052 ]
+INFO:root:Average loss=0.06613142123025965
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2674 | LCC = 0.9061 | SRCC = 0.9062 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.3939 | LCC = 0.7498 | SRCC = 0.7340 ]

results_nisqa/w2v2_large_xlsr53/layer12/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_large_xlsr53/layer12/run4/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:694ba0b51da430b7ff1ab1751e8e95ee5cb3b03858a7e6372ad770efc2dfad7d
+size 3403339

results_nisqa/w2v2_large_xlsr53/layer12/run4/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1f504d1611a237665b0ec2e7916374c3b065d5e9794e853956d774c20afe6ed
+size 3405062

results_nisqa/w2v2_large_xlsr53/layer12/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=0.41416040579707636
+INFO:root:
+[Valid][0][UTT][ MSE = 0.2857 | LCC = 0.8947 | SRCC = 0.8971 ]
+INFO:root:Average loss=0.2352335542612682
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3078 | LCC = 0.9061 | SRCC = 0.9061 ]
+INFO:root:Average loss=0.21054434268116262
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2669 | LCC = 0.9082 | SRCC = 0.9082 ]
+INFO:root:Average loss=0.19278501925957686
+INFO:root:
+[Valid][3][UTT][ MSE = 0.3156 | LCC = 0.9098 | SRCC = 0.9104 ]
+INFO:root:Average loss=0.18586226918793827
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3069 | LCC = 0.9132 | SRCC = 0.9128 ]
+INFO:root:Average loss=0.17585081477902528
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2477 | LCC = 0.9121 | SRCC = 0.9132 ]
+INFO:root:Average loss=0.17084179168319427
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2410 | LCC = 0.9142 | SRCC = 0.9143 ]
+INFO:root:Average loss=0.16671691055414994
+INFO:root:
+[Valid][7][UTT][ MSE = 0.3057 | LCC = 0.9133 | SRCC = 0.9148 ]
+INFO:root:Average loss=0.1580005151601885
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2383 | LCC = 0.9167 | SRCC = 0.9163 ]
+INFO:root:Average loss=0.15242971746907757
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2643 | LCC = 0.9160 | SRCC = 0.9158 ]
+INFO:root:Average loss=0.14962458046357754
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2439 | LCC = 0.9152 | SRCC = 0.9157 ]
+INFO:root:Average loss=0.13922389114350942
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2734 | LCC = 0.9158 | SRCC = 0.9152 ]
+INFO:root:Average loss=0.13483966779329873
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2460 | LCC = 0.9164 | SRCC = 0.9160 ]
+INFO:root:Average loss=0.13005666262497104
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2298 | LCC = 0.9153 | SRCC = 0.9151 ]
+INFO:root:Average loss=0.12223739444003628
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2550 | LCC = 0.9157 | SRCC = 0.9152 ]
+INFO:root:Average loss=0.1214045338289586
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2336 | LCC = 0.9141 | SRCC = 0.9139 ]
+INFO:root:Average loss=0.1117067652070798
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2364 | LCC = 0.9138 | SRCC = 0.9134 ]
+INFO:root:Average loss=0.1091026964818122
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2527 | LCC = 0.9140 | SRCC = 0.9141 ]
+INFO:root:Average loss=0.10411690276740604
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2501 | LCC = 0.9145 | SRCC = 0.9139 ]
+INFO:root:Average loss=0.09875141241388515
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2520 | LCC = 0.9133 | SRCC = 0.9129 ]
+INFO:root:Average loss=0.10760923882009667
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2396 | LCC = 0.9145 | SRCC = 0.9138 ]
+INFO:root:Average loss=0.09167333849506572
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2329 | LCC = 0.9127 | SRCC = 0.9131 ]
+INFO:root:Average loss=0.08998436317106202
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2366 | LCC = 0.9136 | SRCC = 0.9131 ]
+INFO:root:Average loss=0.0842951462318787
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2380 | LCC = 0.9125 | SRCC = 0.9125 ]
+INFO:root:Average loss=0.08139305824489262
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2627 | LCC = 0.9126 | SRCC = 0.9122 ]
+INFO:root:Average loss=0.07829309293644966
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2313 | LCC = 0.9108 | SRCC = 0.9114 ]
+INFO:root:Average loss=0.0745239981121755
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2509 | LCC = 0.9116 | SRCC = 0.9111 ]
+INFO:root:Average loss=0.07258215023344652
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2394 | LCC = 0.9106 | SRCC = 0.9108 ]
+INFO:root:Average loss=0.07194543371483081
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2633 | LCC = 0.9105 | SRCC = 0.9098 ]
+INFO:root:Average loss=0.06775791381049708
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2808 | LCC = 0.9091 | SRCC = 0.9093 ]
+INFO:root:Average loss=0.06346831925084136
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2386 | LCC = 0.9110 | SRCC = 0.9105 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4112 | LCC = 0.7391 | SRCC = 0.7277 ]

results_nisqa/w2v2_large_xlsr53/layer14/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_large_xlsr53/layer14/run0/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:907210f164c4f5daa8564df691f481cf4c319967e43846fe4ac25961f661d8fa
+size 3403339

results_nisqa/w2v2_large_xlsr53/layer14/run0/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a20b8c90de19e7f7df73e1f0edcc56afcc212f41743a722953e7996869fbd12
+size 3405062

results_nisqa/w2v2_large_xlsr53/layer14/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=0.3791770879932911
+INFO:root:
+[Valid][0][UTT][ MSE = 0.4098 | LCC = 0.8955 | SRCC = 0.8981 ]
+INFO:root:Average loss=0.22926518110009286
+INFO:root:
+[Valid][1][UTT][ MSE = 0.2793 | LCC = 0.9013 | SRCC = 0.9037 ]
+INFO:root:Average loss=0.2162648769300108
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2743 | LCC = 0.9099 | SRCC = 0.9095 ]
+INFO:root:Average loss=0.20525923263647652
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2399 | LCC = 0.9101 | SRCC = 0.9103 ]
+INFO:root:Average loss=0.18892772034930355
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3938 | LCC = 0.9120 | SRCC = 0.9128 ]
+INFO:root:Average loss=0.18254828931106998
+INFO:root:
+[Valid][5][UTT][ MSE = 0.3286 | LCC = 0.9130 | SRCC = 0.9136 ]
+INFO:root:Average loss=0.17078013304685583
+INFO:root:
+[Valid][6][UTT][ MSE = 0.3138 | LCC = 0.9154 | SRCC = 0.9157 ]
+INFO:root:Average loss=0.1684514973204949
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2551 | LCC = 0.9166 | SRCC = 0.9164 ]
+INFO:root:Average loss=0.16073711453317907
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2758 | LCC = 0.9175 | SRCC = 0.9173 ]
+INFO:root:Average loss=0.15306419125526627
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2873 | LCC = 0.9160 | SRCC = 0.9161 ]
+INFO:root:Average loss=0.14727837695724014
+INFO:root:
+[Valid][10][UTT][ MSE = 0.3409 | LCC = 0.9129 | SRCC = 0.9159 ]
+INFO:root:Average loss=0.15242519440678504
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2713 | LCC = 0.9175 | SRCC = 0.9171 ]
+INFO:root:Average loss=0.1383383222613376
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2360 | LCC = 0.9171 | SRCC = 0.9172 ]
+INFO:root:Average loss=0.13247131633793
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2460 | LCC = 0.9162 | SRCC = 0.9156 ]
+INFO:root:Average loss=0.13252894851685948
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2460 | LCC = 0.9167 | SRCC = 0.9167 ]
+INFO:root:Average loss=0.1252457243255797
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2640 | LCC = 0.9153 | SRCC = 0.9159 ]
+INFO:root:Average loss=0.11886505985018835
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2294 | LCC = 0.9157 | SRCC = 0.9160 ]
+INFO:root:Average loss=0.11873209063028324
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2445 | LCC = 0.9166 | SRCC = 0.9173 ]
+INFO:root:Average loss=0.11207933680859604
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2917 | LCC = 0.9163 | SRCC = 0.9167 ]
+INFO:root:Average loss=0.10586290722551374
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2675 | LCC = 0.9141 | SRCC = 0.9139 ]
+INFO:root:Average loss=0.10274997174825971
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2431 | LCC = 0.9144 | SRCC = 0.9143 ]
+INFO:root:Average loss=0.10379111116504394
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2405 | LCC = 0.9153 | SRCC = 0.9150 ]
+INFO:root:Average loss=0.09900554067137614
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2481 | LCC = 0.9157 | SRCC = 0.9153 ]
+INFO:root:Average loss=0.09268046810017155
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2587 | LCC = 0.9139 | SRCC = 0.9145 ]
+INFO:root:Average loss=0.09012907052229595
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2534 | LCC = 0.9114 | SRCC = 0.9116 ]
+INFO:root:Average loss=0.08545121195257743
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2532 | LCC = 0.9126 | SRCC = 0.9123 ]
+INFO:root:Average loss=0.08638116256522306
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2429 | LCC = 0.9142 | SRCC = 0.9139 ]
+INFO:root:Average loss=0.08175373676023043
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2413 | LCC = 0.9118 | SRCC = 0.9120 ]
+INFO:root:Average loss=0.07706572438423344
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2464 | LCC = 0.9129 | SRCC = 0.9124 ]
+INFO:root:Average loss=0.0743916753270378
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2350 | LCC = 0.9135 | SRCC = 0.9128 ]
+INFO:root:Average loss=0.0726413448594209
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2590 | LCC = 0.9128 | SRCC = 0.9126 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4257 | LCC = 0.7352 | SRCC = 0.7218 ]

results_nisqa/w2v2_large_xlsr53/layer14/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_large_xlsr53/layer14/run1/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36b91cce00ca18f88e62fb7b1aba67a1442e74b66d5e6a60532f6e4e39f7f7f1
+size 3403339

results_nisqa/w2v2_large_xlsr53/layer14/run1/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3fda2b91bafcd832116e76645a16958328a8f7f61cc35f6ae7903ba9cefbd05
+size 3405062

results_nisqa/w2v2_large_xlsr53/layer14/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=0.4261210528067771
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3374 | LCC = 0.8923 | SRCC = 0.8931 ]
+INFO:root:Average loss=0.2311271959646589
+INFO:root:
+[Valid][1][UTT][ MSE = 0.2998 | LCC = 0.9030 | SRCC = 0.9037 ]
+INFO:root:Average loss=0.21287726839153753
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3002 | LCC = 0.9080 | SRCC = 0.9085 ]
+INFO:root:Average loss=0.20020959259285404
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2851 | LCC = 0.9107 | SRCC = 0.9112 ]
+INFO:root:Average loss=0.1868785330325882
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3467 | LCC = 0.9081 | SRCC = 0.9115 ]
+INFO:root:Average loss=0.1789675059518373
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2492 | LCC = 0.9138 | SRCC = 0.9139 ]
+INFO:root:Average loss=0.17277968584457573
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2453 | LCC = 0.9134 | SRCC = 0.9143 ]
+INFO:root:Average loss=0.16428255173512277
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2953 | LCC = 0.9117 | SRCC = 0.9140 ]
+INFO:root:Average loss=0.16041882297379434
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2408 | LCC = 0.9150 | SRCC = 0.9144 ]
+INFO:root:Average loss=0.1528915340414626
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2394 | LCC = 0.9153 | SRCC = 0.9156 ]
+INFO:root:Average loss=0.1471505684056723
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2480 | LCC = 0.9150 | SRCC = 0.9150 ]
+INFO:root:Average loss=0.15120059402519567
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2460 | LCC = 0.9143 | SRCC = 0.9152 ]
+INFO:root:Average loss=0.14327160385302726
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2539 | LCC = 0.9146 | SRCC = 0.9152 ]
+INFO:root:Average loss=0.13278373188710627
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2486 | LCC = 0.9155 | SRCC = 0.9161 ]
+INFO:root:Average loss=0.12705664263467567
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2676 | LCC = 0.9131 | SRCC = 0.9147 ]
+INFO:root:Average loss=0.12663086839666257
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2446 | LCC = 0.9123 | SRCC = 0.9143 ]
+INFO:root:Average loss=0.11722304753829978
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2437 | LCC = 0.9129 | SRCC = 0.9142 ]
+INFO:root:Average loss=0.11366021889396485
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2559 | LCC = 0.9135 | SRCC = 0.9132 ]
+INFO:root:Average loss=0.11358722872127687
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2545 | LCC = 0.9095 | SRCC = 0.9114 ]
+INFO:root:Average loss=0.10673875008532077
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2284 | LCC = 0.9128 | SRCC = 0.9127 ]
+INFO:root:Average loss=0.106672309683582
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2817 | LCC = 0.9100 | SRCC = 0.9109 ]
+INFO:root:Average loss=0.10157084581307593
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2364 | LCC = 0.9115 | SRCC = 0.9115 ]
+INFO:root:Average loss=0.09500453237853299
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2495 | LCC = 0.9083 | SRCC = 0.9102 ]
+INFO:root:Average loss=0.09273233475712682
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2435 | LCC = 0.9094 | SRCC = 0.9099 ]
+INFO:root:Average loss=0.08883656708878933
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2440 | LCC = 0.9095 | SRCC = 0.9103 ]
+INFO:root:Average loss=0.084707157675586
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2426 | LCC = 0.9099 | SRCC = 0.9110 ]
+INFO:root:Average loss=0.08089624058907431
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2367 | LCC = 0.9104 | SRCC = 0.9103 ]
+INFO:root:Average loss=0.07952034202097469
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2577 | LCC = 0.9105 | SRCC = 0.9104 ]
+INFO:root:Average loss=0.07809555099878697
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2477 | LCC = 0.9095 | SRCC = 0.9093 ]
+INFO:root:Average loss=0.07676186440566371
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2332 | LCC = 0.9095 | SRCC = 0.9092 ]
+INFO:root:Average loss=0.07511110605962704
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2446 | LCC = 0.9087 | SRCC = 0.9091 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4352 | LCC = 0.7288 | SRCC = 0.7170 ]

results_nisqa/w2v2_large_xlsr53/layer14/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_large_xlsr53/layer14/run2/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ac78493b6857fc4da5a11e9102b310abc8c6ccfba317de3ea1265e1783f626a
+size 3403339

results_nisqa/w2v2_large_xlsr53/layer14/run2/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c94a456b1647cada0d9c292e4c78b85995936b8d2ac820efaa196f47ebab46e5
+size 3405062

results_nisqa/w2v2_large_xlsr53/layer14/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=0.4328031659643085
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3669 | LCC = 0.8909 | SRCC = 0.8937 ]
+INFO:root:Average loss=0.23079833371101777
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3341 | LCC = 0.9016 | SRCC = 0.9030 ]
+INFO:root:Average loss=0.21412245395211127
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2782 | LCC = 0.9090 | SRCC = 0.9090 ]
+INFO:root:Average loss=0.2046452695665332
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2705 | LCC = 0.9109 | SRCC = 0.9106 ]
+INFO:root:Average loss=0.18974567394208358
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2551 | LCC = 0.9095 | SRCC = 0.9111 ]
+INFO:root:Average loss=0.18001785099161843
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2900 | LCC = 0.9142 | SRCC = 0.9139 ]
+INFO:root:Average loss=0.17494174204982085
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2540 | LCC = 0.9153 | SRCC = 0.9151 ]
+INFO:root:Average loss=0.16146793496402012
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2375 | LCC = 0.9147 | SRCC = 0.9154 ]
+INFO:root:Average loss=0.15804472656091514
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2681 | LCC = 0.9131 | SRCC = 0.9130 ]
+INFO:root:Average loss=0.15093822264774687
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2812 | LCC = 0.9146 | SRCC = 0.9150 ]
+INFO:root:Average loss=0.14987623700172226
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2802 | LCC = 0.9104 | SRCC = 0.9147 ]
+INFO:root:Average loss=0.1429780126909989
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2469 | LCC = 0.9159 | SRCC = 0.9155 ]
+INFO:root:Average loss=0.13940503857383837
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2684 | LCC = 0.9137 | SRCC = 0.9136 ]
+INFO:root:Average loss=0.13034706989120196
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2579 | LCC = 0.9148 | SRCC = 0.9143 ]
+INFO:root:Average loss=0.1329750814017533
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2738 | LCC = 0.9144 | SRCC = 0.9139 ]
+INFO:root:Average loss=0.1238376032022727
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2588 | LCC = 0.9113 | SRCC = 0.9121 ]
+INFO:root:Average loss=0.11883686000556615
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2638 | LCC = 0.9138 | SRCC = 0.9133 ]
+INFO:root:Average loss=0.11514426457744113
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2559 | LCC = 0.9103 | SRCC = 0.9121 ]
+INFO:root:Average loss=0.10790569944619444
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2473 | LCC = 0.9108 | SRCC = 0.9120 ]
+INFO:root:Average loss=0.10526645398897931
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2435 | LCC = 0.9105 | SRCC = 0.9107 ]
+INFO:root:Average loss=0.10034853870296753
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2541 | LCC = 0.9088 | SRCC = 0.9104 ]
+INFO:root:Average loss=0.09749430626889184
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2579 | LCC = 0.9092 | SRCC = 0.9101 ]
+INFO:root:Average loss=0.09449068491341751
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2426 | LCC = 0.9076 | SRCC = 0.9089 ]
+INFO:root:Average loss=0.09184361057560568
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2771 | LCC = 0.9060 | SRCC = 0.9073 ]
+INFO:root:Average loss=0.09074248348744963
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2822 | LCC = 0.9085 | SRCC = 0.9078 ]
+INFO:root:Average loss=0.088384607473033
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2650 | LCC = 0.9065 | SRCC = 0.9075 ]
+INFO:root:Average loss=0.07886680794847493
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2439 | LCC = 0.9045 | SRCC = 0.9077 ]
+INFO:root:Average loss=0.07903828868114879
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2469 | LCC = 0.9080 | SRCC = 0.9083 ]
+INFO:root:Average loss=0.07793877549426403
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2743 | LCC = 0.9073 | SRCC = 0.9071 ]
+INFO:root:Average loss=0.07448805584383837
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2569 | LCC = 0.9065 | SRCC = 0.9068 ]
+INFO:root:Average loss=0.07315749887140155
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2479 | LCC = 0.9071 | SRCC = 0.9065 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4029 | LCC = 0.7384 | SRCC = 0.7290 ]

results_nisqa/w2v2_large_xlsr53/layer14/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_large_xlsr53/layer14/run3/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:926b72153865ed1ca2b5df9b4e21f44865d4b20c701a1e79d35eb21f3e394fef
+size 3403339

results_nisqa/w2v2_large_xlsr53/layer14/run3/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f463eb53d2d654a744a4ea8dcb963a99d15ed89aeb36a8e29e5be7f32648ebf
+size 3405062

results_nisqa/w2v2_large_xlsr53/layer14/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=0.4066778216403344
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3551 | LCC = 0.8898 | SRCC = 0.8935 ]
+INFO:root:Average loss=0.22849988446414815
+INFO:root:
+[Valid][1][UTT][ MSE = 0.2875 | LCC = 0.9038 | SRCC = 0.9054 ]
+INFO:root:Average loss=0.20319303680706574
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2944 | LCC = 0.9029 | SRCC = 0.9078 ]
+INFO:root:Average loss=0.20117783714408818
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2814 | LCC = 0.9117 | SRCC = 0.9118 ]
+INFO:root:Average loss=0.18344302260117723
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2804 | LCC = 0.9125 | SRCC = 0.9127 ]
+INFO:root:Average loss=0.17964792914817787
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2734 | LCC = 0.9130 | SRCC = 0.9133 ]
+INFO:root:Average loss=0.17289037179912445
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2657 | LCC = 0.9133 | SRCC = 0.9140 ]
+INFO:root:Average loss=0.17087360247538957
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2501 | LCC = 0.9164 | SRCC = 0.9167 ]
+INFO:root:Average loss=0.15723626108872407
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2529 | LCC = 0.9168 | SRCC = 0.9174 ]
+INFO:root:Average loss=0.153476501389735
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2842 | LCC = 0.9147 | SRCC = 0.9161 ]
+INFO:root:Average loss=0.15478032775696993
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2650 | LCC = 0.9163 | SRCC = 0.9168 ]
+INFO:root:Average loss=0.14256984973057157
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2783 | LCC = 0.9167 | SRCC = 0.9163 ]
+INFO:root:Average loss=0.14197137460887777
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2584 | LCC = 0.9137 | SRCC = 0.9155 ]
+INFO:root:Average loss=0.13746371886805991
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2323 | LCC = 0.9168 | SRCC = 0.9171 ]
+INFO:root:Average loss=0.12725851839849714
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2511 | LCC = 0.9126 | SRCC = 0.9145 ]
+INFO:root:Average loss=0.1306926447876616
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2534 | LCC = 0.9124 | SRCC = 0.9151 ]
+INFO:root:Average loss=0.1179143189091903
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2413 | LCC = 0.9168 | SRCC = 0.9168 ]
+INFO:root:Average loss=0.11462032002520699
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2278 | LCC = 0.9151 | SRCC = 0.9155 ]
+INFO:root:Average loss=0.11006037636816157
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2740 | LCC = 0.9155 | SRCC = 0.9154 ]
+INFO:root:Average loss=0.1083012154767279
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2354 | LCC = 0.9133 | SRCC = 0.9154 ]
+INFO:root:Average loss=0.1048788479556238
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2574 | LCC = 0.9144 | SRCC = 0.9151 ]
+INFO:root:Average loss=0.09916292686510637
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2274 | LCC = 0.9140 | SRCC = 0.9149 ]
+INFO:root:Average loss=0.09440900047304314
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2582 | LCC = 0.9119 | SRCC = 0.9139 ]
+INFO:root:Average loss=0.09285229978533839
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2493 | LCC = 0.9139 | SRCC = 0.9137 ]
+INFO:root:Average loss=0.09128010918983834
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2349 | LCC = 0.9156 | SRCC = 0.9153 ]
+INFO:root:Average loss=0.08860236656114545
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2455 | LCC = 0.9124 | SRCC = 0.9136 ]
+INFO:root:Average loss=0.08155541593542678
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2371 | LCC = 0.9119 | SRCC = 0.9136 ]
+INFO:root:Average loss=0.07789533904005337
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2408 | LCC = 0.9118 | SRCC = 0.9122 ]
+INFO:root:Average loss=0.07783280504662866
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2368 | LCC = 0.9120 | SRCC = 0.9130 ]
+INFO:root:Average loss=0.07635035908463374
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2324 | LCC = 0.9128 | SRCC = 0.9128 ]
+INFO:root:Average loss=0.07226880088534658
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2597 | LCC = 0.9102 | SRCC = 0.9109 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4038 | LCC = 0.7372 | SRCC = 0.7222 ]

results_nisqa/w2v2_large_xlsr53/layer14/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_large_xlsr53/layer14/run4/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf5c30a03806916b95cdd02739919237216e0328f04f5761370a2f686fe178fe
+size 3403339

results_nisqa/w2v2_large_xlsr53/layer14/run4/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6df65c08a8cdc32516b377ce6e6e65ccce8d83615faf77188ba3c274ad5305af
+size 3405062

results_nisqa/w2v2_large_xlsr53/layer14/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=0.44836491510013626
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3962 | LCC = 0.8923 | SRCC = 0.8938 ]
+INFO:root:Average loss=0.2375784481019643
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3630 | LCC = 0.9021 | SRCC = 0.9041 ]
+INFO:root:Average loss=0.2115830568908956
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2711 | LCC = 0.9084 | SRCC = 0.9086 ]
+INFO:root:Average loss=0.1946315688291037
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2826 | LCC = 0.9105 | SRCC = 0.9113 ]
+INFO:root:Average loss=0.18088793104266845
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2612 | LCC = 0.9131 | SRCC = 0.9129 ]
+INFO:root:Average loss=0.17729542783402294
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2644 | LCC = 0.9133 | SRCC = 0.9130 ]
+INFO:root:Average loss=0.17127232942622522
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2851 | LCC = 0.9142 | SRCC = 0.9142 ]
+INFO:root:Average loss=0.16652182287218942
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2440 | LCC = 0.9152 | SRCC = 0.9149 ]
+INFO:root:Average loss=0.16227951408982966
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2511 | LCC = 0.9155 | SRCC = 0.9151 ]
+INFO:root:Average loss=0.15696514811302195
+INFO:root:
+[Valid][9][UTT][ MSE = 0.3046 | LCC = 0.9136 | SRCC = 0.9148 ]
+INFO:root:Average loss=0.1476474644217877
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2728 | LCC = 0.9138 | SRCC = 0.9143 ]
+INFO:root:Average loss=0.14164253617125439
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2487 | LCC = 0.9140 | SRCC = 0.9135 ]
+INFO:root:Average loss=0.13532037632486035
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2812 | LCC = 0.9144 | SRCC = 0.9142 ]
+INFO:root:Average loss=0.1321494513853437
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2423 | LCC = 0.9140 | SRCC = 0.9137 ]
+INFO:root:Average loss=0.12624686086453454
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2520 | LCC = 0.9145 | SRCC = 0.9147 ]
+INFO:root:Average loss=0.12294504955152556
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2314 | LCC = 0.9128 | SRCC = 0.9129 ]
+INFO:root:Average loss=0.12332842688512251
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2403 | LCC = 0.9149 | SRCC = 0.9145 ]
+INFO:root:Average loss=0.11501647872215061
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2198 | LCC = 0.9127 | SRCC = 0.9127 ]
+INFO:root:Average loss=0.10831912016937498
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2595 | LCC = 0.9106 | SRCC = 0.9119 ]
+INFO:root:Average loss=0.10690468338700389
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2392 | LCC = 0.9136 | SRCC = 0.9128 ]
+INFO:root:Average loss=0.10141312044566077
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2471 | LCC = 0.9125 | SRCC = 0.9123 ]
+INFO:root:Average loss=0.09841810974340907
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2532 | LCC = 0.9111 | SRCC = 0.9112 ]
+INFO:root:Average loss=0.09649028710891745
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2517 | LCC = 0.9122 | SRCC = 0.9123 ]
+INFO:root:Average loss=0.08909589683130986
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2439 | LCC = 0.9094 | SRCC = 0.9105 ]
+INFO:root:Average loss=0.0915194389792536
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2309 | LCC = 0.9114 | SRCC = 0.9107 ]
+INFO:root:Average loss=0.08605860914453606
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2402 | LCC = 0.9110 | SRCC = 0.9117 ]
+INFO:root:Average loss=0.0831039734830746
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2578 | LCC = 0.9083 | SRCC = 0.9087 ]
+INFO:root:Average loss=0.08027082843156909
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2592 | LCC = 0.9089 | SRCC = 0.9101 ]
+INFO:root:Average loss=0.07702724973676521
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2563 | LCC = 0.9086 | SRCC = 0.9092 ]
+INFO:root:Average loss=0.07454945274084979
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2429 | LCC = 0.9089 | SRCC = 0.9084 ]
+INFO:root:Average loss=0.07197922842867802
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2347 | LCC = 0.9098 | SRCC = 0.9097 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.4137 | LCC = 0.7373 | SRCC = 0.7246 ]

results_nisqa/w2v2_large_xlsr53/layer16/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_large_xlsr53/layer16/run0/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c56691a16585285b6578029871f69b4548c71cc2d1ae16ce8f9048d6e1fecd6a
+size 3403339

results_nisqa/w2v2_large_xlsr53/layer16/run0/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6f6e69d1dba0ecdcd23d744885fe82535caabc01c180fd46cc90719a4c443c6
+size 3405062