Upload SSL Layer MOS checkpoints (part 3)

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer0/run4/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 4974
+INFO:root:Num val speech clips: 1066
+INFO:root:Num test speech clips: 1066
+INFO:root:Device=cuda
+INFO:root:Average loss=0.535094215701788
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3842 | LCC = 0.7488 | SRCC = 0.7526 ]
+INFO:root:Average loss=0.29584163962266385
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3277 | LCC = 0.8040 | SRCC = 0.8071 ]
+INFO:root:Average loss=0.261383328300256
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3275 | LCC = 0.8241 | SRCC = 0.8261 ]
+INFO:root:Average loss=0.2500763161060138
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2768 | LCC = 0.8237 | SRCC = 0.8287 ]
+INFO:root:Average loss=0.22942434002955756
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2744 | LCC = 0.8337 | SRCC = 0.8325 ]
+INFO:root:Average loss=0.21366153036554655
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2711 | LCC = 0.8356 | SRCC = 0.8348 ]
+INFO:root:Average loss=0.1999274050960174
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2892 | LCC = 0.8433 | SRCC = 0.8431 ]
+INFO:root:Average loss=0.1926165959582879
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2597 | LCC = 0.8372 | SRCC = 0.8387 ]
+INFO:root:Average loss=0.18232641139855751
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2566 | LCC = 0.8441 | SRCC = 0.8439 ]
+INFO:root:Average loss=0.16798540166555306
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2583 | LCC = 0.8462 | SRCC = 0.8452 ]
+INFO:root:Average loss=0.16106477570839417
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2451 | LCC = 0.8465 | SRCC = 0.8461 ]
+INFO:root:Average loss=0.15572912628069902
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2486 | LCC = 0.8534 | SRCC = 0.8523 ]
+INFO:root:Average loss=0.16287972462865022
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2368 | LCC = 0.8494 | SRCC = 0.8497 ]
+INFO:root:Average loss=0.14000631782871026
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2341 | LCC = 0.8523 | SRCC = 0.8508 ]
+INFO:root:Average loss=0.13539448752999306
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2411 | LCC = 0.8472 | SRCC = 0.8475 ]
+INFO:root:Average loss=0.1342860907316208
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2522 | LCC = 0.8499 | SRCC = 0.8499 ]
+INFO:root:Average loss=0.12398458272218704
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2593 | LCC = 0.8503 | SRCC = 0.8515 ]
+INFO:root:Average loss=0.12395244598006591
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2810 | LCC = 0.8483 | SRCC = 0.8488 ]
+INFO:root:Average loss=0.11546754569579394
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2350 | LCC = 0.8511 | SRCC = 0.8520 ]
+INFO:root:Average loss=0.1048146407955732
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2434 | LCC = 0.8534 | SRCC = 0.8525 ]
+INFO:root:Average loss=0.1139298324019481
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2414 | LCC = 0.8502 | SRCC = 0.8479 ]
+INFO:root:Average loss=0.1061294471892791
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2416 | LCC = 0.8532 | SRCC = 0.8519 ]
+INFO:root:Average loss=0.10086686135484622
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2341 | LCC = 0.8540 | SRCC = 0.8543 ]
+INFO:root:Average loss=0.09473796456288068
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2398 | LCC = 0.8571 | SRCC = 0.8542 ]
+INFO:root:Average loss=0.09539232890193279
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2321 | LCC = 0.8554 | SRCC = 0.8546 ]
+INFO:root:Average loss=0.08975056835856193
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2351 | LCC = 0.8560 | SRCC = 0.8550 ]
+INFO:root:Average loss=0.08621321427516448
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2405 | LCC = 0.8486 | SRCC = 0.8482 ]
+INFO:root:Average loss=0.08316171355545521
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2407 | LCC = 0.8505 | SRCC = 0.8477 ]
+INFO:root:Average loss=0.0816903845526469
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2393 | LCC = 0.8546 | SRCC = 0.8528 ]
+INFO:root:Average loss=0.08002073800143523
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2402 | LCC = 0.8501 | SRCC = 0.8511 ]
+INFO:root:Average loss=0.07424558823307355
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2353 | LCC = 0.8522 | SRCC = 0.8511 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.2507 | LCC = 0.8542 | SRCC = 0.8528 ]

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run0/config.gin

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

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run0/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83707ae7e026bc38645ef7b66be6edf3f700d0a934aeafa84086212548c94877
+size 2747979

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run0/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c38695247b7bde782abd69802d97a48da16d37bbac8d04d4764c524c37f67b0
+size 2749702

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run0/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 4974
+INFO:root:Num val speech clips: 1066
+INFO:root:Num test speech clips: 1066
+INFO:root:Device=cuda
+INFO:root:Average loss=0.4579593210648268
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3317 | LCC = 0.7894 | SRCC = 0.7982 ]
+INFO:root:Average loss=0.2636304598970291
+INFO:root:
+[Valid][1][UTT][ MSE = 0.2871 | LCC = 0.8262 | SRCC = 0.8316 ]
+INFO:root:Average loss=0.23933588980864257
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2538 | LCC = 0.8485 | SRCC = 0.8497 ]
+INFO:root:Average loss=0.21618451598363045
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2379 | LCC = 0.8550 | SRCC = 0.8563 ]
+INFO:root:Average loss=0.20248253204119512
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2463 | LCC = 0.8576 | SRCC = 0.8598 ]
+INFO:root:Average loss=0.18415622499126655
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2249 | LCC = 0.8643 | SRCC = 0.8635 ]
+INFO:root:Average loss=0.17019824874706757
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2349 | LCC = 0.8703 | SRCC = 0.8698 ]
+INFO:root:Average loss=0.16008747092042214
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2346 | LCC = 0.8668 | SRCC = 0.8669 ]
+INFO:root:Average loss=0.15423229556435195
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2285 | LCC = 0.8652 | SRCC = 0.8663 ]
+INFO:root:Average loss=0.14377070342501005
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2259 | LCC = 0.8704 | SRCC = 0.8698 ]
+INFO:root:Average loss=0.13622465137487802
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2213 | LCC = 0.8665 | SRCC = 0.8661 ]
+INFO:root:Average loss=0.1270195396664815
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2703 | LCC = 0.8627 | SRCC = 0.8678 ]
+INFO:root:Average loss=0.11590135536896877
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2276 | LCC = 0.8687 | SRCC = 0.8693 ]
+INFO:root:Average loss=0.11665745041309258
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2237 | LCC = 0.8682 | SRCC = 0.8683 ]
+INFO:root:Average loss=0.10896223200819431
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2068 | LCC = 0.8725 | SRCC = 0.8744 ]
+INFO:root:Average loss=0.09942773428673927
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2140 | LCC = 0.8730 | SRCC = 0.8733 ]
+INFO:root:Average loss=0.10112338221799104
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2134 | LCC = 0.8728 | SRCC = 0.8741 ]
+INFO:root:Average loss=0.09263984352732316
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2332 | LCC = 0.8680 | SRCC = 0.8689 ]
+INFO:root:Average loss=0.09141098483441731
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2508 | LCC = 0.8667 | SRCC = 0.8691 ]
+INFO:root:Average loss=0.08199414028189121
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2179 | LCC = 0.8701 | SRCC = 0.8712 ]
+INFO:root:Average loss=0.07928904661765465
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2148 | LCC = 0.8712 | SRCC = 0.8726 ]
+INFO:root:Average loss=0.07552609296563344
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2072 | LCC = 0.8723 | SRCC = 0.8726 ]
+INFO:root:Average loss=0.0743144827011304
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2219 | LCC = 0.8695 | SRCC = 0.8708 ]
+INFO:root:Average loss=0.07293196237431122
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2231 | LCC = 0.8715 | SRCC = 0.8732 ]
+INFO:root:Average loss=0.07622888860985255
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2320 | LCC = 0.8689 | SRCC = 0.8730 ]
+INFO:root:Average loss=0.06728216910209411
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2217 | LCC = 0.8704 | SRCC = 0.8734 ]
+INFO:root:Average loss=0.06447688805369231
+INFO:root:
+[Valid][26][UTT][ MSE = 0.1995 | LCC = 0.8726 | SRCC = 0.8747 ]
+INFO:root:Average loss=0.06138085411527218
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2278 | LCC = 0.8717 | SRCC = 0.8733 ]
+INFO:root:Average loss=0.060740136565306246
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2234 | LCC = 0.8719 | SRCC = 0.8738 ]
+INFO:root:Average loss=0.06006152732058977
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2505 | LCC = 0.8666 | SRCC = 0.8701 ]
+INFO:root:Average loss=0.05618765596777965
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2154 | LCC = 0.8702 | SRCC = 0.8715 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.2533 | LCC = 0.8534 | SRCC = 0.8568 ]

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run1/config.gin

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

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run1/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bcf18d371a39cc8e99834efd83bca8e6612123bb36b50b9b9cfbc75379cb4ee2
+size 2747979

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run1/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2f13099fe68a0b06c2064d64a56fd47d6a13d3163ba4ed103d8983f8961c73c
+size 2749702

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run1/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 4974
+INFO:root:Num val speech clips: 1066
+INFO:root:Num test speech clips: 1066
+INFO:root:Device=cuda
+INFO:root:Average loss=0.472901660662431
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3769 | LCC = 0.7891 | SRCC = 0.7979 ]
+INFO:root:Average loss=0.277625910173624
+INFO:root:
+[Valid][1][UTT][ MSE = 0.2818 | LCC = 0.8286 | SRCC = 0.8308 ]
+INFO:root:Average loss=0.23368344532373625
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2876 | LCC = 0.8393 | SRCC = 0.8427 ]
+INFO:root:Average loss=0.21330232068132132
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2472 | LCC = 0.8549 | SRCC = 0.8539 ]
+INFO:root:Average loss=0.19755786466292846
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3097 | LCC = 0.8498 | SRCC = 0.8559 ]
+INFO:root:Average loss=0.1885475592735486
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2433 | LCC = 0.8633 | SRCC = 0.8624 ]
+INFO:root:Average loss=0.17586563212367204
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2377 | LCC = 0.8593 | SRCC = 0.8606 ]
+INFO:root:Average loss=0.1659340140147087
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2378 | LCC = 0.8615 | SRCC = 0.8613 ]
+INFO:root:Average loss=0.1488923263282348
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2663 | LCC = 0.8628 | SRCC = 0.8683 ]
+INFO:root:Average loss=0.15275137881055856
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2214 | LCC = 0.8642 | SRCC = 0.8657 ]
+INFO:root:Average loss=0.13369808355585122
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2617 | LCC = 0.8617 | SRCC = 0.8639 ]
+INFO:root:Average loss=0.13001980708959776
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2206 | LCC = 0.8684 | SRCC = 0.8684 ]
+INFO:root:Average loss=0.12487195126521282
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2604 | LCC = 0.8652 | SRCC = 0.8672 ]
+INFO:root:Average loss=0.1163460753667049
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2327 | LCC = 0.8630 | SRCC = 0.8645 ]
+INFO:root:Average loss=0.11134161312992756
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2354 | LCC = 0.8673 | SRCC = 0.8682 ]
+INFO:root:Average loss=0.10704351149690457
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2161 | LCC = 0.8674 | SRCC = 0.8677 ]
+INFO:root:Average loss=0.09886064767264403
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2221 | LCC = 0.8691 | SRCC = 0.8678 ]
+INFO:root:Average loss=0.09969308657141832
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2135 | LCC = 0.8706 | SRCC = 0.8712 ]
+INFO:root:Average loss=0.08704848606617023
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2210 | LCC = 0.8675 | SRCC = 0.8682 ]
+INFO:root:Average loss=0.08401912899735646
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2291 | LCC = 0.8651 | SRCC = 0.8669 ]
+INFO:root:Average loss=0.0786556480691219
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2226 | LCC = 0.8644 | SRCC = 0.8660 ]
+INFO:root:Average loss=0.08213168043547715
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2155 | LCC = 0.8654 | SRCC = 0.8665 ]
+INFO:root:Average loss=0.07721420019291915
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2155 | LCC = 0.8670 | SRCC = 0.8674 ]
+INFO:root:Average loss=0.07011394207485211
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2249 | LCC = 0.8628 | SRCC = 0.8641 ]
+INFO:root:Average loss=0.0725587058143738
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2239 | LCC = 0.8662 | SRCC = 0.8658 ]
+INFO:root:Average loss=0.06930257551945172
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2309 | LCC = 0.8661 | SRCC = 0.8669 ]
+INFO:root:Average loss=0.06451827975419852
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2640 | LCC = 0.8643 | SRCC = 0.8649 ]
+INFO:root:Average loss=0.06259272032632278
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2344 | LCC = 0.8678 | SRCC = 0.8670 ]
+INFO:root:Average loss=0.06423783187682812
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2112 | LCC = 0.8678 | SRCC = 0.8673 ]
+INFO:root:Average loss=0.06117730936369835
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2160 | LCC = 0.8653 | SRCC = 0.8665 ]
+INFO:root:Average loss=0.06286063513312584
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2296 | LCC = 0.8656 | SRCC = 0.8654 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.2231 | LCC = 0.8698 | SRCC = 0.8702 ]

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run2/config.gin

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

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run2/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5698ccc79830bf8d2ec4d180ca68ae59ec105762e6f3504c08e09d4802b51c75
+size 2747979

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run2/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a97c7380bd4c9abafd283364d20a5f92cdb2494d7b7f59496eb84af83d40692
+size 2749702

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run2/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 4974
+INFO:root:Num val speech clips: 1066
+INFO:root:Num test speech clips: 1066
+INFO:root:Device=cuda
+INFO:root:Average loss=0.456562588421198
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3457 | LCC = 0.7809 | SRCC = 0.7896 ]
+INFO:root:Average loss=0.27176349801130784
+INFO:root:
+[Valid][1][UTT][ MSE = 0.2868 | LCC = 0.8273 | SRCC = 0.8294 ]
+INFO:root:Average loss=0.2241117415519861
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3013 | LCC = 0.8404 | SRCC = 0.8424 ]
+INFO:root:Average loss=0.20278623757454065
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2491 | LCC = 0.8525 | SRCC = 0.8544 ]
+INFO:root:Average loss=0.1871246367883988
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2346 | LCC = 0.8585 | SRCC = 0.8594 ]
+INFO:root:Average loss=0.1779566798836757
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2373 | LCC = 0.8591 | SRCC = 0.8592 ]
+INFO:root:Average loss=0.17168924174247643
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2454 | LCC = 0.8611 | SRCC = 0.8645 ]
+INFO:root:Average loss=0.15440419946725553
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2183 | LCC = 0.8686 | SRCC = 0.8689 ]
+INFO:root:Average loss=0.14360607663790384
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2176 | LCC = 0.8673 | SRCC = 0.8680 ]
+INFO:root:Average loss=0.14387893571685523
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2319 | LCC = 0.8635 | SRCC = 0.8670 ]
+INFO:root:Average loss=0.13322913064024386
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2264 | LCC = 0.8681 | SRCC = 0.8699 ]
+INFO:root:Average loss=0.12076364944760616
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2313 | LCC = 0.8682 | SRCC = 0.8676 ]
+INFO:root:Average loss=0.11942552851560788
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2117 | LCC = 0.8689 | SRCC = 0.8685 ]
+INFO:root:Average loss=0.10900448539700264
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2188 | LCC = 0.8718 | SRCC = 0.8730 ]
+INFO:root:Average loss=0.10810819392402966
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2287 | LCC = 0.8662 | SRCC = 0.8682 ]
+INFO:root:Average loss=0.09844396683650139
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2190 | LCC = 0.8689 | SRCC = 0.8698 ]
+INFO:root:Average loss=0.09757571619672653
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2521 | LCC = 0.8655 | SRCC = 0.8682 ]
+INFO:root:Average loss=0.0925496702010815
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2077 | LCC = 0.8732 | SRCC = 0.8733 ]
+INFO:root:Average loss=0.09199598765908143
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2235 | LCC = 0.8685 | SRCC = 0.8689 ]
+INFO:root:Average loss=0.08461821456559193
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2049 | LCC = 0.8709 | SRCC = 0.8720 ]
+INFO:root:Average loss=0.07709161961116852
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2244 | LCC = 0.8666 | SRCC = 0.8685 ]
+INFO:root:Average loss=0.07737652537150261
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2106 | LCC = 0.8705 | SRCC = 0.8717 ]
+INFO:root:Average loss=0.07965540465636131
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2147 | LCC = 0.8681 | SRCC = 0.8702 ]
+INFO:root:Average loss=0.07509194949689584
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2256 | LCC = 0.8672 | SRCC = 0.8680 ]
+INFO:root:Average loss=0.06980926901675187
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2195 | LCC = 0.8704 | SRCC = 0.8706 ]
+INFO:root:Average loss=0.06937091277004817
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2093 | LCC = 0.8695 | SRCC = 0.8700 ]
+INFO:root:Average loss=0.06473130279053481
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2109 | LCC = 0.8704 | SRCC = 0.8702 ]
+INFO:root:Average loss=0.060205467188587554
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2164 | LCC = 0.8678 | SRCC = 0.8670 ]
+INFO:root:Average loss=0.062438682008248106
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2175 | LCC = 0.8685 | SRCC = 0.8691 ]
+INFO:root:Average loss=0.05673838244416775
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2158 | LCC = 0.8696 | SRCC = 0.8690 ]
+INFO:root:Average loss=0.056596591256749935
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2119 | LCC = 0.8704 | SRCC = 0.8711 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.2388 | LCC = 0.8582 | SRCC = 0.8600 ]

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run3/config.gin

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

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run3/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0eba625a314422aa88e9a1ecd81632d7937bd76362eb9ea56ead7772f523209
+size 2747979

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run3/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61f48af2af80f57e58dda2b5963a877ed8c312c34e6999ed5d8ff82ddb890627
+size 2749702

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run3/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 4974
+INFO:root:Num val speech clips: 1066
+INFO:root:Num test speech clips: 1066
+INFO:root:Device=cuda
+INFO:root:Average loss=0.4943407733853047
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3524 | LCC = 0.7888 | SRCC = 0.7867 ]
+INFO:root:Average loss=0.27113553709708726
+INFO:root:
+[Valid][1][UTT][ MSE = 0.2766 | LCC = 0.8275 | SRCC = 0.8295 ]
+INFO:root:Average loss=0.22598646409236467
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2490 | LCC = 0.8449 | SRCC = 0.8433 ]
+INFO:root:Average loss=0.20969594824008453
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2584 | LCC = 0.8529 | SRCC = 0.8518 ]
+INFO:root:Average loss=0.19203148696285027
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2432 | LCC = 0.8536 | SRCC = 0.8512 ]
+INFO:root:Average loss=0.17385386847532713
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2296 | LCC = 0.8589 | SRCC = 0.8576 ]
+INFO:root:Average loss=0.17610548025904557
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2539 | LCC = 0.8637 | SRCC = 0.8610 ]
+INFO:root:Average loss=0.1517601083868589
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2255 | LCC = 0.8622 | SRCC = 0.8612 ]
+INFO:root:Average loss=0.14952213164323416
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2327 | LCC = 0.8691 | SRCC = 0.8680 ]
+INFO:root:Average loss=0.13857012022382173
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2259 | LCC = 0.8649 | SRCC = 0.8642 ]
+INFO:root:Average loss=0.12525158346845552
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2173 | LCC = 0.8620 | SRCC = 0.8614 ]
+INFO:root:Average loss=0.12006656911510688
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2160 | LCC = 0.8677 | SRCC = 0.8668 ]
+INFO:root:Average loss=0.1218708648513525
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2141 | LCC = 0.8687 | SRCC = 0.8678 ]
+INFO:root:Average loss=0.11266429378436162
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2422 | LCC = 0.8687 | SRCC = 0.8682 ]
+INFO:root:Average loss=0.106711987262735
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2282 | LCC = 0.8607 | SRCC = 0.8608 ]
+INFO:root:Average loss=0.10513852952191463
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2131 | LCC = 0.8668 | SRCC = 0.8666 ]
+INFO:root:Average loss=0.09303858049978049
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2116 | LCC = 0.8670 | SRCC = 0.8668 ]
+INFO:root:Average loss=0.09211676195263863
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2060 | LCC = 0.8680 | SRCC = 0.8681 ]
+INFO:root:Average loss=0.08704451170678322
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2225 | LCC = 0.8684 | SRCC = 0.8677 ]
+INFO:root:Average loss=0.08433384123520973
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2149 | LCC = 0.8686 | SRCC = 0.8672 ]
+INFO:root:Average loss=0.08346272360246915
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2287 | LCC = 0.8673 | SRCC = 0.8672 ]
+INFO:root:Average loss=0.07591132619060002
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2127 | LCC = 0.8707 | SRCC = 0.8693 ]
+INFO:root:Average loss=0.07149577881090152
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2210 | LCC = 0.8657 | SRCC = 0.8657 ]
+INFO:root:Average loss=0.07121308506108247
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2178 | LCC = 0.8677 | SRCC = 0.8670 ]
+INFO:root:Average loss=0.0649825891909691
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2114 | LCC = 0.8667 | SRCC = 0.8651 ]
+INFO:root:Average loss=0.06367795747251083
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2120 | LCC = 0.8686 | SRCC = 0.8674 ]
+INFO:root:Average loss=0.07200747379698814
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2109 | LCC = 0.8666 | SRCC = 0.8662 ]
+INFO:root:Average loss=0.0621874925131217
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2297 | LCC = 0.8669 | SRCC = 0.8683 ]
+INFO:root:Average loss=0.0624477541886079
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2122 | LCC = 0.8678 | SRCC = 0.8683 ]
+INFO:root:Average loss=0.056709306744428783
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2117 | LCC = 0.8681 | SRCC = 0.8678 ]
+INFO:root:Average loss=0.05619487495949635
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2009 | LCC = 0.8730 | SRCC = 0.8729 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.2482 | LCC = 0.8523 | SRCC = 0.8560 ]

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run4/config.gin

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

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run4/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02a77faff56de2cb9601f358c6a22be1bc537428629a76cd52f06a0a1b0d1807
+size 2747979

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run4/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d371876e0229d5fc44c92ea47e96c4ee8ffce8a7c0f19cc19e09a506c739c43
+size 2749702

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer1/run4/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 4974
+INFO:root:Num val speech clips: 1066
+INFO:root:Num test speech clips: 1066
+INFO:root:Device=cuda
+INFO:root:Average loss=0.45782110687249744
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3364 | LCC = 0.7890 | SRCC = 0.7910 ]
+INFO:root:Average loss=0.2619916040163774
+INFO:root:
+[Valid][1][UTT][ MSE = 0.2988 | LCC = 0.8240 | SRCC = 0.8316 ]
+INFO:root:Average loss=0.22947889069716135
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2622 | LCC = 0.8463 | SRCC = 0.8454 ]
+INFO:root:Average loss=0.2089968497554461
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2518 | LCC = 0.8477 | SRCC = 0.8487 ]
+INFO:root:Average loss=0.1862188675082647
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2489 | LCC = 0.8596 | SRCC = 0.8589 ]
+INFO:root:Average loss=0.18314770083778945
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2721 | LCC = 0.8529 | SRCC = 0.8594 ]
+INFO:root:Average loss=0.17004422193918473
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2385 | LCC = 0.8630 | SRCC = 0.8632 ]
+INFO:root:Average loss=0.15617927106527182
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2328 | LCC = 0.8601 | SRCC = 0.8619 ]
+INFO:root:Average loss=0.14268474424114594
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2152 | LCC = 0.8698 | SRCC = 0.8701 ]
+INFO:root:Average loss=0.14190625714567992
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2288 | LCC = 0.8638 | SRCC = 0.8654 ]
+INFO:root:Average loss=0.13102695603783315
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2433 | LCC = 0.8660 | SRCC = 0.8668 ]
+INFO:root:Average loss=0.12423438301835305
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2445 | LCC = 0.8723 | SRCC = 0.8697 ]
+INFO:root:Average loss=0.122484947148806
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2148 | LCC = 0.8700 | SRCC = 0.8683 ]
+INFO:root:Average loss=0.1195953864699755
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2504 | LCC = 0.8649 | SRCC = 0.8659 ]
+INFO:root:Average loss=0.10542827290602219
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2221 | LCC = 0.8692 | SRCC = 0.8688 ]
+INFO:root:Average loss=0.10004530102014542
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2044 | LCC = 0.8710 | SRCC = 0.8700 ]
+INFO:root:Average loss=0.0951573085039854
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2163 | LCC = 0.8693 | SRCC = 0.8683 ]
+INFO:root:Average loss=0.09048761613667011
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2099 | LCC = 0.8701 | SRCC = 0.8685 ]
+INFO:root:Average loss=0.08737292069082077
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2151 | LCC = 0.8721 | SRCC = 0.8714 ]
+INFO:root:Average loss=0.08500232160664521
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2429 | LCC = 0.8642 | SRCC = 0.8667 ]
+INFO:root:Average loss=0.08119577207626441
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2391 | LCC = 0.8627 | SRCC = 0.8635 ]
+INFO:root:Average loss=0.07703164883722098
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2173 | LCC = 0.8674 | SRCC = 0.8678 ]
+INFO:root:Average loss=0.07261426412524322
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2099 | LCC = 0.8668 | SRCC = 0.8661 ]
+INFO:root:Average loss=0.07048633059438987
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2188 | LCC = 0.8637 | SRCC = 0.8639 ]
+INFO:root:Average loss=0.06686967668625024
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2222 | LCC = 0.8639 | SRCC = 0.8653 ]
+INFO:root:Average loss=0.06369337406104957
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2176 | LCC = 0.8690 | SRCC = 0.8680 ]
+INFO:root:Average loss=0.06694034291192508
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2187 | LCC = 0.8712 | SRCC = 0.8703 ]
+INFO:root:Average loss=0.061048720843898945
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2092 | LCC = 0.8686 | SRCC = 0.8680 ]
+INFO:root:Average loss=0.05901365908674705
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2183 | LCC = 0.8690 | SRCC = 0.8683 ]
+INFO:root:Average loss=0.061393342195795134
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2120 | LCC = 0.8744 | SRCC = 0.8735 ]
+INFO:root:Average loss=0.05789379297922819
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2063 | LCC = 0.8724 | SRCC = 0.8716 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.2476 | LCC = 0.8631 | SRCC = 0.8640 ]

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run0/config.gin

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

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run0/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7855e332683aa76d3b5be5834e3fc7c041067e5a51ff750804aa8e5b0b7507fd
+size 2747979

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run0/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e279789e2f65683e4ee41efa9edb3396389ceefaac008562f048901f7136fe8
+size 2749702

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run0/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 4974
+INFO:root:Num val speech clips: 1066
+INFO:root:Num test speech clips: 1066
+INFO:root:Device=cuda
+INFO:root:Average loss=0.35324714829524356
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3867 | LCC = 0.8282 | SRCC = 0.8278 ]
+INFO:root:Average loss=0.24930995989304322
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3611 | LCC = 0.8364 | SRCC = 0.8430 ]
+INFO:root:Average loss=0.23973685273757347
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3314 | LCC = 0.8430 | SRCC = 0.8420 ]
+INFO:root:Average loss=0.22885072460541359
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2966 | LCC = 0.8469 | SRCC = 0.8476 ]
+INFO:root:Average loss=0.22210966432705903
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3708 | LCC = 0.8439 | SRCC = 0.8506 ]
+INFO:root:Average loss=0.22060811863495752
+INFO:root:
+[Valid][5][UTT][ MSE = 0.3311 | LCC = 0.8595 | SRCC = 0.8568 ]
+INFO:root:Average loss=0.22270624855389962
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2953 | LCC = 0.8612 | SRCC = 0.8592 ]
+INFO:root:Average loss=0.20937233055249238
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2898 | LCC = 0.8619 | SRCC = 0.8623 ]
+INFO:root:Average loss=0.21042930583159128
+INFO:root:
+[Valid][8][UTT][ MSE = 0.3252 | LCC = 0.8604 | SRCC = 0.8590 ]
+INFO:root:Average loss=0.2150851547335967
+INFO:root:
+[Valid][9][UTT][ MSE = 0.3789 | LCC = 0.8601 | SRCC = 0.8623 ]
+INFO:root:Average loss=0.21118132865581757
+INFO:root:
+[Valid][10][UTT][ MSE = 0.3025 | LCC = 0.8604 | SRCC = 0.8638 ]
+INFO:root:Average loss=0.19872835182990783
+INFO:root:
+[Valid][11][UTT][ MSE = 0.3028 | LCC = 0.8637 | SRCC = 0.8611 ]
+INFO:root:Average loss=0.19554700817053133
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2758 | LCC = 0.8614 | SRCC = 0.8638 ]
+INFO:root:Average loss=0.19753452925345835
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2998 | LCC = 0.8589 | SRCC = 0.8631 ]
+INFO:root:Average loss=0.1991382343455767
+INFO:root:
+[Valid][14][UTT][ MSE = 0.3042 | LCC = 0.8594 | SRCC = 0.8608 ]
+INFO:root:Average loss=0.19103276347502685
+INFO:root:
+[Valid][15][UTT][ MSE = 0.3140 | LCC = 0.8643 | SRCC = 0.8620 ]
+INFO:root:Average loss=0.1848798216535495
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2835 | LCC = 0.8646 | SRCC = 0.8644 ]
+INFO:root:Average loss=0.18802716783606088
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2759 | LCC = 0.8637 | SRCC = 0.8650 ]
+INFO:root:Average loss=0.1843139182489652
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2932 | LCC = 0.8660 | SRCC = 0.8645 ]
+INFO:root:Average loss=0.1860717388872917
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2532 | LCC = 0.8625 | SRCC = 0.8660 ]
+INFO:root:Average loss=0.179396929171605
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2935 | LCC = 0.8668 | SRCC = 0.8657 ]
+INFO:root:Average loss=0.17457233312038276
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2671 | LCC = 0.8640 | SRCC = 0.8652 ]
+INFO:root:Average loss=0.1704600256605026
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2466 | LCC = 0.8642 | SRCC = 0.8668 ]
+INFO:root:Average loss=0.17291488508001351
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2811 | LCC = 0.8631 | SRCC = 0.8636 ]
+INFO:root:Average loss=0.1721521261601876
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2653 | LCC = 0.8657 | SRCC = 0.8651 ]
+INFO:root:Average loss=0.17146906065635192
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2940 | LCC = 0.8628 | SRCC = 0.8620 ]
+INFO:root:Average loss=0.1732069048552941
+INFO:root:
+[Valid][26][UTT][ MSE = 0.3067 | LCC = 0.8639 | SRCC = 0.8646 ]
+INFO:root:Average loss=0.16534294818456358
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2942 | LCC = 0.8665 | SRCC = 0.8649 ]
+INFO:root:Average loss=0.16810460035235453
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2733 | LCC = 0.8702 | SRCC = 0.8692 ]
+INFO:root:Average loss=0.1651376108519542
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2612 | LCC = 0.8649 | SRCC = 0.8674 ]
+INFO:root:Average loss=0.16043257235716551
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2681 | LCC = 0.8640 | SRCC = 0.8641 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.2663 | LCC = 0.8632 | SRCC = 0.8647 ]

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run1/config.gin

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

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run1/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:025648c1443cfec8343e325108a504abe1aaa78924862060cf31247479eb7d0c
+size 2747979

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run1/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9cf33e76e72c6b1970c247f95db3cc9ac7146b17ee8d41c356ab26c99ec55c42
+size 2749702

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run1/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 4974
+INFO:root:Num val speech clips: 1066
+INFO:root:Num test speech clips: 1066
+INFO:root:Device=cuda
+INFO:root:Average loss=0.3174068969793809
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3576 | LCC = 0.8231 | SRCC = 0.8222 ]
+INFO:root:Average loss=0.2588573136390784
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3749 | LCC = 0.8406 | SRCC = 0.8386 ]
+INFO:root:Average loss=0.2442090860926188
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3411 | LCC = 0.8382 | SRCC = 0.8405 ]
+INFO:root:Average loss=0.242265957288253
+INFO:root:
+[Valid][3][UTT][ MSE = 0.3309 | LCC = 0.8506 | SRCC = 0.8491 ]
+INFO:root:Average loss=0.22892723996669817
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3126 | LCC = 0.8479 | SRCC = 0.8493 ]
+INFO:root:Average loss=0.22753513814547124
+INFO:root:
+[Valid][5][UTT][ MSE = 0.3182 | LCC = 0.8536 | SRCC = 0.8520 ]
+INFO:root:Average loss=0.22084839871296516
+INFO:root:
+[Valid][6][UTT][ MSE = 0.3279 | LCC = 0.8569 | SRCC = 0.8551 ]
+INFO:root:Average loss=0.22092963239321342
+INFO:root:
+[Valid][7][UTT][ MSE = 0.3281 | LCC = 0.8614 | SRCC = 0.8591 ]
+INFO:root:Average loss=0.21114806830883026
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2746 | LCC = 0.8561 | SRCC = 0.8599 ]
+INFO:root:Average loss=0.21660359891561362
+INFO:root:
+[Valid][9][UTT][ MSE = 0.3215 | LCC = 0.8591 | SRCC = 0.8591 ]
+INFO:root:Average loss=0.2115032353844398
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2897 | LCC = 0.8616 | SRCC = 0.8603 ]
+INFO:root:Average loss=0.19968528281419706
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2724 | LCC = 0.8636 | SRCC = 0.8632 ]
+INFO:root:Average loss=0.20283982348747742
+INFO:root:
+[Valid][12][UTT][ MSE = 0.3356 | LCC = 0.8662 | SRCC = 0.8637 ]
+INFO:root:Average loss=0.20301509123200026
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2896 | LCC = 0.8636 | SRCC = 0.8628 ]
+INFO:root:Average loss=0.19392374377602187
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2697 | LCC = 0.8634 | SRCC = 0.8640 ]
+INFO:root:Average loss=0.1967244171179258
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2897 | LCC = 0.8649 | SRCC = 0.8657 ]
+INFO:root:Average loss=0.19277676042073813
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2672 | LCC = 0.8631 | SRCC = 0.8649 ]
+INFO:root:Average loss=0.18900141673974502
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2854 | LCC = 0.8669 | SRCC = 0.8681 ]
+INFO:root:Average loss=0.1896263198592724
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2882 | LCC = 0.8669 | SRCC = 0.8674 ]
+INFO:root:Average loss=0.18029660722002006
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2709 | LCC = 0.8642 | SRCC = 0.8683 ]
+INFO:root:Average loss=0.18510208517695084
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2649 | LCC = 0.8674 | SRCC = 0.8686 ]
+INFO:root:Average loss=0.18069163958231607
+INFO:root:
+[Valid][21][UTT][ MSE = 0.3011 | LCC = 0.8656 | SRCC = 0.8662 ]
+INFO:root:Average loss=0.18128735505235502
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2878 | LCC = 0.8646 | SRCC = 0.8660 ]
+INFO:root:Average loss=0.18800209577266985
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2614 | LCC = 0.8638 | SRCC = 0.8663 ]
+INFO:root:Average loss=0.17238368590672812
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2676 | LCC = 0.8688 | SRCC = 0.8696 ]
+INFO:root:Average loss=0.17248299058813316
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2626 | LCC = 0.8616 | SRCC = 0.8674 ]
+INFO:root:Average loss=0.17391084316067207
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2791 | LCC = 0.8660 | SRCC = 0.8675 ]
+INFO:root:Average loss=0.17232279985761031
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2713 | LCC = 0.8670 | SRCC = 0.8677 ]
+INFO:root:Average loss=0.16440686717247352
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2772 | LCC = 0.8658 | SRCC = 0.8672 ]
+INFO:root:Average loss=0.16360459018212098
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2662 | LCC = 0.8632 | SRCC = 0.8652 ]
+INFO:root:Average loss=0.1604236720655209
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2527 | LCC = 0.8659 | SRCC = 0.8682 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.2689 | LCC = 0.8565 | SRCC = 0.8595 ]

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run2/config.gin

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

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run2/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e2ea65ffa4da02a6470e6212da2d23521a072eacf68d5269d09fae8f38f043f
+size 2747979

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run2/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d83565dbaa3144282845251019596a4f2c5cbf885296d24db1d05c033c2d14d1
+size 2749702

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run2/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 4974
+INFO:root:Num val speech clips: 1066
+INFO:root:Num test speech clips: 1066
+INFO:root:Device=cuda
+INFO:root:Average loss=0.3141508494050075
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3923 | LCC = 0.8184 | SRCC = 0.8285 ]
+INFO:root:Average loss=0.2554331055054298
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3284 | LCC = 0.8396 | SRCC = 0.8396 ]
+INFO:root:Average loss=0.24269820195742142
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3514 | LCC = 0.8523 | SRCC = 0.8499 ]
+INFO:root:Average loss=0.23832531464405549
+INFO:root:
+[Valid][3][UTT][ MSE = 0.3120 | LCC = 0.8541 | SRCC = 0.8504 ]
+INFO:root:Average loss=0.23004326759240565
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3032 | LCC = 0.8579 | SRCC = 0.8555 ]
+INFO:root:Average loss=0.22105907400449118
+INFO:root:
+[Valid][5][UTT][ MSE = 0.3209 | LCC = 0.8551 | SRCC = 0.8531 ]
+INFO:root:Average loss=0.2214812878996898
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2872 | LCC = 0.8490 | SRCC = 0.8500 ]
+INFO:root:Average loss=0.21716442780616957
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2854 | LCC = 0.8513 | SRCC = 0.8523 ]
+INFO:root:Average loss=0.21841917167871427
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2818 | LCC = 0.8596 | SRCC = 0.8600 ]
+INFO:root:Average loss=0.2112439840267866
+INFO:root:
+[Valid][9][UTT][ MSE = 0.3440 | LCC = 0.8598 | SRCC = 0.8599 ]
+INFO:root:Average loss=0.21325892257766846
+INFO:root:
+[Valid][10][UTT][ MSE = 0.3217 | LCC = 0.8542 | SRCC = 0.8535 ]
+INFO:root:Average loss=0.21124051358455267
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2747 | LCC = 0.8650 | SRCC = 0.8646 ]
+INFO:root:Average loss=0.1968130994683657
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2713 | LCC = 0.8587 | SRCC = 0.8634 ]
+INFO:root:Average loss=0.19356833159541473
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2894 | LCC = 0.8633 | SRCC = 0.8632 ]
+INFO:root:Average loss=0.19821430895573053
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2775 | LCC = 0.8674 | SRCC = 0.8664 ]
+INFO:root:Average loss=0.2002851986923279
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2740 | LCC = 0.8679 | SRCC = 0.8664 ]
+INFO:root:Average loss=0.1913907012114158
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2871 | LCC = 0.8606 | SRCC = 0.8614 ]
+INFO:root:Average loss=0.19164187346513456
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2830 | LCC = 0.8692 | SRCC = 0.8669 ]
+INFO:root:Average loss=0.18783364225274476
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2903 | LCC = 0.8663 | SRCC = 0.8658 ]
+INFO:root:Average loss=0.1885135512894545
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2650 | LCC = 0.8640 | SRCC = 0.8640 ]
+INFO:root:Average loss=0.18124018838772407
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2583 | LCC = 0.8667 | SRCC = 0.8669 ]
+INFO:root:Average loss=0.18142137542749062
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2521 | LCC = 0.8619 | SRCC = 0.8626 ]
+INFO:root:Average loss=0.18358278303192213
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2609 | LCC = 0.8651 | SRCC = 0.8633 ]
+INFO:root:Average loss=0.17670650570056376
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2851 | LCC = 0.8628 | SRCC = 0.8602 ]
+INFO:root:Average loss=0.17500572498792258
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2729 | LCC = 0.8625 | SRCC = 0.8616 ]
+INFO:root:Average loss=0.17408147664406362
+INFO:root:
+[Valid][25][UTT][ MSE = 0.3144 | LCC = 0.8629 | SRCC = 0.8630 ]
+INFO:root:Average loss=0.17899597054108596
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2503 | LCC = 0.8634 | SRCC = 0.8646 ]
+INFO:root:Average loss=0.16612830137213072
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2560 | LCC = 0.8637 | SRCC = 0.8630 ]
+INFO:root:Average loss=0.1669961369763582
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2636 | LCC = 0.8639 | SRCC = 0.8647 ]
+INFO:root:Average loss=0.16910032335764322
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2576 | LCC = 0.8609 | SRCC = 0.8626 ]
+INFO:root:Average loss=0.16751875518224177
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2563 | LCC = 0.8542 | SRCC = 0.8557 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.2817 | LCC = 0.8629 | SRCC = 0.8658 ]

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run3/config.gin

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

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run3/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ad47deba415f4f454db1aa85e4c54abc8885c8ac975cbfc9cbfc6ddc8ab67795
+size 2747979

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run3/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:402bb1c181bf3ea0bc45fb73825d31536a27bea1d15dccf12d2e6b328db2d6ad
+size 2749702

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run3/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 4974
+INFO:root:Num val speech clips: 1066
+INFO:root:Num test speech clips: 1066
+INFO:root:Device=cuda
+INFO:root:Average loss=0.34359422612648743
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3381 | LCC = 0.8327 | SRCC = 0.8335 ]
+INFO:root:Average loss=0.25268365882146054
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3041 | LCC = 0.8411 | SRCC = 0.8399 ]
+INFO:root:Average loss=0.24073684922395608
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3230 | LCC = 0.8512 | SRCC = 0.8501 ]
+INFO:root:Average loss=0.2348229088462316
+INFO:root:
+[Valid][3][UTT][ MSE = 0.2999 | LCC = 0.8548 | SRCC = 0.8524 ]
+INFO:root:Average loss=0.235323510490931
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3086 | LCC = 0.8544 | SRCC = 0.8571 ]
+INFO:root:Average loss=0.22614780565102896
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2810 | LCC = 0.8524 | SRCC = 0.8514 ]
+INFO:root:Average loss=0.2186295346189768
+INFO:root:
+[Valid][6][UTT][ MSE = 0.2780 | LCC = 0.8590 | SRCC = 0.8589 ]
+INFO:root:Average loss=0.21325671061491355
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2519 | LCC = 0.8589 | SRCC = 0.8602 ]
+INFO:root:Average loss=0.2090612978507311
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2709 | LCC = 0.8612 | SRCC = 0.8603 ]
+INFO:root:Average loss=0.20827738482218522
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2621 | LCC = 0.8588 | SRCC = 0.8582 ]
+INFO:root:Average loss=0.2062528269031109
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2415 | LCC = 0.8616 | SRCC = 0.8613 ]
+INFO:root:Average loss=0.197889989003157
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2550 | LCC = 0.8632 | SRCC = 0.8612 ]
+INFO:root:Average loss=0.19881469240555397
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2649 | LCC = 0.8652 | SRCC = 0.8644 ]
+INFO:root:Average loss=0.19561036427815756
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2773 | LCC = 0.8651 | SRCC = 0.8631 ]
+INFO:root:Average loss=0.19029056471891892
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2441 | LCC = 0.8643 | SRCC = 0.8638 ]
+INFO:root:Average loss=0.1908005829422902
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2589 | LCC = 0.8649 | SRCC = 0.8657 ]
+INFO:root:Average loss=0.18912337643977922
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2542 | LCC = 0.8658 | SRCC = 0.8647 ]
+INFO:root:Average loss=0.19007801226316354
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2460 | LCC = 0.8642 | SRCC = 0.8631 ]
+INFO:root:Average loss=0.18176348832173225
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2564 | LCC = 0.8661 | SRCC = 0.8651 ]
+INFO:root:Average loss=0.1788630990836865
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2397 | LCC = 0.8640 | SRCC = 0.8662 ]
+INFO:root:Average loss=0.17991739702530396
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2408 | LCC = 0.8635 | SRCC = 0.8670 ]
+INFO:root:Average loss=0.1741629709991125
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2399 | LCC = 0.8669 | SRCC = 0.8685 ]
+INFO:root:Average loss=0.17182968671505267
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2403 | LCC = 0.8698 | SRCC = 0.8691 ]
+INFO:root:Average loss=0.17087039007590368
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2568 | LCC = 0.8661 | SRCC = 0.8643 ]
+INFO:root:Average loss=0.17264685989954534
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2772 | LCC = 0.8684 | SRCC = 0.8661 ]
+INFO:root:Average loss=0.16902533393257704
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2580 | LCC = 0.8659 | SRCC = 0.8658 ]
+INFO:root:Average loss=0.17027832032778326
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2435 | LCC = 0.8680 | SRCC = 0.8693 ]
+INFO:root:Average loss=0.16568645128072837
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2514 | LCC = 0.8690 | SRCC = 0.8695 ]
+INFO:root:Average loss=0.16094743938018113
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2324 | LCC = 0.8647 | SRCC = 0.8669 ]
+INFO:root:Average loss=0.1598964351683091
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2724 | LCC = 0.8676 | SRCC = 0.8667 ]
+INFO:root:Average loss=0.1596166241245392
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2283 | LCC = 0.8673 | SRCC = 0.8687 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.2485 | LCC = 0.8601 | SRCC = 0.8622 ]

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run4/config.gin

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

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run4/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfa1566f376a6968e46fdded79b9a9d30f4e98a681ee6884790bc80821e5b9c4
+size 2747979

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run4/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a623dda0c661b21e1456f5a03c03a4f605304923bbcb4f0de0a67efc525b548
+size 2749702

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer10/run4/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 4974
+INFO:root:Num val speech clips: 1066
+INFO:root:Num test speech clips: 1066
+INFO:root:Device=cuda
+INFO:root:Average loss=0.3513447588835007
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3417 | LCC = 0.8276 | SRCC = 0.8304 ]
+INFO:root:Average loss=0.25473815221817064
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3410 | LCC = 0.8433 | SRCC = 0.8437 ]
+INFO:root:Average loss=0.2443691394650019
+INFO:root:
+[Valid][2][UTT][ MSE = 0.2765 | LCC = 0.8447 | SRCC = 0.8477 ]
+INFO:root:Average loss=0.2353168160487444
+INFO:root:
+[Valid][3][UTT][ MSE = 0.3346 | LCC = 0.8560 | SRCC = 0.8527 ]
+INFO:root:Average loss=0.2238629882534345
+INFO:root:
+[Valid][4][UTT][ MSE = 0.2785 | LCC = 0.8559 | SRCC = 0.8575 ]
+INFO:root:Average loss=0.2155209752993706
+INFO:root:
+[Valid][5][UTT][ MSE = 0.2880 | LCC = 0.8587 | SRCC = 0.8590 ]
+INFO:root:Average loss=0.21565459630428216
+INFO:root:
+[Valid][6][UTT][ MSE = 0.3298 | LCC = 0.8565 | SRCC = 0.8595 ]
+INFO:root:Average loss=0.21336795523380622
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2992 | LCC = 0.8569 | SRCC = 0.8610 ]
+INFO:root:Average loss=0.2078454213646742
+INFO:root:
+[Valid][8][UTT][ MSE = 0.2593 | LCC = 0.8567 | SRCC = 0.8599 ]
+INFO:root:Average loss=0.20342941735035333
+INFO:root:
+[Valid][9][UTT][ MSE = 0.2910 | LCC = 0.8631 | SRCC = 0.8626 ]
+INFO:root:Average loss=0.2024319343841993
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2872 | LCC = 0.8618 | SRCC = 0.8637 ]
+INFO:root:Average loss=0.20285591301627648
+INFO:root:
+[Valid][11][UTT][ MSE = 0.2462 | LCC = 0.8623 | SRCC = 0.8655 ]
+INFO:root:Average loss=0.20194479804008436
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2472 | LCC = 0.8576 | SRCC = 0.8611 ]
+INFO:root:Average loss=0.19600733178548324
+INFO:root:
+[Valid][13][UTT][ MSE = 0.2877 | LCC = 0.8634 | SRCC = 0.8633 ]
+INFO:root:Average loss=0.19065598054574087
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2626 | LCC = 0.8643 | SRCC = 0.8658 ]
+INFO:root:Average loss=0.19070384431725892
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2757 | LCC = 0.8670 | SRCC = 0.8663 ]
+INFO:root:Average loss=0.18763475922437814
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2372 | LCC = 0.8641 | SRCC = 0.8665 ]
+INFO:root:Average loss=0.183171934233262
+INFO:root:
+[Valid][17][UTT][ MSE = 0.3073 | LCC = 0.8655 | SRCC = 0.8645 ]
+INFO:root:Average loss=0.1805742650459974
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2959 | LCC = 0.8655 | SRCC = 0.8659 ]
+INFO:root:Average loss=0.17919325016630003
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2687 | LCC = 0.8681 | SRCC = 0.8682 ]
+INFO:root:Average loss=0.17685179431469011
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2754 | LCC = 0.8618 | SRCC = 0.8654 ]
+INFO:root:Average loss=0.1755190059924737
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2870 | LCC = 0.8667 | SRCC = 0.8662 ]
+INFO:root:Average loss=0.17320412789017725
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2651 | LCC = 0.8644 | SRCC = 0.8630 ]
+INFO:root:Average loss=0.17252024960441467
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2881 | LCC = 0.8630 | SRCC = 0.8663 ]
+INFO:root:Average loss=0.1731349052144931
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2851 | LCC = 0.8594 | SRCC = 0.8622 ]
+INFO:root:Average loss=0.16869943321515352
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2651 | LCC = 0.8621 | SRCC = 0.8620 ]
+INFO:root:Average loss=0.16282657724924576
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2416 | LCC = 0.8639 | SRCC = 0.8650 ]
+INFO:root:Average loss=0.16413819732574317
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2694 | LCC = 0.8627 | SRCC = 0.8625 ]
+INFO:root:Average loss=0.160474983163369
+INFO:root:
+[Valid][28][UTT][ MSE = 0.2371 | LCC = 0.8634 | SRCC = 0.8649 ]
+INFO:root:Average loss=0.159651718269556
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2544 | LCC = 0.8606 | SRCC = 0.8618 ]
+INFO:root:Average loss=0.1596557400547541
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2655 | LCC = 0.8603 | SRCC = 0.8627 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.2828 | LCC = 0.8589 | SRCC = 0.8639 ]

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer11/run0/config.gin

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

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer11/run0/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d01b4451e54e22c9c9aac64ac367210df6ee5506cfed3fda0df890d2544a3d9
+size 2747979

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer11/run0/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f09cbf2e22bdc818bcb0e3841beede25be35ab2dedf89ad95dc3964888aae7f1
+size 2749702

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer11/run0/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 4974
+INFO:root:Num val speech clips: 1066
+INFO:root:Num test speech clips: 1066
+INFO:root:Device=cuda
+INFO:root:Average loss=0.33078792691230774
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3967 | LCC = 0.8191 | SRCC = 0.8127 ]
+INFO:root:Average loss=0.27789780440238804
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3707 | LCC = 0.8283 | SRCC = 0.8245 ]
+INFO:root:Average loss=0.26373726091323757
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3493 | LCC = 0.8369 | SRCC = 0.8353 ]
+INFO:root:Average loss=0.25764327267041576
+INFO:root:
+[Valid][3][UTT][ MSE = 0.3964 | LCC = 0.8469 | SRCC = 0.8444 ]
+INFO:root:Average loss=0.25970415274302167
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3312 | LCC = 0.8428 | SRCC = 0.8408 ]
+INFO:root:Average loss=0.2419386041852144
+INFO:root:
+[Valid][5][UTT][ MSE = 0.3288 | LCC = 0.8389 | SRCC = 0.8439 ]
+INFO:root:Average loss=0.24131066447649246
+INFO:root:
+[Valid][6][UTT][ MSE = 0.3517 | LCC = 0.8399 | SRCC = 0.8431 ]
+INFO:root:Average loss=0.24209309560366166
+INFO:root:
+[Valid][7][UTT][ MSE = 0.3284 | LCC = 0.8506 | SRCC = 0.8494 ]
+INFO:root:Average loss=0.2312618557077188
+INFO:root:
+[Valid][8][UTT][ MSE = 0.3576 | LCC = 0.8539 | SRCC = 0.8505 ]
+INFO:root:Average loss=0.2371799586675106
+INFO:root:
+[Valid][9][UTT][ MSE = 0.3117 | LCC = 0.8467 | SRCC = 0.8461 ]
+INFO:root:Average loss=0.2394746320369916
+INFO:root:
+[Valid][10][UTT][ MSE = 0.3395 | LCC = 0.8519 | SRCC = 0.8501 ]
+INFO:root:Average loss=0.22599803923796385
+INFO:root:
+[Valid][11][UTT][ MSE = 0.3133 | LCC = 0.8537 | SRCC = 0.8537 ]
+INFO:root:Average loss=0.22792509752206314
+INFO:root:
+[Valid][12][UTT][ MSE = 0.3548 | LCC = 0.8512 | SRCC = 0.8539 ]
+INFO:root:Average loss=0.22745798805203193
+INFO:root:
+[Valid][13][UTT][ MSE = 0.3159 | LCC = 0.8572 | SRCC = 0.8545 ]
+INFO:root:Average loss=0.22512680273025465
+INFO:root:
+[Valid][14][UTT][ MSE = 0.3208 | LCC = 0.8575 | SRCC = 0.8553 ]
+INFO:root:Average loss=0.22657086929449669
+INFO:root:
+[Valid][15][UTT][ MSE = 0.3144 | LCC = 0.8528 | SRCC = 0.8521 ]
+INFO:root:Average loss=0.21680787740609583
+INFO:root:
+[Valid][16][UTT][ MSE = 0.3172 | LCC = 0.8582 | SRCC = 0.8557 ]
+INFO:root:Average loss=0.22560458238690329
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2977 | LCC = 0.8565 | SRCC = 0.8568 ]
+INFO:root:Average loss=0.21467241090841782
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2953 | LCC = 0.8576 | SRCC = 0.8583 ]
+INFO:root:Average loss=0.21918566009173027
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2863 | LCC = 0.8585 | SRCC = 0.8582 ]
+INFO:root:Average loss=0.20934437387264693
+INFO:root:
+[Valid][20][UTT][ MSE = 0.3032 | LCC = 0.8609 | SRCC = 0.8580 ]
+INFO:root:Average loss=0.21085804777267653
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2974 | LCC = 0.8591 | SRCC = 0.8600 ]
+INFO:root:Average loss=0.21469746224391154
+INFO:root:
+[Valid][22][UTT][ MSE = 0.3163 | LCC = 0.8597 | SRCC = 0.8579 ]
+INFO:root:Average loss=0.21678824531726348
+INFO:root:
+[Valid][23][UTT][ MSE = 0.3375 | LCC = 0.8598 | SRCC = 0.8597 ]
+INFO:root:Average loss=0.21253518779308367
+INFO:root:
+[Valid][24][UTT][ MSE = 0.3090 | LCC = 0.8624 | SRCC = 0.8613 ]
+INFO:root:Average loss=0.20803204990732363
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2852 | LCC = 0.8608 | SRCC = 0.8610 ]
+INFO:root:Average loss=0.2030117167876317
+INFO:root:
+[Valid][26][UTT][ MSE = 0.2944 | LCC = 0.8617 | SRCC = 0.8611 ]
+INFO:root:Average loss=0.2059562731629763
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2873 | LCC = 0.8624 | SRCC = 0.8609 ]
+INFO:root:Average loss=0.21658280415412706
+INFO:root:
+[Valid][28][UTT][ MSE = 0.3111 | LCC = 0.8616 | SRCC = 0.8612 ]
+INFO:root:Average loss=0.20796056340138117
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2915 | LCC = 0.8640 | SRCC = 0.8621 ]
+INFO:root:Average loss=0.20105138449714735
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2922 | LCC = 0.8605 | SRCC = 0.8597 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.2717 | LCC = 0.8615 | SRCC = 0.8630 ]

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer11/run1/config.gin

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

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer11/run1/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98452ebb1345f88a5e91108016a92662ce8327e33a92f63715bf15a2377878d5
+size 2747979

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer11/run1/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:855db1b44d86923a0504edc0e932dd50fc60ad913adb683729f013134e0e04de
+size 2749638

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer11/run1/train.log

@@ -0,0 +1,99 @@
+INFO:root:Num train speech clips: 4974
+INFO:root:Num val speech clips: 1066
+INFO:root:Num test speech clips: 1066
+INFO:root:Device=cuda
+INFO:root:Average loss=0.32513359360969984
+INFO:root:
+[Valid][0][UTT][ MSE = 0.3488 | LCC = 0.8163 | SRCC = 0.8176 ]
+INFO:root:Average loss=0.2751318552555182
+INFO:root:
+[Valid][1][UTT][ MSE = 0.3305 | LCC = 0.8224 | SRCC = 0.8278 ]
+INFO:root:Average loss=0.266876003299004
+INFO:root:
+[Valid][2][UTT][ MSE = 0.3343 | LCC = 0.8360 | SRCC = 0.8354 ]
+INFO:root:Average loss=0.25654739408920973
+INFO:root:
+[Valid][3][UTT][ MSE = 0.3230 | LCC = 0.8402 | SRCC = 0.8444 ]
+INFO:root:Average loss=0.25022137031341213
+INFO:root:
+[Valid][4][UTT][ MSE = 0.3637 | LCC = 0.8288 | SRCC = 0.8409 ]
+INFO:root:Average loss=0.24391714808268425
+INFO:root:
+[Valid][5][UTT][ MSE = 0.3139 | LCC = 0.8413 | SRCC = 0.8424 ]
+INFO:root:Average loss=0.24498498745453665
+INFO:root:
+[Valid][6][UTT][ MSE = 0.3378 | LCC = 0.8471 | SRCC = 0.8459 ]
+INFO:root:Average loss=0.24725066583890182
+INFO:root:
+[Valid][7][UTT][ MSE = 0.2778 | LCC = 0.8446 | SRCC = 0.8459 ]
+INFO:root:Average loss=0.2337179992061395
+INFO:root:
+[Valid][8][UTT][ MSE = 0.3004 | LCC = 0.8492 | SRCC = 0.8486 ]
+INFO:root:Average loss=0.23140103064286402
+INFO:root:
+[Valid][9][UTT][ MSE = 0.3304 | LCC = 0.8477 | SRCC = 0.8528 ]
+INFO:root:Average loss=0.23007153681455514
+INFO:root:
+[Valid][10][UTT][ MSE = 0.2887 | LCC = 0.8547 | SRCC = 0.8525 ]
+INFO:root:Average loss=0.22550642108305907
+INFO:root:
+[Valid][11][UTT][ MSE = 0.3106 | LCC = 0.8449 | SRCC = 0.8516 ]
+INFO:root:Average loss=0.22466648045258644
+INFO:root:
+[Valid][12][UTT][ MSE = 0.2777 | LCC = 0.8539 | SRCC = 0.8561 ]
+INFO:root:Average loss=0.2247244686079331
+INFO:root:
+[Valid][13][UTT][ MSE = 0.3254 | LCC = 0.8336 | SRCC = 0.8429 ]
+INFO:root:Average loss=0.22946318487326303
+INFO:root:
+[Valid][14][UTT][ MSE = 0.2903 | LCC = 0.8522 | SRCC = 0.8529 ]
+INFO:root:Average loss=0.2135695755863801
+INFO:root:
+[Valid][15][UTT][ MSE = 0.2898 | LCC = 0.8583 | SRCC = 0.8584 ]
+INFO:root:Average loss=0.22677059070422098
+INFO:root:
+[Valid][16][UTT][ MSE = 0.2731 | LCC = 0.8488 | SRCC = 0.8538 ]
+INFO:root:Average loss=0.22322779645522436
+INFO:root:
+[Valid][17][UTT][ MSE = 0.2682 | LCC = 0.8475 | SRCC = 0.8524 ]
+INFO:root:Average loss=0.21604020855365655
+INFO:root:
+[Valid][18][UTT][ MSE = 0.2931 | LCC = 0.8534 | SRCC = 0.8542 ]
+INFO:root:Average loss=0.21095065371348307
+INFO:root:
+[Valid][19][UTT][ MSE = 0.2887 | LCC = 0.8579 | SRCC = 0.8583 ]
+INFO:root:Average loss=0.21357266681316572
+INFO:root:
+[Valid][20][UTT][ MSE = 0.2916 | LCC = 0.8598 | SRCC = 0.8584 ]
+INFO:root:Average loss=0.2227043398679831
+INFO:root:
+[Valid][21][UTT][ MSE = 0.2846 | LCC = 0.8516 | SRCC = 0.8568 ]
+INFO:root:Average loss=0.21614293028146792
+INFO:root:
+[Valid][22][UTT][ MSE = 0.2744 | LCC = 0.8600 | SRCC = 0.8583 ]
+INFO:root:Average loss=0.2121183169193757
+INFO:root:
+[Valid][23][UTT][ MSE = 0.2664 | LCC = 0.8540 | SRCC = 0.8584 ]
+INFO:root:Average loss=0.20658476134905449
+INFO:root:
+[Valid][24][UTT][ MSE = 0.2692 | LCC = 0.8579 | SRCC = 0.8588 ]
+INFO:root:Average loss=0.20228580117989808
+INFO:root:
+[Valid][25][UTT][ MSE = 0.2800 | LCC = 0.8580 | SRCC = 0.8596 ]
+INFO:root:Average loss=0.20654089290362138
+INFO:root:
+[Valid][26][UTT][ MSE = 0.3005 | LCC = 0.8612 | SRCC = 0.8594 ]
+INFO:root:Average loss=0.2029284200607202
+INFO:root:
+[Valid][27][UTT][ MSE = 0.2808 | LCC = 0.8602 | SRCC = 0.8626 ]
+INFO:root:Average loss=0.20251721306106982
+INFO:root:
+[Valid][28][UTT][ MSE = 0.3137 | LCC = 0.8626 | SRCC = 0.8630 ]
+INFO:root:Average loss=0.20582877605771407
+INFO:root:
+[Valid][29][UTT][ MSE = 0.2828 | LCC = 0.8641 | SRCC = 0.8637 ]
+INFO:root:Average loss=0.20313372577612215
+INFO:root:
+[Valid][30][UTT][ MSE = 0.2643 | LCC = 0.8533 | SRCC = 0.8589 ]
+INFO:root:
+[Test][31][UTT][ MSE = 0.2771 | LCC = 0.8658 | SRCC = 0.8662 ]

results_bvcc/w2v2_base_bvcc_finetuned_10epoch/layer11/run2/config.gin

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