diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5cf5a1a15d92851b12df160a155d9d1ce941d951 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e6b50c9387dcbe58e205f9a57677b09388adccb43bdcedee1f59d0478f4c8355 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..8255e492481d5087976c302bf60bba1c41627f79 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b99ff50d28a48dacfe0bf472bec9104893ed3d08d396214babc54bebba8b268b +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3f214bfc393246b5803258ffb3214aa15049e64e --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/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=0.3602916034750449 +INFO:root: +[Valid][0][UTT][ MSE = 0.3017 | LCC = 0.8297 | SRCC = 0.8342 ] +INFO:root:Average loss=0.22860305699018332 +INFO:root: +[Valid][1][UTT][ MSE = 0.2704 | LCC = 0.8552 | SRCC = 0.8523 ] +INFO:root:Average loss=0.2008502225463207 +INFO:root: +[Valid][2][UTT][ MSE = 0.2308 | LCC = 0.8624 | SRCC = 0.8625 ] +INFO:root:Average loss=0.185303858648508 +INFO:root: +[Valid][3][UTT][ MSE = 0.2760 | LCC = 0.8650 | SRCC = 0.8637 ] +INFO:root:Average loss=0.17254548395673433 +INFO:root: +[Valid][4][UTT][ MSE = 0.2495 | LCC = 0.8698 | SRCC = 0.8676 ] +INFO:root:Average loss=0.15667101721733045 +INFO:root: +[Valid][5][UTT][ MSE = 0.2468 | LCC = 0.8667 | SRCC = 0.8644 ] +INFO:root:Average loss=0.1475692986486814 +INFO:root: +[Valid][6][UTT][ MSE = 0.2355 | LCC = 0.8645 | SRCC = 0.8632 ] +INFO:root:Average loss=0.13897179391903755 +INFO:root: +[Valid][7][UTT][ MSE = 0.2490 | LCC = 0.8602 | SRCC = 0.8608 ] +INFO:root:Average loss=0.1250226007631192 +INFO:root: +[Valid][8][UTT][ MSE = 0.2292 | LCC = 0.8599 | SRCC = 0.8590 ] +INFO:root:Average loss=0.1178060746154724 +INFO:root: +[Valid][9][UTT][ MSE = 0.2492 | LCC = 0.8604 | SRCC = 0.8591 ] +INFO:root:Average loss=0.10731051575679046 +INFO:root: +[Valid][10][UTT][ MSE = 0.2352 | LCC = 0.8585 | SRCC = 0.8561 ] +INFO:root:Average loss=0.10309349201046504 +INFO:root: +[Valid][11][UTT][ MSE = 0.2484 | LCC = 0.8597 | SRCC = 0.8569 ] +INFO:root:Average loss=0.09537047563264003 +INFO:root: +[Valid][12][UTT][ MSE = 0.2379 | LCC = 0.8544 | SRCC = 0.8522 ] +INFO:root:Average loss=0.08652413985094963 +INFO:root: +[Valid][13][UTT][ MSE = 0.2674 | LCC = 0.8531 | SRCC = 0.8496 ] +INFO:root:Average loss=0.0837886471969959 +INFO:root: +[Valid][14][UTT][ MSE = 0.2418 | LCC = 0.8534 | SRCC = 0.8531 ] +INFO:root:Average loss=0.07590657128737523 +INFO:root: +[Valid][15][UTT][ MSE = 0.2412 | LCC = 0.8523 | SRCC = 0.8515 ] +INFO:root:Average loss=0.07344822246485795 +INFO:root: +[Valid][16][UTT][ MSE = 0.2441 | LCC = 0.8499 | SRCC = 0.8483 ] +INFO:root:Average loss=0.07031746958501828 +INFO:root: +[Valid][17][UTT][ MSE = 0.2497 | LCC = 0.8503 | SRCC = 0.8488 ] +INFO:root:Average loss=0.0652227365435698 +INFO:root: +[Valid][18][UTT][ MSE = 0.2388 | LCC = 0.8523 | SRCC = 0.8509 ] +INFO:root:Average loss=0.06416319434841473 +INFO:root: +[Valid][19][UTT][ MSE = 0.2686 | LCC = 0.8484 | SRCC = 0.8486 ] +INFO:root:Average loss=0.06138015863222954 +INFO:root: +[Valid][20][UTT][ MSE = 0.2499 | LCC = 0.8480 | SRCC = 0.8456 ] +INFO:root:Average loss=0.05661991114417712 +INFO:root: +[Valid][21][UTT][ MSE = 0.2570 | LCC = 0.8495 | SRCC = 0.8489 ] +INFO:root:Average loss=0.05485928946962723 +INFO:root: +[Valid][22][UTT][ MSE = 0.2561 | LCC = 0.8467 | SRCC = 0.8468 ] +INFO:root:Average loss=0.05111394521708672 +INFO:root: +[Valid][23][UTT][ MSE = 0.2419 | LCC = 0.8487 | SRCC = 0.8462 ] +INFO:root:Average loss=0.05058705133314316 +INFO:root: +[Valid][24][UTT][ MSE = 0.2489 | LCC = 0.8469 | SRCC = 0.8445 ] +INFO:root:Average loss=0.04935907319379158 +INFO:root: +[Valid][25][UTT][ MSE = 0.2400 | LCC = 0.8457 | SRCC = 0.8447 ] +INFO:root:Average loss=0.04663730957187139 +INFO:root: +[Valid][26][UTT][ MSE = 0.2469 | LCC = 0.8478 | SRCC = 0.8456 ] +INFO:root:Average loss=0.04633009780007295 +INFO:root: +[Valid][27][UTT][ MSE = 0.2476 | LCC = 0.8467 | SRCC = 0.8461 ] +INFO:root:Average loss=0.04485997715248512 +INFO:root: +[Valid][28][UTT][ MSE = 0.2528 | LCC = 0.8471 | SRCC = 0.8462 ] +INFO:root:Average loss=0.04893341567367315 +INFO:root: +[Valid][29][UTT][ MSE = 0.2455 | LCC = 0.8453 | SRCC = 0.8441 ] +INFO:root:Average loss=0.04172439204576688 +INFO:root: +[Valid][30][UTT][ MSE = 0.2377 | LCC = 0.8469 | SRCC = 0.8457 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2426 | LCC = 0.8652 | SRCC = 0.8651 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5d9d92d46aafc9b2232843a6108c72903f11854e --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:69884306410746ecb8677bfefed6eb0b64c36c6dc02b6daca2cad34720312c19 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..73d0cd3661cccb7dfcf4b8ebf322a4a2bb6cb943 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4bc767021def3f224f03464651aad93bce7804f579a71f1711116c8bafb7ab1a +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6829a06255621145e5d9131341b0071069b840e0 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/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=0.362190163288361 +INFO:root: +[Valid][0][UTT][ MSE = 0.3009 | LCC = 0.8301 | SRCC = 0.8381 ] +INFO:root:Average loss=0.2284203271071116 +INFO:root: +[Valid][1][UTT][ MSE = 0.2673 | LCC = 0.8539 | SRCC = 0.8531 ] +INFO:root:Average loss=0.20273739557999831 +INFO:root: +[Valid][2][UTT][ MSE = 0.2613 | LCC = 0.8658 | SRCC = 0.8636 ] +INFO:root:Average loss=0.18265455999435523 +INFO:root: +[Valid][3][UTT][ MSE = 0.2410 | LCC = 0.8670 | SRCC = 0.8649 ] +INFO:root:Average loss=0.17725740506862983 +INFO:root: +[Valid][4][UTT][ MSE = 0.2656 | LCC = 0.8689 | SRCC = 0.8669 ] +INFO:root:Average loss=0.1582477321036351 +INFO:root: +[Valid][5][UTT][ MSE = 0.2515 | LCC = 0.8697 | SRCC = 0.8684 ] +INFO:root:Average loss=0.14821337889402342 +INFO:root: +[Valid][6][UTT][ MSE = 0.2149 | LCC = 0.8709 | SRCC = 0.8709 ] +INFO:root:Average loss=0.14272961555383143 +INFO:root: +[Valid][7][UTT][ MSE = 0.2318 | LCC = 0.8722 | SRCC = 0.8715 ] +INFO:root:Average loss=0.12586095403784361 +INFO:root: +[Valid][8][UTT][ MSE = 0.2161 | LCC = 0.8669 | SRCC = 0.8674 ] +INFO:root:Average loss=0.12062422826122014 +INFO:root: +[Valid][9][UTT][ MSE = 0.2265 | LCC = 0.8707 | SRCC = 0.8704 ] +INFO:root:Average loss=0.11422893978082217 +INFO:root: +[Valid][10][UTT][ MSE = 0.2333 | LCC = 0.8650 | SRCC = 0.8649 ] +INFO:root:Average loss=0.10612992885021064 +INFO:root: +[Valid][11][UTT][ MSE = 0.2235 | LCC = 0.8654 | SRCC = 0.8653 ] +INFO:root:Average loss=0.09691693519170468 +INFO:root: +[Valid][12][UTT][ MSE = 0.2206 | LCC = 0.8645 | SRCC = 0.8653 ] +INFO:root:Average loss=0.09305107727264747 +INFO:root: +[Valid][13][UTT][ MSE = 0.2303 | LCC = 0.8633 | SRCC = 0.8627 ] +INFO:root:Average loss=0.0835542368869751 +INFO:root: +[Valid][14][UTT][ MSE = 0.2274 | LCC = 0.8627 | SRCC = 0.8636 ] +INFO:root:Average loss=0.08460547426381172 +INFO:root: +[Valid][15][UTT][ MSE = 0.2450 | LCC = 0.8617 | SRCC = 0.8615 ] +INFO:root:Average loss=0.07684859504493383 +INFO:root: +[Valid][16][UTT][ MSE = 0.2259 | LCC = 0.8592 | SRCC = 0.8591 ] +INFO:root:Average loss=0.07778260527321926 +INFO:root: +[Valid][17][UTT][ MSE = 0.2414 | LCC = 0.8588 | SRCC = 0.8589 ] +INFO:root:Average loss=0.06990714949101974 +INFO:root: +[Valid][18][UTT][ MSE = 0.2353 | LCC = 0.8605 | SRCC = 0.8596 ] +INFO:root:Average loss=0.06567355474600425 +INFO:root: +[Valid][19][UTT][ MSE = 0.2199 | LCC = 0.8611 | SRCC = 0.8613 ] +INFO:root:Average loss=0.06059241590973658 +INFO:root: +[Valid][20][UTT][ MSE = 0.2355 | LCC = 0.8560 | SRCC = 0.8553 ] +INFO:root:Average loss=0.058565773404179476 +INFO:root: +[Valid][21][UTT][ MSE = 0.2288 | LCC = 0.8588 | SRCC = 0.8592 ] +INFO:root:Average loss=0.05620270083921078 +INFO:root: +[Valid][22][UTT][ MSE = 0.2361 | LCC = 0.8612 | SRCC = 0.8620 ] +INFO:root:Average loss=0.05402580209267445 +INFO:root: +[Valid][23][UTT][ MSE = 0.2293 | LCC = 0.8571 | SRCC = 0.8579 ] +INFO:root:Average loss=0.0521805359958074 +INFO:root: +[Valid][24][UTT][ MSE = 0.2434 | LCC = 0.8537 | SRCC = 0.8538 ] +INFO:root:Average loss=0.0494673982835733 +INFO:root: +[Valid][25][UTT][ MSE = 0.2347 | LCC = 0.8553 | SRCC = 0.8554 ] +INFO:root:Average loss=0.047855974055635624 +INFO:root: +[Valid][26][UTT][ MSE = 0.2351 | LCC = 0.8556 | SRCC = 0.8555 ] +INFO:root:Average loss=0.045182778691061035 +INFO:root: +[Valid][27][UTT][ MSE = 0.2422 | LCC = 0.8575 | SRCC = 0.8570 ] +INFO:root:Average loss=0.04461886026920416 +INFO:root: +[Valid][28][UTT][ MSE = 0.2312 | LCC = 0.8586 | SRCC = 0.8585 ] +INFO:root:Average loss=0.04863246611486643 +INFO:root: +[Valid][29][UTT][ MSE = 0.2319 | LCC = 0.8594 | SRCC = 0.8594 ] +INFO:root:Average loss=0.04162825722820484 +INFO:root: +[Valid][30][UTT][ MSE = 0.2430 | LCC = 0.8544 | SRCC = 0.8547 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2513 | LCC = 0.8650 | SRCC = 0.8636 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..015f000d68cae322066ec51b1dc1906f9a124d8f --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc0bad2deb9bb1684da3dabc3f58b3b06de9ec6bf59170afbc4d9e27cf501f80 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d7f853b887924ef06fcfb30fd25b1e9aef04eef0 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:532c95da86fd5cb69f939eae1706a57c03831d7bf58fbf91b39978c90f6bbfd0 +size 2749638 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0164bb8cf31c38258bf07929002f00f8593f3b9f --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer5/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=0.42477776454045224 +INFO:root: +[Valid][0][UTT][ MSE = 0.3218 | LCC = 0.8238 | SRCC = 0.8294 ] +INFO:root:Average loss=0.23074349073263314 +INFO:root: +[Valid][1][UTT][ MSE = 0.2693 | LCC = 0.8502 | SRCC = 0.8498 ] +INFO:root:Average loss=0.21214958547781676 +INFO:root: +[Valid][2][UTT][ MSE = 0.2734 | LCC = 0.8590 | SRCC = 0.8568 ] +INFO:root:Average loss=0.18804034265952232 +INFO:root: +[Valid][3][UTT][ MSE = 0.2621 | LCC = 0.8589 | SRCC = 0.8581 ] +INFO:root:Average loss=0.16857376904823843 +INFO:root: +[Valid][4][UTT][ MSE = 0.2477 | LCC = 0.8623 | SRCC = 0.8611 ] +INFO:root:Average loss=0.15654911922338682 +INFO:root: +[Valid][5][UTT][ MSE = 0.2614 | LCC = 0.8590 | SRCC = 0.8580 ] +INFO:root:Average loss=0.144945139113145 +INFO:root: +[Valid][6][UTT][ MSE = 0.2632 | LCC = 0.8569 | SRCC = 0.8559 ] +INFO:root:Average loss=0.13225655763959274 +INFO:root: +[Valid][7][UTT][ MSE = 0.2594 | LCC = 0.8559 | SRCC = 0.8553 ] +INFO:root:Average loss=0.1209458742195215 +INFO:root: +[Valid][8][UTT][ MSE = 0.2373 | LCC = 0.8545 | SRCC = 0.8548 ] +INFO:root:Average loss=0.11163341435484397 +INFO:root: +[Valid][9][UTT][ MSE = 0.2365 | LCC = 0.8569 | SRCC = 0.8569 ] +INFO:root:Average loss=0.10707223711487575 +INFO:root: +[Valid][10][UTT][ MSE = 0.2548 | LCC = 0.8537 | SRCC = 0.8537 ] +INFO:root:Average loss=0.09564200382775222 +INFO:root: +[Valid][11][UTT][ MSE = 0.2528 | LCC = 0.8518 | SRCC = 0.8520 ] +INFO:root:Average loss=0.09240232288646393 +INFO:root: +[Valid][12][UTT][ MSE = 0.2525 | LCC = 0.8513 | SRCC = 0.8511 ] +INFO:root:Average loss=0.08303699680627921 +INFO:root: +[Valid][13][UTT][ MSE = 0.2726 | LCC = 0.8468 | SRCC = 0.8463 ] +INFO:root:Average loss=0.08223164812303506 +INFO:root: +[Valid][14][UTT][ MSE = 0.2445 | LCC = 0.8524 | SRCC = 0.8526 ] +INFO:root:Average loss=0.07442832212799634 +INFO:root: +[Valid][15][UTT][ MSE = 0.2502 | LCC = 0.8453 | SRCC = 0.8476 ] +INFO:root:Average loss=0.07438317820047721 +INFO:root: +[Valid][16][UTT][ MSE = 0.2584 | LCC = 0.8472 | SRCC = 0.8471 ] +INFO:root:Average loss=0.06566222432332161 +INFO:root: +[Valid][17][UTT][ MSE = 0.2409 | LCC = 0.8496 | SRCC = 0.8500 ] +INFO:root:Average loss=0.0633346375842125 +INFO:root: +[Valid][18][UTT][ MSE = 0.2424 | LCC = 0.8464 | SRCC = 0.8473 ] +INFO:root:Average loss=0.062345233507072315 +INFO:root: +[Valid][19][UTT][ MSE = 0.2516 | LCC = 0.8499 | SRCC = 0.8480 ] +INFO:root:Average loss=0.06149311549961567 +INFO:root: +[Valid][20][UTT][ MSE = 0.2590 | LCC = 0.8481 | SRCC = 0.8481 ] +INFO:root:Average loss=0.057354543883448995 +INFO:root: +[Valid][21][UTT][ MSE = 0.2502 | LCC = 0.8477 | SRCC = 0.8473 ] +INFO:root:Average loss=0.05565970307454849 +INFO:root: +[Valid][22][UTT][ MSE = 0.2667 | LCC = 0.8479 | SRCC = 0.8477 ] +INFO:root:Average loss=0.05516067046958666 +INFO:root: +[Valid][23][UTT][ MSE = 0.2483 | LCC = 0.8464 | SRCC = 0.8475 ] +INFO:root:Average loss=0.04986106446729256 +INFO:root: +[Valid][24][UTT][ MSE = 0.2603 | LCC = 0.8465 | SRCC = 0.8480 ] +INFO:root:Average loss=0.050457364425827295 +INFO:root: +[Valid][25][UTT][ MSE = 0.2401 | LCC = 0.8512 | SRCC = 0.8521 ] +INFO:root:Average loss=0.04614992647503431 +INFO:root: +[Valid][26][UTT][ MSE = 0.2470 | LCC = 0.8502 | SRCC = 0.8497 ] +INFO:root:Average loss=0.04329039013156524 +INFO:root: +[Valid][27][UTT][ MSE = 0.2521 | LCC = 0.8483 | SRCC = 0.8483 ] +INFO:root:Average loss=0.0468870301802571 +INFO:root: +[Valid][28][UTT][ MSE = 0.2459 | LCC = 0.8471 | SRCC = 0.8469 ] +INFO:root:Average loss=0.041542455864449344 +INFO:root: +[Valid][29][UTT][ MSE = 0.2544 | LCC = 0.8480 | SRCC = 0.8482 ] +INFO:root:Average loss=0.04111270451297363 +INFO:root: +[Valid][30][UTT][ MSE = 0.2684 | LCC = 0.8404 | SRCC = 0.8402 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2458 | LCC = 0.8641 | SRCC = 0.8660 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d5ad32bd763c99d734da6e613e8c3e03be83d6f9 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e4ce8cbfd7b297a9bc73cf79d8557e7f9b95f747a18d6c73a5b0c34fd755f043 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..01906d8d25bf77880c9c6ff7e06c58f599fadb55 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9fc80e491c15e22e6525277680d7e91ea160cb9d89a4e88013c7bc5c8dafb32b +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..cc7a24509b147e384be51dbef2a35bbf4ff9528e --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/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=0.37467708037449765 +INFO:root: +[Valid][0][UTT][ MSE = 0.2990 | LCC = 0.8281 | SRCC = 0.8343 ] +INFO:root:Average loss=0.2333670946267935 +INFO:root: +[Valid][1][UTT][ MSE = 0.3143 | LCC = 0.8528 | SRCC = 0.8531 ] +INFO:root:Average loss=0.20628967842994592 +INFO:root: +[Valid][2][UTT][ MSE = 0.2688 | LCC = 0.8577 | SRCC = 0.8596 ] +INFO:root:Average loss=0.18531899908796334 +INFO:root: +[Valid][3][UTT][ MSE = 0.2608 | LCC = 0.8651 | SRCC = 0.8653 ] +INFO:root:Average loss=0.17028685458577597 +INFO:root: +[Valid][4][UTT][ MSE = 0.2490 | LCC = 0.8692 | SRCC = 0.8693 ] +INFO:root:Average loss=0.15651398706130493 +INFO:root: +[Valid][5][UTT][ MSE = 0.2452 | LCC = 0.8712 | SRCC = 0.8696 ] +INFO:root:Average loss=0.14767660764165413 +INFO:root: +[Valid][6][UTT][ MSE = 0.2434 | LCC = 0.8708 | SRCC = 0.8693 ] +INFO:root:Average loss=0.1454870343590394 +INFO:root: +[Valid][7][UTT][ MSE = 0.2332 | LCC = 0.8654 | SRCC = 0.8666 ] +INFO:root:Average loss=0.13086320966100082 +INFO:root: +[Valid][8][UTT][ MSE = 0.2651 | LCC = 0.8646 | SRCC = 0.8659 ] +INFO:root:Average loss=0.13209669798230514 +INFO:root: +[Valid][9][UTT][ MSE = 0.2353 | LCC = 0.8653 | SRCC = 0.8648 ] +INFO:root:Average loss=0.11594731981555621 +INFO:root: +[Valid][10][UTT][ MSE = 0.2512 | LCC = 0.8638 | SRCC = 0.8637 ] +INFO:root:Average loss=0.10751021524461415 +INFO:root: +[Valid][11][UTT][ MSE = 0.2630 | LCC = 0.8627 | SRCC = 0.8650 ] +INFO:root:Average loss=0.10847543781766525 +INFO:root: +[Valid][12][UTT][ MSE = 0.2342 | LCC = 0.8638 | SRCC = 0.8646 ] +INFO:root:Average loss=0.09495962396837197 +INFO:root: +[Valid][13][UTT][ MSE = 0.2558 | LCC = 0.8581 | SRCC = 0.8585 ] +INFO:root:Average loss=0.09150164330808017 +INFO:root: +[Valid][14][UTT][ MSE = 0.2514 | LCC = 0.8579 | SRCC = 0.8582 ] +INFO:root:Average loss=0.08520652950765231 +INFO:root: +[Valid][15][UTT][ MSE = 0.2599 | LCC = 0.8562 | SRCC = 0.8576 ] +INFO:root:Average loss=0.07926359218664658 +INFO:root: +[Valid][16][UTT][ MSE = 0.2289 | LCC = 0.8552 | SRCC = 0.8571 ] +INFO:root:Average loss=0.07587479799985886 +INFO:root: +[Valid][17][UTT][ MSE = 0.2405 | LCC = 0.8530 | SRCC = 0.8549 ] +INFO:root:Average loss=0.07495530842779538 +INFO:root: +[Valid][18][UTT][ MSE = 0.2702 | LCC = 0.8545 | SRCC = 0.8561 ] +INFO:root:Average loss=0.06786195890834698 +INFO:root: +[Valid][19][UTT][ MSE = 0.2363 | LCC = 0.8514 | SRCC = 0.8527 ] +INFO:root:Average loss=0.06390005674881813 +INFO:root: +[Valid][20][UTT][ MSE = 0.2880 | LCC = 0.8501 | SRCC = 0.8523 ] +INFO:root:Average loss=0.06459033708923902 +INFO:root: +[Valid][21][UTT][ MSE = 0.2764 | LCC = 0.8492 | SRCC = 0.8524 ] +INFO:root:Average loss=0.0626679030366433 +INFO:root: +[Valid][22][UTT][ MSE = 0.2336 | LCC = 0.8538 | SRCC = 0.8561 ] +INFO:root:Average loss=0.057929873609772094 +INFO:root: +[Valid][23][UTT][ MSE = 0.2522 | LCC = 0.8498 | SRCC = 0.8517 ] +INFO:root:Average loss=0.054576437609891094 +INFO:root: +[Valid][24][UTT][ MSE = 0.2479 | LCC = 0.8518 | SRCC = 0.8535 ] +INFO:root:Average loss=0.05767752265987488 +INFO:root: +[Valid][25][UTT][ MSE = 0.2480 | LCC = 0.8496 | SRCC = 0.8512 ] +INFO:root:Average loss=0.05206704005981103 +INFO:root: +[Valid][26][UTT][ MSE = 0.2765 | LCC = 0.8493 | SRCC = 0.8518 ] +INFO:root:Average loss=0.04931540763339935 +INFO:root: +[Valid][27][UTT][ MSE = 0.2568 | LCC = 0.8494 | SRCC = 0.8503 ] +INFO:root:Average loss=0.04757789340920937 +INFO:root: +[Valid][28][UTT][ MSE = 0.2551 | LCC = 0.8521 | SRCC = 0.8531 ] +INFO:root:Average loss=0.04776382002119835 +INFO:root: +[Valid][29][UTT][ MSE = 0.2409 | LCC = 0.8512 | SRCC = 0.8522 ] +INFO:root:Average loss=0.045480646479588285 +INFO:root: +[Valid][30][UTT][ MSE = 0.2519 | LCC = 0.8506 | SRCC = 0.8530 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2367 | LCC = 0.8643 | SRCC = 0.8639 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5a38ab52a27a65ff5c8d05ef8fcce8a756991c6a --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4da41d337bab409df9120bda5f858c3242ee8a66301e702c145e23f333a0aa28 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..86d340d5a892d95511d5f783e4342797e07c3b87 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:56b570322548468a0e5d392aae1ef40f04c7c330bd8bbc864435d326f85340d6 +size 2749638 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d5839dfd37ca4b0701a5fdea5f954307217efafa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/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=0.3797180706109756 +INFO:root: +[Valid][0][UTT][ MSE = 0.3046 | LCC = 0.8346 | SRCC = 0.8365 ] +INFO:root:Average loss=0.2305016590234561 +INFO:root: +[Valid][1][UTT][ MSE = 0.2970 | LCC = 0.8560 | SRCC = 0.8567 ] +INFO:root:Average loss=0.21048951024810472 +INFO:root: +[Valid][2][UTT][ MSE = 0.2487 | LCC = 0.8620 | SRCC = 0.8611 ] +INFO:root:Average loss=0.1981869487044139 +INFO:root: +[Valid][3][UTT][ MSE = 0.2646 | LCC = 0.8666 | SRCC = 0.8650 ] +INFO:root:Average loss=0.18076091221509835 +INFO:root: +[Valid][4][UTT][ MSE = 0.2333 | LCC = 0.8649 | SRCC = 0.8658 ] +INFO:root:Average loss=0.16797982777158418 +INFO:root: +[Valid][5][UTT][ MSE = 0.2682 | LCC = 0.8642 | SRCC = 0.8651 ] +INFO:root:Average loss=0.1563210419546335 +INFO:root: +[Valid][6][UTT][ MSE = 0.2478 | LCC = 0.8661 | SRCC = 0.8650 ] +INFO:root:Average loss=0.1436955794118918 +INFO:root: +[Valid][7][UTT][ MSE = 0.2476 | LCC = 0.8625 | SRCC = 0.8622 ] +INFO:root:Average loss=0.1337132291534008 +INFO:root: +[Valid][8][UTT][ MSE = 0.2725 | LCC = 0.8544 | SRCC = 0.8546 ] +INFO:root:Average loss=0.13001818515551397 +INFO:root: +[Valid][9][UTT][ MSE = 0.2334 | LCC = 0.8595 | SRCC = 0.8588 ] +INFO:root:Average loss=0.11326501470727798 +INFO:root: +[Valid][10][UTT][ MSE = 0.2451 | LCC = 0.8579 | SRCC = 0.8586 ] +INFO:root:Average loss=0.10822764717233486 +INFO:root: +[Valid][11][UTT][ MSE = 0.2424 | LCC = 0.8543 | SRCC = 0.8550 ] +INFO:root:Average loss=0.10454240384010169 +INFO:root: +[Valid][12][UTT][ MSE = 0.2552 | LCC = 0.8536 | SRCC = 0.8525 ] +INFO:root:Average loss=0.09520155578278579 +INFO:root: +[Valid][13][UTT][ MSE = 0.2694 | LCC = 0.8526 | SRCC = 0.8515 ] +INFO:root:Average loss=0.08673519899065678 +INFO:root: +[Valid][14][UTT][ MSE = 0.2543 | LCC = 0.8538 | SRCC = 0.8536 ] +INFO:root:Average loss=0.08294013228553992 +INFO:root: +[Valid][15][UTT][ MSE = 0.2458 | LCC = 0.8495 | SRCC = 0.8504 ] +INFO:root:Average loss=0.07720066425509942 +INFO:root: +[Valid][16][UTT][ MSE = 0.2380 | LCC = 0.8510 | SRCC = 0.8518 ] +INFO:root:Average loss=0.07973640894469543 +INFO:root: +[Valid][17][UTT][ MSE = 0.2383 | LCC = 0.8531 | SRCC = 0.8532 ] +INFO:root:Average loss=0.07244081924167964 +INFO:root: +[Valid][18][UTT][ MSE = 0.2476 | LCC = 0.8460 | SRCC = 0.8452 ] +INFO:root:Average loss=0.06955408964019555 +INFO:root: +[Valid][19][UTT][ MSE = 0.2450 | LCC = 0.8496 | SRCC = 0.8494 ] +INFO:root:Average loss=0.06167036619706032 +INFO:root: +[Valid][20][UTT][ MSE = 0.2696 | LCC = 0.8440 | SRCC = 0.8436 ] +INFO:root:Average loss=0.06148498204465096 +INFO:root: +[Valid][21][UTT][ MSE = 0.2755 | LCC = 0.8439 | SRCC = 0.8430 ] +INFO:root:Average loss=0.06650885165883945 +INFO:root: +[Valid][22][UTT][ MSE = 0.2547 | LCC = 0.8477 | SRCC = 0.8471 ] +INFO:root:Average loss=0.057192399859046325 +INFO:root: +[Valid][23][UTT][ MSE = 0.2527 | LCC = 0.8450 | SRCC = 0.8445 ] +INFO:root:Average loss=0.05703940180440744 +INFO:root: +[Valid][24][UTT][ MSE = 0.2541 | LCC = 0.8491 | SRCC = 0.8501 ] +INFO:root:Average loss=0.05327783687374531 +INFO:root: +[Valid][25][UTT][ MSE = 0.2460 | LCC = 0.8481 | SRCC = 0.8479 ] +INFO:root:Average loss=0.049764264231690995 +INFO:root: +[Valid][26][UTT][ MSE = 0.2582 | LCC = 0.8453 | SRCC = 0.8447 ] +INFO:root:Average loss=0.047981671296442166 +INFO:root: +[Valid][27][UTT][ MSE = 0.2657 | LCC = 0.8437 | SRCC = 0.8419 ] +INFO:root:Average loss=0.04447451112075494 +INFO:root: +[Valid][28][UTT][ MSE = 0.2835 | LCC = 0.8452 | SRCC = 0.8450 ] +INFO:root:Average loss=0.04357468714125645 +INFO:root: +[Valid][29][UTT][ MSE = 0.2536 | LCC = 0.8481 | SRCC = 0.8473 ] +INFO:root:Average loss=0.0441230527865581 +INFO:root: +[Valid][30][UTT][ MSE = 0.2505 | LCC = 0.8434 | SRCC = 0.8419 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2562 | LCC = 0.8669 | SRCC = 0.8678 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9480f8a97351920663638e82b0b136f3fe3fba69 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c3a340a3c9b00ee154b8cc01d4c70a8607131a1149c8cc7465131a4d8f472467 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d28665b7e08e5b22a03481d740502572904103ba --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e949f710450eb03c290163913d30d804fd3d0b850d3fd878d4e4f335753ea46f +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..04316ec823ee5913f5cf2e7d9e7ab4e27259507c --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/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=0.35811989085796553 +INFO:root: +[Valid][0][UTT][ MSE = 0.2855 | LCC = 0.8356 | SRCC = 0.8394 ] +INFO:root:Average loss=0.23174308660702828 +INFO:root: +[Valid][1][UTT][ MSE = 0.2696 | LCC = 0.8587 | SRCC = 0.8571 ] +INFO:root:Average loss=0.20694455829186317 +INFO:root: +[Valid][2][UTT][ MSE = 0.2411 | LCC = 0.8652 | SRCC = 0.8645 ] +INFO:root:Average loss=0.19412387449007767 +INFO:root: +[Valid][3][UTT][ MSE = 0.2507 | LCC = 0.8690 | SRCC = 0.8672 ] +INFO:root:Average loss=0.17476236848877028 +INFO:root: +[Valid][4][UTT][ MSE = 0.2181 | LCC = 0.8662 | SRCC = 0.8685 ] +INFO:root:Average loss=0.16277041458166563 +INFO:root: +[Valid][5][UTT][ MSE = 0.2422 | LCC = 0.8689 | SRCC = 0.8669 ] +INFO:root:Average loss=0.14799939946104318 +INFO:root: +[Valid][6][UTT][ MSE = 0.2368 | LCC = 0.8692 | SRCC = 0.8682 ] +INFO:root:Average loss=0.14131302386522293 +INFO:root: +[Valid][7][UTT][ MSE = 0.2461 | LCC = 0.8671 | SRCC = 0.8668 ] +INFO:root:Average loss=0.12920137915091637 +INFO:root: +[Valid][8][UTT][ MSE = 0.2390 | LCC = 0.8635 | SRCC = 0.8640 ] +INFO:root:Average loss=0.12187185606513268 +INFO:root: +[Valid][9][UTT][ MSE = 0.2271 | LCC = 0.8669 | SRCC = 0.8675 ] +INFO:root:Average loss=0.1094062059926681 +INFO:root: +[Valid][10][UTT][ MSE = 0.2259 | LCC = 0.8613 | SRCC = 0.8616 ] +INFO:root:Average loss=0.10354185056609985 +INFO:root: +[Valid][11][UTT][ MSE = 0.2506 | LCC = 0.8603 | SRCC = 0.8608 ] +INFO:root:Average loss=0.09743791331465428 +INFO:root: +[Valid][12][UTT][ MSE = 0.2289 | LCC = 0.8613 | SRCC = 0.8616 ] +INFO:root:Average loss=0.09324750246909949 +INFO:root: +[Valid][13][UTT][ MSE = 0.2974 | LCC = 0.8539 | SRCC = 0.8557 ] +INFO:root:Average loss=0.08660409962519622 +INFO:root: +[Valid][14][UTT][ MSE = 0.2674 | LCC = 0.8559 | SRCC = 0.8572 ] +INFO:root:Average loss=0.0802232102992443 +INFO:root: +[Valid][15][UTT][ MSE = 0.2495 | LCC = 0.8566 | SRCC = 0.8564 ] +INFO:root:Average loss=0.07793291405034371 +INFO:root: +[Valid][16][UTT][ MSE = 0.2467 | LCC = 0.8566 | SRCC = 0.8556 ] +INFO:root:Average loss=0.07416404583133183 +INFO:root: +[Valid][17][UTT][ MSE = 0.2503 | LCC = 0.8575 | SRCC = 0.8589 ] +INFO:root:Average loss=0.07246936599795635 +INFO:root: +[Valid][18][UTT][ MSE = 0.2352 | LCC = 0.8545 | SRCC = 0.8548 ] +INFO:root:Average loss=0.06706997064443734 +INFO:root: +[Valid][19][UTT][ MSE = 0.2268 | LCC = 0.8561 | SRCC = 0.8555 ] +INFO:root:Average loss=0.06369093447350539 +INFO:root: +[Valid][20][UTT][ MSE = 0.2389 | LCC = 0.8555 | SRCC = 0.8549 ] +INFO:root:Average loss=0.057506588072730944 +INFO:root: +[Valid][21][UTT][ MSE = 0.2496 | LCC = 0.8510 | SRCC = 0.8510 ] +INFO:root:Average loss=0.06347753059787628 +INFO:root: +[Valid][22][UTT][ MSE = 0.2309 | LCC = 0.8569 | SRCC = 0.8567 ] +INFO:root:Average loss=0.055077012532796614 +INFO:root: +[Valid][23][UTT][ MSE = 0.2539 | LCC = 0.8537 | SRCC = 0.8541 ] +INFO:root:Average loss=0.05302810387160534 +INFO:root: +[Valid][24][UTT][ MSE = 0.2626 | LCC = 0.8524 | SRCC = 0.8529 ] +INFO:root:Average loss=0.05048807672201059 +INFO:root: +[Valid][25][UTT][ MSE = 0.2428 | LCC = 0.8513 | SRCC = 0.8510 ] +INFO:root:Average loss=0.05132855040331682 +INFO:root: +[Valid][26][UTT][ MSE = 0.2463 | LCC = 0.8525 | SRCC = 0.8521 ] +INFO:root:Average loss=0.046166109470411755 +INFO:root: +[Valid][27][UTT][ MSE = 0.2623 | LCC = 0.8507 | SRCC = 0.8507 ] +INFO:root:Average loss=0.04295750248890657 +INFO:root: +[Valid][28][UTT][ MSE = 0.2365 | LCC = 0.8521 | SRCC = 0.8527 ] +INFO:root:Average loss=0.04421266321188364 +INFO:root: +[Valid][29][UTT][ MSE = 0.2431 | LCC = 0.8536 | SRCC = 0.8542 ] +INFO:root:Average loss=0.04202679036041865 +INFO:root: +[Valid][30][UTT][ MSE = 0.2473 | LCC = 0.8542 | SRCC = 0.8547 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2523 | LCC = 0.8611 | SRCC = 0.8646 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..80b074d7cf29a5c98731a0628919cf2393432878 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:90a5e9750e159716c95c09fb2e3ad10009d90142d3ecc0dc5c7137a955376e50 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..efb03549d25443784dd2562b2be56aa68d69c7fd --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b4d858653d04b8c64d1c8ce6ece083d1c3e874c6c955e93f2d0707ce68f7482e +size 2749638 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3b47ea592cf4fd45301b5c38bb4fcc5d3c238651 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/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=0.3604837806943135 +INFO:root: +[Valid][0][UTT][ MSE = 0.3241 | LCC = 0.8343 | SRCC = 0.8350 ] +INFO:root:Average loss=0.2331581534101413 +INFO:root: +[Valid][1][UTT][ MSE = 0.2768 | LCC = 0.8506 | SRCC = 0.8496 ] +INFO:root:Average loss=0.2057274090938079 +INFO:root: +[Valid][2][UTT][ MSE = 0.2501 | LCC = 0.8595 | SRCC = 0.8580 ] +INFO:root:Average loss=0.191218614196166 +INFO:root: +[Valid][3][UTT][ MSE = 0.2560 | LCC = 0.8631 | SRCC = 0.8616 ] +INFO:root:Average loss=0.17719015383567566 +INFO:root: +[Valid][4][UTT][ MSE = 0.2326 | LCC = 0.8634 | SRCC = 0.8630 ] +INFO:root:Average loss=0.17022792326334196 +INFO:root: +[Valid][5][UTT][ MSE = 0.2758 | LCC = 0.8614 | SRCC = 0.8604 ] +INFO:root:Average loss=0.1579711240453598 +INFO:root: +[Valid][6][UTT][ MSE = 0.2327 | LCC = 0.8606 | SRCC = 0.8610 ] +INFO:root:Average loss=0.14690540969753876 +INFO:root: +[Valid][7][UTT][ MSE = 0.2559 | LCC = 0.8593 | SRCC = 0.8603 ] +INFO:root:Average loss=0.13715836730522987 +INFO:root: +[Valid][8][UTT][ MSE = 0.2581 | LCC = 0.8587 | SRCC = 0.8578 ] +INFO:root:Average loss=0.12865965249828804 +INFO:root: +[Valid][9][UTT][ MSE = 0.2611 | LCC = 0.8561 | SRCC = 0.8558 ] +INFO:root:Average loss=0.11679353803778306 +INFO:root: +[Valid][10][UTT][ MSE = 0.2572 | LCC = 0.8575 | SRCC = 0.8581 ] +INFO:root:Average loss=0.1120085115902699 +INFO:root: +[Valid][11][UTT][ MSE = 0.2507 | LCC = 0.8551 | SRCC = 0.8558 ] +INFO:root:Average loss=0.10751390299544884 +INFO:root: +[Valid][12][UTT][ MSE = 0.2619 | LCC = 0.8516 | SRCC = 0.8521 ] +INFO:root:Average loss=0.10514366779571925 +INFO:root: +[Valid][13][UTT][ MSE = 0.2485 | LCC = 0.8506 | SRCC = 0.8511 ] +INFO:root:Average loss=0.09369600846026188 +INFO:root: +[Valid][14][UTT][ MSE = 0.2562 | LCC = 0.8512 | SRCC = 0.8518 ] +INFO:root:Average loss=0.08671288693753573 +INFO:root: +[Valid][15][UTT][ MSE = 0.2629 | LCC = 0.8497 | SRCC = 0.8518 ] +INFO:root:Average loss=0.08395195432389393 +INFO:root: +[Valid][16][UTT][ MSE = 0.2715 | LCC = 0.8498 | SRCC = 0.8515 ] +INFO:root:Average loss=0.08351552882828774 +INFO:root: +[Valid][17][UTT][ MSE = 0.2751 | LCC = 0.8487 | SRCC = 0.8502 ] +INFO:root:Average loss=0.08010196704895069 +INFO:root: +[Valid][18][UTT][ MSE = 0.2838 | LCC = 0.8459 | SRCC = 0.8481 ] +INFO:root:Average loss=0.0744340203415889 +INFO:root: +[Valid][19][UTT][ MSE = 0.2526 | LCC = 0.8501 | SRCC = 0.8512 ] +INFO:root:Average loss=0.06690540408285764 +INFO:root: +[Valid][20][UTT][ MSE = 0.2752 | LCC = 0.8450 | SRCC = 0.8464 ] +INFO:root:Average loss=0.07163178252104001 +INFO:root: +[Valid][21][UTT][ MSE = 0.2750 | LCC = 0.8497 | SRCC = 0.8523 ] +INFO:root:Average loss=0.06305020106717563 +INFO:root: +[Valid][22][UTT][ MSE = 0.2566 | LCC = 0.8493 | SRCC = 0.8493 ] +INFO:root:Average loss=0.05843058061332275 +INFO:root: +[Valid][23][UTT][ MSE = 0.2525 | LCC = 0.8448 | SRCC = 0.8452 ] +INFO:root:Average loss=0.05796974214414755 +INFO:root: +[Valid][24][UTT][ MSE = 0.2625 | LCC = 0.8451 | SRCC = 0.8460 ] +INFO:root:Average loss=0.05792016645845694 +INFO:root: +[Valid][25][UTT][ MSE = 0.2687 | LCC = 0.8429 | SRCC = 0.8440 ] +INFO:root:Average loss=0.0553492755175401 +INFO:root: +[Valid][26][UTT][ MSE = 0.2491 | LCC = 0.8500 | SRCC = 0.8518 ] +INFO:root:Average loss=0.05292717873668059 +INFO:root: +[Valid][27][UTT][ MSE = 0.2502 | LCC = 0.8471 | SRCC = 0.8483 ] +INFO:root:Average loss=0.04914245057182434 +INFO:root: +[Valid][28][UTT][ MSE = 0.2660 | LCC = 0.8468 | SRCC = 0.8488 ] +INFO:root:Average loss=0.046933587974844836 +INFO:root: +[Valid][29][UTT][ MSE = 0.2531 | LCC = 0.8474 | SRCC = 0.8484 ] +INFO:root:Average loss=0.04748835655836723 +INFO:root: +[Valid][30][UTT][ MSE = 0.2588 | LCC = 0.8448 | SRCC = 0.8453 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2281 | LCC = 0.8657 | SRCC = 0.8675 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f4c744ea79659144318c35469963cf8098db88a9 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:397f58628d505f682ce8eef89e81c31956362c4e51ff7457d979973ae6441cfe +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..51654d8cac29c274a149d83b548bcdbd78d85ff5 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5aed4c0e3a4ffc26b47e238b8ba399c924795cf8d7087008e04d50b87280e34e +size 2749638 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2376909621bfee5dd9669db9d92c0f9c407348d5 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer6/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=0.3770978162303949 +INFO:root: +[Valid][0][UTT][ MSE = 0.3275 | LCC = 0.8323 | SRCC = 0.8332 ] +INFO:root:Average loss=0.2387025990547278 +INFO:root: +[Valid][1][UTT][ MSE = 0.2596 | LCC = 0.8497 | SRCC = 0.8498 ] +INFO:root:Average loss=0.21004770352290228 +INFO:root: +[Valid][2][UTT][ MSE = 0.2532 | LCC = 0.8611 | SRCC = 0.8606 ] +INFO:root:Average loss=0.1903307160887963 +INFO:root: +[Valid][3][UTT][ MSE = 0.2562 | LCC = 0.8654 | SRCC = 0.8620 ] +INFO:root:Average loss=0.17257102416493955 +INFO:root: +[Valid][4][UTT][ MSE = 0.2605 | LCC = 0.8643 | SRCC = 0.8634 ] +INFO:root:Average loss=0.16535658886035284 +INFO:root: +[Valid][5][UTT][ MSE = 0.2388 | LCC = 0.8655 | SRCC = 0.8653 ] +INFO:root:Average loss=0.15282277906170258 +INFO:root: +[Valid][6][UTT][ MSE = 0.2265 | LCC = 0.8653 | SRCC = 0.8646 ] +INFO:root:Average loss=0.1427481387479183 +INFO:root: +[Valid][7][UTT][ MSE = 0.2672 | LCC = 0.8649 | SRCC = 0.8616 ] +INFO:root:Average loss=0.13475456872047523 +INFO:root: +[Valid][8][UTT][ MSE = 0.2567 | LCC = 0.8611 | SRCC = 0.8585 ] +INFO:root:Average loss=0.12519202104363686 +INFO:root: +[Valid][9][UTT][ MSE = 0.2382 | LCC = 0.8615 | SRCC = 0.8595 ] +INFO:root:Average loss=0.11535744235301629 +INFO:root: +[Valid][10][UTT][ MSE = 0.2624 | LCC = 0.8616 | SRCC = 0.8614 ] +INFO:root:Average loss=0.10666486000021298 +INFO:root: +[Valid][11][UTT][ MSE = 0.2487 | LCC = 0.8593 | SRCC = 0.8570 ] +INFO:root:Average loss=0.1015381939136065 +INFO:root: +[Valid][12][UTT][ MSE = 0.2378 | LCC = 0.8577 | SRCC = 0.8554 ] +INFO:root:Average loss=0.0960148461163044 +INFO:root: +[Valid][13][UTT][ MSE = 0.2706 | LCC = 0.8567 | SRCC = 0.8539 ] +INFO:root:Average loss=0.09119731350204884 +INFO:root: +[Valid][14][UTT][ MSE = 0.2399 | LCC = 0.8560 | SRCC = 0.8547 ] +INFO:root:Average loss=0.081835809808511 +INFO:root: +[Valid][15][UTT][ MSE = 0.2524 | LCC = 0.8541 | SRCC = 0.8513 ] +INFO:root:Average loss=0.08057321904179378 +INFO:root: +[Valid][16][UTT][ MSE = 0.2529 | LCC = 0.8497 | SRCC = 0.8479 ] +INFO:root:Average loss=0.07568612451163623 +INFO:root: +[Valid][17][UTT][ MSE = 0.2479 | LCC = 0.8510 | SRCC = 0.8491 ] +INFO:root:Average loss=0.070625869748302 +INFO:root: +[Valid][18][UTT][ MSE = 0.2463 | LCC = 0.8553 | SRCC = 0.8540 ] +INFO:root:Average loss=0.06824355677534373 +INFO:root: +[Valid][19][UTT][ MSE = 0.2423 | LCC = 0.8536 | SRCC = 0.8517 ] +INFO:root:Average loss=0.06662077552232987 +INFO:root: +[Valid][20][UTT][ MSE = 0.2452 | LCC = 0.8556 | SRCC = 0.8540 ] +INFO:root:Average loss=0.059250764978619724 +INFO:root: +[Valid][21][UTT][ MSE = 0.2390 | LCC = 0.8555 | SRCC = 0.8540 ] +INFO:root:Average loss=0.058074076349536576 +INFO:root: +[Valid][22][UTT][ MSE = 0.2421 | LCC = 0.8544 | SRCC = 0.8522 ] +INFO:root:Average loss=0.05762129979064831 +INFO:root: +[Valid][23][UTT][ MSE = 0.2617 | LCC = 0.8510 | SRCC = 0.8492 ] +INFO:root:Average loss=0.056809222230162375 +INFO:root: +[Valid][24][UTT][ MSE = 0.2373 | LCC = 0.8537 | SRCC = 0.8508 ] +INFO:root:Average loss=0.05110459559811996 +INFO:root: +[Valid][25][UTT][ MSE = 0.2363 | LCC = 0.8545 | SRCC = 0.8540 ] +INFO:root:Average loss=0.05058353465910141 +INFO:root: +[Valid][26][UTT][ MSE = 0.2560 | LCC = 0.8498 | SRCC = 0.8479 ] +INFO:root:Average loss=0.04859237482723517 +INFO:root: +[Valid][27][UTT][ MSE = 0.2396 | LCC = 0.8527 | SRCC = 0.8519 ] +INFO:root:Average loss=0.04709915105157938 +INFO:root: +[Valid][28][UTT][ MSE = 0.2621 | LCC = 0.8506 | SRCC = 0.8497 ] +INFO:root:Average loss=0.04966300873993299 +INFO:root: +[Valid][29][UTT][ MSE = 0.2490 | LCC = 0.8528 | SRCC = 0.8523 ] +INFO:root:Average loss=0.04379731996987875 +INFO:root: +[Valid][30][UTT][ MSE = 0.2431 | LCC = 0.8532 | SRCC = 0.8525 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2391 | LCC = 0.8677 | SRCC = 0.8670 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2b6e9a61be2dd30e3d7d63a00a28f3824fcf1637 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bcb23afd3177b4134fd8992391887bb35558ec5236e983a5ea6ef140639c1514 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0217e4f434d31e413ebe7ce0a88553ba41c6cec7 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d95505fe3bca9aafb0219284b64b790e45615f1a520ecad914672b76849e621 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..db10216a633bf91c9c4d3540698b6b9367ce99ec --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/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=0.3469079965964342 +INFO:root: +[Valid][0][UTT][ MSE = 0.2989 | LCC = 0.8382 | SRCC = 0.8401 ] +INFO:root:Average loss=0.23198478783552462 +INFO:root: +[Valid][1][UTT][ MSE = 0.2585 | LCC = 0.8558 | SRCC = 0.8529 ] +INFO:root:Average loss=0.20541685781417748 +INFO:root: +[Valid][2][UTT][ MSE = 0.2483 | LCC = 0.8641 | SRCC = 0.8621 ] +INFO:root:Average loss=0.19130761921405792 +INFO:root: +[Valid][3][UTT][ MSE = 0.2713 | LCC = 0.8677 | SRCC = 0.8651 ] +INFO:root:Average loss=0.18461942663177466 +INFO:root: +[Valid][4][UTT][ MSE = 0.2428 | LCC = 0.8685 | SRCC = 0.8679 ] +INFO:root:Average loss=0.16500769097071427 +INFO:root: +[Valid][5][UTT][ MSE = 0.2516 | LCC = 0.8690 | SRCC = 0.8660 ] +INFO:root:Average loss=0.15804373329648605 +INFO:root: +[Valid][6][UTT][ MSE = 0.2455 | LCC = 0.8693 | SRCC = 0.8679 ] +INFO:root:Average loss=0.15088937603510344 +INFO:root: +[Valid][7][UTT][ MSE = 0.2357 | LCC = 0.8691 | SRCC = 0.8673 ] +INFO:root:Average loss=0.1387404889250413 +INFO:root: +[Valid][8][UTT][ MSE = 0.2327 | LCC = 0.8666 | SRCC = 0.8658 ] +INFO:root:Average loss=0.13043687473504972 +INFO:root: +[Valid][9][UTT][ MSE = 0.2570 | LCC = 0.8636 | SRCC = 0.8635 ] +INFO:root:Average loss=0.11843642649742273 +INFO:root: +[Valid][10][UTT][ MSE = 0.2293 | LCC = 0.8645 | SRCC = 0.8642 ] +INFO:root:Average loss=0.12040004864908181 +INFO:root: +[Valid][11][UTT][ MSE = 0.2339 | LCC = 0.8614 | SRCC = 0.8606 ] +INFO:root:Average loss=0.11031159605735387 +INFO:root: +[Valid][12][UTT][ MSE = 0.2628 | LCC = 0.8545 | SRCC = 0.8545 ] +INFO:root:Average loss=0.10428104143685256 +INFO:root: +[Valid][13][UTT][ MSE = 0.2402 | LCC = 0.8574 | SRCC = 0.8572 ] +INFO:root:Average loss=0.09424971884641892 +INFO:root: +[Valid][14][UTT][ MSE = 0.2438 | LCC = 0.8565 | SRCC = 0.8566 ] +INFO:root:Average loss=0.08903348732453126 +INFO:root: +[Valid][15][UTT][ MSE = 0.2415 | LCC = 0.8514 | SRCC = 0.8501 ] +INFO:root:Average loss=0.08564536870481113 +INFO:root: +[Valid][16][UTT][ MSE = 0.2446 | LCC = 0.8547 | SRCC = 0.8534 ] +INFO:root:Average loss=0.08497493747526254 +INFO:root: +[Valid][17][UTT][ MSE = 0.2353 | LCC = 0.8549 | SRCC = 0.8540 ] +INFO:root:Average loss=0.07745299349801663 +INFO:root: +[Valid][18][UTT][ MSE = 0.2783 | LCC = 0.8477 | SRCC = 0.8468 ] +INFO:root:Average loss=0.07135221438530163 +INFO:root: +[Valid][19][UTT][ MSE = 0.2526 | LCC = 0.8502 | SRCC = 0.8495 ] +INFO:root:Average loss=0.07131569780027255 +INFO:root: +[Valid][20][UTT][ MSE = 0.2605 | LCC = 0.8463 | SRCC = 0.8451 ] +INFO:root:Average loss=0.06440688287600493 +INFO:root: +[Valid][21][UTT][ MSE = 0.2764 | LCC = 0.8484 | SRCC = 0.8474 ] +INFO:root:Average loss=0.0627313334112748 +INFO:root: +[Valid][22][UTT][ MSE = 0.2873 | LCC = 0.8452 | SRCC = 0.8444 ] +INFO:root:Average loss=0.05825639129258119 +INFO:root: +[Valid][23][UTT][ MSE = 0.2651 | LCC = 0.8423 | SRCC = 0.8421 ] +INFO:root:Average loss=0.056449636578177795 +INFO:root: +[Valid][24][UTT][ MSE = 0.2629 | LCC = 0.8450 | SRCC = 0.8436 ] +INFO:root:Average loss=0.055312554948987104 +INFO:root: +[Valid][25][UTT][ MSE = 0.2510 | LCC = 0.8484 | SRCC = 0.8479 ] +INFO:root:Average loss=0.051834007199758135 +INFO:root: +[Valid][26][UTT][ MSE = 0.2352 | LCC = 0.8502 | SRCC = 0.8500 ] +INFO:root:Average loss=0.04881216726528528 +INFO:root: +[Valid][27][UTT][ MSE = 0.2513 | LCC = 0.8447 | SRCC = 0.8440 ] +INFO:root:Average loss=0.04795687951338597 +INFO:root: +[Valid][28][UTT][ MSE = 0.2557 | LCC = 0.8469 | SRCC = 0.8456 ] +INFO:root:Average loss=0.045337436434168085 +INFO:root: +[Valid][29][UTT][ MSE = 0.2537 | LCC = 0.8483 | SRCC = 0.8473 ] +INFO:root:Average loss=0.04486588393457425 +INFO:root: +[Valid][30][UTT][ MSE = 0.2551 | LCC = 0.8463 | SRCC = 0.8444 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2494 | LCC = 0.8673 | SRCC = 0.8665 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6d27f6d96880035e756ee1b9068374db5d813c3f --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:53753beeb216822b7a44963f08a62439e29026986a49876121c9e28db9c01ac6 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..99dbe5028a8bf5144d8a64e18cf1079eb8211db4 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:22e32b5b8e6cff4723c3767c557a1cf8849e21f40cde73a0161d2cbd38cc58cc +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..191da5f5ba63d5f2e88dca6076a8f1dda030f97c --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/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=0.35092105181553424 +INFO:root: +[Valid][0][UTT][ MSE = 0.3467 | LCC = 0.8330 | SRCC = 0.8389 ] +INFO:root:Average loss=0.22391803409808722 +INFO:root: +[Valid][1][UTT][ MSE = 0.2580 | LCC = 0.8565 | SRCC = 0.8579 ] +INFO:root:Average loss=0.2077186908095311 +INFO:root: +[Valid][2][UTT][ MSE = 0.2649 | LCC = 0.8597 | SRCC = 0.8611 ] +INFO:root:Average loss=0.1873732989606185 +INFO:root: +[Valid][3][UTT][ MSE = 0.2662 | LCC = 0.8629 | SRCC = 0.8674 ] +INFO:root:Average loss=0.1828128002010859 +INFO:root: +[Valid][4][UTT][ MSE = 0.2472 | LCC = 0.8694 | SRCC = 0.8691 ] +INFO:root:Average loss=0.17032867536330834 +INFO:root: +[Valid][5][UTT][ MSE = 0.2253 | LCC = 0.8675 | SRCC = 0.8682 ] +INFO:root:Average loss=0.15470075215666723 +INFO:root: +[Valid][6][UTT][ MSE = 0.2232 | LCC = 0.8677 | SRCC = 0.8674 ] +INFO:root:Average loss=0.14759624739870048 +INFO:root: +[Valid][7][UTT][ MSE = 0.2272 | LCC = 0.8673 | SRCC = 0.8694 ] +INFO:root:Average loss=0.14010418645846537 +INFO:root: +[Valid][8][UTT][ MSE = 0.2238 | LCC = 0.8671 | SRCC = 0.8693 ] +INFO:root:Average loss=0.1272563649675785 +INFO:root: +[Valid][9][UTT][ MSE = 0.2293 | LCC = 0.8661 | SRCC = 0.8664 ] +INFO:root:Average loss=0.12291097459502709 +INFO:root: +[Valid][10][UTT][ MSE = 0.2458 | LCC = 0.8646 | SRCC = 0.8627 ] +INFO:root:Average loss=0.11045287401439288 +INFO:root: +[Valid][11][UTT][ MSE = 0.2423 | LCC = 0.8621 | SRCC = 0.8611 ] +INFO:root:Average loss=0.10909775281563783 +INFO:root: +[Valid][12][UTT][ MSE = 0.2404 | LCC = 0.8589 | SRCC = 0.8595 ] +INFO:root:Average loss=0.09805169773216431 +INFO:root: +[Valid][13][UTT][ MSE = 0.2434 | LCC = 0.8584 | SRCC = 0.8595 ] +INFO:root:Average loss=0.09028928206326106 +INFO:root: +[Valid][14][UTT][ MSE = 0.2443 | LCC = 0.8571 | SRCC = 0.8571 ] +INFO:root:Average loss=0.08622201809134239 +INFO:root: +[Valid][15][UTT][ MSE = 0.2455 | LCC = 0.8500 | SRCC = 0.8495 ] +INFO:root:Average loss=0.08073516118411835 +INFO:root: +[Valid][16][UTT][ MSE = 0.2566 | LCC = 0.8536 | SRCC = 0.8546 ] +INFO:root:Average loss=0.07807559548662259 +INFO:root: +[Valid][17][UTT][ MSE = 0.2554 | LCC = 0.8526 | SRCC = 0.8529 ] +INFO:root:Average loss=0.07593226824433376 +INFO:root: +[Valid][18][UTT][ MSE = 0.2307 | LCC = 0.8567 | SRCC = 0.8576 ] +INFO:root:Average loss=0.07104522601152077 +INFO:root: +[Valid][19][UTT][ MSE = 0.2343 | LCC = 0.8569 | SRCC = 0.8568 ] +INFO:root:Average loss=0.0637034338254195 +INFO:root: +[Valid][20][UTT][ MSE = 0.2489 | LCC = 0.8500 | SRCC = 0.8502 ] +INFO:root:Average loss=0.06301592171001129 +INFO:root: +[Valid][21][UTT][ MSE = 0.2420 | LCC = 0.8489 | SRCC = 0.8498 ] +INFO:root:Average loss=0.05648809709610083 +INFO:root: +[Valid][22][UTT][ MSE = 0.2530 | LCC = 0.8474 | SRCC = 0.8473 ] +INFO:root:Average loss=0.0569760552010475 +INFO:root: +[Valid][23][UTT][ MSE = 0.2716 | LCC = 0.8537 | SRCC = 0.8538 ] +INFO:root:Average loss=0.054729500642189614 +INFO:root: +[Valid][24][UTT][ MSE = 0.2478 | LCC = 0.8465 | SRCC = 0.8482 ] +INFO:root:Average loss=0.05050178380826345 +INFO:root: +[Valid][25][UTT][ MSE = 0.2516 | LCC = 0.8517 | SRCC = 0.8526 ] +INFO:root:Average loss=0.0497330358872811 +INFO:root: +[Valid][26][UTT][ MSE = 0.2371 | LCC = 0.8499 | SRCC = 0.8501 ] +INFO:root:Average loss=0.04755664182205995 +INFO:root: +[Valid][27][UTT][ MSE = 0.2439 | LCC = 0.8511 | SRCC = 0.8500 ] +INFO:root:Average loss=0.046851443532758795 +INFO:root: +[Valid][28][UTT][ MSE = 0.2498 | LCC = 0.8515 | SRCC = 0.8519 ] +INFO:root:Average loss=0.04367782999403202 +INFO:root: +[Valid][29][UTT][ MSE = 0.2481 | LCC = 0.8487 | SRCC = 0.8492 ] +INFO:root:Average loss=0.04227986633300017 +INFO:root: +[Valid][30][UTT][ MSE = 0.2561 | LCC = 0.8521 | SRCC = 0.8512 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2591 | LCC = 0.8659 | SRCC = 0.8697 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..05e9ea615867160a89634550ddd845c9a0aceb79 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:808f410352b17bde41dd47121471350a76295d454545e44e38a25e69823b3dff +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..90efa2e4c9837cd0424b556614cf7828f4a9ca8c --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e50b120f4936b41a4e400e3d81050ab5235850aa90ab0561a6fd79bd6b6b76f9 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..cfa55868d6b33b38a1acf8db4fe7ce7b1687c936 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/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=0.3361775741363183 +INFO:root: +[Valid][0][UTT][ MSE = 0.3013 | LCC = 0.8406 | SRCC = 0.8431 ] +INFO:root:Average loss=0.22454409377697188 +INFO:root: +[Valid][1][UTT][ MSE = 0.3010 | LCC = 0.8625 | SRCC = 0.8594 ] +INFO:root:Average loss=0.20117125765253335 +INFO:root: +[Valid][2][UTT][ MSE = 0.2633 | LCC = 0.8674 | SRCC = 0.8654 ] +INFO:root:Average loss=0.18968014877576095 +INFO:root: +[Valid][3][UTT][ MSE = 0.2349 | LCC = 0.8672 | SRCC = 0.8676 ] +INFO:root:Average loss=0.17348546171799684 +INFO:root: +[Valid][4][UTT][ MSE = 0.2443 | LCC = 0.8684 | SRCC = 0.8675 ] +INFO:root:Average loss=0.1605497519366252 +INFO:root: +[Valid][5][UTT][ MSE = 0.2410 | LCC = 0.8680 | SRCC = 0.8677 ] +INFO:root:Average loss=0.16413550499158028 +INFO:root: +[Valid][6][UTT][ MSE = 0.2729 | LCC = 0.8683 | SRCC = 0.8689 ] +INFO:root:Average loss=0.1432249536498999 +INFO:root: +[Valid][7][UTT][ MSE = 0.2575 | LCC = 0.8671 | SRCC = 0.8662 ] +INFO:root:Average loss=0.13707019159427056 +INFO:root: +[Valid][8][UTT][ MSE = 0.2333 | LCC = 0.8670 | SRCC = 0.8674 ] +INFO:root:Average loss=0.12744018387718078 +INFO:root: +[Valid][9][UTT][ MSE = 0.2412 | LCC = 0.8634 | SRCC = 0.8633 ] +INFO:root:Average loss=0.11622359966620421 +INFO:root: +[Valid][10][UTT][ MSE = 0.2643 | LCC = 0.8639 | SRCC = 0.8639 ] +INFO:root:Average loss=0.11292111424681468 +INFO:root: +[Valid][11][UTT][ MSE = 0.2803 | LCC = 0.8580 | SRCC = 0.8582 ] +INFO:root:Average loss=0.10526881997401898 +INFO:root: +[Valid][12][UTT][ MSE = 0.2425 | LCC = 0.8621 | SRCC = 0.8636 ] +INFO:root:Average loss=0.10095537816866851 +INFO:root: +[Valid][13][UTT][ MSE = 0.2605 | LCC = 0.8557 | SRCC = 0.8590 ] +INFO:root:Average loss=0.09372408124498832 +INFO:root: +[Valid][14][UTT][ MSE = 0.2570 | LCC = 0.8530 | SRCC = 0.8542 ] +INFO:root:Average loss=0.08658124324985039 +INFO:root: +[Valid][15][UTT][ MSE = 0.2685 | LCC = 0.8596 | SRCC = 0.8607 ] +INFO:root:Average loss=0.0811407636755552 +INFO:root: +[Valid][16][UTT][ MSE = 0.2395 | LCC = 0.8570 | SRCC = 0.8582 ] +INFO:root:Average loss=0.07503594157214348 +INFO:root: +[Valid][17][UTT][ MSE = 0.2475 | LCC = 0.8559 | SRCC = 0.8550 ] +INFO:root:Average loss=0.07276743564468163 +INFO:root: +[Valid][18][UTT][ MSE = 0.2805 | LCC = 0.8557 | SRCC = 0.8550 ] +INFO:root:Average loss=0.07363468394256555 +INFO:root: +[Valid][19][UTT][ MSE = 0.2336 | LCC = 0.8569 | SRCC = 0.8587 ] +INFO:root:Average loss=0.07126202931006749 +INFO:root: +[Valid][20][UTT][ MSE = 0.2502 | LCC = 0.8539 | SRCC = 0.8541 ] +INFO:root:Average loss=0.0614911320213324 +INFO:root: +[Valid][21][UTT][ MSE = 0.2507 | LCC = 0.8570 | SRCC = 0.8577 ] +INFO:root:Average loss=0.06123488305661923 +INFO:root: +[Valid][22][UTT][ MSE = 0.2405 | LCC = 0.8536 | SRCC = 0.8550 ] +INFO:root:Average loss=0.05606937422775305 +INFO:root: +[Valid][23][UTT][ MSE = 0.2599 | LCC = 0.8507 | SRCC = 0.8510 ] +INFO:root:Average loss=0.05564926383204949 +INFO:root: +[Valid][24][UTT][ MSE = 0.2626 | LCC = 0.8530 | SRCC = 0.8536 ] +INFO:root:Average loss=0.05060609640219273 +INFO:root: +[Valid][25][UTT][ MSE = 0.2654 | LCC = 0.8502 | SRCC = 0.8512 ] +INFO:root:Average loss=0.05456724017858505 +INFO:root: +[Valid][26][UTT][ MSE = 0.2618 | LCC = 0.8515 | SRCC = 0.8526 ] +INFO:root:Average loss=0.05297986059807814 +INFO:root: +[Valid][27][UTT][ MSE = 0.2483 | LCC = 0.8501 | SRCC = 0.8511 ] +INFO:root:Average loss=0.04690793209159986 +INFO:root: +[Valid][28][UTT][ MSE = 0.2490 | LCC = 0.8527 | SRCC = 0.8530 ] +INFO:root:Average loss=0.04672756762458728 +INFO:root: +[Valid][29][UTT][ MSE = 0.2377 | LCC = 0.8532 | SRCC = 0.8547 ] +INFO:root:Average loss=0.051660896016237065 +INFO:root: +[Valid][30][UTT][ MSE = 0.2572 | LCC = 0.8520 | SRCC = 0.8522 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2382 | LCC = 0.8607 | SRCC = 0.8628 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..951c7061f4feeee79d2f24403cb65ea8fa0f8c5e --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:49619c4ec3b71f6a7b998d5fb817bd1de77e54dd5338a2cba438d42e85a6f417 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c02ad21cda8a5f74fbde9ba7e6b3cdcc9c7548c4 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9f6d6aff373ff50a73cacdb16fce4025d6a77eaf15c43ae9bd6556f1aa0d0cc2 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3ab6d3bf8fecca22871dd5ff0af86a53e8e9be80 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/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=0.36031487526801914 +INFO:root: +[Valid][0][UTT][ MSE = 0.2909 | LCC = 0.8361 | SRCC = 0.8392 ] +INFO:root:Average loss=0.22986766352103308 +INFO:root: +[Valid][1][UTT][ MSE = 0.2835 | LCC = 0.8539 | SRCC = 0.8506 ] +INFO:root:Average loss=0.21281702033220193 +INFO:root: +[Valid][2][UTT][ MSE = 0.2674 | LCC = 0.8612 | SRCC = 0.8583 ] +INFO:root:Average loss=0.19140471250582963 +INFO:root: +[Valid][3][UTT][ MSE = 0.2410 | LCC = 0.8623 | SRCC = 0.8600 ] +INFO:root:Average loss=0.18375769677834633 +INFO:root: +[Valid][4][UTT][ MSE = 0.2631 | LCC = 0.8599 | SRCC = 0.8586 ] +INFO:root:Average loss=0.17224495705121604 +INFO:root: +[Valid][5][UTT][ MSE = 0.2364 | LCC = 0.8620 | SRCC = 0.8609 ] +INFO:root:Average loss=0.15913706100904024 +INFO:root: +[Valid][6][UTT][ MSE = 0.2423 | LCC = 0.8638 | SRCC = 0.8621 ] +INFO:root:Average loss=0.155670099820082 +INFO:root: +[Valid][7][UTT][ MSE = 0.2358 | LCC = 0.8633 | SRCC = 0.8627 ] +INFO:root:Average loss=0.14069017166128525 +INFO:root: +[Valid][8][UTT][ MSE = 0.2496 | LCC = 0.8571 | SRCC = 0.8553 ] +INFO:root:Average loss=0.1286107660868229 +INFO:root: +[Valid][9][UTT][ MSE = 0.2474 | LCC = 0.8587 | SRCC = 0.8572 ] +INFO:root:Average loss=0.12176207672709073 +INFO:root: +[Valid][10][UTT][ MSE = 0.2495 | LCC = 0.8587 | SRCC = 0.8576 ] +INFO:root:Average loss=0.11548341027437112 +INFO:root: +[Valid][11][UTT][ MSE = 0.2463 | LCC = 0.8580 | SRCC = 0.8565 ] +INFO:root:Average loss=0.10905310291892444 +INFO:root: +[Valid][12][UTT][ MSE = 0.2554 | LCC = 0.8566 | SRCC = 0.8552 ] +INFO:root:Average loss=0.10277998423538147 +INFO:root: +[Valid][13][UTT][ MSE = 0.2511 | LCC = 0.8490 | SRCC = 0.8482 ] +INFO:root:Average loss=0.09820752328213973 +INFO:root: +[Valid][14][UTT][ MSE = 0.2500 | LCC = 0.8500 | SRCC = 0.8498 ] +INFO:root:Average loss=0.09047077357386932 +INFO:root: +[Valid][15][UTT][ MSE = 0.2468 | LCC = 0.8519 | SRCC = 0.8514 ] +INFO:root:Average loss=0.08370262503815003 +INFO:root: +[Valid][16][UTT][ MSE = 0.2386 | LCC = 0.8501 | SRCC = 0.8502 ] +INFO:root:Average loss=0.0783373729731792 +INFO:root: +[Valid][17][UTT][ MSE = 0.2434 | LCC = 0.8489 | SRCC = 0.8481 ] +INFO:root:Average loss=0.07786965699723133 +INFO:root: +[Valid][18][UTT][ MSE = 0.2333 | LCC = 0.8554 | SRCC = 0.8550 ] +INFO:root:Average loss=0.07153331612547238 +INFO:root: +[Valid][19][UTT][ MSE = 0.2397 | LCC = 0.8520 | SRCC = 0.8514 ] +INFO:root:Average loss=0.06711864270842992 +INFO:root: +[Valid][20][UTT][ MSE = 0.2385 | LCC = 0.8527 | SRCC = 0.8528 ] +INFO:root:Average loss=0.06474789690512878 +INFO:root: +[Valid][21][UTT][ MSE = 0.2565 | LCC = 0.8452 | SRCC = 0.8442 ] +INFO:root:Average loss=0.06640106401382348 +INFO:root: +[Valid][22][UTT][ MSE = 0.2527 | LCC = 0.8476 | SRCC = 0.8468 ] +INFO:root:Average loss=0.0617793684777541 +INFO:root: +[Valid][23][UTT][ MSE = 0.2416 | LCC = 0.8485 | SRCC = 0.8480 ] +INFO:root:Average loss=0.05555884659481354 +INFO:root: +[Valid][24][UTT][ MSE = 0.2528 | LCC = 0.8471 | SRCC = 0.8463 ] +INFO:root:Average loss=0.05438259654702284 +INFO:root: +[Valid][25][UTT][ MSE = 0.2425 | LCC = 0.8511 | SRCC = 0.8501 ] +INFO:root:Average loss=0.050841256307485774 +INFO:root: +[Valid][26][UTT][ MSE = 0.2370 | LCC = 0.8497 | SRCC = 0.8503 ] +INFO:root:Average loss=0.05048329660143608 +INFO:root: +[Valid][27][UTT][ MSE = 0.2482 | LCC = 0.8499 | SRCC = 0.8490 ] +INFO:root:Average loss=0.047865212871096074 +INFO:root: +[Valid][28][UTT][ MSE = 0.2448 | LCC = 0.8464 | SRCC = 0.8446 ] +INFO:root:Average loss=0.04802763521766815 +INFO:root: +[Valid][29][UTT][ MSE = 0.2605 | LCC = 0.8441 | SRCC = 0.8437 ] +INFO:root:Average loss=0.0466090687431204 +INFO:root: +[Valid][30][UTT][ MSE = 0.2672 | LCC = 0.8404 | SRCC = 0.8388 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2572 | LCC = 0.8572 | SRCC = 0.8603 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2e61d38e39dd8b89a14e7b245c1203afabe606d1 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b91b223ae33e3bae2145beadebba1e4d53b06115e58c1332019eace278d7cdca +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d29ce6a3428f2739bc49ff3eef16da8a575bcd57 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5b64ab0e2aaa9ab1dff702b1235d92f51ae9897c2669acc2fb94579a4617f413 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5e663605259ab6f2ae8f5e303b87d02d487b79ce --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer7/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=0.3935772079305771 +INFO:root: +[Valid][0][UTT][ MSE = 0.3136 | LCC = 0.8309 | SRCC = 0.8344 ] +INFO:root:Average loss=0.2340518199862578 +INFO:root: +[Valid][1][UTT][ MSE = 0.2768 | LCC = 0.8522 | SRCC = 0.8543 ] +INFO:root:Average loss=0.21616570880779853 +INFO:root: +[Valid][2][UTT][ MSE = 0.2632 | LCC = 0.8638 | SRCC = 0.8625 ] +INFO:root:Average loss=0.18913919765215653 +INFO:root: +[Valid][3][UTT][ MSE = 0.2265 | LCC = 0.8659 | SRCC = 0.8662 ] +INFO:root:Average loss=0.17506762441152182 +INFO:root: +[Valid][4][UTT][ MSE = 0.2546 | LCC = 0.8694 | SRCC = 0.8684 ] +INFO:root:Average loss=0.16724137159494254 +INFO:root: +[Valid][5][UTT][ MSE = 0.2255 | LCC = 0.8716 | SRCC = 0.8705 ] +INFO:root:Average loss=0.1591491005741633 +INFO:root: +[Valid][6][UTT][ MSE = 0.2369 | LCC = 0.8715 | SRCC = 0.8715 ] +INFO:root:Average loss=0.14699988172222406 +INFO:root: +[Valid][7][UTT][ MSE = 0.2114 | LCC = 0.8686 | SRCC = 0.8705 ] +INFO:root:Average loss=0.14021307459244362 +INFO:root: +[Valid][8][UTT][ MSE = 0.2282 | LCC = 0.8693 | SRCC = 0.8700 ] +INFO:root:Average loss=0.12701662161793464 +INFO:root: +[Valid][9][UTT][ MSE = 0.2454 | LCC = 0.8655 | SRCC = 0.8663 ] +INFO:root:Average loss=0.11901020994171119 +INFO:root: +[Valid][10][UTT][ MSE = 0.2447 | LCC = 0.8637 | SRCC = 0.8670 ] +INFO:root:Average loss=0.11120298237372668 +INFO:root: +[Valid][11][UTT][ MSE = 0.2200 | LCC = 0.8665 | SRCC = 0.8682 ] +INFO:root:Average loss=0.10354154442365353 +INFO:root: +[Valid][12][UTT][ MSE = 0.2430 | LCC = 0.8585 | SRCC = 0.8598 ] +INFO:root:Average loss=0.0965094521928292 +INFO:root: +[Valid][13][UTT][ MSE = 0.2175 | LCC = 0.8660 | SRCC = 0.8667 ] +INFO:root:Average loss=0.08877817923441911 +INFO:root: +[Valid][14][UTT][ MSE = 0.2222 | LCC = 0.8638 | SRCC = 0.8645 ] +INFO:root:Average loss=0.08564599498342244 +INFO:root: +[Valid][15][UTT][ MSE = 0.2287 | LCC = 0.8556 | SRCC = 0.8575 ] +INFO:root:Average loss=0.07916431396435468 +INFO:root: +[Valid][16][UTT][ MSE = 0.2377 | LCC = 0.8565 | SRCC = 0.8574 ] +INFO:root:Average loss=0.07731060382838432 +INFO:root: +[Valid][17][UTT][ MSE = 0.2253 | LCC = 0.8576 | SRCC = 0.8586 ] +INFO:root:Average loss=0.07711132157307404 +INFO:root: +[Valid][18][UTT][ MSE = 0.2365 | LCC = 0.8540 | SRCC = 0.8567 ] +INFO:root:Average loss=0.06977713915208976 +INFO:root: +[Valid][19][UTT][ MSE = 0.2250 | LCC = 0.8579 | SRCC = 0.8587 ] +INFO:root:Average loss=0.06773408483236264 +INFO:root: +[Valid][20][UTT][ MSE = 0.2319 | LCC = 0.8556 | SRCC = 0.8576 ] +INFO:root:Average loss=0.062366763989512734 +INFO:root: +[Valid][21][UTT][ MSE = 0.2296 | LCC = 0.8541 | SRCC = 0.8562 ] +INFO:root:Average loss=0.05969893999206714 +INFO:root: +[Valid][22][UTT][ MSE = 0.2382 | LCC = 0.8543 | SRCC = 0.8566 ] +INFO:root:Average loss=0.05845110025256872 +INFO:root: +[Valid][23][UTT][ MSE = 0.2374 | LCC = 0.8508 | SRCC = 0.8546 ] +INFO:root:Average loss=0.0568846740688269 +INFO:root: +[Valid][24][UTT][ MSE = 0.2354 | LCC = 0.8495 | SRCC = 0.8524 ] +INFO:root:Average loss=0.05424533278132097 +INFO:root: +[Valid][25][UTT][ MSE = 0.2745 | LCC = 0.8416 | SRCC = 0.8439 ] +INFO:root:Average loss=0.053723059905072056 +INFO:root: +[Valid][26][UTT][ MSE = 0.2385 | LCC = 0.8517 | SRCC = 0.8522 ] +INFO:root:Average loss=0.05074396973045973 +INFO:root: +[Valid][27][UTT][ MSE = 0.2348 | LCC = 0.8531 | SRCC = 0.8542 ] +INFO:root:Average loss=0.05129267003100652 +INFO:root: +[Valid][28][UTT][ MSE = 0.2438 | LCC = 0.8519 | SRCC = 0.8536 ] +INFO:root:Average loss=0.046840745000503003 +INFO:root: +[Valid][29][UTT][ MSE = 0.2485 | LCC = 0.8503 | SRCC = 0.8514 ] +INFO:root:Average loss=0.04367403065164884 +INFO:root: +[Valid][30][UTT][ MSE = 0.2409 | LCC = 0.8479 | SRCC = 0.8511 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2414 | LCC = 0.8629 | SRCC = 0.8635 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4209b24ffcd36eb97420bb5cd1debaaa94d62617 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1913daf00e6341bfbf0bbdc03b6f5d660fd3582ff2aad561e39afc071f967592 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2c3fb611869f6df1efbbc989ebb661560e81bb1e --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be09ef74e4d54f418b709ac79940a142262a57ba3ee08075ec1178a892fa6d29 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..163634233b6fa7ffb042b1db2e54a38acc576852 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/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=0.36618207433284855 +INFO:root: +[Valid][0][UTT][ MSE = 0.3115 | LCC = 0.8374 | SRCC = 0.8441 ] +INFO:root:Average loss=0.23035683272740778 +INFO:root: +[Valid][1][UTT][ MSE = 0.2507 | LCC = 0.8566 | SRCC = 0.8570 ] +INFO:root:Average loss=0.2076960611037719 +INFO:root: +[Valid][2][UTT][ MSE = 0.3217 | LCC = 0.8657 | SRCC = 0.8648 ] +INFO:root:Average loss=0.19742264197422907 +INFO:root: +[Valid][3][UTT][ MSE = 0.2570 | LCC = 0.8683 | SRCC = 0.8680 ] +INFO:root:Average loss=0.1825676085666204 +INFO:root: +[Valid][4][UTT][ MSE = 0.2422 | LCC = 0.8659 | SRCC = 0.8671 ] +INFO:root:Average loss=0.17696284263944015 +INFO:root: +[Valid][5][UTT][ MSE = 0.2451 | LCC = 0.8708 | SRCC = 0.8697 ] +INFO:root:Average loss=0.1629517748951912 +INFO:root: +[Valid][6][UTT][ MSE = 0.2518 | LCC = 0.8675 | SRCC = 0.8684 ] +INFO:root:Average loss=0.16286234557628632 +INFO:root: +[Valid][7][UTT][ MSE = 0.2884 | LCC = 0.8700 | SRCC = 0.8718 ] +INFO:root:Average loss=0.1491119971451087 +INFO:root: +[Valid][8][UTT][ MSE = 0.2194 | LCC = 0.8675 | SRCC = 0.8685 ] +INFO:root:Average loss=0.14084815157529634 +INFO:root: +[Valid][9][UTT][ MSE = 0.2424 | LCC = 0.8673 | SRCC = 0.8683 ] +INFO:root:Average loss=0.13155510333868173 +INFO:root: +[Valid][10][UTT][ MSE = 0.2442 | LCC = 0.8605 | SRCC = 0.8613 ] +INFO:root:Average loss=0.1232248183626395 +INFO:root: +[Valid][11][UTT][ MSE = 0.2115 | LCC = 0.8640 | SRCC = 0.8678 ] +INFO:root:Average loss=0.12307575650704212 +INFO:root: +[Valid][12][UTT][ MSE = 0.2356 | LCC = 0.8619 | SRCC = 0.8621 ] +INFO:root:Average loss=0.1198196261165998 +INFO:root: +[Valid][13][UTT][ MSE = 0.2491 | LCC = 0.8575 | SRCC = 0.8583 ] +INFO:root:Average loss=0.11264313738315533 +INFO:root: +[Valid][14][UTT][ MSE = 0.2544 | LCC = 0.8525 | SRCC = 0.8541 ] +INFO:root:Average loss=0.10313622104242826 +INFO:root: +[Valid][15][UTT][ MSE = 0.2420 | LCC = 0.8554 | SRCC = 0.8564 ] +INFO:root:Average loss=0.09933174840914898 +INFO:root: +[Valid][16][UTT][ MSE = 0.2666 | LCC = 0.8548 | SRCC = 0.8563 ] +INFO:root:Average loss=0.09304687056021813 +INFO:root: +[Valid][17][UTT][ MSE = 0.2540 | LCC = 0.8563 | SRCC = 0.8564 ] +INFO:root:Average loss=0.08673851129909356 +INFO:root: +[Valid][18][UTT][ MSE = 0.2380 | LCC = 0.8527 | SRCC = 0.8540 ] +INFO:root:Average loss=0.08070615511865188 +INFO:root: +[Valid][19][UTT][ MSE = 0.2473 | LCC = 0.8512 | SRCC = 0.8508 ] +INFO:root:Average loss=0.08580885426356243 +INFO:root: +[Valid][20][UTT][ MSE = 0.2589 | LCC = 0.8553 | SRCC = 0.8559 ] +INFO:root:Average loss=0.07939003622875763 +INFO:root: +[Valid][21][UTT][ MSE = 0.2446 | LCC = 0.8517 | SRCC = 0.8538 ] +INFO:root:Average loss=0.0744810639283596 +INFO:root: +[Valid][22][UTT][ MSE = 0.2306 | LCC = 0.8543 | SRCC = 0.8564 ] +INFO:root:Average loss=0.07538428878745972 +INFO:root: +[Valid][23][UTT][ MSE = 0.2552 | LCC = 0.8456 | SRCC = 0.8460 ] +INFO:root:Average loss=0.06613916660157534 +INFO:root: +[Valid][24][UTT][ MSE = 0.2471 | LCC = 0.8523 | SRCC = 0.8535 ] +INFO:root:Average loss=0.06895109199178524 +INFO:root: +[Valid][25][UTT][ MSE = 0.2579 | LCC = 0.8485 | SRCC = 0.8493 ] +INFO:root:Average loss=0.06527905997175437 +INFO:root: +[Valid][26][UTT][ MSE = 0.2450 | LCC = 0.8469 | SRCC = 0.8490 ] +INFO:root:Average loss=0.05938053083343384 +INFO:root: +[Valid][27][UTT][ MSE = 0.2450 | LCC = 0.8493 | SRCC = 0.8497 ] +INFO:root:Average loss=0.060862778805387326 +INFO:root: +[Valid][28][UTT][ MSE = 0.2437 | LCC = 0.8498 | SRCC = 0.8523 ] +INFO:root:Average loss=0.056953951000021055 +INFO:root: +[Valid][29][UTT][ MSE = 0.2401 | LCC = 0.8506 | SRCC = 0.8527 ] +INFO:root:Average loss=0.05623827669291924 +INFO:root: +[Valid][30][UTT][ MSE = 0.2447 | LCC = 0.8509 | SRCC = 0.8522 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2399 | LCC = 0.8682 | SRCC = 0.8716 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7cf2ae29b483c330c25204cdbc0bc2e0e9c47c4d --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:07ef5cfed7998f179a652c1f2c033a0faa34abdac8945f9e955d2ee9953199c7 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..35721b6f23f7808355a2911994c9869a6aa49c00 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7af20bcb19a9e9f8eda6d8aec1c0544e75dc96b5fc95fcda5edb17fa024364ab +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..215f7604d029124108d9dafc0599346598ec50e4 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/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=0.3815968882961151 +INFO:root: +[Valid][0][UTT][ MSE = 0.2968 | LCC = 0.8400 | SRCC = 0.8404 ] +INFO:root:Average loss=0.2380546312301587 +INFO:root: +[Valid][1][UTT][ MSE = 0.2770 | LCC = 0.8576 | SRCC = 0.8572 ] +INFO:root:Average loss=0.22155761642333788 +INFO:root: +[Valid][2][UTT][ MSE = 0.2382 | LCC = 0.8570 | SRCC = 0.8589 ] +INFO:root:Average loss=0.20129955005951417 +INFO:root: +[Valid][3][UTT][ MSE = 0.2357 | LCC = 0.8594 | SRCC = 0.8607 ] +INFO:root:Average loss=0.19517251428885338 +INFO:root: +[Valid][4][UTT][ MSE = 0.2651 | LCC = 0.8678 | SRCC = 0.8657 ] +INFO:root:Average loss=0.17721360864547583 +INFO:root: +[Valid][5][UTT][ MSE = 0.2326 | LCC = 0.8646 | SRCC = 0.8636 ] +INFO:root:Average loss=0.17024213992632353 +INFO:root: +[Valid][6][UTT][ MSE = 0.2257 | LCC = 0.8673 | SRCC = 0.8671 ] +INFO:root:Average loss=0.16586820647502556 +INFO:root: +[Valid][7][UTT][ MSE = 0.2257 | LCC = 0.8685 | SRCC = 0.8682 ] +INFO:root:Average loss=0.15893090783785552 +INFO:root: +[Valid][8][UTT][ MSE = 0.2417 | LCC = 0.8706 | SRCC = 0.8693 ] +INFO:root:Average loss=0.14977524534631997 +INFO:root: +[Valid][9][UTT][ MSE = 0.2452 | LCC = 0.8687 | SRCC = 0.8685 ] +INFO:root:Average loss=0.1430939763593368 +INFO:root: +[Valid][10][UTT][ MSE = 0.2386 | LCC = 0.8660 | SRCC = 0.8653 ] +INFO:root:Average loss=0.12922520744494903 +INFO:root: +[Valid][11][UTT][ MSE = 0.2636 | LCC = 0.8632 | SRCC = 0.8599 ] +INFO:root:Average loss=0.12632224498650965 +INFO:root: +[Valid][12][UTT][ MSE = 0.2488 | LCC = 0.8653 | SRCC = 0.8633 ] +INFO:root:Average loss=0.12183572265964288 +INFO:root: +[Valid][13][UTT][ MSE = 0.2253 | LCC = 0.8635 | SRCC = 0.8647 ] +INFO:root:Average loss=0.11197917784253757 +INFO:root: +[Valid][14][UTT][ MSE = 0.2405 | LCC = 0.8614 | SRCC = 0.8601 ] +INFO:root:Average loss=0.10845722019290313 +INFO:root: +[Valid][15][UTT][ MSE = 0.2319 | LCC = 0.8603 | SRCC = 0.8613 ] +INFO:root:Average loss=0.1039438019387233 +INFO:root: +[Valid][16][UTT][ MSE = 0.2366 | LCC = 0.8595 | SRCC = 0.8593 ] +INFO:root:Average loss=0.09766694110555527 +INFO:root: +[Valid][17][UTT][ MSE = 0.2320 | LCC = 0.8563 | SRCC = 0.8566 ] +INFO:root:Average loss=0.08778332775601974 +INFO:root: +[Valid][18][UTT][ MSE = 0.2316 | LCC = 0.8577 | SRCC = 0.8589 ] +INFO:root:Average loss=0.09022905730093136 +INFO:root: +[Valid][19][UTT][ MSE = 0.2479 | LCC = 0.8560 | SRCC = 0.8542 ] +INFO:root:Average loss=0.08479707654661094 +INFO:root: +[Valid][20][UTT][ MSE = 0.2627 | LCC = 0.8475 | SRCC = 0.8474 ] +INFO:root:Average loss=0.08220539156060952 +INFO:root: +[Valid][21][UTT][ MSE = 0.2467 | LCC = 0.8564 | SRCC = 0.8544 ] +INFO:root:Average loss=0.07367289109298816 +INFO:root: +[Valid][22][UTT][ MSE = 0.2364 | LCC = 0.8579 | SRCC = 0.8577 ] +INFO:root:Average loss=0.07258540086257152 +INFO:root: +[Valid][23][UTT][ MSE = 0.2412 | LCC = 0.8540 | SRCC = 0.8537 ] +INFO:root:Average loss=0.07410742022479191 +INFO:root: +[Valid][24][UTT][ MSE = 0.2376 | LCC = 0.8541 | SRCC = 0.8537 ] +INFO:root:Average loss=0.07239103345916821 +INFO:root: +[Valid][25][UTT][ MSE = 0.2445 | LCC = 0.8517 | SRCC = 0.8506 ] +INFO:root:Average loss=0.06621793280236232 +INFO:root: +[Valid][26][UTT][ MSE = 0.2428 | LCC = 0.8502 | SRCC = 0.8509 ] +INFO:root:Average loss=0.06520524721306103 +INFO:root: +[Valid][27][UTT][ MSE = 0.2416 | LCC = 0.8526 | SRCC = 0.8519 ] +INFO:root:Average loss=0.062145787983750686 +INFO:root: +[Valid][28][UTT][ MSE = 0.2339 | LCC = 0.8556 | SRCC = 0.8554 ] +INFO:root:Average loss=0.05895796652214649 +INFO:root: +[Valid][29][UTT][ MSE = 0.2420 | LCC = 0.8517 | SRCC = 0.8510 ] +INFO:root:Average loss=0.05381425102360737 +INFO:root: +[Valid][30][UTT][ MSE = 0.2482 | LCC = 0.8506 | SRCC = 0.8495 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2414 | LCC = 0.8645 | SRCC = 0.8630 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..80332c1e4fe5cafafd7cec4bef5a8b7b5a234dca --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:50d69449e466f8cba1870df0b6870e14bc00297997d8af979666290e12d5e56c +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e740ef78b032d391897047d6a8122a4756f6a06a --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:36ff244e91d140388a3f4fe400931bb0edc7bcc55c42c1c72d1da3ddaab27bc6 +size 2749638 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c860ba415ab62980b9a7341f7cf07191350278fc --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/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=0.38624010292383343 +INFO:root: +[Valid][0][UTT][ MSE = 0.3145 | LCC = 0.8218 | SRCC = 0.8307 ] +INFO:root:Average loss=0.23327801873286566 +INFO:root: +[Valid][1][UTT][ MSE = 0.2849 | LCC = 0.8541 | SRCC = 0.8511 ] +INFO:root:Average loss=0.21395382648095107 +INFO:root: +[Valid][2][UTT][ MSE = 0.2704 | LCC = 0.8575 | SRCC = 0.8565 ] +INFO:root:Average loss=0.20186396153309408 +INFO:root: +[Valid][3][UTT][ MSE = 0.2697 | LCC = 0.8661 | SRCC = 0.8639 ] +INFO:root:Average loss=0.19400966205658057 +INFO:root: +[Valid][4][UTT][ MSE = 0.2680 | LCC = 0.8667 | SRCC = 0.8642 ] +INFO:root:Average loss=0.1747581745760563 +INFO:root: +[Valid][5][UTT][ MSE = 0.2596 | LCC = 0.8597 | SRCC = 0.8604 ] +INFO:root:Average loss=0.16902096617298248 +INFO:root: +[Valid][6][UTT][ MSE = 0.2527 | LCC = 0.8634 | SRCC = 0.8620 ] +INFO:root:Average loss=0.1604572149614493 +INFO:root: +[Valid][7][UTT][ MSE = 0.2420 | LCC = 0.8647 | SRCC = 0.8643 ] +INFO:root:Average loss=0.1524528848628203 +INFO:root: +[Valid][8][UTT][ MSE = 0.2458 | LCC = 0.8648 | SRCC = 0.8641 ] +INFO:root:Average loss=0.14067993733363274 +INFO:root: +[Valid][9][UTT][ MSE = 0.2654 | LCC = 0.8588 | SRCC = 0.8593 ] +INFO:root:Average loss=0.13395280429185966 +INFO:root: +[Valid][10][UTT][ MSE = 0.2431 | LCC = 0.8591 | SRCC = 0.8592 ] +INFO:root:Average loss=0.1349294881981153 +INFO:root: +[Valid][11][UTT][ MSE = 0.2544 | LCC = 0.8614 | SRCC = 0.8599 ] +INFO:root:Average loss=0.1255310636300307 +INFO:root: +[Valid][12][UTT][ MSE = 0.2321 | LCC = 0.8562 | SRCC = 0.8558 ] +INFO:root:Average loss=0.11244090426808749 +INFO:root: +[Valid][13][UTT][ MSE = 0.2511 | LCC = 0.8555 | SRCC = 0.8544 ] +INFO:root:Average loss=0.11026650476150024 +INFO:root: +[Valid][14][UTT][ MSE = 0.2458 | LCC = 0.8574 | SRCC = 0.8566 ] +INFO:root:Average loss=0.10423517914918753 +INFO:root: +[Valid][15][UTT][ MSE = 0.2486 | LCC = 0.8547 | SRCC = 0.8534 ] +INFO:root:Average loss=0.0963779003956379 +INFO:root: +[Valid][16][UTT][ MSE = 0.2414 | LCC = 0.8567 | SRCC = 0.8575 ] +INFO:root:Average loss=0.09325948753991188 +INFO:root: +[Valid][17][UTT][ MSE = 0.2692 | LCC = 0.8522 | SRCC = 0.8542 ] +INFO:root:Average loss=0.0934847181615157 +INFO:root: +[Valid][18][UTT][ MSE = 0.2646 | LCC = 0.8522 | SRCC = 0.8521 ] +INFO:root:Average loss=0.08583457715427265 +INFO:root: +[Valid][19][UTT][ MSE = 0.2531 | LCC = 0.8525 | SRCC = 0.8507 ] +INFO:root:Average loss=0.08116927021780075 +INFO:root: +[Valid][20][UTT][ MSE = 0.2302 | LCC = 0.8536 | SRCC = 0.8532 ] +INFO:root:Average loss=0.07833102617699367 +INFO:root: +[Valid][21][UTT][ MSE = 0.2498 | LCC = 0.8487 | SRCC = 0.8489 ] +INFO:root:Average loss=0.07603566152736163 +INFO:root: +[Valid][22][UTT][ MSE = 0.2435 | LCC = 0.8577 | SRCC = 0.8582 ] +INFO:root:Average loss=0.07386757051333404 +INFO:root: +[Valid][23][UTT][ MSE = 0.2538 | LCC = 0.8516 | SRCC = 0.8515 ] +INFO:root:Average loss=0.06637126952409744 +INFO:root: +[Valid][24][UTT][ MSE = 0.2590 | LCC = 0.8477 | SRCC = 0.8481 ] +INFO:root:Average loss=0.06700911936469567 +INFO:root: +[Valid][25][UTT][ MSE = 0.2647 | LCC = 0.8450 | SRCC = 0.8441 ] +INFO:root:Average loss=0.062100743636106834 +INFO:root: +[Valid][26][UTT][ MSE = 0.2504 | LCC = 0.8489 | SRCC = 0.8492 ] +INFO:root:Average loss=0.06089728478437815 +INFO:root: +[Valid][27][UTT][ MSE = 0.2563 | LCC = 0.8505 | SRCC = 0.8504 ] +INFO:root:Average loss=0.059881048324780584 +INFO:root: +[Valid][28][UTT][ MSE = 0.2599 | LCC = 0.8465 | SRCC = 0.8454 ] +INFO:root:Average loss=0.05888143081504565 +INFO:root: +[Valid][29][UTT][ MSE = 0.2522 | LCC = 0.8487 | SRCC = 0.8464 ] +INFO:root:Average loss=0.05399362150675211 +INFO:root: +[Valid][30][UTT][ MSE = 0.2545 | LCC = 0.8505 | SRCC = 0.8503 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2618 | LCC = 0.8691 | SRCC = 0.8683 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4f880b0533dd147f4854f94f1ac66fefd8f125b5 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7093060d3e07bbcf402e754d0e2d6294924d858ef8615918f673ba6fe1c6263a +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..58966bfd6810ffc403f4f326d507cc10b0dccc27 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e5885a4080a87b55b810fd7473de07434629701d87121fb0b654d3af0d3c1555 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..acda6402d9e06da3946ecc95a6af22912919870a --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/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=0.32878030435397076 +INFO:root: +[Valid][0][UTT][ MSE = 0.3769 | LCC = 0.8385 | SRCC = 0.8383 ] +INFO:root:Average loss=0.23124123842288286 +INFO:root: +[Valid][1][UTT][ MSE = 0.2991 | LCC = 0.8565 | SRCC = 0.8544 ] +INFO:root:Average loss=0.21163751796270028 +INFO:root: +[Valid][2][UTT][ MSE = 0.3129 | LCC = 0.8591 | SRCC = 0.8565 ] +INFO:root:Average loss=0.1948734507537805 +INFO:root: +[Valid][3][UTT][ MSE = 0.2956 | LCC = 0.8634 | SRCC = 0.8618 ] +INFO:root:Average loss=0.19122098577328217 +INFO:root: +[Valid][4][UTT][ MSE = 0.2724 | LCC = 0.8667 | SRCC = 0.8650 ] +INFO:root:Average loss=0.17582659547527632 +INFO:root: +[Valid][5][UTT][ MSE = 0.2863 | LCC = 0.8656 | SRCC = 0.8667 ] +INFO:root:Average loss=0.17432949472314271 +INFO:root: +[Valid][6][UTT][ MSE = 0.2617 | LCC = 0.8688 | SRCC = 0.8674 ] +INFO:root:Average loss=0.169759202366456 +INFO:root: +[Valid][7][UTT][ MSE = 0.2947 | LCC = 0.8645 | SRCC = 0.8628 ] +INFO:root:Average loss=0.14853652480703133 +INFO:root: +[Valid][8][UTT][ MSE = 0.2975 | LCC = 0.8626 | SRCC = 0.8637 ] +INFO:root:Average loss=0.14974183933092997 +INFO:root: +[Valid][9][UTT][ MSE = 0.2820 | LCC = 0.8621 | SRCC = 0.8606 ] +INFO:root:Average loss=0.13967383882174125 +INFO:root: +[Valid][10][UTT][ MSE = 0.2471 | LCC = 0.8633 | SRCC = 0.8618 ] +INFO:root:Average loss=0.13777399426087356 +INFO:root: +[Valid][11][UTT][ MSE = 0.2473 | LCC = 0.8587 | SRCC = 0.8573 ] +INFO:root:Average loss=0.1268659087900932 +INFO:root: +[Valid][12][UTT][ MSE = 0.2913 | LCC = 0.8597 | SRCC = 0.8621 ] +INFO:root:Average loss=0.12431207385200721 +INFO:root: +[Valid][13][UTT][ MSE = 0.2555 | LCC = 0.8594 | SRCC = 0.8588 ] +INFO:root:Average loss=0.11059511987826763 +INFO:root: +[Valid][14][UTT][ MSE = 0.3181 | LCC = 0.8551 | SRCC = 0.8548 ] +INFO:root:Average loss=0.11133420018431468 +INFO:root: +[Valid][15][UTT][ MSE = 0.3235 | LCC = 0.8541 | SRCC = 0.8545 ] +INFO:root:Average loss=0.104342583184823 +INFO:root: +[Valid][16][UTT][ MSE = 0.3072 | LCC = 0.8524 | SRCC = 0.8521 ] +INFO:root:Average loss=0.0999179309568344 +INFO:root: +[Valid][17][UTT][ MSE = 0.2629 | LCC = 0.8563 | SRCC = 0.8546 ] +INFO:root:Average loss=0.09097244346944186 +INFO:root: +[Valid][18][UTT][ MSE = 0.2801 | LCC = 0.8552 | SRCC = 0.8528 ] +INFO:root:Average loss=0.0894679606725008 +INFO:root: +[Valid][19][UTT][ MSE = 0.2802 | LCC = 0.8553 | SRCC = 0.8540 ] +INFO:root:Average loss=0.0854200243185728 +INFO:root: +[Valid][20][UTT][ MSE = 0.2849 | LCC = 0.8511 | SRCC = 0.8498 ] +INFO:root:Average loss=0.07749182076599354 +INFO:root: +[Valid][21][UTT][ MSE = 0.2464 | LCC = 0.8509 | SRCC = 0.8489 ] +INFO:root:Average loss=0.08018121018241613 +INFO:root: +[Valid][22][UTT][ MSE = 0.2577 | LCC = 0.8496 | SRCC = 0.8470 ] +INFO:root:Average loss=0.07675094138353299 +INFO:root: +[Valid][23][UTT][ MSE = 0.2565 | LCC = 0.8559 | SRCC = 0.8550 ] +INFO:root:Average loss=0.07114777957590726 +INFO:root: +[Valid][24][UTT][ MSE = 0.2500 | LCC = 0.8572 | SRCC = 0.8556 ] +INFO:root:Average loss=0.07218407180446845 +INFO:root: +[Valid][25][UTT][ MSE = 0.2724 | LCC = 0.8542 | SRCC = 0.8529 ] +INFO:root:Average loss=0.06565765515925029 +INFO:root: +[Valid][26][UTT][ MSE = 0.3200 | LCC = 0.8497 | SRCC = 0.8491 ] +INFO:root:Average loss=0.06484050790850933 +INFO:root: +[Valid][27][UTT][ MSE = 0.2821 | LCC = 0.8533 | SRCC = 0.8530 ] +INFO:root:Average loss=0.05910569073584599 +INFO:root: +[Valid][28][UTT][ MSE = 0.2573 | LCC = 0.8541 | SRCC = 0.8521 ] +INFO:root:Average loss=0.05720124665934306 +INFO:root: +[Valid][29][UTT][ MSE = 0.2467 | LCC = 0.8526 | SRCC = 0.8506 ] +INFO:root:Average loss=0.0585036365649639 +INFO:root: +[Valid][30][UTT][ MSE = 0.2456 | LCC = 0.8515 | SRCC = 0.8498 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2466 | LCC = 0.8655 | SRCC = 0.8666 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6702d3bf520ac170e563ac37e09a603181b32b94 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fae8fc863c2d4237c216da62bd6222524435b74bfe472f19e09b0ca3dd8ee230 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..53ab7f6d5cab99fd12b4f7605b035c17eecc2f8d --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1fc98a44e337c5477d04f9d0bc8389abf35e97d8971d11acddb6b7fa4894a06b +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..43560856f9dd984504a20fe054c53e0768863e63 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer8/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=0.33539584355476576 +INFO:root: +[Valid][0][UTT][ MSE = 0.2892 | LCC = 0.8424 | SRCC = 0.8412 ] +INFO:root:Average loss=0.22963937371969223 +INFO:root: +[Valid][1][UTT][ MSE = 0.2614 | LCC = 0.8533 | SRCC = 0.8522 ] +INFO:root:Average loss=0.21006104082633287 +INFO:root: +[Valid][2][UTT][ MSE = 0.2506 | LCC = 0.8597 | SRCC = 0.8595 ] +INFO:root:Average loss=0.19327130015843955 +INFO:root: +[Valid][3][UTT][ MSE = 0.2521 | LCC = 0.8633 | SRCC = 0.8611 ] +INFO:root:Average loss=0.18400177111228308 +INFO:root: +[Valid][4][UTT][ MSE = 0.2669 | LCC = 0.8609 | SRCC = 0.8616 ] +INFO:root:Average loss=0.17801389403832266 +INFO:root: +[Valid][5][UTT][ MSE = 0.2565 | LCC = 0.8663 | SRCC = 0.8666 ] +INFO:root:Average loss=0.1696953281569175 +INFO:root: +[Valid][6][UTT][ MSE = 0.2300 | LCC = 0.8676 | SRCC = 0.8671 ] +INFO:root:Average loss=0.156595951376053 +INFO:root: +[Valid][7][UTT][ MSE = 0.2327 | LCC = 0.8658 | SRCC = 0.8649 ] +INFO:root:Average loss=0.1464702450694182 +INFO:root: +[Valid][8][UTT][ MSE = 0.2442 | LCC = 0.8610 | SRCC = 0.8639 ] +INFO:root:Average loss=0.14080490382053915 +INFO:root: +[Valid][9][UTT][ MSE = 0.2273 | LCC = 0.8643 | SRCC = 0.8638 ] +INFO:root:Average loss=0.13742113419068167 +INFO:root: +[Valid][10][UTT][ MSE = 0.2259 | LCC = 0.8614 | SRCC = 0.8604 ] +INFO:root:Average loss=0.13125244337014663 +INFO:root: +[Valid][11][UTT][ MSE = 0.2652 | LCC = 0.8564 | SRCC = 0.8585 ] +INFO:root:Average loss=0.11927727055855286 +INFO:root: +[Valid][12][UTT][ MSE = 0.2540 | LCC = 0.8543 | SRCC = 0.8556 ] +INFO:root:Average loss=0.11544651108292434 +INFO:root: +[Valid][13][UTT][ MSE = 0.2508 | LCC = 0.8545 | SRCC = 0.8558 ] +INFO:root:Average loss=0.10529698183139165 +INFO:root: +[Valid][14][UTT][ MSE = 0.2744 | LCC = 0.8485 | SRCC = 0.8503 ] +INFO:root:Average loss=0.10135286330030514 +INFO:root: +[Valid][15][UTT][ MSE = 0.2316 | LCC = 0.8547 | SRCC = 0.8551 ] +INFO:root:Average loss=0.09705155820418628 +INFO:root: +[Valid][16][UTT][ MSE = 0.2882 | LCC = 0.8458 | SRCC = 0.8477 ] +INFO:root:Average loss=0.09080392695390262 +INFO:root: +[Valid][17][UTT][ MSE = 0.2567 | LCC = 0.8461 | SRCC = 0.8452 ] +INFO:root:Average loss=0.09107914184912658 +INFO:root: +[Valid][18][UTT][ MSE = 0.2631 | LCC = 0.8466 | SRCC = 0.8459 ] +INFO:root:Average loss=0.08465558376449805 +INFO:root: +[Valid][19][UTT][ MSE = 0.2470 | LCC = 0.8475 | SRCC = 0.8476 ] +INFO:root:Average loss=0.08160628407047345 +INFO:root: +[Valid][20][UTT][ MSE = 0.2487 | LCC = 0.8487 | SRCC = 0.8501 ] +INFO:root:Average loss=0.07418107867049865 +INFO:root: +[Valid][21][UTT][ MSE = 0.2351 | LCC = 0.8498 | SRCC = 0.8499 ] +INFO:root:Average loss=0.07649286545048921 +INFO:root: +[Valid][22][UTT][ MSE = 0.2521 | LCC = 0.8479 | SRCC = 0.8479 ] +INFO:root:Average loss=0.07194491781485386 +INFO:root: +[Valid][23][UTT][ MSE = 0.2587 | LCC = 0.8459 | SRCC = 0.8494 ] +INFO:root:Average loss=0.06491727262544326 +INFO:root: +[Valid][24][UTT][ MSE = 0.2498 | LCC = 0.8437 | SRCC = 0.8447 ] +INFO:root:Average loss=0.06615112712367988 +INFO:root: +[Valid][25][UTT][ MSE = 0.2459 | LCC = 0.8458 | SRCC = 0.8476 ] +INFO:root:Average loss=0.06572413368102832 +INFO:root: +[Valid][26][UTT][ MSE = 0.2563 | LCC = 0.8432 | SRCC = 0.8428 ] +INFO:root:Average loss=0.05975987141331037 +INFO:root: +[Valid][27][UTT][ MSE = 0.2578 | LCC = 0.8422 | SRCC = 0.8423 ] +INFO:root:Average loss=0.05931108855666258 +INFO:root: +[Valid][28][UTT][ MSE = 0.2537 | LCC = 0.8452 | SRCC = 0.8453 ] +INFO:root:Average loss=0.05625369583662504 +INFO:root: +[Valid][29][UTT][ MSE = 0.2529 | LCC = 0.8445 | SRCC = 0.8445 ] +INFO:root:Average loss=0.05521214356980263 +INFO:root: +[Valid][30][UTT][ MSE = 0.2645 | LCC = 0.8389 | SRCC = 0.8407 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2459 | LCC = 0.8632 | SRCC = 0.8672 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5dba528acd3a6fb0865eadddba97e3b8e794dcaf --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bed671d415e046380f471b0c34c0259f0ddbe846a9a5fa28148f9e754bdf5249 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..1b6157773d31b802f71b0a44a4fd5cbd3acf2f23 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a9d3936ed249a93e678c506621d052bbad0d442a0ce6ee0026e4ec24d4aaa60 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..400520ba63a43b7fb62e91c5516d03f60158e80f --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/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=0.3374752224637912 +INFO:root: +[Valid][0][UTT][ MSE = 0.3146 | LCC = 0.8393 | SRCC = 0.8408 ] +INFO:root:Average loss=0.24100803220883393 +INFO:root: +[Valid][1][UTT][ MSE = 0.2624 | LCC = 0.8445 | SRCC = 0.8504 ] +INFO:root:Average loss=0.22273944643063423 +INFO:root: +[Valid][2][UTT][ MSE = 0.2659 | LCC = 0.8607 | SRCC = 0.8597 ] +INFO:root:Average loss=0.21658872812986374 +INFO:root: +[Valid][3][UTT][ MSE = 0.2449 | LCC = 0.8618 | SRCC = 0.8602 ] +INFO:root:Average loss=0.21207503382212076 +INFO:root: +[Valid][4][UTT][ MSE = 0.2670 | LCC = 0.8656 | SRCC = 0.8643 ] +INFO:root:Average loss=0.19782311908709696 +INFO:root: +[Valid][5][UTT][ MSE = 0.2639 | LCC = 0.8616 | SRCC = 0.8624 ] +INFO:root:Average loss=0.19538876452507117 +INFO:root: +[Valid][6][UTT][ MSE = 0.2385 | LCC = 0.8620 | SRCC = 0.8616 ] +INFO:root:Average loss=0.18935415129630995 +INFO:root: +[Valid][7][UTT][ MSE = 0.2593 | LCC = 0.8641 | SRCC = 0.8649 ] +INFO:root:Average loss=0.18374603251234078 +INFO:root: +[Valid][8][UTT][ MSE = 0.2574 | LCC = 0.8717 | SRCC = 0.8718 ] +INFO:root:Average loss=0.18476076395465776 +INFO:root: +[Valid][9][UTT][ MSE = 0.2314 | LCC = 0.8692 | SRCC = 0.8703 ] +INFO:root:Average loss=0.17293868662837225 +INFO:root: +[Valid][10][UTT][ MSE = 0.2340 | LCC = 0.8697 | SRCC = 0.8730 ] +INFO:root:Average loss=0.16513366624712944 +INFO:root: +[Valid][11][UTT][ MSE = 0.2674 | LCC = 0.8704 | SRCC = 0.8697 ] +INFO:root:Average loss=0.1649537303317816 +INFO:root: +[Valid][12][UTT][ MSE = 0.2464 | LCC = 0.8702 | SRCC = 0.8693 ] +INFO:root:Average loss=0.1648052232578779 +INFO:root: +[Valid][13][UTT][ MSE = 0.2225 | LCC = 0.8681 | SRCC = 0.8685 ] +INFO:root:Average loss=0.15617140115071565 +INFO:root: +[Valid][14][UTT][ MSE = 0.2264 | LCC = 0.8625 | SRCC = 0.8653 ] +INFO:root:Average loss=0.15531296684191778 +INFO:root: +[Valid][15][UTT][ MSE = 0.2413 | LCC = 0.8667 | SRCC = 0.8638 ] +INFO:root:Average loss=0.14556145486541283 +INFO:root: +[Valid][16][UTT][ MSE = 0.2397 | LCC = 0.8683 | SRCC = 0.8674 ] +INFO:root:Average loss=0.14212462802728018 +INFO:root: +[Valid][17][UTT][ MSE = 0.2463 | LCC = 0.8647 | SRCC = 0.8617 ] +INFO:root:Average loss=0.13961797331770262 +INFO:root: +[Valid][18][UTT][ MSE = 0.2472 | LCC = 0.8598 | SRCC = 0.8623 ] +INFO:root:Average loss=0.13960601303439873 +INFO:root: +[Valid][19][UTT][ MSE = 0.2509 | LCC = 0.8659 | SRCC = 0.8650 ] +INFO:root:Average loss=0.13250696468047607 +INFO:root: +[Valid][20][UTT][ MSE = 0.2402 | LCC = 0.8648 | SRCC = 0.8659 ] +INFO:root:Average loss=0.1268687353302271 +INFO:root: +[Valid][21][UTT][ MSE = 0.2502 | LCC = 0.8619 | SRCC = 0.8617 ] +INFO:root:Average loss=0.12302642478010593 +INFO:root: +[Valid][22][UTT][ MSE = 0.2439 | LCC = 0.8607 | SRCC = 0.8625 ] +INFO:root:Average loss=0.11567003996326373 +INFO:root: +[Valid][23][UTT][ MSE = 0.2388 | LCC = 0.8593 | SRCC = 0.8605 ] +INFO:root:Average loss=0.11809138275491886 +INFO:root: +[Valid][24][UTT][ MSE = 0.2453 | LCC = 0.8621 | SRCC = 0.8634 ] +INFO:root:Average loss=0.11257699514046693 +INFO:root: +[Valid][25][UTT][ MSE = 0.2526 | LCC = 0.8544 | SRCC = 0.8575 ] +INFO:root:Average loss=0.10806516070778553 +INFO:root: +[Valid][26][UTT][ MSE = 0.2513 | LCC = 0.8533 | SRCC = 0.8514 ] +INFO:root:Average loss=0.10643826807156587 +INFO:root: +[Valid][27][UTT][ MSE = 0.2286 | LCC = 0.8585 | SRCC = 0.8588 ] +INFO:root:Average loss=0.10464804858351365 +INFO:root: +[Valid][28][UTT][ MSE = 0.2541 | LCC = 0.8540 | SRCC = 0.8523 ] +INFO:root:Average loss=0.1070517995991768 +INFO:root: +[Valid][29][UTT][ MSE = 0.2532 | LCC = 0.8532 | SRCC = 0.8533 ] +INFO:root:Average loss=0.09922454286462222 +INFO:root: +[Valid][30][UTT][ MSE = 0.2702 | LCC = 0.8518 | SRCC = 0.8491 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2845 | LCC = 0.8630 | SRCC = 0.8639 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..790f03d7b7de9a7ba5763f0c1e6eff9d25317235 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2aeb454bc574651872f4d74023cf18dd013c09d6819c7c09cec1cf15a18a0257 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b88ba2251e318a1b91eae76d188086d5a45299e5 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:756f9abb48294e4fced1aba8290a992319eddedeb7b34521bc00521157576ba5 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..4ebfc8c59ad1adcea5d497b124db1b1e61933d75 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/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=0.32314074555268657 +INFO:root: +[Valid][0][UTT][ MSE = 0.2742 | LCC = 0.8417 | SRCC = 0.8452 ] +INFO:root:Average loss=0.2379928273268235 +INFO:root: +[Valid][1][UTT][ MSE = 0.3197 | LCC = 0.8530 | SRCC = 0.8523 ] +INFO:root:Average loss=0.22616034536025462 +INFO:root: +[Valid][2][UTT][ MSE = 0.2606 | LCC = 0.8636 | SRCC = 0.8622 ] +INFO:root:Average loss=0.2150207088352778 +INFO:root: +[Valid][3][UTT][ MSE = 0.2736 | LCC = 0.8653 | SRCC = 0.8639 ] +INFO:root:Average loss=0.20934493323931327 +INFO:root: +[Valid][4][UTT][ MSE = 0.2923 | LCC = 0.8704 | SRCC = 0.8687 ] +INFO:root:Average loss=0.20556371009502655 +INFO:root: +[Valid][5][UTT][ MSE = 0.2517 | LCC = 0.8708 | SRCC = 0.8691 ] +INFO:root:Average loss=0.19261134797946 +INFO:root: +[Valid][6][UTT][ MSE = 0.2272 | LCC = 0.8666 | SRCC = 0.8676 ] +INFO:root:Average loss=0.18724369028439888 +INFO:root: +[Valid][7][UTT][ MSE = 0.2694 | LCC = 0.8717 | SRCC = 0.8712 ] +INFO:root:Average loss=0.18891055853321 +INFO:root: +[Valid][8][UTT][ MSE = 0.2408 | LCC = 0.8717 | SRCC = 0.8699 ] +INFO:root:Average loss=0.18382692623596925 +INFO:root: +[Valid][9][UTT][ MSE = 0.2749 | LCC = 0.8739 | SRCC = 0.8731 ] +INFO:root:Average loss=0.17674915120005608 +INFO:root: +[Valid][10][UTT][ MSE = 0.2235 | LCC = 0.8701 | SRCC = 0.8702 ] +INFO:root:Average loss=0.17748770041343492 +INFO:root: +[Valid][11][UTT][ MSE = 0.2362 | LCC = 0.8721 | SRCC = 0.8746 ] +INFO:root:Average loss=0.1707268425096304 +INFO:root: +[Valid][12][UTT][ MSE = 0.2198 | LCC = 0.8751 | SRCC = 0.8754 ] +INFO:root:Average loss=0.15984378554500067 +INFO:root: +[Valid][13][UTT][ MSE = 0.2260 | LCC = 0.8729 | SRCC = 0.8741 ] +INFO:root:Average loss=0.16092918192346892 +INFO:root: +[Valid][14][UTT][ MSE = 0.2434 | LCC = 0.8704 | SRCC = 0.8692 ] +INFO:root:Average loss=0.15854897818122154 +INFO:root: +[Valid][15][UTT][ MSE = 0.2498 | LCC = 0.8733 | SRCC = 0.8719 ] +INFO:root:Average loss=0.15239549504640776 +INFO:root: +[Valid][16][UTT][ MSE = 0.2497 | LCC = 0.8731 | SRCC = 0.8710 ] +INFO:root:Average loss=0.14545007546742758 +INFO:root: +[Valid][17][UTT][ MSE = 0.2411 | LCC = 0.8660 | SRCC = 0.8661 ] +INFO:root:Average loss=0.14091199044233713 +INFO:root: +[Valid][18][UTT][ MSE = 0.2521 | LCC = 0.8661 | SRCC = 0.8638 ] +INFO:root:Average loss=0.13393390560761476 +INFO:root: +[Valid][19][UTT][ MSE = 0.2436 | LCC = 0.8688 | SRCC = 0.8676 ] +INFO:root:Average loss=0.13241209729741782 +INFO:root: +[Valid][20][UTT][ MSE = 0.2396 | LCC = 0.8645 | SRCC = 0.8644 ] +INFO:root:Average loss=0.13016460759517473 +INFO:root: +[Valid][21][UTT][ MSE = 0.2317 | LCC = 0.8603 | SRCC = 0.8606 ] +INFO:root:Average loss=0.12432436415782341 +INFO:root: +[Valid][22][UTT][ MSE = 0.2347 | LCC = 0.8634 | SRCC = 0.8640 ] +INFO:root:Average loss=0.12528884974427712 +INFO:root: +[Valid][23][UTT][ MSE = 0.2506 | LCC = 0.8557 | SRCC = 0.8543 ] +INFO:root:Average loss=0.11570115721760652 +INFO:root: +[Valid][24][UTT][ MSE = 0.2154 | LCC = 0.8667 | SRCC = 0.8676 ] +INFO:root:Average loss=0.11184593232778403 +INFO:root: +[Valid][25][UTT][ MSE = 0.2517 | LCC = 0.8620 | SRCC = 0.8608 ] +INFO:root:Average loss=0.1071312461908047 +INFO:root: +[Valid][26][UTT][ MSE = 0.2274 | LCC = 0.8618 | SRCC = 0.8611 ] +INFO:root:Average loss=0.10893874615430832 +INFO:root: +[Valid][27][UTT][ MSE = 0.2338 | LCC = 0.8618 | SRCC = 0.8621 ] +INFO:root:Average loss=0.1064663537992881 +INFO:root: +[Valid][28][UTT][ MSE = 0.2352 | LCC = 0.8588 | SRCC = 0.8577 ] +INFO:root:Average loss=0.10233036065712953 +INFO:root: +[Valid][29][UTT][ MSE = 0.2581 | LCC = 0.8558 | SRCC = 0.8525 ] +INFO:root:Average loss=0.10014588701037261 +INFO:root: +[Valid][30][UTT][ MSE = 0.2733 | LCC = 0.8520 | SRCC = 0.8509 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2416 | LCC = 0.8646 | SRCC = 0.8697 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2c5a66bce9e1a1623acfb4cc91932b6dabbc4bff --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:73f46272cb67e765b07eb3d0807cb4803a0f1351756d49b93624f66a2b32a8cc +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..565b609ea29e3209a5ef855e7ce2bf095e930ee4 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:31331fcea176600d921ff7c8d635dafa255311bd27714bf0e897c4bf31d1e31f +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a7257fe17191c852161c3ebf11e40d29a08781cc --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/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=0.3147280596387692 +INFO:root: +[Valid][0][UTT][ MSE = 0.3801 | LCC = 0.8338 | SRCC = 0.8386 ] +INFO:root:Average loss=0.24670883325430062 +INFO:root: +[Valid][1][UTT][ MSE = 0.2904 | LCC = 0.8457 | SRCC = 0.8491 ] +INFO:root:Average loss=0.22737064537329552 +INFO:root: +[Valid][2][UTT][ MSE = 0.2588 | LCC = 0.8572 | SRCC = 0.8558 ] +INFO:root:Average loss=0.21514038206675115 +INFO:root: +[Valid][3][UTT][ MSE = 0.3197 | LCC = 0.8572 | SRCC = 0.8620 ] +INFO:root:Average loss=0.21232245614131293 +INFO:root: +[Valid][4][UTT][ MSE = 0.2502 | LCC = 0.8653 | SRCC = 0.8655 ] +INFO:root:Average loss=0.21098156817830527 +INFO:root: +[Valid][5][UTT][ MSE = 0.2982 | LCC = 0.8677 | SRCC = 0.8681 ] +INFO:root:Average loss=0.2008906153914256 +INFO:root: +[Valid][6][UTT][ MSE = 0.2612 | LCC = 0.8652 | SRCC = 0.8651 ] +INFO:root:Average loss=0.1928392272347059 +INFO:root: +[Valid][7][UTT][ MSE = 0.2642 | LCC = 0.8708 | SRCC = 0.8695 ] +INFO:root:Average loss=0.18984544754792482 +INFO:root: +[Valid][8][UTT][ MSE = 0.2517 | LCC = 0.8696 | SRCC = 0.8705 ] +INFO:root:Average loss=0.18416791991927686 +INFO:root: +[Valid][9][UTT][ MSE = 0.2761 | LCC = 0.8712 | SRCC = 0.8696 ] +INFO:root:Average loss=0.17867028350249314 +INFO:root: +[Valid][10][UTT][ MSE = 0.2360 | LCC = 0.8679 | SRCC = 0.8700 ] +INFO:root:Average loss=0.17545425557555297 +INFO:root: +[Valid][11][UTT][ MSE = 0.2773 | LCC = 0.8704 | SRCC = 0.8695 ] +INFO:root:Average loss=0.16568840505220953 +INFO:root: +[Valid][12][UTT][ MSE = 0.2221 | LCC = 0.8720 | SRCC = 0.8730 ] +INFO:root:Average loss=0.16872546946009 +INFO:root: +[Valid][13][UTT][ MSE = 0.2435 | LCC = 0.8686 | SRCC = 0.8695 ] +INFO:root:Average loss=0.15833164837497932 +INFO:root: +[Valid][14][UTT][ MSE = 0.2784 | LCC = 0.8692 | SRCC = 0.8685 ] +INFO:root:Average loss=0.16117834557707494 +INFO:root: +[Valid][15][UTT][ MSE = 0.2645 | LCC = 0.8671 | SRCC = 0.8722 ] +INFO:root:Average loss=0.158066943860971 +INFO:root: +[Valid][16][UTT][ MSE = 0.2761 | LCC = 0.8690 | SRCC = 0.8701 ] +INFO:root:Average loss=0.1530397287928141 +INFO:root: +[Valid][17][UTT][ MSE = 0.2477 | LCC = 0.8685 | SRCC = 0.8714 ] +INFO:root:Average loss=0.14696078126629195 +INFO:root: +[Valid][18][UTT][ MSE = 0.2706 | LCC = 0.8687 | SRCC = 0.8678 ] +INFO:root:Average loss=0.14053354517389566 +INFO:root: +[Valid][19][UTT][ MSE = 0.2797 | LCC = 0.8702 | SRCC = 0.8691 ] +INFO:root:Average loss=0.13607226388576704 +INFO:root: +[Valid][20][UTT][ MSE = 0.2260 | LCC = 0.8618 | SRCC = 0.8639 ] +INFO:root:Average loss=0.13224574913963294 +INFO:root: +[Valid][21][UTT][ MSE = 0.2564 | LCC = 0.8662 | SRCC = 0.8668 ] +INFO:root:Average loss=0.13053781118912575 +INFO:root: +[Valid][22][UTT][ MSE = 0.2607 | LCC = 0.8714 | SRCC = 0.8726 ] +INFO:root:Average loss=0.12534345629123542 +INFO:root: +[Valid][23][UTT][ MSE = 0.2453 | LCC = 0.8631 | SRCC = 0.8645 ] +INFO:root:Average loss=0.12277458168757267 +INFO:root: +[Valid][24][UTT][ MSE = 0.2438 | LCC = 0.8645 | SRCC = 0.8660 ] +INFO:root:Average loss=0.12218824086280969 +INFO:root: +[Valid][25][UTT][ MSE = 0.2798 | LCC = 0.8652 | SRCC = 0.8653 ] +INFO:root:Average loss=0.11199262824196082 +INFO:root: +[Valid][26][UTT][ MSE = 0.2339 | LCC = 0.8660 | SRCC = 0.8676 ] +INFO:root:Average loss=0.11037853618080799 +INFO:root: +[Valid][27][UTT][ MSE = 0.2383 | LCC = 0.8649 | SRCC = 0.8644 ] +INFO:root:Average loss=0.10889232120452783 +INFO:root: +[Valid][28][UTT][ MSE = 0.2562 | LCC = 0.8614 | SRCC = 0.8607 ] +INFO:root:Average loss=0.10779202643495339 +INFO:root: +[Valid][29][UTT][ MSE = 0.2389 | LCC = 0.8626 | SRCC = 0.8629 ] +INFO:root:Average loss=0.10634962182778579 +INFO:root: +[Valid][30][UTT][ MSE = 0.2463 | LCC = 0.8657 | SRCC = 0.8646 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2429 | LCC = 0.8604 | SRCC = 0.8638 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..250c402c9cda5bc8a1e2f757152e36212c86e1c1 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:69aa71f643a064370dc3c24d18da8260d42a781b2a701461ceb0d2c88608adc0 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b666991dba406e5b39707f0e9a6c6738db9ed882 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:54751452b4dc8f5be93f94aa74dde308860dc9c9ba435f869c2d0f8ec9be2b1e +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2ea0e1d4106b6023b6509e3b6131469059f18819 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/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=0.3566495657731325 +INFO:root: +[Valid][0][UTT][ MSE = 0.3297 | LCC = 0.8317 | SRCC = 0.8364 ] +INFO:root:Average loss=0.24202038634281892 +INFO:root: +[Valid][1][UTT][ MSE = 0.3206 | LCC = 0.8527 | SRCC = 0.8527 ] +INFO:root:Average loss=0.22760958234087014 +INFO:root: +[Valid][2][UTT][ MSE = 0.3096 | LCC = 0.8611 | SRCC = 0.8592 ] +INFO:root:Average loss=0.21062747369974089 +INFO:root: +[Valid][3][UTT][ MSE = 0.3188 | LCC = 0.8648 | SRCC = 0.8661 ] +INFO:root:Average loss=0.20836834743236884 +INFO:root: +[Valid][4][UTT][ MSE = 0.2622 | LCC = 0.8661 | SRCC = 0.8668 ] +INFO:root:Average loss=0.20127289035381415 +INFO:root: +[Valid][5][UTT][ MSE = 0.2988 | LCC = 0.8698 | SRCC = 0.8689 ] +INFO:root:Average loss=0.20009648436919236 +INFO:root: +[Valid][6][UTT][ MSE = 0.2975 | LCC = 0.8700 | SRCC = 0.8686 ] +INFO:root:Average loss=0.19689018584978887 +INFO:root: +[Valid][7][UTT][ MSE = 0.2587 | LCC = 0.8688 | SRCC = 0.8696 ] +INFO:root:Average loss=0.18380980250927118 +INFO:root: +[Valid][8][UTT][ MSE = 0.2397 | LCC = 0.8688 | SRCC = 0.8706 ] +INFO:root:Average loss=0.18432054792841276 +INFO:root: +[Valid][9][UTT][ MSE = 0.2902 | LCC = 0.8637 | SRCC = 0.8653 ] +INFO:root:Average loss=0.17528958408496317 +INFO:root: +[Valid][10][UTT][ MSE = 0.2292 | LCC = 0.8691 | SRCC = 0.8725 ] +INFO:root:Average loss=0.17396186597836324 +INFO:root: +[Valid][11][UTT][ MSE = 0.2741 | LCC = 0.8703 | SRCC = 0.8711 ] +INFO:root:Average loss=0.16358904015177336 +INFO:root: +[Valid][12][UTT][ MSE = 0.2740 | LCC = 0.8688 | SRCC = 0.8689 ] +INFO:root:Average loss=0.1600622258698329 +INFO:root: +[Valid][13][UTT][ MSE = 0.2543 | LCC = 0.8703 | SRCC = 0.8716 ] +INFO:root:Average loss=0.15712316534840143 +INFO:root: +[Valid][14][UTT][ MSE = 0.2605 | LCC = 0.8676 | SRCC = 0.8687 ] +INFO:root:Average loss=0.15368947530022034 +INFO:root: +[Valid][15][UTT][ MSE = 0.2788 | LCC = 0.8663 | SRCC = 0.8655 ] +INFO:root:Average loss=0.14922164170405802 +INFO:root: +[Valid][16][UTT][ MSE = 0.2514 | LCC = 0.8695 | SRCC = 0.8705 ] +INFO:root:Average loss=0.14638779417444497 +INFO:root: +[Valid][17][UTT][ MSE = 0.3053 | LCC = 0.8665 | SRCC = 0.8673 ] +INFO:root:Average loss=0.1427592018093818 +INFO:root: +[Valid][18][UTT][ MSE = 0.2502 | LCC = 0.8654 | SRCC = 0.8662 ] +INFO:root:Average loss=0.1349960346825612 +INFO:root: +[Valid][19][UTT][ MSE = 0.2737 | LCC = 0.8662 | SRCC = 0.8674 ] +INFO:root:Average loss=0.12976453099877405 +INFO:root: +[Valid][20][UTT][ MSE = 0.2788 | LCC = 0.8624 | SRCC = 0.8618 ] +INFO:root:Average loss=0.1309770250167602 +INFO:root: +[Valid][21][UTT][ MSE = 0.2877 | LCC = 0.8631 | SRCC = 0.8648 ] +INFO:root:Average loss=0.11989522696687625 +INFO:root: +[Valid][22][UTT][ MSE = 0.2748 | LCC = 0.8561 | SRCC = 0.8574 ] +INFO:root:Average loss=0.11669113515661313 +INFO:root: +[Valid][23][UTT][ MSE = 0.2545 | LCC = 0.8607 | SRCC = 0.8620 ] +INFO:root:Average loss=0.1116175244633968 +INFO:root: +[Valid][24][UTT][ MSE = 0.2485 | LCC = 0.8542 | SRCC = 0.8551 ] +INFO:root:Average loss=0.11187973561195227 +INFO:root: +[Valid][25][UTT][ MSE = 0.2814 | LCC = 0.8586 | SRCC = 0.8590 ] +INFO:root:Average loss=0.10822046767824735 +INFO:root: +[Valid][26][UTT][ MSE = 0.2633 | LCC = 0.8540 | SRCC = 0.8559 ] +INFO:root:Average loss=0.10317983440099618 +INFO:root: +[Valid][27][UTT][ MSE = 0.2819 | LCC = 0.8549 | SRCC = 0.8549 ] +INFO:root:Average loss=0.10666134065160385 +INFO:root: +[Valid][28][UTT][ MSE = 0.2694 | LCC = 0.8541 | SRCC = 0.8536 ] +INFO:root:Average loss=0.10008100496652798 +INFO:root: +[Valid][29][UTT][ MSE = 0.2596 | LCC = 0.8501 | SRCC = 0.8523 ] +INFO:root:Average loss=0.0943373994758496 +INFO:root: +[Valid][30][UTT][ MSE = 0.2610 | LCC = 0.8453 | SRCC = 0.8470 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2548 | LCC = 0.8662 | SRCC = 0.8652 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..08095b3775f035a8d9a5e8df7ebc25b18e9f9be4 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e20bf2b4b2ba6f21ea9dbaad02ff85c62b89f27e2b064e1b16c56815891ef36 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b88cdc2a43a9c801b87761246c2053fc88fde715 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0604f254ff00741fc3488beadaa5dcdccec9d65819fc1fc8d31aa22e21ba5109 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1fd60cd11f72dc214256b0ed1088129bee9e2a7a --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_8epoch/layer9/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=0.30180098039981645 +INFO:root: +[Valid][0][UTT][ MSE = 0.3151 | LCC = 0.8377 | SRCC = 0.8381 ] +INFO:root:Average loss=0.23587484629108355 +INFO:root: +[Valid][1][UTT][ MSE = 0.2809 | LCC = 0.8512 | SRCC = 0.8523 ] +INFO:root:Average loss=0.2299358892517212 +INFO:root: +[Valid][2][UTT][ MSE = 0.2677 | LCC = 0.8585 | SRCC = 0.8574 ] +INFO:root:Average loss=0.2133918780929003 +INFO:root: +[Valid][3][UTT][ MSE = 0.2562 | LCC = 0.8578 | SRCC = 0.8586 ] +INFO:root:Average loss=0.21536630449386743 +INFO:root: +[Valid][4][UTT][ MSE = 0.2885 | LCC = 0.8608 | SRCC = 0.8597 ] +INFO:root:Average loss=0.20662457916216972 +INFO:root: +[Valid][5][UTT][ MSE = 0.2700 | LCC = 0.8638 | SRCC = 0.8611 ] +INFO:root:Average loss=0.20109384373212472 +INFO:root: +[Valid][6][UTT][ MSE = 0.2616 | LCC = 0.8682 | SRCC = 0.8678 ] +INFO:root:Average loss=0.1967502956589063 +INFO:root: +[Valid][7][UTT][ MSE = 0.2525 | LCC = 0.8689 | SRCC = 0.8678 ] +INFO:root:Average loss=0.18299058500008705 +INFO:root: +[Valid][8][UTT][ MSE = 0.2815 | LCC = 0.8720 | SRCC = 0.8710 ] +INFO:root:Average loss=0.18346101609178078 +INFO:root: +[Valid][9][UTT][ MSE = 0.2925 | LCC = 0.8693 | SRCC = 0.8700 ] +INFO:root:Average loss=0.17857612325594976 +INFO:root: +[Valid][10][UTT][ MSE = 0.2457 | LCC = 0.8676 | SRCC = 0.8703 ] +INFO:root:Average loss=0.1703458764614203 +INFO:root: +[Valid][11][UTT][ MSE = 0.2298 | LCC = 0.8669 | SRCC = 0.8704 ] +INFO:root:Average loss=0.17088847177533004 +INFO:root: +[Valid][12][UTT][ MSE = 0.2855 | LCC = 0.8666 | SRCC = 0.8666 ] +INFO:root:Average loss=0.16352194280196458 +INFO:root: +[Valid][13][UTT][ MSE = 0.2825 | LCC = 0.8694 | SRCC = 0.8684 ] +INFO:root:Average loss=0.1622056238926374 +INFO:root: +[Valid][14][UTT][ MSE = 0.2245 | LCC = 0.8681 | SRCC = 0.8711 ] +INFO:root:Average loss=0.15217787753313017 +INFO:root: +[Valid][15][UTT][ MSE = 0.2382 | LCC = 0.8698 | SRCC = 0.8703 ] +INFO:root:Average loss=0.14353318674824175 +INFO:root: +[Valid][16][UTT][ MSE = 0.2362 | LCC = 0.8706 | SRCC = 0.8724 ] +INFO:root:Average loss=0.14853763131377024 +INFO:root: +[Valid][17][UTT][ MSE = 0.2442 | LCC = 0.8645 | SRCC = 0.8659 ] +INFO:root:Average loss=0.14218709054283607 +INFO:root: +[Valid][18][UTT][ MSE = 0.2358 | LCC = 0.8669 | SRCC = 0.8689 ] +INFO:root:Average loss=0.14258197045479065 +INFO:root: +[Valid][19][UTT][ MSE = 0.2214 | LCC = 0.8688 | SRCC = 0.8705 ] +INFO:root:Average loss=0.13407244705236876 +INFO:root: +[Valid][20][UTT][ MSE = 0.2319 | LCC = 0.8660 | SRCC = 0.8676 ] +INFO:root:Average loss=0.13213319226335257 +INFO:root: +[Valid][21][UTT][ MSE = 0.2596 | LCC = 0.8650 | SRCC = 0.8655 ] +INFO:root:Average loss=0.12491960307726493 +INFO:root: +[Valid][22][UTT][ MSE = 0.2470 | LCC = 0.8703 | SRCC = 0.8686 ] +INFO:root:Average loss=0.12052172126296239 +INFO:root: +[Valid][23][UTT][ MSE = 0.2897 | LCC = 0.8615 | SRCC = 0.8643 ] +INFO:root:Average loss=0.12742517630641276 +INFO:root: +[Valid][24][UTT][ MSE = 0.2692 | LCC = 0.8651 | SRCC = 0.8655 ] +INFO:root:Average loss=0.11616714279621075 +INFO:root: +[Valid][25][UTT][ MSE = 0.2498 | LCC = 0.8660 | SRCC = 0.8677 ] +INFO:root:Average loss=0.11356173255122624 +INFO:root: +[Valid][26][UTT][ MSE = 0.2622 | LCC = 0.8623 | SRCC = 0.8625 ] +INFO:root:Average loss=0.1127744204340837 +INFO:root: +[Valid][27][UTT][ MSE = 0.2473 | LCC = 0.8670 | SRCC = 0.8669 ] +INFO:root:Average loss=0.10642012313772471 +INFO:root: +[Valid][28][UTT][ MSE = 0.2583 | LCC = 0.8633 | SRCC = 0.8636 ] +INFO:root:Average loss=0.10105553866387942 +INFO:root: +[Valid][29][UTT][ MSE = 0.2381 | LCC = 0.8614 | SRCC = 0.8625 ] +INFO:root:Average loss=0.101273332698605 +INFO:root: +[Valid][30][UTT][ MSE = 0.2579 | LCC = 0.8631 | SRCC = 0.8657 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2797 | LCC = 0.8655 | SRCC = 0.8650 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9eefe1ff5ce2dab52e6610a122aaaf2624be6791 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0be51baa92e5c73b41e7bf8e2ef14e517506acc7645ae8725f285307879ad314 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..11cf5abe121c794676259fceafc45e408aeeafa0 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:80b208843da0668a40b962154a7fabec85209ef01ae360293178c93ee6055e8c +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..55f3585a72ca05cdfa981a50a4e1a105c48a8908 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/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=0.5069931436043519 +INFO:root: +[Valid][0][UTT][ MSE = 0.3876 | LCC = 0.7611 | SRCC = 0.7684 ] +INFO:root:Average loss=0.2981251866007463 +INFO:root: +[Valid][1][UTT][ MSE = 0.3297 | LCC = 0.8034 | SRCC = 0.8061 ] +INFO:root:Average loss=0.2599213209289771 +INFO:root: +[Valid][2][UTT][ MSE = 0.3180 | LCC = 0.8178 | SRCC = 0.8209 ] +INFO:root:Average loss=0.25045126389998656 +INFO:root: +[Valid][3][UTT][ MSE = 0.2841 | LCC = 0.8283 | SRCC = 0.8271 ] +INFO:root:Average loss=0.2242409619383323 +INFO:root: +[Valid][4][UTT][ MSE = 0.2923 | LCC = 0.8324 | SRCC = 0.8315 ] +INFO:root:Average loss=0.21899355260225442 +INFO:root: +[Valid][5][UTT][ MSE = 0.2792 | LCC = 0.8331 | SRCC = 0.8346 ] +INFO:root:Average loss=0.20133660771907905 +INFO:root: +[Valid][6][UTT][ MSE = 0.2627 | LCC = 0.8404 | SRCC = 0.8416 ] +INFO:root:Average loss=0.186894341921195 +INFO:root: +[Valid][7][UTT][ MSE = 0.2633 | LCC = 0.8378 | SRCC = 0.8392 ] +INFO:root:Average loss=0.18249704726995566 +INFO:root: +[Valid][8][UTT][ MSE = 0.2626 | LCC = 0.8465 | SRCC = 0.8451 ] +INFO:root:Average loss=0.17568492411802977 +INFO:root: +[Valid][9][UTT][ MSE = 0.2619 | LCC = 0.8463 | SRCC = 0.8439 ] +INFO:root:Average loss=0.16606610803267893 +INFO:root: +[Valid][10][UTT][ MSE = 0.2571 | LCC = 0.8451 | SRCC = 0.8460 ] +INFO:root:Average loss=0.15945641658244988 +INFO:root: +[Valid][11][UTT][ MSE = 0.2471 | LCC = 0.8456 | SRCC = 0.8469 ] +INFO:root:Average loss=0.15543542974270308 +INFO:root: +[Valid][12][UTT][ MSE = 0.2491 | LCC = 0.8494 | SRCC = 0.8495 ] +INFO:root:Average loss=0.15529696777080879 +INFO:root: +[Valid][13][UTT][ MSE = 0.2507 | LCC = 0.8496 | SRCC = 0.8492 ] +INFO:root:Average loss=0.13937369466592103 +INFO:root: +[Valid][14][UTT][ MSE = 0.2932 | LCC = 0.8538 | SRCC = 0.8508 ] +INFO:root:Average loss=0.1356993307096836 +INFO:root: +[Valid][15][UTT][ MSE = 0.2624 | LCC = 0.8508 | SRCC = 0.8478 ] +INFO:root:Average loss=0.1307106754527642 +INFO:root: +[Valid][16][UTT][ MSE = 0.2610 | LCC = 0.8449 | SRCC = 0.8473 ] +INFO:root:Average loss=0.13352255179331854 +INFO:root: +[Valid][17][UTT][ MSE = 0.2430 | LCC = 0.8527 | SRCC = 0.8512 ] +INFO:root:Average loss=0.11678477758780503 +INFO:root: +[Valid][18][UTT][ MSE = 0.2481 | LCC = 0.8531 | SRCC = 0.8504 ] +INFO:root:Average loss=0.11832216506203015 +INFO:root: +[Valid][19][UTT][ MSE = 0.2614 | LCC = 0.8525 | SRCC = 0.8512 ] +INFO:root:Average loss=0.11552679395446411 +INFO:root: +[Valid][20][UTT][ MSE = 0.2586 | LCC = 0.8440 | SRCC = 0.8442 ] +INFO:root:Average loss=0.11134337586088058 +INFO:root: +[Valid][21][UTT][ MSE = 0.2897 | LCC = 0.8520 | SRCC = 0.8506 ] +INFO:root:Average loss=0.11000683110875961 +INFO:root: +[Valid][22][UTT][ MSE = 0.2437 | LCC = 0.8546 | SRCC = 0.8515 ] +INFO:root:Average loss=0.10187494153013596 +INFO:root: +[Valid][23][UTT][ MSE = 0.2461 | LCC = 0.8469 | SRCC = 0.8469 ] +INFO:root:Average loss=0.09864137187982217 +INFO:root: +[Valid][24][UTT][ MSE = 0.2540 | LCC = 0.8404 | SRCC = 0.8440 ] +INFO:root:Average loss=0.10714348434255673 +INFO:root: +[Valid][25][UTT][ MSE = 0.2431 | LCC = 0.8482 | SRCC = 0.8486 ] +INFO:root:Average loss=0.09008235221680923 +INFO:root: +[Valid][26][UTT][ MSE = 0.2877 | LCC = 0.8503 | SRCC = 0.8505 ] +INFO:root:Average loss=0.0932086800248959 +INFO:root: +[Valid][27][UTT][ MSE = 0.2523 | LCC = 0.8453 | SRCC = 0.8430 ] +INFO:root:Average loss=0.08813748666300224 +INFO:root: +[Valid][28][UTT][ MSE = 0.2472 | LCC = 0.8486 | SRCC = 0.8467 ] +INFO:root:Average loss=0.09062106678119072 +INFO:root: +[Valid][29][UTT][ MSE = 0.2455 | LCC = 0.8522 | SRCC = 0.8501 ] +INFO:root:Average loss=0.07973790627259475 +INFO:root: +[Valid][30][UTT][ MSE = 0.2405 | LCC = 0.8477 | SRCC = 0.8477 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2570 | LCC = 0.8506 | SRCC = 0.8491 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..912d5be9e713ad31a5ba8f641614c2482d8aa60b --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:76476bc7682e3581fa0ee6846655bc14c1d00912d6c4006c76fc0a1f5671b517 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e4e925d8242f5ffcfa35a03a9080ed81e9e73415 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6302d32ef448f048a80cbfa23112c4b789e388b654060c4887e201cc1c20939c +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..55f19ce33b27fa883a67c3d0f985a14df59ebbec --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/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=0.5999320764572192 +INFO:root: +[Valid][0][UTT][ MSE = 0.4250 | LCC = 0.7425 | SRCC = 0.7434 ] +INFO:root:Average loss=0.312521572296436 +INFO:root: +[Valid][1][UTT][ MSE = 0.3184 | LCC = 0.7988 | SRCC = 0.8033 ] +INFO:root:Average loss=0.2618719450174234 +INFO:root: +[Valid][2][UTT][ MSE = 0.2920 | LCC = 0.8197 | SRCC = 0.8204 ] +INFO:root:Average loss=0.23923145139064544 +INFO:root: +[Valid][3][UTT][ MSE = 0.2766 | LCC = 0.8233 | SRCC = 0.8281 ] +INFO:root:Average loss=0.22291145091637588 +INFO:root: +[Valid][4][UTT][ MSE = 0.2956 | LCC = 0.8371 | SRCC = 0.8384 ] +INFO:root:Average loss=0.21545908810236516 +INFO:root: +[Valid][5][UTT][ MSE = 0.2578 | LCC = 0.8415 | SRCC = 0.8408 ] +INFO:root:Average loss=0.1952956791680593 +INFO:root: +[Valid][6][UTT][ MSE = 0.3090 | LCC = 0.8416 | SRCC = 0.8468 ] +INFO:root:Average loss=0.19075065058393356 +INFO:root: +[Valid][7][UTT][ MSE = 0.2538 | LCC = 0.8475 | SRCC = 0.8505 ] +INFO:root:Average loss=0.18499762603105643 +INFO:root: +[Valid][8][UTT][ MSE = 0.2520 | LCC = 0.8493 | SRCC = 0.8499 ] +INFO:root:Average loss=0.17532505133213142 +INFO:root: +[Valid][9][UTT][ MSE = 0.2392 | LCC = 0.8523 | SRCC = 0.8542 ] +INFO:root:Average loss=0.1627808555196493 +INFO:root: +[Valid][10][UTT][ MSE = 0.2358 | LCC = 0.8565 | SRCC = 0.8576 ] +INFO:root:Average loss=0.15520545945335656 +INFO:root: +[Valid][11][UTT][ MSE = 0.2427 | LCC = 0.8518 | SRCC = 0.8545 ] +INFO:root:Average loss=0.14771947427055773 +INFO:root: +[Valid][12][UTT][ MSE = 0.2586 | LCC = 0.8501 | SRCC = 0.8533 ] +INFO:root:Average loss=0.13974995423968023 +INFO:root: +[Valid][13][UTT][ MSE = 0.2764 | LCC = 0.8537 | SRCC = 0.8558 ] +INFO:root:Average loss=0.13275694837554908 +INFO:root: +[Valid][14][UTT][ MSE = 0.2610 | LCC = 0.8582 | SRCC = 0.8584 ] +INFO:root:Average loss=0.12979425652286944 +INFO:root: +[Valid][15][UTT][ MSE = 0.2324 | LCC = 0.8536 | SRCC = 0.8567 ] +INFO:root:Average loss=0.12364373699976848 +INFO:root: +[Valid][16][UTT][ MSE = 0.2278 | LCC = 0.8603 | SRCC = 0.8622 ] +INFO:root:Average loss=0.1150916334337149 +INFO:root: +[Valid][17][UTT][ MSE = 0.2463 | LCC = 0.8587 | SRCC = 0.8590 ] +INFO:root:Average loss=0.12071888177440716 +INFO:root: +[Valid][18][UTT][ MSE = 0.2513 | LCC = 0.8574 | SRCC = 0.8591 ] +INFO:root:Average loss=0.10744189862639476 +INFO:root: +[Valid][19][UTT][ MSE = 0.2256 | LCC = 0.8594 | SRCC = 0.8602 ] +INFO:root:Average loss=0.10598772363020824 +INFO:root: +[Valid][20][UTT][ MSE = 0.2215 | LCC = 0.8569 | SRCC = 0.8585 ] +INFO:root:Average loss=0.09707762831105636 +INFO:root: +[Valid][21][UTT][ MSE = 0.2650 | LCC = 0.8580 | SRCC = 0.8614 ] +INFO:root:Average loss=0.09692081331442563 +INFO:root: +[Valid][22][UTT][ MSE = 0.2247 | LCC = 0.8598 | SRCC = 0.8606 ] +INFO:root:Average loss=0.10188132505386303 +INFO:root: +[Valid][23][UTT][ MSE = 0.2549 | LCC = 0.8598 | SRCC = 0.8597 ] +INFO:root:Average loss=0.10317622946623044 +INFO:root: +[Valid][24][UTT][ MSE = 0.2501 | LCC = 0.8600 | SRCC = 0.8602 ] +INFO:root:Average loss=0.09158509344053574 +INFO:root: +[Valid][25][UTT][ MSE = 0.2216 | LCC = 0.8603 | SRCC = 0.8606 ] +INFO:root:Average loss=0.0862731783149334 +INFO:root: +[Valid][26][UTT][ MSE = 0.2207 | LCC = 0.8617 | SRCC = 0.8619 ] +INFO:root:Average loss=0.08408229387341401 +INFO:root: +[Valid][27][UTT][ MSE = 0.2565 | LCC = 0.8549 | SRCC = 0.8574 ] +INFO:root:Average loss=0.08516920859423968 +INFO:root: +[Valid][28][UTT][ MSE = 0.2286 | LCC = 0.8576 | SRCC = 0.8582 ] +INFO:root:Average loss=0.0801646198408726 +INFO:root: +[Valid][29][UTT][ MSE = 0.2618 | LCC = 0.8523 | SRCC = 0.8556 ] +INFO:root:Average loss=0.08514498393887128 +INFO:root: +[Valid][30][UTT][ MSE = 0.2821 | LCC = 0.8539 | SRCC = 0.8573 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2286 | LCC = 0.8562 | SRCC = 0.8559 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ed2b8151dc39a25179fdee4cee2fd77d2b0ca0ef --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bb08f71694a03342130b9a870edd12260f60abeedda85e5748f6ec91d3e90f8e +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..831064a5281a0aa65ebcc954ed9f2d9b6b780743 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:863bd96cf1b441116a5b3c79e07efb1082faecaf9463f7cdb5bef0feca037688 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5a00e5ec3f5abbf3de6a0477912adea428774d08 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/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=0.5357908073526162 +INFO:root: +[Valid][0][UTT][ MSE = 0.3902 | LCC = 0.7474 | SRCC = 0.7593 ] +INFO:root:Average loss=0.30399971417127514 +INFO:root: +[Valid][1][UTT][ MSE = 0.3284 | LCC = 0.8060 | SRCC = 0.8123 ] +INFO:root:Average loss=0.2582993635382408 +INFO:root: +[Valid][2][UTT][ MSE = 0.2864 | LCC = 0.8240 | SRCC = 0.8251 ] +INFO:root:Average loss=0.24106423938885713 +INFO:root: +[Valid][3][UTT][ MSE = 0.2800 | LCC = 0.8280 | SRCC = 0.8296 ] +INFO:root:Average loss=0.222515404319916 +INFO:root: +[Valid][4][UTT][ MSE = 0.2717 | LCC = 0.8350 | SRCC = 0.8352 ] +INFO:root:Average loss=0.20424967450209153 +INFO:root: +[Valid][5][UTT][ MSE = 0.2538 | LCC = 0.8407 | SRCC = 0.8410 ] +INFO:root:Average loss=0.19998688976734114 +INFO:root: +[Valid][6][UTT][ MSE = 0.2582 | LCC = 0.8490 | SRCC = 0.8469 ] +INFO:root:Average loss=0.18204938477048507 +INFO:root: +[Valid][7][UTT][ MSE = 0.2876 | LCC = 0.8498 | SRCC = 0.8490 ] +INFO:root:Average loss=0.17833140960488564 +INFO:root: +[Valid][8][UTT][ MSE = 0.2426 | LCC = 0.8480 | SRCC = 0.8473 ] +INFO:root:Average loss=0.1625711923608413 +INFO:root: +[Valid][9][UTT][ MSE = 0.2509 | LCC = 0.8474 | SRCC = 0.8479 ] +INFO:root:Average loss=0.15974046633793756 +INFO:root: +[Valid][10][UTT][ MSE = 0.2549 | LCC = 0.8485 | SRCC = 0.8472 ] +INFO:root:Average loss=0.16068860582816294 +INFO:root: +[Valid][11][UTT][ MSE = 0.2457 | LCC = 0.8515 | SRCC = 0.8493 ] +INFO:root:Average loss=0.14271195758229646 +INFO:root: +[Valid][12][UTT][ MSE = 0.2290 | LCC = 0.8525 | SRCC = 0.8521 ] +INFO:root:Average loss=0.1337090314198763 +INFO:root: +[Valid][13][UTT][ MSE = 0.2440 | LCC = 0.8481 | SRCC = 0.8478 ] +INFO:root:Average loss=0.13768815354276925 +INFO:root: +[Valid][14][UTT][ MSE = 0.2331 | LCC = 0.8481 | SRCC = 0.8488 ] +INFO:root:Average loss=0.1242494750290345 +INFO:root: +[Valid][15][UTT][ MSE = 0.2390 | LCC = 0.8497 | SRCC = 0.8498 ] +INFO:root:Average loss=0.12353736620682937 +INFO:root: +[Valid][16][UTT][ MSE = 0.2440 | LCC = 0.8523 | SRCC = 0.8500 ] +INFO:root:Average loss=0.1131704368461401 +INFO:root: +[Valid][17][UTT][ MSE = 0.2544 | LCC = 0.8473 | SRCC = 0.8476 ] +INFO:root:Average loss=0.11062274195062809 +INFO:root: +[Valid][18][UTT][ MSE = 0.2272 | LCC = 0.8576 | SRCC = 0.8545 ] +INFO:root:Average loss=0.10761226837833722 +INFO:root: +[Valid][19][UTT][ MSE = 0.2358 | LCC = 0.8576 | SRCC = 0.8572 ] +INFO:root:Average loss=0.10005882716713807 +INFO:root: +[Valid][20][UTT][ MSE = 0.2319 | LCC = 0.8518 | SRCC = 0.8516 ] +INFO:root:Average loss=0.10330712078855588 +INFO:root: +[Valid][21][UTT][ MSE = 0.2341 | LCC = 0.8575 | SRCC = 0.8567 ] +INFO:root:Average loss=0.09903657565323207 +INFO:root: +[Valid][22][UTT][ MSE = 0.2242 | LCC = 0.8564 | SRCC = 0.8542 ] +INFO:root:Average loss=0.09434961183713032 +INFO:root: +[Valid][23][UTT][ MSE = 0.2263 | LCC = 0.8552 | SRCC = 0.8528 ] +INFO:root:Average loss=0.09345405004345454 +INFO:root: +[Valid][24][UTT][ MSE = 0.2380 | LCC = 0.8492 | SRCC = 0.8502 ] +INFO:root:Average loss=0.0874756673971812 +INFO:root: +[Valid][25][UTT][ MSE = 0.2342 | LCC = 0.8564 | SRCC = 0.8537 ] +INFO:root:Average loss=0.08444902439339039 +INFO:root: +[Valid][26][UTT][ MSE = 0.2264 | LCC = 0.8553 | SRCC = 0.8543 ] +INFO:root:Average loss=0.08287427846628886 +INFO:root: +[Valid][27][UTT][ MSE = 0.2504 | LCC = 0.8528 | SRCC = 0.8515 ] +INFO:root:Average loss=0.07855635055173667 +INFO:root: +[Valid][28][UTT][ MSE = 0.2769 | LCC = 0.8490 | SRCC = 0.8506 ] +INFO:root:Average loss=0.08311196063191463 +INFO:root: +[Valid][29][UTT][ MSE = 0.2421 | LCC = 0.8515 | SRCC = 0.8525 ] +INFO:root:Average loss=0.08378847062778778 +INFO:root: +[Valid][30][UTT][ MSE = 0.2325 | LCC = 0.8560 | SRCC = 0.8551 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2483 | LCC = 0.8506 | SRCC = 0.8523 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..91e1e6823ed61d10967c529275c6eb03c2b603a7 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:96fe685b8345299e46ef6566939745402edb6e3b1acb5be6b9c7ee1ecab7a251 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..3e9d0f0361ec50aebc043a36b5159f0ef56f1f90 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:17b4b3ca9b9568a97a6f9b63cf9c1df0f54d676c40c4cfddfb9ffa8543c63c5f +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..83108bf34c746da2dacfe82493f82b869e083898 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/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=0.6326022239831778 +INFO:root: +[Valid][0][UTT][ MSE = 0.4050 | LCC = 0.7455 | SRCC = 0.7489 ] +INFO:root:Average loss=0.3040032392511001 +INFO:root: +[Valid][1][UTT][ MSE = 0.3394 | LCC = 0.8059 | SRCC = 0.8066 ] +INFO:root:Average loss=0.2585008306763111 +INFO:root: +[Valid][2][UTT][ MSE = 0.2933 | LCC = 0.8169 | SRCC = 0.8176 ] +INFO:root:Average loss=0.24072994826695857 +INFO:root: +[Valid][3][UTT][ MSE = 0.2925 | LCC = 0.8295 | SRCC = 0.8308 ] +INFO:root:Average loss=0.21928981787119156 +INFO:root: +[Valid][4][UTT][ MSE = 0.2750 | LCC = 0.8333 | SRCC = 0.8347 ] +INFO:root:Average loss=0.20601657978617227 +INFO:root: +[Valid][5][UTT][ MSE = 0.2578 | LCC = 0.8384 | SRCC = 0.8383 ] +INFO:root:Average loss=0.2082905969940699 +INFO:root: +[Valid][6][UTT][ MSE = 0.2587 | LCC = 0.8401 | SRCC = 0.8382 ] +INFO:root:Average loss=0.18655373987097007 +INFO:root: +[Valid][7][UTT][ MSE = 0.2610 | LCC = 0.8398 | SRCC = 0.8415 ] +INFO:root:Average loss=0.17351614387753683 +INFO:root: +[Valid][8][UTT][ MSE = 0.2714 | LCC = 0.8372 | SRCC = 0.8400 ] +INFO:root:Average loss=0.1677280536446816 +INFO:root: +[Valid][9][UTT][ MSE = 0.2563 | LCC = 0.8428 | SRCC = 0.8436 ] +INFO:root:Average loss=0.16234781067722884 +INFO:root: +[Valid][10][UTT][ MSE = 0.2689 | LCC = 0.8484 | SRCC = 0.8472 ] +INFO:root:Average loss=0.15420835417432663 +INFO:root: +[Valid][11][UTT][ MSE = 0.2548 | LCC = 0.8476 | SRCC = 0.8463 ] +INFO:root:Average loss=0.14618327210728937 +INFO:root: +[Valid][12][UTT][ MSE = 0.2601 | LCC = 0.8475 | SRCC = 0.8503 ] +INFO:root:Average loss=0.13727191081031775 +INFO:root: +[Valid][13][UTT][ MSE = 0.2574 | LCC = 0.8479 | SRCC = 0.8479 ] +INFO:root:Average loss=0.131830619313778 +INFO:root: +[Valid][14][UTT][ MSE = 0.2472 | LCC = 0.8509 | SRCC = 0.8505 ] +INFO:root:Average loss=0.12845951977830666 +INFO:root: +[Valid][15][UTT][ MSE = 0.2590 | LCC = 0.8441 | SRCC = 0.8466 ] +INFO:root:Average loss=0.11967451827457318 +INFO:root: +[Valid][16][UTT][ MSE = 0.2348 | LCC = 0.8546 | SRCC = 0.8533 ] +INFO:root:Average loss=0.11335323359339665 +INFO:root: +[Valid][17][UTT][ MSE = 0.2511 | LCC = 0.8494 | SRCC = 0.8504 ] +INFO:root:Average loss=0.12114587225593053 +INFO:root: +[Valid][18][UTT][ MSE = 0.2358 | LCC = 0.8502 | SRCC = 0.8501 ] +INFO:root:Average loss=0.11147226488743073 +INFO:root: +[Valid][19][UTT][ MSE = 0.2495 | LCC = 0.8549 | SRCC = 0.8552 ] +INFO:root:Average loss=0.10108528362634854 +INFO:root: +[Valid][20][UTT][ MSE = 0.2478 | LCC = 0.8516 | SRCC = 0.8512 ] +INFO:root:Average loss=0.10079269789350338 +INFO:root: +[Valid][21][UTT][ MSE = 0.2337 | LCC = 0.8528 | SRCC = 0.8516 ] +INFO:root:Average loss=0.09792775221360035 +INFO:root: +[Valid][22][UTT][ MSE = 0.2297 | LCC = 0.8526 | SRCC = 0.8518 ] +INFO:root:Average loss=0.0903405984147237 +INFO:root: +[Valid][23][UTT][ MSE = 0.2321 | LCC = 0.8508 | SRCC = 0.8520 ] +INFO:root:Average loss=0.08602060549534284 +INFO:root: +[Valid][24][UTT][ MSE = 0.2288 | LCC = 0.8546 | SRCC = 0.8528 ] +INFO:root:Average loss=0.08962540658047566 +INFO:root: +[Valid][25][UTT][ MSE = 0.2356 | LCC = 0.8526 | SRCC = 0.8523 ] +INFO:root:Average loss=0.08327144136031468 +INFO:root: +[Valid][26][UTT][ MSE = 0.2416 | LCC = 0.8522 | SRCC = 0.8526 ] +INFO:root:Average loss=0.08359434403097019 +INFO:root: +[Valid][27][UTT][ MSE = 0.2457 | LCC = 0.8491 | SRCC = 0.8508 ] +INFO:root:Average loss=0.07995457641589336 +INFO:root: +[Valid][28][UTT][ MSE = 0.2355 | LCC = 0.8556 | SRCC = 0.8537 ] +INFO:root:Average loss=0.07752647976844738 +INFO:root: +[Valid][29][UTT][ MSE = 0.2327 | LCC = 0.8508 | SRCC = 0.8505 ] +INFO:root:Average loss=0.07854311378338398 +INFO:root: +[Valid][30][UTT][ MSE = 0.2429 | LCC = 0.8540 | SRCC = 0.8532 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2454 | LCC = 0.8489 | SRCC = 0.8476 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0a4d3a64aeaf904978c89ef4233e77120379e340 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6d9a5169d82c80c26af533234e8da5c9c830dd50a05652e55684388be52dee31 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..5ca92c840f1c7039aa140801a5f368709a5bd5fd --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0c6f438641286eff847c24448aaa7c7a3d9b7bcf70ffafab1476aa5dd169842e +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer0/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f04f6ebec1f3c90b3016d8903a54e0bed695f7bd --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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=0.4807775064538687 +INFO:root: +[Valid][0][UTT][ MSE = 0.3542 | LCC = 0.7704 | SRCC = 0.7686 ] +INFO:root:Average loss=0.2910503034408276 +INFO:root: +[Valid][1][UTT][ MSE = 0.3118 | LCC = 0.8056 | SRCC = 0.8064 ] +INFO:root:Average loss=0.26803146188075727 +INFO:root: +[Valid][2][UTT][ MSE = 0.2940 | LCC = 0.8175 | SRCC = 0.8177 ] +INFO:root:Average loss=0.23489070187012354 +INFO:root: +[Valid][3][UTT][ MSE = 0.3058 | LCC = 0.8273 | SRCC = 0.8270 ] +INFO:root:Average loss=0.22185274987266615 +INFO:root: +[Valid][4][UTT][ MSE = 0.2692 | LCC = 0.8322 | SRCC = 0.8321 ] +INFO:root:Average loss=0.2107711399021821 +INFO:root: +[Valid][5][UTT][ MSE = 0.2687 | LCC = 0.8359 | SRCC = 0.8340 ] +INFO:root:Average loss=0.19336327795798963 +INFO:root: +[Valid][6][UTT][ MSE = 0.2794 | LCC = 0.8413 | SRCC = 0.8402 ] +INFO:root:Average loss=0.19014353133164918 +INFO:root: +[Valid][7][UTT][ MSE = 0.2670 | LCC = 0.8395 | SRCC = 0.8349 ] +INFO:root:Average loss=0.1754734483666909 +INFO:root: +[Valid][8][UTT][ MSE = 0.2648 | LCC = 0.8438 | SRCC = 0.8434 ] +INFO:root:Average loss=0.16573844573054558 +INFO:root: +[Valid][9][UTT][ MSE = 0.2576 | LCC = 0.8431 | SRCC = 0.8441 ] +INFO:root:Average loss=0.1579093674245553 +INFO:root: +[Valid][10][UTT][ MSE = 0.2439 | LCC = 0.8457 | SRCC = 0.8464 ] +INFO:root:Average loss=0.14688257863506293 +INFO:root: +[Valid][11][UTT][ MSE = 0.2509 | LCC = 0.8513 | SRCC = 0.8501 ] +INFO:root:Average loss=0.14789746071283633 +INFO:root: +[Valid][12][UTT][ MSE = 0.2412 | LCC = 0.8491 | SRCC = 0.8473 ] +INFO:root:Average loss=0.13511872807374367 +INFO:root: +[Valid][13][UTT][ MSE = 0.2380 | LCC = 0.8476 | SRCC = 0.8496 ] +INFO:root:Average loss=0.12961522728586808 +INFO:root: +[Valid][14][UTT][ MSE = 0.2742 | LCC = 0.8383 | SRCC = 0.8446 ] +INFO:root:Average loss=0.129057460774978 +INFO:root: +[Valid][15][UTT][ MSE = 0.2381 | LCC = 0.8528 | SRCC = 0.8532 ] +INFO:root:Average loss=0.12040965335491376 +INFO:root: +[Valid][16][UTT][ MSE = 0.2307 | LCC = 0.8524 | SRCC = 0.8541 ] +INFO:root:Average loss=0.13171847355671418 +INFO:root: +[Valid][17][UTT][ MSE = 0.2457 | LCC = 0.8459 | SRCC = 0.8479 ] +INFO:root:Average loss=0.11267534925196415 +INFO:root: +[Valid][18][UTT][ MSE = 0.2446 | LCC = 0.8468 | SRCC = 0.8494 ] +INFO:root:Average loss=0.11891437837710747 +INFO:root: +[Valid][19][UTT][ MSE = 0.2368 | LCC = 0.8487 | SRCC = 0.8509 ] +INFO:root:Average loss=0.10405926263103119 +INFO:root: +[Valid][20][UTT][ MSE = 0.2335 | LCC = 0.8574 | SRCC = 0.8569 ] +INFO:root:Average loss=0.09940782189369202 +INFO:root: +[Valid][21][UTT][ MSE = 0.2425 | LCC = 0.8527 | SRCC = 0.8529 ] +INFO:root:Average loss=0.10201344648614907 +INFO:root: +[Valid][22][UTT][ MSE = 0.2288 | LCC = 0.8590 | SRCC = 0.8579 ] +INFO:root:Average loss=0.11110573620177232 +INFO:root: +[Valid][23][UTT][ MSE = 0.2327 | LCC = 0.8585 | SRCC = 0.8577 ] +INFO:root:Average loss=0.0966342680920393 +INFO:root: +[Valid][24][UTT][ MSE = 0.2314 | LCC = 0.8548 | SRCC = 0.8546 ] +INFO:root:Average loss=0.08376982430808055 +INFO:root: +[Valid][25][UTT][ MSE = 0.2423 | LCC = 0.8504 | SRCC = 0.8532 ] +INFO:root:Average loss=0.09791710915473792 +INFO:root: +[Valid][26][UTT][ MSE = 0.2438 | LCC = 0.8585 | SRCC = 0.8571 ] +INFO:root:Average loss=0.08460496466320294 +INFO:root: +[Valid][27][UTT][ MSE = 0.2366 | LCC = 0.8575 | SRCC = 0.8564 ] +INFO:root:Average loss=0.07886409234160031 +INFO:root: +[Valid][28][UTT][ MSE = 0.2266 | LCC = 0.8551 | SRCC = 0.8561 ] +INFO:root:Average loss=0.08231285276512305 +INFO:root: +[Valid][29][UTT][ MSE = 0.2259 | LCC = 0.8569 | SRCC = 0.8563 ] +INFO:root:Average loss=0.07654100016523631 +INFO:root: +[Valid][30][UTT][ MSE = 0.2230 | LCC = 0.8574 | SRCC = 0.8577 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2452 | LCC = 0.8528 | SRCC = 0.8524 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..db95233158f4021754c974abb0054c0bb527c13d --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c84efbca7fe8fb967aac47cfa90da14af8e32d0e5ad37262eb905d89fc272b54 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c431ecae75924e9117a265118d3ad22ac9746b4b --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e7e55fbd03f32c613d19147ec39c1a493b560a1e877385687c5a8862299b476 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1cc1a5b4afe8a54989da0f5e2d0615b57c7d9563 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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=0.47339991747568816 +INFO:root: +[Valid][0][UTT][ MSE = 0.3395 | LCC = 0.7831 | SRCC = 0.7890 ] +INFO:root:Average loss=0.2660357453502141 +INFO:root: +[Valid][1][UTT][ MSE = 0.2942 | LCC = 0.8271 | SRCC = 0.8312 ] +INFO:root:Average loss=0.2346986553225762 +INFO:root: +[Valid][2][UTT][ MSE = 0.3371 | LCC = 0.8379 | SRCC = 0.8468 ] +INFO:root:Average loss=0.20961687790277678 +INFO:root: +[Valid][3][UTT][ MSE = 0.2790 | LCC = 0.8542 | SRCC = 0.8530 ] +INFO:root:Average loss=0.19631312157099062 +INFO:root: +[Valid][4][UTT][ MSE = 0.2379 | LCC = 0.8583 | SRCC = 0.8573 ] +INFO:root:Average loss=0.1770884851232553 +INFO:root: +[Valid][5][UTT][ MSE = 0.2298 | LCC = 0.8656 | SRCC = 0.8641 ] +INFO:root:Average loss=0.1642588067513246 +INFO:root: +[Valid][6][UTT][ MSE = 0.2346 | LCC = 0.8623 | SRCC = 0.8601 ] +INFO:root:Average loss=0.15856572918784925 +INFO:root: +[Valid][7][UTT][ MSE = 0.2455 | LCC = 0.8587 | SRCC = 0.8597 ] +INFO:root:Average loss=0.14522142259356302 +INFO:root: +[Valid][8][UTT][ MSE = 0.2315 | LCC = 0.8656 | SRCC = 0.8647 ] +INFO:root:Average loss=0.14037065274822405 +INFO:root: +[Valid][9][UTT][ MSE = 0.2609 | LCC = 0.8623 | SRCC = 0.8623 ] +INFO:root:Average loss=0.13070840722857377 +INFO:root: +[Valid][10][UTT][ MSE = 0.2640 | LCC = 0.8641 | SRCC = 0.8650 ] +INFO:root:Average loss=0.12016755485763916 +INFO:root: +[Valid][11][UTT][ MSE = 0.2198 | LCC = 0.8664 | SRCC = 0.8656 ] +INFO:root:Average loss=0.11315561897861652 +INFO:root: +[Valid][12][UTT][ MSE = 0.2870 | LCC = 0.8601 | SRCC = 0.8637 ] +INFO:root:Average loss=0.11433318214347729 +INFO:root: +[Valid][13][UTT][ MSE = 0.2164 | LCC = 0.8661 | SRCC = 0.8634 ] +INFO:root:Average loss=0.10701431295810601 +INFO:root: +[Valid][14][UTT][ MSE = 0.2241 | LCC = 0.8644 | SRCC = 0.8624 ] +INFO:root:Average loss=0.09860481503300178 +INFO:root: +[Valid][15][UTT][ MSE = 0.2148 | LCC = 0.8676 | SRCC = 0.8654 ] +INFO:root:Average loss=0.09749089071574883 +INFO:root: +[Valid][16][UTT][ MSE = 0.2220 | LCC = 0.8672 | SRCC = 0.8659 ] +INFO:root:Average loss=0.09555440166821846 +INFO:root: +[Valid][17][UTT][ MSE = 0.2168 | LCC = 0.8677 | SRCC = 0.8662 ] +INFO:root:Average loss=0.08701210430799386 +INFO:root: +[Valid][18][UTT][ MSE = 0.2206 | LCC = 0.8672 | SRCC = 0.8667 ] +INFO:root:Average loss=0.08422763067751358 +INFO:root: +[Valid][19][UTT][ MSE = 0.2437 | LCC = 0.8675 | SRCC = 0.8665 ] +INFO:root:Average loss=0.0792574181388586 +INFO:root: +[Valid][20][UTT][ MSE = 0.2655 | LCC = 0.8641 | SRCC = 0.8660 ] +INFO:root:Average loss=0.07600637477559921 +INFO:root: +[Valid][21][UTT][ MSE = 0.2362 | LCC = 0.8685 | SRCC = 0.8683 ] +INFO:root:Average loss=0.07402763983760124 +INFO:root: +[Valid][22][UTT][ MSE = 0.2323 | LCC = 0.8641 | SRCC = 0.8643 ] +INFO:root:Average loss=0.07229215761598869 +INFO:root: +[Valid][23][UTT][ MSE = 0.2166 | LCC = 0.8683 | SRCC = 0.8662 ] +INFO:root:Average loss=0.07403236073561204 +INFO:root: +[Valid][24][UTT][ MSE = 0.2267 | LCC = 0.8686 | SRCC = 0.8671 ] +INFO:root:Average loss=0.07580415713481414 +INFO:root: +[Valid][25][UTT][ MSE = 0.2430 | LCC = 0.8703 | SRCC = 0.8698 ] +INFO:root:Average loss=0.07028491088213065 +INFO:root: +[Valid][26][UTT][ MSE = 0.2431 | LCC = 0.8701 | SRCC = 0.8696 ] +INFO:root:Average loss=0.06327947415411472 +INFO:root: +[Valid][27][UTT][ MSE = 0.2235 | LCC = 0.8675 | SRCC = 0.8658 ] +INFO:root:Average loss=0.061367071544130646 +INFO:root: +[Valid][28][UTT][ MSE = 0.2145 | LCC = 0.8695 | SRCC = 0.8664 ] +INFO:root:Average loss=0.06509465576173404 +INFO:root: +[Valid][29][UTT][ MSE = 0.2485 | LCC = 0.8665 | SRCC = 0.8641 ] +INFO:root:Average loss=0.057367341354107246 +INFO:root: +[Valid][30][UTT][ MSE = 0.2489 | LCC = 0.8664 | SRCC = 0.8633 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2318 | LCC = 0.8620 | SRCC = 0.8595 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c1747030cad83c905621d964f3e91c0cf0e6f4ed --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e5f1c1313fab7e04c608747008fb865b3a60ad54fd9679fae9b18e437d6ce48e +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7beaa9ea31fa4e06c86ff3d5f4cd8e7ae40ef8c2 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0d86e9d0f17a421711b8784169aa1e7c4ab1567b9ab366eda5aaf3ec78468921 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..870f74b95bf4b6ed7b7df575b687204c688975df --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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=0.4877508712502626 +INFO:root: +[Valid][0][UTT][ MSE = 0.3714 | LCC = 0.7867 | SRCC = 0.7918 ] +INFO:root:Average loss=0.2688652578836832 +INFO:root: +[Valid][1][UTT][ MSE = 0.2925 | LCC = 0.8243 | SRCC = 0.8307 ] +INFO:root:Average loss=0.22399364832120064 +INFO:root: +[Valid][2][UTT][ MSE = 0.2671 | LCC = 0.8410 | SRCC = 0.8414 ] +INFO:root:Average loss=0.2103912326005789 +INFO:root: +[Valid][3][UTT][ MSE = 0.2762 | LCC = 0.8499 | SRCC = 0.8525 ] +INFO:root:Average loss=0.2010067825516065 +INFO:root: +[Valid][4][UTT][ MSE = 0.2629 | LCC = 0.8550 | SRCC = 0.8574 ] +INFO:root:Average loss=0.1779301976546263 +INFO:root: +[Valid][5][UTT][ MSE = 0.2394 | LCC = 0.8602 | SRCC = 0.8595 ] +INFO:root:Average loss=0.16274713763059714 +INFO:root: +[Valid][6][UTT][ MSE = 0.2262 | LCC = 0.8643 | SRCC = 0.8646 ] +INFO:root:Average loss=0.15953125900182968 +INFO:root: +[Valid][7][UTT][ MSE = 0.2319 | LCC = 0.8608 | SRCC = 0.8606 ] +INFO:root:Average loss=0.1455148357229355 +INFO:root: +[Valid][8][UTT][ MSE = 0.2404 | LCC = 0.8594 | SRCC = 0.8583 ] +INFO:root:Average loss=0.13526040172347656 +INFO:root: +[Valid][9][UTT][ MSE = 0.2387 | LCC = 0.8648 | SRCC = 0.8660 ] +INFO:root:Average loss=0.1330603342025708 +INFO:root: +[Valid][10][UTT][ MSE = 0.2245 | LCC = 0.8687 | SRCC = 0.8677 ] +INFO:root:Average loss=0.12492259868826622 +INFO:root: +[Valid][11][UTT][ MSE = 0.2177 | LCC = 0.8686 | SRCC = 0.8672 ] +INFO:root:Average loss=0.12084566285976997 +INFO:root: +[Valid][12][UTT][ MSE = 0.2263 | LCC = 0.8657 | SRCC = 0.8653 ] +INFO:root:Average loss=0.11387013643980026 +INFO:root: +[Valid][13][UTT][ MSE = 0.2220 | LCC = 0.8662 | SRCC = 0.8653 ] +INFO:root:Average loss=0.10679911779096493 +INFO:root: +[Valid][14][UTT][ MSE = 0.2076 | LCC = 0.8682 | SRCC = 0.8680 ] +INFO:root:Average loss=0.10548305291778003 +INFO:root: +[Valid][15][UTT][ MSE = 0.2269 | LCC = 0.8688 | SRCC = 0.8671 ] +INFO:root:Average loss=0.09277055035225856 +INFO:root: +[Valid][16][UTT][ MSE = 0.2138 | LCC = 0.8698 | SRCC = 0.8672 ] +INFO:root:Average loss=0.09021762944757938 +INFO:root: +[Valid][17][UTT][ MSE = 0.2145 | LCC = 0.8705 | SRCC = 0.8686 ] +INFO:root:Average loss=0.0879230834543705 +INFO:root: +[Valid][18][UTT][ MSE = 0.2168 | LCC = 0.8692 | SRCC = 0.8671 ] +INFO:root:Average loss=0.08154703051042862 +INFO:root: +[Valid][19][UTT][ MSE = 0.2554 | LCC = 0.8608 | SRCC = 0.8613 ] +INFO:root:Average loss=0.08526290747790764 +INFO:root: +[Valid][20][UTT][ MSE = 0.2302 | LCC = 0.8629 | SRCC = 0.8615 ] +INFO:root:Average loss=0.07860288778558755 +INFO:root: +[Valid][21][UTT][ MSE = 0.2279 | LCC = 0.8676 | SRCC = 0.8672 ] +INFO:root:Average loss=0.07305197331767815 +INFO:root: +[Valid][22][UTT][ MSE = 0.2161 | LCC = 0.8636 | SRCC = 0.8620 ] +INFO:root:Average loss=0.07591358452844314 +INFO:root: +[Valid][23][UTT][ MSE = 0.2366 | LCC = 0.8645 | SRCC = 0.8630 ] +INFO:root:Average loss=0.06911553996495712 +INFO:root: +[Valid][24][UTT][ MSE = 0.2078 | LCC = 0.8696 | SRCC = 0.8661 ] +INFO:root:Average loss=0.0689961300828518 +INFO:root: +[Valid][25][UTT][ MSE = 0.2143 | LCC = 0.8695 | SRCC = 0.8662 ] +INFO:root:Average loss=0.06746718884469607 +INFO:root: +[Valid][26][UTT][ MSE = 0.2100 | LCC = 0.8689 | SRCC = 0.8665 ] +INFO:root:Average loss=0.062315337073344454 +INFO:root: +[Valid][27][UTT][ MSE = 0.2129 | LCC = 0.8665 | SRCC = 0.8668 ] +INFO:root:Average loss=0.06087251609334579 +INFO:root: +[Valid][28][UTT][ MSE = 0.2304 | LCC = 0.8647 | SRCC = 0.8630 ] +INFO:root:Average loss=0.05713790865280689 +INFO:root: +[Valid][29][UTT][ MSE = 0.2180 | LCC = 0.8654 | SRCC = 0.8654 ] +INFO:root:Average loss=0.05885102203450142 +INFO:root: +[Valid][30][UTT][ MSE = 0.2273 | LCC = 0.8683 | SRCC = 0.8663 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2543 | LCC = 0.8559 | SRCC = 0.8563 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d69792436a6032706a0f7b8e2d52d846d58c3905 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:73d2cd4b41011cbdfed422ec8a8c93006df924857a0e3d3dc4bf59475e2e8dae +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0ece235fa1b10a983871777b0652372015f70974 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5b78d3c014a96ab75eb2271f8d5d150c5ed8a3024fb0e7f40057800e22bf2378 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a23697e46a2a7e95e39a68794536ca88429dde01 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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=0.5146210953975335 +INFO:root: +[Valid][0][UTT][ MSE = 0.3553 | LCC = 0.7741 | SRCC = 0.7867 ] +INFO:root:Average loss=0.2620560536399866 +INFO:root: +[Valid][1][UTT][ MSE = 0.2833 | LCC = 0.8303 | SRCC = 0.8322 ] +INFO:root:Average loss=0.22576805509817907 +INFO:root: +[Valid][2][UTT][ MSE = 0.2936 | LCC = 0.8453 | SRCC = 0.8470 ] +INFO:root:Average loss=0.2012460398941468 +INFO:root: +[Valid][3][UTT][ MSE = 0.2441 | LCC = 0.8549 | SRCC = 0.8544 ] +INFO:root:Average loss=0.19280230979888868 +INFO:root: +[Valid][4][UTT][ MSE = 0.2603 | LCC = 0.8521 | SRCC = 0.8537 ] +INFO:root:Average loss=0.18910527191100976 +INFO:root: +[Valid][5][UTT][ MSE = 0.2756 | LCC = 0.8492 | SRCC = 0.8525 ] +INFO:root:Average loss=0.1740478199835007 +INFO:root: +[Valid][6][UTT][ MSE = 0.2419 | LCC = 0.8598 | SRCC = 0.8607 ] +INFO:root:Average loss=0.1603351627022792 +INFO:root: +[Valid][7][UTT][ MSE = 0.2305 | LCC = 0.8618 | SRCC = 0.8620 ] +INFO:root:Average loss=0.14666154961555433 +INFO:root: +[Valid][8][UTT][ MSE = 0.2283 | LCC = 0.8628 | SRCC = 0.8624 ] +INFO:root:Average loss=0.13640004568375075 +INFO:root: +[Valid][9][UTT][ MSE = 0.2190 | LCC = 0.8643 | SRCC = 0.8658 ] +INFO:root:Average loss=0.1264215300862606 +INFO:root: +[Valid][10][UTT][ MSE = 0.2558 | LCC = 0.8651 | SRCC = 0.8658 ] +INFO:root:Average loss=0.12649285019590303 +INFO:root: +[Valid][11][UTT][ MSE = 0.2334 | LCC = 0.8640 | SRCC = 0.8635 ] +INFO:root:Average loss=0.12729215755676612 +INFO:root: +[Valid][12][UTT][ MSE = 0.2241 | LCC = 0.8640 | SRCC = 0.8640 ] +INFO:root:Average loss=0.11027339349190395 +INFO:root: +[Valid][13][UTT][ MSE = 0.2191 | LCC = 0.8637 | SRCC = 0.8631 ] +INFO:root:Average loss=0.10470017112600498 +INFO:root: +[Valid][14][UTT][ MSE = 0.2183 | LCC = 0.8679 | SRCC = 0.8678 ] +INFO:root:Average loss=0.09851260339984527 +INFO:root: +[Valid][15][UTT][ MSE = 0.2181 | LCC = 0.8661 | SRCC = 0.8652 ] +INFO:root:Average loss=0.09618414021455325 +INFO:root: +[Valid][16][UTT][ MSE = 0.2160 | LCC = 0.8657 | SRCC = 0.8652 ] +INFO:root:Average loss=0.09128400874443543 +INFO:root: +[Valid][17][UTT][ MSE = 0.2383 | LCC = 0.8658 | SRCC = 0.8642 ] +INFO:root:Average loss=0.09202707114701088 +INFO:root: +[Valid][18][UTT][ MSE = 0.2151 | LCC = 0.8667 | SRCC = 0.8661 ] +INFO:root:Average loss=0.0866872913753375 +INFO:root: +[Valid][19][UTT][ MSE = 0.2230 | LCC = 0.8629 | SRCC = 0.8618 ] +INFO:root:Average loss=0.07900473064719102 +INFO:root: +[Valid][20][UTT][ MSE = 0.2280 | LCC = 0.8643 | SRCC = 0.8628 ] +INFO:root:Average loss=0.07915956756243339 +INFO:root: +[Valid][21][UTT][ MSE = 0.2170 | LCC = 0.8631 | SRCC = 0.8606 ] +INFO:root:Average loss=0.07335187666691266 +INFO:root: +[Valid][22][UTT][ MSE = 0.2130 | LCC = 0.8650 | SRCC = 0.8648 ] +INFO:root:Average loss=0.06920367742005067 +INFO:root: +[Valid][23][UTT][ MSE = 0.2428 | LCC = 0.8566 | SRCC = 0.8568 ] +INFO:root:Average loss=0.06791466221404381 +INFO:root: +[Valid][24][UTT][ MSE = 0.2251 | LCC = 0.8607 | SRCC = 0.8594 ] +INFO:root:Average loss=0.0649139130822359 +INFO:root: +[Valid][25][UTT][ MSE = 0.2526 | LCC = 0.8559 | SRCC = 0.8570 ] +INFO:root:Average loss=0.06890325386745808 +INFO:root: +[Valid][26][UTT][ MSE = 0.2204 | LCC = 0.8653 | SRCC = 0.8643 ] +INFO:root:Average loss=0.06430077920548427 +INFO:root: +[Valid][27][UTT][ MSE = 0.2253 | LCC = 0.8643 | SRCC = 0.8634 ] +INFO:root:Average loss=0.05794162995731219 +INFO:root: +[Valid][28][UTT][ MSE = 0.2333 | LCC = 0.8612 | SRCC = 0.8613 ] +INFO:root:Average loss=0.05609353450246346 +INFO:root: +[Valid][29][UTT][ MSE = 0.2123 | LCC = 0.8650 | SRCC = 0.8637 ] +INFO:root:Average loss=0.05754668781390557 +INFO:root: +[Valid][30][UTT][ MSE = 0.2162 | LCC = 0.8653 | SRCC = 0.8652 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2507 | LCC = 0.8640 | SRCC = 0.8645 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..25048b4c36dcf0e366d3d8a6b6e2147e1b13a972 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:938625aae251623433d96b42506ab25a4c566a855093f13b75aa6ecb743e5508 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..8c59f6573e9b6669a53f3232aee043e5e04e4003 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a35791b2162342b55129735e5586b5d4ed1b65c1fed8761d77aebbd16d4ccc9d +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9ee4bebd5e40a6a213589c001b71206500e7d695 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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=0.49896956712771684 +INFO:root: +[Valid][0][UTT][ MSE = 0.3656 | LCC = 0.7782 | SRCC = 0.7899 ] +INFO:root:Average loss=0.260722741102561 +INFO:root: +[Valid][1][UTT][ MSE = 0.2664 | LCC = 0.8326 | SRCC = 0.8379 ] +INFO:root:Average loss=0.2165628493978427 +INFO:root: +[Valid][2][UTT][ MSE = 0.2503 | LCC = 0.8467 | SRCC = 0.8491 ] +INFO:root:Average loss=0.1954620264661618 +INFO:root: +[Valid][3][UTT][ MSE = 0.2441 | LCC = 0.8549 | SRCC = 0.8581 ] +INFO:root:Average loss=0.17865248205952156 +INFO:root: +[Valid][4][UTT][ MSE = 0.2448 | LCC = 0.8588 | SRCC = 0.8596 ] +INFO:root:Average loss=0.17316942633344576 +INFO:root: +[Valid][5][UTT][ MSE = 0.2312 | LCC = 0.8625 | SRCC = 0.8641 ] +INFO:root:Average loss=0.15597939644104394 +INFO:root: +[Valid][6][UTT][ MSE = 0.2225 | LCC = 0.8637 | SRCC = 0.8636 ] +INFO:root:Average loss=0.1454695548193577 +INFO:root: +[Valid][7][UTT][ MSE = 0.2206 | LCC = 0.8663 | SRCC = 0.8672 ] +INFO:root:Average loss=0.14175692286628944 +INFO:root: +[Valid][8][UTT][ MSE = 0.2176 | LCC = 0.8701 | SRCC = 0.8696 ] +INFO:root:Average loss=0.1262234257390866 +INFO:root: +[Valid][9][UTT][ MSE = 0.2218 | LCC = 0.8685 | SRCC = 0.8680 ] +INFO:root:Average loss=0.1266207600442263 +INFO:root: +[Valid][10][UTT][ MSE = 0.2365 | LCC = 0.8682 | SRCC = 0.8670 ] +INFO:root:Average loss=0.11432418461220387 +INFO:root: +[Valid][11][UTT][ MSE = 0.2171 | LCC = 0.8644 | SRCC = 0.8647 ] +INFO:root:Average loss=0.10649947344492643 +INFO:root: +[Valid][12][UTT][ MSE = 0.2134 | LCC = 0.8674 | SRCC = 0.8662 ] +INFO:root:Average loss=0.09611140387371564 +INFO:root: +[Valid][13][UTT][ MSE = 0.2121 | LCC = 0.8667 | SRCC = 0.8669 ] +INFO:root:Average loss=0.09753179058241539 +INFO:root: +[Valid][14][UTT][ MSE = 0.2362 | LCC = 0.8677 | SRCC = 0.8669 ] +INFO:root:Average loss=0.0960755420800967 +INFO:root: +[Valid][15][UTT][ MSE = 0.2314 | LCC = 0.8678 | SRCC = 0.8676 ] +INFO:root:Average loss=0.08434346566597621 +INFO:root: +[Valid][16][UTT][ MSE = 0.2200 | LCC = 0.8665 | SRCC = 0.8652 ] +INFO:root:Average loss=0.08117021123568217 +INFO:root: +[Valid][17][UTT][ MSE = 0.2152 | LCC = 0.8693 | SRCC = 0.8678 ] +INFO:root:Average loss=0.07887874792019527 +INFO:root: +[Valid][18][UTT][ MSE = 0.2174 | LCC = 0.8662 | SRCC = 0.8642 ] +INFO:root:Average loss=0.07721581811515185 +INFO:root: +[Valid][19][UTT][ MSE = 0.2115 | LCC = 0.8648 | SRCC = 0.8643 ] +INFO:root:Average loss=0.07451632222495018 +INFO:root: +[Valid][20][UTT][ MSE = 0.2163 | LCC = 0.8651 | SRCC = 0.8634 ] +INFO:root:Average loss=0.06870564254812705 +INFO:root: +[Valid][21][UTT][ MSE = 0.2219 | LCC = 0.8649 | SRCC = 0.8631 ] +INFO:root:Average loss=0.06506461917589872 +INFO:root: +[Valid][22][UTT][ MSE = 0.2267 | LCC = 0.8632 | SRCC = 0.8617 ] +INFO:root:Average loss=0.06348940935463478 +INFO:root: +[Valid][23][UTT][ MSE = 0.2123 | LCC = 0.8669 | SRCC = 0.8652 ] +INFO:root:Average loss=0.06016293884469913 +INFO:root: +[Valid][24][UTT][ MSE = 0.2200 | LCC = 0.8672 | SRCC = 0.8656 ] +INFO:root:Average loss=0.05723005280089684 +INFO:root: +[Valid][25][UTT][ MSE = 0.2148 | LCC = 0.8675 | SRCC = 0.8666 ] +INFO:root:Average loss=0.05811821406659407 +INFO:root: +[Valid][26][UTT][ MSE = 0.2212 | LCC = 0.8680 | SRCC = 0.8673 ] +INFO:root:Average loss=0.054717885616880194 +INFO:root: +[Valid][27][UTT][ MSE = 0.2174 | LCC = 0.8666 | SRCC = 0.8656 ] +INFO:root:Average loss=0.06072963769428241 +INFO:root: +[Valid][28][UTT][ MSE = 0.2214 | LCC = 0.8692 | SRCC = 0.8674 ] +INFO:root:Average loss=0.05013536384854561 +INFO:root: +[Valid][29][UTT][ MSE = 0.2202 | LCC = 0.8637 | SRCC = 0.8627 ] +INFO:root:Average loss=0.04911236713329951 +INFO:root: +[Valid][30][UTT][ MSE = 0.2080 | LCC = 0.8694 | SRCC = 0.8687 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2521 | LCC = 0.8571 | SRCC = 0.8589 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b09a0f5204b1b2c5ae22dbb00260ad3ad6c54caa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3e4c894e34440ec3381d93f3adb4ab91cd18f6088b1028f7c87925e47a834182 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d021361b17e61b169d898eea48db713b114dd6c3 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9fd636c67220d6823d1020987ed5426d8553299b03263d409d377990815c5576 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer1/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..895f6065b34baf4679828b057f95ec3ba3170afb --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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=0.7882658904179548 +INFO:root: +[Valid][0][UTT][ MSE = 0.5161 | LCC = 0.6857 | SRCC = 0.6041 ] +INFO:root:Average loss=0.32763589651156694 +INFO:root: +[Valid][1][UTT][ MSE = 0.3088 | LCC = 0.8032 | SRCC = 0.8136 ] +INFO:root:Average loss=0.2391107548505832 +INFO:root: +[Valid][2][UTT][ MSE = 0.2813 | LCC = 0.8401 | SRCC = 0.8430 ] +INFO:root:Average loss=0.21215550563274285 +INFO:root: +[Valid][3][UTT][ MSE = 0.2482 | LCC = 0.8517 | SRCC = 0.8549 ] +INFO:root:Average loss=0.18965810193465307 +INFO:root: +[Valid][4][UTT][ MSE = 0.2729 | LCC = 0.8526 | SRCC = 0.8591 ] +INFO:root:Average loss=0.17972760647535324 +INFO:root: +[Valid][5][UTT][ MSE = 0.2329 | LCC = 0.8615 | SRCC = 0.8620 ] +INFO:root:Average loss=0.16228475679571813 +INFO:root: +[Valid][6][UTT][ MSE = 0.2535 | LCC = 0.8610 | SRCC = 0.8622 ] +INFO:root:Average loss=0.15546004875348166 +INFO:root: +[Valid][7][UTT][ MSE = 0.2277 | LCC = 0.8647 | SRCC = 0.8663 ] +INFO:root:Average loss=0.14072415242210412 +INFO:root: +[Valid][8][UTT][ MSE = 0.2267 | LCC = 0.8683 | SRCC = 0.8676 ] +INFO:root:Average loss=0.1375259866890235 +INFO:root: +[Valid][9][UTT][ MSE = 0.2151 | LCC = 0.8662 | SRCC = 0.8675 ] +INFO:root:Average loss=0.1310846674709748 +INFO:root: +[Valid][10][UTT][ MSE = 0.2129 | LCC = 0.8694 | SRCC = 0.8706 ] +INFO:root:Average loss=0.12169306543775094 +INFO:root: +[Valid][11][UTT][ MSE = 0.2299 | LCC = 0.8631 | SRCC = 0.8657 ] +INFO:root:Average loss=0.11664533261687328 +INFO:root: +[Valid][12][UTT][ MSE = 0.2588 | LCC = 0.8622 | SRCC = 0.8673 ] +INFO:root:Average loss=0.1239874643775133 +INFO:root: +[Valid][13][UTT][ MSE = 0.2258 | LCC = 0.8700 | SRCC = 0.8698 ] +INFO:root:Average loss=0.1039493550092746 +INFO:root: +[Valid][14][UTT][ MSE = 0.2155 | LCC = 0.8687 | SRCC = 0.8697 ] +INFO:root:Average loss=0.09668220068590763 +INFO:root: +[Valid][15][UTT][ MSE = 0.2044 | LCC = 0.8701 | SRCC = 0.8710 ] +INFO:root:Average loss=0.08994530709699178 +INFO:root: +[Valid][16][UTT][ MSE = 0.2226 | LCC = 0.8706 | SRCC = 0.8701 ] +INFO:root:Average loss=0.08947438393265773 +INFO:root: +[Valid][17][UTT][ MSE = 0.2108 | LCC = 0.8682 | SRCC = 0.8690 ] +INFO:root:Average loss=0.08547647058581695 +INFO:root: +[Valid][18][UTT][ MSE = 0.2120 | LCC = 0.8683 | SRCC = 0.8695 ] +INFO:root:Average loss=0.07987360718349616 +INFO:root: +[Valid][19][UTT][ MSE = 0.2158 | LCC = 0.8704 | SRCC = 0.8694 ] +INFO:root:Average loss=0.07886096792152295 +INFO:root: +[Valid][20][UTT][ MSE = 0.2398 | LCC = 0.8683 | SRCC = 0.8676 ] +INFO:root:Average loss=0.07840089027125102 +INFO:root: +[Valid][21][UTT][ MSE = 0.2162 | LCC = 0.8679 | SRCC = 0.8689 ] +INFO:root:Average loss=0.07248592529541407 +INFO:root: +[Valid][22][UTT][ MSE = 0.2142 | LCC = 0.8704 | SRCC = 0.8704 ] +INFO:root:Average loss=0.0699743276509719 +INFO:root: +[Valid][23][UTT][ MSE = 0.2158 | LCC = 0.8707 | SRCC = 0.8702 ] +INFO:root:Average loss=0.07237623894634919 +INFO:root: +[Valid][24][UTT][ MSE = 0.2131 | LCC = 0.8674 | SRCC = 0.8683 ] +INFO:root:Average loss=0.062000421090767935 +INFO:root: +[Valid][25][UTT][ MSE = 0.2159 | LCC = 0.8671 | SRCC = 0.8677 ] +INFO:root:Average loss=0.05922548525417463 +INFO:root: +[Valid][26][UTT][ MSE = 0.2129 | LCC = 0.8678 | SRCC = 0.8674 ] +INFO:root:Average loss=0.05878572471630879 +INFO:root: +[Valid][27][UTT][ MSE = 0.2103 | LCC = 0.8666 | SRCC = 0.8672 ] +INFO:root:Average loss=0.06050671961827156 +INFO:root: +[Valid][28][UTT][ MSE = 0.2198 | LCC = 0.8679 | SRCC = 0.8669 ] +INFO:root:Average loss=0.0593263296267161 +INFO:root: +[Valid][29][UTT][ MSE = 0.2134 | LCC = 0.8676 | SRCC = 0.8677 ] +INFO:root:Average loss=0.05373074625356075 +INFO:root: +[Valid][30][UTT][ MSE = 0.2189 | LCC = 0.8678 | SRCC = 0.8671 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2507 | LCC = 0.8604 | SRCC = 0.8597 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6da8346100619706e51b45276ccc10ab669060fc --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4c4dbfa29985ac6805031debd6b637ef195f1864bbba518a4b3a10cd78dce839 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..22fb005646afba91942ef9aa804475cb8ac75dae --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:12a1fd3ff201967683b15eefaf136aefd7e8cc181ff0bc22b11d7936deb3c416 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..591ee8ea8b6f68d4ed4ca8c4a61e18c2ed30e8ee --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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=0.3308795826175274 +INFO:root: +[Valid][0][UTT][ MSE = 0.3270 | LCC = 0.8314 | SRCC = 0.8351 ] +INFO:root:Average loss=0.25928314450459605 +INFO:root: +[Valid][1][UTT][ MSE = 0.3167 | LCC = 0.8424 | SRCC = 0.8399 ] +INFO:root:Average loss=0.24739255316746542 +INFO:root: +[Valid][2][UTT][ MSE = 0.3161 | LCC = 0.8501 | SRCC = 0.8484 ] +INFO:root:Average loss=0.23635679617142066 +INFO:root: +[Valid][3][UTT][ MSE = 0.2884 | LCC = 0.8536 | SRCC = 0.8528 ] +INFO:root:Average loss=0.22552493061774817 +INFO:root: +[Valid][4][UTT][ MSE = 0.2768 | LCC = 0.8539 | SRCC = 0.8538 ] +INFO:root:Average loss=0.22048801126388404 +INFO:root: +[Valid][5][UTT][ MSE = 0.2614 | LCC = 0.8587 | SRCC = 0.8583 ] +INFO:root:Average loss=0.22608128419289222 +INFO:root: +[Valid][6][UTT][ MSE = 0.2846 | LCC = 0.8556 | SRCC = 0.8589 ] +INFO:root:Average loss=0.21168289696558928 +INFO:root: +[Valid][7][UTT][ MSE = 0.2730 | LCC = 0.8611 | SRCC = 0.8601 ] +INFO:root:Average loss=0.21605420322754446 +INFO:root: +[Valid][8][UTT][ MSE = 0.2546 | LCC = 0.8610 | SRCC = 0.8601 ] +INFO:root:Average loss=0.20996736142879877 +INFO:root: +[Valid][9][UTT][ MSE = 0.2942 | LCC = 0.8558 | SRCC = 0.8560 ] +INFO:root:Average loss=0.20433499301091218 +INFO:root: +[Valid][10][UTT][ MSE = 0.2766 | LCC = 0.8605 | SRCC = 0.8599 ] +INFO:root:Average loss=0.20285835002477354 +INFO:root: +[Valid][11][UTT][ MSE = 0.2576 | LCC = 0.8551 | SRCC = 0.8585 ] +INFO:root:Average loss=0.19898967043711588 +INFO:root: +[Valid][12][UTT][ MSE = 0.2576 | LCC = 0.8582 | SRCC = 0.8583 ] +INFO:root:Average loss=0.1936720508413437 +INFO:root: +[Valid][13][UTT][ MSE = 0.2752 | LCC = 0.8596 | SRCC = 0.8623 ] +INFO:root:Average loss=0.19183988429797003 +INFO:root: +[Valid][14][UTT][ MSE = 0.2484 | LCC = 0.8561 | SRCC = 0.8600 ] +INFO:root:Average loss=0.19582822441290587 +INFO:root: +[Valid][15][UTT][ MSE = 0.2559 | LCC = 0.8567 | SRCC = 0.8598 ] +INFO:root:Average loss=0.18961058327784905 +INFO:root: +[Valid][16][UTT][ MSE = 0.2457 | LCC = 0.8586 | SRCC = 0.8634 ] +INFO:root:Average loss=0.18980329693892065 +INFO:root: +[Valid][17][UTT][ MSE = 0.2406 | LCC = 0.8612 | SRCC = 0.8639 ] +INFO:root:Average loss=0.18514621544342774 +INFO:root: +[Valid][18][UTT][ MSE = 0.2849 | LCC = 0.8606 | SRCC = 0.8581 ] +INFO:root:Average loss=0.18831180532773337 +INFO:root: +[Valid][19][UTT][ MSE = 0.2605 | LCC = 0.8588 | SRCC = 0.8601 ] +INFO:root:Average loss=0.18325862957116884 +INFO:root: +[Valid][20][UTT][ MSE = 0.2571 | LCC = 0.8580 | SRCC = 0.8588 ] +INFO:root:Average loss=0.17952390225269857 +INFO:root: +[Valid][21][UTT][ MSE = 0.2439 | LCC = 0.8539 | SRCC = 0.8589 ] +INFO:root:Average loss=0.17488141806843954 +INFO:root: +[Valid][22][UTT][ MSE = 0.2565 | LCC = 0.8566 | SRCC = 0.8568 ] +INFO:root:Average loss=0.1734400620827308 +INFO:root: +[Valid][23][UTT][ MSE = 0.2939 | LCC = 0.8576 | SRCC = 0.8579 ] +INFO:root:Average loss=0.17110599481906646 +INFO:root: +[Valid][24][UTT][ MSE = 0.2585 | LCC = 0.8541 | SRCC = 0.8551 ] +INFO:root:Average loss=0.17413157320175415 +INFO:root: +[Valid][25][UTT][ MSE = 0.2602 | LCC = 0.8570 | SRCC = 0.8548 ] +INFO:root:Average loss=0.16771893059978119 +INFO:root: +[Valid][26][UTT][ MSE = 0.2526 | LCC = 0.8562 | SRCC = 0.8544 ] +INFO:root:Average loss=0.1652707373484587 +INFO:root: +[Valid][27][UTT][ MSE = 0.2878 | LCC = 0.8543 | SRCC = 0.8506 ] +INFO:root:Average loss=0.16280121919818413 +INFO:root: +[Valid][28][UTT][ MSE = 0.2538 | LCC = 0.8525 | SRCC = 0.8549 ] +INFO:root:Average loss=0.16092636426671958 +INFO:root: +[Valid][29][UTT][ MSE = 0.2362 | LCC = 0.8543 | SRCC = 0.8554 ] +INFO:root:Average loss=0.15667874566637552 +INFO:root: +[Valid][30][UTT][ MSE = 0.2529 | LCC = 0.8557 | SRCC = 0.8537 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2499 | LCC = 0.8596 | SRCC = 0.8644 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f8c21d3c748418188e6facc64c7573da4aa2b809 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c3a483f756efb6551001d43aaf9d0c911ed66608e6483fafa290c985e4581ad2 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..42da1d9c6cf37ec91a735b884ea469eb306198ff --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:122708d1ac20feb7fa27e15e06c8a3597a2669b7d4beccb209c9c0aa4036463d +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3f8c86fbd804689bdb4446341040f53d0c8a4215 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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=0.3246852866350076 +INFO:root: +[Valid][0][UTT][ MSE = 0.3404 | LCC = 0.8278 | SRCC = 0.8250 ] +INFO:root:Average loss=0.25517930071323347 +INFO:root: +[Valid][1][UTT][ MSE = 0.3270 | LCC = 0.8415 | SRCC = 0.8398 ] +INFO:root:Average loss=0.2348356406467083 +INFO:root: +[Valid][2][UTT][ MSE = 0.3160 | LCC = 0.8497 | SRCC = 0.8489 ] +INFO:root:Average loss=0.22735713231257904 +INFO:root: +[Valid][3][UTT][ MSE = 0.3173 | LCC = 0.8507 | SRCC = 0.8501 ] +INFO:root:Average loss=0.22333444120028081 +INFO:root: +[Valid][4][UTT][ MSE = 0.3120 | LCC = 0.8582 | SRCC = 0.8546 ] +INFO:root:Average loss=0.2221529847727372 +INFO:root: +[Valid][5][UTT][ MSE = 0.2781 | LCC = 0.8563 | SRCC = 0.8584 ] +INFO:root:Average loss=0.21297723398758814 +INFO:root: +[Valid][6][UTT][ MSE = 0.3000 | LCC = 0.8607 | SRCC = 0.8603 ] +INFO:root:Average loss=0.22382618219424516 +INFO:root: +[Valid][7][UTT][ MSE = 0.2855 | LCC = 0.8558 | SRCC = 0.8551 ] +INFO:root:Average loss=0.2085074239816421 +INFO:root: +[Valid][8][UTT][ MSE = 0.3322 | LCC = 0.8632 | SRCC = 0.8606 ] +INFO:root:Average loss=0.2069582448173792 +INFO:root: +[Valid][9][UTT][ MSE = 0.2504 | LCC = 0.8573 | SRCC = 0.8586 ] +INFO:root:Average loss=0.20313230863748452 +INFO:root: +[Valid][10][UTT][ MSE = 0.2782 | LCC = 0.8590 | SRCC = 0.8595 ] +INFO:root:Average loss=0.2031775364317955 +INFO:root: +[Valid][11][UTT][ MSE = 0.3068 | LCC = 0.8611 | SRCC = 0.8581 ] +INFO:root:Average loss=0.19881932991437423 +INFO:root: +[Valid][12][UTT][ MSE = 0.2594 | LCC = 0.8541 | SRCC = 0.8591 ] +INFO:root:Average loss=0.1959917270220243 +INFO:root: +[Valid][13][UTT][ MSE = 0.3065 | LCC = 0.8638 | SRCC = 0.8603 ] +INFO:root:Average loss=0.1911041654455356 +INFO:root: +[Valid][14][UTT][ MSE = 0.2614 | LCC = 0.8631 | SRCC = 0.8627 ] +INFO:root:Average loss=0.18861697575984857 +INFO:root: +[Valid][15][UTT][ MSE = 0.2645 | LCC = 0.8639 | SRCC = 0.8631 ] +INFO:root:Average loss=0.1860273459400886 +INFO:root: +[Valid][16][UTT][ MSE = 0.2547 | LCC = 0.8614 | SRCC = 0.8629 ] +INFO:root:Average loss=0.18769671653325742 +INFO:root: +[Valid][17][UTT][ MSE = 0.2862 | LCC = 0.8582 | SRCC = 0.8611 ] +INFO:root:Average loss=0.1835964471101761 +INFO:root: +[Valid][18][UTT][ MSE = 0.2686 | LCC = 0.8644 | SRCC = 0.8642 ] +INFO:root:Average loss=0.18307214421339524 +INFO:root: +[Valid][19][UTT][ MSE = 0.2904 | LCC = 0.8655 | SRCC = 0.8639 ] +INFO:root:Average loss=0.18091119768527839 +INFO:root: +[Valid][20][UTT][ MSE = 0.2857 | LCC = 0.8623 | SRCC = 0.8660 ] +INFO:root:Average loss=0.17749278858686104 +INFO:root: +[Valid][21][UTT][ MSE = 0.2954 | LCC = 0.8626 | SRCC = 0.8592 ] +INFO:root:Average loss=0.17297363434082422 +INFO:root: +[Valid][22][UTT][ MSE = 0.2774 | LCC = 0.8607 | SRCC = 0.8585 ] +INFO:root:Average loss=0.17143952798766968 +INFO:root: +[Valid][23][UTT][ MSE = 0.2685 | LCC = 0.8636 | SRCC = 0.8622 ] +INFO:root:Average loss=0.1672597786363883 +INFO:root: +[Valid][24][UTT][ MSE = 0.2929 | LCC = 0.8603 | SRCC = 0.8596 ] +INFO:root:Average loss=0.1660705289015403 +INFO:root: +[Valid][25][UTT][ MSE = 0.3033 | LCC = 0.8626 | SRCC = 0.8633 ] +INFO:root:Average loss=0.17029586711373085 +INFO:root: +[Valid][26][UTT][ MSE = 0.2976 | LCC = 0.8575 | SRCC = 0.8599 ] +INFO:root:Average loss=0.16238107771063462 +INFO:root: +[Valid][27][UTT][ MSE = 0.2562 | LCC = 0.8535 | SRCC = 0.8573 ] +INFO:root:Average loss=0.16015864555270243 +INFO:root: +[Valid][28][UTT][ MSE = 0.2770 | LCC = 0.8599 | SRCC = 0.8589 ] +INFO:root:Average loss=0.158100991581495 +INFO:root: +[Valid][29][UTT][ MSE = 0.2722 | LCC = 0.8524 | SRCC = 0.8562 ] +INFO:root:Average loss=0.15285842617352804 +INFO:root: +[Valid][30][UTT][ MSE = 0.2632 | LCC = 0.8560 | SRCC = 0.8595 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2825 | LCC = 0.8626 | SRCC = 0.8604 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2d1660c1c676c59dce96c10c2c536f99077bf743 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7d39aeccc685159697a170ef723996745f164ad71f6737bfd1a456d9079dc7d2 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..11fa137b5c5fb9782cae115f2e514d532916cbb1 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f580b74d6dc052c24e6ae24e07d0aa9eacbdd878b93135e5dd20006cfb01c38b +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..8e249deda92869438d6f3140cbdde1cd65f7d93f --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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=0.30923270071164155 +INFO:root: +[Valid][0][UTT][ MSE = 0.3278 | LCC = 0.8400 | SRCC = 0.8384 ] +INFO:root:Average loss=0.25104309885929793 +INFO:root: +[Valid][1][UTT][ MSE = 0.3144 | LCC = 0.8450 | SRCC = 0.8452 ] +INFO:root:Average loss=0.23939690127586707 +INFO:root: +[Valid][2][UTT][ MSE = 0.3037 | LCC = 0.8526 | SRCC = 0.8509 ] +INFO:root:Average loss=0.22865668665140104 +INFO:root: +[Valid][3][UTT][ MSE = 0.2776 | LCC = 0.8573 | SRCC = 0.8554 ] +INFO:root:Average loss=0.22834542450996545 +INFO:root: +[Valid][4][UTT][ MSE = 0.2829 | LCC = 0.8541 | SRCC = 0.8562 ] +INFO:root:Average loss=0.22117490531542364 +INFO:root: +[Valid][5][UTT][ MSE = 0.2775 | LCC = 0.8484 | SRCC = 0.8515 ] +INFO:root:Average loss=0.21959481168633851 +INFO:root: +[Valid][6][UTT][ MSE = 0.2935 | LCC = 0.8612 | SRCC = 0.8612 ] +INFO:root:Average loss=0.22723054675719676 +INFO:root: +[Valid][7][UTT][ MSE = 0.2750 | LCC = 0.8563 | SRCC = 0.8556 ] +INFO:root:Average loss=0.21153454692699972 +INFO:root: +[Valid][8][UTT][ MSE = 0.3055 | LCC = 0.8612 | SRCC = 0.8591 ] +INFO:root:Average loss=0.21248760828987145 +INFO:root: +[Valid][9][UTT][ MSE = 0.2851 | LCC = 0.8591 | SRCC = 0.8600 ] +INFO:root:Average loss=0.2057060688160933 +INFO:root: +[Valid][10][UTT][ MSE = 0.2698 | LCC = 0.8611 | SRCC = 0.8608 ] +INFO:root:Average loss=0.19716712966179237 +INFO:root: +[Valid][11][UTT][ MSE = 0.2737 | LCC = 0.8602 | SRCC = 0.8610 ] +INFO:root:Average loss=0.19774731477865806 +INFO:root: +[Valid][12][UTT][ MSE = 0.3047 | LCC = 0.8637 | SRCC = 0.8602 ] +INFO:root:Average loss=0.1979124435247519 +INFO:root: +[Valid][13][UTT][ MSE = 0.2750 | LCC = 0.8639 | SRCC = 0.8622 ] +INFO:root:Average loss=0.19404557079840928 +INFO:root: +[Valid][14][UTT][ MSE = 0.2629 | LCC = 0.8577 | SRCC = 0.8602 ] +INFO:root:Average loss=0.19254546717573434 +INFO:root: +[Valid][15][UTT][ MSE = 0.2562 | LCC = 0.8605 | SRCC = 0.8625 ] +INFO:root:Average loss=0.19092896790840688 +INFO:root: +[Valid][16][UTT][ MSE = 0.3034 | LCC = 0.8637 | SRCC = 0.8602 ] +INFO:root:Average loss=0.18788010760759696 +INFO:root: +[Valid][17][UTT][ MSE = 0.2725 | LCC = 0.8587 | SRCC = 0.8602 ] +INFO:root:Average loss=0.18451068884669206 +INFO:root: +[Valid][18][UTT][ MSE = 0.2456 | LCC = 0.8608 | SRCC = 0.8612 ] +INFO:root:Average loss=0.18371053317036384 +INFO:root: +[Valid][19][UTT][ MSE = 0.2681 | LCC = 0.8627 | SRCC = 0.8599 ] +INFO:root:Average loss=0.18538475466462281 +INFO:root: +[Valid][20][UTT][ MSE = 0.2456 | LCC = 0.8603 | SRCC = 0.8605 ] +INFO:root:Average loss=0.17917845350427505 +INFO:root: +[Valid][21][UTT][ MSE = 0.2891 | LCC = 0.8616 | SRCC = 0.8584 ] +INFO:root:Average loss=0.17483650520443916 +INFO:root: +[Valid][22][UTT][ MSE = 0.2701 | LCC = 0.8605 | SRCC = 0.8592 ] +INFO:root:Average loss=0.18093127728654787 +INFO:root: +[Valid][23][UTT][ MSE = 0.2857 | LCC = 0.8581 | SRCC = 0.8618 ] +INFO:root:Average loss=0.17316666818582094 +INFO:root: +[Valid][24][UTT][ MSE = 0.2765 | LCC = 0.8596 | SRCC = 0.8554 ] +INFO:root:Average loss=0.16680500856958902 +INFO:root: +[Valid][25][UTT][ MSE = 0.2779 | LCC = 0.8622 | SRCC = 0.8597 ] +INFO:root:Average loss=0.16870299774484757 +INFO:root: +[Valid][26][UTT][ MSE = 0.2551 | LCC = 0.8593 | SRCC = 0.8576 ] +INFO:root:Average loss=0.16118149745922822 +INFO:root: +[Valid][27][UTT][ MSE = 0.2539 | LCC = 0.8591 | SRCC = 0.8568 ] +INFO:root:Average loss=0.1595595597456663 +INFO:root: +[Valid][28][UTT][ MSE = 0.2677 | LCC = 0.8595 | SRCC = 0.8564 ] +INFO:root:Average loss=0.1655737812129351 +INFO:root: +[Valid][29][UTT][ MSE = 0.2478 | LCC = 0.8599 | SRCC = 0.8589 ] +INFO:root:Average loss=0.1586895581239309 +INFO:root: +[Valid][30][UTT][ MSE = 0.3096 | LCC = 0.8550 | SRCC = 0.8567 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2730 | LCC = 0.8624 | SRCC = 0.8636 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bfe2b377d7f90772208d6ece11e5bce698d2e1de --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e49002f3c8a075a0e525036da1c4a1d13253137db6a62c407a5ab929e17351c1 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..cbc9d9ce379e1884fc8684b105beb943626eb30f --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd8d7d98296925cc004b359c50981fb1d66203cfd20553c67fd54ec85d15d1e1 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a0f16f99d01fa8f4264a8096967aab4ad9a9719d --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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=0.35207090182946277 +INFO:root: +[Valid][0][UTT][ MSE = 0.4253 | LCC = 0.8362 | SRCC = 0.8328 ] +INFO:root:Average loss=0.25445687006681394 +INFO:root: +[Valid][1][UTT][ MSE = 0.4840 | LCC = 0.8274 | SRCC = 0.8373 ] +INFO:root:Average loss=0.24416390577187905 +INFO:root: +[Valid][2][UTT][ MSE = 0.3491 | LCC = 0.8473 | SRCC = 0.8463 ] +INFO:root:Average loss=0.22852604301312032 +INFO:root: +[Valid][3][UTT][ MSE = 0.4171 | LCC = 0.8439 | SRCC = 0.8492 ] +INFO:root:Average loss=0.22777678741094393 +INFO:root: +[Valid][4][UTT][ MSE = 0.3552 | LCC = 0.8530 | SRCC = 0.8512 ] +INFO:root:Average loss=0.22641356938924545 +INFO:root: +[Valid][5][UTT][ MSE = 0.3537 | LCC = 0.8450 | SRCC = 0.8541 ] +INFO:root:Average loss=0.22101333240667978 +INFO:root: +[Valid][6][UTT][ MSE = 0.3764 | LCC = 0.8558 | SRCC = 0.8574 ] +INFO:root:Average loss=0.21980507862873566 +INFO:root: +[Valid][7][UTT][ MSE = 0.3385 | LCC = 0.8565 | SRCC = 0.8572 ] +INFO:root:Average loss=0.20774678198190835 +INFO:root: +[Valid][8][UTT][ MSE = 0.3180 | LCC = 0.8589 | SRCC = 0.8584 ] +INFO:root:Average loss=0.20892594907528314 +INFO:root: +[Valid][9][UTT][ MSE = 0.3155 | LCC = 0.8533 | SRCC = 0.8566 ] +INFO:root:Average loss=0.20345410504020178 +INFO:root: +[Valid][10][UTT][ MSE = 0.3433 | LCC = 0.8548 | SRCC = 0.8583 ] +INFO:root:Average loss=0.2033866496804433 +INFO:root: +[Valid][11][UTT][ MSE = 0.3021 | LCC = 0.8619 | SRCC = 0.8614 ] +INFO:root:Average loss=0.19862065655298722 +INFO:root: +[Valid][12][UTT][ MSE = 0.2651 | LCC = 0.8655 | SRCC = 0.8649 ] +INFO:root:Average loss=0.20523379819515425 +INFO:root: +[Valid][13][UTT][ MSE = 0.2844 | LCC = 0.8619 | SRCC = 0.8608 ] +INFO:root:Average loss=0.19357587779179597 +INFO:root: +[Valid][14][UTT][ MSE = 0.3007 | LCC = 0.8523 | SRCC = 0.8630 ] +INFO:root:Average loss=0.19317989672223726 +INFO:root: +[Valid][15][UTT][ MSE = 0.3004 | LCC = 0.8588 | SRCC = 0.8626 ] +INFO:root:Average loss=0.18832878338602874 +INFO:root: +[Valid][16][UTT][ MSE = 0.3823 | LCC = 0.8619 | SRCC = 0.8623 ] +INFO:root:Average loss=0.19393357462607896 +INFO:root: +[Valid][17][UTT][ MSE = 0.2866 | LCC = 0.8614 | SRCC = 0.8624 ] +INFO:root:Average loss=0.18365310457272407 +INFO:root: +[Valid][18][UTT][ MSE = 0.3237 | LCC = 0.8622 | SRCC = 0.8614 ] +INFO:root:Average loss=0.18611807720019266 +INFO:root: +[Valid][19][UTT][ MSE = 0.2951 | LCC = 0.8616 | SRCC = 0.8623 ] +INFO:root:Average loss=0.181145678823575 +INFO:root: +[Valid][20][UTT][ MSE = 0.2885 | LCC = 0.8620 | SRCC = 0.8627 ] +INFO:root:Average loss=0.1764710104236236 +INFO:root: +[Valid][21][UTT][ MSE = 0.2987 | LCC = 0.8631 | SRCC = 0.8601 ] +INFO:root:Average loss=0.17525963456584856 +INFO:root: +[Valid][22][UTT][ MSE = 0.3050 | LCC = 0.8636 | SRCC = 0.8622 ] +INFO:root:Average loss=0.17550748548446557 +INFO:root: +[Valid][23][UTT][ MSE = 0.3186 | LCC = 0.8636 | SRCC = 0.8621 ] +INFO:root:Average loss=0.17471898404451516 +INFO:root: +[Valid][24][UTT][ MSE = 0.3027 | LCC = 0.8602 | SRCC = 0.8601 ] +INFO:root:Average loss=0.1690089025367529 +INFO:root: +[Valid][25][UTT][ MSE = 0.2795 | LCC = 0.8590 | SRCC = 0.8656 ] +INFO:root:Average loss=0.17415051783124605 +INFO:root: +[Valid][26][UTT][ MSE = 0.2641 | LCC = 0.8637 | SRCC = 0.8627 ] +INFO:root:Average loss=0.1648880342642466 +INFO:root: +[Valid][27][UTT][ MSE = 0.3048 | LCC = 0.8592 | SRCC = 0.8567 ] +INFO:root:Average loss=0.1629031809667746 +INFO:root: +[Valid][28][UTT][ MSE = 0.2997 | LCC = 0.8556 | SRCC = 0.8630 ] +INFO:root:Average loss=0.1666495878344927 +INFO:root: +[Valid][29][UTT][ MSE = 0.2970 | LCC = 0.8605 | SRCC = 0.8619 ] +INFO:root:Average loss=0.15785632406671843 +INFO:root: +[Valid][30][UTT][ MSE = 0.3281 | LCC = 0.8570 | SRCC = 0.8610 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2579 | LCC = 0.8607 | SRCC = 0.8633 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a881da5db667521c60c26fc956191f36a16eacb4 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af774d7c768a62c53c9d3d5a2e3309795d7b666f54181d7d75da795189439144 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b02be188d16d67cd3ea40bc66c9b51383108a1c0 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4e08d2055a536ba1990d4bbf8518dba7b88f63a40a02061c04a6e49ffd0b35e5 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer10/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0d3659d39f8b164b952c6250cc066b4c26748373 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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=0.3120487406849861 +INFO:root: +[Valid][0][UTT][ MSE = 0.3685 | LCC = 0.8322 | SRCC = 0.8346 ] +INFO:root:Average loss=0.25050701143649906 +INFO:root: +[Valid][1][UTT][ MSE = 0.2785 | LCC = 0.8404 | SRCC = 0.8467 ] +INFO:root:Average loss=0.24524648048174688 +INFO:root: +[Valid][2][UTT][ MSE = 0.3192 | LCC = 0.8503 | SRCC = 0.8516 ] +INFO:root:Average loss=0.23233794440061617 +INFO:root: +[Valid][3][UTT][ MSE = 0.3059 | LCC = 0.8552 | SRCC = 0.8531 ] +INFO:root:Average loss=0.22374157015329751 +INFO:root: +[Valid][4][UTT][ MSE = 0.2596 | LCC = 0.8549 | SRCC = 0.8567 ] +INFO:root:Average loss=0.22189512122900057 +INFO:root: +[Valid][5][UTT][ MSE = 0.2679 | LCC = 0.8554 | SRCC = 0.8571 ] +INFO:root:Average loss=0.21690141161282858 +INFO:root: +[Valid][6][UTT][ MSE = 0.2866 | LCC = 0.8597 | SRCC = 0.8584 ] +INFO:root:Average loss=0.2156898381236272 +INFO:root: +[Valid][7][UTT][ MSE = 0.2945 | LCC = 0.8595 | SRCC = 0.8627 ] +INFO:root:Average loss=0.2099729638833266 +INFO:root: +[Valid][8][UTT][ MSE = 0.2623 | LCC = 0.8562 | SRCC = 0.8589 ] +INFO:root:Average loss=0.20977254383839095 +INFO:root: +[Valid][9][UTT][ MSE = 0.2641 | LCC = 0.8565 | SRCC = 0.8615 ] +INFO:root:Average loss=0.2048671929500042 +INFO:root: +[Valid][10][UTT][ MSE = 0.2592 | LCC = 0.8641 | SRCC = 0.8649 ] +INFO:root:Average loss=0.20311110247022066 +INFO:root: +[Valid][11][UTT][ MSE = 0.2620 | LCC = 0.8635 | SRCC = 0.8636 ] +INFO:root:Average loss=0.2009956236833181 +INFO:root: +[Valid][12][UTT][ MSE = 0.2682 | LCC = 0.8625 | SRCC = 0.8641 ] +INFO:root:Average loss=0.20089152799202845 +INFO:root: +[Valid][13][UTT][ MSE = 0.2741 | LCC = 0.8600 | SRCC = 0.8635 ] +INFO:root:Average loss=0.19873928221372458 +INFO:root: +[Valid][14][UTT][ MSE = 0.2635 | LCC = 0.8624 | SRCC = 0.8632 ] +INFO:root:Average loss=0.19114753231406212 +INFO:root: +[Valid][15][UTT][ MSE = 0.2891 | LCC = 0.8643 | SRCC = 0.8623 ] +INFO:root:Average loss=0.19391632958864555 +INFO:root: +[Valid][16][UTT][ MSE = 0.2804 | LCC = 0.8632 | SRCC = 0.8636 ] +INFO:root:Average loss=0.19001176284673887 +INFO:root: +[Valid][17][UTT][ MSE = 0.2844 | LCC = 0.8614 | SRCC = 0.8642 ] +INFO:root:Average loss=0.19607973328003517 +INFO:root: +[Valid][18][UTT][ MSE = 0.2699 | LCC = 0.8618 | SRCC = 0.8627 ] +INFO:root:Average loss=0.18593846366573602 +INFO:root: +[Valid][19][UTT][ MSE = 0.2804 | LCC = 0.8612 | SRCC = 0.8648 ] +INFO:root:Average loss=0.1822114592561355 +INFO:root: +[Valid][20][UTT][ MSE = 0.2792 | LCC = 0.8620 | SRCC = 0.8633 ] +INFO:root:Average loss=0.18213469898089385 +INFO:root: +[Valid][21][UTT][ MSE = 0.2565 | LCC = 0.8550 | SRCC = 0.8612 ] +INFO:root:Average loss=0.17925947981003004 +INFO:root: +[Valid][22][UTT][ MSE = 0.2415 | LCC = 0.8571 | SRCC = 0.8623 ] +INFO:root:Average loss=0.18110982099404702 +INFO:root: +[Valid][23][UTT][ MSE = 0.3182 | LCC = 0.8611 | SRCC = 0.8606 ] +INFO:root:Average loss=0.17502323203744033 +INFO:root: +[Valid][24][UTT][ MSE = 0.2562 | LCC = 0.8605 | SRCC = 0.8616 ] +INFO:root:Average loss=0.1695346172230366 +INFO:root: +[Valid][25][UTT][ MSE = 0.2680 | LCC = 0.8626 | SRCC = 0.8637 ] +INFO:root:Average loss=0.1686308124126532 +INFO:root: +[Valid][26][UTT][ MSE = 0.2605 | LCC = 0.8620 | SRCC = 0.8618 ] +INFO:root:Average loss=0.16711103247526365 +INFO:root: +[Valid][27][UTT][ MSE = 0.2628 | LCC = 0.8608 | SRCC = 0.8620 ] +INFO:root:Average loss=0.16651450852170968 +INFO:root: +[Valid][28][UTT][ MSE = 0.2545 | LCC = 0.8544 | SRCC = 0.8596 ] +INFO:root:Average loss=0.16207221198158386 +INFO:root: +[Valid][29][UTT][ MSE = 0.2812 | LCC = 0.8575 | SRCC = 0.8565 ] +INFO:root:Average loss=0.16496230509036627 +INFO:root: +[Valid][30][UTT][ MSE = 0.2593 | LCC = 0.8622 | SRCC = 0.8633 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2803 | LCC = 0.8627 | SRCC = 0.8619 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..52f049472fb0d251abf49a099e6b0955e0b50e38 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:efde597a2555f07af20278f3015c095f8aac65174d43b10a17141e8cdb696b46 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0a0e83ac67c0cf91f3c7a6fcbb777068e744d46b --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fdafe02cacd1d5d299b3230cc242ba29214e1e06645f86f7e10e4fea12b339f1 +size 2749638 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..47425907cd098d50855b920dae0fb39c99c3b011 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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=0.4532192104902023 +INFO:root: +[Valid][0][UTT][ MSE = 0.3944 | LCC = 0.8110 | SRCC = 0.8143 ] +INFO:root:Average loss=0.2853809058284148 +INFO:root: +[Valid][1][UTT][ MSE = 0.3765 | LCC = 0.8293 | SRCC = 0.8253 ] +INFO:root:Average loss=0.26625718634862167 +INFO:root: +[Valid][2][UTT][ MSE = 0.3561 | LCC = 0.8391 | SRCC = 0.8369 ] +INFO:root:Average loss=0.25990709318564487 +INFO:root: +[Valid][3][UTT][ MSE = 0.3481 | LCC = 0.8396 | SRCC = 0.8372 ] +INFO:root:Average loss=0.25026978666965777 +INFO:root: +[Valid][4][UTT][ MSE = 0.3619 | LCC = 0.8348 | SRCC = 0.8382 ] +INFO:root:Average loss=0.24460979016163412 +INFO:root: +[Valid][5][UTT][ MSE = 0.3269 | LCC = 0.8442 | SRCC = 0.8436 ] +INFO:root:Average loss=0.2489587916777684 +INFO:root: +[Valid][6][UTT][ MSE = 0.3498 | LCC = 0.8403 | SRCC = 0.8404 ] +INFO:root:Average loss=0.2438163894873399 +INFO:root: +[Valid][7][UTT][ MSE = 0.3059 | LCC = 0.8444 | SRCC = 0.8449 ] +INFO:root:Average loss=0.231833700186167 +INFO:root: +[Valid][8][UTT][ MSE = 0.3257 | LCC = 0.8479 | SRCC = 0.8475 ] +INFO:root:Average loss=0.23653438534492102 +INFO:root: +[Valid][9][UTT][ MSE = 0.3096 | LCC = 0.8466 | SRCC = 0.8487 ] +INFO:root:Average loss=0.23169876367617875 +INFO:root: +[Valid][10][UTT][ MSE = 0.3071 | LCC = 0.8510 | SRCC = 0.8500 ] +INFO:root:Average loss=0.23079887338173696 +INFO:root: +[Valid][11][UTT][ MSE = 0.3203 | LCC = 0.8559 | SRCC = 0.8536 ] +INFO:root:Average loss=0.2253613599982017 +INFO:root: +[Valid][12][UTT][ MSE = 0.2811 | LCC = 0.8519 | SRCC = 0.8538 ] +INFO:root:Average loss=0.22143562118976545 +INFO:root: +[Valid][13][UTT][ MSE = 0.2870 | LCC = 0.8507 | SRCC = 0.8514 ] +INFO:root:Average loss=0.2174448353739885 +INFO:root: +[Valid][14][UTT][ MSE = 0.3010 | LCC = 0.8541 | SRCC = 0.8547 ] +INFO:root:Average loss=0.22295742405530733 +INFO:root: +[Valid][15][UTT][ MSE = 0.3019 | LCC = 0.8436 | SRCC = 0.8495 ] +INFO:root:Average loss=0.22591596421522972 +INFO:root: +[Valid][16][UTT][ MSE = 0.2953 | LCC = 0.8527 | SRCC = 0.8552 ] +INFO:root:Average loss=0.21691885609657335 +INFO:root: +[Valid][17][UTT][ MSE = 0.2843 | LCC = 0.8499 | SRCC = 0.8533 ] +INFO:root:Average loss=0.21558100949877348 +INFO:root: +[Valid][18][UTT][ MSE = 0.3042 | LCC = 0.8533 | SRCC = 0.8533 ] +INFO:root:Average loss=0.21333952553761312 +INFO:root: +[Valid][19][UTT][ MSE = 0.3267 | LCC = 0.8559 | SRCC = 0.8564 ] +INFO:root:Average loss=0.2131421006260774 +INFO:root: +[Valid][20][UTT][ MSE = 0.3042 | LCC = 0.8597 | SRCC = 0.8593 ] +INFO:root:Average loss=0.21659544492379212 +INFO:root: +[Valid][21][UTT][ MSE = 0.3354 | LCC = 0.8545 | SRCC = 0.8537 ] +INFO:root:Average loss=0.21213401223604494 +INFO:root: +[Valid][22][UTT][ MSE = 0.2856 | LCC = 0.8566 | SRCC = 0.8565 ] +INFO:root:Average loss=0.21031709072681573 +INFO:root: +[Valid][23][UTT][ MSE = 0.2817 | LCC = 0.8581 | SRCC = 0.8576 ] +INFO:root:Average loss=0.2129031349069033 +INFO:root: +[Valid][24][UTT][ MSE = 0.3002 | LCC = 0.8554 | SRCC = 0.8559 ] +INFO:root:Average loss=0.2092898652339593 +INFO:root: +[Valid][25][UTT][ MSE = 0.2823 | LCC = 0.8555 | SRCC = 0.8561 ] +INFO:root:Average loss=0.20808723511604163 +INFO:root: +[Valid][26][UTT][ MSE = 0.2998 | LCC = 0.8587 | SRCC = 0.8571 ] +INFO:root:Average loss=0.2084498014778663 +INFO:root: +[Valid][27][UTT][ MSE = 0.3130 | LCC = 0.8511 | SRCC = 0.8535 ] +INFO:root:Average loss=0.2067728659663445 +INFO:root: +[Valid][28][UTT][ MSE = 0.2741 | LCC = 0.8551 | SRCC = 0.8559 ] +INFO:root:Average loss=0.20294080932553 +INFO:root: +[Valid][29][UTT][ MSE = 0.2891 | LCC = 0.8565 | SRCC = 0.8557 ] +INFO:root:Average loss=0.20201842372234052 +INFO:root: +[Valid][30][UTT][ MSE = 0.2753 | LCC = 0.8551 | SRCC = 0.8570 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2847 | LCC = 0.8640 | SRCC = 0.8671 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b5f74a0f2497bb87322aebe7adee452e900405a2 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:46cd8acf884442f0e68876e8a6922f01d362e246248c747400e528838ae617e4 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..3f843f8ef65533e11760503fff2e351679726f08 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6051e0601e9a229694debb25d318123547875cf2389584fa345bc089034a368a +size 2749638 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..91f9733d7595ca39245c1d62cb6ee6bf3d135032 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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=0.3414864469415102 +INFO:root: +[Valid][0][UTT][ MSE = 0.3664 | LCC = 0.8183 | SRCC = 0.8139 ] +INFO:root:Average loss=0.2799505889415741 +INFO:root: +[Valid][1][UTT][ MSE = 0.3276 | LCC = 0.8237 | SRCC = 0.8232 ] +INFO:root:Average loss=0.2708041360362982 +INFO:root: +[Valid][2][UTT][ MSE = 0.3387 | LCC = 0.8346 | SRCC = 0.8337 ] +INFO:root:Average loss=0.2656720477419022 +INFO:root: +[Valid][3][UTT][ MSE = 0.3134 | LCC = 0.8362 | SRCC = 0.8364 ] +INFO:root:Average loss=0.2531787663316115 +INFO:root: +[Valid][4][UTT][ MSE = 0.2888 | LCC = 0.8355 | SRCC = 0.8392 ] +INFO:root:Average loss=0.24457821536522645 +INFO:root: +[Valid][5][UTT][ MSE = 0.2818 | LCC = 0.8412 | SRCC = 0.8419 ] +INFO:root:Average loss=0.24407632381488115 +INFO:root: +[Valid][6][UTT][ MSE = 0.3209 | LCC = 0.8479 | SRCC = 0.8464 ] +INFO:root:Average loss=0.23862082425218362 +INFO:root: +[Valid][7][UTT][ MSE = 0.3223 | LCC = 0.8490 | SRCC = 0.8483 ] +INFO:root:Average loss=0.23258551554037973 +INFO:root: +[Valid][8][UTT][ MSE = 0.3116 | LCC = 0.8495 | SRCC = 0.8500 ] +INFO:root:Average loss=0.22815328234663376 +INFO:root: +[Valid][9][UTT][ MSE = 0.2861 | LCC = 0.8520 | SRCC = 0.8509 ] +INFO:root:Average loss=0.23054251170311219 +INFO:root: +[Valid][10][UTT][ MSE = 0.2867 | LCC = 0.8516 | SRCC = 0.8523 ] +INFO:root:Average loss=0.23037156844750428 +INFO:root: +[Valid][11][UTT][ MSE = 0.2903 | LCC = 0.8504 | SRCC = 0.8531 ] +INFO:root:Average loss=0.2263024833339911 +INFO:root: +[Valid][12][UTT][ MSE = 0.2660 | LCC = 0.8481 | SRCC = 0.8509 ] +INFO:root:Average loss=0.2233288272833213 +INFO:root: +[Valid][13][UTT][ MSE = 0.2787 | LCC = 0.8511 | SRCC = 0.8516 ] +INFO:root:Average loss=0.22684401999681425 +INFO:root: +[Valid][14][UTT][ MSE = 0.3030 | LCC = 0.8498 | SRCC = 0.8523 ] +INFO:root:Average loss=0.23110827593467173 +INFO:root: +[Valid][15][UTT][ MSE = 0.3330 | LCC = 0.8538 | SRCC = 0.8560 ] +INFO:root:Average loss=0.21632092522505003 +INFO:root: +[Valid][16][UTT][ MSE = 0.2783 | LCC = 0.8525 | SRCC = 0.8536 ] +INFO:root:Average loss=0.21572734377323052 +INFO:root: +[Valid][17][UTT][ MSE = 0.2678 | LCC = 0.8493 | SRCC = 0.8535 ] +INFO:root:Average loss=0.21733841930444425 +INFO:root: +[Valid][18][UTT][ MSE = 0.2994 | LCC = 0.8514 | SRCC = 0.8547 ] +INFO:root:Average loss=0.21751118565981203 +INFO:root: +[Valid][19][UTT][ MSE = 0.2936 | LCC = 0.8550 | SRCC = 0.8540 ] +INFO:root:Average loss=0.21457694375362152 +INFO:root: +[Valid][20][UTT][ MSE = 0.2850 | LCC = 0.8525 | SRCC = 0.8586 ] +INFO:root:Average loss=0.22275107067364913 +INFO:root: +[Valid][21][UTT][ MSE = 0.2580 | LCC = 0.8550 | SRCC = 0.8552 ] +INFO:root:Average loss=0.21775192538133034 +INFO:root: +[Valid][22][UTT][ MSE = 0.2649 | LCC = 0.8558 | SRCC = 0.8573 ] +INFO:root:Average loss=0.212500556348226 +INFO:root: +[Valid][23][UTT][ MSE = 0.2880 | LCC = 0.8564 | SRCC = 0.8567 ] +INFO:root:Average loss=0.20610233052418783 +INFO:root: +[Valid][24][UTT][ MSE = 0.2777 | LCC = 0.8582 | SRCC = 0.8565 ] +INFO:root:Average loss=0.20787887370739228 +INFO:root: +[Valid][25][UTT][ MSE = 0.2734 | LCC = 0.8534 | SRCC = 0.8587 ] +INFO:root:Average loss=0.20735528759467295 +INFO:root: +[Valid][26][UTT][ MSE = 0.2883 | LCC = 0.8576 | SRCC = 0.8590 ] +INFO:root:Average loss=0.20822564569803384 +INFO:root: +[Valid][27][UTT][ MSE = 0.2700 | LCC = 0.8556 | SRCC = 0.8580 ] +INFO:root:Average loss=0.21004762443212363 +INFO:root: +[Valid][28][UTT][ MSE = 0.2785 | LCC = 0.8598 | SRCC = 0.8577 ] +INFO:root:Average loss=0.2038485791820746 +INFO:root: +[Valid][29][UTT][ MSE = 0.2655 | LCC = 0.8554 | SRCC = 0.8587 ] +INFO:root:Average loss=0.20266110946734747 +INFO:root: +[Valid][30][UTT][ MSE = 0.2611 | LCC = 0.8556 | SRCC = 0.8571 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2814 | LCC = 0.8616 | SRCC = 0.8651 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3aacad02eab9c395bff5b29a271dc741d9aede18 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b52138be9ff1b638ae74629fd8775ac002c9aaf8c9cb95755eb22773f0c524ee +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..120254b954b95017371a090ac84366258c87a9e9 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:506a0c0310978bb507180924605dabd62a3af2f4e6c4717a46633a58727cd850 +size 2749638 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..fe96a8f2bd861033e6550287ae2fe0434ce25057 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/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=0.32960711152125627 +INFO:root: +[Valid][0][UTT][ MSE = 0.4145 | LCC = 0.8161 | SRCC = 0.8160 ] +INFO:root:Average loss=0.2753743750926776 +INFO:root: +[Valid][1][UTT][ MSE = 0.4049 | LCC = 0.8197 | SRCC = 0.8290 ] +INFO:root:Average loss=0.27025715681986934 +INFO:root: +[Valid][2][UTT][ MSE = 0.3814 | LCC = 0.8378 | SRCC = 0.8350 ] +INFO:root:Average loss=0.25425670162225383 +INFO:root: +[Valid][3][UTT][ MSE = 0.3386 | LCC = 0.8394 | SRCC = 0.8403 ] +INFO:root:Average loss=0.24310960391393074 +INFO:root: +[Valid][4][UTT][ MSE = 0.3737 | LCC = 0.8429 | SRCC = 0.8414 ] +INFO:root:Average loss=0.24205551391992813 +INFO:root: +[Valid][5][UTT][ MSE = 0.3133 | LCC = 0.8448 | SRCC = 0.8451 ] +INFO:root:Average loss=0.23830199318054396 +INFO:root: +[Valid][6][UTT][ MSE = 0.2970 | LCC = 0.8479 | SRCC = 0.8472 ] +INFO:root:Average loss=0.24285172384518844 +INFO:root: +[Valid][7][UTT][ MSE = 0.3661 | LCC = 0.8488 | SRCC = 0.8479 ] +INFO:root:Average loss=0.23581334299001938 +INFO:root: +[Valid][8][UTT][ MSE = 0.3601 | LCC = 0.8473 | SRCC = 0.8502 ] +INFO:root:Average loss=0.2317460257655535 +INFO:root: +[Valid][9][UTT][ MSE = 0.2919 | LCC = 0.8519 | SRCC = 0.8527 ] +INFO:root:Average loss=0.22634849840631852 +INFO:root: +[Valid][10][UTT][ MSE = 0.3301 | LCC = 0.8515 | SRCC = 0.8499 ] +INFO:root:Average loss=0.23746592761614385 +INFO:root: +[Valid][11][UTT][ MSE = 0.3359 | LCC = 0.8534 | SRCC = 0.8530 ] +INFO:root:Average loss=0.2312026770833211 +INFO:root: +[Valid][12][UTT][ MSE = 0.3017 | LCC = 0.8504 | SRCC = 0.8500 ] +INFO:root:Average loss=0.23210391153891882 +INFO:root: +[Valid][13][UTT][ MSE = 0.2873 | LCC = 0.8497 | SRCC = 0.8525 ] +INFO:root:Average loss=0.22659973418101287 +INFO:root: +[Valid][14][UTT][ MSE = 0.3349 | LCC = 0.8563 | SRCC = 0.8571 ] +INFO:root:Average loss=0.22245604716814482 +INFO:root: +[Valid][15][UTT][ MSE = 0.3189 | LCC = 0.8549 | SRCC = 0.8538 ] +INFO:root:Average loss=0.22014366204922015 +INFO:root: +[Valid][16][UTT][ MSE = 0.2739 | LCC = 0.8512 | SRCC = 0.8533 ] +INFO:root:Average loss=0.21343828336550638 +INFO:root: +[Valid][17][UTT][ MSE = 0.3015 | LCC = 0.8556 | SRCC = 0.8542 ] +INFO:root:Average loss=0.22088520725568137 +INFO:root: +[Valid][18][UTT][ MSE = 0.3101 | LCC = 0.8515 | SRCC = 0.8519 ] +INFO:root:Average loss=0.21339474436946404 +INFO:root: +[Valid][19][UTT][ MSE = 0.3201 | LCC = 0.8568 | SRCC = 0.8560 ] +INFO:root:Average loss=0.21238505553740722 +INFO:root: +[Valid][20][UTT][ MSE = 0.3302 | LCC = 0.8516 | SRCC = 0.8503 ] +INFO:root:Average loss=0.20909397972699922 +INFO:root: +[Valid][21][UTT][ MSE = 0.3011 | LCC = 0.8575 | SRCC = 0.8567 ] +INFO:root:Average loss=0.21133586945824134 +INFO:root: +[Valid][22][UTT][ MSE = 0.2743 | LCC = 0.8523 | SRCC = 0.8549 ] +INFO:root:Average loss=0.20910177341638467 +INFO:root: +[Valid][23][UTT][ MSE = 0.2985 | LCC = 0.8548 | SRCC = 0.8546 ] +INFO:root:Average loss=0.2079564753251198 +INFO:root: +[Valid][24][UTT][ MSE = 0.2875 | LCC = 0.8562 | SRCC = 0.8590 ] +INFO:root:Average loss=0.2156316876793519 +INFO:root: +[Valid][25][UTT][ MSE = 0.2591 | LCC = 0.8513 | SRCC = 0.8549 ] +INFO:root:Average loss=0.20833934041169974 +INFO:root: +[Valid][26][UTT][ MSE = 0.2933 | LCC = 0.8533 | SRCC = 0.8541 ] +INFO:root:Average loss=0.20628936913533089 +INFO:root: +[Valid][27][UTT][ MSE = 0.2983 | LCC = 0.8530 | SRCC = 0.8570 ] +INFO:root:Average loss=0.20384962961841852 +INFO:root: +[Valid][28][UTT][ MSE = 0.3357 | LCC = 0.8539 | SRCC = 0.8527 ] +INFO:root:Average loss=0.2013640112410753 +INFO:root: +[Valid][29][UTT][ MSE = 0.2862 | LCC = 0.8555 | SRCC = 0.8551 ] +INFO:root:Average loss=0.2017555133654521 +INFO:root: +[Valid][30][UTT][ MSE = 0.3086 | LCC = 0.8596 | SRCC = 0.8590 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2804 | LCC = 0.8664 | SRCC = 0.8647 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..cedb5a769172e02e06609629f5fb7e06405596d5 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:32b1f6d5dff8959d5abfd1fed05eda4811a8f7009b37f73bf79dac1c1fd2cf78 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..273072727e6ce57c85b09b021d2ceccb9b5d8551 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f5d13e822f46a0d0f0e1e47eb35586e3f61a864779fd7147362494a106dfc930 +size 2749638 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..4e385d689a7aef918a919797a48486ccb57416e3 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/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=0.33540038688060564 +INFO:root: +[Valid][0][UTT][ MSE = 0.3836 | LCC = 0.8211 | SRCC = 0.8188 ] +INFO:root:Average loss=0.27259066509894836 +INFO:root: +[Valid][1][UTT][ MSE = 0.3774 | LCC = 0.8357 | SRCC = 0.8329 ] +INFO:root:Average loss=0.26478802852141553 +INFO:root: +[Valid][2][UTT][ MSE = 0.3345 | LCC = 0.8268 | SRCC = 0.8319 ] +INFO:root:Average loss=0.2608690158678935 +INFO:root: +[Valid][3][UTT][ MSE = 0.3569 | LCC = 0.8413 | SRCC = 0.8419 ] +INFO:root:Average loss=0.24157170645701578 +INFO:root: +[Valid][4][UTT][ MSE = 0.3258 | LCC = 0.8454 | SRCC = 0.8452 ] +INFO:root:Average loss=0.2433326693299489 +INFO:root: +[Valid][5][UTT][ MSE = 0.3519 | LCC = 0.8502 | SRCC = 0.8486 ] +INFO:root:Average loss=0.24424330832866523 +INFO:root: +[Valid][6][UTT][ MSE = 0.3239 | LCC = 0.8485 | SRCC = 0.8488 ] +INFO:root:Average loss=0.23971542257529038 +INFO:root: +[Valid][7][UTT][ MSE = 0.3129 | LCC = 0.8457 | SRCC = 0.8467 ] +INFO:root:Average loss=0.24198373368917367 +INFO:root: +[Valid][8][UTT][ MSE = 0.3203 | LCC = 0.8485 | SRCC = 0.8497 ] +INFO:root:Average loss=0.22992006823038444 +INFO:root: +[Valid][9][UTT][ MSE = 0.3215 | LCC = 0.8487 | SRCC = 0.8505 ] +INFO:root:Average loss=0.2292407575326088 +INFO:root: +[Valid][10][UTT][ MSE = 0.3012 | LCC = 0.8471 | SRCC = 0.8499 ] +INFO:root:Average loss=0.2277332271138827 +INFO:root: +[Valid][11][UTT][ MSE = 0.3410 | LCC = 0.8576 | SRCC = 0.8551 ] +INFO:root:Average loss=0.23650136341651282 +INFO:root: +[Valid][12][UTT][ MSE = 0.3020 | LCC = 0.8507 | SRCC = 0.8523 ] +INFO:root:Average loss=0.22721074731686178 +INFO:root: +[Valid][13][UTT][ MSE = 0.3112 | LCC = 0.8527 | SRCC = 0.8526 ] +INFO:root:Average loss=0.2252418345365769 +INFO:root: +[Valid][14][UTT][ MSE = 0.3515 | LCC = 0.8502 | SRCC = 0.8516 ] +INFO:root:Average loss=0.21861772468456855 +INFO:root: +[Valid][15][UTT][ MSE = 0.3239 | LCC = 0.8519 | SRCC = 0.8520 ] +INFO:root:Average loss=0.21813878627159658 +INFO:root: +[Valid][16][UTT][ MSE = 0.2911 | LCC = 0.8512 | SRCC = 0.8533 ] +INFO:root:Average loss=0.2159977152179449 +INFO:root: +[Valid][17][UTT][ MSE = 0.3266 | LCC = 0.8533 | SRCC = 0.8517 ] +INFO:root:Average loss=0.2219454858165521 +INFO:root: +[Valid][18][UTT][ MSE = 0.3108 | LCC = 0.8560 | SRCC = 0.8551 ] +INFO:root:Average loss=0.21609310223123965 +INFO:root: +[Valid][19][UTT][ MSE = 0.3055 | LCC = 0.8558 | SRCC = 0.8564 ] +INFO:root:Average loss=0.2117533198533914 +INFO:root: +[Valid][20][UTT][ MSE = 0.3332 | LCC = 0.8566 | SRCC = 0.8565 ] +INFO:root:Average loss=0.21728441806939933 +INFO:root: +[Valid][21][UTT][ MSE = 0.3167 | LCC = 0.8525 | SRCC = 0.8527 ] +INFO:root:Average loss=0.21311091956419823 +INFO:root: +[Valid][22][UTT][ MSE = 0.2878 | LCC = 0.8547 | SRCC = 0.8547 ] +INFO:root:Average loss=0.20646097969550353 +INFO:root: +[Valid][23][UTT][ MSE = 0.3003 | LCC = 0.8580 | SRCC = 0.8556 ] +INFO:root:Average loss=0.21453195619277465 +INFO:root: +[Valid][24][UTT][ MSE = 0.2852 | LCC = 0.8512 | SRCC = 0.8560 ] +INFO:root:Average loss=0.2070385542435524 +INFO:root: +[Valid][25][UTT][ MSE = 0.2835 | LCC = 0.8605 | SRCC = 0.8588 ] +INFO:root:Average loss=0.21043433057956207 +INFO:root: +[Valid][26][UTT][ MSE = 0.2945 | LCC = 0.8579 | SRCC = 0.8584 ] +INFO:root:Average loss=0.2058912294033246 +INFO:root: +[Valid][27][UTT][ MSE = 0.3204 | LCC = 0.8574 | SRCC = 0.8562 ] +INFO:root:Average loss=0.2021433256375484 +INFO:root: +[Valid][28][UTT][ MSE = 0.2954 | LCC = 0.8593 | SRCC = 0.8579 ] +INFO:root:Average loss=0.20159115327092317 +INFO:root: +[Valid][29][UTT][ MSE = 0.2888 | LCC = 0.8569 | SRCC = 0.8585 ] +INFO:root:Average loss=0.19780898418946144 +INFO:root: +[Valid][30][UTT][ MSE = 0.2794 | LCC = 0.8591 | SRCC = 0.8572 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2666 | LCC = 0.8621 | SRCC = 0.8651 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..11ff55acac786c6fcfb0704b57036407ed268c3f --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:80d6c461f7c85650a4d77255e9a3aee81cf894603f537a74ec80170da5570240 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..343176e4066f13bf8e6dd76857d9eb8f15b38ad5 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:37a9520100eb7f09f26dcb8d15290cd1ee384264381e594bbd12ddd2d1fc45f2 +size 2749638 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1fd31e8d41bbb78f43d9381cdbb75961d6decb98 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer11/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=0.3785460079327608 +INFO:root: +[Valid][0][UTT][ MSE = 0.4465 | LCC = 0.8223 | SRCC = 0.8167 ] +INFO:root:Average loss=0.2792622219675627 +INFO:root: +[Valid][1][UTT][ MSE = 0.5065 | LCC = 0.8165 | SRCC = 0.8261 ] +INFO:root:Average loss=0.26278477143018675 +INFO:root: +[Valid][2][UTT][ MSE = 0.4239 | LCC = 0.8411 | SRCC = 0.8391 ] +INFO:root:Average loss=0.246562102666268 +INFO:root: +[Valid][3][UTT][ MSE = 0.4693 | LCC = 0.8398 | SRCC = 0.8387 ] +INFO:root:Average loss=0.2447187751531601 +INFO:root: +[Valid][4][UTT][ MSE = 0.3796 | LCC = 0.8485 | SRCC = 0.8463 ] +INFO:root:Average loss=0.24058388383724752 +INFO:root: +[Valid][5][UTT][ MSE = 0.3893 | LCC = 0.8430 | SRCC = 0.8444 ] +INFO:root:Average loss=0.23746383839692825 +INFO:root: +[Valid][6][UTT][ MSE = 0.3731 | LCC = 0.8498 | SRCC = 0.8492 ] +INFO:root:Average loss=0.23611326611194855 +INFO:root: +[Valid][7][UTT][ MSE = 0.3900 | LCC = 0.8484 | SRCC = 0.8502 ] +INFO:root:Average loss=0.2259309966212664 +INFO:root: +[Valid][8][UTT][ MSE = 0.3502 | LCC = 0.8455 | SRCC = 0.8467 ] +INFO:root:Average loss=0.22928206278727606 +INFO:root: +[Valid][9][UTT][ MSE = 0.3598 | LCC = 0.8512 | SRCC = 0.8502 ] +INFO:root:Average loss=0.2288763633905313 +INFO:root: +[Valid][10][UTT][ MSE = 0.3260 | LCC = 0.8535 | SRCC = 0.8530 ] +INFO:root:Average loss=0.22591212067084435 +INFO:root: +[Valid][11][UTT][ MSE = 0.3596 | LCC = 0.8507 | SRCC = 0.8523 ] +INFO:root:Average loss=0.22300321685198027 +INFO:root: +[Valid][12][UTT][ MSE = 0.3815 | LCC = 0.8506 | SRCC = 0.8537 ] +INFO:root:Average loss=0.22146038558238593 +INFO:root: +[Valid][13][UTT][ MSE = 0.3765 | LCC = 0.8516 | SRCC = 0.8510 ] +INFO:root:Average loss=0.21896930516530305 +INFO:root: +[Valid][14][UTT][ MSE = 0.3550 | LCC = 0.8534 | SRCC = 0.8529 ] +INFO:root:Average loss=0.22646953127323052 +INFO:root: +[Valid][15][UTT][ MSE = 0.3643 | LCC = 0.8519 | SRCC = 0.8516 ] +INFO:root:Average loss=0.21822745200151053 +INFO:root: +[Valid][16][UTT][ MSE = 0.3555 | LCC = 0.8550 | SRCC = 0.8552 ] +INFO:root:Average loss=0.21372156227246308 +INFO:root: +[Valid][17][UTT][ MSE = 0.2991 | LCC = 0.8562 | SRCC = 0.8576 ] +INFO:root:Average loss=0.21027573541953012 +INFO:root: +[Valid][18][UTT][ MSE = 0.3263 | LCC = 0.8562 | SRCC = 0.8565 ] +INFO:root:Average loss=0.21555930662613648 +INFO:root: +[Valid][19][UTT][ MSE = 0.3615 | LCC = 0.8483 | SRCC = 0.8548 ] +INFO:root:Average loss=0.21341991004271385 +INFO:root: +[Valid][20][UTT][ MSE = 0.3014 | LCC = 0.8584 | SRCC = 0.8578 ] +INFO:root:Average loss=0.21686284377788886 +INFO:root: +[Valid][21][UTT][ MSE = 0.3568 | LCC = 0.8531 | SRCC = 0.8546 ] +INFO:root:Average loss=0.2149986240726251 +INFO:root: +[Valid][22][UTT][ MSE = 0.3033 | LCC = 0.8563 | SRCC = 0.8565 ] +INFO:root:Average loss=0.20660228033860525 +INFO:root: +[Valid][23][UTT][ MSE = 0.3167 | LCC = 0.8582 | SRCC = 0.8572 ] +INFO:root:Average loss=0.20763335950099504 +INFO:root: +[Valid][24][UTT][ MSE = 0.3019 | LCC = 0.8562 | SRCC = 0.8577 ] +INFO:root:Average loss=0.20287774102045938 +INFO:root: +[Valid][25][UTT][ MSE = 0.3380 | LCC = 0.8586 | SRCC = 0.8578 ] +INFO:root:Average loss=0.20943365761866936 +INFO:root: +[Valid][26][UTT][ MSE = 0.3342 | LCC = 0.8618 | SRCC = 0.8598 ] +INFO:root:Average loss=0.2138086524911416 +INFO:root: +[Valid][27][UTT][ MSE = 0.3549 | LCC = 0.8546 | SRCC = 0.8601 ] +INFO:root:Average loss=0.2010330289411239 +INFO:root: +[Valid][28][UTT][ MSE = 0.3248 | LCC = 0.8592 | SRCC = 0.8586 ] +INFO:root:Average loss=0.20346214345250374 +INFO:root: +[Valid][29][UTT][ MSE = 0.2957 | LCC = 0.8557 | SRCC = 0.8595 ] +INFO:root:Average loss=0.2044158310462267 +INFO:root: +[Valid][30][UTT][ MSE = 0.3344 | LCC = 0.8561 | SRCC = 0.8554 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3062 | LCC = 0.8643 | SRCC = 0.8645 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..060a82e65613491e97831f5ae25f7a2bd80eeda7 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4f55054a6998aa1e69e68cd6b8a1f8f42844bfc27984b6059ed6cbff7cda8b33 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a1562c811cec083863a29c3ca0ea847e9e899828 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8aaeb8113ff79255c6c7c7e03ded983960619d094d6f24ccad008dcb98261089 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ad43a6d50e96334da5be895bcb7cb527c12f15e8 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/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=0.39415592222641677 +INFO:root: +[Valid][0][UTT][ MSE = 0.2941 | LCC = 0.8182 | SRCC = 0.8295 ] +INFO:root:Average loss=0.23284295755319107 +INFO:root: +[Valid][1][UTT][ MSE = 0.2502 | LCC = 0.8508 | SRCC = 0.8524 ] +INFO:root:Average loss=0.20535496049202406 +INFO:root: +[Valid][2][UTT][ MSE = 0.2626 | LCC = 0.8588 | SRCC = 0.8581 ] +INFO:root:Average loss=0.18826627387450293 +INFO:root: +[Valid][3][UTT][ MSE = 0.2664 | LCC = 0.8650 | SRCC = 0.8633 ] +INFO:root:Average loss=0.17693574717029548 +INFO:root: +[Valid][4][UTT][ MSE = 0.2787 | LCC = 0.8672 | SRCC = 0.8667 ] +INFO:root:Average loss=0.16609821019646448 +INFO:root: +[Valid][5][UTT][ MSE = 0.2240 | LCC = 0.8693 | SRCC = 0.8691 ] +INFO:root:Average loss=0.15746110677719116 +INFO:root: +[Valid][6][UTT][ MSE = 0.2479 | LCC = 0.8709 | SRCC = 0.8705 ] +INFO:root:Average loss=0.14863241884188774 +INFO:root: +[Valid][7][UTT][ MSE = 0.2376 | LCC = 0.8733 | SRCC = 0.8734 ] +INFO:root:Average loss=0.14144992341215795 +INFO:root: +[Valid][8][UTT][ MSE = 0.2145 | LCC = 0.8721 | SRCC = 0.8719 ] +INFO:root:Average loss=0.12851081416010857 +INFO:root: +[Valid][9][UTT][ MSE = 0.2320 | LCC = 0.8680 | SRCC = 0.8686 ] +INFO:root:Average loss=0.12797482684254646 +INFO:root: +[Valid][10][UTT][ MSE = 0.2366 | LCC = 0.8741 | SRCC = 0.8737 ] +INFO:root:Average loss=0.11701420780557853 +INFO:root: +[Valid][11][UTT][ MSE = 0.2329 | LCC = 0.8667 | SRCC = 0.8675 ] +INFO:root:Average loss=0.10974441449611615 +INFO:root: +[Valid][12][UTT][ MSE = 0.2395 | LCC = 0.8712 | SRCC = 0.8698 ] +INFO:root:Average loss=0.10804665413422462 +INFO:root: +[Valid][13][UTT][ MSE = 0.2222 | LCC = 0.8768 | SRCC = 0.8751 ] +INFO:root:Average loss=0.0991703209777673 +INFO:root: +[Valid][14][UTT][ MSE = 0.2061 | LCC = 0.8756 | SRCC = 0.8747 ] +INFO:root:Average loss=0.09668676335460101 +INFO:root: +[Valid][15][UTT][ MSE = 0.2042 | LCC = 0.8699 | SRCC = 0.8704 ] +INFO:root:Average loss=0.09351040346500201 +INFO:root: +[Valid][16][UTT][ MSE = 0.2193 | LCC = 0.8734 | SRCC = 0.8722 ] +INFO:root:Average loss=0.08515675270404571 +INFO:root: +[Valid][17][UTT][ MSE = 0.2250 | LCC = 0.8731 | SRCC = 0.8705 ] +INFO:root:Average loss=0.07880854391707824 +INFO:root: +[Valid][18][UTT][ MSE = 0.2395 | LCC = 0.8689 | SRCC = 0.8674 ] +INFO:root:Average loss=0.0768394420066705 +INFO:root: +[Valid][19][UTT][ MSE = 0.2222 | LCC = 0.8700 | SRCC = 0.8698 ] +INFO:root:Average loss=0.07490547655675656 +INFO:root: +[Valid][20][UTT][ MSE = 0.2302 | LCC = 0.8709 | SRCC = 0.8695 ] +INFO:root:Average loss=0.08149352225546654 +INFO:root: +[Valid][21][UTT][ MSE = 0.2359 | LCC = 0.8692 | SRCC = 0.8698 ] +INFO:root:Average loss=0.07166705734263627 +INFO:root: +[Valid][22][UTT][ MSE = 0.2211 | LCC = 0.8693 | SRCC = 0.8681 ] +INFO:root:Average loss=0.06749512293399909 +INFO:root: +[Valid][23][UTT][ MSE = 0.2275 | LCC = 0.8678 | SRCC = 0.8678 ] +INFO:root:Average loss=0.06370113159601505 +INFO:root: +[Valid][24][UTT][ MSE = 0.2110 | LCC = 0.8726 | SRCC = 0.8728 ] +INFO:root:Average loss=0.061607940695606746 +INFO:root: +[Valid][25][UTT][ MSE = 0.2223 | LCC = 0.8687 | SRCC = 0.8680 ] +INFO:root:Average loss=0.05899640325552378 +INFO:root: +[Valid][26][UTT][ MSE = 0.2068 | LCC = 0.8692 | SRCC = 0.8691 ] +INFO:root:Average loss=0.05724171630274027 +INFO:root: +[Valid][27][UTT][ MSE = 0.2167 | LCC = 0.8692 | SRCC = 0.8686 ] +INFO:root:Average loss=0.058020406092206635 +INFO:root: +[Valid][28][UTT][ MSE = 0.2170 | LCC = 0.8683 | SRCC = 0.8673 ] +INFO:root:Average loss=0.06311979999717994 +INFO:root: +[Valid][29][UTT][ MSE = 0.2103 | LCC = 0.8711 | SRCC = 0.8704 ] +INFO:root:Average loss=0.05340368315004385 +INFO:root: +[Valid][30][UTT][ MSE = 0.2481 | LCC = 0.8687 | SRCC = 0.8686 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2582 | LCC = 0.8654 | SRCC = 0.8663 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a908e27dd1513a65055debe1a5ec7764a3ae6187 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1c30fe7ca659a797998f024293df332e72ef6f156c820e2f4dff6b63b2d4590b +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7d4755a8d815e4e3e815710f994c8fe026320686 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ebe667e139b649053d02d98f60a1d16bdc1f4a0dfc95982240d07b66a165d9d3 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..eb22ead1201218f60b9c6efaab8c9ef47e07b116 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/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=0.4488563550970493 +INFO:root: +[Valid][0][UTT][ MSE = 0.3094 | LCC = 0.8191 | SRCC = 0.8285 ] +INFO:root:Average loss=0.23703111306979105 +INFO:root: +[Valid][1][UTT][ MSE = 0.2661 | LCC = 0.8517 | SRCC = 0.8538 ] +INFO:root:Average loss=0.20717393492276853 +INFO:root: +[Valid][2][UTT][ MSE = 0.2757 | LCC = 0.8607 | SRCC = 0.8628 ] +INFO:root:Average loss=0.1892788257354345 +INFO:root: +[Valid][3][UTT][ MSE = 0.2465 | LCC = 0.8659 | SRCC = 0.8674 ] +INFO:root:Average loss=0.1784378124926335 +INFO:root: +[Valid][4][UTT][ MSE = 0.2376 | LCC = 0.8694 | SRCC = 0.8718 ] +INFO:root:Average loss=0.1647373222005673 +INFO:root: +[Valid][5][UTT][ MSE = 0.2359 | LCC = 0.8755 | SRCC = 0.8755 ] +INFO:root:Average loss=0.15470614628149912 +INFO:root: +[Valid][6][UTT][ MSE = 0.2260 | LCC = 0.8761 | SRCC = 0.8759 ] +INFO:root:Average loss=0.14281380472657007 +INFO:root: +[Valid][7][UTT][ MSE = 0.2099 | LCC = 0.8747 | SRCC = 0.8741 ] +INFO:root:Average loss=0.1398324010272821 +INFO:root: +[Valid][8][UTT][ MSE = 0.2328 | LCC = 0.8762 | SRCC = 0.8752 ] +INFO:root:Average loss=0.1324755206513099 +INFO:root: +[Valid][9][UTT][ MSE = 0.2294 | LCC = 0.8728 | SRCC = 0.8729 ] +INFO:root:Average loss=0.1256799146724053 +INFO:root: +[Valid][10][UTT][ MSE = 0.2107 | LCC = 0.8776 | SRCC = 0.8758 ] +INFO:root:Average loss=0.11984698617687592 +INFO:root: +[Valid][11][UTT][ MSE = 0.2627 | LCC = 0.8728 | SRCC = 0.8747 ] +INFO:root:Average loss=0.11299752186124142 +INFO:root: +[Valid][12][UTT][ MSE = 0.2172 | LCC = 0.8770 | SRCC = 0.8748 ] +INFO:root:Average loss=0.10866084599342102 +INFO:root: +[Valid][13][UTT][ MSE = 0.2183 | LCC = 0.8750 | SRCC = 0.8733 ] +INFO:root:Average loss=0.0996365602581929 +INFO:root: +[Valid][14][UTT][ MSE = 0.2041 | LCC = 0.8737 | SRCC = 0.8732 ] +INFO:root:Average loss=0.09614932460662647 +INFO:root: +[Valid][15][UTT][ MSE = 0.2202 | LCC = 0.8736 | SRCC = 0.8723 ] +INFO:root:Average loss=0.09051781649199817 +INFO:root: +[Valid][16][UTT][ MSE = 0.2221 | LCC = 0.8726 | SRCC = 0.8705 ] +INFO:root:Average loss=0.08748181723058224 +INFO:root: +[Valid][17][UTT][ MSE = 0.2172 | LCC = 0.8707 | SRCC = 0.8692 ] +INFO:root:Average loss=0.08426437493509208 +INFO:root: +[Valid][18][UTT][ MSE = 0.2165 | LCC = 0.8711 | SRCC = 0.8707 ] +INFO:root:Average loss=0.07784430248041947 +INFO:root: +[Valid][19][UTT][ MSE = 0.2395 | LCC = 0.8729 | SRCC = 0.8725 ] +INFO:root:Average loss=0.07790932665841702 +INFO:root: +[Valid][20][UTT][ MSE = 0.2300 | LCC = 0.8717 | SRCC = 0.8720 ] +INFO:root:Average loss=0.07237955512335667 +INFO:root: +[Valid][21][UTT][ MSE = 0.2154 | LCC = 0.8714 | SRCC = 0.8702 ] +INFO:root:Average loss=0.06789290456053539 +INFO:root: +[Valid][22][UTT][ MSE = 0.2251 | LCC = 0.8707 | SRCC = 0.8691 ] +INFO:root:Average loss=0.06880376220513613 +INFO:root: +[Valid][23][UTT][ MSE = 0.2141 | LCC = 0.8712 | SRCC = 0.8696 ] +INFO:root:Average loss=0.06691365593518966 +INFO:root: +[Valid][24][UTT][ MSE = 0.2138 | LCC = 0.8722 | SRCC = 0.8711 ] +INFO:root:Average loss=0.06475625468943363 +INFO:root: +[Valid][25][UTT][ MSE = 0.2226 | LCC = 0.8742 | SRCC = 0.8714 ] +INFO:root:Average loss=0.060989546470153026 +INFO:root: +[Valid][26][UTT][ MSE = 0.2135 | LCC = 0.8736 | SRCC = 0.8728 ] +INFO:root:Average loss=0.06058739713178231 +INFO:root: +[Valid][27][UTT][ MSE = 0.2199 | LCC = 0.8744 | SRCC = 0.8733 ] +INFO:root:Average loss=0.05712242410160028 +INFO:root: +[Valid][28][UTT][ MSE = 0.2250 | LCC = 0.8718 | SRCC = 0.8729 ] +INFO:root:Average loss=0.058669290147148646 +INFO:root: +[Valid][29][UTT][ MSE = 0.2137 | LCC = 0.8732 | SRCC = 0.8722 ] +INFO:root:Average loss=0.052133089647843286 +INFO:root: +[Valid][30][UTT][ MSE = 0.2232 | LCC = 0.8718 | SRCC = 0.8709 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2270 | LCC = 0.8670 | SRCC = 0.8678 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8c52f1bb9446352440e390cbc13e048b213567c8 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:990bee8cf82dd1fc7c706c023997d0c03696aee126dcd996a4c4de2b3915a3ef +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..40246d4d38fe780f76cd250eb0c7f013c3f4f29a --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:afa4364c4ef11618c9df308256b690db7c65eee5763b3f4ec9e79fdea90a3e46 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..324bae08492c7409a752fa054446178c47b4f5f4 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/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=0.3997827456929745 +INFO:root: +[Valid][0][UTT][ MSE = 0.2875 | LCC = 0.8222 | SRCC = 0.8294 ] +INFO:root:Average loss=0.2224613661185289 +INFO:root: +[Valid][1][UTT][ MSE = 0.2638 | LCC = 0.8536 | SRCC = 0.8556 ] +INFO:root:Average loss=0.2016164950835399 +INFO:root: +[Valid][2][UTT][ MSE = 0.2429 | LCC = 0.8595 | SRCC = 0.8609 ] +INFO:root:Average loss=0.18070614414337355 +INFO:root: +[Valid][3][UTT][ MSE = 0.2503 | LCC = 0.8692 | SRCC = 0.8670 ] +INFO:root:Average loss=0.1653951948078779 +INFO:root: +[Valid][4][UTT][ MSE = 0.2174 | LCC = 0.8712 | SRCC = 0.8712 ] +INFO:root:Average loss=0.15783441410614893 +INFO:root: +[Valid][5][UTT][ MSE = 0.2237 | LCC = 0.8712 | SRCC = 0.8703 ] +INFO:root:Average loss=0.1485423629100506 +INFO:root: +[Valid][6][UTT][ MSE = 0.2491 | LCC = 0.8658 | SRCC = 0.8690 ] +INFO:root:Average loss=0.14441857802180144 +INFO:root: +[Valid][7][UTT][ MSE = 0.2119 | LCC = 0.8709 | SRCC = 0.8732 ] +INFO:root:Average loss=0.13125511889274305 +INFO:root: +[Valid][8][UTT][ MSE = 0.2140 | LCC = 0.8754 | SRCC = 0.8742 ] +INFO:root:Average loss=0.13023759329166168 +INFO:root: +[Valid][9][UTT][ MSE = 0.2079 | LCC = 0.8716 | SRCC = 0.8734 ] +INFO:root:Average loss=0.1237821034513987 +INFO:root: +[Valid][10][UTT][ MSE = 0.2197 | LCC = 0.8742 | SRCC = 0.8739 ] +INFO:root:Average loss=0.11343929835428031 +INFO:root: +[Valid][11][UTT][ MSE = 0.2052 | LCC = 0.8714 | SRCC = 0.8721 ] +INFO:root:Average loss=0.10467363664737114 +INFO:root: +[Valid][12][UTT][ MSE = 0.2154 | LCC = 0.8697 | SRCC = 0.8692 ] +INFO:root:Average loss=0.10184905945490567 +INFO:root: +[Valid][13][UTT][ MSE = 0.2372 | LCC = 0.8690 | SRCC = 0.8672 ] +INFO:root:Average loss=0.09432135465053412 +INFO:root: +[Valid][14][UTT][ MSE = 0.2188 | LCC = 0.8663 | SRCC = 0.8670 ] +INFO:root:Average loss=0.09077877150132106 +INFO:root: +[Valid][15][UTT][ MSE = 0.2158 | LCC = 0.8672 | SRCC = 0.8669 ] +INFO:root:Average loss=0.08591071884028423 +INFO:root: +[Valid][16][UTT][ MSE = 0.2256 | LCC = 0.8688 | SRCC = 0.8673 ] +INFO:root:Average loss=0.0883284192532301 +INFO:root: +[Valid][17][UTT][ MSE = 0.2299 | LCC = 0.8661 | SRCC = 0.8673 ] +INFO:root:Average loss=0.08155904108515152 +INFO:root: +[Valid][18][UTT][ MSE = 0.2141 | LCC = 0.8675 | SRCC = 0.8675 ] +INFO:root:Average loss=0.07822417644545054 +INFO:root: +[Valid][19][UTT][ MSE = 0.2210 | LCC = 0.8685 | SRCC = 0.8670 ] +INFO:root:Average loss=0.07221397561713672 +INFO:root: +[Valid][20][UTT][ MSE = 0.2233 | LCC = 0.8664 | SRCC = 0.8653 ] +INFO:root:Average loss=0.07134366962007987 +INFO:root: +[Valid][21][UTT][ MSE = 0.2201 | LCC = 0.8647 | SRCC = 0.8664 ] +INFO:root:Average loss=0.0664091496131359 +INFO:root: +[Valid][22][UTT][ MSE = 0.2210 | LCC = 0.8648 | SRCC = 0.8644 ] +INFO:root:Average loss=0.06227574838946263 +INFO:root: +[Valid][23][UTT][ MSE = 0.2285 | LCC = 0.8631 | SRCC = 0.8633 ] +INFO:root:Average loss=0.05992810948727987 +INFO:root: +[Valid][24][UTT][ MSE = 0.2153 | LCC = 0.8672 | SRCC = 0.8669 ] +INFO:root:Average loss=0.05871230421157984 +INFO:root: +[Valid][25][UTT][ MSE = 0.2200 | LCC = 0.8655 | SRCC = 0.8655 ] +INFO:root:Average loss=0.06159316554952126 +INFO:root: +[Valid][26][UTT][ MSE = 0.2214 | LCC = 0.8667 | SRCC = 0.8646 ] +INFO:root:Average loss=0.053979346433129065 +INFO:root: +[Valid][27][UTT][ MSE = 0.2143 | LCC = 0.8679 | SRCC = 0.8669 ] +INFO:root:Average loss=0.058740780354501344 +INFO:root: +[Valid][28][UTT][ MSE = 0.2138 | LCC = 0.8696 | SRCC = 0.8688 ] +INFO:root:Average loss=0.05082789092109753 +INFO:root: +[Valid][29][UTT][ MSE = 0.2174 | LCC = 0.8664 | SRCC = 0.8655 ] +INFO:root:Average loss=0.048485805017825886 +INFO:root: +[Valid][30][UTT][ MSE = 0.2184 | LCC = 0.8676 | SRCC = 0.8672 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2204 | LCC = 0.8724 | SRCC = 0.8717 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4661c78916573a3bc7469468817dc89dc0de2382 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f615d0ae3316c0acbe87b590ab2d2ba58db7eee2f403a2e52633d3d3475c122d +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a5013a48085fd899401dc679c8b3a1479e400093 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c33305cb82a86defb899006d84af0da7b4d8e63416534e64cb7c90ef7d009cc +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0ce5bf1f8779c789a315901b4c37b134db07f76e --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/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=0.41279081178781313 +INFO:root: +[Valid][0][UTT][ MSE = 0.2914 | LCC = 0.8178 | SRCC = 0.8249 ] +INFO:root:Average loss=0.23485865272008455 +INFO:root: +[Valid][1][UTT][ MSE = 0.2569 | LCC = 0.8528 | SRCC = 0.8545 ] +INFO:root:Average loss=0.20550835438263723 +INFO:root: +[Valid][2][UTT][ MSE = 0.2585 | LCC = 0.8601 | SRCC = 0.8623 ] +INFO:root:Average loss=0.18909212650778967 +INFO:root: +[Valid][3][UTT][ MSE = 0.2318 | LCC = 0.8655 | SRCC = 0.8652 ] +INFO:root:Average loss=0.17555323329109412 +INFO:root: +[Valid][4][UTT][ MSE = 0.2227 | LCC = 0.8694 | SRCC = 0.8701 ] +INFO:root:Average loss=0.16886752805648705 +INFO:root: +[Valid][5][UTT][ MSE = 0.2402 | LCC = 0.8691 | SRCC = 0.8722 ] +INFO:root:Average loss=0.15659662804160362 +INFO:root: +[Valid][6][UTT][ MSE = 0.2135 | LCC = 0.8742 | SRCC = 0.8738 ] +INFO:root:Average loss=0.15044130814763215 +INFO:root: +[Valid][7][UTT][ MSE = 0.2275 | LCC = 0.8710 | SRCC = 0.8729 ] +INFO:root:Average loss=0.14117358739559466 +INFO:root: +[Valid][8][UTT][ MSE = 0.2106 | LCC = 0.8773 | SRCC = 0.8770 ] +INFO:root:Average loss=0.12773076139199427 +INFO:root: +[Valid][9][UTT][ MSE = 0.2162 | LCC = 0.8713 | SRCC = 0.8739 ] +INFO:root:Average loss=0.13365286569564772 +INFO:root: +[Valid][10][UTT][ MSE = 0.2307 | LCC = 0.8704 | SRCC = 0.8722 ] +INFO:root:Average loss=0.11729115658463576 +INFO:root: +[Valid][11][UTT][ MSE = 0.2110 | LCC = 0.8722 | SRCC = 0.8734 ] +INFO:root:Average loss=0.11072220357182698 +INFO:root: +[Valid][12][UTT][ MSE = 0.2222 | LCC = 0.8712 | SRCC = 0.8716 ] +INFO:root:Average loss=0.10019177270050232 +INFO:root: +[Valid][13][UTT][ MSE = 0.2259 | LCC = 0.8725 | SRCC = 0.8723 ] +INFO:root:Average loss=0.10057638337214787 +INFO:root: +[Valid][14][UTT][ MSE = 0.2108 | LCC = 0.8682 | SRCC = 0.8711 ] +INFO:root:Average loss=0.09041582611508858 +INFO:root: +[Valid][15][UTT][ MSE = 0.2236 | LCC = 0.8675 | SRCC = 0.8684 ] +INFO:root:Average loss=0.0877060032903384 +INFO:root: +[Valid][16][UTT][ MSE = 0.2137 | LCC = 0.8692 | SRCC = 0.8701 ] +INFO:root:Average loss=0.08308262697970256 +INFO:root: +[Valid][17][UTT][ MSE = 0.2150 | LCC = 0.8660 | SRCC = 0.8672 ] +INFO:root:Average loss=0.07990660093342647 +INFO:root: +[Valid][18][UTT][ MSE = 0.2194 | LCC = 0.8665 | SRCC = 0.8672 ] +INFO:root:Average loss=0.0740860103128048 +INFO:root: +[Valid][19][UTT][ MSE = 0.2056 | LCC = 0.8689 | SRCC = 0.8701 ] +INFO:root:Average loss=0.07093427592936234 +INFO:root: +[Valid][20][UTT][ MSE = 0.2231 | LCC = 0.8650 | SRCC = 0.8654 ] +INFO:root:Average loss=0.07494398865562218 +INFO:root: +[Valid][21][UTT][ MSE = 0.2278 | LCC = 0.8681 | SRCC = 0.8702 ] +INFO:root:Average loss=0.06586341402278496 +INFO:root: +[Valid][22][UTT][ MSE = 0.2175 | LCC = 0.8673 | SRCC = 0.8668 ] +INFO:root:Average loss=0.06286342227115081 +INFO:root: +[Valid][23][UTT][ MSE = 0.2078 | LCC = 0.8673 | SRCC = 0.8676 ] +INFO:root:Average loss=0.0624709415416687 +INFO:root: +[Valid][24][UTT][ MSE = 0.2225 | LCC = 0.8640 | SRCC = 0.8635 ] +INFO:root:Average loss=0.0595596973808148 +INFO:root: +[Valid][25][UTT][ MSE = 0.2259 | LCC = 0.8636 | SRCC = 0.8640 ] +INFO:root:Average loss=0.06088248793131266 +INFO:root: +[Valid][26][UTT][ MSE = 0.2166 | LCC = 0.8655 | SRCC = 0.8652 ] +INFO:root:Average loss=0.056731962718260594 +INFO:root: +[Valid][27][UTT][ MSE = 0.2224 | LCC = 0.8660 | SRCC = 0.8649 ] +INFO:root:Average loss=0.05322434643331247 +INFO:root: +[Valid][28][UTT][ MSE = 0.2112 | LCC = 0.8666 | SRCC = 0.8664 ] +INFO:root:Average loss=0.05480873804443922 +INFO:root: +[Valid][29][UTT][ MSE = 0.2197 | LCC = 0.8663 | SRCC = 0.8651 ] +INFO:root:Average loss=0.05122745896761234 +INFO:root: +[Valid][30][UTT][ MSE = 0.2145 | LCC = 0.8673 | SRCC = 0.8664 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2415 | LCC = 0.8674 | SRCC = 0.8691 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7d535ba9126d375b4978a398037bbcda68218d98 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:427bbc940e830177145303b4bacec74b5957688b5df83ba35ce44b393f804d78 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4123a2569d587f30962205fe50992104d4c8ccdf --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:86ddd2aed266daac9a408ea6b549bd6e00954981249519524c74c619ccbd9785 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..8879e4bf8a47d9791bed6050413559210b5ec35f --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer2/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=0.4251187453285242 +INFO:root: +[Valid][0][UTT][ MSE = 0.2884 | LCC = 0.8219 | SRCC = 0.8294 ] +INFO:root:Average loss=0.23859414840355897 +INFO:root: +[Valid][1][UTT][ MSE = 0.2607 | LCC = 0.8502 | SRCC = 0.8531 ] +INFO:root:Average loss=0.2091229916192018 +INFO:root: +[Valid][2][UTT][ MSE = 0.2378 | LCC = 0.8602 | SRCC = 0.8619 ] +INFO:root:Average loss=0.193088483160887 +INFO:root: +[Valid][3][UTT][ MSE = 0.2390 | LCC = 0.8673 | SRCC = 0.8665 ] +INFO:root:Average loss=0.17249932713233507 +INFO:root: +[Valid][4][UTT][ MSE = 0.2290 | LCC = 0.8738 | SRCC = 0.8723 ] +INFO:root:Average loss=0.16408894688655168 +INFO:root: +[Valid][5][UTT][ MSE = 0.2310 | LCC = 0.8754 | SRCC = 0.8740 ] +INFO:root:Average loss=0.15443196243200547 +INFO:root: +[Valid][6][UTT][ MSE = 0.2246 | LCC = 0.8739 | SRCC = 0.8724 ] +INFO:root:Average loss=0.1412004929704544 +INFO:root: +[Valid][7][UTT][ MSE = 0.2221 | LCC = 0.8698 | SRCC = 0.8713 ] +INFO:root:Average loss=0.1345912710023232 +INFO:root: +[Valid][8][UTT][ MSE = 0.2203 | LCC = 0.8699 | SRCC = 0.8693 ] +INFO:root:Average loss=0.12103556669675387 +INFO:root: +[Valid][9][UTT][ MSE = 0.2123 | LCC = 0.8719 | SRCC = 0.8723 ] +INFO:root:Average loss=0.11560756951952592 +INFO:root: +[Valid][10][UTT][ MSE = 0.2172 | LCC = 0.8773 | SRCC = 0.8767 ] +INFO:root:Average loss=0.12119797531228799 +INFO:root: +[Valid][11][UTT][ MSE = 0.2241 | LCC = 0.8732 | SRCC = 0.8731 ] +INFO:root:Average loss=0.1080723443092444 +INFO:root: +[Valid][12][UTT][ MSE = 0.2125 | LCC = 0.8741 | SRCC = 0.8734 ] +INFO:root:Average loss=0.1012131620485049 +INFO:root: +[Valid][13][UTT][ MSE = 0.2424 | LCC = 0.8714 | SRCC = 0.8717 ] +INFO:root:Average loss=0.09329690325718659 +INFO:root: +[Valid][14][UTT][ MSE = 0.2276 | LCC = 0.8739 | SRCC = 0.8740 ] +INFO:root:Average loss=0.09376056721577278 +INFO:root: +[Valid][15][UTT][ MSE = 0.2132 | LCC = 0.8735 | SRCC = 0.8733 ] +INFO:root:Average loss=0.09141857420595792 +INFO:root: +[Valid][16][UTT][ MSE = 0.2047 | LCC = 0.8720 | SRCC = 0.8729 ] +INFO:root:Average loss=0.08347902384897073 +INFO:root: +[Valid][17][UTT][ MSE = 0.2235 | LCC = 0.8712 | SRCC = 0.8711 ] +INFO:root:Average loss=0.07752436953477371 +INFO:root: +[Valid][18][UTT][ MSE = 0.2170 | LCC = 0.8706 | SRCC = 0.8699 ] +INFO:root:Average loss=0.0754955638295565 +INFO:root: +[Valid][19][UTT][ MSE = 0.2118 | LCC = 0.8700 | SRCC = 0.8695 ] +INFO:root:Average loss=0.07094017010277663 +INFO:root: +[Valid][20][UTT][ MSE = 0.2146 | LCC = 0.8675 | SRCC = 0.8681 ] +INFO:root:Average loss=0.07294006870152095 +INFO:root: +[Valid][21][UTT][ MSE = 0.2119 | LCC = 0.8713 | SRCC = 0.8712 ] +INFO:root:Average loss=0.06754512927280022 +INFO:root: +[Valid][22][UTT][ MSE = 0.2184 | LCC = 0.8662 | SRCC = 0.8664 ] +INFO:root:Average loss=0.06301354678968589 +INFO:root: +[Valid][23][UTT][ MSE = 0.2133 | LCC = 0.8670 | SRCC = 0.8667 ] +INFO:root:Average loss=0.06317993845695104 +INFO:root: +[Valid][24][UTT][ MSE = 0.2245 | LCC = 0.8636 | SRCC = 0.8668 ] +INFO:root:Average loss=0.06329992809929909 +INFO:root: +[Valid][25][UTT][ MSE = 0.2197 | LCC = 0.8703 | SRCC = 0.8703 ] +INFO:root:Average loss=0.056796218101412825 +INFO:root: +[Valid][26][UTT][ MSE = 0.2100 | LCC = 0.8694 | SRCC = 0.8696 ] +INFO:root:Average loss=0.05270393665593404 +INFO:root: +[Valid][27][UTT][ MSE = 0.2110 | LCC = 0.8674 | SRCC = 0.8680 ] +INFO:root:Average loss=0.05342982771495978 +INFO:root: +[Valid][28][UTT][ MSE = 0.2077 | LCC = 0.8705 | SRCC = 0.8700 ] +INFO:root:Average loss=0.05555527545034121 +INFO:root: +[Valid][29][UTT][ MSE = 0.2305 | LCC = 0.8719 | SRCC = 0.8715 ] +INFO:root:Average loss=0.04760929383337498 +INFO:root: +[Valid][30][UTT][ MSE = 0.2308 | LCC = 0.8708 | SRCC = 0.8700 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2563 | LCC = 0.8641 | SRCC = 0.8642 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e48aaf4c1507a12d92b930eb3b1297d56c231086 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cfb0311493a342eedba3a90cb3e2f8b3ddf30637771ce62af18ee8fb08dd0ac0 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b097a16b483d0eaf583e7106ebc56e0a1bc5dd3c --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2eb6b0def4ad9ebc8d37246da45931c4a9ec820a9141e413d2cf2d0235cd9dd6 +size 2749638 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..91a3cc5cc7b898e5c5097a7d0a70c6673e8bb163 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/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=0.40118071838067126 +INFO:root: +[Valid][0][UTT][ MSE = 0.3725 | LCC = 0.8154 | SRCC = 0.8295 ] +INFO:root:Average loss=0.23140481935861784 +INFO:root: +[Valid][1][UTT][ MSE = 0.2870 | LCC = 0.8565 | SRCC = 0.8538 ] +INFO:root:Average loss=0.2035154582789311 +INFO:root: +[Valid][2][UTT][ MSE = 0.2784 | LCC = 0.8620 | SRCC = 0.8604 ] +INFO:root:Average loss=0.18285511338557953 +INFO:root: +[Valid][3][UTT][ MSE = 0.2761 | LCC = 0.8641 | SRCC = 0.8636 ] +INFO:root:Average loss=0.1741818276544412 +INFO:root: +[Valid][4][UTT][ MSE = 0.2408 | LCC = 0.8736 | SRCC = 0.8716 ] +INFO:root:Average loss=0.161564718072231 +INFO:root: +[Valid][5][UTT][ MSE = 0.2734 | LCC = 0.8688 | SRCC = 0.8673 ] +INFO:root:Average loss=0.14657480928760308 +INFO:root: +[Valid][6][UTT][ MSE = 0.2907 | LCC = 0.8675 | SRCC = 0.8678 ] +INFO:root:Average loss=0.13502229902988824 +INFO:root: +[Valid][7][UTT][ MSE = 0.2247 | LCC = 0.8712 | SRCC = 0.8693 ] +INFO:root:Average loss=0.1301136267108795 +INFO:root: +[Valid][8][UTT][ MSE = 0.2657 | LCC = 0.8671 | SRCC = 0.8668 ] +INFO:root:Average loss=0.12521331594922605 +INFO:root: +[Valid][9][UTT][ MSE = 0.2912 | LCC = 0.8609 | SRCC = 0.8619 ] +INFO:root:Average loss=0.11774486733170655 +INFO:root: +[Valid][10][UTT][ MSE = 0.2337 | LCC = 0.8660 | SRCC = 0.8662 ] +INFO:root:Average loss=0.10804048610421327 +INFO:root: +[Valid][11][UTT][ MSE = 0.2614 | LCC = 0.8652 | SRCC = 0.8645 ] +INFO:root:Average loss=0.1107531033265285 +INFO:root: +[Valid][12][UTT][ MSE = 0.2581 | LCC = 0.8647 | SRCC = 0.8630 ] +INFO:root:Average loss=0.09555720227460067 +INFO:root: +[Valid][13][UTT][ MSE = 0.2386 | LCC = 0.8617 | SRCC = 0.8615 ] +INFO:root:Average loss=0.09509992962464309 +INFO:root: +[Valid][14][UTT][ MSE = 0.2596 | LCC = 0.8625 | SRCC = 0.8617 ] +INFO:root:Average loss=0.09518847418710208 +INFO:root: +[Valid][15][UTT][ MSE = 0.2832 | LCC = 0.8584 | SRCC = 0.8568 ] +INFO:root:Average loss=0.08312136718095878 +INFO:root: +[Valid][16][UTT][ MSE = 0.2650 | LCC = 0.8602 | SRCC = 0.8584 ] +INFO:root:Average loss=0.08416180092936908 +INFO:root: +[Valid][17][UTT][ MSE = 0.2640 | LCC = 0.8604 | SRCC = 0.8594 ] +INFO:root:Average loss=0.0779145943144193 +INFO:root: +[Valid][18][UTT][ MSE = 0.3039 | LCC = 0.8599 | SRCC = 0.8594 ] +INFO:root:Average loss=0.07276589838931194 +INFO:root: +[Valid][19][UTT][ MSE = 0.2447 | LCC = 0.8637 | SRCC = 0.8624 ] +INFO:root:Average loss=0.07462622196628498 +INFO:root: +[Valid][20][UTT][ MSE = 0.2832 | LCC = 0.8596 | SRCC = 0.8601 ] +INFO:root:Average loss=0.06725372756138825 +INFO:root: +[Valid][21][UTT][ MSE = 0.2681 | LCC = 0.8598 | SRCC = 0.8583 ] +INFO:root:Average loss=0.06442850823394763 +INFO:root: +[Valid][22][UTT][ MSE = 0.2469 | LCC = 0.8621 | SRCC = 0.8599 ] +INFO:root:Average loss=0.06097599782813818 +INFO:root: +[Valid][23][UTT][ MSE = 0.2527 | LCC = 0.8596 | SRCC = 0.8588 ] +INFO:root:Average loss=0.05818404811315048 +INFO:root: +[Valid][24][UTT][ MSE = 0.2803 | LCC = 0.8573 | SRCC = 0.8563 ] +INFO:root:Average loss=0.0618847651550403 +INFO:root: +[Valid][25][UTT][ MSE = 0.2337 | LCC = 0.8602 | SRCC = 0.8590 ] +INFO:root:Average loss=0.05852745210704131 +INFO:root: +[Valid][26][UTT][ MSE = 0.2887 | LCC = 0.8588 | SRCC = 0.8596 ] +INFO:root:Average loss=0.05249256487840261 +INFO:root: +[Valid][27][UTT][ MSE = 0.2402 | LCC = 0.8602 | SRCC = 0.8591 ] +INFO:root:Average loss=0.0499483574516116 +INFO:root: +[Valid][28][UTT][ MSE = 0.2981 | LCC = 0.8582 | SRCC = 0.8595 ] +INFO:root:Average loss=0.052667513298682675 +INFO:root: +[Valid][29][UTT][ MSE = 0.2536 | LCC = 0.8622 | SRCC = 0.8621 ] +INFO:root:Average loss=0.049637572218974434 +INFO:root: +[Valid][30][UTT][ MSE = 0.2493 | LCC = 0.8617 | SRCC = 0.8619 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2252 | LCC = 0.8750 | SRCC = 0.8757 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e218a29ce200ae81dedbdf8f518ec2e9e0de9ebf --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8446f9e3578a42e237ec36ab9658063f533a8a8dead3f6b77d83260a44cc8d58 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a07e9a7598130c0689c7095459a6b6b2a3e12fc9 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:46a2cf4e4079c6ed59fa8dc269bfcb1eda29cb82a84a53805e4bd11ebc0043cb +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c6a7dcd606df60666bb8ca6ad205b9e702197978 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/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=0.39526093980440724 +INFO:root: +[Valid][0][UTT][ MSE = 0.3230 | LCC = 0.8323 | SRCC = 0.8365 ] +INFO:root:Average loss=0.22008343289295831 +INFO:root: +[Valid][1][UTT][ MSE = 0.2656 | LCC = 0.8587 | SRCC = 0.8574 ] +INFO:root:Average loss=0.19370023810710663 +INFO:root: +[Valid][2][UTT][ MSE = 0.2394 | LCC = 0.8705 | SRCC = 0.8679 ] +INFO:root:Average loss=0.17828997367849717 +INFO:root: +[Valid][3][UTT][ MSE = 0.2660 | LCC = 0.8718 | SRCC = 0.8690 ] +INFO:root:Average loss=0.16998623407039887 +INFO:root: +[Valid][4][UTT][ MSE = 0.2294 | LCC = 0.8706 | SRCC = 0.8681 ] +INFO:root:Average loss=0.15825005143116683 +INFO:root: +[Valid][5][UTT][ MSE = 0.2471 | LCC = 0.8757 | SRCC = 0.8729 ] +INFO:root:Average loss=0.1506192123469634 +INFO:root: +[Valid][6][UTT][ MSE = 0.2338 | LCC = 0.8758 | SRCC = 0.8739 ] +INFO:root:Average loss=0.13730382680510864 +INFO:root: +[Valid][7][UTT][ MSE = 0.2347 | LCC = 0.8700 | SRCC = 0.8690 ] +INFO:root:Average loss=0.13988425133702082 +INFO:root: +[Valid][8][UTT][ MSE = 0.2563 | LCC = 0.8720 | SRCC = 0.8729 ] +INFO:root:Average loss=0.12482770446401376 +INFO:root: +[Valid][9][UTT][ MSE = 0.2327 | LCC = 0.8702 | SRCC = 0.8676 ] +INFO:root:Average loss=0.11796962164151363 +INFO:root: +[Valid][10][UTT][ MSE = 0.2218 | LCC = 0.8706 | SRCC = 0.8692 ] +INFO:root:Average loss=0.11141344943107703 +INFO:root: +[Valid][11][UTT][ MSE = 0.2348 | LCC = 0.8712 | SRCC = 0.8701 ] +INFO:root:Average loss=0.10409993301026332 +INFO:root: +[Valid][12][UTT][ MSE = 0.2185 | LCC = 0.8691 | SRCC = 0.8684 ] +INFO:root:Average loss=0.09982156132658322 +INFO:root: +[Valid][13][UTT][ MSE = 0.2298 | LCC = 0.8701 | SRCC = 0.8690 ] +INFO:root:Average loss=0.09408930469399844 +INFO:root: +[Valid][14][UTT][ MSE = 0.2655 | LCC = 0.8688 | SRCC = 0.8685 ] +INFO:root:Average loss=0.09511247219947669 +INFO:root: +[Valid][15][UTT][ MSE = 0.2230 | LCC = 0.8659 | SRCC = 0.8645 ] +INFO:root:Average loss=0.08501430753714 +INFO:root: +[Valid][16][UTT][ MSE = 0.2318 | LCC = 0.8673 | SRCC = 0.8653 ] +INFO:root:Average loss=0.08107466255434048 +INFO:root: +[Valid][17][UTT][ MSE = 0.2409 | LCC = 0.8652 | SRCC = 0.8642 ] +INFO:root:Average loss=0.09437653088034728 +INFO:root: +[Valid][18][UTT][ MSE = 0.3110 | LCC = 0.8638 | SRCC = 0.8651 ] +INFO:root:Average loss=0.07526444385831173 +INFO:root: +[Valid][19][UTT][ MSE = 0.2462 | LCC = 0.8642 | SRCC = 0.8632 ] +INFO:root:Average loss=0.07282389528476275 +INFO:root: +[Valid][20][UTT][ MSE = 0.2348 | LCC = 0.8626 | SRCC = 0.8618 ] +INFO:root:Average loss=0.06776874722578587 +INFO:root: +[Valid][21][UTT][ MSE = 0.2428 | LCC = 0.8673 | SRCC = 0.8666 ] +INFO:root:Average loss=0.06378505570002091 +INFO:root: +[Valid][22][UTT][ MSE = 0.2519 | LCC = 0.8617 | SRCC = 0.8598 ] +INFO:root:Average loss=0.06252320617055282 +INFO:root: +[Valid][23][UTT][ MSE = 0.2522 | LCC = 0.8625 | SRCC = 0.8609 ] +INFO:root:Average loss=0.06303991511082038 +INFO:root: +[Valid][24][UTT][ MSE = 0.2660 | LCC = 0.8612 | SRCC = 0.8614 ] +INFO:root:Average loss=0.058082286411753066 +INFO:root: +[Valid][25][UTT][ MSE = 0.2306 | LCC = 0.8655 | SRCC = 0.8645 ] +INFO:root:Average loss=0.05515077041509824 +INFO:root: +[Valid][26][UTT][ MSE = 0.2243 | LCC = 0.8636 | SRCC = 0.8624 ] +INFO:root:Average loss=0.05446696290985132 +INFO:root: +[Valid][27][UTT][ MSE = 0.2353 | LCC = 0.8604 | SRCC = 0.8586 ] +INFO:root:Average loss=0.05033818465203811 +INFO:root: +[Valid][28][UTT][ MSE = 0.2524 | LCC = 0.8600 | SRCC = 0.8584 ] +INFO:root:Average loss=0.05240022118848104 +INFO:root: +[Valid][29][UTT][ MSE = 0.2480 | LCC = 0.8614 | SRCC = 0.8624 ] +INFO:root:Average loss=0.050213645570553266 +INFO:root: +[Valid][30][UTT][ MSE = 0.2347 | LCC = 0.8604 | SRCC = 0.8602 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2288 | LCC = 0.8726 | SRCC = 0.8728 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f2c99444dcc8671cd3aec6513fe92f8f80b2ea85 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:affc8e491af4ea790313323ef9895e85de9e4db0c66c0e2cf714e7878b4af257 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a525ca6e00a2ccf20d8c96cd0193e57c86e4c267 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aa91b4d6df7ae92e589ebb5dc31ab41406d98db3bd1be7929f6f506d5996ec9d +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..18705268bac5998db4d94ea3072d2414013277c2 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/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=0.5024450274232106 +INFO:root: +[Valid][0][UTT][ MSE = 0.3371 | LCC = 0.8159 | SRCC = 0.8249 ] +INFO:root:Average loss=0.23574041040279928 +INFO:root: +[Valid][1][UTT][ MSE = 0.2530 | LCC = 0.8503 | SRCC = 0.8535 ] +INFO:root:Average loss=0.20422680389422637 +INFO:root: +[Valid][2][UTT][ MSE = 0.2427 | LCC = 0.8643 | SRCC = 0.8630 ] +INFO:root:Average loss=0.18158734293702322 +INFO:root: +[Valid][3][UTT][ MSE = 0.2623 | LCC = 0.8696 | SRCC = 0.8689 ] +INFO:root:Average loss=0.17116506235339704 +INFO:root: +[Valid][4][UTT][ MSE = 0.2419 | LCC = 0.8697 | SRCC = 0.8691 ] +INFO:root:Average loss=0.1552977647918921 +INFO:root: +[Valid][5][UTT][ MSE = 0.2223 | LCC = 0.8734 | SRCC = 0.8735 ] +INFO:root:Average loss=0.1497966501957331 +INFO:root: +[Valid][6][UTT][ MSE = 0.2465 | LCC = 0.8718 | SRCC = 0.8725 ] +INFO:root:Average loss=0.13697268708776206 +INFO:root: +[Valid][7][UTT][ MSE = 0.2374 | LCC = 0.8744 | SRCC = 0.8725 ] +INFO:root:Average loss=0.12340728408442093 +INFO:root: +[Valid][8][UTT][ MSE = 0.2332 | LCC = 0.8727 | SRCC = 0.8731 ] +INFO:root:Average loss=0.12134550932126167 +INFO:root: +[Valid][9][UTT][ MSE = 0.2261 | LCC = 0.8726 | SRCC = 0.8733 ] +INFO:root:Average loss=0.10963514600044642 +INFO:root: +[Valid][10][UTT][ MSE = 0.2405 | LCC = 0.8695 | SRCC = 0.8707 ] +INFO:root:Average loss=0.10599954817921688 +INFO:root: +[Valid][11][UTT][ MSE = 0.2352 | LCC = 0.8700 | SRCC = 0.8724 ] +INFO:root:Average loss=0.09778048069431232 +INFO:root: +[Valid][12][UTT][ MSE = 0.2385 | LCC = 0.8686 | SRCC = 0.8665 ] +INFO:root:Average loss=0.09478043096187787 +INFO:root: +[Valid][13][UTT][ MSE = 0.2502 | LCC = 0.8700 | SRCC = 0.8704 ] +INFO:root:Average loss=0.08963746591829337 +INFO:root: +[Valid][14][UTT][ MSE = 0.3096 | LCC = 0.8647 | SRCC = 0.8654 ] +INFO:root:Average loss=0.09014999236051853 +INFO:root: +[Valid][15][UTT][ MSE = 0.2472 | LCC = 0.8662 | SRCC = 0.8664 ] +INFO:root:Average loss=0.08145293698478968 +INFO:root: +[Valid][16][UTT][ MSE = 0.2643 | LCC = 0.8652 | SRCC = 0.8650 ] +INFO:root:Average loss=0.07584134317361392 +INFO:root: +[Valid][17][UTT][ MSE = 0.2410 | LCC = 0.8652 | SRCC = 0.8656 ] +INFO:root:Average loss=0.07070707458143051 +INFO:root: +[Valid][18][UTT][ MSE = 0.2465 | LCC = 0.8661 | SRCC = 0.8658 ] +INFO:root:Average loss=0.0692152586311866 +INFO:root: +[Valid][19][UTT][ MSE = 0.2198 | LCC = 0.8670 | SRCC = 0.8673 ] +INFO:root:Average loss=0.06249605527577492 +INFO:root: +[Valid][20][UTT][ MSE = 0.2717 | LCC = 0.8663 | SRCC = 0.8684 ] +INFO:root:Average loss=0.05892562216673142 +INFO:root: +[Valid][21][UTT][ MSE = 0.2261 | LCC = 0.8665 | SRCC = 0.8660 ] +INFO:root:Average loss=0.06155821012380796 +INFO:root: +[Valid][22][UTT][ MSE = 0.2516 | LCC = 0.8648 | SRCC = 0.8663 ] +INFO:root:Average loss=0.05725243983742518 +INFO:root: +[Valid][23][UTT][ MSE = 0.2278 | LCC = 0.8648 | SRCC = 0.8648 ] +INFO:root:Average loss=0.05375116944122009 +INFO:root: +[Valid][24][UTT][ MSE = 0.2865 | LCC = 0.8595 | SRCC = 0.8617 ] +INFO:root:Average loss=0.05137403863362777 +INFO:root: +[Valid][25][UTT][ MSE = 0.2433 | LCC = 0.8641 | SRCC = 0.8634 ] +INFO:root:Average loss=0.05077397503340856 +INFO:root: +[Valid][26][UTT][ MSE = 0.2305 | LCC = 0.8620 | SRCC = 0.8616 ] +INFO:root:Average loss=0.05042035502787584 +INFO:root: +[Valid][27][UTT][ MSE = 0.2458 | LCC = 0.8613 | SRCC = 0.8634 ] +INFO:root:Average loss=0.046247853754231565 +INFO:root: +[Valid][28][UTT][ MSE = 0.2411 | LCC = 0.8628 | SRCC = 0.8632 ] +INFO:root:Average loss=0.04364005777125175 +INFO:root: +[Valid][29][UTT][ MSE = 0.2440 | LCC = 0.8631 | SRCC = 0.8621 ] +INFO:root:Average loss=0.045386010446609594 +INFO:root: +[Valid][30][UTT][ MSE = 0.2587 | LCC = 0.8649 | SRCC = 0.8659 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2364 | LCC = 0.8679 | SRCC = 0.8675 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c151d592049f4a23ca26faddcab92957e17d9671 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1ba8262341cd4834703ed9e08e54dc889e20f7be030fd355b1929bcb2f83014e +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6d20e913ab3645337f34beb3b46360e98bce9963 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e89190ec99fca8ba426e11ed1114b347e9359fecd05e8dacf7decd07c7ad647 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..fd047e21f395fb23866333b0470509f5693e6ff4 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/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=0.4293087730422998 +INFO:root: +[Valid][0][UTT][ MSE = 0.3158 | LCC = 0.8208 | SRCC = 0.8291 ] +INFO:root:Average loss=0.23242939722079498 +INFO:root: +[Valid][1][UTT][ MSE = 0.2704 | LCC = 0.8505 | SRCC = 0.8542 ] +INFO:root:Average loss=0.20306410716894346 +INFO:root: +[Valid][2][UTT][ MSE = 0.2346 | LCC = 0.8638 | SRCC = 0.8636 ] +INFO:root:Average loss=0.1896814861549781 +INFO:root: +[Valid][3][UTT][ MSE = 0.2244 | LCC = 0.8712 | SRCC = 0.8712 ] +INFO:root:Average loss=0.1778171276434874 +INFO:root: +[Valid][4][UTT][ MSE = 0.2159 | LCC = 0.8736 | SRCC = 0.8740 ] +INFO:root:Average loss=0.1601995057784594 +INFO:root: +[Valid][5][UTT][ MSE = 0.2309 | LCC = 0.8720 | SRCC = 0.8719 ] +INFO:root:Average loss=0.1485131432612737 +INFO:root: +[Valid][6][UTT][ MSE = 0.2152 | LCC = 0.8751 | SRCC = 0.8735 ] +INFO:root:Average loss=0.14302474919420022 +INFO:root: +[Valid][7][UTT][ MSE = 0.2204 | LCC = 0.8754 | SRCC = 0.8749 ] +INFO:root:Average loss=0.13017404079437256 +INFO:root: +[Valid][8][UTT][ MSE = 0.2115 | LCC = 0.8752 | SRCC = 0.8746 ] +INFO:root:Average loss=0.12560147180771217 +INFO:root: +[Valid][9][UTT][ MSE = 0.2142 | LCC = 0.8748 | SRCC = 0.8744 ] +INFO:root:Average loss=0.11206755729822013 +INFO:root: +[Valid][10][UTT][ MSE = 0.2130 | LCC = 0.8673 | SRCC = 0.8696 ] +INFO:root:Average loss=0.10892442699808341 +INFO:root: +[Valid][11][UTT][ MSE = 0.2082 | LCC = 0.8686 | SRCC = 0.8702 ] +INFO:root:Average loss=0.10202286354242227 +INFO:root: +[Valid][12][UTT][ MSE = 0.2080 | LCC = 0.8690 | SRCC = 0.8692 ] +INFO:root:Average loss=0.09450924625763527 +INFO:root: +[Valid][13][UTT][ MSE = 0.2116 | LCC = 0.8681 | SRCC = 0.8695 ] +INFO:root:Average loss=0.09340116897454628 +INFO:root: +[Valid][14][UTT][ MSE = 0.2380 | LCC = 0.8694 | SRCC = 0.8685 ] +INFO:root:Average loss=0.09209802326483604 +INFO:root: +[Valid][15][UTT][ MSE = 0.2205 | LCC = 0.8673 | SRCC = 0.8667 ] +INFO:root:Average loss=0.08095935904062711 +INFO:root: +[Valid][16][UTT][ MSE = 0.2180 | LCC = 0.8675 | SRCC = 0.8683 ] +INFO:root:Average loss=0.07623278426054196 +INFO:root: +[Valid][17][UTT][ MSE = 0.2102 | LCC = 0.8657 | SRCC = 0.8666 ] +INFO:root:Average loss=0.07289596962241027 +INFO:root: +[Valid][18][UTT][ MSE = 0.2119 | LCC = 0.8676 | SRCC = 0.8676 ] +INFO:root:Average loss=0.0695502625252956 +INFO:root: +[Valid][19][UTT][ MSE = 0.2222 | LCC = 0.8642 | SRCC = 0.8653 ] +INFO:root:Average loss=0.06732951329113582 +INFO:root: +[Valid][20][UTT][ MSE = 0.2250 | LCC = 0.8660 | SRCC = 0.8662 ] +INFO:root:Average loss=0.06399268141159645 +INFO:root: +[Valid][21][UTT][ MSE = 0.2182 | LCC = 0.8645 | SRCC = 0.8639 ] +INFO:root:Average loss=0.061854612989685476 +INFO:root: +[Valid][22][UTT][ MSE = 0.2211 | LCC = 0.8647 | SRCC = 0.8652 ] +INFO:root:Average loss=0.05990061822992105 +INFO:root: +[Valid][23][UTT][ MSE = 0.2320 | LCC = 0.8590 | SRCC = 0.8590 ] +INFO:root:Average loss=0.061777617113712505 +INFO:root: +[Valid][24][UTT][ MSE = 0.2310 | LCC = 0.8651 | SRCC = 0.8656 ] +INFO:root:Average loss=0.05495775089814113 +INFO:root: +[Valid][25][UTT][ MSE = 0.2193 | LCC = 0.8654 | SRCC = 0.8665 ] +INFO:root:Average loss=0.05392577962424511 +INFO:root: +[Valid][26][UTT][ MSE = 0.2116 | LCC = 0.8653 | SRCC = 0.8669 ] +INFO:root:Average loss=0.05223704635714873 +INFO:root: +[Valid][27][UTT][ MSE = 0.2306 | LCC = 0.8628 | SRCC = 0.8633 ] +INFO:root:Average loss=0.052556678222922176 +INFO:root: +[Valid][28][UTT][ MSE = 0.2165 | LCC = 0.8644 | SRCC = 0.8642 ] +INFO:root:Average loss=0.05005755691001049 +INFO:root: +[Valid][29][UTT][ MSE = 0.2169 | LCC = 0.8625 | SRCC = 0.8623 ] +INFO:root:Average loss=0.04892897567687891 +INFO:root: +[Valid][30][UTT][ MSE = 0.2162 | LCC = 0.8631 | SRCC = 0.8634 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2452 | LCC = 0.8689 | SRCC = 0.8719 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d84574470558cdf88b1047f6a213c993ad8a3d05 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9da2ebf532606862e218982abb67c4be3f7a654b441b976420ffe1e538b8d709 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..14c86c8e1ada40edcac832151a02f5c3e39425f4 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d89ebfc58a3beae715f1adcdf357762d03318c1e37fd7b367086a300b0e202b2 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b0e86f35bfc42810b56e0c2609ef51575d10cd70 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer3/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=0.46167608063954574 +INFO:root: +[Valid][0][UTT][ MSE = 0.3094 | LCC = 0.8079 | SRCC = 0.8154 ] +INFO:root:Average loss=0.231427666086417 +INFO:root: +[Valid][1][UTT][ MSE = 0.2703 | LCC = 0.8552 | SRCC = 0.8550 ] +INFO:root:Average loss=0.19970792054365843 +INFO:root: +[Valid][2][UTT][ MSE = 0.2429 | LCC = 0.8677 | SRCC = 0.8666 ] +INFO:root:Average loss=0.18254924938082695 +INFO:root: +[Valid][3][UTT][ MSE = 0.2463 | LCC = 0.8713 | SRCC = 0.8719 ] +INFO:root:Average loss=0.17089494929099694 +INFO:root: +[Valid][4][UTT][ MSE = 0.2470 | LCC = 0.8759 | SRCC = 0.8729 ] +INFO:root:Average loss=0.15709735634617317 +INFO:root: +[Valid][5][UTT][ MSE = 0.2222 | LCC = 0.8745 | SRCC = 0.8734 ] +INFO:root:Average loss=0.14892849966119498 +INFO:root: +[Valid][6][UTT][ MSE = 0.2212 | LCC = 0.8739 | SRCC = 0.8720 ] +INFO:root:Average loss=0.14104623863330254 +INFO:root: +[Valid][7][UTT][ MSE = 0.2234 | LCC = 0.8738 | SRCC = 0.8719 ] +INFO:root:Average loss=0.12613695258131394 +INFO:root: +[Valid][8][UTT][ MSE = 0.2094 | LCC = 0.8770 | SRCC = 0.8754 ] +INFO:root:Average loss=0.12352442951538624 +INFO:root: +[Valid][9][UTT][ MSE = 0.2190 | LCC = 0.8720 | SRCC = 0.8705 ] +INFO:root:Average loss=0.11420529765578416 +INFO:root: +[Valid][10][UTT][ MSE = 0.2549 | LCC = 0.8694 | SRCC = 0.8696 ] +INFO:root:Average loss=0.10757553692047413 +INFO:root: +[Valid][11][UTT][ MSE = 0.2180 | LCC = 0.8696 | SRCC = 0.8695 ] +INFO:root:Average loss=0.10033179628543365 +INFO:root: +[Valid][12][UTT][ MSE = 0.2222 | LCC = 0.8692 | SRCC = 0.8687 ] +INFO:root:Average loss=0.09563998142496133 +INFO:root: +[Valid][13][UTT][ MSE = 0.2355 | LCC = 0.8694 | SRCC = 0.8695 ] +INFO:root:Average loss=0.0917367219256285 +INFO:root: +[Valid][14][UTT][ MSE = 0.2116 | LCC = 0.8686 | SRCC = 0.8684 ] +INFO:root:Average loss=0.08874211274087429 +INFO:root: +[Valid][15][UTT][ MSE = 0.2121 | LCC = 0.8725 | SRCC = 0.8718 ] +INFO:root:Average loss=0.0815517473488282 +INFO:root: +[Valid][16][UTT][ MSE = 0.2059 | LCC = 0.8702 | SRCC = 0.8701 ] +INFO:root:Average loss=0.07893724978352204 +INFO:root: +[Valid][17][UTT][ MSE = 0.2145 | LCC = 0.8684 | SRCC = 0.8678 ] +INFO:root:Average loss=0.0757449062493367 +INFO:root: +[Valid][18][UTT][ MSE = 0.2303 | LCC = 0.8700 | SRCC = 0.8694 ] +INFO:root:Average loss=0.07038048349129848 +INFO:root: +[Valid][19][UTT][ MSE = 0.2441 | LCC = 0.8675 | SRCC = 0.8678 ] +INFO:root:Average loss=0.07097534018640335 +INFO:root: +[Valid][20][UTT][ MSE = 0.2134 | LCC = 0.8673 | SRCC = 0.8690 ] +INFO:root:Average loss=0.06880145233411056 +INFO:root: +[Valid][21][UTT][ MSE = 0.2128 | LCC = 0.8672 | SRCC = 0.8670 ] +INFO:root:Average loss=0.06282092716831428 +INFO:root: +[Valid][22][UTT][ MSE = 0.2105 | LCC = 0.8671 | SRCC = 0.8681 ] +INFO:root:Average loss=0.0620267391204834 +INFO:root: +[Valid][23][UTT][ MSE = 0.2220 | LCC = 0.8663 | SRCC = 0.8674 ] +INFO:root:Average loss=0.057826548910293825 +INFO:root: +[Valid][24][UTT][ MSE = 0.2133 | LCC = 0.8665 | SRCC = 0.8664 ] +INFO:root:Average loss=0.05490911045135596 +INFO:root: +[Valid][25][UTT][ MSE = 0.2183 | LCC = 0.8623 | SRCC = 0.8629 ] +INFO:root:Average loss=0.05557106631115461 +INFO:root: +[Valid][26][UTT][ MSE = 0.2193 | LCC = 0.8646 | SRCC = 0.8639 ] +INFO:root:Average loss=0.0529148426766579 +INFO:root: +[Valid][27][UTT][ MSE = 0.2167 | LCC = 0.8634 | SRCC = 0.8641 ] +INFO:root:Average loss=0.051990470251975916 +INFO:root: +[Valid][28][UTT][ MSE = 0.2142 | LCC = 0.8652 | SRCC = 0.8659 ] +INFO:root:Average loss=0.04928527932423048 +INFO:root: +[Valid][29][UTT][ MSE = 0.2204 | LCC = 0.8655 | SRCC = 0.8651 ] +INFO:root:Average loss=0.05013318672680702 +INFO:root: +[Valid][30][UTT][ MSE = 0.2124 | LCC = 0.8664 | SRCC = 0.8668 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2216 | LCC = 0.8699 | SRCC = 0.8710 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d3786f427e3eb68a4549d5447730eb7b0645a0d0 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:87fcca01a9d714c3cfabe406100fbb47525fbdc6b76c3a15289326abb60a7876 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f1af712cd6d52e156331eebb0c846f10c7488158 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:40f08c679e409afe1b814699776bf9f833c9acccfc46bc039607a910f0a31be0 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e29b1c0fa5ac13afc69de4a7b3ac1e00dffd5af9 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/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=0.39070606231689453 +INFO:root: +[Valid][0][UTT][ MSE = 0.2956 | LCC = 0.8346 | SRCC = 0.8402 ] +INFO:root:Average loss=0.22447705536316603 +INFO:root: +[Valid][1][UTT][ MSE = 0.2774 | LCC = 0.8651 | SRCC = 0.8622 ] +INFO:root:Average loss=0.20570342242717743 +INFO:root: +[Valid][2][UTT][ MSE = 0.2572 | LCC = 0.8730 | SRCC = 0.8703 ] +INFO:root:Average loss=0.18220939468114805 +INFO:root: +[Valid][3][UTT][ MSE = 0.2374 | LCC = 0.8759 | SRCC = 0.8739 ] +INFO:root:Average loss=0.16930523820412466 +INFO:root: +[Valid][4][UTT][ MSE = 0.2373 | LCC = 0.8743 | SRCC = 0.8727 ] +INFO:root:Average loss=0.15425335596769285 +INFO:root: +[Valid][5][UTT][ MSE = 0.2248 | LCC = 0.8782 | SRCC = 0.8756 ] +INFO:root:Average loss=0.1470878061193686 +INFO:root: +[Valid][6][UTT][ MSE = 0.2613 | LCC = 0.8771 | SRCC = 0.8740 ] +INFO:root:Average loss=0.13821052864958078 +INFO:root: +[Valid][7][UTT][ MSE = 0.2427 | LCC = 0.8716 | SRCC = 0.8694 ] +INFO:root:Average loss=0.127827577961561 +INFO:root: +[Valid][8][UTT][ MSE = 0.2235 | LCC = 0.8714 | SRCC = 0.8689 ] +INFO:root:Average loss=0.12015475084384282 +INFO:root: +[Valid][9][UTT][ MSE = 0.2286 | LCC = 0.8700 | SRCC = 0.8679 ] +INFO:root:Average loss=0.11122395270145856 +INFO:root: +[Valid][10][UTT][ MSE = 0.2445 | LCC = 0.8683 | SRCC = 0.8658 ] +INFO:root:Average loss=0.10834319631640728 +INFO:root: +[Valid][11][UTT][ MSE = 0.2293 | LCC = 0.8680 | SRCC = 0.8654 ] +INFO:root:Average loss=0.10114161269022869 +INFO:root: +[Valid][12][UTT][ MSE = 0.2309 | LCC = 0.8669 | SRCC = 0.8667 ] +INFO:root:Average loss=0.09023799417683712 +INFO:root: +[Valid][13][UTT][ MSE = 0.2267 | LCC = 0.8661 | SRCC = 0.8649 ] +INFO:root:Average loss=0.08530046427861238 +INFO:root: +[Valid][14][UTT][ MSE = 0.2207 | LCC = 0.8625 | SRCC = 0.8608 ] +INFO:root:Average loss=0.07803452119995387 +INFO:root: +[Valid][15][UTT][ MSE = 0.2632 | LCC = 0.8579 | SRCC = 0.8564 ] +INFO:root:Average loss=0.08514710081120332 +INFO:root: +[Valid][16][UTT][ MSE = 0.2363 | LCC = 0.8631 | SRCC = 0.8602 ] +INFO:root:Average loss=0.07096443793330437 +INFO:root: +[Valid][17][UTT][ MSE = 0.2493 | LCC = 0.8636 | SRCC = 0.8618 ] +INFO:root:Average loss=0.07061165786133362 +INFO:root: +[Valid][18][UTT][ MSE = 0.2271 | LCC = 0.8648 | SRCC = 0.8637 ] +INFO:root:Average loss=0.06606489644409755 +INFO:root: +[Valid][19][UTT][ MSE = 0.2376 | LCC = 0.8643 | SRCC = 0.8617 ] +INFO:root:Average loss=0.06405561431669272 +INFO:root: +[Valid][20][UTT][ MSE = 0.2353 | LCC = 0.8588 | SRCC = 0.8581 ] +INFO:root:Average loss=0.06327959849761847 +INFO:root: +[Valid][21][UTT][ MSE = 0.2255 | LCC = 0.8636 | SRCC = 0.8640 ] +INFO:root:Average loss=0.05869640634419062 +INFO:root: +[Valid][22][UTT][ MSE = 0.2348 | LCC = 0.8628 | SRCC = 0.8614 ] +INFO:root:Average loss=0.05457077390299393 +INFO:root: +[Valid][23][UTT][ MSE = 0.2278 | LCC = 0.8626 | SRCC = 0.8623 ] +INFO:root:Average loss=0.05006869633992513 +INFO:root: +[Valid][24][UTT][ MSE = 0.2436 | LCC = 0.8592 | SRCC = 0.8581 ] +INFO:root:Average loss=0.051547071777093105 +INFO:root: +[Valid][25][UTT][ MSE = 0.2287 | LCC = 0.8625 | SRCC = 0.8607 ] +INFO:root:Average loss=0.0490439714720616 +INFO:root: +[Valid][26][UTT][ MSE = 0.2332 | LCC = 0.8611 | SRCC = 0.8600 ] +INFO:root:Average loss=0.04719956909330227 +INFO:root: +[Valid][27][UTT][ MSE = 0.2266 | LCC = 0.8639 | SRCC = 0.8632 ] +INFO:root:Average loss=0.05132829550749216 +INFO:root: +[Valid][28][UTT][ MSE = 0.2409 | LCC = 0.8591 | SRCC = 0.8583 ] +INFO:root:Average loss=0.04626798496032373 +INFO:root: +[Valid][29][UTT][ MSE = 0.2276 | LCC = 0.8630 | SRCC = 0.8626 ] +INFO:root:Average loss=0.044534252263987675 +INFO:root: +[Valid][30][UTT][ MSE = 0.2320 | LCC = 0.8631 | SRCC = 0.8624 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2381 | LCC = 0.8715 | SRCC = 0.8714 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9aa2e307129caadb24a1ab7e1dbb67daa351a056 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:22bfc2f0484ce4a71df6882d8956a4200ae294c8e493e738e8d6d5b30c8f92ea +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e5dc20fb7fec09aeb69ff78ce234b0adae73b5dc --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0bfe68c47a798258a559f1e3f6c4f4dc46d7949e696e101eca29e51f9c3c6f4e +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ee10bfa2563a7702d862630df2b00882ea9d9207 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/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=0.41382645796506834 +INFO:root: +[Valid][0][UTT][ MSE = 0.3288 | LCC = 0.8236 | SRCC = 0.8325 ] +INFO:root:Average loss=0.22901040583084792 +INFO:root: +[Valid][1][UTT][ MSE = 0.2706 | LCC = 0.8546 | SRCC = 0.8562 ] +INFO:root:Average loss=0.19952639784568396 +INFO:root: +[Valid][2][UTT][ MSE = 0.2547 | LCC = 0.8653 | SRCC = 0.8636 ] +INFO:root:Average loss=0.1895627920062114 +INFO:root: +[Valid][3][UTT][ MSE = 0.2452 | LCC = 0.8707 | SRCC = 0.8691 ] +INFO:root:Average loss=0.1738398369306173 +INFO:root: +[Valid][4][UTT][ MSE = 0.2435 | LCC = 0.8670 | SRCC = 0.8682 ] +INFO:root:Average loss=0.15643102599260134 +INFO:root: +[Valid][5][UTT][ MSE = 0.2299 | LCC = 0.8753 | SRCC = 0.8740 ] +INFO:root:Average loss=0.14948061442910096 +INFO:root: +[Valid][6][UTT][ MSE = 0.2230 | LCC = 0.8705 | SRCC = 0.8707 ] +INFO:root:Average loss=0.13898896406858396 +INFO:root: +[Valid][7][UTT][ MSE = 0.2486 | LCC = 0.8711 | SRCC = 0.8696 ] +INFO:root:Average loss=0.1291853207617234 +INFO:root: +[Valid][8][UTT][ MSE = 0.2209 | LCC = 0.8705 | SRCC = 0.8702 ] +INFO:root:Average loss=0.11830809273016758 +INFO:root: +[Valid][9][UTT][ MSE = 0.2282 | LCC = 0.8720 | SRCC = 0.8714 ] +INFO:root:Average loss=0.11295829092462857 +INFO:root: +[Valid][10][UTT][ MSE = 0.2265 | LCC = 0.8693 | SRCC = 0.8692 ] +INFO:root:Average loss=0.10388922237623961 +INFO:root: +[Valid][11][UTT][ MSE = 0.2213 | LCC = 0.8658 | SRCC = 0.8661 ] +INFO:root:Average loss=0.09452481739796124 +INFO:root: +[Valid][12][UTT][ MSE = 0.2524 | LCC = 0.8614 | SRCC = 0.8600 ] +INFO:root:Average loss=0.0893687550456096 +INFO:root: +[Valid][13][UTT][ MSE = 0.2277 | LCC = 0.8648 | SRCC = 0.8635 ] +INFO:root:Average loss=0.08558994894608474 +INFO:root: +[Valid][14][UTT][ MSE = 0.2500 | LCC = 0.8622 | SRCC = 0.8619 ] +INFO:root:Average loss=0.07922579042422466 +INFO:root: +[Valid][15][UTT][ MSE = 0.2352 | LCC = 0.8619 | SRCC = 0.8606 ] +INFO:root:Average loss=0.074670611283718 +INFO:root: +[Valid][16][UTT][ MSE = 0.2293 | LCC = 0.8576 | SRCC = 0.8557 ] +INFO:root:Average loss=0.0733890742636644 +INFO:root: +[Valid][17][UTT][ MSE = 0.2507 | LCC = 0.8583 | SRCC = 0.8588 ] +INFO:root:Average loss=0.07004839678605397 +INFO:root: +[Valid][18][UTT][ MSE = 0.2252 | LCC = 0.8615 | SRCC = 0.8615 ] +INFO:root:Average loss=0.06636198084705915 +INFO:root: +[Valid][19][UTT][ MSE = 0.2389 | LCC = 0.8605 | SRCC = 0.8577 ] +INFO:root:Average loss=0.06404952074472721 +INFO:root: +[Valid][20][UTT][ MSE = 0.2561 | LCC = 0.8599 | SRCC = 0.8596 ] +INFO:root:Average loss=0.06137077763485603 +INFO:root: +[Valid][21][UTT][ MSE = 0.2361 | LCC = 0.8581 | SRCC = 0.8564 ] +INFO:root:Average loss=0.05868253560784536 +INFO:root: +[Valid][22][UTT][ MSE = 0.2456 | LCC = 0.8569 | SRCC = 0.8545 ] +INFO:root:Average loss=0.05748657481028484 +INFO:root: +[Valid][23][UTT][ MSE = 0.2399 | LCC = 0.8543 | SRCC = 0.8546 ] +INFO:root:Average loss=0.05215535714076115 +INFO:root: +[Valid][24][UTT][ MSE = 0.2633 | LCC = 0.8565 | SRCC = 0.8568 ] +INFO:root:Average loss=0.05165512458636211 +INFO:root: +[Valid][25][UTT][ MSE = 0.2310 | LCC = 0.8576 | SRCC = 0.8575 ] +INFO:root:Average loss=0.05489631770895077 +INFO:root: +[Valid][26][UTT][ MSE = 0.2325 | LCC = 0.8559 | SRCC = 0.8558 ] +INFO:root:Average loss=0.05147262072811524 +INFO:root: +[Valid][27][UTT][ MSE = 0.2555 | LCC = 0.8584 | SRCC = 0.8571 ] +INFO:root:Average loss=0.046991667280403465 +INFO:root: +[Valid][28][UTT][ MSE = 0.2379 | LCC = 0.8569 | SRCC = 0.8561 ] +INFO:root:Average loss=0.042602925728528925 +INFO:root: +[Valid][29][UTT][ MSE = 0.2477 | LCC = 0.8542 | SRCC = 0.8541 ] +INFO:root:Average loss=0.04191091907425569 +INFO:root: +[Valid][30][UTT][ MSE = 0.2506 | LCC = 0.8540 | SRCC = 0.8533 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2305 | LCC = 0.8709 | SRCC = 0.8710 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8b5af1ac43ef535fb6ec547a1bebf649ecd98c56 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:791efbe752cd91f577d65cf8059a47548360c7f4ab0a9b9fddc94ec3bd2ddbfc +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..76cb9bb86d0441ab6b5f93e6048360c08fd2441f --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:31ddf50a3d5fb3e586b52b50383db3ee7ed9439c236c379a34cd4a3c8f287fb5 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f780cf1db61926e3c5ed0d6c437b5c3f99d3e2ac --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/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=0.3469563548763593 +INFO:root: +[Valid][0][UTT][ MSE = 0.2915 | LCC = 0.8401 | SRCC = 0.8424 ] +INFO:root:Average loss=0.22782965443837336 +INFO:root: +[Valid][1][UTT][ MSE = 0.2551 | LCC = 0.8551 | SRCC = 0.8552 ] +INFO:root:Average loss=0.20564006832547677 +INFO:root: +[Valid][2][UTT][ MSE = 0.2439 | LCC = 0.8645 | SRCC = 0.8638 ] +INFO:root:Average loss=0.1848632510847006 +INFO:root: +[Valid][3][UTT][ MSE = 0.2464 | LCC = 0.8687 | SRCC = 0.8678 ] +INFO:root:Average loss=0.17175757655730614 +INFO:root: +[Valid][4][UTT][ MSE = 0.2321 | LCC = 0.8693 | SRCC = 0.8681 ] +INFO:root:Average loss=0.16200700182563219 +INFO:root: +[Valid][5][UTT][ MSE = 0.2351 | LCC = 0.8695 | SRCC = 0.8694 ] +INFO:root:Average loss=0.15539379513416535 +INFO:root: +[Valid][6][UTT][ MSE = 0.2291 | LCC = 0.8725 | SRCC = 0.8712 ] +INFO:root:Average loss=0.14236059746681115 +INFO:root: +[Valid][7][UTT][ MSE = 0.2218 | LCC = 0.8708 | SRCC = 0.8699 ] +INFO:root:Average loss=0.12803342680518442 +INFO:root: +[Valid][8][UTT][ MSE = 0.2194 | LCC = 0.8693 | SRCC = 0.8681 ] +INFO:root:Average loss=0.12114565580701217 +INFO:root: +[Valid][9][UTT][ MSE = 0.2482 | LCC = 0.8650 | SRCC = 0.8642 ] +INFO:root:Average loss=0.12353137441170521 +INFO:root: +[Valid][10][UTT][ MSE = 0.2304 | LCC = 0.8650 | SRCC = 0.8639 ] +INFO:root:Average loss=0.10533048623265365 +INFO:root: +[Valid][11][UTT][ MSE = 0.2416 | LCC = 0.8607 | SRCC = 0.8598 ] +INFO:root:Average loss=0.10142372462611932 +INFO:root: +[Valid][12][UTT][ MSE = 0.2267 | LCC = 0.8624 | SRCC = 0.8615 ] +INFO:root:Average loss=0.09906589765197192 +INFO:root: +[Valid][13][UTT][ MSE = 0.2360 | LCC = 0.8638 | SRCC = 0.8616 ] +INFO:root:Average loss=0.0890628868379654 +INFO:root: +[Valid][14][UTT][ MSE = 0.2227 | LCC = 0.8600 | SRCC = 0.8592 ] +INFO:root:Average loss=0.08281335568962953 +INFO:root: +[Valid][15][UTT][ MSE = 0.2259 | LCC = 0.8612 | SRCC = 0.8600 ] +INFO:root:Average loss=0.08320613028720403 +INFO:root: +[Valid][16][UTT][ MSE = 0.2351 | LCC = 0.8567 | SRCC = 0.8551 ] +INFO:root:Average loss=0.07743480371741149 +INFO:root: +[Valid][17][UTT][ MSE = 0.2259 | LCC = 0.8583 | SRCC = 0.8578 ] +INFO:root:Average loss=0.07216754326453576 +INFO:root: +[Valid][18][UTT][ MSE = 0.2441 | LCC = 0.8514 | SRCC = 0.8500 ] +INFO:root:Average loss=0.06927430734802516 +INFO:root: +[Valid][19][UTT][ MSE = 0.2286 | LCC = 0.8572 | SRCC = 0.8551 ] +INFO:root:Average loss=0.06602833797343266 +INFO:root: +[Valid][20][UTT][ MSE = 0.2247 | LCC = 0.8586 | SRCC = 0.8576 ] +INFO:root:Average loss=0.061307888764601484 +INFO:root: +[Valid][21][UTT][ MSE = 0.2353 | LCC = 0.8558 | SRCC = 0.8549 ] +INFO:root:Average loss=0.06296748629747292 +INFO:root: +[Valid][22][UTT][ MSE = 0.2326 | LCC = 0.8574 | SRCC = 0.8569 ] +INFO:root:Average loss=0.057513836675729506 +INFO:root: +[Valid][23][UTT][ MSE = 0.2372 | LCC = 0.8546 | SRCC = 0.8535 ] +INFO:root:Average loss=0.05522278490930031 +INFO:root: +[Valid][24][UTT][ MSE = 0.2316 | LCC = 0.8555 | SRCC = 0.8546 ] +INFO:root:Average loss=0.05347588123419346 +INFO:root: +[Valid][25][UTT][ MSE = 0.2317 | LCC = 0.8551 | SRCC = 0.8540 ] +INFO:root:Average loss=0.05038572733218853 +INFO:root: +[Valid][26][UTT][ MSE = 0.2331 | LCC = 0.8568 | SRCC = 0.8559 ] +INFO:root:Average loss=0.05020139743693364 +INFO:root: +[Valid][27][UTT][ MSE = 0.2298 | LCC = 0.8571 | SRCC = 0.8561 ] +INFO:root:Average loss=0.04635571301556551 +INFO:root: +[Valid][28][UTT][ MSE = 0.2376 | LCC = 0.8563 | SRCC = 0.8556 ] +INFO:root:Average loss=0.04594200658492553 +INFO:root: +[Valid][29][UTT][ MSE = 0.2274 | LCC = 0.8560 | SRCC = 0.8552 ] +INFO:root:Average loss=0.046133052271146044 +INFO:root: +[Valid][30][UTT][ MSE = 0.2390 | LCC = 0.8543 | SRCC = 0.8533 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2409 | LCC = 0.8652 | SRCC = 0.8670 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8f0b78786ca3dfa86bda2c09a647e4ed9564f783 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5a6096f112cad775b72db36b90296965f200e7c023cfae35ddafc3d0e55bf521 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..055666167b9182afad49ae5ccae65e9648c3d4cb --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d9288a801cf2c3416130e7e401f348062468647beca779bcd6ac7ba0930fe468 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0a14d7e560452c170e30555d0accc0776290f94e --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/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=0.3815225334121631 +INFO:root: +[Valid][0][UTT][ MSE = 0.2916 | LCC = 0.8319 | SRCC = 0.8394 ] +INFO:root:Average loss=0.22421150387097627 +INFO:root: +[Valid][1][UTT][ MSE = 0.2582 | LCC = 0.8608 | SRCC = 0.8606 ] +INFO:root:Average loss=0.19904962392189565 +INFO:root: +[Valid][2][UTT][ MSE = 0.2439 | LCC = 0.8702 | SRCC = 0.8693 ] +INFO:root:Average loss=0.17859859220110452 +INFO:root: +[Valid][3][UTT][ MSE = 0.2189 | LCC = 0.8732 | SRCC = 0.8724 ] +INFO:root:Average loss=0.16671495168254927 +INFO:root: +[Valid][4][UTT][ MSE = 0.2317 | LCC = 0.8742 | SRCC = 0.8731 ] +INFO:root:Average loss=0.15855873481203347 +INFO:root: +[Valid][5][UTT][ MSE = 0.2303 | LCC = 0.8753 | SRCC = 0.8736 ] +INFO:root:Average loss=0.14435220385591188 +INFO:root: +[Valid][6][UTT][ MSE = 0.2421 | LCC = 0.8707 | SRCC = 0.8691 ] +INFO:root:Average loss=0.13224162447911042 +INFO:root: +[Valid][7][UTT][ MSE = 0.2333 | LCC = 0.8694 | SRCC = 0.8669 ] +INFO:root:Average loss=0.1255853943144664 +INFO:root: +[Valid][8][UTT][ MSE = 0.2384 | LCC = 0.8689 | SRCC = 0.8667 ] +INFO:root:Average loss=0.11199452269535798 +INFO:root: +[Valid][9][UTT][ MSE = 0.2306 | LCC = 0.8680 | SRCC = 0.8660 ] +INFO:root:Average loss=0.11306363497024928 +INFO:root: +[Valid][10][UTT][ MSE = 0.2410 | LCC = 0.8641 | SRCC = 0.8613 ] +INFO:root:Average loss=0.10192532173525064 +INFO:root: +[Valid][11][UTT][ MSE = 0.2176 | LCC = 0.8679 | SRCC = 0.8661 ] +INFO:root:Average loss=0.09500166525443395 +INFO:root: +[Valid][12][UTT][ MSE = 0.2308 | LCC = 0.8645 | SRCC = 0.8635 ] +INFO:root:Average loss=0.09073156720170608 +INFO:root: +[Valid][13][UTT][ MSE = 0.2434 | LCC = 0.8636 | SRCC = 0.8605 ] +INFO:root:Average loss=0.08433127002074169 +INFO:root: +[Valid][14][UTT][ MSE = 0.2313 | LCC = 0.8647 | SRCC = 0.8623 ] +INFO:root:Average loss=0.08260872568457554 +INFO:root: +[Valid][15][UTT][ MSE = 0.2507 | LCC = 0.8594 | SRCC = 0.8555 ] +INFO:root:Average loss=0.07827913216673411 +INFO:root: +[Valid][16][UTT][ MSE = 0.2323 | LCC = 0.8613 | SRCC = 0.8585 ] +INFO:root:Average loss=0.07546127620988931 +INFO:root: +[Valid][17][UTT][ MSE = 0.2271 | LCC = 0.8603 | SRCC = 0.8570 ] +INFO:root:Average loss=0.06926539960579994 +INFO:root: +[Valid][18][UTT][ MSE = 0.2441 | LCC = 0.8599 | SRCC = 0.8576 ] +INFO:root:Average loss=0.06891322475022231 +INFO:root: +[Valid][19][UTT][ MSE = 0.2291 | LCC = 0.8595 | SRCC = 0.8562 ] +INFO:root:Average loss=0.062278188000886865 +INFO:root: +[Valid][20][UTT][ MSE = 0.2398 | LCC = 0.8570 | SRCC = 0.8529 ] +INFO:root:Average loss=0.06377345113418041 +INFO:root: +[Valid][21][UTT][ MSE = 0.2243 | LCC = 0.8605 | SRCC = 0.8574 ] +INFO:root:Average loss=0.05906577618458332 +INFO:root: +[Valid][22][UTT][ MSE = 0.2214 | LCC = 0.8616 | SRCC = 0.8585 ] +INFO:root:Average loss=0.05487806655657597 +INFO:root: +[Valid][23][UTT][ MSE = 0.2547 | LCC = 0.8542 | SRCC = 0.8497 ] +INFO:root:Average loss=0.05343935896570866 +INFO:root: +[Valid][24][UTT][ MSE = 0.2361 | LCC = 0.8560 | SRCC = 0.8510 ] +INFO:root:Average loss=0.05040177144110203 +INFO:root: +[Valid][25][UTT][ MSE = 0.2397 | LCC = 0.8578 | SRCC = 0.8547 ] +INFO:root:Average loss=0.05169765104372532 +INFO:root: +[Valid][26][UTT][ MSE = 0.2338 | LCC = 0.8563 | SRCC = 0.8526 ] +INFO:root:Average loss=0.046884946477336764 +INFO:root: +[Valid][27][UTT][ MSE = 0.2343 | LCC = 0.8531 | SRCC = 0.8492 ] +INFO:root:Average loss=0.046133783765328236 +INFO:root: +[Valid][28][UTT][ MSE = 0.2246 | LCC = 0.8596 | SRCC = 0.8566 ] +INFO:root:Average loss=0.04486478576197838 +INFO:root: +[Valid][29][UTT][ MSE = 0.2361 | LCC = 0.8576 | SRCC = 0.8541 ] +INFO:root:Average loss=0.04342797362747101 +INFO:root: +[Valid][30][UTT][ MSE = 0.2488 | LCC = 0.8532 | SRCC = 0.8497 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2426 | LCC = 0.8683 | SRCC = 0.8675 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5b03187b0a56db39fa4609aab5425015655832f8 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d56e1a92faaa259208f9aa5ec4522b94de8e7ac4366757fac8b5a5e4defbe496 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..93905d237e6217807c5f1f5f63ebf67db3f30860 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b711f3a808209f796f6d3448f1c92e07c062f3c7f5b23af1d1dfc1f4cdb00697 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3f394e7dd51219a5de09745c9a085808298461a7 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer4/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=0.36418532943114257 +INFO:root: +[Valid][0][UTT][ MSE = 0.3234 | LCC = 0.8266 | SRCC = 0.8384 ] +INFO:root:Average loss=0.23175318023333183 +INFO:root: +[Valid][1][UTT][ MSE = 0.2649 | LCC = 0.8585 | SRCC = 0.8582 ] +INFO:root:Average loss=0.20113022959767243 +INFO:root: +[Valid][2][UTT][ MSE = 0.2589 | LCC = 0.8639 | SRCC = 0.8666 ] +INFO:root:Average loss=0.19017713268597922 +INFO:root: +[Valid][3][UTT][ MSE = 0.2393 | LCC = 0.8731 | SRCC = 0.8724 ] +INFO:root:Average loss=0.17137887357519224 +INFO:root: +[Valid][4][UTT][ MSE = 0.2346 | LCC = 0.8739 | SRCC = 0.8739 ] +INFO:root:Average loss=0.1562641639357958 +INFO:root: +[Valid][5][UTT][ MSE = 0.2287 | LCC = 0.8725 | SRCC = 0.8732 ] +INFO:root:Average loss=0.14685720730668458 +INFO:root: +[Valid][6][UTT][ MSE = 0.2160 | LCC = 0.8759 | SRCC = 0.8753 ] +INFO:root:Average loss=0.13743910460900038 +INFO:root: +[Valid][7][UTT][ MSE = 0.2231 | LCC = 0.8732 | SRCC = 0.8726 ] +INFO:root:Average loss=0.1270499940101917 +INFO:root: +[Valid][8][UTT][ MSE = 0.2091 | LCC = 0.8759 | SRCC = 0.8763 ] +INFO:root:Average loss=0.12044027466804554 +INFO:root: +[Valid][9][UTT][ MSE = 0.2157 | LCC = 0.8728 | SRCC = 0.8727 ] +INFO:root:Average loss=0.11327958202514893 +INFO:root: +[Valid][10][UTT][ MSE = 0.2217 | LCC = 0.8726 | SRCC = 0.8730 ] +INFO:root:Average loss=0.10424418709216973 +INFO:root: +[Valid][11][UTT][ MSE = 0.2448 | LCC = 0.8696 | SRCC = 0.8701 ] +INFO:root:Average loss=0.10329522975744346 +INFO:root: +[Valid][12][UTT][ MSE = 0.2285 | LCC = 0.8712 | SRCC = 0.8709 ] +INFO:root:Average loss=0.09408677001603138 +INFO:root: +[Valid][13][UTT][ MSE = 0.2227 | LCC = 0.8716 | SRCC = 0.8710 ] +INFO:root:Average loss=0.08586978324903892 +INFO:root: +[Valid][14][UTT][ MSE = 0.2191 | LCC = 0.8680 | SRCC = 0.8674 ] +INFO:root:Average loss=0.08400350742233105 +INFO:root: +[Valid][15][UTT][ MSE = 0.2454 | LCC = 0.8662 | SRCC = 0.8673 ] +INFO:root:Average loss=0.07641620709536931 +INFO:root: +[Valid][16][UTT][ MSE = 0.2260 | LCC = 0.8690 | SRCC = 0.8712 ] +INFO:root:Average loss=0.07844252731555547 +INFO:root: +[Valid][17][UTT][ MSE = 0.2523 | LCC = 0.8656 | SRCC = 0.8665 ] +INFO:root:Average loss=0.07231389697736655 +INFO:root: +[Valid][18][UTT][ MSE = 0.2448 | LCC = 0.8632 | SRCC = 0.8638 ] +INFO:root:Average loss=0.06881730439953315 +INFO:root: +[Valid][19][UTT][ MSE = 0.2383 | LCC = 0.8649 | SRCC = 0.8645 ] +INFO:root:Average loss=0.06529162881466058 +INFO:root: +[Valid][20][UTT][ MSE = 0.2416 | LCC = 0.8661 | SRCC = 0.8669 ] +INFO:root:Average loss=0.05817731274053072 +INFO:root: +[Valid][21][UTT][ MSE = 0.2349 | LCC = 0.8650 | SRCC = 0.8648 ] +INFO:root:Average loss=0.05935792629726422 +INFO:root: +[Valid][22][UTT][ MSE = 0.2222 | LCC = 0.8680 | SRCC = 0.8676 ] +INFO:root:Average loss=0.05460362298748432 +INFO:root: +[Valid][23][UTT][ MSE = 0.2333 | LCC = 0.8642 | SRCC = 0.8633 ] +INFO:root:Average loss=0.05575832289954027 +INFO:root: +[Valid][24][UTT][ MSE = 0.2258 | LCC = 0.8653 | SRCC = 0.8630 ] +INFO:root:Average loss=0.056319306341883466 +INFO:root: +[Valid][25][UTT][ MSE = 0.2460 | LCC = 0.8604 | SRCC = 0.8603 ] +INFO:root:Average loss=0.04851069852996331 +INFO:root: +[Valid][26][UTT][ MSE = 0.2408 | LCC = 0.8628 | SRCC = 0.8615 ] +INFO:root:Average loss=0.04836251928160588 +INFO:root: +[Valid][27][UTT][ MSE = 0.2393 | LCC = 0.8621 | SRCC = 0.8618 ] +INFO:root:Average loss=0.047687627136325225 +INFO:root: +[Valid][28][UTT][ MSE = 0.2223 | LCC = 0.8617 | SRCC = 0.8616 ] +INFO:root:Average loss=0.0470732861222365 +INFO:root: +[Valid][29][UTT][ MSE = 0.2706 | LCC = 0.8603 | SRCC = 0.8599 ] +INFO:root:Average loss=0.0447518687026623 +INFO:root: +[Valid][30][UTT][ MSE = 0.2513 | LCC = 0.8604 | SRCC = 0.8596 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2299 | LCC = 0.8685 | SRCC = 0.8697 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7362ec16d6b001185534ba1eeec52a2d4ba2d652 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3b2ab96dda2bd2212ff09e8c2d07938a470c03615140107c1701ee6b03569b79 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..80fab55749cd10f4b05fed4063b9e13db5b7df34 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d78d96c67ef2116461ce8b30e406920e8540ba42885ff82b412261724285d3ba +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..39a2d5b79a1097ecc2d3aae71e9de510d318fb3b --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/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=0.3506660681122389 +INFO:root: +[Valid][0][UTT][ MSE = 0.2879 | LCC = 0.8413 | SRCC = 0.8441 ] +INFO:root:Average loss=0.2239122346807749 +INFO:root: +[Valid][1][UTT][ MSE = 0.2525 | LCC = 0.8585 | SRCC = 0.8585 ] +INFO:root:Average loss=0.19635647325179514 +INFO:root: +[Valid][2][UTT][ MSE = 0.2874 | LCC = 0.8685 | SRCC = 0.8677 ] +INFO:root:Average loss=0.18216681107878685 +INFO:root: +[Valid][3][UTT][ MSE = 0.2295 | LCC = 0.8651 | SRCC = 0.8682 ] +INFO:root:Average loss=0.17164985147806314 +INFO:root: +[Valid][4][UTT][ MSE = 0.2525 | LCC = 0.8700 | SRCC = 0.8697 ] +INFO:root:Average loss=0.1562244404011812 +INFO:root: +[Valid][5][UTT][ MSE = 0.2229 | LCC = 0.8668 | SRCC = 0.8663 ] +INFO:root:Average loss=0.15114206142532519 +INFO:root: +[Valid][6][UTT][ MSE = 0.2816 | LCC = 0.8659 | SRCC = 0.8650 ] +INFO:root:Average loss=0.13366828285730803 +INFO:root: +[Valid][7][UTT][ MSE = 0.2625 | LCC = 0.8607 | SRCC = 0.8623 ] +INFO:root:Average loss=0.12557650261964554 +INFO:root: +[Valid][8][UTT][ MSE = 0.2357 | LCC = 0.8601 | SRCC = 0.8599 ] +INFO:root:Average loss=0.12397540828738457 +INFO:root: +[Valid][9][UTT][ MSE = 0.2321 | LCC = 0.8594 | SRCC = 0.8597 ] +INFO:root:Average loss=0.10975578685219471 +INFO:root: +[Valid][10][UTT][ MSE = 0.2324 | LCC = 0.8604 | SRCC = 0.8599 ] +INFO:root:Average loss=0.10825055407790038 +INFO:root: +[Valid][11][UTT][ MSE = 0.2310 | LCC = 0.8602 | SRCC = 0.8608 ] +INFO:root:Average loss=0.10084281785365863 +INFO:root: +[Valid][12][UTT][ MSE = 0.2248 | LCC = 0.8553 | SRCC = 0.8560 ] +INFO:root:Average loss=0.09024754883005069 +INFO:root: +[Valid][13][UTT][ MSE = 0.2453 | LCC = 0.8515 | SRCC = 0.8502 ] +INFO:root:Average loss=0.0872609772456762 +INFO:root: +[Valid][14][UTT][ MSE = 0.2417 | LCC = 0.8546 | SRCC = 0.8552 ] +INFO:root:Average loss=0.0809895490320065 +INFO:root: +[Valid][15][UTT][ MSE = 0.2417 | LCC = 0.8531 | SRCC = 0.8531 ] +INFO:root:Average loss=0.0782187684224202 +INFO:root: +[Valid][16][UTT][ MSE = 0.2351 | LCC = 0.8532 | SRCC = 0.8536 ] +INFO:root:Average loss=0.07323666623769662 +INFO:root: +[Valid][17][UTT][ MSE = 0.2340 | LCC = 0.8521 | SRCC = 0.8535 ] +INFO:root:Average loss=0.07233706933374588 +INFO:root: +[Valid][18][UTT][ MSE = 0.2467 | LCC = 0.8481 | SRCC = 0.8475 ] +INFO:root:Average loss=0.06440826531690665 +INFO:root: +[Valid][19][UTT][ MSE = 0.2448 | LCC = 0.8490 | SRCC = 0.8494 ] +INFO:root:Average loss=0.061791205778717995 +INFO:root: +[Valid][20][UTT][ MSE = 0.2373 | LCC = 0.8505 | SRCC = 0.8502 ] +INFO:root:Average loss=0.05936571010030233 +INFO:root: +[Valid][21][UTT][ MSE = 0.2387 | LCC = 0.8533 | SRCC = 0.8531 ] +INFO:root:Average loss=0.057636064501144946 +INFO:root: +[Valid][22][UTT][ MSE = 0.2416 | LCC = 0.8516 | SRCC = 0.8506 ] +INFO:root:Average loss=0.05481955303977697 +INFO:root: +[Valid][23][UTT][ MSE = 0.2475 | LCC = 0.8484 | SRCC = 0.8478 ] +INFO:root:Average loss=0.05235941278246733 +INFO:root: +[Valid][24][UTT][ MSE = 0.2507 | LCC = 0.8461 | SRCC = 0.8457 ] +INFO:root:Average loss=0.04917696712968441 +INFO:root: +[Valid][25][UTT][ MSE = 0.2370 | LCC = 0.8502 | SRCC = 0.8496 ] +INFO:root:Average loss=0.04853120565605469 +INFO:root: +[Valid][26][UTT][ MSE = 0.2438 | LCC = 0.8486 | SRCC = 0.8476 ] +INFO:root:Average loss=0.0482374593997613 +INFO:root: +[Valid][27][UTT][ MSE = 0.2398 | LCC = 0.8480 | SRCC = 0.8478 ] +INFO:root:Average loss=0.047138168213841244 +INFO:root: +[Valid][28][UTT][ MSE = 0.2651 | LCC = 0.8498 | SRCC = 0.8501 ] +INFO:root:Average loss=0.044254112702149614 +INFO:root: +[Valid][29][UTT][ MSE = 0.2501 | LCC = 0.8482 | SRCC = 0.8480 ] +INFO:root:Average loss=0.041749684689327694 +INFO:root: +[Valid][30][UTT][ MSE = 0.2448 | LCC = 0.8477 | SRCC = 0.8464 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2885 | LCC = 0.8659 | SRCC = 0.8681 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b2b6b963d9989f03c9274e335d9eef305e857e45 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4589ada76bfd10b8f11595ab54c3a293b842e72577d1c01a80c381a4ffd2ad90 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..42343ffc3235403cc3412148c0bd5ec41cf61c48 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:487ede7f6ff93e18eb14d2004a267ca680d0d7fd14a40a1376f10e5408ce7190 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..22fa3585e8eddc0045e93e02fe8e73867e71a06c --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/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=0.34860952313129717 +INFO:root: +[Valid][0][UTT][ MSE = 0.3306 | LCC = 0.8400 | SRCC = 0.8419 ] +INFO:root:Average loss=0.21788202627347067 +INFO:root: +[Valid][1][UTT][ MSE = 0.2951 | LCC = 0.8616 | SRCC = 0.8587 ] +INFO:root:Average loss=0.193148618325209 +INFO:root: +[Valid][2][UTT][ MSE = 0.2585 | LCC = 0.8665 | SRCC = 0.8638 ] +INFO:root:Average loss=0.17964933010248038 +INFO:root: +[Valid][3][UTT][ MSE = 0.3071 | LCC = 0.8673 | SRCC = 0.8645 ] +INFO:root:Average loss=0.16656268626833573 +INFO:root: +[Valid][4][UTT][ MSE = 0.2472 | LCC = 0.8694 | SRCC = 0.8685 ] +INFO:root:Average loss=0.15258954436733171 +INFO:root: +[Valid][5][UTT][ MSE = 0.2640 | LCC = 0.8664 | SRCC = 0.8649 ] +INFO:root:Average loss=0.14635997695418504 +INFO:root: +[Valid][6][UTT][ MSE = 0.2459 | LCC = 0.8680 | SRCC = 0.8663 ] +INFO:root:Average loss=0.13176286086822167 +INFO:root: +[Valid][7][UTT][ MSE = 0.2493 | LCC = 0.8652 | SRCC = 0.8643 ] +INFO:root:Average loss=0.12344094747916246 +INFO:root: +[Valid][8][UTT][ MSE = 0.2472 | LCC = 0.8591 | SRCC = 0.8576 ] +INFO:root:Average loss=0.1162449782475447 +INFO:root: +[Valid][9][UTT][ MSE = 0.2632 | LCC = 0.8590 | SRCC = 0.8593 ] +INFO:root:Average loss=0.10676347464323044 +INFO:root: +[Valid][10][UTT][ MSE = 0.2809 | LCC = 0.8546 | SRCC = 0.8550 ] +INFO:root:Average loss=0.10205524142545003 +INFO:root: +[Valid][11][UTT][ MSE = 0.2466 | LCC = 0.8558 | SRCC = 0.8552 ] +INFO:root:Average loss=0.09310868855279225 +INFO:root: +[Valid][12][UTT][ MSE = 0.2732 | LCC = 0.8529 | SRCC = 0.8511 ] +INFO:root:Average loss=0.08824093338961785 +INFO:root: +[Valid][13][UTT][ MSE = 0.2603 | LCC = 0.8502 | SRCC = 0.8487 ] +INFO:root:Average loss=0.08301931762924561 +INFO:root: +[Valid][14][UTT][ MSE = 0.2643 | LCC = 0.8534 | SRCC = 0.8513 ] +INFO:root:Average loss=0.08093626591830681 +INFO:root: +[Valid][15][UTT][ MSE = 0.2577 | LCC = 0.8491 | SRCC = 0.8468 ] +INFO:root:Average loss=0.07044779089016792 +INFO:root: +[Valid][16][UTT][ MSE = 0.2598 | LCC = 0.8485 | SRCC = 0.8465 ] +INFO:root:Average loss=0.07225989937209167 +INFO:root: +[Valid][17][UTT][ MSE = 0.2779 | LCC = 0.8467 | SRCC = 0.8450 ] +INFO:root:Average loss=0.0672268676929749 +INFO:root: +[Valid][18][UTT][ MSE = 0.2519 | LCC = 0.8484 | SRCC = 0.8465 ] +INFO:root:Average loss=0.06249176094738337 +INFO:root: +[Valid][19][UTT][ MSE = 0.2753 | LCC = 0.8489 | SRCC = 0.8461 ] +INFO:root:Average loss=0.06006383762145654 +INFO:root: +[Valid][20][UTT][ MSE = 0.2614 | LCC = 0.8472 | SRCC = 0.8447 ] +INFO:root:Average loss=0.05742910905526234 +INFO:root: +[Valid][21][UTT][ MSE = 0.2692 | LCC = 0.8468 | SRCC = 0.8453 ] +INFO:root:Average loss=0.056038545396847605 +INFO:root: +[Valid][22][UTT][ MSE = 0.2477 | LCC = 0.8475 | SRCC = 0.8468 ] +INFO:root:Average loss=0.05194484093823494 +INFO:root: +[Valid][23][UTT][ MSE = 0.2573 | LCC = 0.8464 | SRCC = 0.8444 ] +INFO:root:Average loss=0.04964481408779438 +INFO:root: +[Valid][24][UTT][ MSE = 0.2655 | LCC = 0.8503 | SRCC = 0.8480 ] +INFO:root:Average loss=0.048205056896385476 +INFO:root: +[Valid][25][UTT][ MSE = 0.3117 | LCC = 0.8456 | SRCC = 0.8445 ] +INFO:root:Average loss=0.047491592092391774 +INFO:root: +[Valid][26][UTT][ MSE = 0.2716 | LCC = 0.8475 | SRCC = 0.8456 ] +INFO:root:Average loss=0.045105395384897023 +INFO:root: +[Valid][27][UTT][ MSE = 0.2540 | LCC = 0.8500 | SRCC = 0.8486 ] +INFO:root:Average loss=0.04391550275091177 +INFO:root: +[Valid][28][UTT][ MSE = 0.2662 | LCC = 0.8446 | SRCC = 0.8431 ] +INFO:root:Average loss=0.04489046765061525 +INFO:root: +[Valid][29][UTT][ MSE = 0.2705 | LCC = 0.8460 | SRCC = 0.8442 ] +INFO:root:Average loss=0.041163261979818344 +INFO:root: +[Valid][30][UTT][ MSE = 0.2541 | LCC = 0.8479 | SRCC = 0.8460 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2356 | LCC = 0.8670 | SRCC = 0.8656 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2f29f3751886cfd2c592b5a0db55ffc2f360db84 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:77bb32d241070a35a984190490bc72b61cbbc6a7519a3bd433555f7c3b3163df +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..07ac944f1d3825700212002f4889fa6003692429 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5aed44ce024f56652c2e3a6ea95e3797139bebec6216b529d1c2fad5dd4534e8 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ea91691562ec49c13118924d3eff7411f85760d4 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/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=0.36550134076521945 +INFO:root: +[Valid][0][UTT][ MSE = 0.3226 | LCC = 0.8316 | SRCC = 0.8388 ] +INFO:root:Average loss=0.22743672801133913 +INFO:root: +[Valid][1][UTT][ MSE = 0.2660 | LCC = 0.8597 | SRCC = 0.8571 ] +INFO:root:Average loss=0.20518067326301184 +INFO:root: +[Valid][2][UTT][ MSE = 0.2485 | LCC = 0.8665 | SRCC = 0.8636 ] +INFO:root:Average loss=0.18579649170621848 +INFO:root: +[Valid][3][UTT][ MSE = 0.2340 | LCC = 0.8680 | SRCC = 0.8679 ] +INFO:root:Average loss=0.17097741269912475 +INFO:root: +[Valid][4][UTT][ MSE = 0.2351 | LCC = 0.8716 | SRCC = 0.8699 ] +INFO:root:Average loss=0.15770126659518632 +INFO:root: +[Valid][5][UTT][ MSE = 0.2346 | LCC = 0.8691 | SRCC = 0.8695 ] +INFO:root:Average loss=0.14602753863884851 +INFO:root: +[Valid][6][UTT][ MSE = 0.2373 | LCC = 0.8712 | SRCC = 0.8681 ] +INFO:root:Average loss=0.13873623091823015 +INFO:root: +[Valid][7][UTT][ MSE = 0.2252 | LCC = 0.8710 | SRCC = 0.8690 ] +INFO:root:Average loss=0.12935440481091157 +INFO:root: +[Valid][8][UTT][ MSE = 0.2316 | LCC = 0.8664 | SRCC = 0.8660 ] +INFO:root:Average loss=0.12060795113062248 +INFO:root: +[Valid][9][UTT][ MSE = 0.2521 | LCC = 0.8627 | SRCC = 0.8601 ] +INFO:root:Average loss=0.11010828595130871 +INFO:root: +[Valid][10][UTT][ MSE = 0.2514 | LCC = 0.8592 | SRCC = 0.8559 ] +INFO:root:Average loss=0.10108953083936985 +INFO:root: +[Valid][11][UTT][ MSE = 0.2290 | LCC = 0.8579 | SRCC = 0.8559 ] +INFO:root:Average loss=0.09476434105099776 +INFO:root: +[Valid][12][UTT][ MSE = 0.2854 | LCC = 0.8576 | SRCC = 0.8557 ] +INFO:root:Average loss=0.09219642150669526 +INFO:root: +[Valid][13][UTT][ MSE = 0.2503 | LCC = 0.8568 | SRCC = 0.8536 ] +INFO:root:Average loss=0.08468488732782695 +INFO:root: +[Valid][14][UTT][ MSE = 0.2466 | LCC = 0.8538 | SRCC = 0.8510 ] +INFO:root:Average loss=0.08272089837835385 +INFO:root: +[Valid][15][UTT][ MSE = 0.2446 | LCC = 0.8535 | SRCC = 0.8505 ] +INFO:root:Average loss=0.07756284471505727 +INFO:root: +[Valid][16][UTT][ MSE = 0.2372 | LCC = 0.8558 | SRCC = 0.8531 ] +INFO:root:Average loss=0.07178546268588458 +INFO:root: +[Valid][17][UTT][ MSE = 0.2386 | LCC = 0.8567 | SRCC = 0.8557 ] +INFO:root:Average loss=0.06816431913429345 +INFO:root: +[Valid][18][UTT][ MSE = 0.2701 | LCC = 0.8527 | SRCC = 0.8514 ] +INFO:root:Average loss=0.06552931766670483 +INFO:root: +[Valid][19][UTT][ MSE = 0.2335 | LCC = 0.8537 | SRCC = 0.8532 ] +INFO:root:Average loss=0.06198607977384176 +INFO:root: +[Valid][20][UTT][ MSE = 0.2349 | LCC = 0.8547 | SRCC = 0.8528 ] +INFO:root:Average loss=0.06000591652133526 +INFO:root: +[Valid][21][UTT][ MSE = 0.2320 | LCC = 0.8545 | SRCC = 0.8531 ] +INFO:root:Average loss=0.05451665804363214 +INFO:root: +[Valid][22][UTT][ MSE = 0.2382 | LCC = 0.8548 | SRCC = 0.8530 ] +INFO:root:Average loss=0.05283194245436253 +INFO:root: +[Valid][23][UTT][ MSE = 0.2318 | LCC = 0.8549 | SRCC = 0.8535 ] +INFO:root:Average loss=0.04977195900984299 +INFO:root: +[Valid][24][UTT][ MSE = 0.2489 | LCC = 0.8500 | SRCC = 0.8467 ] +INFO:root:Average loss=0.05085209582765133 +INFO:root: +[Valid][25][UTT][ MSE = 0.2372 | LCC = 0.8535 | SRCC = 0.8506 ] +INFO:root:Average loss=0.04573140121423281 +INFO:root: +[Valid][26][UTT][ MSE = 0.2347 | LCC = 0.8535 | SRCC = 0.8518 ] +INFO:root:Average loss=0.04259916769865996 +INFO:root: +[Valid][27][UTT][ MSE = 0.2454 | LCC = 0.8530 | SRCC = 0.8508 ] +INFO:root:Average loss=0.043112704578118444 +INFO:root: +[Valid][28][UTT][ MSE = 0.2318 | LCC = 0.8565 | SRCC = 0.8548 ] +INFO:root:Average loss=0.04460999372009283 +INFO:root: +[Valid][29][UTT][ MSE = 0.2359 | LCC = 0.8547 | SRCC = 0.8538 ] +INFO:root:Average loss=0.04205459263175726 +INFO:root: +[Valid][30][UTT][ MSE = 0.2543 | LCC = 0.8510 | SRCC = 0.8490 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2344 | LCC = 0.8667 | SRCC = 0.8646 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..38a1cf4598c5a420ad0f2bd8c8daa66c30427eaf --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5f39b7dbce49af8e639e17780a7382126690229ef9160bc770fcc8d7377caf99 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..48669094b28202a07c46e6ec1380848a904e33c2 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d4c335f669951a83bd21461cb73e5be22a03191691d02baf4b35acae79669088 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..20bf259d37a2563bf22e43afb15a727b0f22feee --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/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=0.3568229608428784 +INFO:root: +[Valid][0][UTT][ MSE = 0.3013 | LCC = 0.8375 | SRCC = 0.8406 ] +INFO:root:Average loss=0.23326290589876664 +INFO:root: +[Valid][1][UTT][ MSE = 0.2702 | LCC = 0.8575 | SRCC = 0.8555 ] +INFO:root:Average loss=0.20395341993142396 +INFO:root: +[Valid][2][UTT][ MSE = 0.2610 | LCC = 0.8656 | SRCC = 0.8645 ] +INFO:root:Average loss=0.18683569391186422 +INFO:root: +[Valid][3][UTT][ MSE = 0.2465 | LCC = 0.8699 | SRCC = 0.8677 ] +INFO:root:Average loss=0.1720204403003057 +INFO:root: +[Valid][4][UTT][ MSE = 0.2360 | LCC = 0.8722 | SRCC = 0.8702 ] +INFO:root:Average loss=0.15676654903934553 +INFO:root: +[Valid][5][UTT][ MSE = 0.2423 | LCC = 0.8673 | SRCC = 0.8659 ] +INFO:root:Average loss=0.149453118061408 +INFO:root: +[Valid][6][UTT][ MSE = 0.2365 | LCC = 0.8697 | SRCC = 0.8670 ] +INFO:root:Average loss=0.14060711106046653 +INFO:root: +[Valid][7][UTT][ MSE = 0.2228 | LCC = 0.8703 | SRCC = 0.8691 ] +INFO:root:Average loss=0.12718118679447052 +INFO:root: +[Valid][8][UTT][ MSE = 0.2252 | LCC = 0.8651 | SRCC = 0.8647 ] +INFO:root:Average loss=0.11635448067234112 +INFO:root: +[Valid][9][UTT][ MSE = 0.2281 | LCC = 0.8653 | SRCC = 0.8645 ] +INFO:root:Average loss=0.11618690899549386 +INFO:root: +[Valid][10][UTT][ MSE = 0.2361 | LCC = 0.8672 | SRCC = 0.8648 ] +INFO:root:Average loss=0.10554092520704636 +INFO:root: +[Valid][11][UTT][ MSE = 0.2388 | LCC = 0.8632 | SRCC = 0.8604 ] +INFO:root:Average loss=0.09794734174815509 +INFO:root: +[Valid][12][UTT][ MSE = 0.2234 | LCC = 0.8660 | SRCC = 0.8645 ] +INFO:root:Average loss=0.08685126260687144 +INFO:root: +[Valid][13][UTT][ MSE = 0.2470 | LCC = 0.8588 | SRCC = 0.8565 ] +INFO:root:Average loss=0.08246368585297695 +INFO:root: +[Valid][14][UTT][ MSE = 0.2213 | LCC = 0.8626 | SRCC = 0.8616 ] +INFO:root:Average loss=0.07784797337192756 +INFO:root: +[Valid][15][UTT][ MSE = 0.2376 | LCC = 0.8566 | SRCC = 0.8549 ] +INFO:root:Average loss=0.07904398852051833 +INFO:root: +[Valid][16][UTT][ MSE = 0.2377 | LCC = 0.8612 | SRCC = 0.8577 ] +INFO:root:Average loss=0.07124616430164912 +INFO:root: +[Valid][17][UTT][ MSE = 0.2382 | LCC = 0.8560 | SRCC = 0.8534 ] +INFO:root:Average loss=0.07084866732549973 +INFO:root: +[Valid][18][UTT][ MSE = 0.2335 | LCC = 0.8575 | SRCC = 0.8548 ] +INFO:root:Average loss=0.06428851316181514 +INFO:root: +[Valid][19][UTT][ MSE = 0.2469 | LCC = 0.8552 | SRCC = 0.8531 ] +INFO:root:Average loss=0.06520013859829842 +INFO:root: +[Valid][20][UTT][ MSE = 0.2450 | LCC = 0.8557 | SRCC = 0.8521 ] +INFO:root:Average loss=0.05688501163744009 +INFO:root: +[Valid][21][UTT][ MSE = 0.2494 | LCC = 0.8538 | SRCC = 0.8526 ] +INFO:root:Average loss=0.05489346342018017 +INFO:root: +[Valid][22][UTT][ MSE = 0.2396 | LCC = 0.8524 | SRCC = 0.8494 ] +INFO:root:Average loss=0.052350979966995045 +INFO:root: +[Valid][23][UTT][ MSE = 0.2437 | LCC = 0.8544 | SRCC = 0.8517 ] +INFO:root:Average loss=0.05123212988464496 +INFO:root: +[Valid][24][UTT][ MSE = 0.2669 | LCC = 0.8499 | SRCC = 0.8450 ] +INFO:root:Average loss=0.05136392451822758 +INFO:root: +[Valid][25][UTT][ MSE = 0.2431 | LCC = 0.8533 | SRCC = 0.8509 ] +INFO:root:Average loss=0.046154085642252214 +INFO:root: +[Valid][26][UTT][ MSE = 0.2549 | LCC = 0.8500 | SRCC = 0.8474 ] +INFO:root:Average loss=0.04427948861550062 +INFO:root: +[Valid][27][UTT][ MSE = 0.2333 | LCC = 0.8569 | SRCC = 0.8535 ] +INFO:root:Average loss=0.04695770388039259 +INFO:root: +[Valid][28][UTT][ MSE = 0.2421 | LCC = 0.8526 | SRCC = 0.8494 ] +INFO:root:Average loss=0.04160770086141733 +INFO:root: +[Valid][29][UTT][ MSE = 0.2389 | LCC = 0.8542 | SRCC = 0.8510 ] +INFO:root:Average loss=0.04196249428563393 +INFO:root: +[Valid][30][UTT][ MSE = 0.2306 | LCC = 0.8552 | SRCC = 0.8529 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2419 | LCC = 0.8679 | SRCC = 0.8667 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e7fde7bef5a3fc4b3894c7c5423659bb6929ff11 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5d7c76b89a8f62934811ea3cba32861d04ddbd22156893d6301e694f720802fe +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d4dc7a2a090f733dcb5ddf04cd4966b53d5879a5 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:79b94d64ea8942b3129fb8a5f5087d077670c0db09ab4cc1643e7ac437726cd7 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..cecbf02e0531a2d76c4286c820ab393d1f476ed1 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer5/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=0.3595262229060515 +INFO:root: +[Valid][0][UTT][ MSE = 0.3454 | LCC = 0.8393 | SRCC = 0.8380 ] +INFO:root:Average loss=0.22473877171675363 +INFO:root: +[Valid][1][UTT][ MSE = 0.3070 | LCC = 0.8618 | SRCC = 0.8574 ] +INFO:root:Average loss=0.19946333699119398 +INFO:root: +[Valid][2][UTT][ MSE = 0.2556 | LCC = 0.8678 | SRCC = 0.8656 ] +INFO:root:Average loss=0.17979559923211733 +INFO:root: +[Valid][3][UTT][ MSE = 0.2698 | LCC = 0.8679 | SRCC = 0.8679 ] +INFO:root:Average loss=0.1718667365419559 +INFO:root: +[Valid][4][UTT][ MSE = 0.2494 | LCC = 0.8696 | SRCC = 0.8677 ] +INFO:root:Average loss=0.15568304434418678 +INFO:root: +[Valid][5][UTT][ MSE = 0.2519 | LCC = 0.8696 | SRCC = 0.8676 ] +INFO:root:Average loss=0.1490375858086806 +INFO:root: +[Valid][6][UTT][ MSE = 0.2497 | LCC = 0.8683 | SRCC = 0.8675 ] +INFO:root:Average loss=0.13177544203324196 +INFO:root: +[Valid][7][UTT][ MSE = 0.2251 | LCC = 0.8676 | SRCC = 0.8678 ] +INFO:root:Average loss=0.12794954052720314 +INFO:root: +[Valid][8][UTT][ MSE = 0.2442 | LCC = 0.8672 | SRCC = 0.8656 ] +INFO:root:Average loss=0.11589938258895507 +INFO:root: +[Valid][9][UTT][ MSE = 0.2516 | LCC = 0.8621 | SRCC = 0.8606 ] +INFO:root:Average loss=0.11088716773650585 +INFO:root: +[Valid][10][UTT][ MSE = 0.2397 | LCC = 0.8630 | SRCC = 0.8617 ] +INFO:root:Average loss=0.1023254431784153 +INFO:root: +[Valid][11][UTT][ MSE = 0.2553 | LCC = 0.8625 | SRCC = 0.8611 ] +INFO:root:Average loss=0.10059604215889405 +INFO:root: +[Valid][12][UTT][ MSE = 0.2716 | LCC = 0.8608 | SRCC = 0.8589 ] +INFO:root:Average loss=0.0900332535115572 +INFO:root: +[Valid][13][UTT][ MSE = 0.2597 | LCC = 0.8605 | SRCC = 0.8606 ] +INFO:root:Average loss=0.08557469636583939 +INFO:root: +[Valid][14][UTT][ MSE = 0.2416 | LCC = 0.8593 | SRCC = 0.8587 ] +INFO:root:Average loss=0.07883443931738536 +INFO:root: +[Valid][15][UTT][ MSE = 0.2638 | LCC = 0.8571 | SRCC = 0.8561 ] +INFO:root:Average loss=0.07571529511075753 +INFO:root: +[Valid][16][UTT][ MSE = 0.2318 | LCC = 0.8611 | SRCC = 0.8600 ] +INFO:root:Average loss=0.07118511639344387 +INFO:root: +[Valid][17][UTT][ MSE = 0.2670 | LCC = 0.8594 | SRCC = 0.8571 ] +INFO:root:Average loss=0.06968598913114804 +INFO:root: +[Valid][18][UTT][ MSE = 0.2449 | LCC = 0.8626 | SRCC = 0.8624 ] +INFO:root:Average loss=0.0640985907939 +INFO:root: +[Valid][19][UTT][ MSE = 0.2633 | LCC = 0.8594 | SRCC = 0.8590 ] +INFO:root:Average loss=0.06558285247630034 +INFO:root: +[Valid][20][UTT][ MSE = 0.2800 | LCC = 0.8577 | SRCC = 0.8585 ] +INFO:root:Average loss=0.06112644702005081 +INFO:root: +[Valid][21][UTT][ MSE = 0.2432 | LCC = 0.8591 | SRCC = 0.8589 ] +INFO:root:Average loss=0.05628125264476507 +INFO:root: +[Valid][22][UTT][ MSE = 0.2348 | LCC = 0.8565 | SRCC = 0.8544 ] +INFO:root:Average loss=0.05396719706746248 +INFO:root: +[Valid][23][UTT][ MSE = 0.2539 | LCC = 0.8575 | SRCC = 0.8556 ] +INFO:root:Average loss=0.055747028034275926 +INFO:root: +[Valid][24][UTT][ MSE = 0.2592 | LCC = 0.8593 | SRCC = 0.8571 ] +INFO:root:Average loss=0.05354839532325665 +INFO:root: +[Valid][25][UTT][ MSE = 0.2508 | LCC = 0.8583 | SRCC = 0.8570 ] +INFO:root:Average loss=0.05082804915041496 +INFO:root: +[Valid][26][UTT][ MSE = 0.2587 | LCC = 0.8588 | SRCC = 0.8573 ] +INFO:root:Average loss=0.04764386868247619 +INFO:root: +[Valid][27][UTT][ MSE = 0.2321 | LCC = 0.8594 | SRCC = 0.8577 ] +INFO:root:Average loss=0.04476834766757794 +INFO:root: +[Valid][28][UTT][ MSE = 0.2474 | LCC = 0.8574 | SRCC = 0.8549 ] +INFO:root:Average loss=0.043770779927189536 +INFO:root: +[Valid][29][UTT][ MSE = 0.2332 | LCC = 0.8583 | SRCC = 0.8570 ] +INFO:root:Average loss=0.043522827207851104 +INFO:root: +[Valid][30][UTT][ MSE = 0.2533 | LCC = 0.8573 | SRCC = 0.8557 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2421 | LCC = 0.8651 | SRCC = 0.8631 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..fcacaa795b527880b16d87a45a5f6bca4e2e6aa9 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c0258688d2596f449ad62816e4b44cb18bc27c4db1195f59778206cbbc4c4b6 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..3050856079db5576b691cdb0dffe77e4b769ec40 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24c3a0c98fe36d016509d4b11dca77315675d46f976508d59e2aef2cd68ee550 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e4f0ec8f9647f83737ad359767a0a435a2fe316f --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/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=0.37662898347927976 +INFO:root: +[Valid][0][UTT][ MSE = 0.3374 | LCC = 0.8416 | SRCC = 0.8428 ] +INFO:root:Average loss=0.22872272076515052 +INFO:root: +[Valid][1][UTT][ MSE = 0.2906 | LCC = 0.8625 | SRCC = 0.8619 ] +INFO:root:Average loss=0.20082451240756574 +INFO:root: +[Valid][2][UTT][ MSE = 0.2407 | LCC = 0.8717 | SRCC = 0.8706 ] +INFO:root:Average loss=0.19368504828367478 +INFO:root: +[Valid][3][UTT][ MSE = 0.2613 | LCC = 0.8726 | SRCC = 0.8704 ] +INFO:root:Average loss=0.17128495757396406 +INFO:root: +[Valid][4][UTT][ MSE = 0.2457 | LCC = 0.8743 | SRCC = 0.8712 ] +INFO:root:Average loss=0.16314771570838416 +INFO:root: +[Valid][5][UTT][ MSE = 0.2583 | LCC = 0.8717 | SRCC = 0.8701 ] +INFO:root:Average loss=0.15198849074733564 +INFO:root: +[Valid][6][UTT][ MSE = 0.2530 | LCC = 0.8726 | SRCC = 0.8707 ] +INFO:root:Average loss=0.13924745155068544 +INFO:root: +[Valid][7][UTT][ MSE = 0.2590 | LCC = 0.8699 | SRCC = 0.8681 ] +INFO:root:Average loss=0.12893828100118881 +INFO:root: +[Valid][8][UTT][ MSE = 0.2512 | LCC = 0.8645 | SRCC = 0.8626 ] +INFO:root:Average loss=0.12568333783210853 +INFO:root: +[Valid][9][UTT][ MSE = 0.2660 | LCC = 0.8660 | SRCC = 0.8637 ] +INFO:root:Average loss=0.12102874072316365 +INFO:root: +[Valid][10][UTT][ MSE = 0.2498 | LCC = 0.8679 | SRCC = 0.8656 ] +INFO:root:Average loss=0.10584202905495961 +INFO:root: +[Valid][11][UTT][ MSE = 0.2439 | LCC = 0.8640 | SRCC = 0.8621 ] +INFO:root:Average loss=0.10930917693827397 +INFO:root: +[Valid][12][UTT][ MSE = 0.2740 | LCC = 0.8613 | SRCC = 0.8600 ] +INFO:root:Average loss=0.09695326403165475 +INFO:root: +[Valid][13][UTT][ MSE = 0.2518 | LCC = 0.8635 | SRCC = 0.8621 ] +INFO:root:Average loss=0.08869052525514211 +INFO:root: +[Valid][14][UTT][ MSE = 0.2473 | LCC = 0.8583 | SRCC = 0.8566 ] +INFO:root:Average loss=0.0822978658744922 +INFO:root: +[Valid][15][UTT][ MSE = 0.2488 | LCC = 0.8543 | SRCC = 0.8518 ] +INFO:root:Average loss=0.07713037285094078 +INFO:root: +[Valid][16][UTT][ MSE = 0.2390 | LCC = 0.8626 | SRCC = 0.8611 ] +INFO:root:Average loss=0.07291681515291715 +INFO:root: +[Valid][17][UTT][ MSE = 0.2408 | LCC = 0.8632 | SRCC = 0.8611 ] +INFO:root:Average loss=0.06736221408041623 +INFO:root: +[Valid][18][UTT][ MSE = 0.2465 | LCC = 0.8571 | SRCC = 0.8561 ] +INFO:root:Average loss=0.06570938473137525 +INFO:root: +[Valid][19][UTT][ MSE = 0.2447 | LCC = 0.8584 | SRCC = 0.8571 ] +INFO:root:Average loss=0.06754968138650441 +INFO:root: +[Valid][20][UTT][ MSE = 0.2503 | LCC = 0.8571 | SRCC = 0.8561 ] +INFO:root:Average loss=0.06006206347583196 +INFO:root: +[Valid][21][UTT][ MSE = 0.2384 | LCC = 0.8598 | SRCC = 0.8579 ] +INFO:root:Average loss=0.05699872808196606 +INFO:root: +[Valid][22][UTT][ MSE = 0.2530 | LCC = 0.8540 | SRCC = 0.8517 ] +INFO:root:Average loss=0.05451706744348391 +INFO:root: +[Valid][23][UTT][ MSE = 0.2399 | LCC = 0.8597 | SRCC = 0.8584 ] +INFO:root:Average loss=0.05882871007690063 +INFO:root: +[Valid][24][UTT][ MSE = 0.2506 | LCC = 0.8547 | SRCC = 0.8538 ] +INFO:root:Average loss=0.052178205349124394 +INFO:root: +[Valid][25][UTT][ MSE = 0.2619 | LCC = 0.8545 | SRCC = 0.8534 ] +INFO:root:Average loss=0.048751841681316875 +INFO:root: +[Valid][26][UTT][ MSE = 0.2582 | LCC = 0.8536 | SRCC = 0.8523 ] +INFO:root:Average loss=0.04571998587403542 +INFO:root: +[Valid][27][UTT][ MSE = 0.2628 | LCC = 0.8529 | SRCC = 0.8516 ] +INFO:root:Average loss=0.046422197173039116 +INFO:root: +[Valid][28][UTT][ MSE = 0.2464 | LCC = 0.8589 | SRCC = 0.8577 ] +INFO:root:Average loss=0.04452958416480284 +INFO:root: +[Valid][29][UTT][ MSE = 0.2601 | LCC = 0.8495 | SRCC = 0.8482 ] +INFO:root:Average loss=0.04420116956894978 +INFO:root: +[Valid][30][UTT][ MSE = 0.2668 | LCC = 0.8534 | SRCC = 0.8536 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2483 | LCC = 0.8663 | SRCC = 0.8664 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..faffd4fa99407e8f9e7c30819573ebb51f8f89a0 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e21e5391eeaa8768e3082479b6fb9d18a024e95c50e9318a4801745db58d0200 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e03078e84501b77dcb0b1dcf2fa5768101a50f72 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d4b9406221723fbdceeea1bd9fa8bc3bd6aa1d7d43719729c59e1607c36d33d2 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e923b3d4699ffa2c9dc0f23d263a75922a8a8627 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/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=0.38016679061528963 +INFO:root: +[Valid][0][UTT][ MSE = 0.3300 | LCC = 0.8364 | SRCC = 0.8364 ] +INFO:root:Average loss=0.23010518764838195 +INFO:root: +[Valid][1][UTT][ MSE = 0.2657 | LCC = 0.8567 | SRCC = 0.8573 ] +INFO:root:Average loss=0.2049266875554354 +INFO:root: +[Valid][2][UTT][ MSE = 0.2541 | LCC = 0.8651 | SRCC = 0.8645 ] +INFO:root:Average loss=0.19715865090107307 +INFO:root: +[Valid][3][UTT][ MSE = 0.2644 | LCC = 0.8671 | SRCC = 0.8659 ] +INFO:root:Average loss=0.18033413025431144 +INFO:root: +[Valid][4][UTT][ MSE = 0.2626 | LCC = 0.8670 | SRCC = 0.8660 ] +INFO:root:Average loss=0.1659525410296061 +INFO:root: +[Valid][5][UTT][ MSE = 0.2505 | LCC = 0.8642 | SRCC = 0.8631 ] +INFO:root:Average loss=0.15832179956711256 +INFO:root: +[Valid][6][UTT][ MSE = 0.2916 | LCC = 0.8658 | SRCC = 0.8641 ] +INFO:root:Average loss=0.14731614540020624 +INFO:root: +[Valid][7][UTT][ MSE = 0.2548 | LCC = 0.8631 | SRCC = 0.8609 ] +INFO:root:Average loss=0.14135649523291832 +INFO:root: +[Valid][8][UTT][ MSE = 0.2460 | LCC = 0.8627 | SRCC = 0.8624 ] +INFO:root:Average loss=0.1258084043287314 +INFO:root: +[Valid][9][UTT][ MSE = 0.2489 | LCC = 0.8647 | SRCC = 0.8641 ] +INFO:root:Average loss=0.12144845714553809 +INFO:root: +[Valid][10][UTT][ MSE = 0.2380 | LCC = 0.8613 | SRCC = 0.8611 ] +INFO:root:Average loss=0.11188555957797246 +INFO:root: +[Valid][11][UTT][ MSE = 0.2367 | LCC = 0.8628 | SRCC = 0.8616 ] +INFO:root:Average loss=0.11534176566279851 +INFO:root: +[Valid][12][UTT][ MSE = 0.3187 | LCC = 0.8545 | SRCC = 0.8569 ] +INFO:root:Average loss=0.10706581075031024 +INFO:root: +[Valid][13][UTT][ MSE = 0.2603 | LCC = 0.8588 | SRCC = 0.8584 ] +INFO:root:Average loss=0.09298180726667245 +INFO:root: +[Valid][14][UTT][ MSE = 0.2568 | LCC = 0.8567 | SRCC = 0.8558 ] +INFO:root:Average loss=0.0894466707817255 +INFO:root: +[Valid][15][UTT][ MSE = 0.2491 | LCC = 0.8539 | SRCC = 0.8536 ] +INFO:root:Average loss=0.08470619020935817 +INFO:root: +[Valid][16][UTT][ MSE = 0.2377 | LCC = 0.8560 | SRCC = 0.8557 ] +INFO:root:Average loss=0.07915342040359974 +INFO:root: +[Valid][17][UTT][ MSE = 0.2590 | LCC = 0.8519 | SRCC = 0.8507 ] +INFO:root:Average loss=0.07489363615138409 +INFO:root: +[Valid][18][UTT][ MSE = 0.2538 | LCC = 0.8488 | SRCC = 0.8489 ] +INFO:root:Average loss=0.06871900038841443 +INFO:root: +[Valid][19][UTT][ MSE = 0.2479 | LCC = 0.8518 | SRCC = 0.8504 ] +INFO:root:Average loss=0.06596600106702401 +INFO:root: +[Valid][20][UTT][ MSE = 0.2727 | LCC = 0.8491 | SRCC = 0.8475 ] +INFO:root:Average loss=0.06183138800164064 +INFO:root: +[Valid][21][UTT][ MSE = 0.2516 | LCC = 0.8508 | SRCC = 0.8488 ] +INFO:root:Average loss=0.0613967564721138 +INFO:root: +[Valid][22][UTT][ MSE = 0.2848 | LCC = 0.8506 | SRCC = 0.8494 ] +INFO:root:Average loss=0.058816186271798916 +INFO:root: +[Valid][23][UTT][ MSE = 0.2438 | LCC = 0.8542 | SRCC = 0.8535 ] +INFO:root:Average loss=0.06132974523382309 +INFO:root: +[Valid][24][UTT][ MSE = 0.2485 | LCC = 0.8527 | SRCC = 0.8512 ] +INFO:root:Average loss=0.05529741570353508 +INFO:root: +[Valid][25][UTT][ MSE = 0.2550 | LCC = 0.8490 | SRCC = 0.8478 ] +INFO:root:Average loss=0.05354660732719379 +INFO:root: +[Valid][26][UTT][ MSE = 0.2702 | LCC = 0.8464 | SRCC = 0.8448 ] +INFO:root:Average loss=0.050062165571711004 +INFO:root: +[Valid][27][UTT][ MSE = 0.2600 | LCC = 0.8474 | SRCC = 0.8454 ] +INFO:root:Average loss=0.048995111089868426 +INFO:root: +[Valid][28][UTT][ MSE = 0.2556 | LCC = 0.8478 | SRCC = 0.8465 ] +INFO:root:Average loss=0.04909681495374594 +INFO:root: +[Valid][29][UTT][ MSE = 0.2875 | LCC = 0.8447 | SRCC = 0.8436 ] +INFO:root:Average loss=0.049055086639829174 +INFO:root: +[Valid][30][UTT][ MSE = 0.2527 | LCC = 0.8495 | SRCC = 0.8474 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2617 | LCC = 0.8653 | SRCC = 0.8671 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f58b96049eb10566897927ebfc0053b7db4d9ea9 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4326f489bc22e413d12a9c6ed9806ff78dacd4319db2112c67b81e870264352c +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ff1153f1fbebece34d57d1117956b8423debf2ea --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f096b4a445c42faecae050a724202e60b1ccb3d01b414c61c12a6296fe8012ba +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..04556dd4eb52e95f001d09b6ff3caea8df9a9ead --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/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=0.33275369707590496 +INFO:root: +[Valid][0][UTT][ MSE = 0.2967 | LCC = 0.8377 | SRCC = 0.8437 ] +INFO:root:Average loss=0.21851457359317022 +INFO:root: +[Valid][1][UTT][ MSE = 0.2690 | LCC = 0.8619 | SRCC = 0.8600 ] +INFO:root:Average loss=0.20862483156797212 +INFO:root: +[Valid][2][UTT][ MSE = 0.2630 | LCC = 0.8681 | SRCC = 0.8656 ] +INFO:root:Average loss=0.19034708176667875 +INFO:root: +[Valid][3][UTT][ MSE = 0.2288 | LCC = 0.8688 | SRCC = 0.8681 ] +INFO:root:Average loss=0.17898023462830445 +INFO:root: +[Valid][4][UTT][ MSE = 0.2789 | LCC = 0.8685 | SRCC = 0.8670 ] +INFO:root:Average loss=0.17256294486996454 +INFO:root: +[Valid][5][UTT][ MSE = 0.2481 | LCC = 0.8714 | SRCC = 0.8698 ] +INFO:root:Average loss=0.15613886752189735 +INFO:root: +[Valid][6][UTT][ MSE = 0.2566 | LCC = 0.8695 | SRCC = 0.8671 ] +INFO:root:Average loss=0.14984425166860604 +INFO:root: +[Valid][7][UTT][ MSE = 0.2659 | LCC = 0.8660 | SRCC = 0.8658 ] +INFO:root:Average loss=0.1322043136908458 +INFO:root: +[Valid][8][UTT][ MSE = 0.2269 | LCC = 0.8663 | SRCC = 0.8654 ] +INFO:root:Average loss=0.13058045535133436 +INFO:root: +[Valid][9][UTT][ MSE = 0.2421 | LCC = 0.8664 | SRCC = 0.8654 ] +INFO:root:Average loss=0.11448775508846992 +INFO:root: +[Valid][10][UTT][ MSE = 0.2315 | LCC = 0.8629 | SRCC = 0.8622 ] +INFO:root:Average loss=0.11448502874909303 +INFO:root: +[Valid][11][UTT][ MSE = 0.2442 | LCC = 0.8635 | SRCC = 0.8623 ] +INFO:root:Average loss=0.10621776021061799 +INFO:root: +[Valid][12][UTT][ MSE = 0.2356 | LCC = 0.8627 | SRCC = 0.8625 ] +INFO:root:Average loss=0.09489946368222053 +INFO:root: +[Valid][13][UTT][ MSE = 0.2353 | LCC = 0.8608 | SRCC = 0.8596 ] +INFO:root:Average loss=0.08694203446308772 +INFO:root: +[Valid][14][UTT][ MSE = 0.2293 | LCC = 0.8613 | SRCC = 0.8608 ] +INFO:root:Average loss=0.08192989340004249 +INFO:root: +[Valid][15][UTT][ MSE = 0.2208 | LCC = 0.8625 | SRCC = 0.8629 ] +INFO:root:Average loss=0.07956037273964821 +INFO:root: +[Valid][16][UTT][ MSE = 0.2360 | LCC = 0.8580 | SRCC = 0.8582 ] +INFO:root:Average loss=0.07526221828391919 +INFO:root: +[Valid][17][UTT][ MSE = 0.2346 | LCC = 0.8611 | SRCC = 0.8595 ] +INFO:root:Average loss=0.07061882531986786 +INFO:root: +[Valid][18][UTT][ MSE = 0.2632 | LCC = 0.8559 | SRCC = 0.8551 ] +INFO:root:Average loss=0.06583132317815071 +INFO:root: +[Valid][19][UTT][ MSE = 0.2294 | LCC = 0.8584 | SRCC = 0.8580 ] +INFO:root:Average loss=0.06522104339912915 +INFO:root: +[Valid][20][UTT][ MSE = 0.2468 | LCC = 0.8564 | SRCC = 0.8571 ] +INFO:root:Average loss=0.05993116838045609 +INFO:root: +[Valid][21][UTT][ MSE = 0.2471 | LCC = 0.8557 | SRCC = 0.8547 ] +INFO:root:Average loss=0.05827635168456114 +INFO:root: +[Valid][22][UTT][ MSE = 0.2402 | LCC = 0.8563 | SRCC = 0.8553 ] +INFO:root:Average loss=0.05669754074934202 +INFO:root: +[Valid][23][UTT][ MSE = 0.2419 | LCC = 0.8537 | SRCC = 0.8538 ] +INFO:root:Average loss=0.052084068552805826 +INFO:root: +[Valid][24][UTT][ MSE = 0.2409 | LCC = 0.8550 | SRCC = 0.8535 ] +INFO:root:Average loss=0.05173896379673328 +INFO:root: +[Valid][25][UTT][ MSE = 0.2291 | LCC = 0.8554 | SRCC = 0.8549 ] +INFO:root:Average loss=0.050798057936705075 +INFO:root: +[Valid][26][UTT][ MSE = 0.2378 | LCC = 0.8544 | SRCC = 0.8532 ] +INFO:root:Average loss=0.047485315025998995 +INFO:root: +[Valid][27][UTT][ MSE = 0.2410 | LCC = 0.8533 | SRCC = 0.8526 ] +INFO:root:Average loss=0.044455126739847355 +INFO:root: +[Valid][28][UTT][ MSE = 0.2414 | LCC = 0.8548 | SRCC = 0.8551 ] +INFO:root:Average loss=0.04412764418297089 +INFO:root: +[Valid][29][UTT][ MSE = 0.2613 | LCC = 0.8497 | SRCC = 0.8504 ] +INFO:root:Average loss=0.04266225548986441 +INFO:root: +[Valid][30][UTT][ MSE = 0.2320 | LCC = 0.8568 | SRCC = 0.8572 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2678 | LCC = 0.8654 | SRCC = 0.8673 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c11b01255fa922b84bae01c174eb0599ca834597 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:71d62d6122cbc7092d151da060046df9edef03e36449900d7872983c50523456 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..49e15f079ab2ac5814e1f15f388ead56c23a7909 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:47309ec4b07b3e46d2df62efd97ce243d0ce613c2398d7b4e38e94deba9b42f3 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ee9e67f8d6779286d632c622d082f21be651a11b --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/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=0.5448146605720887 +INFO:root: +[Valid][0][UTT][ MSE = 0.3526 | LCC = 0.8140 | SRCC = 0.8258 ] +INFO:root:Average loss=0.24388784284775072 +INFO:root: +[Valid][1][UTT][ MSE = 0.2695 | LCC = 0.8466 | SRCC = 0.8514 ] +INFO:root:Average loss=0.2123973863438154 +INFO:root: +[Valid][2][UTT][ MSE = 0.2954 | LCC = 0.8666 | SRCC = 0.8630 ] +INFO:root:Average loss=0.18696158054547432 +INFO:root: +[Valid][3][UTT][ MSE = 0.2324 | LCC = 0.8681 | SRCC = 0.8683 ] +INFO:root:Average loss=0.18060644544087923 +INFO:root: +[Valid][4][UTT][ MSE = 0.2517 | LCC = 0.8701 | SRCC = 0.8671 ] +INFO:root:Average loss=0.16954588250089914 +INFO:root: +[Valid][5][UTT][ MSE = 0.2384 | LCC = 0.8674 | SRCC = 0.8672 ] +INFO:root:Average loss=0.15501732379198074 +INFO:root: +[Valid][6][UTT][ MSE = 0.2615 | LCC = 0.8664 | SRCC = 0.8638 ] +INFO:root:Average loss=0.145430438602582 +INFO:root: +[Valid][7][UTT][ MSE = 0.2568 | LCC = 0.8656 | SRCC = 0.8641 ] +INFO:root:Average loss=0.13371641933918 +INFO:root: +[Valid][8][UTT][ MSE = 0.2275 | LCC = 0.8638 | SRCC = 0.8644 ] +INFO:root:Average loss=0.1187549609786425 +INFO:root: +[Valid][9][UTT][ MSE = 0.2292 | LCC = 0.8632 | SRCC = 0.8622 ] +INFO:root:Average loss=0.1153817851191912 +INFO:root: +[Valid][10][UTT][ MSE = 0.2453 | LCC = 0.8598 | SRCC = 0.8604 ] +INFO:root:Average loss=0.10624120852504021 +INFO:root: +[Valid][11][UTT][ MSE = 0.2291 | LCC = 0.8573 | SRCC = 0.8579 ] +INFO:root:Average loss=0.09952080722611684 +INFO:root: +[Valid][12][UTT][ MSE = 0.2290 | LCC = 0.8615 | SRCC = 0.8599 ] +INFO:root:Average loss=0.09190802285686517 +INFO:root: +[Valid][13][UTT][ MSE = 0.2290 | LCC = 0.8603 | SRCC = 0.8603 ] +INFO:root:Average loss=0.0850571755033273 +INFO:root: +[Valid][14][UTT][ MSE = 0.2272 | LCC = 0.8571 | SRCC = 0.8571 ] +INFO:root:Average loss=0.0855944188646017 +INFO:root: +[Valid][15][UTT][ MSE = 0.2342 | LCC = 0.8552 | SRCC = 0.8550 ] +INFO:root:Average loss=0.07442066398186561 +INFO:root: +[Valid][16][UTT][ MSE = 0.2453 | LCC = 0.8563 | SRCC = 0.8539 ] +INFO:root:Average loss=0.07493224281531113 +INFO:root: +[Valid][17][UTT][ MSE = 0.2330 | LCC = 0.8595 | SRCC = 0.8602 ] +INFO:root:Average loss=0.07165318751373352 +INFO:root: +[Valid][18][UTT][ MSE = 0.2284 | LCC = 0.8580 | SRCC = 0.8574 ] +INFO:root:Average loss=0.07231290413974187 +INFO:root: +[Valid][19][UTT][ MSE = 0.2339 | LCC = 0.8548 | SRCC = 0.8542 ] +INFO:root:Average loss=0.06138812922514402 +INFO:root: +[Valid][20][UTT][ MSE = 0.2394 | LCC = 0.8549 | SRCC = 0.8547 ] +INFO:root:Average loss=0.05940250617762407 +INFO:root: +[Valid][21][UTT][ MSE = 0.2343 | LCC = 0.8557 | SRCC = 0.8554 ] +INFO:root:Average loss=0.05733059680996797 +INFO:root: +[Valid][22][UTT][ MSE = 0.2373 | LCC = 0.8553 | SRCC = 0.8555 ] +INFO:root:Average loss=0.06024368785512753 +INFO:root: +[Valid][23][UTT][ MSE = 0.2304 | LCC = 0.8589 | SRCC = 0.8586 ] +INFO:root:Average loss=0.05112675415017666 +INFO:root: +[Valid][24][UTT][ MSE = 0.2320 | LCC = 0.8537 | SRCC = 0.8539 ] +INFO:root:Average loss=0.04913635606853626 +INFO:root: +[Valid][25][UTT][ MSE = 0.2373 | LCC = 0.8551 | SRCC = 0.8537 ] +INFO:root:Average loss=0.047845299331805646 +INFO:root: +[Valid][26][UTT][ MSE = 0.2381 | LCC = 0.8536 | SRCC = 0.8530 ] +INFO:root:Average loss=0.04648768032590548 +INFO:root: +[Valid][27][UTT][ MSE = 0.2315 | LCC = 0.8544 | SRCC = 0.8545 ] +INFO:root:Average loss=0.04346613927433888 +INFO:root: +[Valid][28][UTT][ MSE = 0.2375 | LCC = 0.8532 | SRCC = 0.8517 ] +INFO:root:Average loss=0.045977200668018595 +INFO:root: +[Valid][29][UTT][ MSE = 0.2455 | LCC = 0.8507 | SRCC = 0.8499 ] +INFO:root:Average loss=0.0445131384409391 +INFO:root: +[Valid][30][UTT][ MSE = 0.2319 | LCC = 0.8537 | SRCC = 0.8534 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2407 | LCC = 0.8687 | SRCC = 0.8695 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c5c38e1fed006e0e8f09afc70b32405afe9030ef --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0532cb3afae100fcbe15f1053fefc3aafb459a0cfa0dc636a3ccbd75914f77cb +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..57a7c19b98355c50ae4e0afe1535ecfc8fd6bbe9 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:baa7f03522d5938cf6129bd4c348627e91b02d7722ff1f21239d6c69734f7288 +size 2749638 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2fc16835d857d07a7a0d1bb4cf2ef4b185e72a86 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer6/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=0.3796987508734067 +INFO:root: +[Valid][0][UTT][ MSE = 0.2898 | LCC = 0.8306 | SRCC = 0.8350 ] +INFO:root:Average loss=0.2322843380463429 +INFO:root: +[Valid][1][UTT][ MSE = 0.3115 | LCC = 0.8552 | SRCC = 0.8533 ] +INFO:root:Average loss=0.20829598873089522 +INFO:root: +[Valid][2][UTT][ MSE = 0.2656 | LCC = 0.8629 | SRCC = 0.8603 ] +INFO:root:Average loss=0.19324811433370298 +INFO:root: +[Valid][3][UTT][ MSE = 0.2593 | LCC = 0.8643 | SRCC = 0.8618 ] +INFO:root:Average loss=0.17637640858689943 +INFO:root: +[Valid][4][UTT][ MSE = 0.2854 | LCC = 0.8618 | SRCC = 0.8609 ] +INFO:root:Average loss=0.1660911986270012 +INFO:root: +[Valid][5][UTT][ MSE = 0.2607 | LCC = 0.8654 | SRCC = 0.8647 ] +INFO:root:Average loss=0.150049786346081 +INFO:root: +[Valid][6][UTT][ MSE = 0.2497 | LCC = 0.8634 | SRCC = 0.8626 ] +INFO:root:Average loss=0.14749946187321955 +INFO:root: +[Valid][7][UTT][ MSE = 0.2655 | LCC = 0.8618 | SRCC = 0.8618 ] +INFO:root:Average loss=0.13700921843067193 +INFO:root: +[Valid][8][UTT][ MSE = 0.2522 | LCC = 0.8604 | SRCC = 0.8591 ] +INFO:root:Average loss=0.12604393972418246 +INFO:root: +[Valid][9][UTT][ MSE = 0.2378 | LCC = 0.8595 | SRCC = 0.8594 ] +INFO:root:Average loss=0.11384631464114556 +INFO:root: +[Valid][10][UTT][ MSE = 0.2582 | LCC = 0.8579 | SRCC = 0.8566 ] +INFO:root:Average loss=0.1108385891868518 +INFO:root: +[Valid][11][UTT][ MSE = 0.2557 | LCC = 0.8593 | SRCC = 0.8571 ] +INFO:root:Average loss=0.10535197074596699 +INFO:root: +[Valid][12][UTT][ MSE = 0.2503 | LCC = 0.8562 | SRCC = 0.8565 ] +INFO:root:Average loss=0.09321501516760924 +INFO:root: +[Valid][13][UTT][ MSE = 0.2349 | LCC = 0.8591 | SRCC = 0.8571 ] +INFO:root:Average loss=0.09009855579680358 +INFO:root: +[Valid][14][UTT][ MSE = 0.2394 | LCC = 0.8586 | SRCC = 0.8581 ] +INFO:root:Average loss=0.0828036710810967 +INFO:root: +[Valid][15][UTT][ MSE = 0.2434 | LCC = 0.8563 | SRCC = 0.8570 ] +INFO:root:Average loss=0.07825194084300445 +INFO:root: +[Valid][16][UTT][ MSE = 0.2336 | LCC = 0.8584 | SRCC = 0.8589 ] +INFO:root:Average loss=0.07904966820317966 +INFO:root: +[Valid][17][UTT][ MSE = 0.2343 | LCC = 0.8572 | SRCC = 0.8586 ] +INFO:root:Average loss=0.06892322213985981 +INFO:root: +[Valid][18][UTT][ MSE = 0.2437 | LCC = 0.8561 | SRCC = 0.8554 ] +INFO:root:Average loss=0.0649212462684283 +INFO:root: +[Valid][19][UTT][ MSE = 0.2442 | LCC = 0.8507 | SRCC = 0.8506 ] +INFO:root:Average loss=0.06838943565694186 +INFO:root: +[Valid][20][UTT][ MSE = 0.2458 | LCC = 0.8585 | SRCC = 0.8567 ] +INFO:root:Average loss=0.06170700418834503 +INFO:root: +[Valid][21][UTT][ MSE = 0.2573 | LCC = 0.8492 | SRCC = 0.8470 ] +INFO:root:Average loss=0.05720063026707906 +INFO:root: +[Valid][22][UTT][ MSE = 0.2370 | LCC = 0.8554 | SRCC = 0.8554 ] +INFO:root:Average loss=0.0587134306342938 +INFO:root: +[Valid][23][UTT][ MSE = 0.2271 | LCC = 0.8579 | SRCC = 0.8582 ] +INFO:root:Average loss=0.053077140011084385 +INFO:root: +[Valid][24][UTT][ MSE = 0.2486 | LCC = 0.8541 | SRCC = 0.8553 ] +INFO:root:Average loss=0.05021835018235903 +INFO:root: +[Valid][25][UTT][ MSE = 0.2413 | LCC = 0.8531 | SRCC = 0.8534 ] +INFO:root:Average loss=0.05103945175711161 +INFO:root: +[Valid][26][UTT][ MSE = 0.2532 | LCC = 0.8496 | SRCC = 0.8501 ] +INFO:root:Average loss=0.04693874114981064 +INFO:root: +[Valid][27][UTT][ MSE = 0.2341 | LCC = 0.8533 | SRCC = 0.8539 ] +INFO:root:Average loss=0.04879284038757666 +INFO:root: +[Valid][28][UTT][ MSE = 0.2426 | LCC = 0.8500 | SRCC = 0.8502 ] +INFO:root:Average loss=0.04624080989891902 +INFO:root: +[Valid][29][UTT][ MSE = 0.2697 | LCC = 0.8464 | SRCC = 0.8468 ] +INFO:root:Average loss=0.04596287826410471 +INFO:root: +[Valid][30][UTT][ MSE = 0.2528 | LCC = 0.8516 | SRCC = 0.8514 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2428 | LCC = 0.8695 | SRCC = 0.8692 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7b2ce020f0e9e13b69ebd304412b8e5c957a2311 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ccf773f7d95cd709a2ae82ba92efdbee97683eb98c3adb55a9c8cc23b0f6afb5 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f6a36b2ef24113dd2d25e741c662aa4abd281e5b --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df55219296196a1831da778cf2ed0bcd6da6483363369b060aca81a26119bbd7 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f6530b8393c34628a8e2c9300b959e5b90a8996f --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/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=0.33742717863657534 +INFO:root: +[Valid][0][UTT][ MSE = 0.3357 | LCC = 0.8436 | SRCC = 0.8455 ] +INFO:root:Average loss=0.22780233850845924 +INFO:root: +[Valid][1][UTT][ MSE = 0.2786 | LCC = 0.8628 | SRCC = 0.8598 ] +INFO:root:Average loss=0.2093371739372229 +INFO:root: +[Valid][2][UTT][ MSE = 0.2401 | LCC = 0.8663 | SRCC = 0.8658 ] +INFO:root:Average loss=0.1910275570475138 +INFO:root: +[Valid][3][UTT][ MSE = 0.2911 | LCC = 0.8686 | SRCC = 0.8667 ] +INFO:root:Average loss=0.18114006051268333 +INFO:root: +[Valid][4][UTT][ MSE = 0.2252 | LCC = 0.8733 | SRCC = 0.8717 ] +INFO:root:Average loss=0.16884568572426453 +INFO:root: +[Valid][5][UTT][ MSE = 0.2373 | LCC = 0.8716 | SRCC = 0.8700 ] +INFO:root:Average loss=0.15413505498033303 +INFO:root: +[Valid][6][UTT][ MSE = 0.2467 | LCC = 0.8701 | SRCC = 0.8676 ] +INFO:root:Average loss=0.14714046816031137 +INFO:root: +[Valid][7][UTT][ MSE = 0.2936 | LCC = 0.8721 | SRCC = 0.8692 ] +INFO:root:Average loss=0.13716195485530755 +INFO:root: +[Valid][8][UTT][ MSE = 0.2438 | LCC = 0.8717 | SRCC = 0.8693 ] +INFO:root:Average loss=0.1331830030450454 +INFO:root: +[Valid][9][UTT][ MSE = 0.2359 | LCC = 0.8690 | SRCC = 0.8673 ] +INFO:root:Average loss=0.12008567020679131 +INFO:root: +[Valid][10][UTT][ MSE = 0.2542 | LCC = 0.8685 | SRCC = 0.8667 ] +INFO:root:Average loss=0.11143728374288632 +INFO:root: +[Valid][11][UTT][ MSE = 0.2570 | LCC = 0.8655 | SRCC = 0.8630 ] +INFO:root:Average loss=0.10472652975183266 +INFO:root: +[Valid][12][UTT][ MSE = 0.2617 | LCC = 0.8653 | SRCC = 0.8631 ] +INFO:root:Average loss=0.0989552135937489 +INFO:root: +[Valid][13][UTT][ MSE = 0.2428 | LCC = 0.8666 | SRCC = 0.8651 ] +INFO:root:Average loss=0.09453364786429283 +INFO:root: +[Valid][14][UTT][ MSE = 0.2413 | LCC = 0.8657 | SRCC = 0.8656 ] +INFO:root:Average loss=0.0869763793471532 +INFO:root: +[Valid][15][UTT][ MSE = 0.2446 | LCC = 0.8587 | SRCC = 0.8552 ] +INFO:root:Average loss=0.08471637290830795 +INFO:root: +[Valid][16][UTT][ MSE = 0.2277 | LCC = 0.8639 | SRCC = 0.8626 ] +INFO:root:Average loss=0.07649376563345775 +INFO:root: +[Valid][17][UTT][ MSE = 0.2495 | LCC = 0.8600 | SRCC = 0.8559 ] +INFO:root:Average loss=0.0781239263044718 +INFO:root: +[Valid][18][UTT][ MSE = 0.2988 | LCC = 0.8596 | SRCC = 0.8580 ] +INFO:root:Average loss=0.07254668153249301 +INFO:root: +[Valid][19][UTT][ MSE = 0.2723 | LCC = 0.8569 | SRCC = 0.8518 ] +INFO:root:Average loss=0.06897442458340755 +INFO:root: +[Valid][20][UTT][ MSE = 0.2310 | LCC = 0.8589 | SRCC = 0.8565 ] +INFO:root:Average loss=0.06120508789825134 +INFO:root: +[Valid][21][UTT][ MSE = 0.2240 | LCC = 0.8639 | SRCC = 0.8624 ] +INFO:root:Average loss=0.0624264137676129 +INFO:root: +[Valid][22][UTT][ MSE = 0.2464 | LCC = 0.8609 | SRCC = 0.8583 ] +INFO:root:Average loss=0.05848711256224375 +INFO:root: +[Valid][23][UTT][ MSE = 0.2355 | LCC = 0.8601 | SRCC = 0.8567 ] +INFO:root:Average loss=0.055372774171141476 +INFO:root: +[Valid][24][UTT][ MSE = 0.2322 | LCC = 0.8586 | SRCC = 0.8552 ] +INFO:root:Average loss=0.05495965194243651 +INFO:root: +[Valid][25][UTT][ MSE = 0.2230 | LCC = 0.8585 | SRCC = 0.8573 ] +INFO:root:Average loss=0.05193904004035852 +INFO:root: +[Valid][26][UTT][ MSE = 0.2477 | LCC = 0.8541 | SRCC = 0.8512 ] +INFO:root:Average loss=0.053872195430673085 +INFO:root: +[Valid][27][UTT][ MSE = 0.2231 | LCC = 0.8603 | SRCC = 0.8586 ] +INFO:root:Average loss=0.052220724523067474 +INFO:root: +[Valid][28][UTT][ MSE = 0.2386 | LCC = 0.8576 | SRCC = 0.8540 ] +INFO:root:Average loss=0.04630086819330851 +INFO:root: +[Valid][29][UTT][ MSE = 0.2449 | LCC = 0.8537 | SRCC = 0.8492 ] +INFO:root:Average loss=0.046450643274837576 +INFO:root: +[Valid][30][UTT][ MSE = 0.2405 | LCC = 0.8569 | SRCC = 0.8547 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2370 | LCC = 0.8635 | SRCC = 0.8647 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2a71e8748c4d92644019b51781b6446cda2958c4 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8068b91b871e316a262eef93bc18fc4752fa4714288b19a3ca6e882f706139c0 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..70cf67229703a0b5173ea0493bb31df3933a8908 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cb9fe2998d0745ae9b9635383da4fe50761d7662ac241362deae88a8200f0c3c +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..10fad2126315e39943ef7f9e1b6f20afb81005e0 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/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=0.3640528684243178 +INFO:root: +[Valid][0][UTT][ MSE = 0.3054 | LCC = 0.8425 | SRCC = 0.8436 ] +INFO:root:Average loss=0.22665134912882096 +INFO:root: +[Valid][1][UTT][ MSE = 0.2420 | LCC = 0.8636 | SRCC = 0.8639 ] +INFO:root:Average loss=0.2019966823550371 +INFO:root: +[Valid][2][UTT][ MSE = 0.2549 | LCC = 0.8738 | SRCC = 0.8715 ] +INFO:root:Average loss=0.1896469687613157 +INFO:root: +[Valid][3][UTT][ MSE = 0.2712 | LCC = 0.8759 | SRCC = 0.8736 ] +INFO:root:Average loss=0.17658125752439865 +INFO:root: +[Valid][4][UTT][ MSE = 0.2615 | LCC = 0.8741 | SRCC = 0.8715 ] +INFO:root:Average loss=0.16378516780260283 +INFO:root: +[Valid][5][UTT][ MSE = 0.2390 | LCC = 0.8741 | SRCC = 0.8715 ] +INFO:root:Average loss=0.15591328524244139 +INFO:root: +[Valid][6][UTT][ MSE = 0.2501 | LCC = 0.8703 | SRCC = 0.8689 ] +INFO:root:Average loss=0.1446750428623114 +INFO:root: +[Valid][7][UTT][ MSE = 0.2491 | LCC = 0.8639 | SRCC = 0.8641 ] +INFO:root:Average loss=0.13367679772468713 +INFO:root: +[Valid][8][UTT][ MSE = 0.2300 | LCC = 0.8713 | SRCC = 0.8693 ] +INFO:root:Average loss=0.12653808028270036 +INFO:root: +[Valid][9][UTT][ MSE = 0.2278 | LCC = 0.8678 | SRCC = 0.8667 ] +INFO:root:Average loss=0.11818161797829163 +INFO:root: +[Valid][10][UTT][ MSE = 0.2331 | LCC = 0.8681 | SRCC = 0.8660 ] +INFO:root:Average loss=0.10836845784424207 +INFO:root: +[Valid][11][UTT][ MSE = 0.2338 | LCC = 0.8639 | SRCC = 0.8620 ] +INFO:root:Average loss=0.10197275232237119 +INFO:root: +[Valid][12][UTT][ MSE = 0.2373 | LCC = 0.8625 | SRCC = 0.8600 ] +INFO:root:Average loss=0.0979276335774324 +INFO:root: +[Valid][13][UTT][ MSE = 0.2308 | LCC = 0.8592 | SRCC = 0.8580 ] +INFO:root:Average loss=0.0931238646690662 +INFO:root: +[Valid][14][UTT][ MSE = 0.2347 | LCC = 0.8617 | SRCC = 0.8595 ] +INFO:root:Average loss=0.08422505865112329 +INFO:root: +[Valid][15][UTT][ MSE = 0.2306 | LCC = 0.8589 | SRCC = 0.8579 ] +INFO:root:Average loss=0.0842766255999987 +INFO:root: +[Valid][16][UTT][ MSE = 0.2581 | LCC = 0.8536 | SRCC = 0.8538 ] +INFO:root:Average loss=0.07932516316381785 +INFO:root: +[Valid][17][UTT][ MSE = 0.2390 | LCC = 0.8554 | SRCC = 0.8539 ] +INFO:root:Average loss=0.07161090169579555 +INFO:root: +[Valid][18][UTT][ MSE = 0.2378 | LCC = 0.8549 | SRCC = 0.8526 ] +INFO:root:Average loss=0.07143093177523369 +INFO:root: +[Valid][19][UTT][ MSE = 0.2452 | LCC = 0.8563 | SRCC = 0.8528 ] +INFO:root:Average loss=0.07060055449031867 +INFO:root: +[Valid][20][UTT][ MSE = 0.2432 | LCC = 0.8548 | SRCC = 0.8532 ] +INFO:root:Average loss=0.06629876983471406 +INFO:root: +[Valid][21][UTT][ MSE = 0.2487 | LCC = 0.8529 | SRCC = 0.8499 ] +INFO:root:Average loss=0.06311473178748901 +INFO:root: +[Valid][22][UTT][ MSE = 0.2340 | LCC = 0.8540 | SRCC = 0.8518 ] +INFO:root:Average loss=0.06055045872926712 +INFO:root: +[Valid][23][UTT][ MSE = 0.2266 | LCC = 0.8563 | SRCC = 0.8547 ] +INFO:root:Average loss=0.053039177965659365 +INFO:root: +[Valid][24][UTT][ MSE = 0.2621 | LCC = 0.8490 | SRCC = 0.8462 ] +INFO:root:Average loss=0.0535769275843333 +INFO:root: +[Valid][25][UTT][ MSE = 0.2463 | LCC = 0.8522 | SRCC = 0.8490 ] +INFO:root:Average loss=0.053156220712340795 +INFO:root: +[Valid][26][UTT][ MSE = 0.2307 | LCC = 0.8549 | SRCC = 0.8532 ] +INFO:root:Average loss=0.05270926601802691 +INFO:root: +[Valid][27][UTT][ MSE = 0.2657 | LCC = 0.8468 | SRCC = 0.8446 ] +INFO:root:Average loss=0.05033771228045225 +INFO:root: +[Valid][28][UTT][ MSE = 0.2401 | LCC = 0.8523 | SRCC = 0.8498 ] +INFO:root:Average loss=0.0460592704371382 +INFO:root: +[Valid][29][UTT][ MSE = 0.2403 | LCC = 0.8531 | SRCC = 0.8506 ] +INFO:root:Average loss=0.04465719930923138 +INFO:root: +[Valid][30][UTT][ MSE = 0.2378 | LCC = 0.8554 | SRCC = 0.8536 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2743 | LCC = 0.8650 | SRCC = 0.8623 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ed5f21d07e10b685c260b4781f493c2af8c1a8eb --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:45528fa4411f5931318818bc8751121f16302f47d19b94eb306406dc85f3ef12 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..debb69cf4ca4e5935deca7b6d2a7d5e1ced64de2 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1f8b5029aa6ca07dabc6480a7223b0a48b2105610451affb6915437990322216 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..08f9fd437432be1756be3db61c8b9562a2dc33d2 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/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=0.35264090429513884 +INFO:root: +[Valid][0][UTT][ MSE = 0.2787 | LCC = 0.8387 | SRCC = 0.8450 ] +INFO:root:Average loss=0.22579400470623603 +INFO:root: +[Valid][1][UTT][ MSE = 0.2880 | LCC = 0.8641 | SRCC = 0.8629 ] +INFO:root:Average loss=0.20927019503254157 +INFO:root: +[Valid][2][UTT][ MSE = 0.2758 | LCC = 0.8642 | SRCC = 0.8643 ] +INFO:root:Average loss=0.19606755387324554 +INFO:root: +[Valid][3][UTT][ MSE = 0.2584 | LCC = 0.8682 | SRCC = 0.8670 ] +INFO:root:Average loss=0.17514645976897997 +INFO:root: +[Valid][4][UTT][ MSE = 0.2503 | LCC = 0.8698 | SRCC = 0.8673 ] +INFO:root:Average loss=0.1644989435489361 +INFO:root: +[Valid][5][UTT][ MSE = 0.2476 | LCC = 0.8688 | SRCC = 0.8683 ] +INFO:root:Average loss=0.15743551919093499 +INFO:root: +[Valid][6][UTT][ MSE = 0.2352 | LCC = 0.8740 | SRCC = 0.8735 ] +INFO:root:Average loss=0.14311946718356547 +INFO:root: +[Valid][7][UTT][ MSE = 0.2523 | LCC = 0.8684 | SRCC = 0.8667 ] +INFO:root:Average loss=0.13802544361887834 +INFO:root: +[Valid][8][UTT][ MSE = 0.2365 | LCC = 0.8713 | SRCC = 0.8704 ] +INFO:root:Average loss=0.12927090510343894 +INFO:root: +[Valid][9][UTT][ MSE = 0.2301 | LCC = 0.8688 | SRCC = 0.8685 ] +INFO:root:Average loss=0.11858809643831009 +INFO:root: +[Valid][10][UTT][ MSE = 0.2575 | LCC = 0.8693 | SRCC = 0.8679 ] +INFO:root:Average loss=0.11000682442234112 +INFO:root: +[Valid][11][UTT][ MSE = 0.2323 | LCC = 0.8680 | SRCC = 0.8679 ] +INFO:root:Average loss=0.10003466397905961 +INFO:root: +[Valid][12][UTT][ MSE = 0.2458 | LCC = 0.8667 | SRCC = 0.8663 ] +INFO:root:Average loss=0.09775910860834977 +INFO:root: +[Valid][13][UTT][ MSE = 0.2591 | LCC = 0.8664 | SRCC = 0.8650 ] +INFO:root:Average loss=0.0921578149860486 +INFO:root: +[Valid][14][UTT][ MSE = 0.2429 | LCC = 0.8584 | SRCC = 0.8573 ] +INFO:root:Average loss=0.08584270497354177 +INFO:root: +[Valid][15][UTT][ MSE = 0.2595 | LCC = 0.8609 | SRCC = 0.8592 ] +INFO:root:Average loss=0.08427284228113982 +INFO:root: +[Valid][16][UTT][ MSE = 0.2807 | LCC = 0.8623 | SRCC = 0.8622 ] +INFO:root:Average loss=0.0752900937715402 +INFO:root: +[Valid][17][UTT][ MSE = 0.2428 | LCC = 0.8620 | SRCC = 0.8619 ] +INFO:root:Average loss=0.0706941213649817 +INFO:root: +[Valid][18][UTT][ MSE = 0.2338 | LCC = 0.8622 | SRCC = 0.8623 ] +INFO:root:Average loss=0.06802802924544384 +INFO:root: +[Valid][19][UTT][ MSE = 0.2483 | LCC = 0.8565 | SRCC = 0.8556 ] +INFO:root:Average loss=0.06524293229748042 +INFO:root: +[Valid][20][UTT][ MSE = 0.2425 | LCC = 0.8596 | SRCC = 0.8587 ] +INFO:root:Average loss=0.06009290806758098 +INFO:root: +[Valid][21][UTT][ MSE = 0.2459 | LCC = 0.8573 | SRCC = 0.8570 ] +INFO:root:Average loss=0.05809716536448552 +INFO:root: +[Valid][22][UTT][ MSE = 0.2605 | LCC = 0.8586 | SRCC = 0.8581 ] +INFO:root:Average loss=0.055300687511379905 +INFO:root: +[Valid][23][UTT][ MSE = 0.2365 | LCC = 0.8601 | SRCC = 0.8583 ] +INFO:root:Average loss=0.05821262705975618 +INFO:root: +[Valid][24][UTT][ MSE = 0.2485 | LCC = 0.8595 | SRCC = 0.8583 ] +INFO:root:Average loss=0.053312672063326225 +INFO:root: +[Valid][25][UTT][ MSE = 0.2545 | LCC = 0.8542 | SRCC = 0.8518 ] +INFO:root:Average loss=0.04945038495442042 +INFO:root: +[Valid][26][UTT][ MSE = 0.2456 | LCC = 0.8588 | SRCC = 0.8575 ] +INFO:root:Average loss=0.047712452470874175 +INFO:root: +[Valid][27][UTT][ MSE = 0.2566 | LCC = 0.8581 | SRCC = 0.8574 ] +INFO:root:Average loss=0.04874761056346007 +INFO:root: +[Valid][28][UTT][ MSE = 0.2531 | LCC = 0.8523 | SRCC = 0.8509 ] +INFO:root:Average loss=0.04350034486597929 +INFO:root: +[Valid][29][UTT][ MSE = 0.2447 | LCC = 0.8553 | SRCC = 0.8543 ] +INFO:root:Average loss=0.04486984305847914 +INFO:root: +[Valid][30][UTT][ MSE = 0.2439 | LCC = 0.8600 | SRCC = 0.8584 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2511 | LCC = 0.8623 | SRCC = 0.8630 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..10ae9e471bbefaf434caab7623124789bfd7ac83 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c7864be64e91baa24c70e8761466d8267c2de366926819d533bfe214444f1182 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..78fa4f8971600ee1b69f9ec496fd62001da62667 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eaa7017964bd38823a6f2c26844b09c069a4a22929097042cdcc28bcbd86e4df +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..cfd922e4afc10279b8405c3af4d34496c3ddd97d --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/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=0.3807700282106033 +INFO:root: +[Valid][0][UTT][ MSE = 0.3247 | LCC = 0.8426 | SRCC = 0.8420 ] +INFO:root:Average loss=0.22242442002663246 +INFO:root: +[Valid][1][UTT][ MSE = 0.3106 | LCC = 0.8629 | SRCC = 0.8601 ] +INFO:root:Average loss=0.21464128104540017 +INFO:root: +[Valid][2][UTT][ MSE = 0.2867 | LCC = 0.8667 | SRCC = 0.8646 ] +INFO:root:Average loss=0.18898341422661757 +INFO:root: +[Valid][3][UTT][ MSE = 0.2681 | LCC = 0.8685 | SRCC = 0.8676 ] +INFO:root:Average loss=0.17702438023227912 +INFO:root: +[Valid][4][UTT][ MSE = 0.3252 | LCC = 0.8680 | SRCC = 0.8693 ] +INFO:root:Average loss=0.16927827913791704 +INFO:root: +[Valid][5][UTT][ MSE = 0.2738 | LCC = 0.8694 | SRCC = 0.8671 ] +INFO:root:Average loss=0.1539739214648039 +INFO:root: +[Valid][6][UTT][ MSE = 0.2558 | LCC = 0.8676 | SRCC = 0.8671 ] +INFO:root:Average loss=0.1438266441034965 +INFO:root: +[Valid][7][UTT][ MSE = 0.2369 | LCC = 0.8686 | SRCC = 0.8689 ] +INFO:root:Average loss=0.14414093777155265 +INFO:root: +[Valid][8][UTT][ MSE = 0.2303 | LCC = 0.8692 | SRCC = 0.8692 ] +INFO:root:Average loss=0.12784585108359656 +INFO:root: +[Valid][9][UTT][ MSE = 0.2575 | LCC = 0.8632 | SRCC = 0.8634 ] +INFO:root:Average loss=0.12108528327483398 +INFO:root: +[Valid][10][UTT][ MSE = 0.2515 | LCC = 0.8665 | SRCC = 0.8655 ] +INFO:root:Average loss=0.11366128396147336 +INFO:root: +[Valid][11][UTT][ MSE = 0.2312 | LCC = 0.8654 | SRCC = 0.8652 ] +INFO:root:Average loss=0.10246202454735072 +INFO:root: +[Valid][12][UTT][ MSE = 0.2421 | LCC = 0.8674 | SRCC = 0.8662 ] +INFO:root:Average loss=0.09443127975249901 +INFO:root: +[Valid][13][UTT][ MSE = 0.2437 | LCC = 0.8607 | SRCC = 0.8596 ] +INFO:root:Average loss=0.09249770880127564 +INFO:root: +[Valid][14][UTT][ MSE = 0.2628 | LCC = 0.8590 | SRCC = 0.8575 ] +INFO:root:Average loss=0.08867674020047371 +INFO:root: +[Valid][15][UTT][ MSE = 0.2366 | LCC = 0.8554 | SRCC = 0.8572 ] +INFO:root:Average loss=0.084106489442862 +INFO:root: +[Valid][16][UTT][ MSE = 0.2477 | LCC = 0.8566 | SRCC = 0.8557 ] +INFO:root:Average loss=0.07651336328723492 +INFO:root: +[Valid][17][UTT][ MSE = 0.2435 | LCC = 0.8623 | SRCC = 0.8606 ] +INFO:root:Average loss=0.07061393730915509 +INFO:root: +[Valid][18][UTT][ MSE = 0.2397 | LCC = 0.8560 | SRCC = 0.8548 ] +INFO:root:Average loss=0.06837016788239662 +INFO:root: +[Valid][19][UTT][ MSE = 0.2421 | LCC = 0.8531 | SRCC = 0.8524 ] +INFO:root:Average loss=0.06632582103021634 +INFO:root: +[Valid][20][UTT][ MSE = 0.2677 | LCC = 0.8560 | SRCC = 0.8548 ] +INFO:root:Average loss=0.06333858558000663 +INFO:root: +[Valid][21][UTT][ MSE = 0.2354 | LCC = 0.8581 | SRCC = 0.8572 ] +INFO:root:Average loss=0.05795361493260433 +INFO:root: +[Valid][22][UTT][ MSE = 0.2627 | LCC = 0.8570 | SRCC = 0.8561 ] +INFO:root:Average loss=0.059403940462149106 +INFO:root: +[Valid][23][UTT][ MSE = 0.2555 | LCC = 0.8539 | SRCC = 0.8525 ] +INFO:root:Average loss=0.05359882656007241 +INFO:root: +[Valid][24][UTT][ MSE = 0.2390 | LCC = 0.8517 | SRCC = 0.8503 ] +INFO:root:Average loss=0.049026874061196275 +INFO:root: +[Valid][25][UTT][ MSE = 0.2891 | LCC = 0.8472 | SRCC = 0.8445 ] +INFO:root:Average loss=0.050250446901489526 +INFO:root: +[Valid][26][UTT][ MSE = 0.2906 | LCC = 0.8505 | SRCC = 0.8491 ] +INFO:root:Average loss=0.05165304057300091 +INFO:root: +[Valid][27][UTT][ MSE = 0.2427 | LCC = 0.8542 | SRCC = 0.8523 ] +INFO:root:Average loss=0.04407712435111021 +INFO:root: +[Valid][28][UTT][ MSE = 0.2476 | LCC = 0.8509 | SRCC = 0.8489 ] +INFO:root:Average loss=0.0410588234424209 +INFO:root: +[Valid][29][UTT][ MSE = 0.2542 | LCC = 0.8502 | SRCC = 0.8484 ] +INFO:root:Average loss=0.042960193318625294 +INFO:root: +[Valid][30][UTT][ MSE = 0.2498 | LCC = 0.8497 | SRCC = 0.8473 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2537 | LCC = 0.8654 | SRCC = 0.8656 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a13f1a50565fbec736580cf98bf30da287aa9295 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6a166494957494a869223ba0eedccb0a559b0a6716742f5b8a6e9a8df14ca22 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a6d672cc1d643fd108072661da942f8b45181f44 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a592db2c9a4315d68a2e4d1c22e115eb9ce2dc5b1c0b10ff71db4c09bb54fbdb +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..71a797c83d311f2a956c1bee2df7fe02da1f6e12 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer7/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=0.3103841965397199 +INFO:root: +[Valid][0][UTT][ MSE = 0.3163 | LCC = 0.8453 | SRCC = 0.8436 ] +INFO:root:Average loss=0.22787650025043732 +INFO:root: +[Valid][1][UTT][ MSE = 0.2857 | LCC = 0.8585 | SRCC = 0.8576 ] +INFO:root:Average loss=0.19957687724859285 +INFO:root: +[Valid][2][UTT][ MSE = 0.2914 | LCC = 0.8698 | SRCC = 0.8659 ] +INFO:root:Average loss=0.18803542174207857 +INFO:root: +[Valid][3][UTT][ MSE = 0.2505 | LCC = 0.8682 | SRCC = 0.8663 ] +INFO:root:Average loss=0.17382121535065848 +INFO:root: +[Valid][4][UTT][ MSE = 0.2511 | LCC = 0.8699 | SRCC = 0.8678 ] +INFO:root:Average loss=0.16009522210329008 +INFO:root: +[Valid][5][UTT][ MSE = 0.2708 | LCC = 0.8693 | SRCC = 0.8647 ] +INFO:root:Average loss=0.1501860301464032 +INFO:root: +[Valid][6][UTT][ MSE = 0.2662 | LCC = 0.8665 | SRCC = 0.8667 ] +INFO:root:Average loss=0.1442657180894644 +INFO:root: +[Valid][7][UTT][ MSE = 0.2427 | LCC = 0.8671 | SRCC = 0.8650 ] +INFO:root:Average loss=0.13680985703682289 +INFO:root: +[Valid][8][UTT][ MSE = 0.2273 | LCC = 0.8680 | SRCC = 0.8666 ] +INFO:root:Average loss=0.1256762601626225 +INFO:root: +[Valid][9][UTT][ MSE = 0.2732 | LCC = 0.8638 | SRCC = 0.8622 ] +INFO:root:Average loss=0.1218717882457452 +INFO:root: +[Valid][10][UTT][ MSE = 0.2505 | LCC = 0.8662 | SRCC = 0.8623 ] +INFO:root:Average loss=0.10787580272135062 +INFO:root: +[Valid][11][UTT][ MSE = 0.2622 | LCC = 0.8632 | SRCC = 0.8602 ] +INFO:root:Average loss=0.09967985777900769 +INFO:root: +[Valid][12][UTT][ MSE = 0.2438 | LCC = 0.8600 | SRCC = 0.8573 ] +INFO:root:Average loss=0.09670201641244766 +INFO:root: +[Valid][13][UTT][ MSE = 0.2467 | LCC = 0.8524 | SRCC = 0.8500 ] +INFO:root:Average loss=0.09034859795027818 +INFO:root: +[Valid][14][UTT][ MSE = 0.2491 | LCC = 0.8590 | SRCC = 0.8567 ] +INFO:root:Average loss=0.08391022968750733 +INFO:root: +[Valid][15][UTT][ MSE = 0.2319 | LCC = 0.8601 | SRCC = 0.8581 ] +INFO:root:Average loss=0.08291059947357728 +INFO:root: +[Valid][16][UTT][ MSE = 0.2661 | LCC = 0.8580 | SRCC = 0.8539 ] +INFO:root:Average loss=0.07393119173745315 +INFO:root: +[Valid][17][UTT][ MSE = 0.2465 | LCC = 0.8566 | SRCC = 0.8542 ] +INFO:root:Average loss=0.07319305240152738 +INFO:root: +[Valid][18][UTT][ MSE = 0.2795 | LCC = 0.8545 | SRCC = 0.8515 ] +INFO:root:Average loss=0.06967244937251775 +INFO:root: +[Valid][19][UTT][ MSE = 0.2478 | LCC = 0.8535 | SRCC = 0.8517 ] +INFO:root:Average loss=0.06337515723246795 +INFO:root: +[Valid][20][UTT][ MSE = 0.2477 | LCC = 0.8547 | SRCC = 0.8515 ] +INFO:root:Average loss=0.06187615591364029 +INFO:root: +[Valid][21][UTT][ MSE = 0.2583 | LCC = 0.8520 | SRCC = 0.8499 ] +INFO:root:Average loss=0.06013964725515017 +INFO:root: +[Valid][22][UTT][ MSE = 0.2365 | LCC = 0.8541 | SRCC = 0.8534 ] +INFO:root:Average loss=0.061029720812653884 +INFO:root: +[Valid][23][UTT][ MSE = 0.2410 | LCC = 0.8534 | SRCC = 0.8516 ] +INFO:root:Average loss=0.057862770003386035 +INFO:root: +[Valid][24][UTT][ MSE = 0.2505 | LCC = 0.8519 | SRCC = 0.8499 ] +INFO:root:Average loss=0.05434349250908081 +INFO:root: +[Valid][25][UTT][ MSE = 0.2388 | LCC = 0.8526 | SRCC = 0.8506 ] +INFO:root:Average loss=0.05097115073257532 +INFO:root: +[Valid][26][UTT][ MSE = 0.2463 | LCC = 0.8555 | SRCC = 0.8526 ] +INFO:root:Average loss=0.049035497702276096 +INFO:root: +[Valid][27][UTT][ MSE = 0.2561 | LCC = 0.8490 | SRCC = 0.8459 ] +INFO:root:Average loss=0.047571454579249405 +INFO:root: +[Valid][28][UTT][ MSE = 0.2468 | LCC = 0.8504 | SRCC = 0.8478 ] +INFO:root:Average loss=0.04738145727568712 +INFO:root: +[Valid][29][UTT][ MSE = 0.2429 | LCC = 0.8491 | SRCC = 0.8467 ] +INFO:root:Average loss=0.04637191220162771 +INFO:root: +[Valid][30][UTT][ MSE = 0.2575 | LCC = 0.8471 | SRCC = 0.8446 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2351 | LCC = 0.8661 | SRCC = 0.8656 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..302988c2d1560c4f87e58f0eb1c6d4d9fcf91a79 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e1f31bea8dc739e4cc08c4497df68870364f497da590421016fffc00e5d66d6 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6641649a877137489e5befaf079eb5d7864424c9 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:472ca956e00736d272708ba46b67e730705f6b495b99e5e90d58b256445d0e85 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..7214fb73d5e1e4714ee356f8bd93909a2ffc669b --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/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=0.33209514923584765 +INFO:root: +[Valid][0][UTT][ MSE = 0.2922 | LCC = 0.8464 | SRCC = 0.8449 ] +INFO:root:Average loss=0.23119913175320014 +INFO:root: +[Valid][1][UTT][ MSE = 0.2601 | LCC = 0.8620 | SRCC = 0.8611 ] +INFO:root:Average loss=0.21146307522669816 +INFO:root: +[Valid][2][UTT][ MSE = 0.2592 | LCC = 0.8710 | SRCC = 0.8694 ] +INFO:root:Average loss=0.19596996884315443 +INFO:root: +[Valid][3][UTT][ MSE = 0.2283 | LCC = 0.8719 | SRCC = 0.8715 ] +INFO:root:Average loss=0.18383596761104387 +INFO:root: +[Valid][4][UTT][ MSE = 0.2404 | LCC = 0.8765 | SRCC = 0.8751 ] +INFO:root:Average loss=0.17239006828421202 +INFO:root: +[Valid][5][UTT][ MSE = 0.2441 | LCC = 0.8774 | SRCC = 0.8760 ] +INFO:root:Average loss=0.16338440374686167 +INFO:root: +[Valid][6][UTT][ MSE = 0.2400 | LCC = 0.8805 | SRCC = 0.8783 ] +INFO:root:Average loss=0.1566480721991796 +INFO:root: +[Valid][7][UTT][ MSE = 0.2294 | LCC = 0.8800 | SRCC = 0.8796 ] +INFO:root:Average loss=0.14447636902332306 +INFO:root: +[Valid][8][UTT][ MSE = 0.2388 | LCC = 0.8757 | SRCC = 0.8806 ] +INFO:root:Average loss=0.14666066442926726 +INFO:root: +[Valid][9][UTT][ MSE = 0.2401 | LCC = 0.8791 | SRCC = 0.8783 ] +INFO:root:Average loss=0.1300454401434996 +INFO:root: +[Valid][10][UTT][ MSE = 0.2617 | LCC = 0.8711 | SRCC = 0.8726 ] +INFO:root:Average loss=0.12747505450477967 +INFO:root: +[Valid][11][UTT][ MSE = 0.2316 | LCC = 0.8778 | SRCC = 0.8771 ] +INFO:root:Average loss=0.12376054080250935 +INFO:root: +[Valid][12][UTT][ MSE = 0.2227 | LCC = 0.8749 | SRCC = 0.8730 ] +INFO:root:Average loss=0.11567415540608075 +INFO:root: +[Valid][13][UTT][ MSE = 0.2157 | LCC = 0.8738 | SRCC = 0.8740 ] +INFO:root:Average loss=0.10927531190025501 +INFO:root: +[Valid][14][UTT][ MSE = 0.2418 | LCC = 0.8731 | SRCC = 0.8733 ] +INFO:root:Average loss=0.10167597477825788 +INFO:root: +[Valid][15][UTT][ MSE = 0.2520 | LCC = 0.8673 | SRCC = 0.8658 ] +INFO:root:Average loss=0.09934514120985301 +INFO:root: +[Valid][16][UTT][ MSE = 0.2106 | LCC = 0.8749 | SRCC = 0.8751 ] +INFO:root:Average loss=0.1052641949783533 +INFO:root: +[Valid][17][UTT][ MSE = 0.2190 | LCC = 0.8743 | SRCC = 0.8740 ] +INFO:root:Average loss=0.08443397942643899 +INFO:root: +[Valid][18][UTT][ MSE = 0.2327 | LCC = 0.8690 | SRCC = 0.8697 ] +INFO:root:Average loss=0.08708556669835861 +INFO:root: +[Valid][19][UTT][ MSE = 0.2271 | LCC = 0.8731 | SRCC = 0.8723 ] +INFO:root:Average loss=0.08496936773642516 +INFO:root: +[Valid][20][UTT][ MSE = 0.2261 | LCC = 0.8750 | SRCC = 0.8748 ] +INFO:root:Average loss=0.07817776864155745 +INFO:root: +[Valid][21][UTT][ MSE = 0.2223 | LCC = 0.8707 | SRCC = 0.8697 ] +INFO:root:Average loss=0.07650488672348169 +INFO:root: +[Valid][22][UTT][ MSE = 0.2160 | LCC = 0.8699 | SRCC = 0.8694 ] +INFO:root:Average loss=0.076873696098725 +INFO:root: +[Valid][23][UTT][ MSE = 0.2437 | LCC = 0.8702 | SRCC = 0.8693 ] +INFO:root:Average loss=0.06920917336948407 +INFO:root: +[Valid][24][UTT][ MSE = 0.2525 | LCC = 0.8684 | SRCC = 0.8679 ] +INFO:root:Average loss=0.06485115860899289 +INFO:root: +[Valid][25][UTT][ MSE = 0.2364 | LCC = 0.8682 | SRCC = 0.8672 ] +INFO:root:Average loss=0.0644174377219035 +INFO:root: +[Valid][26][UTT][ MSE = 0.2342 | LCC = 0.8657 | SRCC = 0.8645 ] +INFO:root:Average loss=0.061564551952939764 +INFO:root: +[Valid][27][UTT][ MSE = 0.2238 | LCC = 0.8726 | SRCC = 0.8711 ] +INFO:root:Average loss=0.05801534284956944 +INFO:root: +[Valid][28][UTT][ MSE = 0.2129 | LCC = 0.8705 | SRCC = 0.8687 ] +INFO:root:Average loss=0.06398360361942114 +INFO:root: +[Valid][29][UTT][ MSE = 0.2050 | LCC = 0.8732 | SRCC = 0.8727 ] +INFO:root:Average loss=0.057463854957276427 +INFO:root: +[Valid][30][UTT][ MSE = 0.2278 | LCC = 0.8660 | SRCC = 0.8660 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2434 | LCC = 0.8655 | SRCC = 0.8659 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4912eebabe674563c410ceb1224ab42431e1c22f --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:80ea4877e6df74ed633c1097e1b0a75a6546f87deeddf4df790077da9dcb5027 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ea12d9018c5c0067b25fb63cc91b043006bb951e --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8df5ac0ee9ded20e5bfe6037a35089401fdd014cf71afdc639d182f69f8c977a +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..384b015b765c56540fb8bd8f01e43a2fea4c5ff5 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/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=0.4083507544337175 +INFO:root: +[Valid][0][UTT][ MSE = 0.3246 | LCC = 0.8295 | SRCC = 0.8373 ] +INFO:root:Average loss=0.22988853775537932 +INFO:root: +[Valid][1][UTT][ MSE = 0.3035 | LCC = 0.8534 | SRCC = 0.8542 ] +INFO:root:Average loss=0.2090420669470078 +INFO:root: +[Valid][2][UTT][ MSE = 0.2838 | LCC = 0.8630 | SRCC = 0.8630 ] +INFO:root:Average loss=0.19832462588181862 +INFO:root: +[Valid][3][UTT][ MSE = 0.2743 | LCC = 0.8660 | SRCC = 0.8676 ] +INFO:root:Average loss=0.18739020499663475 +INFO:root: +[Valid][4][UTT][ MSE = 0.2733 | LCC = 0.8674 | SRCC = 0.8696 ] +INFO:root:Average loss=0.17449302598834038 +INFO:root: +[Valid][5][UTT][ MSE = 0.2833 | LCC = 0.8688 | SRCC = 0.8675 ] +INFO:root:Average loss=0.16785271847859407 +INFO:root: +[Valid][6][UTT][ MSE = 0.2330 | LCC = 0.8728 | SRCC = 0.8746 ] +INFO:root:Average loss=0.15945666866042676 +INFO:root: +[Valid][7][UTT][ MSE = 0.2967 | LCC = 0.8641 | SRCC = 0.8723 ] +INFO:root:Average loss=0.15009602283438048 +INFO:root: +[Valid][8][UTT][ MSE = 0.2532 | LCC = 0.8702 | SRCC = 0.8701 ] +INFO:root:Average loss=0.14021324700652024 +INFO:root: +[Valid][9][UTT][ MSE = 0.2302 | LCC = 0.8738 | SRCC = 0.8738 ] +INFO:root:Average loss=0.13093974173833162 +INFO:root: +[Valid][10][UTT][ MSE = 0.2586 | LCC = 0.8683 | SRCC = 0.8694 ] +INFO:root:Average loss=0.12561985640189585 +INFO:root: +[Valid][11][UTT][ MSE = 0.2791 | LCC = 0.8731 | SRCC = 0.8722 ] +INFO:root:Average loss=0.11863762589219289 +INFO:root: +[Valid][12][UTT][ MSE = 0.2679 | LCC = 0.8702 | SRCC = 0.8717 ] +INFO:root:Average loss=0.1120705019014004 +INFO:root: +[Valid][13][UTT][ MSE = 0.2671 | LCC = 0.8693 | SRCC = 0.8708 ] +INFO:root:Average loss=0.11265706003476413 +INFO:root: +[Valid][14][UTT][ MSE = 0.3003 | LCC = 0.8672 | SRCC = 0.8677 ] +INFO:root:Average loss=0.10677495751625453 +INFO:root: +[Valid][15][UTT][ MSE = 0.2867 | LCC = 0.8686 | SRCC = 0.8702 ] +INFO:root:Average loss=0.09619949113291043 +INFO:root: +[Valid][16][UTT][ MSE = 0.2412 | LCC = 0.8634 | SRCC = 0.8632 ] +INFO:root:Average loss=0.09118822006843029 +INFO:root: +[Valid][17][UTT][ MSE = 0.2469 | LCC = 0.8637 | SRCC = 0.8652 ] +INFO:root:Average loss=0.08906554187146518 +INFO:root: +[Valid][18][UTT][ MSE = 0.2267 | LCC = 0.8630 | SRCC = 0.8629 ] +INFO:root:Average loss=0.08841762352639283 +INFO:root: +[Valid][19][UTT][ MSE = 0.2599 | LCC = 0.8652 | SRCC = 0.8661 ] +INFO:root:Average loss=0.07697582168456836 +INFO:root: +[Valid][20][UTT][ MSE = 0.2764 | LCC = 0.8602 | SRCC = 0.8600 ] +INFO:root:Average loss=0.07488159887874737 +INFO:root: +[Valid][21][UTT][ MSE = 0.2591 | LCC = 0.8618 | SRCC = 0.8626 ] +INFO:root:Average loss=0.07035090029239655 +INFO:root: +[Valid][22][UTT][ MSE = 0.2528 | LCC = 0.8612 | SRCC = 0.8615 ] +INFO:root:Average loss=0.06865156575655326 +INFO:root: +[Valid][23][UTT][ MSE = 0.2456 | LCC = 0.8603 | SRCC = 0.8596 ] +INFO:root:Average loss=0.06990358161811645 +INFO:root: +[Valid][24][UTT][ MSE = 0.2732 | LCC = 0.8596 | SRCC = 0.8604 ] +INFO:root:Average loss=0.06458569509096634 +INFO:root: +[Valid][25][UTT][ MSE = 0.2236 | LCC = 0.8604 | SRCC = 0.8602 ] +INFO:root:Average loss=0.06148592029244472 +INFO:root: +[Valid][26][UTT][ MSE = 0.2565 | LCC = 0.8563 | SRCC = 0.8567 ] +INFO:root:Average loss=0.058851511002733156 +INFO:root: +[Valid][27][UTT][ MSE = 0.2473 | LCC = 0.8608 | SRCC = 0.8606 ] +INFO:root:Average loss=0.06372656568120687 +INFO:root: +[Valid][28][UTT][ MSE = 0.2626 | LCC = 0.8570 | SRCC = 0.8571 ] +INFO:root:Average loss=0.05708372960679042 +INFO:root: +[Valid][29][UTT][ MSE = 0.2726 | LCC = 0.8591 | SRCC = 0.8602 ] +INFO:root:Average loss=0.05494349072567928 +INFO:root: +[Valid][30][UTT][ MSE = 0.2592 | LCC = 0.8553 | SRCC = 0.8563 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2223 | LCC = 0.8662 | SRCC = 0.8678 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..10f5fba6b5f5d5aa592bdb03561bf6af6b646903 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c6d3501bc5dcf0fe558c88fdbe182eb7c666ad18c926d5b50cdf4a8bbc9f9ad6 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7189d9e4214ef7cf782191c2c450dfac96f5eac1 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:edd04f76f39ed5ae4d4558eccc89c8e9e3d7365ad5ff8c31f0f85ecc6f8b08f4 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..373d62fb14f30e537b314e86bc1bc1ae9280cc70 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/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=0.34225472043722105 +INFO:root: +[Valid][0][UTT][ MSE = 0.3332 | LCC = 0.8417 | SRCC = 0.8413 ] +INFO:root:Average loss=0.23146407497234833 +INFO:root: +[Valid][1][UTT][ MSE = 0.3342 | LCC = 0.8557 | SRCC = 0.8537 ] +INFO:root:Average loss=0.21144697261162293 +INFO:root: +[Valid][2][UTT][ MSE = 0.2837 | LCC = 0.8643 | SRCC = 0.8653 ] +INFO:root:Average loss=0.19960581339322603 +INFO:root: +[Valid][3][UTT][ MSE = 0.2744 | LCC = 0.8679 | SRCC = 0.8657 ] +INFO:root:Average loss=0.18693701712748942 +INFO:root: +[Valid][4][UTT][ MSE = 0.2621 | LCC = 0.8693 | SRCC = 0.8683 ] +INFO:root:Average loss=0.18738638485471407 +INFO:root: +[Valid][5][UTT][ MSE = 0.2449 | LCC = 0.8697 | SRCC = 0.8696 ] +INFO:root:Average loss=0.16930593741245759 +INFO:root: +[Valid][6][UTT][ MSE = 0.3039 | LCC = 0.8720 | SRCC = 0.8707 ] +INFO:root:Average loss=0.16024679566423097 +INFO:root: +[Valid][7][UTT][ MSE = 0.2906 | LCC = 0.8678 | SRCC = 0.8665 ] +INFO:root:Average loss=0.15469886429416826 +INFO:root: +[Valid][8][UTT][ MSE = 0.2920 | LCC = 0.8713 | SRCC = 0.8711 ] +INFO:root:Average loss=0.14705462543628153 +INFO:root: +[Valid][9][UTT][ MSE = 0.2580 | LCC = 0.8686 | SRCC = 0.8691 ] +INFO:root:Average loss=0.13125126493664888 +INFO:root: +[Valid][10][UTT][ MSE = 0.2461 | LCC = 0.8659 | SRCC = 0.8632 ] +INFO:root:Average loss=0.12729217713842025 +INFO:root: +[Valid][11][UTT][ MSE = 0.2398 | LCC = 0.8635 | SRCC = 0.8645 ] +INFO:root:Average loss=0.1226180894061541 +INFO:root: +[Valid][12][UTT][ MSE = 0.2377 | LCC = 0.8621 | SRCC = 0.8612 ] +INFO:root:Average loss=0.11683518850268462 +INFO:root: +[Valid][13][UTT][ MSE = 0.2524 | LCC = 0.8635 | SRCC = 0.8618 ] +INFO:root:Average loss=0.1063418787641403 +INFO:root: +[Valid][14][UTT][ MSE = 0.2677 | LCC = 0.8595 | SRCC = 0.8570 ] +INFO:root:Average loss=0.10943473321505082 +INFO:root: +[Valid][15][UTT][ MSE = 0.2455 | LCC = 0.8564 | SRCC = 0.8558 ] +INFO:root:Average loss=0.0986725190320076 +INFO:root: +[Valid][16][UTT][ MSE = 0.2675 | LCC = 0.8603 | SRCC = 0.8585 ] +INFO:root:Average loss=0.09478403341311675 +INFO:root: +[Valid][17][UTT][ MSE = 0.2535 | LCC = 0.8595 | SRCC = 0.8578 ] +INFO:root:Average loss=0.08725700713694096 +INFO:root: +[Valid][18][UTT][ MSE = 0.2552 | LCC = 0.8620 | SRCC = 0.8597 ] +INFO:root:Average loss=0.08628312746683757 +INFO:root: +[Valid][19][UTT][ MSE = 0.2709 | LCC = 0.8486 | SRCC = 0.8455 ] +INFO:root:Average loss=0.08038867761691411 +INFO:root: +[Valid][20][UTT][ MSE = 0.2478 | LCC = 0.8589 | SRCC = 0.8581 ] +INFO:root:Average loss=0.07686445241173108 +INFO:root: +[Valid][21][UTT][ MSE = 0.2717 | LCC = 0.8527 | SRCC = 0.8527 ] +INFO:root:Average loss=0.07937888624385381 +INFO:root: +[Valid][22][UTT][ MSE = 0.2734 | LCC = 0.8547 | SRCC = 0.8536 ] +INFO:root:Average loss=0.06978532213431138 +INFO:root: +[Valid][23][UTT][ MSE = 0.2673 | LCC = 0.8517 | SRCC = 0.8500 ] +INFO:root:Average loss=0.0700823416790137 +INFO:root: +[Valid][24][UTT][ MSE = 0.2627 | LCC = 0.8512 | SRCC = 0.8489 ] +INFO:root:Average loss=0.06454286647912784 +INFO:root: +[Valid][25][UTT][ MSE = 0.2484 | LCC = 0.8561 | SRCC = 0.8539 ] +INFO:root:Average loss=0.0638640707788559 +INFO:root: +[Valid][26][UTT][ MSE = 0.2417 | LCC = 0.8568 | SRCC = 0.8551 ] +INFO:root:Average loss=0.058738514685477965 +INFO:root: +[Valid][27][UTT][ MSE = 0.2685 | LCC = 0.8571 | SRCC = 0.8534 ] +INFO:root:Average loss=0.05837503553200991 +INFO:root: +[Valid][28][UTT][ MSE = 0.2774 | LCC = 0.8543 | SRCC = 0.8529 ] +INFO:root:Average loss=0.05663338346550098 +INFO:root: +[Valid][29][UTT][ MSE = 0.2402 | LCC = 0.8547 | SRCC = 0.8535 ] +INFO:root:Average loss=0.060771323836002596 +INFO:root: +[Valid][30][UTT][ MSE = 0.2471 | LCC = 0.8537 | SRCC = 0.8526 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2725 | LCC = 0.8695 | SRCC = 0.8653 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8535813e69d73fe01b1f2663fafeb1b4b7e1ebd1 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:178c41d14555d9466f8d0149cd20a69743204612d3ea4c991fb5ef61d43c295b +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..dc76d7131a3f03ba6004dfb7dea8ff8e046f369d --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fda197c5f98bf560aea36eb08b464da6bf1b3ca44ca2981867fe295b383326a5 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..84c035fc4bd64a0683d2964f1871f971937b0cbb --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/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=0.36308635007112455 +INFO:root: +[Valid][0][UTT][ MSE = 0.2929 | LCC = 0.8365 | SRCC = 0.8418 ] +INFO:root:Average loss=0.23273047384543297 +INFO:root: +[Valid][1][UTT][ MSE = 0.2784 | LCC = 0.8575 | SRCC = 0.8590 ] +INFO:root:Average loss=0.21231638391812643 +INFO:root: +[Valid][2][UTT][ MSE = 0.2416 | LCC = 0.8663 | SRCC = 0.8658 ] +INFO:root:Average loss=0.19903084960503456 +INFO:root: +[Valid][3][UTT][ MSE = 0.2532 | LCC = 0.8728 | SRCC = 0.8704 ] +INFO:root:Average loss=0.1852286436045781 +INFO:root: +[Valid][4][UTT][ MSE = 0.2363 | LCC = 0.8733 | SRCC = 0.8719 ] +INFO:root:Average loss=0.1752839520191535 +INFO:root: +[Valid][5][UTT][ MSE = 0.2300 | LCC = 0.8735 | SRCC = 0.8729 ] +INFO:root:Average loss=0.16991740789933082 +INFO:root: +[Valid][6][UTT][ MSE = 0.2204 | LCC = 0.8715 | SRCC = 0.8734 ] +INFO:root:Average loss=0.1603295012162282 +INFO:root: +[Valid][7][UTT][ MSE = 0.2355 | LCC = 0.8752 | SRCC = 0.8738 ] +INFO:root:Average loss=0.1520897045922585 +INFO:root: +[Valid][8][UTT][ MSE = 0.2185 | LCC = 0.8737 | SRCC = 0.8724 ] +INFO:root:Average loss=0.14967505901287764 +INFO:root: +[Valid][9][UTT][ MSE = 0.2234 | LCC = 0.8769 | SRCC = 0.8752 ] +INFO:root:Average loss=0.13342537511235628 +INFO:root: +[Valid][10][UTT][ MSE = 0.2290 | LCC = 0.8722 | SRCC = 0.8721 ] +INFO:root:Average loss=0.12937093612093192 +INFO:root: +[Valid][11][UTT][ MSE = 0.2263 | LCC = 0.8739 | SRCC = 0.8711 ] +INFO:root:Average loss=0.12084119872022898 +INFO:root: +[Valid][12][UTT][ MSE = 0.2284 | LCC = 0.8711 | SRCC = 0.8694 ] +INFO:root:Average loss=0.116198145903838 +INFO:root: +[Valid][13][UTT][ MSE = 0.2247 | LCC = 0.8693 | SRCC = 0.8668 ] +INFO:root:Average loss=0.11124976915426743 +INFO:root: +[Valid][14][UTT][ MSE = 0.2288 | LCC = 0.8673 | SRCC = 0.8660 ] +INFO:root:Average loss=0.1026406643482355 +INFO:root: +[Valid][15][UTT][ MSE = 0.2199 | LCC = 0.8702 | SRCC = 0.8679 ] +INFO:root:Average loss=0.10225202596913545 +INFO:root: +[Valid][16][UTT][ MSE = 0.2351 | LCC = 0.8687 | SRCC = 0.8662 ] +INFO:root:Average loss=0.09559804893648013 +INFO:root: +[Valid][17][UTT][ MSE = 0.2160 | LCC = 0.8688 | SRCC = 0.8667 ] +INFO:root:Average loss=0.09463410795881198 +INFO:root: +[Valid][18][UTT][ MSE = 0.2283 | LCC = 0.8669 | SRCC = 0.8639 ] +INFO:root:Average loss=0.08699418136324638 +INFO:root: +[Valid][19][UTT][ MSE = 0.2410 | LCC = 0.8637 | SRCC = 0.8619 ] +INFO:root:Average loss=0.08389166317498073 +INFO:root: +[Valid][20][UTT][ MSE = 0.2311 | LCC = 0.8670 | SRCC = 0.8641 ] +INFO:root:Average loss=0.07778497790105832 +INFO:root: +[Valid][21][UTT][ MSE = 0.2271 | LCC = 0.8662 | SRCC = 0.8643 ] +INFO:root:Average loss=0.0766339359852748 +INFO:root: +[Valid][22][UTT][ MSE = 0.2269 | LCC = 0.8622 | SRCC = 0.8615 ] +INFO:root:Average loss=0.07388052735955287 +INFO:root: +[Valid][23][UTT][ MSE = 0.2259 | LCC = 0.8651 | SRCC = 0.8614 ] +INFO:root:Average loss=0.07026607992175297 +INFO:root: +[Valid][24][UTT][ MSE = 0.2370 | LCC = 0.8634 | SRCC = 0.8602 ] +INFO:root:Average loss=0.06345869309436052 +INFO:root: +[Valid][25][UTT][ MSE = 0.2435 | LCC = 0.8596 | SRCC = 0.8594 ] +INFO:root:Average loss=0.06345017249576557 +INFO:root: +[Valid][26][UTT][ MSE = 0.2351 | LCC = 0.8599 | SRCC = 0.8546 ] +INFO:root:Average loss=0.06284101498432648 +INFO:root: +[Valid][27][UTT][ MSE = 0.2419 | LCC = 0.8569 | SRCC = 0.8566 ] +INFO:root:Average loss=0.059364851707449325 +INFO:root: +[Valid][28][UTT][ MSE = 0.2336 | LCC = 0.8619 | SRCC = 0.8594 ] +INFO:root:Average loss=0.06138869351110397 +INFO:root: +[Valid][29][UTT][ MSE = 0.2448 | LCC = 0.8627 | SRCC = 0.8605 ] +INFO:root:Average loss=0.05563218308947025 +INFO:root: +[Valid][30][UTT][ MSE = 0.2161 | LCC = 0.8654 | SRCC = 0.8622 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2353 | LCC = 0.8670 | SRCC = 0.8673 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..21dbbd9bd93877d07f8da48f3e79a9db59d502ac --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1badc9722fd2785dd60d6289ae9238871ac101cf886086fba95d28477b9c8449 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2ec1ae161b61fe15d1614a0ee7d35c0f37332a0f --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f393913851aff5e15f1a5460e007c4dadf73a330c1eaedd4885d71041db27482 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3bba6fa5c5e667a389fc415766c0ab4c91fc67e8 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer8/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=0.3655484813528183 +INFO:root: +[Valid][0][UTT][ MSE = 0.3117 | LCC = 0.8402 | SRCC = 0.8386 ] +INFO:root:Average loss=0.23739252315881926 +INFO:root: +[Valid][1][UTT][ MSE = 0.2580 | LCC = 0.8525 | SRCC = 0.8535 ] +INFO:root:Average loss=0.21671955860578096 +INFO:root: +[Valid][2][UTT][ MSE = 0.2756 | LCC = 0.8627 | SRCC = 0.8602 ] +INFO:root:Average loss=0.20920587522097123 +INFO:root: +[Valid][3][UTT][ MSE = 0.2635 | LCC = 0.8663 | SRCC = 0.8635 ] +INFO:root:Average loss=0.18723595142364502 +INFO:root: +[Valid][4][UTT][ MSE = 0.2552 | LCC = 0.8660 | SRCC = 0.8630 ] +INFO:root:Average loss=0.17711245899017042 +INFO:root: +[Valid][5][UTT][ MSE = 0.2453 | LCC = 0.8726 | SRCC = 0.8706 ] +INFO:root:Average loss=0.16815125837157935 +INFO:root: +[Valid][6][UTT][ MSE = 0.2215 | LCC = 0.8707 | SRCC = 0.8720 ] +INFO:root:Average loss=0.15616603759236825 +INFO:root: +[Valid][7][UTT][ MSE = 0.2228 | LCC = 0.8711 | SRCC = 0.8711 ] +INFO:root:Average loss=0.14601245245490319 +INFO:root: +[Valid][8][UTT][ MSE = 0.2350 | LCC = 0.8736 | SRCC = 0.8732 ] +INFO:root:Average loss=0.15115068891109565 +INFO:root: +[Valid][9][UTT][ MSE = 0.2466 | LCC = 0.8711 | SRCC = 0.8703 ] +INFO:root:Average loss=0.1297964931298525 +INFO:root: +[Valid][10][UTT][ MSE = 0.2227 | LCC = 0.8698 | SRCC = 0.8727 ] +INFO:root:Average loss=0.1277084939945967 +INFO:root: +[Valid][11][UTT][ MSE = 0.2165 | LCC = 0.8737 | SRCC = 0.8729 ] +INFO:root:Average loss=0.11655938434295166 +INFO:root: +[Valid][12][UTT][ MSE = 0.2324 | LCC = 0.8698 | SRCC = 0.8697 ] +INFO:root:Average loss=0.11419948085378377 +INFO:root: +[Valid][13][UTT][ MSE = 0.2293 | LCC = 0.8696 | SRCC = 0.8676 ] +INFO:root:Average loss=0.10401832192945175 +INFO:root: +[Valid][14][UTT][ MSE = 0.2413 | LCC = 0.8675 | SRCC = 0.8649 ] +INFO:root:Average loss=0.09972557454155041 +INFO:root: +[Valid][15][UTT][ MSE = 0.2272 | LCC = 0.8698 | SRCC = 0.8683 ] +INFO:root:Average loss=0.09191223205282138 +INFO:root: +[Valid][16][UTT][ MSE = 0.2549 | LCC = 0.8595 | SRCC = 0.8558 ] +INFO:root:Average loss=0.08858995856000827 +INFO:root: +[Valid][17][UTT][ MSE = 0.2189 | LCC = 0.8668 | SRCC = 0.8663 ] +INFO:root:Average loss=0.0853423043512381 +INFO:root: +[Valid][18][UTT][ MSE = 0.2401 | LCC = 0.8639 | SRCC = 0.8618 ] +INFO:root:Average loss=0.07818884927874957 +INFO:root: +[Valid][19][UTT][ MSE = 0.2379 | LCC = 0.8651 | SRCC = 0.8636 ] +INFO:root:Average loss=0.07742864237381862 +INFO:root: +[Valid][20][UTT][ MSE = 0.2355 | LCC = 0.8645 | SRCC = 0.8617 ] +INFO:root:Average loss=0.07126354712706345 +INFO:root: +[Valid][21][UTT][ MSE = 0.2370 | LCC = 0.8597 | SRCC = 0.8578 ] +INFO:root:Average loss=0.07031177199230744 +INFO:root: +[Valid][22][UTT][ MSE = 0.2421 | LCC = 0.8591 | SRCC = 0.8565 ] +INFO:root:Average loss=0.06514986871908872 +INFO:root: +[Valid][23][UTT][ MSE = 0.2325 | LCC = 0.8634 | SRCC = 0.8616 ] +INFO:root:Average loss=0.06358320724505645 +INFO:root: +[Valid][24][UTT][ MSE = 0.2354 | LCC = 0.8615 | SRCC = 0.8602 ] +INFO:root:Average loss=0.05884944184277302 +INFO:root: +[Valid][25][UTT][ MSE = 0.2318 | LCC = 0.8597 | SRCC = 0.8592 ] +INFO:root:Average loss=0.058484464119642206 +INFO:root: +[Valid][26][UTT][ MSE = 0.2345 | LCC = 0.8580 | SRCC = 0.8570 ] +INFO:root:Average loss=0.05421663796863495 +INFO:root: +[Valid][27][UTT][ MSE = 0.2248 | LCC = 0.8609 | SRCC = 0.8593 ] +INFO:root:Average loss=0.05520514398813248 +INFO:root: +[Valid][28][UTT][ MSE = 0.2435 | LCC = 0.8567 | SRCC = 0.8550 ] +INFO:root:Average loss=0.050872160312839046 +INFO:root: +[Valid][29][UTT][ MSE = 0.2303 | LCC = 0.8607 | SRCC = 0.8594 ] +INFO:root:Average loss=0.048508681333026826 +INFO:root: +[Valid][30][UTT][ MSE = 0.2333 | LCC = 0.8601 | SRCC = 0.8584 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2336 | LCC = 0.8609 | SRCC = 0.8610 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run0/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run0/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..de36173aedb4b87a36dc784212a0a98a36363b5b --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:515df6566e80cf07723998c9edfff4b5a6ad1fc3ba8ad2678c4507b03b034d8f +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run0/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2a188a1644cfbb441ae19993fd9a6633c1f53e87 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:686ad0a24d3e7d1937ebdaa6b8b99d84d0a6ff02558e0afdc5cd2a391ccd43c4 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run0/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d08a79110753d95d1b553bcb35c233bea6ddd702 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/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=0.3340061954580821 +INFO:root: +[Valid][0][UTT][ MSE = 0.2896 | LCC = 0.8385 | SRCC = 0.8392 ] +INFO:root:Average loss=0.23982949612232354 +INFO:root: +[Valid][1][UTT][ MSE = 0.3043 | LCC = 0.8556 | SRCC = 0.8530 ] +INFO:root:Average loss=0.22820411049402678 +INFO:root: +[Valid][2][UTT][ MSE = 0.3228 | LCC = 0.8581 | SRCC = 0.8578 ] +INFO:root:Average loss=0.217079829902221 +INFO:root: +[Valid][3][UTT][ MSE = 0.2577 | LCC = 0.8634 | SRCC = 0.8621 ] +INFO:root:Average loss=0.2065199220027679 +INFO:root: +[Valid][4][UTT][ MSE = 0.2428 | LCC = 0.8663 | SRCC = 0.8662 ] +INFO:root:Average loss=0.20292554509181243 +INFO:root: +[Valid][5][UTT][ MSE = 0.2462 | LCC = 0.8709 | SRCC = 0.8684 ] +INFO:root:Average loss=0.1947002338293271 +INFO:root: +[Valid][6][UTT][ MSE = 0.2549 | LCC = 0.8716 | SRCC = 0.8711 ] +INFO:root:Average loss=0.1876861137839464 +INFO:root: +[Valid][7][UTT][ MSE = 0.2435 | LCC = 0.8738 | SRCC = 0.8734 ] +INFO:root:Average loss=0.18504596759493536 +INFO:root: +[Valid][8][UTT][ MSE = 0.2583 | LCC = 0.8758 | SRCC = 0.8732 ] +INFO:root:Average loss=0.17880174078238317 +INFO:root: +[Valid][9][UTT][ MSE = 0.2389 | LCC = 0.8721 | SRCC = 0.8734 ] +INFO:root:Average loss=0.17392623118865183 +INFO:root: +[Valid][10][UTT][ MSE = 0.2635 | LCC = 0.8740 | SRCC = 0.8733 ] +INFO:root:Average loss=0.16947162905946755 +INFO:root: +[Valid][11][UTT][ MSE = 0.2272 | LCC = 0.8778 | SRCC = 0.8772 ] +INFO:root:Average loss=0.16525100496334907 +INFO:root: +[Valid][12][UTT][ MSE = 0.2672 | LCC = 0.8729 | SRCC = 0.8743 ] +INFO:root:Average loss=0.16650699336941427 +INFO:root: +[Valid][13][UTT][ MSE = 0.2307 | LCC = 0.8728 | SRCC = 0.8744 ] +INFO:root:Average loss=0.15559036361101347 +INFO:root: +[Valid][14][UTT][ MSE = 0.2643 | LCC = 0.8738 | SRCC = 0.8708 ] +INFO:root:Average loss=0.15167278080032423 +INFO:root: +[Valid][15][UTT][ MSE = 0.2527 | LCC = 0.8762 | SRCC = 0.8744 ] +INFO:root:Average loss=0.1489467465151579 +INFO:root: +[Valid][16][UTT][ MSE = 0.2165 | LCC = 0.8725 | SRCC = 0.8728 ] +INFO:root:Average loss=0.14302589228520027 +INFO:root: +[Valid][17][UTT][ MSE = 0.2356 | LCC = 0.8706 | SRCC = 0.8721 ] +INFO:root:Average loss=0.14205397149691215 +INFO:root: +[Valid][18][UTT][ MSE = 0.2281 | LCC = 0.8705 | SRCC = 0.8684 ] +INFO:root:Average loss=0.13214732057008988 +INFO:root: +[Valid][19][UTT][ MSE = 0.2504 | LCC = 0.8687 | SRCC = 0.8703 ] +INFO:root:Average loss=0.1310147354618097 +INFO:root: +[Valid][20][UTT][ MSE = 0.2308 | LCC = 0.8722 | SRCC = 0.8723 ] +INFO:root:Average loss=0.12991287224949935 +INFO:root: +[Valid][21][UTT][ MSE = 0.2113 | LCC = 0.8710 | SRCC = 0.8719 ] +INFO:root:Average loss=0.1245213154798899 +INFO:root: +[Valid][22][UTT][ MSE = 0.2358 | LCC = 0.8721 | SRCC = 0.8709 ] +INFO:root:Average loss=0.12395225417537567 +INFO:root: +[Valid][23][UTT][ MSE = 0.2306 | LCC = 0.8677 | SRCC = 0.8674 ] +INFO:root:Average loss=0.1200115768573223 +INFO:root: +[Valid][24][UTT][ MSE = 0.2450 | LCC = 0.8637 | SRCC = 0.8645 ] +INFO:root:Average loss=0.11792380525133549 +INFO:root: +[Valid][25][UTT][ MSE = 0.2392 | LCC = 0.8688 | SRCC = 0.8689 ] +INFO:root:Average loss=0.11034812052280475 +INFO:root: +[Valid][26][UTT][ MSE = 0.2306 | LCC = 0.8648 | SRCC = 0.8646 ] +INFO:root:Average loss=0.10950364984380893 +INFO:root: +[Valid][27][UTT][ MSE = 0.2672 | LCC = 0.8649 | SRCC = 0.8639 ] +INFO:root:Average loss=0.10524722532584117 +INFO:root: +[Valid][28][UTT][ MSE = 0.2633 | LCC = 0.8641 | SRCC = 0.8628 ] +INFO:root:Average loss=0.10511479975703435 +INFO:root: +[Valid][29][UTT][ MSE = 0.2509 | LCC = 0.8655 | SRCC = 0.8661 ] +INFO:root:Average loss=0.09808160775364974 +INFO:root: +[Valid][30][UTT][ MSE = 0.2581 | LCC = 0.8612 | SRCC = 0.8590 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2431 | LCC = 0.8644 | SRCC = 0.8654 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run1/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run1/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..53e63d9cad5d909ae9380f0f6e5b013bdd818910 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:44b60992fae7e80c9992745069061435ef034bd638b783d93e89884a5d10b9b5 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run1/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..084b4e69781db11f49f2891396bee6f521dfc3d6 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c59b5839caa169c7961df596131edd4451a2ffe08b4d65883b43e675d2b2d02d +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run1/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a8f2ab1e51648811e674b8db359bd1985bf166d6 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/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=0.32302946903002566 +INFO:root: +[Valid][0][UTT][ MSE = 0.3317 | LCC = 0.8455 | SRCC = 0.8443 ] +INFO:root:Average loss=0.23899685324002534 +INFO:root: +[Valid][1][UTT][ MSE = 0.3362 | LCC = 0.8497 | SRCC = 0.8531 ] +INFO:root:Average loss=0.22552887770609978 +INFO:root: +[Valid][2][UTT][ MSE = 0.3159 | LCC = 0.8590 | SRCC = 0.8553 ] +INFO:root:Average loss=0.2131164339490426 +INFO:root: +[Valid][3][UTT][ MSE = 0.2851 | LCC = 0.8645 | SRCC = 0.8613 ] +INFO:root:Average loss=0.20299096997731772 +INFO:root: +[Valid][4][UTT][ MSE = 0.2651 | LCC = 0.8658 | SRCC = 0.8637 ] +INFO:root:Average loss=0.2003931890313442 +INFO:root: +[Valid][5][UTT][ MSE = 0.2738 | LCC = 0.8698 | SRCC = 0.8675 ] +INFO:root:Average loss=0.1931111665490346 +INFO:root: +[Valid][6][UTT][ MSE = 0.2708 | LCC = 0.8678 | SRCC = 0.8655 ] +INFO:root:Average loss=0.19017198987496206 +INFO:root: +[Valid][7][UTT][ MSE = 0.2809 | LCC = 0.8707 | SRCC = 0.8674 ] +INFO:root:Average loss=0.17886430216141236 +INFO:root: +[Valid][8][UTT][ MSE = 0.3123 | LCC = 0.8659 | SRCC = 0.8640 ] +INFO:root:Average loss=0.17525717721153528 +INFO:root: +[Valid][9][UTT][ MSE = 0.2938 | LCC = 0.8662 | SRCC = 0.8648 ] +INFO:root:Average loss=0.17661781255633402 +INFO:root: +[Valid][10][UTT][ MSE = 0.2814 | LCC = 0.8692 | SRCC = 0.8673 ] +INFO:root:Average loss=0.173471255084643 +INFO:root: +[Valid][11][UTT][ MSE = 0.2586 | LCC = 0.8692 | SRCC = 0.8661 ] +INFO:root:Average loss=0.16157585659470314 +INFO:root: +[Valid][12][UTT][ MSE = 0.2812 | LCC = 0.8696 | SRCC = 0.8668 ] +INFO:root:Average loss=0.16908670235902834 +INFO:root: +[Valid][13][UTT][ MSE = 0.2880 | LCC = 0.8658 | SRCC = 0.8647 ] +INFO:root:Average loss=0.15776466520933005 +INFO:root: +[Valid][14][UTT][ MSE = 0.2719 | LCC = 0.8667 | SRCC = 0.8651 ] +INFO:root:Average loss=0.15266873964514488 +INFO:root: +[Valid][15][UTT][ MSE = 0.2718 | LCC = 0.8663 | SRCC = 0.8664 ] +INFO:root:Average loss=0.14844792450849825 +INFO:root: +[Valid][16][UTT][ MSE = 0.2547 | LCC = 0.8652 | SRCC = 0.8658 ] +INFO:root:Average loss=0.14794653663650537 +INFO:root: +[Valid][17][UTT][ MSE = 0.2818 | LCC = 0.8597 | SRCC = 0.8591 ] +INFO:root:Average loss=0.14235138329557884 +INFO:root: +[Valid][18][UTT][ MSE = 0.2763 | LCC = 0.8640 | SRCC = 0.8624 ] +INFO:root:Average loss=0.13758635205718187 +INFO:root: +[Valid][19][UTT][ MSE = 0.2774 | LCC = 0.8633 | SRCC = 0.8617 ] +INFO:root:Average loss=0.13258144555565637 +INFO:root: +[Valid][20][UTT][ MSE = 0.2396 | LCC = 0.8631 | SRCC = 0.8627 ] +INFO:root:Average loss=0.12763657783850646 +INFO:root: +[Valid][21][UTT][ MSE = 0.2552 | LCC = 0.8640 | SRCC = 0.8628 ] +INFO:root:Average loss=0.12964414298916474 +INFO:root: +[Valid][22][UTT][ MSE = 0.2630 | LCC = 0.8583 | SRCC = 0.8587 ] +INFO:root:Average loss=0.1214365731829252 +INFO:root: +[Valid][23][UTT][ MSE = 0.2506 | LCC = 0.8595 | SRCC = 0.8584 ] +INFO:root:Average loss=0.1127018136664843 +INFO:root: +[Valid][24][UTT][ MSE = 0.2840 | LCC = 0.8515 | SRCC = 0.8537 ] +INFO:root:Average loss=0.1144135129184295 +INFO:root: +[Valid][25][UTT][ MSE = 0.2683 | LCC = 0.8544 | SRCC = 0.8546 ] +INFO:root:Average loss=0.11241791683893937 +INFO:root: +[Valid][26][UTT][ MSE = 0.2557 | LCC = 0.8604 | SRCC = 0.8588 ] +INFO:root:Average loss=0.10646346736794864 +INFO:root: +[Valid][27][UTT][ MSE = 0.2379 | LCC = 0.8575 | SRCC = 0.8591 ] +INFO:root:Average loss=0.10314549548694721 +INFO:root: +[Valid][28][UTT][ MSE = 0.2395 | LCC = 0.8570 | SRCC = 0.8573 ] +INFO:root:Average loss=0.10001679471670052 +INFO:root: +[Valid][29][UTT][ MSE = 0.2429 | LCC = 0.8592 | SRCC = 0.8580 ] +INFO:root:Average loss=0.09505656313819763 +INFO:root: +[Valid][30][UTT][ MSE = 0.2576 | LCC = 0.8520 | SRCC = 0.8505 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2866 | LCC = 0.8644 | SRCC = 0.8645 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run2/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run2/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..14d943fdf0e4d280557719845e90ce4bd66b56d3 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1c94efb328e062cb2628dad45fc84a629d790bdea11eb7a73065f6fb103e8612 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run2/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..19d3da63880fbea8faf9bb970f5bc2fa57088cd5 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cfeca0506ae0ad559aa8f32091b354bdab1d109cc5b577eb0aa1b6224e8efb17 +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run2/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6984c5c3f619655568922c8620623b30e607a6c7 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/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=0.3199341791944626 +INFO:root: +[Valid][0][UTT][ MSE = 0.3504 | LCC = 0.8392 | SRCC = 0.8380 ] +INFO:root:Average loss=0.24481296997803909 +INFO:root: +[Valid][1][UTT][ MSE = 0.3071 | LCC = 0.8566 | SRCC = 0.8534 ] +INFO:root:Average loss=0.22878569440963942 +INFO:root: +[Valid][2][UTT][ MSE = 0.2753 | LCC = 0.8570 | SRCC = 0.8570 ] +INFO:root:Average loss=0.21610633780558905 +INFO:root: +[Valid][3][UTT][ MSE = 0.3027 | LCC = 0.8640 | SRCC = 0.8637 ] +INFO:root:Average loss=0.21045044446602845 +INFO:root: +[Valid][4][UTT][ MSE = 0.3412 | LCC = 0.8628 | SRCC = 0.8648 ] +INFO:root:Average loss=0.20379171214806727 +INFO:root: +[Valid][5][UTT][ MSE = 0.2658 | LCC = 0.8723 | SRCC = 0.8713 ] +INFO:root:Average loss=0.1929104075026818 +INFO:root: +[Valid][6][UTT][ MSE = 0.2965 | LCC = 0.8722 | SRCC = 0.8709 ] +INFO:root:Average loss=0.199457456668218 +INFO:root: +[Valid][7][UTT][ MSE = 0.3013 | LCC = 0.8690 | SRCC = 0.8708 ] +INFO:root:Average loss=0.18565039862042818 +INFO:root: +[Valid][8][UTT][ MSE = 0.2577 | LCC = 0.8672 | SRCC = 0.8720 ] +INFO:root:Average loss=0.18211751068249726 +INFO:root: +[Valid][9][UTT][ MSE = 0.2764 | LCC = 0.8753 | SRCC = 0.8741 ] +INFO:root:Average loss=0.17925519792315286 +INFO:root: +[Valid][10][UTT][ MSE = 0.2525 | LCC = 0.8747 | SRCC = 0.8764 ] +INFO:root:Average loss=0.17860022225441077 +INFO:root: +[Valid][11][UTT][ MSE = 0.3041 | LCC = 0.8712 | SRCC = 0.8718 ] +INFO:root:Average loss=0.16756709263874933 +INFO:root: +[Valid][12][UTT][ MSE = 0.3142 | LCC = 0.8749 | SRCC = 0.8740 ] +INFO:root:Average loss=0.1684780925130233 +INFO:root: +[Valid][13][UTT][ MSE = 0.2614 | LCC = 0.8742 | SRCC = 0.8760 ] +INFO:root:Average loss=0.15753005187098795 +INFO:root: +[Valid][14][UTT][ MSE = 0.2691 | LCC = 0.8734 | SRCC = 0.8736 ] +INFO:root:Average loss=0.15322947435272047 +INFO:root: +[Valid][15][UTT][ MSE = 0.2512 | LCC = 0.8760 | SRCC = 0.8777 ] +INFO:root:Average loss=0.14968889149335715 +INFO:root: +[Valid][16][UTT][ MSE = 0.2945 | LCC = 0.8669 | SRCC = 0.8678 ] +INFO:root:Average loss=0.14761923864865914 +INFO:root: +[Valid][17][UTT][ MSE = 0.2998 | LCC = 0.8742 | SRCC = 0.8738 ] +INFO:root:Average loss=0.15131688309021485 +INFO:root: +[Valid][18][UTT][ MSE = 0.2808 | LCC = 0.8758 | SRCC = 0.8748 ] +INFO:root:Average loss=0.14053730599773237 +INFO:root: +[Valid][19][UTT][ MSE = 0.2961 | LCC = 0.8724 | SRCC = 0.8719 ] +INFO:root:Average loss=0.13931214112119797 +INFO:root: +[Valid][20][UTT][ MSE = 0.2525 | LCC = 0.8734 | SRCC = 0.8729 ] +INFO:root:Average loss=0.130641985589113 +INFO:root: +[Valid][21][UTT][ MSE = 0.2712 | LCC = 0.8690 | SRCC = 0.8692 ] +INFO:root:Average loss=0.12839872170335206 +INFO:root: +[Valid][22][UTT][ MSE = 0.2569 | LCC = 0.8684 | SRCC = 0.8673 ] +INFO:root:Average loss=0.11959244922185555 +INFO:root: +[Valid][23][UTT][ MSE = 0.2680 | LCC = 0.8636 | SRCC = 0.8628 ] +INFO:root:Average loss=0.11952045569435144 +INFO:root: +[Valid][24][UTT][ MSE = 0.2378 | LCC = 0.8637 | SRCC = 0.8664 ] +INFO:root:Average loss=0.12554272025441512 +INFO:root: +[Valid][25][UTT][ MSE = 0.2802 | LCC = 0.8662 | SRCC = 0.8655 ] +INFO:root:Average loss=0.11278379975985257 +INFO:root: +[Valid][26][UTT][ MSE = 0.2913 | LCC = 0.8657 | SRCC = 0.8661 ] +INFO:root:Average loss=0.11225656582376896 +INFO:root: +[Valid][27][UTT][ MSE = 0.2608 | LCC = 0.8639 | SRCC = 0.8636 ] +INFO:root:Average loss=0.11326493427921565 +INFO:root: +[Valid][28][UTT][ MSE = 0.2776 | LCC = 0.8572 | SRCC = 0.8594 ] +INFO:root:Average loss=0.1072983786654778 +INFO:root: +[Valid][29][UTT][ MSE = 0.2910 | LCC = 0.8633 | SRCC = 0.8627 ] +INFO:root:Average loss=0.10309627499335851 +INFO:root: +[Valid][30][UTT][ MSE = 0.2605 | LCC = 0.8606 | SRCC = 0.8626 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2543 | LCC = 0.8580 | SRCC = 0.8612 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run3/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run3/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f61ccd87bcf714435a9237e46b4c313a8b751f60 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a17f2d183a31fdf858d9906007dd9512d90c7536298009b10f1ccd68e8b472ee +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run3/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d31daeefa4834351efa20b9f29d2351164af0e70 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0652fbbfa837240825327cdb1118975679379116711771d1c41812c41e3b4dfd +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run3/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..77ccf6e140aa6194fb98058914b4c713f46d7272 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/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=0.3259888222584358 +INFO:root: +[Valid][0][UTT][ MSE = 0.3359 | LCC = 0.8401 | SRCC = 0.8399 ] +INFO:root:Average loss=0.2508231751047648 +INFO:root: +[Valid][1][UTT][ MSE = 0.3033 | LCC = 0.8491 | SRCC = 0.8495 ] +INFO:root:Average loss=0.22889740765094757 +INFO:root: +[Valid][2][UTT][ MSE = 0.2882 | LCC = 0.8603 | SRCC = 0.8565 ] +INFO:root:Average loss=0.2151374123417414 +INFO:root: +[Valid][3][UTT][ MSE = 0.3029 | LCC = 0.8622 | SRCC = 0.8617 ] +INFO:root:Average loss=0.21177216848501793 +INFO:root: +[Valid][4][UTT][ MSE = 0.2881 | LCC = 0.8690 | SRCC = 0.8660 ] +INFO:root:Average loss=0.2026370603304643 +INFO:root: +[Valid][5][UTT][ MSE = 0.3210 | LCC = 0.8681 | SRCC = 0.8647 ] +INFO:root:Average loss=0.20138240873049468 +INFO:root: +[Valid][6][UTT][ MSE = 0.2745 | LCC = 0.8694 | SRCC = 0.8675 ] +INFO:root:Average loss=0.19132645074755716 +INFO:root: +[Valid][7][UTT][ MSE = 0.2942 | LCC = 0.8728 | SRCC = 0.8717 ] +INFO:root:Average loss=0.18721264543441626 +INFO:root: +[Valid][8][UTT][ MSE = 0.2606 | LCC = 0.8748 | SRCC = 0.8750 ] +INFO:root:Average loss=0.18200797539873 +INFO:root: +[Valid][9][UTT][ MSE = 0.2441 | LCC = 0.8728 | SRCC = 0.8725 ] +INFO:root:Average loss=0.17753761605574533 +INFO:root: +[Valid][10][UTT][ MSE = 0.2475 | LCC = 0.8736 | SRCC = 0.8741 ] +INFO:root:Average loss=0.16952538155974486 +INFO:root: +[Valid][11][UTT][ MSE = 0.2572 | LCC = 0.8775 | SRCC = 0.8770 ] +INFO:root:Average loss=0.1676890811859033 +INFO:root: +[Valid][12][UTT][ MSE = 0.2804 | LCC = 0.8679 | SRCC = 0.8676 ] +INFO:root:Average loss=0.16532218609100732 +INFO:root: +[Valid][13][UTT][ MSE = 0.2847 | LCC = 0.8750 | SRCC = 0.8717 ] +INFO:root:Average loss=0.15787655143783644 +INFO:root: +[Valid][14][UTT][ MSE = 0.2851 | LCC = 0.8734 | SRCC = 0.8703 ] +INFO:root:Average loss=0.14986074820925027 +INFO:root: +[Valid][15][UTT][ MSE = 0.2758 | LCC = 0.8718 | SRCC = 0.8702 ] +INFO:root:Average loss=0.15006907494404378 +INFO:root: +[Valid][16][UTT][ MSE = 0.3071 | LCC = 0.8748 | SRCC = 0.8737 ] +INFO:root:Average loss=0.1451612128279148 +INFO:root: +[Valid][17][UTT][ MSE = 0.3133 | LCC = 0.8692 | SRCC = 0.8669 ] +INFO:root:Average loss=0.1438258765026545 +INFO:root: +[Valid][18][UTT][ MSE = 0.2597 | LCC = 0.8695 | SRCC = 0.8678 ] +INFO:root:Average loss=0.13574903830885887 +INFO:root: +[Valid][19][UTT][ MSE = 0.2572 | LCC = 0.8688 | SRCC = 0.8690 ] +INFO:root:Average loss=0.13279325935320976 +INFO:root: +[Valid][20][UTT][ MSE = 0.2721 | LCC = 0.8621 | SRCC = 0.8641 ] +INFO:root:Average loss=0.13167221395251077 +INFO:root: +[Valid][21][UTT][ MSE = 0.2723 | LCC = 0.8622 | SRCC = 0.8594 ] +INFO:root:Average loss=0.12268200421180481 +INFO:root: +[Valid][22][UTT][ MSE = 0.2731 | LCC = 0.8697 | SRCC = 0.8671 ] +INFO:root:Average loss=0.12152725200240429 +INFO:root: +[Valid][23][UTT][ MSE = 0.2623 | LCC = 0.8665 | SRCC = 0.8651 ] +INFO:root:Average loss=0.11857297596258995 +INFO:root: +[Valid][24][UTT][ MSE = 0.2622 | LCC = 0.8659 | SRCC = 0.8634 ] +INFO:root:Average loss=0.11688749253367767 +INFO:root: +[Valid][25][UTT][ MSE = 0.3050 | LCC = 0.8644 | SRCC = 0.8611 ] +INFO:root:Average loss=0.11520759780437519 +INFO:root: +[Valid][26][UTT][ MSE = 0.2755 | LCC = 0.8650 | SRCC = 0.8634 ] +INFO:root:Average loss=0.11280997183460456 +INFO:root: +[Valid][27][UTT][ MSE = 0.2739 | LCC = 0.8581 | SRCC = 0.8572 ] +INFO:root:Average loss=0.10442017591916598 +INFO:root: +[Valid][28][UTT][ MSE = 0.2861 | LCC = 0.8628 | SRCC = 0.8609 ] +INFO:root:Average loss=0.10122802447623168 +INFO:root: +[Valid][29][UTT][ MSE = 0.2765 | LCC = 0.8620 | SRCC = 0.8598 ] +INFO:root:Average loss=0.1018157161008089 +INFO:root: +[Valid][30][UTT][ MSE = 0.2732 | LCC = 0.8667 | SRCC = 0.8642 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2498 | LCC = 0.8638 | SRCC = 0.8621 ] diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run4/config.gin b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run4/model.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bb688794163f76e713629a4bd2556311c79e29a3 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4f106bc49e5281096b9a3d958680b67b46faa64f26a80368c1ab4fba708650f8 +size 2747979 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run4/model_best.pt b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b802ce590b358d978cc74073956c1f918ec49e69 --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6e03da00aee1d4186518776a96823a4237a29eb438b798a53342962c2674905a +size 2749702 diff --git a/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run4/train.log b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5dfe1f7dc73677c2eb66e77c8300bfc9b8947aee --- /dev/null +++ b/results_bvcc/w2v2_base_bvcc_finetuned_9epoch/layer9/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=0.34431028289672655 +INFO:root: +[Valid][0][UTT][ MSE = 0.3064 | LCC = 0.8368 | SRCC = 0.8426 ] +INFO:root:Average loss=0.23507611396221015 +INFO:root: +[Valid][1][UTT][ MSE = 0.3139 | LCC = 0.8592 | SRCC = 0.8579 ] +INFO:root:Average loss=0.22626717828023127 +INFO:root: +[Valid][2][UTT][ MSE = 0.2891 | LCC = 0.8625 | SRCC = 0.8615 ] +INFO:root:Average loss=0.2129603156294578 +INFO:root: +[Valid][3][UTT][ MSE = 0.2784 | LCC = 0.8664 | SRCC = 0.8666 ] +INFO:root:Average loss=0.20857188640496668 +INFO:root: +[Valid][4][UTT][ MSE = 0.3069 | LCC = 0.8669 | SRCC = 0.8677 ] +INFO:root:Average loss=0.19979342073202133 +INFO:root: +[Valid][5][UTT][ MSE = 0.2431 | LCC = 0.8710 | SRCC = 0.8713 ] +INFO:root:Average loss=0.19379054659452194 +INFO:root: +[Valid][6][UTT][ MSE = 0.2631 | LCC = 0.8665 | SRCC = 0.8700 ] +INFO:root:Average loss=0.19360903316201308 +INFO:root: +[Valid][7][UTT][ MSE = 0.2692 | LCC = 0.8710 | SRCC = 0.8702 ] +INFO:root:Average loss=0.18025366618083075 +INFO:root: +[Valid][8][UTT][ MSE = 0.2751 | LCC = 0.8681 | SRCC = 0.8710 ] +INFO:root:Average loss=0.17576909552399927 +INFO:root: +[Valid][9][UTT][ MSE = 0.2324 | LCC = 0.8702 | SRCC = 0.8740 ] +INFO:root:Average loss=0.17409889696118158 +INFO:root: +[Valid][10][UTT][ MSE = 0.2556 | LCC = 0.8736 | SRCC = 0.8729 ] +INFO:root:Average loss=0.17222991910500404 +INFO:root: +[Valid][11][UTT][ MSE = 0.2659 | LCC = 0.8704 | SRCC = 0.8733 ] +INFO:root:Average loss=0.16670601012615058 +INFO:root: +[Valid][12][UTT][ MSE = 0.2733 | LCC = 0.8716 | SRCC = 0.8717 ] +INFO:root:Average loss=0.15903242324025202 +INFO:root: +[Valid][13][UTT][ MSE = 0.2511 | LCC = 0.8737 | SRCC = 0.8727 ] +INFO:root:Average loss=0.15571803437211576 +INFO:root: +[Valid][14][UTT][ MSE = 0.2297 | LCC = 0.8678 | SRCC = 0.8712 ] +INFO:root:Average loss=0.14995472677625143 +INFO:root: +[Valid][15][UTT][ MSE = 0.2531 | LCC = 0.8728 | SRCC = 0.8721 ] +INFO:root:Average loss=0.14233685026948267 +INFO:root: +[Valid][16][UTT][ MSE = 0.2475 | LCC = 0.8707 | SRCC = 0.8691 ] +INFO:root:Average loss=0.13633116946006432 +INFO:root: +[Valid][17][UTT][ MSE = 0.2214 | LCC = 0.8674 | SRCC = 0.8686 ] +INFO:root:Average loss=0.1342562697827816 +INFO:root: +[Valid][18][UTT][ MSE = 0.2286 | LCC = 0.8704 | SRCC = 0.8707 ] +INFO:root:Average loss=0.13295730881583998 +INFO:root: +[Valid][19][UTT][ MSE = 0.2466 | LCC = 0.8685 | SRCC = 0.8672 ] +INFO:root:Average loss=0.12768145641073203 +INFO:root: +[Valid][20][UTT][ MSE = 0.2397 | LCC = 0.8698 | SRCC = 0.8691 ] +INFO:root:Average loss=0.12732031759925377 +INFO:root: +[Valid][21][UTT][ MSE = 0.2478 | LCC = 0.8700 | SRCC = 0.8683 ] +INFO:root:Average loss=0.12626235234813812 +INFO:root: +[Valid][22][UTT][ MSE = 0.2424 | LCC = 0.8662 | SRCC = 0.8676 ] +INFO:root:Average loss=0.11797913746573986 +INFO:root: +[Valid][23][UTT][ MSE = 0.2438 | LCC = 0.8649 | SRCC = 0.8638 ] +INFO:root:Average loss=0.11715307683707812 +INFO:root: +[Valid][24][UTT][ MSE = 0.2474 | LCC = 0.8654 | SRCC = 0.8647 ] +INFO:root:Average loss=0.11112744198777737 +INFO:root: +[Valid][25][UTT][ MSE = 0.2368 | LCC = 0.8655 | SRCC = 0.8673 ] +INFO:root:Average loss=0.11261048884345935 +INFO:root: +[Valid][26][UTT][ MSE = 0.2502 | LCC = 0.8609 | SRCC = 0.8621 ] +INFO:root:Average loss=0.10902230116801384 +INFO:root: +[Valid][27][UTT][ MSE = 0.2470 | LCC = 0.8644 | SRCC = 0.8632 ] +INFO:root:Average loss=0.10109574901751983 +INFO:root: +[Valid][28][UTT][ MSE = 0.2446 | LCC = 0.8611 | SRCC = 0.8616 ] +INFO:root:Average loss=0.10116254395017257 +INFO:root: +[Valid][29][UTT][ MSE = 0.2401 | LCC = 0.8597 | SRCC = 0.8616 ] +INFO:root:Average loss=0.09900289539916393 +INFO:root: +[Valid][30][UTT][ MSE = 0.2414 | LCC = 0.8612 | SRCC = 0.8620 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2487 | LCC = 0.8658 | SRCC = 0.8654 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run0/config.gin b/results_bvcc/w2v2_large_asr_de/layer0/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run0/model.pt b/results_bvcc/w2v2_large_asr_de/layer0/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b81acd2ba12216da31559c525f80b65798446ee4 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dd106ec1873fa308416252889fc4ded92b1deaa1acc2fbcdc482561c101a1ff6 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run0/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer0/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..fd5ef0b38b56b19cb4efeb4cee2849abc01fff2d --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f229a80a7e3b812811412e433cefa797d7b303bfc3436071420fb6bfe1deb30e +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run0/train.log b/results_bvcc/w2v2_large_asr_de/layer0/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..667e1fa40596b3cd4a1c8f581ce2d65cb659cc7d --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/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=0.6172447300110108 +INFO:root: +[Valid][0][UTT][ MSE = 0.4743 | LCC = 0.6809 | SRCC = 0.6815 ] +INFO:root:Average loss=0.4128917619968072 +INFO:root: +[Valid][1][UTT][ MSE = 0.4013 | LCC = 0.7242 | SRCC = 0.7247 ] +INFO:root:Average loss=0.3743923060022868 +INFO:root: +[Valid][2][UTT][ MSE = 0.4723 | LCC = 0.7500 | SRCC = 0.7508 ] +INFO:root:Average loss=0.35041309893131256 +INFO:root: +[Valid][3][UTT][ MSE = 0.5224 | LCC = 0.7257 | SRCC = 0.7403 ] +INFO:root:Average loss=0.4082281621984946 +INFO:root: +[Valid][4][UTT][ MSE = 0.3685 | LCC = 0.7580 | SRCC = 0.7552 ] +INFO:root:Average loss=0.31481729466945696 +INFO:root: +[Valid][5][UTT][ MSE = 0.3537 | LCC = 0.7701 | SRCC = 0.7659 ] +INFO:root:Average loss=0.31541060293332124 +INFO:root: +[Valid][6][UTT][ MSE = 0.3411 | LCC = 0.7741 | SRCC = 0.7687 ] +INFO:root:Average loss=0.29828748871118593 +INFO:root: +[Valid][7][UTT][ MSE = 0.3692 | LCC = 0.7709 | SRCC = 0.7700 ] +INFO:root:Average loss=0.2948915939300488 +INFO:root: +[Valid][8][UTT][ MSE = 0.3766 | LCC = 0.7712 | SRCC = 0.7711 ] +INFO:root:Average loss=0.29434095571438473 +INFO:root: +[Valid][9][UTT][ MSE = 0.3736 | LCC = 0.7780 | SRCC = 0.7819 ] +INFO:root:Average loss=0.27653334862910783 +INFO:root: +[Valid][10][UTT][ MSE = 0.3553 | LCC = 0.7833 | SRCC = 0.7847 ] +INFO:root:Average loss=0.260523568934355 +INFO:root: +[Valid][11][UTT][ MSE = 0.3248 | LCC = 0.7964 | SRCC = 0.7924 ] +INFO:root:Average loss=0.25545856757805896 +INFO:root: +[Valid][12][UTT][ MSE = 0.3098 | LCC = 0.8103 | SRCC = 0.8070 ] +INFO:root:Average loss=0.25160193118529445 +INFO:root: +[Valid][13][UTT][ MSE = 0.3082 | LCC = 0.8094 | SRCC = 0.8062 ] +INFO:root:Average loss=0.23903538993535897 +INFO:root: +[Valid][14][UTT][ MSE = 0.3251 | LCC = 0.8056 | SRCC = 0.8058 ] +INFO:root:Average loss=0.23735840370257696 +INFO:root: +[Valid][15][UTT][ MSE = 0.3059 | LCC = 0.8139 | SRCC = 0.8111 ] +INFO:root:Average loss=0.22898403631571013 +INFO:root: +[Valid][16][UTT][ MSE = 0.2988 | LCC = 0.8109 | SRCC = 0.8106 ] +INFO:root:Average loss=0.2241018410676565 +INFO:root: +[Valid][17][UTT][ MSE = 0.2965 | LCC = 0.8104 | SRCC = 0.8062 ] +INFO:root:Average loss=0.2253491930090464 +INFO:root: +[Valid][18][UTT][ MSE = 0.3921 | LCC = 0.7993 | SRCC = 0.8053 ] +INFO:root:Average loss=0.22872768791440207 +INFO:root: +[Valid][19][UTT][ MSE = 0.2857 | LCC = 0.8159 | SRCC = 0.8159 ] +INFO:root:Average loss=0.21060849678439972 +INFO:root: +[Valid][20][UTT][ MSE = 0.3182 | LCC = 0.8089 | SRCC = 0.8046 ] +INFO:root:Average loss=0.20140083115070295 +INFO:root: +[Valid][21][UTT][ MSE = 0.2840 | LCC = 0.8154 | SRCC = 0.8113 ] +INFO:root:Average loss=0.2020682127047808 +INFO:root: +[Valid][22][UTT][ MSE = 0.2819 | LCC = 0.8251 | SRCC = 0.8245 ] +INFO:root:Average loss=0.1950214562507776 +INFO:root: +[Valid][23][UTT][ MSE = 0.2784 | LCC = 0.8202 | SRCC = 0.8173 ] +INFO:root:Average loss=0.19277595222378388 +INFO:root: +[Valid][24][UTT][ MSE = 0.2715 | LCC = 0.8231 | SRCC = 0.8228 ] +INFO:root:Average loss=0.18057630917964837 +INFO:root: +[Valid][25][UTT][ MSE = 0.2918 | LCC = 0.8223 | SRCC = 0.8240 ] +INFO:root:Average loss=0.18107678464207894 +INFO:root: +[Valid][26][UTT][ MSE = 0.2867 | LCC = 0.8182 | SRCC = 0.8196 ] +INFO:root:Average loss=0.17876275179859918 +INFO:root: +[Valid][27][UTT][ MSE = 0.2934 | LCC = 0.8198 | SRCC = 0.8147 ] +INFO:root:Average loss=0.17340032985577217 +INFO:root: +[Valid][28][UTT][ MSE = 0.2893 | LCC = 0.8255 | SRCC = 0.8232 ] +INFO:root:Average loss=0.16838809523062828 +INFO:root: +[Valid][29][UTT][ MSE = 0.2819 | LCC = 0.8209 | SRCC = 0.8229 ] +INFO:root:Average loss=0.16600188746666297 +INFO:root: +[Valid][30][UTT][ MSE = 0.2862 | LCC = 0.8235 | SRCC = 0.8219 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2832 | LCC = 0.8365 | SRCC = 0.8344 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run1/config.gin b/results_bvcc/w2v2_large_asr_de/layer0/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run1/model.pt b/results_bvcc/w2v2_large_asr_de/layer0/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bd8530b21fe3ffb404904197cab3f788fa535806 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a533516cbd7b4a15119db75a859df76fbf607122ceea0bfe57385b13d89df45d +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run1/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer0/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..12465ec034ca158fa264a3c2db0db52ef6bc4fa3 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f8911f7dcf7702d0df031b94893b0f6b51803fe0902e3742cb7c4297071d944f +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run1/train.log b/results_bvcc/w2v2_large_asr_de/layer0/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0f5e1a14b26122e146e1d1073ee591af209b67ff --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/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=0.5526466828126174 +INFO:root: +[Valid][0][UTT][ MSE = 0.4558 | LCC = 0.6908 | SRCC = 0.6889 ] +INFO:root:Average loss=0.4186242482601068 +INFO:root: +[Valid][1][UTT][ MSE = 0.4536 | LCC = 0.7150 | SRCC = 0.7092 ] +INFO:root:Average loss=0.3973239767245757 +INFO:root: +[Valid][2][UTT][ MSE = 0.4047 | LCC = 0.7348 | SRCC = 0.7305 ] +INFO:root:Average loss=0.3552419654069803 +INFO:root: +[Valid][3][UTT][ MSE = 0.3912 | LCC = 0.7477 | SRCC = 0.7407 ] +INFO:root:Average loss=0.3398699449040951 +INFO:root: +[Valid][4][UTT][ MSE = 0.3735 | LCC = 0.7546 | SRCC = 0.7474 ] +INFO:root:Average loss=0.31782706750509065 +INFO:root: +[Valid][5][UTT][ MSE = 0.3971 | LCC = 0.7658 | SRCC = 0.7665 ] +INFO:root:Average loss=0.3220853226689192 +INFO:root: +[Valid][6][UTT][ MSE = 0.3421 | LCC = 0.7792 | SRCC = 0.7735 ] +INFO:root:Average loss=0.2890723592195755 +INFO:root: +[Valid][7][UTT][ MSE = 0.3332 | LCC = 0.7840 | SRCC = 0.7786 ] +INFO:root:Average loss=0.27809147517650556 +INFO:root: +[Valid][8][UTT][ MSE = 0.3473 | LCC = 0.7898 | SRCC = 0.7901 ] +INFO:root:Average loss=0.2858888619603255 +INFO:root: +[Valid][9][UTT][ MSE = 0.3427 | LCC = 0.7993 | SRCC = 0.7976 ] +INFO:root:Average loss=0.285251704737162 +INFO:root: +[Valid][10][UTT][ MSE = 0.3637 | LCC = 0.7941 | SRCC = 0.7890 ] +INFO:root:Average loss=0.2657060947937843 +INFO:root: +[Valid][11][UTT][ MSE = 0.3305 | LCC = 0.7938 | SRCC = 0.7930 ] +INFO:root:Average loss=0.2583668419183829 +INFO:root: +[Valid][12][UTT][ MSE = 0.3627 | LCC = 0.8029 | SRCC = 0.8012 ] +INFO:root:Average loss=0.24498242827562186 +INFO:root: +[Valid][13][UTT][ MSE = 0.3015 | LCC = 0.8053 | SRCC = 0.8022 ] +INFO:root:Average loss=0.25183767481492114 +INFO:root: +[Valid][14][UTT][ MSE = 0.3267 | LCC = 0.8025 | SRCC = 0.8022 ] +INFO:root:Average loss=0.23320174465576807 +INFO:root: +[Valid][15][UTT][ MSE = 0.3005 | LCC = 0.8126 | SRCC = 0.8110 ] +INFO:root:Average loss=0.22714056418492243 +INFO:root: +[Valid][16][UTT][ MSE = 0.3113 | LCC = 0.8066 | SRCC = 0.8060 ] +INFO:root:Average loss=0.22282007069159776 +INFO:root: +[Valid][17][UTT][ MSE = 0.3163 | LCC = 0.8143 | SRCC = 0.8114 ] +INFO:root:Average loss=0.21815717067473975 +INFO:root: +[Valid][18][UTT][ MSE = 0.2838 | LCC = 0.8114 | SRCC = 0.8093 ] +INFO:root:Average loss=0.21844583482314378 +INFO:root: +[Valid][19][UTT][ MSE = 0.3083 | LCC = 0.8119 | SRCC = 0.8123 ] +INFO:root:Average loss=0.20276173051351157 +INFO:root: +[Valid][20][UTT][ MSE = 0.2947 | LCC = 0.8137 | SRCC = 0.8133 ] +INFO:root:Average loss=0.19854514549175897 +INFO:root: +[Valid][21][UTT][ MSE = 0.3540 | LCC = 0.8153 | SRCC = 0.8166 ] +INFO:root:Average loss=0.1942321988634574 +INFO:root: +[Valid][22][UTT][ MSE = 0.3189 | LCC = 0.8161 | SRCC = 0.8170 ] +INFO:root:Average loss=0.1892010665092713 +INFO:root: +[Valid][23][UTT][ MSE = 0.2829 | LCC = 0.8213 | SRCC = 0.8172 ] +INFO:root:Average loss=0.18133346115549406 +INFO:root: +[Valid][24][UTT][ MSE = 0.3335 | LCC = 0.8132 | SRCC = 0.8133 ] +INFO:root:Average loss=0.1834619508530849 +INFO:root: +[Valid][25][UTT][ MSE = 0.2731 | LCC = 0.8282 | SRCC = 0.8264 ] +INFO:root:Average loss=0.1772554065936651 +INFO:root: +[Valid][26][UTT][ MSE = 0.3043 | LCC = 0.8235 | SRCC = 0.8214 ] +INFO:root:Average loss=0.17276299085754615 +INFO:root: +[Valid][27][UTT][ MSE = 0.2723 | LCC = 0.8261 | SRCC = 0.8235 ] +INFO:root:Average loss=0.17461665079761773 +INFO:root: +[Valid][28][UTT][ MSE = 0.2947 | LCC = 0.8238 | SRCC = 0.8198 ] +INFO:root:Average loss=0.17634899704120097 +INFO:root: +[Valid][29][UTT][ MSE = 0.2939 | LCC = 0.8268 | SRCC = 0.8250 ] +INFO:root:Average loss=0.16198038367124704 +INFO:root: +[Valid][30][UTT][ MSE = 0.2786 | LCC = 0.8185 | SRCC = 0.8185 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2899 | LCC = 0.8257 | SRCC = 0.8222 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run2/config.gin b/results_bvcc/w2v2_large_asr_de/layer0/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run2/model.pt b/results_bvcc/w2v2_large_asr_de/layer0/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9dc024948f5cb8b3a19f825fdd4d94133c1173cf --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:36d00b88d002bfadb8bf59898971b2af67b69452dd631a8946b6c258f4991f2b +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run2/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer0/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ff70f8b239e797502fb50dd631a2059db4d983ee --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:01438344be485639d1e8051e6d038ccf0463cb8269bf1c14460268c92a98bcbb +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run2/train.log b/results_bvcc/w2v2_large_asr_de/layer0/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a5c461042ef7aef17c385ae4cbb967598eff4187 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/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=0.6065079122781754 +INFO:root: +[Valid][0][UTT][ MSE = 0.5764 | LCC = 0.6628 | SRCC = 0.6820 ] +INFO:root:Average loss=0.4108311900725731 +INFO:root: +[Valid][1][UTT][ MSE = 0.4626 | LCC = 0.7185 | SRCC = 0.7162 ] +INFO:root:Average loss=0.3689121974584384 +INFO:root: +[Valid][2][UTT][ MSE = 0.5095 | LCC = 0.7243 | SRCC = 0.7213 ] +INFO:root:Average loss=0.3495283688490207 +INFO:root: +[Valid][3][UTT][ MSE = 0.3831 | LCC = 0.7504 | SRCC = 0.7495 ] +INFO:root:Average loss=0.3317761570215225 +INFO:root: +[Valid][4][UTT][ MSE = 0.4027 | LCC = 0.7567 | SRCC = 0.7584 ] +INFO:root:Average loss=0.315953798018969 +INFO:root: +[Valid][5][UTT][ MSE = 0.4996 | LCC = 0.7599 | SRCC = 0.7633 ] +INFO:root:Average loss=0.3136767156613179 +INFO:root: +[Valid][6][UTT][ MSE = 0.3613 | LCC = 0.7789 | SRCC = 0.7776 ] +INFO:root:Average loss=0.29475720933614635 +INFO:root: +[Valid][7][UTT][ MSE = 0.3837 | LCC = 0.7877 | SRCC = 0.7857 ] +INFO:root:Average loss=0.2978175796377353 +INFO:root: +[Valid][8][UTT][ MSE = 0.3454 | LCC = 0.7769 | SRCC = 0.7775 ] +INFO:root:Average loss=0.2946078305443128 +INFO:root: +[Valid][9][UTT][ MSE = 0.3791 | LCC = 0.7860 | SRCC = 0.7884 ] +INFO:root:Average loss=0.27496714660754573 +INFO:root: +[Valid][10][UTT][ MSE = 0.3270 | LCC = 0.7941 | SRCC = 0.7942 ] +INFO:root:Average loss=0.28162058404622936 +INFO:root: +[Valid][11][UTT][ MSE = 0.4987 | LCC = 0.7696 | SRCC = 0.7704 ] +INFO:root:Average loss=0.2633321397961714 +INFO:root: +[Valid][12][UTT][ MSE = 0.3258 | LCC = 0.8022 | SRCC = 0.8008 ] +INFO:root:Average loss=0.2419839415412683 +INFO:root: +[Valid][13][UTT][ MSE = 0.3145 | LCC = 0.8047 | SRCC = 0.8064 ] +INFO:root:Average loss=0.25210349414593136 +INFO:root: +[Valid][14][UTT][ MSE = 0.3149 | LCC = 0.8003 | SRCC = 0.8007 ] +INFO:root:Average loss=0.2388275941977134 +INFO:root: +[Valid][15][UTT][ MSE = 0.3215 | LCC = 0.8031 | SRCC = 0.8014 ] +INFO:root:Average loss=0.23393049377661485 +INFO:root: +[Valid][16][UTT][ MSE = 0.3193 | LCC = 0.8019 | SRCC = 0.7980 ] +INFO:root:Average loss=0.22544086361542726 +INFO:root: +[Valid][17][UTT][ MSE = 0.3134 | LCC = 0.8090 | SRCC = 0.8085 ] +INFO:root:Average loss=0.22271475214988756 +INFO:root: +[Valid][18][UTT][ MSE = 0.3251 | LCC = 0.8115 | SRCC = 0.8121 ] +INFO:root:Average loss=0.217414962939727 +INFO:root: +[Valid][19][UTT][ MSE = 0.2980 | LCC = 0.8163 | SRCC = 0.8155 ] +INFO:root:Average loss=0.21556153702430236 +INFO:root: +[Valid][20][UTT][ MSE = 0.3130 | LCC = 0.8172 | SRCC = 0.8150 ] +INFO:root:Average loss=0.21353446539395896 +INFO:root: +[Valid][21][UTT][ MSE = 0.3158 | LCC = 0.8057 | SRCC = 0.8029 ] +INFO:root:Average loss=0.2105002938172756 +INFO:root: +[Valid][22][UTT][ MSE = 0.3017 | LCC = 0.8186 | SRCC = 0.8188 ] +INFO:root:Average loss=0.20859590363808167 +INFO:root: +[Valid][23][UTT][ MSE = 0.2970 | LCC = 0.8115 | SRCC = 0.8123 ] +INFO:root:Average loss=0.19932512738383734 +INFO:root: +[Valid][24][UTT][ MSE = 0.3177 | LCC = 0.8159 | SRCC = 0.8152 ] +INFO:root:Average loss=0.17993423657921645 +INFO:root: +[Valid][25][UTT][ MSE = 0.2812 | LCC = 0.8198 | SRCC = 0.8187 ] +INFO:root:Average loss=0.18461685627698898 +INFO:root: +[Valid][26][UTT][ MSE = 0.2897 | LCC = 0.8202 | SRCC = 0.8175 ] +INFO:root:Average loss=0.19267338562088135 +INFO:root: +[Valid][27][UTT][ MSE = 0.3020 | LCC = 0.8154 | SRCC = 0.8158 ] +INFO:root:Average loss=0.1701152533865892 +INFO:root: +[Valid][28][UTT][ MSE = 0.2948 | LCC = 0.8141 | SRCC = 0.8123 ] +INFO:root:Average loss=0.1709772870899775 +INFO:root: +[Valid][29][UTT][ MSE = 0.3279 | LCC = 0.8204 | SRCC = 0.8192 ] +INFO:root:Average loss=0.17327634875590986 +INFO:root: +[Valid][30][UTT][ MSE = 0.3710 | LCC = 0.8160 | SRCC = 0.8170 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2734 | LCC = 0.8377 | SRCC = 0.8337 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run3/config.gin b/results_bvcc/w2v2_large_asr_de/layer0/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run3/model.pt b/results_bvcc/w2v2_large_asr_de/layer0/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..50af4f289545946f9015f834a3f6873152991b5e --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2c70cca657362570645e478b276c4c0545e559e952de39bcc80f1a36737ac041 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run3/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer0/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..78dce2481d2eea8bf3921757e2faab85f74ff5e2 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:27efcf1e978c93cf92d589c3b7f45a3baa13d37b89725cae2b5fbae57972750a +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run3/train.log b/results_bvcc/w2v2_large_asr_de/layer0/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0b0c12b2976fab8588a28b130e5719ecfba8eace --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/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=0.5408154118519563 +INFO:root: +[Valid][0][UTT][ MSE = 0.4428 | LCC = 0.6973 | SRCC = 0.6945 ] +INFO:root:Average loss=0.395339629588983 +INFO:root: +[Valid][1][UTT][ MSE = 0.4047 | LCC = 0.7317 | SRCC = 0.7298 ] +INFO:root:Average loss=0.35061968843906355 +INFO:root: +[Valid][2][UTT][ MSE = 0.4039 | LCC = 0.7427 | SRCC = 0.7415 ] +INFO:root:Average loss=0.35119049403911984 +INFO:root: +[Valid][3][UTT][ MSE = 0.3721 | LCC = 0.7518 | SRCC = 0.7507 ] +INFO:root:Average loss=0.3419664249970363 +INFO:root: +[Valid][4][UTT][ MSE = 0.3985 | LCC = 0.7665 | SRCC = 0.7652 ] +INFO:root:Average loss=0.32191711301222825 +INFO:root: +[Valid][5][UTT][ MSE = 0.4017 | LCC = 0.7696 | SRCC = 0.7675 ] +INFO:root:Average loss=0.29814151521676624 +INFO:root: +[Valid][6][UTT][ MSE = 0.3389 | LCC = 0.7852 | SRCC = 0.7834 ] +INFO:root:Average loss=0.29333454809891873 +INFO:root: +[Valid][7][UTT][ MSE = 0.3766 | LCC = 0.7849 | SRCC = 0.7851 ] +INFO:root:Average loss=0.3001756030015456 +INFO:root: +[Valid][8][UTT][ MSE = 0.3459 | LCC = 0.7864 | SRCC = 0.7881 ] +INFO:root:Average loss=0.27399129630663455 +INFO:root: +[Valid][9][UTT][ MSE = 0.3249 | LCC = 0.7875 | SRCC = 0.7894 ] +INFO:root:Average loss=0.272205949593813 +INFO:root: +[Valid][10][UTT][ MSE = 0.3162 | LCC = 0.7953 | SRCC = 0.7978 ] +INFO:root:Average loss=0.27098072816928226 +INFO:root: +[Valid][11][UTT][ MSE = 0.3172 | LCC = 0.7997 | SRCC = 0.7978 ] +INFO:root:Average loss=0.2474273825303102 +INFO:root: +[Valid][12][UTT][ MSE = 0.3232 | LCC = 0.7968 | SRCC = 0.7950 ] +INFO:root:Average loss=0.25278799274028874 +INFO:root: +[Valid][13][UTT][ MSE = 0.3023 | LCC = 0.8003 | SRCC = 0.8006 ] +INFO:root:Average loss=0.2611511026819547 +INFO:root: +[Valid][14][UTT][ MSE = 0.3373 | LCC = 0.8033 | SRCC = 0.8042 ] +INFO:root:Average loss=0.26870559041316694 +INFO:root: +[Valid][15][UTT][ MSE = 0.3045 | LCC = 0.7949 | SRCC = 0.8011 ] +INFO:root:Average loss=0.2382044412004642 +INFO:root: +[Valid][16][UTT][ MSE = 0.3020 | LCC = 0.8072 | SRCC = 0.8086 ] +INFO:root:Average loss=0.22435293480371818 +INFO:root: +[Valid][17][UTT][ MSE = 0.3209 | LCC = 0.8056 | SRCC = 0.8062 ] +INFO:root:Average loss=0.21828115655061525 +INFO:root: +[Valid][18][UTT][ MSE = 0.3163 | LCC = 0.8050 | SRCC = 0.8050 ] +INFO:root:Average loss=0.21096639239635223 +INFO:root: +[Valid][19][UTT][ MSE = 0.3058 | LCC = 0.8012 | SRCC = 0.8034 ] +INFO:root:Average loss=0.2061458035157277 +INFO:root: +[Valid][20][UTT][ MSE = 0.3097 | LCC = 0.8058 | SRCC = 0.8079 ] +INFO:root:Average loss=0.20293491448347384 +INFO:root: +[Valid][21][UTT][ MSE = 0.3115 | LCC = 0.8055 | SRCC = 0.8098 ] +INFO:root:Average loss=0.2002112022959269 +INFO:root: +[Valid][22][UTT][ MSE = 0.3364 | LCC = 0.8017 | SRCC = 0.8076 ] +INFO:root:Average loss=0.1871980510843106 +INFO:root: +[Valid][23][UTT][ MSE = 0.3357 | LCC = 0.8128 | SRCC = 0.8115 ] +INFO:root:Average loss=0.18334462331273618 +INFO:root: +[Valid][24][UTT][ MSE = 0.2986 | LCC = 0.8044 | SRCC = 0.8062 ] +INFO:root:Average loss=0.18153688674553847 +INFO:root: +[Valid][25][UTT][ MSE = 0.2848 | LCC = 0.8181 | SRCC = 0.8174 ] +INFO:root:Average loss=0.17143495371326423 +INFO:root: +[Valid][26][UTT][ MSE = 0.2893 | LCC = 0.8119 | SRCC = 0.8130 ] +INFO:root:Average loss=0.1779894846945237 +INFO:root: +[Valid][27][UTT][ MSE = 0.2879 | LCC = 0.8162 | SRCC = 0.8152 ] +INFO:root:Average loss=0.16031512388816246 +INFO:root: +[Valid][28][UTT][ MSE = 0.2771 | LCC = 0.8159 | SRCC = 0.8182 ] +INFO:root:Average loss=0.15867479967001158 +INFO:root: +[Valid][29][UTT][ MSE = 0.3335 | LCC = 0.8073 | SRCC = 0.8130 ] +INFO:root:Average loss=0.17168127477933198 +INFO:root: +[Valid][30][UTT][ MSE = 0.2907 | LCC = 0.8198 | SRCC = 0.8199 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3070 | LCC = 0.8317 | SRCC = 0.8324 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run4/config.gin b/results_bvcc/w2v2_large_asr_de/layer0/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run4/model.pt b/results_bvcc/w2v2_large_asr_de/layer0/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..376c1dfabd0abb3a42049291bc8955030e3af66c --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:165a4327bd5f64b83934de50063c3d54803e68c9b9bb8be9ca6e627311f20d9f +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run4/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer0/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..afbbe8ac4f22c1630dffac9a077e882529a40a4f --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer0/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d941730b469d57a52216005c721caad491398b9d7f6bfa498f2afca8bf2f3b72 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer0/run4/train.log b/results_bvcc/w2v2_large_asr_de/layer0/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b2627370914f7d873800f7dd12ea6cfd736afc82 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/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=0.5531979378981468 +INFO:root: +[Valid][0][UTT][ MSE = 0.4911 | LCC = 0.6923 | SRCC = 0.6914 ] +INFO:root:Average loss=0.41761438663189226 +INFO:root: +[Valid][1][UTT][ MSE = 0.5765 | LCC = 0.7135 | SRCC = 0.7263 ] +INFO:root:Average loss=0.39313523605083806 +INFO:root: +[Valid][2][UTT][ MSE = 0.4113 | LCC = 0.7416 | SRCC = 0.7347 ] +INFO:root:Average loss=0.3602299959613727 +INFO:root: +[Valid][3][UTT][ MSE = 0.3770 | LCC = 0.7513 | SRCC = 0.7498 ] +INFO:root:Average loss=0.3385853679516377 +INFO:root: +[Valid][4][UTT][ MSE = 0.3761 | LCC = 0.7696 | SRCC = 0.7657 ] +INFO:root:Average loss=0.34570154050985974 +INFO:root: +[Valid][5][UTT][ MSE = 0.3707 | LCC = 0.7626 | SRCC = 0.7568 ] +INFO:root:Average loss=0.3068122359422537 +INFO:root: +[Valid][6][UTT][ MSE = 0.3509 | LCC = 0.7750 | SRCC = 0.7696 ] +INFO:root:Average loss=0.3034033781060806 +INFO:root: +[Valid][7][UTT][ MSE = 0.3372 | LCC = 0.7839 | SRCC = 0.7785 ] +INFO:root:Average loss=0.2868316028362665 +INFO:root: +[Valid][8][UTT][ MSE = 0.3246 | LCC = 0.7889 | SRCC = 0.7832 ] +INFO:root:Average loss=0.28298015682361066 +INFO:root: +[Valid][9][UTT][ MSE = 0.3426 | LCC = 0.7968 | SRCC = 0.7905 ] +INFO:root:Average loss=0.2694217805296947 +INFO:root: +[Valid][10][UTT][ MSE = 0.3454 | LCC = 0.7891 | SRCC = 0.7811 ] +INFO:root:Average loss=0.2620553080088053 +INFO:root: +[Valid][11][UTT][ MSE = 0.3146 | LCC = 0.7956 | SRCC = 0.7922 ] +INFO:root:Average loss=0.24348831272278076 +INFO:root: +[Valid][12][UTT][ MSE = 0.3408 | LCC = 0.7982 | SRCC = 0.7926 ] +INFO:root:Average loss=0.26037104217669904 +INFO:root: +[Valid][13][UTT][ MSE = 0.3093 | LCC = 0.8081 | SRCC = 0.8060 ] +INFO:root:Average loss=0.23838678976664177 +INFO:root: +[Valid][14][UTT][ MSE = 0.3176 | LCC = 0.8154 | SRCC = 0.8130 ] +INFO:root:Average loss=0.23226559773469582 +INFO:root: +[Valid][15][UTT][ MSE = 0.3023 | LCC = 0.8065 | SRCC = 0.8062 ] +INFO:root:Average loss=0.2501130216778853 +INFO:root: +[Valid][16][UTT][ MSE = 0.3248 | LCC = 0.8082 | SRCC = 0.8046 ] +INFO:root:Average loss=0.22388031371893027 +INFO:root: +[Valid][17][UTT][ MSE = 0.3182 | LCC = 0.8154 | SRCC = 0.8155 ] +INFO:root:Average loss=0.2346612056478476 +INFO:root: +[Valid][18][UTT][ MSE = 0.2874 | LCC = 0.8110 | SRCC = 0.8086 ] +INFO:root:Average loss=0.2241645988363486 +INFO:root: +[Valid][19][UTT][ MSE = 0.3071 | LCC = 0.8125 | SRCC = 0.8100 ] +INFO:root:Average loss=0.2361613787137545 +INFO:root: +[Valid][20][UTT][ MSE = 0.2949 | LCC = 0.8150 | SRCC = 0.8125 ] +INFO:root:Average loss=0.20930511542619804 +INFO:root: +[Valid][21][UTT][ MSE = 0.3208 | LCC = 0.8195 | SRCC = 0.8168 ] +INFO:root:Average loss=0.21074356119602156 +INFO:root: +[Valid][22][UTT][ MSE = 0.2976 | LCC = 0.8176 | SRCC = 0.8148 ] +INFO:root:Average loss=0.20035695437437448 +INFO:root: +[Valid][23][UTT][ MSE = 0.3071 | LCC = 0.8178 | SRCC = 0.8170 ] +INFO:root:Average loss=0.19130446857366806 +INFO:root: +[Valid][24][UTT][ MSE = 0.2789 | LCC = 0.8215 | SRCC = 0.8197 ] +INFO:root:Average loss=0.18140169901725572 +INFO:root: +[Valid][25][UTT][ MSE = 0.2900 | LCC = 0.8236 | SRCC = 0.8215 ] +INFO:root:Average loss=0.183766113450894 +INFO:root: +[Valid][26][UTT][ MSE = 0.2771 | LCC = 0.8221 | SRCC = 0.8211 ] +INFO:root:Average loss=0.19223522308927316 +INFO:root: +[Valid][27][UTT][ MSE = 0.2797 | LCC = 0.8246 | SRCC = 0.8224 ] +INFO:root:Average loss=0.1723972353606652 +INFO:root: +[Valid][28][UTT][ MSE = 0.2796 | LCC = 0.8271 | SRCC = 0.8275 ] +INFO:root:Average loss=0.16563425088922182 +INFO:root: +[Valid][29][UTT][ MSE = 0.3251 | LCC = 0.8187 | SRCC = 0.8211 ] +INFO:root:Average loss=0.1697640088506234 +INFO:root: +[Valid][30][UTT][ MSE = 0.2759 | LCC = 0.8317 | SRCC = 0.8325 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3206 | LCC = 0.8230 | SRCC = 0.8244 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run0/config.gin b/results_bvcc/w2v2_large_asr_de/layer10/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run0/model.pt b/results_bvcc/w2v2_large_asr_de/layer10/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..70e00d334aaa3c29470cf091044aa73adaea6431 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer10/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b57ec1f0d3c31d6a5ab90d17c9eb6ba4a91584584d28c279215251ccc677dd5f +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run0/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer10/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..bd9672c636e2e82b13be69ce82bddb32ff219370 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer10/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7a9708087b41baf10c29dbc1238bb7c5b8ce527c91df684db0a0936e06de9ce3 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run0/train.log b/results_bvcc/w2v2_large_asr_de/layer10/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..8fe96ae424546d97927a0d3a750eb5df19e3204a --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/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=0.5020410840710005 +INFO:root: +[Valid][0][UTT][ MSE = 0.4301 | LCC = 0.7738 | SRCC = 0.7819 ] +INFO:root:Average loss=0.31567535262841445 +INFO:root: +[Valid][1][UTT][ MSE = 0.3240 | LCC = 0.8050 | SRCC = 0.8065 ] +INFO:root:Average loss=0.2796969963954045 +INFO:root: +[Valid][2][UTT][ MSE = 0.4522 | LCC = 0.8103 | SRCC = 0.8132 ] +INFO:root:Average loss=0.2623311086342885 +INFO:root: +[Valid][3][UTT][ MSE = 0.3202 | LCC = 0.8207 | SRCC = 0.8195 ] +INFO:root:Average loss=0.22723613450160393 +INFO:root: +[Valid][4][UTT][ MSE = 0.3132 | LCC = 0.8232 | SRCC = 0.8214 ] +INFO:root:Average loss=0.2097283959006652 +INFO:root: +[Valid][5][UTT][ MSE = 0.4270 | LCC = 0.8077 | SRCC = 0.8150 ] +INFO:root:Average loss=0.2035641589990029 +INFO:root: +[Valid][6][UTT][ MSE = 0.3792 | LCC = 0.8173 | SRCC = 0.8207 ] +INFO:root:Average loss=0.18233560799406126 +INFO:root: +[Valid][7][UTT][ MSE = 0.3354 | LCC = 0.8197 | SRCC = 0.8224 ] +INFO:root:Average loss=0.17653203278015822 +INFO:root: +[Valid][8][UTT][ MSE = 0.2878 | LCC = 0.8239 | SRCC = 0.8252 ] +INFO:root:Average loss=0.16415260980526605 +INFO:root: +[Valid][9][UTT][ MSE = 0.3028 | LCC = 0.8256 | SRCC = 0.8249 ] +INFO:root:Average loss=0.1531685089262632 +INFO:root: +[Valid][10][UTT][ MSE = 0.3074 | LCC = 0.8212 | SRCC = 0.8207 ] +INFO:root:Average loss=0.1584326663078406 +INFO:root: +[Valid][11][UTT][ MSE = 0.3512 | LCC = 0.8226 | SRCC = 0.8271 ] +INFO:root:Average loss=0.13584643917588088 +INFO:root: +[Valid][12][UTT][ MSE = 0.3652 | LCC = 0.8123 | SRCC = 0.8168 ] +INFO:root:Average loss=0.13238409161567688 +INFO:root: +[Valid][13][UTT][ MSE = 0.2828 | LCC = 0.8220 | SRCC = 0.8226 ] +INFO:root:Average loss=0.12722358289055336 +INFO:root: +[Valid][14][UTT][ MSE = 0.3547 | LCC = 0.8239 | SRCC = 0.8277 ] +INFO:root:Average loss=0.12115969728582944 +INFO:root: +[Valid][15][UTT][ MSE = 0.4975 | LCC = 0.8112 | SRCC = 0.8207 ] +INFO:root:Average loss=0.12137557279605132 +INFO:root: +[Valid][16][UTT][ MSE = 0.3023 | LCC = 0.8196 | SRCC = 0.8205 ] +INFO:root:Average loss=0.12458381123649769 +INFO:root: +[Valid][17][UTT][ MSE = 0.2816 | LCC = 0.8268 | SRCC = 0.8267 ] +INFO:root:Average loss=0.10001740910303898 +INFO:root: +[Valid][18][UTT][ MSE = 0.2842 | LCC = 0.8226 | SRCC = 0.8205 ] +INFO:root:Average loss=0.09196623032673812 +INFO:root: +[Valid][19][UTT][ MSE = 0.3963 | LCC = 0.8140 | SRCC = 0.8203 ] +INFO:root:Average loss=0.08779883332168445 +INFO:root: +[Valid][20][UTT][ MSE = 0.4118 | LCC = 0.8128 | SRCC = 0.8210 ] +INFO:root:Average loss=0.08260605192910402 +INFO:root: +[Valid][21][UTT][ MSE = 0.3876 | LCC = 0.8184 | SRCC = 0.8238 ] +INFO:root:Average loss=0.0750254870702823 +INFO:root: +[Valid][22][UTT][ MSE = 0.3581 | LCC = 0.8232 | SRCC = 0.8277 ] +INFO:root:Average loss=0.07434255519929604 +INFO:root: +[Valid][23][UTT][ MSE = 0.2915 | LCC = 0.8223 | SRCC = 0.8212 ] +INFO:root:Average loss=0.07379694340320733 +INFO:root: +[Valid][24][UTT][ MSE = 0.3178 | LCC = 0.8250 | SRCC = 0.8247 ] +INFO:root:Average loss=0.06813481698433559 +INFO:root: +[Valid][25][UTT][ MSE = 0.3165 | LCC = 0.8230 | SRCC = 0.8250 ] +INFO:root:Average loss=0.06897854241423118 +INFO:root: +[Valid][26][UTT][ MSE = 0.3101 | LCC = 0.8225 | SRCC = 0.8218 ] +INFO:root:Average loss=0.06258222890587953 +INFO:root: +[Valid][27][UTT][ MSE = 0.3976 | LCC = 0.8148 | SRCC = 0.8216 ] +INFO:root:Average loss=0.06297604825634223 +INFO:root: +[Valid][28][UTT][ MSE = 0.3184 | LCC = 0.8231 | SRCC = 0.8230 ] +INFO:root:Average loss=0.05521759873208327 +INFO:root: +[Valid][29][UTT][ MSE = 0.2809 | LCC = 0.8270 | SRCC = 0.8256 ] +INFO:root:Average loss=0.05770857054262589 +INFO:root: +[Valid][30][UTT][ MSE = 0.3342 | LCC = 0.8224 | SRCC = 0.8238 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2705 | LCC = 0.8276 | SRCC = 0.8270 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run1/config.gin b/results_bvcc/w2v2_large_asr_de/layer10/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run1/model.pt b/results_bvcc/w2v2_large_asr_de/layer10/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5b49e1c3bc2e9a4fd19a476e79304187826921b5 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer10/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3708386e1ef48bc41cb6bc79a4c84b0c8e29c526e23d7d8d89a9872b5e88ce77 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run1/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer10/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..009f55a8e24407b113ee656787e66dadbdeda346 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer10/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:940142e5d84f873e590e8a8290d2834c5ec516a3ebb334723f981984e8e151c9 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run1/train.log b/results_bvcc/w2v2_large_asr_de/layer10/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2d84fa61dcb0a59654e6f78f34c56c3a94eb501a --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/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=0.5553148576082327 +INFO:root: +[Valid][0][UTT][ MSE = 0.3986 | LCC = 0.7566 | SRCC = 0.7593 ] +INFO:root:Average loss=0.336250765583454 +INFO:root: +[Valid][1][UTT][ MSE = 0.3581 | LCC = 0.7947 | SRCC = 0.7973 ] +INFO:root:Average loss=0.28989177006177413 +INFO:root: +[Valid][2][UTT][ MSE = 0.3573 | LCC = 0.8076 | SRCC = 0.8068 ] +INFO:root:Average loss=0.2626003869451009 +INFO:root: +[Valid][3][UTT][ MSE = 0.3153 | LCC = 0.8136 | SRCC = 0.8127 ] +INFO:root:Average loss=0.2400206753458732 +INFO:root: +[Valid][4][UTT][ MSE = 0.3257 | LCC = 0.8197 | SRCC = 0.8177 ] +INFO:root:Average loss=0.23666226176115182 +INFO:root: +[Valid][5][UTT][ MSE = 0.3213 | LCC = 0.8162 | SRCC = 0.8178 ] +INFO:root:Average loss=0.20557998216305023 +INFO:root: +[Valid][6][UTT][ MSE = 0.2921 | LCC = 0.8213 | SRCC = 0.8198 ] +INFO:root:Average loss=0.19359000084491876 +INFO:root: +[Valid][7][UTT][ MSE = 0.3217 | LCC = 0.8165 | SRCC = 0.8200 ] +INFO:root:Average loss=0.1871856516943528 +INFO:root: +[Valid][8][UTT][ MSE = 0.2918 | LCC = 0.8191 | SRCC = 0.8177 ] +INFO:root:Average loss=0.17344086760511765 +INFO:root: +[Valid][9][UTT][ MSE = 0.3514 | LCC = 0.8094 | SRCC = 0.8114 ] +INFO:root:Average loss=0.16376554794036424 +INFO:root: +[Valid][10][UTT][ MSE = 0.2947 | LCC = 0.8127 | SRCC = 0.8110 ] +INFO:root:Average loss=0.17408961497056177 +INFO:root: +[Valid][11][UTT][ MSE = 0.2995 | LCC = 0.8140 | SRCC = 0.8122 ] +INFO:root:Average loss=0.14746656307043174 +INFO:root: +[Valid][12][UTT][ MSE = 0.3251 | LCC = 0.8100 | SRCC = 0.8114 ] +INFO:root:Average loss=0.13751384272025183 +INFO:root: +[Valid][13][UTT][ MSE = 0.2825 | LCC = 0.8219 | SRCC = 0.8216 ] +INFO:root:Average loss=0.13346601993991777 +INFO:root: +[Valid][14][UTT][ MSE = 0.3199 | LCC = 0.8200 | SRCC = 0.8218 ] +INFO:root:Average loss=0.11513922955745305 +INFO:root: +[Valid][15][UTT][ MSE = 0.2858 | LCC = 0.8129 | SRCC = 0.8133 ] +INFO:root:Average loss=0.1239815710637814 +INFO:root: +[Valid][16][UTT][ MSE = 0.2986 | LCC = 0.8124 | SRCC = 0.8156 ] +INFO:root:Average loss=0.11400143085764004 +INFO:root: +[Valid][17][UTT][ MSE = 0.3384 | LCC = 0.8127 | SRCC = 0.8149 ] +INFO:root:Average loss=0.10485056014015125 +INFO:root: +[Valid][18][UTT][ MSE = 0.2818 | LCC = 0.8148 | SRCC = 0.8158 ] +INFO:root:Average loss=0.09312896764813325 +INFO:root: +[Valid][19][UTT][ MSE = 0.2858 | LCC = 0.8111 | SRCC = 0.8143 ] +INFO:root:Average loss=0.08786759988810772 +INFO:root: +[Valid][20][UTT][ MSE = 0.2956 | LCC = 0.8088 | SRCC = 0.8114 ] +INFO:root:Average loss=0.09450433856974809 +INFO:root: +[Valid][21][UTT][ MSE = 0.2958 | LCC = 0.8105 | SRCC = 0.8115 ] +INFO:root:Average loss=0.08375900945602319 +INFO:root: +[Valid][22][UTT][ MSE = 0.2863 | LCC = 0.8122 | SRCC = 0.8134 ] +INFO:root:Average loss=0.08003293092434223 +INFO:root: +[Valid][23][UTT][ MSE = 0.2975 | LCC = 0.8111 | SRCC = 0.8114 ] +INFO:root:Average loss=0.07468575492310218 +INFO:root: +[Valid][24][UTT][ MSE = 0.2861 | LCC = 0.8102 | SRCC = 0.8110 ] +INFO:root:Average loss=0.06894085145531556 +INFO:root: +[Valid][25][UTT][ MSE = 0.3069 | LCC = 0.8108 | SRCC = 0.8125 ] +INFO:root:Average loss=0.07019786450725335 +INFO:root: +[Valid][26][UTT][ MSE = 0.2866 | LCC = 0.8096 | SRCC = 0.8108 ] +INFO:root:Average loss=0.06578636346146083 +INFO:root: +[Valid][27][UTT][ MSE = 0.2893 | LCC = 0.8090 | SRCC = 0.8102 ] +INFO:root:Average loss=0.06992785006952591 +INFO:root: +[Valid][28][UTT][ MSE = 0.3081 | LCC = 0.8098 | SRCC = 0.8137 ] +INFO:root:Average loss=0.06048895982213509 +INFO:root: +[Valid][29][UTT][ MSE = 0.3238 | LCC = 0.8104 | SRCC = 0.8120 ] +INFO:root:Average loss=0.05707599308628302 +INFO:root: +[Valid][30][UTT][ MSE = 0.3091 | LCC = 0.8095 | SRCC = 0.8104 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2719 | LCC = 0.8261 | SRCC = 0.8259 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run2/config.gin b/results_bvcc/w2v2_large_asr_de/layer10/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run2/model.pt b/results_bvcc/w2v2_large_asr_de/layer10/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..32649505f3bed7358cfe61ef0652c988be3b51d1 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer10/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc8351320bf248448432fbed60389061d284dd807550b4cf0bc149a2242c70a2 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run2/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer10/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..bdbd681b773a57bb1bb7655a9051eb17e3000496 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer10/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a59942b03aef037693589aeeff84661a231216bfbb757a82887ec21bcae9c438 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run2/train.log b/results_bvcc/w2v2_large_asr_de/layer10/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..37c188ed683cfe9658753db3a22fd738d0764681 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/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=0.5969993082376627 +INFO:root: +[Valid][0][UTT][ MSE = 0.3990 | LCC = 0.7626 | SRCC = 0.7666 ] +INFO:root:Average loss=0.3351117745041847 +INFO:root: +[Valid][1][UTT][ MSE = 0.3421 | LCC = 0.8017 | SRCC = 0.8020 ] +INFO:root:Average loss=0.27760989772967803 +INFO:root: +[Valid][2][UTT][ MSE = 0.2978 | LCC = 0.8197 | SRCC = 0.8179 ] +INFO:root:Average loss=0.2687001363971295 +INFO:root: +[Valid][3][UTT][ MSE = 0.3039 | LCC = 0.8212 | SRCC = 0.8196 ] +INFO:root:Average loss=0.2425610268345246 +INFO:root: +[Valid][4][UTT][ MSE = 0.3721 | LCC = 0.8225 | SRCC = 0.8244 ] +INFO:root:Average loss=0.21773937516487563 +INFO:root: +[Valid][5][UTT][ MSE = 0.2737 | LCC = 0.8291 | SRCC = 0.8270 ] +INFO:root:Average loss=0.2101348768442105 +INFO:root: +[Valid][6][UTT][ MSE = 0.3662 | LCC = 0.8185 | SRCC = 0.8198 ] +INFO:root:Average loss=0.1998729944611207 +INFO:root: +[Valid][7][UTT][ MSE = 0.2849 | LCC = 0.8250 | SRCC = 0.8238 ] +INFO:root:Average loss=0.1759980610356881 +INFO:root: +[Valid][8][UTT][ MSE = 0.2829 | LCC = 0.8253 | SRCC = 0.8235 ] +INFO:root:Average loss=0.16694274296363196 +INFO:root: +[Valid][9][UTT][ MSE = 0.2878 | LCC = 0.8211 | SRCC = 0.8171 ] +INFO:root:Average loss=0.16531188700061578 +INFO:root: +[Valid][10][UTT][ MSE = 0.3038 | LCC = 0.8211 | SRCC = 0.8202 ] +INFO:root:Average loss=0.1525383240137345 +INFO:root: +[Valid][11][UTT][ MSE = 0.3144 | LCC = 0.8220 | SRCC = 0.8200 ] +INFO:root:Average loss=0.14291705506352279 +INFO:root: +[Valid][12][UTT][ MSE = 0.2925 | LCC = 0.8177 | SRCC = 0.8149 ] +INFO:root:Average loss=0.1315504344036946 +INFO:root: +[Valid][13][UTT][ MSE = 0.2783 | LCC = 0.8225 | SRCC = 0.8213 ] +INFO:root:Average loss=0.1197394905373072 +INFO:root: +[Valid][14][UTT][ MSE = 0.2939 | LCC = 0.8204 | SRCC = 0.8181 ] +INFO:root:Average loss=0.11027407120817746 +INFO:root: +[Valid][15][UTT][ MSE = 0.2810 | LCC = 0.8155 | SRCC = 0.8146 ] +INFO:root:Average loss=0.110151090969642 +INFO:root: +[Valid][16][UTT][ MSE = 0.2887 | LCC = 0.8134 | SRCC = 0.8117 ] +INFO:root:Average loss=0.09951598139909598 +INFO:root: +[Valid][17][UTT][ MSE = 0.2889 | LCC = 0.8158 | SRCC = 0.8137 ] +INFO:root:Average loss=0.0955129778251434 +INFO:root: +[Valid][18][UTT][ MSE = 0.2892 | LCC = 0.8112 | SRCC = 0.8098 ] +INFO:root:Average loss=0.09016812697817118 +INFO:root: +[Valid][19][UTT][ MSE = 0.2790 | LCC = 0.8155 | SRCC = 0.8152 ] +INFO:root:Average loss=0.08550148008343501 +INFO:root: +[Valid][20][UTT][ MSE = 0.2869 | LCC = 0.8097 | SRCC = 0.8090 ] +INFO:root:Average loss=0.0785736173677903 +INFO:root: +[Valid][21][UTT][ MSE = 0.2789 | LCC = 0.8175 | SRCC = 0.8177 ] +INFO:root:Average loss=0.07420058347857915 +INFO:root: +[Valid][22][UTT][ MSE = 0.2773 | LCC = 0.8183 | SRCC = 0.8162 ] +INFO:root:Average loss=0.07864789053415641 +INFO:root: +[Valid][23][UTT][ MSE = 0.2866 | LCC = 0.8180 | SRCC = 0.8178 ] +INFO:root:Average loss=0.06753401883328572 +INFO:root: +[Valid][24][UTT][ MSE = 0.2834 | LCC = 0.8149 | SRCC = 0.8131 ] +INFO:root:Average loss=0.06836182428285098 +INFO:root: +[Valid][25][UTT][ MSE = 0.2829 | LCC = 0.8167 | SRCC = 0.8149 ] +INFO:root:Average loss=0.06368149172228116 +INFO:root: +[Valid][26][UTT][ MSE = 0.2901 | LCC = 0.8095 | SRCC = 0.8075 ] +INFO:root:Average loss=0.06620066761015317 +INFO:root: +[Valid][27][UTT][ MSE = 0.2807 | LCC = 0.8169 | SRCC = 0.8157 ] +INFO:root:Average loss=0.06555320642506465 +INFO:root: +[Valid][28][UTT][ MSE = 0.2907 | LCC = 0.8163 | SRCC = 0.8147 ] +INFO:root:Average loss=0.05875179859308096 +INFO:root: +[Valid][29][UTT][ MSE = 0.2871 | LCC = 0.8158 | SRCC = 0.8150 ] +INFO:root:Average loss=0.05459584400821955 +INFO:root: +[Valid][30][UTT][ MSE = 0.2850 | LCC = 0.8121 | SRCC = 0.8103 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2871 | LCC = 0.8298 | SRCC = 0.8348 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run3/config.gin b/results_bvcc/w2v2_large_asr_de/layer10/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run3/model.pt b/results_bvcc/w2v2_large_asr_de/layer10/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..30ca30664f2d094b9e8f1656efba83e02e95952f --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer10/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d3b60cc55a1ac6a2f40673240e77f91b51dfe038e5f80f70e42f109481656cfd +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run3/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer10/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f1c6158ddcf1fe3d3e8df153aa88fa4b3f7e1ba0 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer10/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16c2b380419536685595f9948f91c3a8d4c0039d0bfafc94da6d0f0f7c4bac5a +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run3/train.log b/results_bvcc/w2v2_large_asr_de/layer10/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..36cf586d8dafff47a6e0355fc480f942528e8a35 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/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=0.5549995635564511 +INFO:root: +[Valid][0][UTT][ MSE = 0.3766 | LCC = 0.7710 | SRCC = 0.7730 ] +INFO:root:Average loss=0.31929760999404466 +INFO:root: +[Valid][1][UTT][ MSE = 0.3439 | LCC = 0.8023 | SRCC = 0.8023 ] +INFO:root:Average loss=0.28425408746951664 +INFO:root: +[Valid][2][UTT][ MSE = 0.3259 | LCC = 0.8142 | SRCC = 0.8141 ] +INFO:root:Average loss=0.2545405312990531 +INFO:root: +[Valid][3][UTT][ MSE = 0.3058 | LCC = 0.8210 | SRCC = 0.8193 ] +INFO:root:Average loss=0.24000383598300126 +INFO:root: +[Valid][4][UTT][ MSE = 0.3391 | LCC = 0.8218 | SRCC = 0.8211 ] +INFO:root:Average loss=0.26011717586945265 +INFO:root: +[Valid][5][UTT][ MSE = 0.3011 | LCC = 0.8265 | SRCC = 0.8256 ] +INFO:root:Average loss=0.22101046851812264 +INFO:root: +[Valid][6][UTT][ MSE = 0.3189 | LCC = 0.8201 | SRCC = 0.8215 ] +INFO:root:Average loss=0.20010922151880386 +INFO:root: +[Valid][7][UTT][ MSE = 0.2928 | LCC = 0.8252 | SRCC = 0.8250 ] +INFO:root:Average loss=0.18898958292527077 +INFO:root: +[Valid][8][UTT][ MSE = 0.2820 | LCC = 0.8254 | SRCC = 0.8242 ] +INFO:root:Average loss=0.18066776075806373 +INFO:root: +[Valid][9][UTT][ MSE = 0.2828 | LCC = 0.8218 | SRCC = 0.8205 ] +INFO:root:Average loss=0.16588104659548172 +INFO:root: +[Valid][10][UTT][ MSE = 0.2868 | LCC = 0.8205 | SRCC = 0.8193 ] +INFO:root:Average loss=0.1639656910720544 +INFO:root: +[Valid][11][UTT][ MSE = 0.3081 | LCC = 0.8220 | SRCC = 0.8202 ] +INFO:root:Average loss=0.15706343261095193 +INFO:root: +[Valid][12][UTT][ MSE = 0.2907 | LCC = 0.8199 | SRCC = 0.8183 ] +INFO:root:Average loss=0.14317488555724805 +INFO:root: +[Valid][13][UTT][ MSE = 0.3172 | LCC = 0.8204 | SRCC = 0.8190 ] +INFO:root:Average loss=0.12452544281497979 +INFO:root: +[Valid][14][UTT][ MSE = 0.3719 | LCC = 0.8197 | SRCC = 0.8222 ] +INFO:root:Average loss=0.12073555741554652 +INFO:root: +[Valid][15][UTT][ MSE = 0.2961 | LCC = 0.8257 | SRCC = 0.8230 ] +INFO:root:Average loss=0.11687750655871171 +INFO:root: +[Valid][16][UTT][ MSE = 0.2909 | LCC = 0.8208 | SRCC = 0.8186 ] +INFO:root:Average loss=0.1129086074920801 +INFO:root: +[Valid][17][UTT][ MSE = 0.2823 | LCC = 0.8184 | SRCC = 0.8162 ] +INFO:root:Average loss=0.09779982679547408 +INFO:root: +[Valid][18][UTT][ MSE = 0.3049 | LCC = 0.8218 | SRCC = 0.8215 ] +INFO:root:Average loss=0.09802473030793361 +INFO:root: +[Valid][19][UTT][ MSE = 0.3072 | LCC = 0.8144 | SRCC = 0.8130 ] +INFO:root:Average loss=0.08803974956465073 +INFO:root: +[Valid][20][UTT][ MSE = 0.3093 | LCC = 0.8165 | SRCC = 0.8149 ] +INFO:root:Average loss=0.08730783867530334 +INFO:root: +[Valid][21][UTT][ MSE = 0.2856 | LCC = 0.8180 | SRCC = 0.8163 ] +INFO:root:Average loss=0.08133714746397275 +INFO:root: +[Valid][22][UTT][ MSE = 0.2973 | LCC = 0.8180 | SRCC = 0.8145 ] +INFO:root:Average loss=0.08341629535723956 +INFO:root: +[Valid][23][UTT][ MSE = 0.3026 | LCC = 0.8172 | SRCC = 0.8150 ] +INFO:root:Average loss=0.07852650080353786 +INFO:root: +[Valid][24][UTT][ MSE = 0.3053 | LCC = 0.8136 | SRCC = 0.8099 ] +INFO:root:Average loss=0.07717900269497664 +INFO:root: +[Valid][25][UTT][ MSE = 0.3177 | LCC = 0.8151 | SRCC = 0.8142 ] +INFO:root:Average loss=0.07134113895396392 +INFO:root: +[Valid][26][UTT][ MSE = 0.3143 | LCC = 0.8118 | SRCC = 0.8089 ] +INFO:root:Average loss=0.06848903955557407 +INFO:root: +[Valid][27][UTT][ MSE = 0.2804 | LCC = 0.8177 | SRCC = 0.8142 ] +INFO:root:Average loss=0.0643348641311511 +INFO:root: +[Valid][28][UTT][ MSE = 0.2852 | LCC = 0.8157 | SRCC = 0.8123 ] +INFO:root:Average loss=0.05930174247194559 +INFO:root: +[Valid][29][UTT][ MSE = 0.2943 | LCC = 0.8169 | SRCC = 0.8145 ] +INFO:root:Average loss=0.057826522642221205 +INFO:root: +[Valid][30][UTT][ MSE = 0.2913 | LCC = 0.8151 | SRCC = 0.8134 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3085 | LCC = 0.8258 | SRCC = 0.8296 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run4/config.gin b/results_bvcc/w2v2_large_asr_de/layer10/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run4/model.pt b/results_bvcc/w2v2_large_asr_de/layer10/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..df14f67d076b65427e47d64b4873943a5fec54af --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer10/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b0b7149ca54a1fdc9c7b337603dfbcf821d29ab11282513899ea3249ca1c703e +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run4/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer10/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..43192f0f90c4988c653734cb5c8347849ae749d9 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer10/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7a1239846095a7de8658e88309bccfc618261b69317a50051b2c4356ffff8678 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer10/run4/train.log b/results_bvcc/w2v2_large_asr_de/layer10/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..db9683b2e19a8e9f19ec1090d5fe964d153e289b --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/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=0.5795584783340112 +INFO:root: +[Valid][0][UTT][ MSE = 0.4041 | LCC = 0.7672 | SRCC = 0.7682 ] +INFO:root:Average loss=0.32929107260245544 +INFO:root: +[Valid][1][UTT][ MSE = 0.3295 | LCC = 0.8030 | SRCC = 0.8029 ] +INFO:root:Average loss=0.2721660335858663 +INFO:root: +[Valid][2][UTT][ MSE = 0.3103 | LCC = 0.8158 | SRCC = 0.8145 ] +INFO:root:Average loss=0.25864943613608676 +INFO:root: +[Valid][3][UTT][ MSE = 0.3136 | LCC = 0.8195 | SRCC = 0.8162 ] +INFO:root:Average loss=0.22812715077247375 +INFO:root: +[Valid][4][UTT][ MSE = 0.2726 | LCC = 0.8243 | SRCC = 0.8231 ] +INFO:root:Average loss=0.2071815469326117 +INFO:root: +[Valid][5][UTT][ MSE = 0.2745 | LCC = 0.8251 | SRCC = 0.8232 ] +INFO:root:Average loss=0.2135248934993377 +INFO:root: +[Valid][6][UTT][ MSE = 0.2778 | LCC = 0.8257 | SRCC = 0.8246 ] +INFO:root:Average loss=0.19224215833804545 +INFO:root: +[Valid][7][UTT][ MSE = 0.3064 | LCC = 0.8232 | SRCC = 0.8216 ] +INFO:root:Average loss=0.19286425564533624 +INFO:root: +[Valid][8][UTT][ MSE = 0.2959 | LCC = 0.8211 | SRCC = 0.8212 ] +INFO:root:Average loss=0.16533010911483031 +INFO:root: +[Valid][9][UTT][ MSE = 0.2830 | LCC = 0.8249 | SRCC = 0.8236 ] +INFO:root:Average loss=0.1481315813576564 +INFO:root: +[Valid][10][UTT][ MSE = 0.2769 | LCC = 0.8245 | SRCC = 0.8214 ] +INFO:root:Average loss=0.14711834175082353 +INFO:root: +[Valid][11][UTT][ MSE = 0.2687 | LCC = 0.8230 | SRCC = 0.8217 ] +INFO:root:Average loss=0.1297961053175804 +INFO:root: +[Valid][12][UTT][ MSE = 0.2778 | LCC = 0.8252 | SRCC = 0.8233 ] +INFO:root:Average loss=0.12998016092639703 +INFO:root: +[Valid][13][UTT][ MSE = 0.2724 | LCC = 0.8265 | SRCC = 0.8242 ] +INFO:root:Average loss=0.11446644891148959 +INFO:root: +[Valid][14][UTT][ MSE = 0.2942 | LCC = 0.8189 | SRCC = 0.8167 ] +INFO:root:Average loss=0.11216129410343292 +INFO:root: +[Valid][15][UTT][ MSE = 0.3465 | LCC = 0.8164 | SRCC = 0.8215 ] +INFO:root:Average loss=0.11183772895198602 +INFO:root: +[Valid][16][UTT][ MSE = 0.2769 | LCC = 0.8199 | SRCC = 0.8182 ] +INFO:root:Average loss=0.09561920070495361 +INFO:root: +[Valid][17][UTT][ MSE = 0.2876 | LCC = 0.8181 | SRCC = 0.8166 ] +INFO:root:Average loss=0.09098167421343999 +INFO:root: +[Valid][18][UTT][ MSE = 0.2806 | LCC = 0.8186 | SRCC = 0.8196 ] +INFO:root:Average loss=0.09964778073705159 +INFO:root: +[Valid][19][UTT][ MSE = 0.2805 | LCC = 0.8210 | SRCC = 0.8210 ] +INFO:root:Average loss=0.08135359953993405 +INFO:root: +[Valid][20][UTT][ MSE = 0.2782 | LCC = 0.8172 | SRCC = 0.8154 ] +INFO:root:Average loss=0.07693772591077365 +INFO:root: +[Valid][21][UTT][ MSE = 0.2802 | LCC = 0.8165 | SRCC = 0.8143 ] +INFO:root:Average loss=0.0826277617747203 +INFO:root: +[Valid][22][UTT][ MSE = 0.2807 | LCC = 0.8193 | SRCC = 0.8176 ] +INFO:root:Average loss=0.06972751250633827 +INFO:root: +[Valid][23][UTT][ MSE = 0.2699 | LCC = 0.8217 | SRCC = 0.8196 ] +INFO:root:Average loss=0.0687195260841877 +INFO:root: +[Valid][24][UTT][ MSE = 0.2842 | LCC = 0.8181 | SRCC = 0.8155 ] +INFO:root:Average loss=0.06477259195003754 +INFO:root: +[Valid][25][UTT][ MSE = 0.2801 | LCC = 0.8173 | SRCC = 0.8156 ] +INFO:root:Average loss=0.0652148185345607 +INFO:root: +[Valid][26][UTT][ MSE = 0.2811 | LCC = 0.8202 | SRCC = 0.8198 ] +INFO:root:Average loss=0.0568422337468618 +INFO:root: +[Valid][27][UTT][ MSE = 0.2796 | LCC = 0.8148 | SRCC = 0.8128 ] +INFO:root:Average loss=0.06077929629156223 +INFO:root: +[Valid][28][UTT][ MSE = 0.2822 | LCC = 0.8154 | SRCC = 0.8128 ] +INFO:root:Average loss=0.05839754736576325 +INFO:root: +[Valid][29][UTT][ MSE = 0.2865 | LCC = 0.8118 | SRCC = 0.8087 ] +INFO:root:Average loss=0.05380082703553713 +INFO:root: +[Valid][30][UTT][ MSE = 0.2758 | LCC = 0.8189 | SRCC = 0.8184 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2850 | LCC = 0.8288 | SRCC = 0.8334 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run0/config.gin b/results_bvcc/w2v2_large_asr_de/layer12/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run0/model.pt b/results_bvcc/w2v2_large_asr_de/layer12/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..38a9c8d19115b56e1ed6f0b50dd6f207f8742545 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43cae129e530e35fa2c7aa63045ce707a6c42dd178164ae356cd869ebf8f741e +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run0/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer12/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a0c3096932434b0053d2825a956ee459a158e144 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:931c9cba7ae2595ad54d637256c5df9d8946489a2c017595dc44a5dbec160a5a +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run0/train.log b/results_bvcc/w2v2_large_asr_de/layer12/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a3810542a5e1e89dabc6f63888f0e5b7031263f0 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/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=0.5553118811968045 +INFO:root: +[Valid][0][UTT][ MSE = 0.4214 | LCC = 0.7437 | SRCC = 0.7452 ] +INFO:root:Average loss=0.34144809058843517 +INFO:root: +[Valid][1][UTT][ MSE = 0.3692 | LCC = 0.7918 | SRCC = 0.7888 ] +INFO:root:Average loss=0.29118738839259517 +INFO:root: +[Valid][2][UTT][ MSE = 0.4127 | LCC = 0.7945 | SRCC = 0.7965 ] +INFO:root:Average loss=0.26623372905529463 +INFO:root: +[Valid][3][UTT][ MSE = 0.3228 | LCC = 0.8091 | SRCC = 0.8061 ] +INFO:root:Average loss=0.23998378465572992 +INFO:root: +[Valid][4][UTT][ MSE = 0.3106 | LCC = 0.8093 | SRCC = 0.8065 ] +INFO:root:Average loss=0.23176316955150703 +INFO:root: +[Valid][5][UTT][ MSE = 0.3105 | LCC = 0.8128 | SRCC = 0.8088 ] +INFO:root:Average loss=0.21945655097564062 +INFO:root: +[Valid][6][UTT][ MSE = 0.3110 | LCC = 0.8154 | SRCC = 0.8122 ] +INFO:root:Average loss=0.20291494826475778 +INFO:root: +[Valid][7][UTT][ MSE = 0.3067 | LCC = 0.8159 | SRCC = 0.8129 ] +INFO:root:Average loss=0.1950025691244847 +INFO:root: +[Valid][8][UTT][ MSE = 0.3022 | LCC = 0.8117 | SRCC = 0.8094 ] +INFO:root:Average loss=0.1759574555624754 +INFO:root: +[Valid][9][UTT][ MSE = 0.3275 | LCC = 0.8026 | SRCC = 0.8005 ] +INFO:root:Average loss=0.16009264362928194 +INFO:root: +[Valid][10][UTT][ MSE = 0.3278 | LCC = 0.8044 | SRCC = 0.8024 ] +INFO:root:Average loss=0.15070938519560373 +INFO:root: +[Valid][11][UTT][ MSE = 0.3107 | LCC = 0.8018 | SRCC = 0.8030 ] +INFO:root:Average loss=0.13960882844642186 +INFO:root: +[Valid][12][UTT][ MSE = 0.3049 | LCC = 0.8071 | SRCC = 0.8043 ] +INFO:root:Average loss=0.13869836878700134 +INFO:root: +[Valid][13][UTT][ MSE = 0.2989 | LCC = 0.8051 | SRCC = 0.8027 ] +INFO:root:Average loss=0.12283329694316937 +INFO:root: +[Valid][14][UTT][ MSE = 0.3514 | LCC = 0.7918 | SRCC = 0.7930 ] +INFO:root:Average loss=0.1193208697801217 +INFO:root: +[Valid][15][UTT][ MSE = 0.3601 | LCC = 0.7916 | SRCC = 0.7933 ] +INFO:root:Average loss=0.10767008712849556 +INFO:root: +[Valid][16][UTT][ MSE = 0.3069 | LCC = 0.7990 | SRCC = 0.7984 ] +INFO:root:Average loss=0.1001695219713908 +INFO:root: +[Valid][17][UTT][ MSE = 0.3285 | LCC = 0.7958 | SRCC = 0.7932 ] +INFO:root:Average loss=0.09714864218273224 +INFO:root: +[Valid][18][UTT][ MSE = 0.3309 | LCC = 0.7883 | SRCC = 0.7862 ] +INFO:root:Average loss=0.09451573208356515 +INFO:root: +[Valid][19][UTT][ MSE = 0.3041 | LCC = 0.7962 | SRCC = 0.7935 ] +INFO:root:Average loss=0.08451993610614385 +INFO:root: +[Valid][20][UTT][ MSE = 0.3135 | LCC = 0.7942 | SRCC = 0.7920 ] +INFO:root:Average loss=0.08115214477173793 +INFO:root: +[Valid][21][UTT][ MSE = 0.3024 | LCC = 0.7981 | SRCC = 0.7956 ] +INFO:root:Average loss=0.07493936203611203 +INFO:root: +[Valid][22][UTT][ MSE = 0.3273 | LCC = 0.7924 | SRCC = 0.7927 ] +INFO:root:Average loss=0.0700382810468093 +INFO:root: +[Valid][23][UTT][ MSE = 0.3155 | LCC = 0.7933 | SRCC = 0.7948 ] +INFO:root:Average loss=0.06825653267785525 +INFO:root: +[Valid][24][UTT][ MSE = 0.3318 | LCC = 0.7908 | SRCC = 0.7917 ] +INFO:root:Average loss=0.06595043412958965 +INFO:root: +[Valid][25][UTT][ MSE = 0.3063 | LCC = 0.7979 | SRCC = 0.7959 ] +INFO:root:Average loss=0.06523191952743591 +INFO:root: +[Valid][26][UTT][ MSE = 0.3102 | LCC = 0.7982 | SRCC = 0.7958 ] +INFO:root:Average loss=0.06083325585589195 +INFO:root: +[Valid][27][UTT][ MSE = 0.3124 | LCC = 0.7938 | SRCC = 0.7922 ] +INFO:root:Average loss=0.06116252578794956 +INFO:root: +[Valid][28][UTT][ MSE = 0.3058 | LCC = 0.7987 | SRCC = 0.7964 ] +INFO:root:Average loss=0.057369483730349786 +INFO:root: +[Valid][29][UTT][ MSE = 0.3076 | LCC = 0.7980 | SRCC = 0.7973 ] +INFO:root:Average loss=0.052267345026708566 +INFO:root: +[Valid][30][UTT][ MSE = 0.3240 | LCC = 0.7899 | SRCC = 0.7899 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2937 | LCC = 0.8255 | SRCC = 0.8257 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run1/config.gin b/results_bvcc/w2v2_large_asr_de/layer12/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run1/model.pt b/results_bvcc/w2v2_large_asr_de/layer12/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..eef7bcfd6e8a1c8312981f07738312a5717e2472 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b1c7ee1935bf5956ec552e7c728b0eaf76b7114317d6df0d5307cb76748cc92e +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run1/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer12/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..23200c2ae26bd63352a946d4e13254cded0707ff --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5b2c64a518a5ae734c400fc208d15c9f3043591559438f337a947b141691a333 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run1/train.log b/results_bvcc/w2v2_large_asr_de/layer12/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..100aacdb4f6dc904050327acfac8aa180ea8a107 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/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=0.5316474372759844 +INFO:root: +[Valid][0][UTT][ MSE = 0.3887 | LCC = 0.7596 | SRCC = 0.7623 ] +INFO:root:Average loss=0.3378170472689164 +INFO:root: +[Valid][1][UTT][ MSE = 0.3668 | LCC = 0.7887 | SRCC = 0.7920 ] +INFO:root:Average loss=0.2865099312785344 +INFO:root: +[Valid][2][UTT][ MSE = 0.3128 | LCC = 0.8056 | SRCC = 0.8047 ] +INFO:root:Average loss=0.2705295303693184 +INFO:root: +[Valid][3][UTT][ MSE = 0.3236 | LCC = 0.8097 | SRCC = 0.8073 ] +INFO:root:Average loss=0.23960204938283333 +INFO:root: +[Valid][4][UTT][ MSE = 0.3028 | LCC = 0.8146 | SRCC = 0.8136 ] +INFO:root:Average loss=0.23216098222212914 +INFO:root: +[Valid][5][UTT][ MSE = 0.2959 | LCC = 0.8152 | SRCC = 0.8147 ] +INFO:root:Average loss=0.21357216055576617 +INFO:root: +[Valid][6][UTT][ MSE = 0.3102 | LCC = 0.8127 | SRCC = 0.8096 ] +INFO:root:Average loss=0.20052701941667458 +INFO:root: +[Valid][7][UTT][ MSE = 0.2893 | LCC = 0.8156 | SRCC = 0.8133 ] +INFO:root:Average loss=0.19980415147848618 +INFO:root: +[Valid][8][UTT][ MSE = 0.3606 | LCC = 0.8172 | SRCC = 0.8154 ] +INFO:root:Average loss=0.18448574000444168 +INFO:root: +[Valid][9][UTT][ MSE = 0.2884 | LCC = 0.8153 | SRCC = 0.8147 ] +INFO:root:Average loss=0.15896772660123995 +INFO:root: +[Valid][10][UTT][ MSE = 0.3031 | LCC = 0.8150 | SRCC = 0.8132 ] +INFO:root:Average loss=0.15668142472322172 +INFO:root: +[Valid][11][UTT][ MSE = 0.2879 | LCC = 0.8115 | SRCC = 0.8112 ] +INFO:root:Average loss=0.1358016643386621 +INFO:root: +[Valid][12][UTT][ MSE = 0.2877 | LCC = 0.8081 | SRCC = 0.8064 ] +INFO:root:Average loss=0.12267208796663162 +INFO:root: +[Valid][13][UTT][ MSE = 0.3149 | LCC = 0.8104 | SRCC = 0.8080 ] +INFO:root:Average loss=0.13075364753603935 +INFO:root: +[Valid][14][UTT][ MSE = 0.2944 | LCC = 0.8064 | SRCC = 0.8045 ] +INFO:root:Average loss=0.11042927329738934 +INFO:root: +[Valid][15][UTT][ MSE = 0.2960 | LCC = 0.8089 | SRCC = 0.8063 ] +INFO:root:Average loss=0.11068413234674014 +INFO:root: +[Valid][16][UTT][ MSE = 0.3056 | LCC = 0.8082 | SRCC = 0.8060 ] +INFO:root:Average loss=0.09538930000211948 +INFO:root: +[Valid][17][UTT][ MSE = 0.3009 | LCC = 0.8070 | SRCC = 0.8052 ] +INFO:root:Average loss=0.1016500804764338 +INFO:root: +[Valid][18][UTT][ MSE = 0.3022 | LCC = 0.8078 | SRCC = 0.8083 ] +INFO:root:Average loss=0.09304089891986969 +INFO:root: +[Valid][19][UTT][ MSE = 0.2900 | LCC = 0.8077 | SRCC = 0.8071 ] +INFO:root:Average loss=0.09416867845142499 +INFO:root: +[Valid][20][UTT][ MSE = 0.2953 | LCC = 0.8041 | SRCC = 0.8039 ] +INFO:root:Average loss=0.07718449582656224 +INFO:root: +[Valid][21][UTT][ MSE = 0.3085 | LCC = 0.8061 | SRCC = 0.8034 ] +INFO:root:Average loss=0.0808523597721106 +INFO:root: +[Valid][22][UTT][ MSE = 0.2836 | LCC = 0.8118 | SRCC = 0.8112 ] +INFO:root:Average loss=0.06746636350185443 +INFO:root: +[Valid][23][UTT][ MSE = 0.2909 | LCC = 0.8061 | SRCC = 0.8057 ] +INFO:root:Average loss=0.07022327093932873 +INFO:root: +[Valid][24][UTT][ MSE = 0.2907 | LCC = 0.8082 | SRCC = 0.8072 ] +INFO:root:Average loss=0.06748059315559192 +INFO:root: +[Valid][25][UTT][ MSE = 0.2869 | LCC = 0.8095 | SRCC = 0.8074 ] +INFO:root:Average loss=0.062150852038310125 +INFO:root: +[Valid][26][UTT][ MSE = 0.3018 | LCC = 0.8029 | SRCC = 0.8034 ] +INFO:root:Average loss=0.05920848995447159 +INFO:root: +[Valid][27][UTT][ MSE = 0.2945 | LCC = 0.8060 | SRCC = 0.8056 ] +INFO:root:Average loss=0.05608322160939375 +INFO:root: +[Valid][28][UTT][ MSE = 0.2876 | LCC = 0.8100 | SRCC = 0.8083 ] +INFO:root:Average loss=0.05689092030605444 +INFO:root: +[Valid][29][UTT][ MSE = 0.2920 | LCC = 0.8080 | SRCC = 0.8079 ] +INFO:root:Average loss=0.054997538383572526 +INFO:root: +[Valid][30][UTT][ MSE = 0.2940 | LCC = 0.8081 | SRCC = 0.8058 ] +INFO:root: +[Test][31][UTT][ MSE = 0.4177 | LCC = 0.8160 | SRCC = 0.8233 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run2/config.gin b/results_bvcc/w2v2_large_asr_de/layer12/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run2/model.pt b/results_bvcc/w2v2_large_asr_de/layer12/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ccf4887d68775f753876bcc31f109cfd87f2f249 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:082c8c343c1978555446f2b749e7f5d94738f790c387817a235d90aa99c7e845 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run2/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer12/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c79f62f928cf84f5bc186901a936b220367fd0d7 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6fa6d178526782fff5e40a04e161f37cfebb2668f3b43e25ff184c1925c3a6d +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run2/train.log b/results_bvcc/w2v2_large_asr_de/layer12/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..eb05f550d99569d339faf8c9051317e1503b0889 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/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=0.5873763920404972 +INFO:root: +[Valid][0][UTT][ MSE = 0.4129 | LCC = 0.7307 | SRCC = 0.7313 ] +INFO:root:Average loss=0.3589370674811877 +INFO:root: +[Valid][1][UTT][ MSE = 0.3619 | LCC = 0.7807 | SRCC = 0.7798 ] +INFO:root:Average loss=0.30185407400131226 +INFO:root: +[Valid][2][UTT][ MSE = 0.3266 | LCC = 0.7944 | SRCC = 0.7929 ] +INFO:root:Average loss=0.27396879115929973 +INFO:root: +[Valid][3][UTT][ MSE = 0.3208 | LCC = 0.7998 | SRCC = 0.7999 ] +INFO:root:Average loss=0.24810503919919333 +INFO:root: +[Valid][4][UTT][ MSE = 0.3500 | LCC = 0.8021 | SRCC = 0.8021 ] +INFO:root:Average loss=0.23683231400373655 +INFO:root: +[Valid][5][UTT][ MSE = 0.3029 | LCC = 0.8069 | SRCC = 0.8052 ] +INFO:root:Average loss=0.20966419883263418 +INFO:root: +[Valid][6][UTT][ MSE = 0.3321 | LCC = 0.8121 | SRCC = 0.8107 ] +INFO:root:Average loss=0.2100936457132682 +INFO:root: +[Valid][7][UTT][ MSE = 0.3008 | LCC = 0.8069 | SRCC = 0.8078 ] +INFO:root:Average loss=0.2102544555106224 +INFO:root: +[Valid][8][UTT][ MSE = 0.3097 | LCC = 0.8094 | SRCC = 0.8088 ] +INFO:root:Average loss=0.18527913819520903 +INFO:root: +[Valid][9][UTT][ MSE = 0.2915 | LCC = 0.8144 | SRCC = 0.8142 ] +INFO:root:Average loss=0.15572170989635664 +INFO:root: +[Valid][10][UTT][ MSE = 0.2912 | LCC = 0.8073 | SRCC = 0.8083 ] +INFO:root:Average loss=0.14570886250107717 +INFO:root: +[Valid][11][UTT][ MSE = 0.2993 | LCC = 0.8097 | SRCC = 0.8102 ] +INFO:root:Average loss=0.14014486853892988 +INFO:root: +[Valid][12][UTT][ MSE = 0.3012 | LCC = 0.8042 | SRCC = 0.8050 ] +INFO:root:Average loss=0.13426593490518057 +INFO:root: +[Valid][13][UTT][ MSE = 0.2920 | LCC = 0.8069 | SRCC = 0.8073 ] +INFO:root:Average loss=0.1315553834518561 +INFO:root: +[Valid][14][UTT][ MSE = 0.2876 | LCC = 0.8104 | SRCC = 0.8115 ] +INFO:root:Average loss=0.11924783809062763 +INFO:root: +[Valid][15][UTT][ MSE = 0.3211 | LCC = 0.8045 | SRCC = 0.8072 ] +INFO:root:Average loss=0.12383828875728142 +INFO:root: +[Valid][16][UTT][ MSE = 0.2971 | LCC = 0.8053 | SRCC = 0.8055 ] +INFO:root:Average loss=0.10321400171289077 +INFO:root: +[Valid][17][UTT][ MSE = 0.2920 | LCC = 0.8053 | SRCC = 0.8064 ] +INFO:root:Average loss=0.09380856218437354 +INFO:root: +[Valid][18][UTT][ MSE = 0.2951 | LCC = 0.8027 | SRCC = 0.8025 ] +INFO:root:Average loss=0.0889916938657944 +INFO:root: +[Valid][19][UTT][ MSE = 0.2975 | LCC = 0.8024 | SRCC = 0.8027 ] +INFO:root:Average loss=0.0944796434770792 +INFO:root: +[Valid][20][UTT][ MSE = 0.2918 | LCC = 0.8046 | SRCC = 0.8062 ] +INFO:root:Average loss=0.08082464886590457 +INFO:root: +[Valid][21][UTT][ MSE = 0.3076 | LCC = 0.8040 | SRCC = 0.8066 ] +INFO:root:Average loss=0.07451083066944893 +INFO:root: +[Valid][22][UTT][ MSE = 0.2999 | LCC = 0.7998 | SRCC = 0.7990 ] +INFO:root:Average loss=0.07280300844174165 +INFO:root: +[Valid][23][UTT][ MSE = 0.3263 | LCC = 0.8058 | SRCC = 0.8061 ] +INFO:root:Average loss=0.06803994754759166 +INFO:root: +[Valid][24][UTT][ MSE = 0.3003 | LCC = 0.8037 | SRCC = 0.8031 ] +INFO:root:Average loss=0.08233287514975438 +INFO:root: +[Valid][25][UTT][ MSE = 0.3015 | LCC = 0.8052 | SRCC = 0.8051 ] +INFO:root:Average loss=0.06521178963474739 +INFO:root: +[Valid][26][UTT][ MSE = 0.2893 | LCC = 0.8091 | SRCC = 0.8094 ] +INFO:root:Average loss=0.05989298468025831 +INFO:root: +[Valid][27][UTT][ MSE = 0.2929 | LCC = 0.8055 | SRCC = 0.8056 ] +INFO:root:Average loss=0.05732595576689793 +INFO:root: +[Valid][28][UTT][ MSE = 0.2965 | LCC = 0.8010 | SRCC = 0.8014 ] +INFO:root:Average loss=0.05644938502556238 +INFO:root: +[Valid][29][UTT][ MSE = 0.2982 | LCC = 0.8018 | SRCC = 0.8018 ] +INFO:root:Average loss=0.05261006755515551 +INFO:root: +[Valid][30][UTT][ MSE = 0.2975 | LCC = 0.8040 | SRCC = 0.8038 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3140 | LCC = 0.8144 | SRCC = 0.8172 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run3/config.gin b/results_bvcc/w2v2_large_asr_de/layer12/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run3/model.pt b/results_bvcc/w2v2_large_asr_de/layer12/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0c9f56c7c0c8094878abfee4b18f973e7ff4c07f --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f9ea0a05069ec71d552276c66c0f38da3546d9c5c350bd5cae1b7ae42851dc79 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run3/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer12/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4c2c5e2e5eb0145edd79563d5df3422b4ed713bc --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1167aa291a6cbd1f84df4a592ff5a79774b1be5c9e33c92605466af92d7674f5 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run3/train.log b/results_bvcc/w2v2_large_asr_de/layer12/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a6cdd4e8950d9cae7fb7f9798109d245f3721743 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/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=0.6036551453364201 +INFO:root: +[Valid][0][UTT][ MSE = 0.4094 | LCC = 0.7282 | SRCC = 0.7324 ] +INFO:root:Average loss=0.35302545321293366 +INFO:root: +[Valid][1][UTT][ MSE = 0.3841 | LCC = 0.7714 | SRCC = 0.7783 ] +INFO:root:Average loss=0.2975278277045641 +INFO:root: +[Valid][2][UTT][ MSE = 0.3227 | LCC = 0.7944 | SRCC = 0.7943 ] +INFO:root:Average loss=0.27441758910814923 +INFO:root: +[Valid][3][UTT][ MSE = 0.3203 | LCC = 0.7980 | SRCC = 0.7951 ] +INFO:root:Average loss=0.2539201588966908 +INFO:root: +[Valid][4][UTT][ MSE = 0.3134 | LCC = 0.8023 | SRCC = 0.8013 ] +INFO:root:Average loss=0.23337777436543733 +INFO:root: +[Valid][5][UTT][ MSE = 0.3080 | LCC = 0.8060 | SRCC = 0.8034 ] +INFO:root:Average loss=0.219025181272091 +INFO:root: +[Valid][6][UTT][ MSE = 0.3017 | LCC = 0.8073 | SRCC = 0.8049 ] +INFO:root:Average loss=0.21745957969090876 +INFO:root: +[Valid][7][UTT][ MSE = 0.3001 | LCC = 0.8100 | SRCC = 0.8095 ] +INFO:root:Average loss=0.18775649464283234 +INFO:root: +[Valid][8][UTT][ MSE = 0.3066 | LCC = 0.8041 | SRCC = 0.8030 ] +INFO:root:Average loss=0.1753363885367528 +INFO:root: +[Valid][9][UTT][ MSE = 0.3294 | LCC = 0.7996 | SRCC = 0.8032 ] +INFO:root:Average loss=0.1586325527765812 +INFO:root: +[Valid][10][UTT][ MSE = 0.3076 | LCC = 0.8051 | SRCC = 0.8028 ] +INFO:root:Average loss=0.17025936289857596 +INFO:root: +[Valid][11][UTT][ MSE = 0.2971 | LCC = 0.8082 | SRCC = 0.8071 ] +INFO:root:Average loss=0.14191955442611986 +INFO:root: +[Valid][12][UTT][ MSE = 0.3107 | LCC = 0.7950 | SRCC = 0.7931 ] +INFO:root:Average loss=0.13679102617196548 +INFO:root: +[Valid][13][UTT][ MSE = 0.3340 | LCC = 0.8014 | SRCC = 0.7995 ] +INFO:root:Average loss=0.12490496488335805 +INFO:root: +[Valid][14][UTT][ MSE = 0.3109 | LCC = 0.7928 | SRCC = 0.7915 ] +INFO:root:Average loss=0.13114627632193077 +INFO:root: +[Valid][15][UTT][ MSE = 0.3073 | LCC = 0.8015 | SRCC = 0.8023 ] +INFO:root:Average loss=0.1117732676749046 +INFO:root: +[Valid][16][UTT][ MSE = 0.3085 | LCC = 0.8018 | SRCC = 0.7998 ] +INFO:root:Average loss=0.10329657554244384 +INFO:root: +[Valid][17][UTT][ MSE = 0.3073 | LCC = 0.7987 | SRCC = 0.7985 ] +INFO:root:Average loss=0.0946111122194009 +INFO:root: +[Valid][18][UTT][ MSE = 0.3130 | LCC = 0.7937 | SRCC = 0.7921 ] +INFO:root:Average loss=0.10111161679602586 +INFO:root: +[Valid][19][UTT][ MSE = 0.3085 | LCC = 0.8025 | SRCC = 0.8009 ] +INFO:root:Average loss=0.09323648081567043 +INFO:root: +[Valid][20][UTT][ MSE = 0.3121 | LCC = 0.7952 | SRCC = 0.7950 ] +INFO:root:Average loss=0.09110645185678433 +INFO:root: +[Valid][21][UTT][ MSE = 0.3169 | LCC = 0.7925 | SRCC = 0.7905 ] +INFO:root:Average loss=0.07778568241076592 +INFO:root: +[Valid][22][UTT][ MSE = 0.3054 | LCC = 0.7957 | SRCC = 0.7929 ] +INFO:root:Average loss=0.07206828323885417 +INFO:root: +[Valid][23][UTT][ MSE = 0.3022 | LCC = 0.7986 | SRCC = 0.7970 ] +INFO:root:Average loss=0.07406607542473537 +INFO:root: +[Valid][24][UTT][ MSE = 0.2920 | LCC = 0.8048 | SRCC = 0.8029 ] +INFO:root:Average loss=0.07104938759062535 +INFO:root: +[Valid][25][UTT][ MSE = 0.2991 | LCC = 0.8019 | SRCC = 0.8021 ] +INFO:root:Average loss=0.10039874557883312 +INFO:root: +[Valid][26][UTT][ MSE = 0.3054 | LCC = 0.8062 | SRCC = 0.8048 ] +INFO:root:Average loss=0.06833225297622192 +INFO:root: +[Valid][27][UTT][ MSE = 0.3026 | LCC = 0.7979 | SRCC = 0.7956 ] +INFO:root:Average loss=0.06302424181157197 +INFO:root: +[Valid][28][UTT][ MSE = 0.3089 | LCC = 0.8000 | SRCC = 0.7996 ] +INFO:root:Average loss=0.07023469955684283 +INFO:root: +[Valid][29][UTT][ MSE = 0.3443 | LCC = 0.7944 | SRCC = 0.7960 ] +INFO:root:Average loss=0.061766424956612095 +INFO:root: +[Valid][30][UTT][ MSE = 0.3007 | LCC = 0.8001 | SRCC = 0.7987 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2829 | LCC = 0.8217 | SRCC = 0.8213 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run4/config.gin b/results_bvcc/w2v2_large_asr_de/layer12/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run4/model.pt b/results_bvcc/w2v2_large_asr_de/layer12/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0ce7d49d984f61b5fedc88f72b328513cc7ef342 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b83e663d287f44fb8b2013dabe1ff9c066a1f77393726b013a670e93c473fee4 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run4/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer12/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ffcf3ba7a386e0fc1f10a97985d7c4db59f8d48c --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eab4825d5848da803e0f013e90d4ee2c2f8a206733b36cebfff1817c91447358 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer12/run4/train.log b/results_bvcc/w2v2_large_asr_de/layer12/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..4deaf6ae4333f9407a8905a700c186b14ef5338e --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer12/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=0.6016858124580139 +INFO:root: +[Valid][0][UTT][ MSE = 0.4452 | LCC = 0.7199 | SRCC = 0.7254 ] +INFO:root:Average loss=0.3559558336169292 +INFO:root: +[Valid][1][UTT][ MSE = 0.3609 | LCC = 0.7861 | SRCC = 0.7864 ] +INFO:root:Average loss=0.2979538564880689 +INFO:root: +[Valid][2][UTT][ MSE = 0.3833 | LCC = 0.7976 | SRCC = 0.8009 ] +INFO:root:Average loss=0.2777838802490479 +INFO:root: +[Valid][3][UTT][ MSE = 0.3202 | LCC = 0.8039 | SRCC = 0.8019 ] +INFO:root:Average loss=0.2550332120213753 +INFO:root: +[Valid][4][UTT][ MSE = 0.3400 | LCC = 0.8100 | SRCC = 0.8064 ] +INFO:root:Average loss=0.23332316180070242 +INFO:root: +[Valid][5][UTT][ MSE = 0.3036 | LCC = 0.8088 | SRCC = 0.8077 ] +INFO:root:Average loss=0.2254006952429429 +INFO:root: +[Valid][6][UTT][ MSE = 0.3321 | LCC = 0.8042 | SRCC = 0.8038 ] +INFO:root:Average loss=0.2207991830431498 +INFO:root: +[Valid][7][UTT][ MSE = 0.4041 | LCC = 0.8014 | SRCC = 0.8035 ] +INFO:root:Average loss=0.20482157782102242 +INFO:root: +[Valid][8][UTT][ MSE = 0.3425 | LCC = 0.8096 | SRCC = 0.8081 ] +INFO:root:Average loss=0.17818406195594713 +INFO:root: +[Valid][9][UTT][ MSE = 0.2885 | LCC = 0.8120 | SRCC = 0.8100 ] +INFO:root:Average loss=0.1679020690230223 +INFO:root: +[Valid][10][UTT][ MSE = 0.3081 | LCC = 0.8087 | SRCC = 0.8065 ] +INFO:root:Average loss=0.16237454164104584 +INFO:root: +[Valid][11][UTT][ MSE = 0.2998 | LCC = 0.8087 | SRCC = 0.8074 ] +INFO:root:Average loss=0.14933632266445038 +INFO:root: +[Valid][12][UTT][ MSE = 0.3244 | LCC = 0.7992 | SRCC = 0.7983 ] +INFO:root:Average loss=0.1364749354811815 +INFO:root: +[Valid][13][UTT][ MSE = 0.3230 | LCC = 0.8046 | SRCC = 0.8052 ] +INFO:root:Average loss=0.1247017849714328 +INFO:root: +[Valid][14][UTT][ MSE = 0.3245 | LCC = 0.7952 | SRCC = 0.7954 ] +INFO:root:Average loss=0.11419759967770332 +INFO:root: +[Valid][15][UTT][ MSE = 0.3073 | LCC = 0.7994 | SRCC = 0.7986 ] +INFO:root:Average loss=0.10872145178608406 +INFO:root: +[Valid][16][UTT][ MSE = 0.3233 | LCC = 0.7947 | SRCC = 0.7954 ] +INFO:root:Average loss=0.10862295277034625 +INFO:root: +[Valid][17][UTT][ MSE = 0.3040 | LCC = 0.7966 | SRCC = 0.7965 ] +INFO:root:Average loss=0.10223656539351512 +INFO:root: +[Valid][18][UTT][ MSE = 0.3132 | LCC = 0.7991 | SRCC = 0.7977 ] +INFO:root:Average loss=0.09706748916934697 +INFO:root: +[Valid][19][UTT][ MSE = 0.2936 | LCC = 0.8046 | SRCC = 0.8041 ] +INFO:root:Average loss=0.08328012014046693 +INFO:root: +[Valid][20][UTT][ MSE = 0.3081 | LCC = 0.7988 | SRCC = 0.7985 ] +INFO:root:Average loss=0.07476714683266786 +INFO:root: +[Valid][21][UTT][ MSE = 0.3194 | LCC = 0.7916 | SRCC = 0.7915 ] +INFO:root:Average loss=0.09059952710492489 +INFO:root: +[Valid][22][UTT][ MSE = 0.3125 | LCC = 0.7987 | SRCC = 0.7980 ] +INFO:root:Average loss=0.07147009307757402 +INFO:root: +[Valid][23][UTT][ MSE = 0.3207 | LCC = 0.7944 | SRCC = 0.7949 ] +INFO:root:Average loss=0.06317038089036942 +INFO:root: +[Valid][24][UTT][ MSE = 0.3092 | LCC = 0.7935 | SRCC = 0.7931 ] +INFO:root:Average loss=0.07445053378931987 +INFO:root: +[Valid][25][UTT][ MSE = 0.3301 | LCC = 0.7950 | SRCC = 0.7960 ] +INFO:root:Average loss=0.05935964613961868 +INFO:root: +[Valid][26][UTT][ MSE = 0.3138 | LCC = 0.7930 | SRCC = 0.7938 ] +INFO:root:Average loss=0.05943986052312912 +INFO:root: +[Valid][27][UTT][ MSE = 0.3078 | LCC = 0.7960 | SRCC = 0.7957 ] +INFO:root:Average loss=0.059264968555325113 +INFO:root: +[Valid][28][UTT][ MSE = 0.3145 | LCC = 0.7924 | SRCC = 0.7937 ] +INFO:root:Average loss=0.05402342139337307 +INFO:root: +[Valid][29][UTT][ MSE = 0.3145 | LCC = 0.7922 | SRCC = 0.7929 ] +INFO:root:Average loss=0.05530006316705392 +INFO:root: +[Valid][30][UTT][ MSE = 0.3164 | LCC = 0.7891 | SRCC = 0.7901 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2963 | LCC = 0.8205 | SRCC = 0.8217 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run0/config.gin b/results_bvcc/w2v2_large_asr_de/layer14/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run0/model.pt b/results_bvcc/w2v2_large_asr_de/layer14/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..41b0ea5e3195ecc66d5151b9685e33b1abe0777b --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e0615ef09dd1bbc842a0d93f417b4f4f60ca21b3067ab0c4dc734874cb3cfff9 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run0/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer14/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7448f505f9c52ce9d987f7113fa24a2bf3f669d7 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:959394542d643a0659429f407939b6a92d670ce020ec238e4854b4fbbaa844b9 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run0/train.log b/results_bvcc/w2v2_large_asr_de/layer14/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..7e7f21b60648b926b7cefbf3815b4bbc4a93d2bb --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/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=0.5887284592176095 +INFO:root: +[Valid][0][UTT][ MSE = 0.4302 | LCC = 0.7324 | SRCC = 0.7347 ] +INFO:root:Average loss=0.3514398651627394 +INFO:root: +[Valid][1][UTT][ MSE = 0.3728 | LCC = 0.7787 | SRCC = 0.7814 ] +INFO:root:Average loss=0.31456478035602814 +INFO:root: +[Valid][2][UTT][ MSE = 0.3715 | LCC = 0.7874 | SRCC = 0.7869 ] +INFO:root:Average loss=0.2842082025913092 +INFO:root: +[Valid][3][UTT][ MSE = 0.3296 | LCC = 0.7992 | SRCC = 0.7986 ] +INFO:root:Average loss=0.25770214161811733 +INFO:root: +[Valid][4][UTT][ MSE = 0.3099 | LCC = 0.8018 | SRCC = 0.7996 ] +INFO:root:Average loss=0.23873171993555167 +INFO:root: +[Valid][5][UTT][ MSE = 0.3359 | LCC = 0.8048 | SRCC = 0.8029 ] +INFO:root:Average loss=0.23849281745079237 +INFO:root: +[Valid][6][UTT][ MSE = 0.3381 | LCC = 0.8058 | SRCC = 0.8038 ] +INFO:root:Average loss=0.23059625847217363 +INFO:root: +[Valid][7][UTT][ MSE = 0.3023 | LCC = 0.8088 | SRCC = 0.8067 ] +INFO:root:Average loss=0.2224571970410836 +INFO:root: +[Valid][8][UTT][ MSE = 0.3209 | LCC = 0.8092 | SRCC = 0.8087 ] +INFO:root:Average loss=0.19488260083091566 +INFO:root: +[Valid][9][UTT][ MSE = 0.3152 | LCC = 0.8064 | SRCC = 0.8052 ] +INFO:root:Average loss=0.1813739873468876 +INFO:root: +[Valid][10][UTT][ MSE = 0.3040 | LCC = 0.8111 | SRCC = 0.8089 ] +INFO:root:Average loss=0.1688056845122423 +INFO:root: +[Valid][11][UTT][ MSE = 0.2952 | LCC = 0.8084 | SRCC = 0.8061 ] +INFO:root:Average loss=0.15739171942457175 +INFO:root: +[Valid][12][UTT][ MSE = 0.3164 | LCC = 0.8036 | SRCC = 0.8015 ] +INFO:root:Average loss=0.15765780372879443 +INFO:root: +[Valid][13][UTT][ MSE = 0.3291 | LCC = 0.8028 | SRCC = 0.8003 ] +INFO:root:Average loss=0.14375393016216081 +INFO:root: +[Valid][14][UTT][ MSE = 0.3175 | LCC = 0.8015 | SRCC = 0.7985 ] +INFO:root:Average loss=0.13201230525588378 +INFO:root: +[Valid][15][UTT][ MSE = 0.3391 | LCC = 0.7960 | SRCC = 0.7940 ] +INFO:root:Average loss=0.13043970920336553 +INFO:root: +[Valid][16][UTT][ MSE = 0.3048 | LCC = 0.8005 | SRCC = 0.7972 ] +INFO:root:Average loss=0.1260214479496846 +INFO:root: +[Valid][17][UTT][ MSE = 0.3364 | LCC = 0.8014 | SRCC = 0.7996 ] +INFO:root:Average loss=0.11570199168263337 +INFO:root: +[Valid][18][UTT][ MSE = 0.3528 | LCC = 0.7989 | SRCC = 0.7969 ] +INFO:root:Average loss=0.12657336269815764 +INFO:root: +[Valid][19][UTT][ MSE = 0.3063 | LCC = 0.7973 | SRCC = 0.7956 ] +INFO:root:Average loss=0.10670522399819814 +INFO:root: +[Valid][20][UTT][ MSE = 0.2997 | LCC = 0.8030 | SRCC = 0.8019 ] +INFO:root:Average loss=0.10423283059245501 +INFO:root: +[Valid][21][UTT][ MSE = 0.3328 | LCC = 0.7983 | SRCC = 0.7966 ] +INFO:root:Average loss=0.09481463395059109 +INFO:root: +[Valid][22][UTT][ MSE = 0.3190 | LCC = 0.7961 | SRCC = 0.7947 ] +INFO:root:Average loss=0.0959304637538317 +INFO:root: +[Valid][23][UTT][ MSE = 0.3060 | LCC = 0.7968 | SRCC = 0.7945 ] +INFO:root:Average loss=0.08814282333239531 +INFO:root: +[Valid][24][UTT][ MSE = 0.2997 | LCC = 0.8010 | SRCC = 0.7994 ] +INFO:root:Average loss=0.08032417330795373 +INFO:root: +[Valid][25][UTT][ MSE = 0.3156 | LCC = 0.7999 | SRCC = 0.7974 ] +INFO:root:Average loss=0.08455421866323704 +INFO:root: +[Valid][26][UTT][ MSE = 0.3072 | LCC = 0.8020 | SRCC = 0.8015 ] +INFO:root:Average loss=0.0817997791350652 +INFO:root: +[Valid][27][UTT][ MSE = 0.3276 | LCC = 0.8027 | SRCC = 0.8019 ] +INFO:root:Average loss=0.07710074198742707 +INFO:root: +[Valid][28][UTT][ MSE = 0.3158 | LCC = 0.7972 | SRCC = 0.7946 ] +INFO:root:Average loss=0.07348147837015298 +INFO:root: +[Valid][29][UTT][ MSE = 0.3071 | LCC = 0.8005 | SRCC = 0.7971 ] +INFO:root:Average loss=0.07126698375512393 +INFO:root: +[Valid][30][UTT][ MSE = 0.3159 | LCC = 0.7972 | SRCC = 0.7953 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2882 | LCC = 0.8142 | SRCC = 0.8140 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run1/config.gin b/results_bvcc/w2v2_large_asr_de/layer14/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run1/model.pt b/results_bvcc/w2v2_large_asr_de/layer14/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d20af6da7a8110882c8e28578bc8754f694c6510 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b13b86b11bcec8a4cd90b4195212097d0a3b33e45dd50ce99934e914bb583e10 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run1/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer14/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7e6e516f2a5fe2804adf0b5cb3e90fd4d2166b34 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:37247ff3fb90bdf75099ef166955b5a35e35c6b825fd5236fa33581b7165df75 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run1/train.log b/results_bvcc/w2v2_large_asr_de/layer14/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9930315c5e4074b2f96e1e5a9a0f6aa67a4c6756 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/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=0.5140194101975515 +INFO:root: +[Valid][0][UTT][ MSE = 0.3949 | LCC = 0.7400 | SRCC = 0.7404 ] +INFO:root:Average loss=0.3633739214677077 +INFO:root: +[Valid][1][UTT][ MSE = 0.3579 | LCC = 0.7722 | SRCC = 0.7746 ] +INFO:root:Average loss=0.31792190269781995 +INFO:root: +[Valid][2][UTT][ MSE = 0.3362 | LCC = 0.7833 | SRCC = 0.7836 ] +INFO:root:Average loss=0.29836817467824006 +INFO:root: +[Valid][3][UTT][ MSE = 0.3302 | LCC = 0.7915 | SRCC = 0.7876 ] +INFO:root:Average loss=0.2646245289689455 +INFO:root: +[Valid][4][UTT][ MSE = 0.3462 | LCC = 0.8011 | SRCC = 0.7975 ] +INFO:root:Average loss=0.24478325133140272 +INFO:root: +[Valid][5][UTT][ MSE = 0.3258 | LCC = 0.8020 | SRCC = 0.7952 ] +INFO:root:Average loss=0.2255932858739144 +INFO:root: +[Valid][6][UTT][ MSE = 0.3508 | LCC = 0.8137 | SRCC = 0.8108 ] +INFO:root:Average loss=0.22795768005725664 +INFO:root: +[Valid][7][UTT][ MSE = 0.3060 | LCC = 0.8091 | SRCC = 0.8061 ] +INFO:root:Average loss=0.20811015634964675 +INFO:root: +[Valid][8][UTT][ MSE = 0.3379 | LCC = 0.8093 | SRCC = 0.8075 ] +INFO:root:Average loss=0.20342672931460234 +INFO:root: +[Valid][9][UTT][ MSE = 0.3097 | LCC = 0.8079 | SRCC = 0.8056 ] +INFO:root:Average loss=0.18492736409489924 +INFO:root: +[Valid][10][UTT][ MSE = 0.2934 | LCC = 0.8054 | SRCC = 0.8046 ] +INFO:root:Average loss=0.18925067849266222 +INFO:root: +[Valid][11][UTT][ MSE = 0.3589 | LCC = 0.8008 | SRCC = 0.8029 ] +INFO:root:Average loss=0.17214773337428385 +INFO:root: +[Valid][12][UTT][ MSE = 0.2854 | LCC = 0.8100 | SRCC = 0.8101 ] +INFO:root:Average loss=0.15926246106242523 +INFO:root: +[Valid][13][UTT][ MSE = 0.2861 | LCC = 0.8106 | SRCC = 0.8111 ] +INFO:root:Average loss=0.14182521164035186 +INFO:root: +[Valid][14][UTT][ MSE = 0.3061 | LCC = 0.8031 | SRCC = 0.8017 ] +INFO:root:Average loss=0.14448515106088075 +INFO:root: +[Valid][15][UTT][ MSE = 0.3214 | LCC = 0.7994 | SRCC = 0.8000 ] +INFO:root:Average loss=0.14263941968480745 +INFO:root: +[Valid][16][UTT][ MSE = 0.2983 | LCC = 0.8030 | SRCC = 0.8025 ] +INFO:root:Average loss=0.12773212551688537 +INFO:root: +[Valid][17][UTT][ MSE = 0.3421 | LCC = 0.8023 | SRCC = 0.8048 ] +INFO:root:Average loss=0.11365202040626453 +INFO:root: +[Valid][18][UTT][ MSE = 0.3046 | LCC = 0.8015 | SRCC = 0.7993 ] +INFO:root:Average loss=0.10959776347646347 +INFO:root: +[Valid][19][UTT][ MSE = 0.2959 | LCC = 0.8043 | SRCC = 0.8037 ] +INFO:root:Average loss=0.10714346213409534 +INFO:root: +[Valid][20][UTT][ MSE = 0.3378 | LCC = 0.8007 | SRCC = 0.7994 ] +INFO:root:Average loss=0.10648287488864018 +INFO:root: +[Valid][21][UTT][ MSE = 0.3268 | LCC = 0.8066 | SRCC = 0.8060 ] +INFO:root:Average loss=0.09514786374683563 +INFO:root: +[Valid][22][UTT][ MSE = 0.2881 | LCC = 0.8075 | SRCC = 0.8070 ] +INFO:root:Average loss=0.08571794184927757 +INFO:root: +[Valid][23][UTT][ MSE = 0.3038 | LCC = 0.8055 | SRCC = 0.8056 ] +INFO:root:Average loss=0.08100777711623754 +INFO:root: +[Valid][24][UTT][ MSE = 0.3116 | LCC = 0.7963 | SRCC = 0.7947 ] +INFO:root:Average loss=0.07904304201022172 +INFO:root: +[Valid][25][UTT][ MSE = 0.2970 | LCC = 0.8060 | SRCC = 0.8055 ] +INFO:root:Average loss=0.07524719022405453 +INFO:root: +[Valid][26][UTT][ MSE = 0.2988 | LCC = 0.8082 | SRCC = 0.8077 ] +INFO:root:Average loss=0.07145083213272767 +INFO:root: +[Valid][27][UTT][ MSE = 0.3016 | LCC = 0.8063 | SRCC = 0.8057 ] +INFO:root:Average loss=0.08284081432681817 +INFO:root: +[Valid][28][UTT][ MSE = 0.3072 | LCC = 0.8045 | SRCC = 0.8070 ] +INFO:root:Average loss=0.07235424688611275 +INFO:root: +[Valid][29][UTT][ MSE = 0.2993 | LCC = 0.8029 | SRCC = 0.8014 ] +INFO:root:Average loss=0.06758380623964164 +INFO:root: +[Valid][30][UTT][ MSE = 0.3026 | LCC = 0.8013 | SRCC = 0.8025 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3990 | LCC = 0.8107 | SRCC = 0.8182 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run2/config.gin b/results_bvcc/w2v2_large_asr_de/layer14/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run2/model.pt b/results_bvcc/w2v2_large_asr_de/layer14/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e98fe4bc1732cdd8dc0ef9629580aadc00568ded --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:828c36ff13da9860bf5e328d6b7b4f048fbd060ce7dcb8e6a7ff18616820dac1 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run2/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer14/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d914fdfff54a0248b7c4b098e096deca785e60e4 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2cfb0ec61e84db3b3981973330bb446e252c2e7788b63ef493891842c1f2bf8f +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run2/train.log b/results_bvcc/w2v2_large_asr_de/layer14/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..153a8882929cfffd8ecaadfe75f7f6329b58f49a --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/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=0.6202406745690566 +INFO:root: +[Valid][0][UTT][ MSE = 0.4474 | LCC = 0.7222 | SRCC = 0.7259 ] +INFO:root:Average loss=0.36719463211603653 +INFO:root: +[Valid][1][UTT][ MSE = 0.3975 | LCC = 0.7695 | SRCC = 0.7683 ] +INFO:root:Average loss=0.308158349723388 +INFO:root: +[Valid][2][UTT][ MSE = 0.3350 | LCC = 0.7896 | SRCC = 0.7878 ] +INFO:root:Average loss=0.27013268035191756 +INFO:root: +[Valid][3][UTT][ MSE = 0.3436 | LCC = 0.7988 | SRCC = 0.7968 ] +INFO:root:Average loss=0.27208725076455337 +INFO:root: +[Valid][4][UTT][ MSE = 0.3113 | LCC = 0.8054 | SRCC = 0.8039 ] +INFO:root:Average loss=0.23416605228796983 +INFO:root: +[Valid][5][UTT][ MSE = 0.3140 | LCC = 0.8048 | SRCC = 0.8011 ] +INFO:root:Average loss=0.21330096706365928 +INFO:root: +[Valid][6][UTT][ MSE = 0.3291 | LCC = 0.8044 | SRCC = 0.8020 ] +INFO:root:Average loss=0.20167142468003127 +INFO:root: +[Valid][7][UTT][ MSE = 0.2984 | LCC = 0.8077 | SRCC = 0.8053 ] +INFO:root:Average loss=0.19324833307510766 +INFO:root: +[Valid][8][UTT][ MSE = 0.3527 | LCC = 0.8063 | SRCC = 0.8026 ] +INFO:root:Average loss=0.17645521863148764 +INFO:root: +[Valid][9][UTT][ MSE = 0.3010 | LCC = 0.8098 | SRCC = 0.8062 ] +INFO:root:Average loss=0.15953802556181565 +INFO:root: +[Valid][10][UTT][ MSE = 0.3024 | LCC = 0.8133 | SRCC = 0.8123 ] +INFO:root:Average loss=0.1559635031108673 +INFO:root: +[Valid][11][UTT][ MSE = 0.3527 | LCC = 0.8011 | SRCC = 0.8029 ] +INFO:root:Average loss=0.14841243175741953 +INFO:root: +[Valid][12][UTT][ MSE = 0.3039 | LCC = 0.8050 | SRCC = 0.8031 ] +INFO:root:Average loss=0.12958651609145677 +INFO:root: +[Valid][13][UTT][ MSE = 0.3507 | LCC = 0.7999 | SRCC = 0.7975 ] +INFO:root:Average loss=0.12360279567730732 +INFO:root: +[Valid][14][UTT][ MSE = 0.3174 | LCC = 0.7991 | SRCC = 0.7990 ] +INFO:root:Average loss=0.1261224631124582 +INFO:root: +[Valid][15][UTT][ MSE = 0.3094 | LCC = 0.8029 | SRCC = 0.8022 ] +INFO:root:Average loss=0.12491992488503456 +INFO:root: +[Valid][16][UTT][ MSE = 0.3465 | LCC = 0.8067 | SRCC = 0.8049 ] +INFO:root:Average loss=0.11184493872599724 +INFO:root: +[Valid][17][UTT][ MSE = 0.3098 | LCC = 0.8035 | SRCC = 0.8020 ] +INFO:root:Average loss=0.1049298225209499 +INFO:root: +[Valid][18][UTT][ MSE = 0.3060 | LCC = 0.7999 | SRCC = 0.7991 ] +INFO:root:Average loss=0.10159172146366192 +INFO:root: +[Valid][19][UTT][ MSE = 0.3158 | LCC = 0.8017 | SRCC = 0.7995 ] +INFO:root:Average loss=0.09756062595316997 +INFO:root: +[Valid][20][UTT][ MSE = 0.3123 | LCC = 0.8024 | SRCC = 0.8012 ] +INFO:root:Average loss=0.08755804378634845 +INFO:root: +[Valid][21][UTT][ MSE = 0.3187 | LCC = 0.7988 | SRCC = 0.7981 ] +INFO:root:Average loss=0.08433608128092228 +INFO:root: +[Valid][22][UTT][ MSE = 0.3530 | LCC = 0.7966 | SRCC = 0.7944 ] +INFO:root:Average loss=0.09079154189198445 +INFO:root: +[Valid][23][UTT][ MSE = 0.3051 | LCC = 0.7999 | SRCC = 0.7979 ] +INFO:root:Average loss=0.07799145951867104 +INFO:root: +[Valid][24][UTT][ MSE = 0.2969 | LCC = 0.8045 | SRCC = 0.8029 ] +INFO:root:Average loss=0.09391102896860012 +INFO:root: +[Valid][25][UTT][ MSE = 0.3001 | LCC = 0.8052 | SRCC = 0.8036 ] +INFO:root:Average loss=0.07810445254047711 +INFO:root: +[Valid][26][UTT][ MSE = 0.3054 | LCC = 0.8017 | SRCC = 0.7994 ] +INFO:root:Average loss=0.06820937804877758 +INFO:root: +[Valid][27][UTT][ MSE = 0.3208 | LCC = 0.7990 | SRCC = 0.7969 ] +INFO:root:Average loss=0.06784825409070039 +INFO:root: +[Valid][28][UTT][ MSE = 0.2981 | LCC = 0.8018 | SRCC = 0.7996 ] +INFO:root:Average loss=0.06058803291465992 +INFO:root: +[Valid][29][UTT][ MSE = 0.3053 | LCC = 0.8007 | SRCC = 0.7987 ] +INFO:root:Average loss=0.05971184200965441 +INFO:root: +[Valid][30][UTT][ MSE = 0.3185 | LCC = 0.7990 | SRCC = 0.7975 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2881 | LCC = 0.8138 | SRCC = 0.8139 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run3/config.gin b/results_bvcc/w2v2_large_asr_de/layer14/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run3/model.pt b/results_bvcc/w2v2_large_asr_de/layer14/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b862e5489e4b8dc3628370a63f6d2b2a21073594 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0f561dd1323ec3bd38154fd594c4b53c00fcd00753abe9a69a32d3dc9aafc9c0 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run3/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer14/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..5052c4637632880876138cee3e22189d65ce0e7f --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:03f94ac935ec0df89e3de6d8a06d3c33f5d280194f658dd2ff21f353f534f9e0 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run3/train.log b/results_bvcc/w2v2_large_asr_de/layer14/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c3bd544f165803b5262bbc787647463b3b909335 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/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=0.5256527536190473 +INFO:root: +[Valid][0][UTT][ MSE = 0.3991 | LCC = 0.7432 | SRCC = 0.7532 ] +INFO:root:Average loss=0.3445301361573048 +INFO:root: +[Valid][1][UTT][ MSE = 0.3979 | LCC = 0.7876 | SRCC = 0.7887 ] +INFO:root:Average loss=0.3026154230420406 +INFO:root: +[Valid][2][UTT][ MSE = 0.3398 | LCC = 0.7997 | SRCC = 0.7989 ] +INFO:root:Average loss=0.27890370385004926 +INFO:root: +[Valid][3][UTT][ MSE = 0.3189 | LCC = 0.8015 | SRCC = 0.8002 ] +INFO:root:Average loss=0.25217417149971694 +INFO:root: +[Valid][4][UTT][ MSE = 0.3059 | LCC = 0.8096 | SRCC = 0.8063 ] +INFO:root:Average loss=0.23731351414552102 +INFO:root: +[Valid][5][UTT][ MSE = 0.3078 | LCC = 0.8055 | SRCC = 0.8012 ] +INFO:root:Average loss=0.2251818657685549 +INFO:root: +[Valid][6][UTT][ MSE = 0.2928 | LCC = 0.8100 | SRCC = 0.8084 ] +INFO:root:Average loss=0.20625614489500338 +INFO:root: +[Valid][7][UTT][ MSE = 0.3250 | LCC = 0.8102 | SRCC = 0.8061 ] +INFO:root:Average loss=0.1951617086545015 +INFO:root: +[Valid][8][UTT][ MSE = 0.3078 | LCC = 0.8123 | SRCC = 0.8099 ] +INFO:root:Average loss=0.17844519296135658 +INFO:root: +[Valid][9][UTT][ MSE = 0.2987 | LCC = 0.8105 | SRCC = 0.8075 ] +INFO:root:Average loss=0.1793060788932519 +INFO:root: +[Valid][10][UTT][ MSE = 0.3030 | LCC = 0.8100 | SRCC = 0.8108 ] +INFO:root:Average loss=0.15866065436066726 +INFO:root: +[Valid][11][UTT][ MSE = 0.2899 | LCC = 0.8081 | SRCC = 0.8050 ] +INFO:root:Average loss=0.1516689461393234 +INFO:root: +[Valid][12][UTT][ MSE = 0.2912 | LCC = 0.8060 | SRCC = 0.8032 ] +INFO:root:Average loss=0.1400534755144364 +INFO:root: +[Valid][13][UTT][ MSE = 0.3063 | LCC = 0.8016 | SRCC = 0.7987 ] +INFO:root:Average loss=0.13752935100824404 +INFO:root: +[Valid][14][UTT][ MSE = 0.2994 | LCC = 0.8045 | SRCC = 0.8024 ] +INFO:root:Average loss=0.11812703215923065 +INFO:root: +[Valid][15][UTT][ MSE = 0.2979 | LCC = 0.8015 | SRCC = 0.7996 ] +INFO:root:Average loss=0.12269354611635208 +INFO:root: +[Valid][16][UTT][ MSE = 0.3415 | LCC = 0.8000 | SRCC = 0.7998 ] +INFO:root:Average loss=0.11137735824554394 +INFO:root: +[Valid][17][UTT][ MSE = 0.3235 | LCC = 0.7968 | SRCC = 0.7945 ] +INFO:root:Average loss=0.10413887547567868 +INFO:root: +[Valid][18][UTT][ MSE = 0.3019 | LCC = 0.8007 | SRCC = 0.7970 ] +INFO:root:Average loss=0.10439454038173725 +INFO:root: +[Valid][19][UTT][ MSE = 0.3016 | LCC = 0.8034 | SRCC = 0.8012 ] +INFO:root:Average loss=0.09432406943195905 +INFO:root: +[Valid][20][UTT][ MSE = 0.3024 | LCC = 0.7995 | SRCC = 0.7975 ] +INFO:root:Average loss=0.08673400632463969 +INFO:root: +[Valid][21][UTT][ MSE = 0.3307 | LCC = 0.7966 | SRCC = 0.7963 ] +INFO:root:Average loss=0.08730770236788651 +INFO:root: +[Valid][22][UTT][ MSE = 0.3067 | LCC = 0.7960 | SRCC = 0.7943 ] +INFO:root:Average loss=0.08994834846219955 +INFO:root: +[Valid][23][UTT][ MSE = 0.3372 | LCC = 0.8020 | SRCC = 0.8014 ] +INFO:root:Average loss=0.0880087402004462 +INFO:root: +[Valid][24][UTT][ MSE = 0.3075 | LCC = 0.7968 | SRCC = 0.7949 ] +INFO:root:Average loss=0.08531015800933044 +INFO:root: +[Valid][25][UTT][ MSE = 0.3206 | LCC = 0.8009 | SRCC = 0.7989 ] +INFO:root:Average loss=0.07920651152156867 +INFO:root: +[Valid][26][UTT][ MSE = 0.3036 | LCC = 0.7966 | SRCC = 0.7941 ] +INFO:root:Average loss=0.06835750805643889 +INFO:root: +[Valid][27][UTT][ MSE = 0.3028 | LCC = 0.7973 | SRCC = 0.7966 ] +INFO:root:Average loss=0.0846090303399624 +INFO:root: +[Valid][28][UTT][ MSE = 0.3408 | LCC = 0.7959 | SRCC = 0.7967 ] +INFO:root:Average loss=0.08307809053132167 +INFO:root: +[Valid][29][UTT][ MSE = 0.3071 | LCC = 0.7934 | SRCC = 0.7918 ] +INFO:root:Average loss=0.06082330553386456 +INFO:root: +[Valid][30][UTT][ MSE = 0.3094 | LCC = 0.7930 | SRCC = 0.7906 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2793 | LCC = 0.8253 | SRCC = 0.8249 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run4/config.gin b/results_bvcc/w2v2_large_asr_de/layer14/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run4/model.pt b/results_bvcc/w2v2_large_asr_de/layer14/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..cf0da9d8e3f5b3d571d3d19d87b25a3a6a1ead49 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b61f781a4c91bddb7f4683ab541686b9893ea6c027f1fcef567935f9a1e3170a +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run4/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer14/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..afab3257d8df546d94f60144906944116f738ba6 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f1dcdbe8b0054d161849d973b9d18d3d11be78f9a7a644aa2fc7e5524e8e2dc2 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer14/run4/train.log b/results_bvcc/w2v2_large_asr_de/layer14/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5569678117edbf6d91e1e94c5e2c816bb11cd6d2 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer14/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=0.5603201255584375 +INFO:root: +[Valid][0][UTT][ MSE = 0.3974 | LCC = 0.7436 | SRCC = 0.7478 ] +INFO:root:Average loss=0.37081605463455886 +INFO:root: +[Valid][1][UTT][ MSE = 0.3740 | LCC = 0.7748 | SRCC = 0.7749 ] +INFO:root:Average loss=0.3150634490526639 +INFO:root: +[Valid][2][UTT][ MSE = 0.3605 | LCC = 0.7882 | SRCC = 0.7861 ] +INFO:root:Average loss=0.2919948876668245 +INFO:root: +[Valid][3][UTT][ MSE = 0.3434 | LCC = 0.7928 | SRCC = 0.7877 ] +INFO:root:Average loss=0.2588868477405646 +INFO:root: +[Valid][4][UTT][ MSE = 0.3574 | LCC = 0.7994 | SRCC = 0.7965 ] +INFO:root:Average loss=0.2470546786983808 +INFO:root: +[Valid][5][UTT][ MSE = 0.3274 | LCC = 0.8048 | SRCC = 0.8008 ] +INFO:root:Average loss=0.23760618154819196 +INFO:root: +[Valid][6][UTT][ MSE = 0.3190 | LCC = 0.8066 | SRCC = 0.8011 ] +INFO:root:Average loss=0.21414438157509535 +INFO:root: +[Valid][7][UTT][ MSE = 0.3171 | LCC = 0.8074 | SRCC = 0.8033 ] +INFO:root:Average loss=0.19779181403991503 +INFO:root: +[Valid][8][UTT][ MSE = 0.3179 | LCC = 0.8052 | SRCC = 0.8021 ] +INFO:root:Average loss=0.19717857470879188 +INFO:root: +[Valid][9][UTT][ MSE = 0.3247 | LCC = 0.8055 | SRCC = 0.8023 ] +INFO:root:Average loss=0.18515536275047523 +INFO:root: +[Valid][10][UTT][ MSE = 0.3185 | LCC = 0.8049 | SRCC = 0.8021 ] +INFO:root:Average loss=0.16536256422599158 +INFO:root: +[Valid][11][UTT][ MSE = 0.3027 | LCC = 0.8046 | SRCC = 0.8019 ] +INFO:root:Average loss=0.1620084124688919 +INFO:root: +[Valid][12][UTT][ MSE = 0.3122 | LCC = 0.8040 | SRCC = 0.8018 ] +INFO:root:Average loss=0.14540286753804255 +INFO:root: +[Valid][13][UTT][ MSE = 0.3006 | LCC = 0.8004 | SRCC = 0.7979 ] +INFO:root:Average loss=0.13656388223171234 +INFO:root: +[Valid][14][UTT][ MSE = 0.3274 | LCC = 0.7995 | SRCC = 0.7973 ] +INFO:root:Average loss=0.13159466133667871 +INFO:root: +[Valid][15][UTT][ MSE = 0.3079 | LCC = 0.8019 | SRCC = 0.8000 ] +INFO:root:Average loss=0.15958268644335943 +INFO:root: +[Valid][16][UTT][ MSE = 0.3092 | LCC = 0.8040 | SRCC = 0.8027 ] +INFO:root:Average loss=0.12109528544048469 +INFO:root: +[Valid][17][UTT][ MSE = 0.3005 | LCC = 0.8029 | SRCC = 0.8003 ] +INFO:root:Average loss=0.11831444033827537 +INFO:root: +[Valid][18][UTT][ MSE = 0.3154 | LCC = 0.7953 | SRCC = 0.7915 ] +INFO:root:Average loss=0.11160208609623787 +INFO:root: +[Valid][19][UTT][ MSE = 0.3416 | LCC = 0.8013 | SRCC = 0.8004 ] +INFO:root:Average loss=0.1112319212884475 +INFO:root: +[Valid][20][UTT][ MSE = 0.3105 | LCC = 0.7995 | SRCC = 0.7970 ] +INFO:root:Average loss=0.0935268400189204 +INFO:root: +[Valid][21][UTT][ MSE = 0.3282 | LCC = 0.7978 | SRCC = 0.7957 ] +INFO:root:Average loss=0.10309107687610847 +INFO:root: +[Valid][22][UTT][ MSE = 0.3077 | LCC = 0.8056 | SRCC = 0.8032 ] +INFO:root:Average loss=0.1006944416425167 +INFO:root: +[Valid][23][UTT][ MSE = 0.3064 | LCC = 0.8046 | SRCC = 0.8033 ] +INFO:root:Average loss=0.0864234902919867 +INFO:root: +[Valid][24][UTT][ MSE = 0.3173 | LCC = 0.8001 | SRCC = 0.7993 ] +INFO:root:Average loss=0.08412263609277897 +INFO:root: +[Valid][25][UTT][ MSE = 0.3158 | LCC = 0.7985 | SRCC = 0.7964 ] +INFO:root:Average loss=0.08929182426669659 +INFO:root: +[Valid][26][UTT][ MSE = 0.3651 | LCC = 0.8023 | SRCC = 0.8011 ] +INFO:root:Average loss=0.08307532808528496 +INFO:root: +[Valid][27][UTT][ MSE = 0.3188 | LCC = 0.8045 | SRCC = 0.8023 ] +INFO:root:Average loss=0.0720879676250311 +INFO:root: +[Valid][28][UTT][ MSE = 0.3019 | LCC = 0.8017 | SRCC = 0.7997 ] +INFO:root:Average loss=0.06961400625415337 +INFO:root: +[Valid][29][UTT][ MSE = 0.3215 | LCC = 0.7970 | SRCC = 0.7958 ] +INFO:root:Average loss=0.06954468557467827 +INFO:root: +[Valid][30][UTT][ MSE = 0.3638 | LCC = 0.7969 | SRCC = 0.7962 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2996 | LCC = 0.8139 | SRCC = 0.8143 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run0/config.gin b/results_bvcc/w2v2_large_asr_de/layer16/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run0/model.pt b/results_bvcc/w2v2_large_asr_de/layer16/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bc854540f55db72f74ddf5b0d3c0ed654305776e --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:86b4a30fe28cd140c7922088e3521f3616692a38def1696c0b17cfb5b61d14c1 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run0/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer16/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d21c3566921fc950f76725ebe16328350abf1c68 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2c58c68a48b0cb8c36b823d8dc9c294d83a9657425833b9a503bda923f452a07 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run0/train.log b/results_bvcc/w2v2_large_asr_de/layer16/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0101a4b378a5b71093f7012a5653fa9f30052573 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/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=0.6247731068959603 +INFO:root: +[Valid][0][UTT][ MSE = 0.4368 | LCC = 0.7329 | SRCC = 0.7366 ] +INFO:root:Average loss=0.37680388050965774 +INFO:root: +[Valid][1][UTT][ MSE = 0.4018 | LCC = 0.7745 | SRCC = 0.7731 ] +INFO:root:Average loss=0.3415707521713697 +INFO:root: +[Valid][2][UTT][ MSE = 0.3477 | LCC = 0.7888 | SRCC = 0.7878 ] +INFO:root:Average loss=0.3121731338592676 +INFO:root: +[Valid][3][UTT][ MSE = 0.3531 | LCC = 0.7897 | SRCC = 0.7883 ] +INFO:root:Average loss=0.29186112586504376 +INFO:root: +[Valid][4][UTT][ MSE = 0.3425 | LCC = 0.7949 | SRCC = 0.7949 ] +INFO:root:Average loss=0.2686758190393448 +INFO:root: +[Valid][5][UTT][ MSE = 0.3108 | LCC = 0.8042 | SRCC = 0.8011 ] +INFO:root:Average loss=0.24518505464761686 +INFO:root: +[Valid][6][UTT][ MSE = 0.3486 | LCC = 0.8029 | SRCC = 0.8019 ] +INFO:root:Average loss=0.2340882109143795 +INFO:root: +[Valid][7][UTT][ MSE = 0.3289 | LCC = 0.8031 | SRCC = 0.8006 ] +INFO:root:Average loss=0.2194988414263114 +INFO:root: +[Valid][8][UTT][ MSE = 0.3235 | LCC = 0.8031 | SRCC = 0.7995 ] +INFO:root:Average loss=0.20697079264582732 +INFO:root: +[Valid][9][UTT][ MSE = 0.3031 | LCC = 0.8028 | SRCC = 0.7998 ] +INFO:root:Average loss=0.2101507496375304 +INFO:root: +[Valid][10][UTT][ MSE = 0.3096 | LCC = 0.8035 | SRCC = 0.8003 ] +INFO:root:Average loss=0.18214382670628718 +INFO:root: +[Valid][11][UTT][ MSE = 0.3084 | LCC = 0.8008 | SRCC = 0.7977 ] +INFO:root:Average loss=0.18028741864821848 +INFO:root: +[Valid][12][UTT][ MSE = 0.3095 | LCC = 0.7994 | SRCC = 0.7946 ] +INFO:root:Average loss=0.16507556518683067 +INFO:root: +[Valid][13][UTT][ MSE = 0.3196 | LCC = 0.7984 | SRCC = 0.7944 ] +INFO:root:Average loss=0.15414444481333098 +INFO:root: +[Valid][14][UTT][ MSE = 0.3070 | LCC = 0.8013 | SRCC = 0.7992 ] +INFO:root:Average loss=0.14852567303639191 +INFO:root: +[Valid][15][UTT][ MSE = 0.3625 | LCC = 0.7920 | SRCC = 0.7914 ] +INFO:root:Average loss=0.14962083989611039 +INFO:root: +[Valid][16][UTT][ MSE = 0.3336 | LCC = 0.7974 | SRCC = 0.7964 ] +INFO:root:Average loss=0.1412992975077568 +INFO:root: +[Valid][17][UTT][ MSE = 0.3023 | LCC = 0.8001 | SRCC = 0.7979 ] +INFO:root:Average loss=0.13432402297472343 +INFO:root: +[Valid][18][UTT][ MSE = 0.3241 | LCC = 0.7970 | SRCC = 0.7986 ] +INFO:root:Average loss=0.11806758961234337 +INFO:root: +[Valid][19][UTT][ MSE = 0.2951 | LCC = 0.8020 | SRCC = 0.8002 ] +INFO:root:Average loss=0.11493394218194179 +INFO:root: +[Valid][20][UTT][ MSE = 0.3063 | LCC = 0.7975 | SRCC = 0.7941 ] +INFO:root:Average loss=0.10748664480753434 +INFO:root: +[Valid][21][UTT][ MSE = 0.3025 | LCC = 0.8005 | SRCC = 0.7965 ] +INFO:root:Average loss=0.10724590432185394 +INFO:root: +[Valid][22][UTT][ MSE = 0.3037 | LCC = 0.7965 | SRCC = 0.7937 ] +INFO:root:Average loss=0.10953135086366764 +INFO:root: +[Valid][23][UTT][ MSE = 0.3069 | LCC = 0.8004 | SRCC = 0.7976 ] +INFO:root:Average loss=0.09712773547149621 +INFO:root: +[Valid][24][UTT][ MSE = 0.3405 | LCC = 0.7900 | SRCC = 0.7880 ] +INFO:root:Average loss=0.09828333255763237 +INFO:root: +[Valid][25][UTT][ MSE = 0.3129 | LCC = 0.7945 | SRCC = 0.7925 ] +INFO:root:Average loss=0.09694748672728355 +INFO:root: +[Valid][26][UTT][ MSE = 0.3352 | LCC = 0.7879 | SRCC = 0.7871 ] +INFO:root:Average loss=0.09703806735193118 +INFO:root: +[Valid][27][UTT][ MSE = 0.3315 | LCC = 0.7950 | SRCC = 0.7959 ] +INFO:root:Average loss=0.08925082214558736 +INFO:root: +[Valid][28][UTT][ MSE = 0.3322 | LCC = 0.7978 | SRCC = 0.7987 ] +INFO:root:Average loss=0.08032925398303913 +INFO:root: +[Valid][29][UTT][ MSE = 0.3401 | LCC = 0.7911 | SRCC = 0.7907 ] +INFO:root:Average loss=0.0882290834799791 +INFO:root: +[Valid][30][UTT][ MSE = 0.3064 | LCC = 0.7951 | SRCC = 0.7935 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3263 | LCC = 0.7997 | SRCC = 0.8017 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run1/config.gin b/results_bvcc/w2v2_large_asr_de/layer16/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run1/model.pt b/results_bvcc/w2v2_large_asr_de/layer16/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0b431e99def65e67c9cade6bd884c11c227a5558 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e6856e56940dc6e94a267a7aa96a104b2e22431b60d57a0b0e5044e9a7affb8 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run1/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer16/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..96e5f936332ec69ac414b5036f437f3f5c221210 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:61c50f43f86764b1821e619cc9dcca6e61909f04fc7f33095f7bd472867db61c +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run1/train.log b/results_bvcc/w2v2_large_asr_de/layer16/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1327323d18acd341c82407b6eb51554d33fba706 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/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=0.6003520297698486 +INFO:root: +[Valid][0][UTT][ MSE = 0.4534 | LCC = 0.7167 | SRCC = 0.7188 ] +INFO:root:Average loss=0.3766224701435138 +INFO:root: +[Valid][1][UTT][ MSE = 0.3925 | LCC = 0.7713 | SRCC = 0.7703 ] +INFO:root:Average loss=0.32407492532944066 +INFO:root: +[Valid][2][UTT][ MSE = 0.3909 | LCC = 0.7737 | SRCC = 0.7718 ] +INFO:root:Average loss=0.2987801616008465 +INFO:root: +[Valid][3][UTT][ MSE = 0.3998 | LCC = 0.7843 | SRCC = 0.7811 ] +INFO:root:Average loss=0.2709105404523703 +INFO:root: +[Valid][4][UTT][ MSE = 0.3284 | LCC = 0.8010 | SRCC = 0.7994 ] +INFO:root:Average loss=0.2701709176867436 +INFO:root: +[Valid][5][UTT][ MSE = 0.3713 | LCC = 0.7957 | SRCC = 0.7953 ] +INFO:root:Average loss=0.2647583862909904 +INFO:root: +[Valid][6][UTT][ MSE = 0.3707 | LCC = 0.7987 | SRCC = 0.7965 ] +INFO:root:Average loss=0.23300214971487337 +INFO:root: +[Valid][7][UTT][ MSE = 0.3314 | LCC = 0.8023 | SRCC = 0.7996 ] +INFO:root:Average loss=0.2178077376805819 +INFO:root: +[Valid][8][UTT][ MSE = 0.3336 | LCC = 0.8040 | SRCC = 0.8023 ] +INFO:root:Average loss=0.2026489059894513 +INFO:root: +[Valid][9][UTT][ MSE = 0.3472 | LCC = 0.8034 | SRCC = 0.8026 ] +INFO:root:Average loss=0.18925152165003312 +INFO:root: +[Valid][10][UTT][ MSE = 0.3099 | LCC = 0.8050 | SRCC = 0.8029 ] +INFO:root:Average loss=0.188047356903553 +INFO:root: +[Valid][11][UTT][ MSE = 0.3083 | LCC = 0.8000 | SRCC = 0.7973 ] +INFO:root:Average loss=0.1848634674380987 +INFO:root: +[Valid][12][UTT][ MSE = 0.3072 | LCC = 0.8062 | SRCC = 0.8041 ] +INFO:root:Average loss=0.18542858480643004 +INFO:root: +[Valid][13][UTT][ MSE = 0.3002 | LCC = 0.8074 | SRCC = 0.8050 ] +INFO:root:Average loss=0.1700604887535939 +INFO:root: +[Valid][14][UTT][ MSE = 0.2968 | LCC = 0.8052 | SRCC = 0.8039 ] +INFO:root:Average loss=0.14934848487759247 +INFO:root: +[Valid][15][UTT][ MSE = 0.3379 | LCC = 0.7952 | SRCC = 0.7934 ] +INFO:root:Average loss=0.15875224730907342 +INFO:root: +[Valid][16][UTT][ MSE = 0.3051 | LCC = 0.8028 | SRCC = 0.7997 ] +INFO:root:Average loss=0.1458238119689318 +INFO:root: +[Valid][17][UTT][ MSE = 0.3236 | LCC = 0.8008 | SRCC = 0.7995 ] +INFO:root:Average loss=0.13690483374282336 +INFO:root: +[Valid][18][UTT][ MSE = 0.3007 | LCC = 0.8051 | SRCC = 0.8032 ] +INFO:root:Average loss=0.1292797465546009 +INFO:root: +[Valid][19][UTT][ MSE = 0.3191 | LCC = 0.7952 | SRCC = 0.7917 ] +INFO:root:Average loss=0.11278171321520439 +INFO:root: +[Valid][20][UTT][ MSE = 0.3248 | LCC = 0.7970 | SRCC = 0.7946 ] +INFO:root:Average loss=0.11563276470853733 +INFO:root: +[Valid][21][UTT][ MSE = 0.3664 | LCC = 0.7985 | SRCC = 0.7969 ] +INFO:root:Average loss=0.10719747124956204 +INFO:root: +[Valid][22][UTT][ MSE = 0.3550 | LCC = 0.7942 | SRCC = 0.7918 ] +INFO:root:Average loss=0.10762533488181922 +INFO:root: +[Valid][23][UTT][ MSE = 0.3184 | LCC = 0.7970 | SRCC = 0.7945 ] +INFO:root:Average loss=0.09470044253155208 +INFO:root: +[Valid][24][UTT][ MSE = 0.3044 | LCC = 0.7997 | SRCC = 0.7976 ] +INFO:root:Average loss=0.09499901514022778 +INFO:root: +[Valid][25][UTT][ MSE = 0.3190 | LCC = 0.7962 | SRCC = 0.7931 ] +INFO:root:Average loss=0.08971433169566669 +INFO:root: +[Valid][26][UTT][ MSE = 0.3243 | LCC = 0.7897 | SRCC = 0.7865 ] +INFO:root:Average loss=0.08632044159831145 +INFO:root: +[Valid][27][UTT][ MSE = 0.3495 | LCC = 0.7967 | SRCC = 0.7959 ] +INFO:root:Average loss=0.08922372404963543 +INFO:root: +[Valid][28][UTT][ MSE = 0.3075 | LCC = 0.7994 | SRCC = 0.7980 ] +INFO:root:Average loss=0.07710069623322059 +INFO:root: +[Valid][29][UTT][ MSE = 0.3088 | LCC = 0.7971 | SRCC = 0.7941 ] +INFO:root:Average loss=0.08294099316191979 +INFO:root: +[Valid][30][UTT][ MSE = 0.3304 | LCC = 0.7971 | SRCC = 0.7957 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2900 | LCC = 0.8116 | SRCC = 0.8117 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run2/config.gin b/results_bvcc/w2v2_large_asr_de/layer16/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run2/model.pt b/results_bvcc/w2v2_large_asr_de/layer16/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..145ef8291af8292c19322bfebb87e949558522ee --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:798b8eb27ebbda9e605400ddcd5028fa1cd2a6dfb07a0ed08697f412fbd53812 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run2/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer16/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..8a59498997f990a44fd022125b6bb5c7057c7e2f --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e7f7baa6e804dd815372c1c3fc97d41ab6cd960c0d1f91227037c4b0f64ff7f1 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run2/train.log b/results_bvcc/w2v2_large_asr_de/layer16/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..636de5dbe260744d3d10bf1a068132d63475017c --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/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=0.6250840498086734 +INFO:root: +[Valid][0][UTT][ MSE = 0.5329 | LCC = 0.7295 | SRCC = 0.7347 ] +INFO:root:Average loss=0.368826150512084 +INFO:root: +[Valid][1][UTT][ MSE = 0.4520 | LCC = 0.7722 | SRCC = 0.7740 ] +INFO:root:Average loss=0.330862759015499 +INFO:root: +[Valid][2][UTT][ MSE = 0.3568 | LCC = 0.7838 | SRCC = 0.7822 ] +INFO:root:Average loss=0.3246920336133394 +INFO:root: +[Valid][3][UTT][ MSE = 0.4815 | LCC = 0.7814 | SRCC = 0.7864 ] +INFO:root:Average loss=0.3030421319298255 +INFO:root: +[Valid][4][UTT][ MSE = 0.3245 | LCC = 0.7921 | SRCC = 0.7907 ] +INFO:root:Average loss=0.2667125912430959 +INFO:root: +[Valid][5][UTT][ MSE = 0.3669 | LCC = 0.7934 | SRCC = 0.7915 ] +INFO:root:Average loss=0.24781602831223074 +INFO:root: +[Valid][6][UTT][ MSE = 0.3481 | LCC = 0.7884 | SRCC = 0.7880 ] +INFO:root:Average loss=0.2410315107076596 +INFO:root: +[Valid][7][UTT][ MSE = 0.4362 | LCC = 0.8036 | SRCC = 0.8033 ] +INFO:root:Average loss=0.21523018467884797 +INFO:root: +[Valid][8][UTT][ MSE = 0.3290 | LCC = 0.8034 | SRCC = 0.8010 ] +INFO:root:Average loss=0.2119864930327122 +INFO:root: +[Valid][9][UTT][ MSE = 0.3378 | LCC = 0.8038 | SRCC = 0.8006 ] +INFO:root:Average loss=0.21060514450073242 +INFO:root: +[Valid][10][UTT][ MSE = 0.3100 | LCC = 0.8001 | SRCC = 0.7984 ] +INFO:root:Average loss=0.18442195625259325 +INFO:root: +[Valid][11][UTT][ MSE = 0.3118 | LCC = 0.8042 | SRCC = 0.8039 ] +INFO:root:Average loss=0.17259460563460985 +INFO:root: +[Valid][12][UTT][ MSE = 0.3284 | LCC = 0.8000 | SRCC = 0.8009 ] +INFO:root:Average loss=0.19263937696814537 +INFO:root: +[Valid][13][UTT][ MSE = 0.3071 | LCC = 0.8024 | SRCC = 0.8021 ] +INFO:root:Average loss=0.15480263798664778 +INFO:root: +[Valid][14][UTT][ MSE = 0.3205 | LCC = 0.8073 | SRCC = 0.8059 ] +INFO:root:Average loss=0.1441481155462754 +INFO:root: +[Valid][15][UTT][ MSE = 0.3273 | LCC = 0.8069 | SRCC = 0.8060 ] +INFO:root:Average loss=0.1395328381122687 +INFO:root: +[Valid][16][UTT][ MSE = 0.3143 | LCC = 0.7993 | SRCC = 0.7983 ] +INFO:root:Average loss=0.1404255031584165 +INFO:root: +[Valid][17][UTT][ MSE = 0.3237 | LCC = 0.8045 | SRCC = 0.8022 ] +INFO:root:Average loss=0.1350811919531761 +INFO:root: +[Valid][18][UTT][ MSE = 0.3387 | LCC = 0.8029 | SRCC = 0.8015 ] +INFO:root:Average loss=0.12946516829423416 +INFO:root: +[Valid][19][UTT][ MSE = 0.3564 | LCC = 0.8045 | SRCC = 0.8028 ] +INFO:root:Average loss=0.13117332173845705 +INFO:root: +[Valid][20][UTT][ MSE = 0.3152 | LCC = 0.7997 | SRCC = 0.7987 ] +INFO:root:Average loss=0.11076740108621426 +INFO:root: +[Valid][21][UTT][ MSE = 0.3225 | LCC = 0.8024 | SRCC = 0.8004 ] +INFO:root:Average loss=0.1282612128326526 +INFO:root: +[Valid][22][UTT][ MSE = 0.3706 | LCC = 0.8050 | SRCC = 0.8061 ] +INFO:root:Average loss=0.10546119816792317 +INFO:root: +[Valid][23][UTT][ MSE = 0.3439 | LCC = 0.8051 | SRCC = 0.8044 ] +INFO:root:Average loss=0.09865652893980344 +INFO:root: +[Valid][24][UTT][ MSE = 0.3020 | LCC = 0.8044 | SRCC = 0.8039 ] +INFO:root:Average loss=0.08918962016319618 +INFO:root: +[Valid][25][UTT][ MSE = 0.2954 | LCC = 0.8077 | SRCC = 0.8061 ] +INFO:root:Average loss=0.09823543974795403 +INFO:root: +[Valid][26][UTT][ MSE = 0.2981 | LCC = 0.8090 | SRCC = 0.8078 ] +INFO:root:Average loss=0.09301110643606919 +INFO:root: +[Valid][27][UTT][ MSE = 0.3235 | LCC = 0.8074 | SRCC = 0.8056 ] +INFO:root:Average loss=0.08571803899338612 +INFO:root: +[Valid][28][UTT][ MSE = 0.2980 | LCC = 0.8084 | SRCC = 0.8070 ] +INFO:root:Average loss=0.07799852567796524 +INFO:root: +[Valid][29][UTT][ MSE = 0.3037 | LCC = 0.8024 | SRCC = 0.8029 ] +INFO:root:Average loss=0.07721459755721764 +INFO:root: +[Valid][30][UTT][ MSE = 0.2975 | LCC = 0.8061 | SRCC = 0.8042 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3087 | LCC = 0.7978 | SRCC = 0.7982 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run3/config.gin b/results_bvcc/w2v2_large_asr_de/layer16/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run3/model.pt b/results_bvcc/w2v2_large_asr_de/layer16/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..88525fb34726cb5418d5588981e6b8cf8b8453aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5070e18f8cc9de1d795dcd35cbbd7daf35d5593df37e9c6f6682b145d1faf29a +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run3/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer16/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6865c5ea076de2e820c98ee2a29423ae44d43625 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d243e2bebb993eef8b8b944ab4fe2dd39611100211b3712326420f3f74767f64 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run3/train.log b/results_bvcc/w2v2_large_asr_de/layer16/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..afbc7376d6fe559ed89c534924d62d9b17a7596a --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/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=0.5677066109119318 +INFO:root: +[Valid][0][UTT][ MSE = 0.4210 | LCC = 0.7300 | SRCC = 0.7335 ] +INFO:root:Average loss=0.38967104217945003 +INFO:root: +[Valid][1][UTT][ MSE = 0.3964 | LCC = 0.7636 | SRCC = 0.7632 ] +INFO:root:Average loss=0.31829489118013626 +INFO:root: +[Valid][2][UTT][ MSE = 0.3763 | LCC = 0.7884 | SRCC = 0.7887 ] +INFO:root:Average loss=0.29417148729165393 +INFO:root: +[Valid][3][UTT][ MSE = 0.3270 | LCC = 0.7928 | SRCC = 0.7895 ] +INFO:root:Average loss=0.2887208446478232 +INFO:root: +[Valid][4][UTT][ MSE = 0.3544 | LCC = 0.7974 | SRCC = 0.7946 ] +INFO:root:Average loss=0.2608552649617195 +INFO:root: +[Valid][5][UTT][ MSE = 0.3418 | LCC = 0.8063 | SRCC = 0.8023 ] +INFO:root:Average loss=0.2364271288880935 +INFO:root: +[Valid][6][UTT][ MSE = 0.3251 | LCC = 0.8033 | SRCC = 0.8009 ] +INFO:root:Average loss=0.25509140487664783 +INFO:root: +[Valid][7][UTT][ MSE = 0.3052 | LCC = 0.8066 | SRCC = 0.8023 ] +INFO:root:Average loss=0.22582589242702875 +INFO:root: +[Valid][8][UTT][ MSE = 0.3100 | LCC = 0.7979 | SRCC = 0.7932 ] +INFO:root:Average loss=0.20637991776069006 +INFO:root: +[Valid][9][UTT][ MSE = 0.3159 | LCC = 0.8035 | SRCC = 0.7991 ] +INFO:root:Average loss=0.20483296546034324 +INFO:root: +[Valid][10][UTT][ MSE = 0.2998 | LCC = 0.8130 | SRCC = 0.8110 ] +INFO:root:Average loss=0.19791897997642174 +INFO:root: +[Valid][11][UTT][ MSE = 0.3208 | LCC = 0.8084 | SRCC = 0.8063 ] +INFO:root:Average loss=0.1713402396402298 +INFO:root: +[Valid][12][UTT][ MSE = 0.3031 | LCC = 0.8110 | SRCC = 0.8083 ] +INFO:root:Average loss=0.1707815322547387 +INFO:root: +[Valid][13][UTT][ MSE = 0.2936 | LCC = 0.8110 | SRCC = 0.8079 ] +INFO:root:Average loss=0.1630601514226351 +INFO:root: +[Valid][14][UTT][ MSE = 0.3456 | LCC = 0.8061 | SRCC = 0.8046 ] +INFO:root:Average loss=0.1574166470613235 +INFO:root: +[Valid][15][UTT][ MSE = 0.3053 | LCC = 0.8094 | SRCC = 0.8067 ] +INFO:root:Average loss=0.15079497192532587 +INFO:root: +[Valid][16][UTT][ MSE = 0.3056 | LCC = 0.8063 | SRCC = 0.8044 ] +INFO:root:Average loss=0.1472923192076194 +INFO:root: +[Valid][17][UTT][ MSE = 0.3269 | LCC = 0.8035 | SRCC = 0.8021 ] +INFO:root:Average loss=0.13758231460666046 +INFO:root: +[Valid][18][UTT][ MSE = 0.3018 | LCC = 0.8083 | SRCC = 0.8065 ] +INFO:root:Average loss=0.14641106911958793 +INFO:root: +[Valid][19][UTT][ MSE = 0.2980 | LCC = 0.8102 | SRCC = 0.8084 ] +INFO:root:Average loss=0.12628652145847297 +INFO:root: +[Valid][20][UTT][ MSE = 0.2907 | LCC = 0.8103 | SRCC = 0.8094 ] +INFO:root:Average loss=0.10971651407770622 +INFO:root: +[Valid][21][UTT][ MSE = 0.3109 | LCC = 0.8085 | SRCC = 0.8077 ] +INFO:root:Average loss=0.11612961040093349 +INFO:root: +[Valid][22][UTT][ MSE = 0.2995 | LCC = 0.8047 | SRCC = 0.8027 ] +INFO:root:Average loss=0.10912918858230114 +INFO:root: +[Valid][23][UTT][ MSE = 0.2991 | LCC = 0.8044 | SRCC = 0.8036 ] +INFO:root:Average loss=0.11126169667412074 +INFO:root: +[Valid][24][UTT][ MSE = 0.3324 | LCC = 0.7989 | SRCC = 0.7985 ] +INFO:root:Average loss=0.10607839495134659 +INFO:root: +[Valid][25][UTT][ MSE = 0.3449 | LCC = 0.7985 | SRCC = 0.7987 ] +INFO:root:Average loss=0.0908988753381448 +INFO:root: +[Valid][26][UTT][ MSE = 0.2906 | LCC = 0.8067 | SRCC = 0.8047 ] +INFO:root:Average loss=0.10024463566832054 +INFO:root: +[Valid][27][UTT][ MSE = 0.4031 | LCC = 0.8001 | SRCC = 0.8027 ] +INFO:root:Average loss=0.08439009493360153 +INFO:root: +[Valid][28][UTT][ MSE = 0.3014 | LCC = 0.8044 | SRCC = 0.8023 ] +INFO:root:Average loss=0.08465025908289811 +INFO:root: +[Valid][29][UTT][ MSE = 0.2962 | LCC = 0.8047 | SRCC = 0.8030 ] +INFO:root:Average loss=0.08554124292463829 +INFO:root: +[Valid][30][UTT][ MSE = 0.2987 | LCC = 0.8049 | SRCC = 0.8028 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3138 | LCC = 0.7993 | SRCC = 0.8019 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run4/config.gin b/results_bvcc/w2v2_large_asr_de/layer16/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run4/model.pt b/results_bvcc/w2v2_large_asr_de/layer16/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bad977a60c3c2c234610c8d162d72366a5d9ac82 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:915aeab6d70830527413f5753e08017eda60fa80fb582df242cd867ec1505401 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run4/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer16/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2453f33fadc170c86fbe21cf92623ed40666c2c4 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:02cbe004d573f1f03646a08a8db6ba3a4fed2b1c49a3479159ffb73c23509ab6 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer16/run4/train.log b/results_bvcc/w2v2_large_asr_de/layer16/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..36e0322f506c334988eb28314cfcaa979dfd4f0d --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer16/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=0.5389253902129638 +INFO:root: +[Valid][0][UTT][ MSE = 0.3978 | LCC = 0.7487 | SRCC = 0.7520 ] +INFO:root:Average loss=0.35723062814810336 +INFO:root: +[Valid][1][UTT][ MSE = 0.3379 | LCC = 0.7797 | SRCC = 0.7825 ] +INFO:root:Average loss=0.33311747243771184 +INFO:root: +[Valid][2][UTT][ MSE = 0.3680 | LCC = 0.7933 | SRCC = 0.7898 ] +INFO:root:Average loss=0.28619858909111756 +INFO:root: +[Valid][3][UTT][ MSE = 0.3254 | LCC = 0.7997 | SRCC = 0.7957 ] +INFO:root:Average loss=0.27818460304003495 +INFO:root: +[Valid][4][UTT][ MSE = 0.3321 | LCC = 0.7978 | SRCC = 0.7929 ] +INFO:root:Average loss=0.2761309863283084 +INFO:root: +[Valid][5][UTT][ MSE = 0.3138 | LCC = 0.8016 | SRCC = 0.7983 ] +INFO:root:Average loss=0.2371363401030883 +INFO:root: +[Valid][6][UTT][ MSE = 0.3056 | LCC = 0.8032 | SRCC = 0.7996 ] +INFO:root:Average loss=0.2281729131937027 +INFO:root: +[Valid][7][UTT][ MSE = 0.2951 | LCC = 0.8114 | SRCC = 0.8080 ] +INFO:root:Average loss=0.22560934798839766 +INFO:root: +[Valid][8][UTT][ MSE = 0.3303 | LCC = 0.8007 | SRCC = 0.8014 ] +INFO:root:Average loss=0.22843127907850805 +INFO:root: +[Valid][9][UTT][ MSE = 0.3033 | LCC = 0.8079 | SRCC = 0.8072 ] +INFO:root:Average loss=0.20993703661056665 +INFO:root: +[Valid][10][UTT][ MSE = 0.3414 | LCC = 0.7956 | SRCC = 0.7949 ] +INFO:root:Average loss=0.2036919692196907 +INFO:root: +[Valid][11][UTT][ MSE = 0.3275 | LCC = 0.8043 | SRCC = 0.8034 ] +INFO:root:Average loss=0.17632257575408006 +INFO:root: +[Valid][12][UTT][ MSE = 0.3257 | LCC = 0.8027 | SRCC = 0.8000 ] +INFO:root:Average loss=0.1755575788899874 +INFO:root: +[Valid][13][UTT][ MSE = 0.3078 | LCC = 0.8075 | SRCC = 0.8041 ] +INFO:root:Average loss=0.174609269373692 +INFO:root: +[Valid][14][UTT][ MSE = 0.3096 | LCC = 0.8033 | SRCC = 0.8000 ] +INFO:root:Average loss=0.15696673563275582 +INFO:root: +[Valid][15][UTT][ MSE = 0.3022 | LCC = 0.8079 | SRCC = 0.8042 ] +INFO:root:Average loss=0.15490486377324814 +INFO:root: +[Valid][16][UTT][ MSE = 0.3031 | LCC = 0.8043 | SRCC = 0.8014 ] +INFO:root:Average loss=0.13920761721256453 +INFO:root: +[Valid][17][UTT][ MSE = 0.3254 | LCC = 0.8063 | SRCC = 0.8026 ] +INFO:root:Average loss=0.13404206759654558 +INFO:root: +[Valid][18][UTT][ MSE = 0.2991 | LCC = 0.8045 | SRCC = 0.8029 ] +INFO:root:Average loss=0.12747386795205948 +INFO:root: +[Valid][19][UTT][ MSE = 0.3038 | LCC = 0.8027 | SRCC = 0.8009 ] +INFO:root:Average loss=0.12092314259364055 +INFO:root: +[Valid][20][UTT][ MSE = 0.3021 | LCC = 0.8020 | SRCC = 0.7987 ] +INFO:root:Average loss=0.12304624513937877 +INFO:root: +[Valid][21][UTT][ MSE = 0.3074 | LCC = 0.7984 | SRCC = 0.7965 ] +INFO:root:Average loss=0.1064494451842247 +INFO:root: +[Valid][22][UTT][ MSE = 0.2985 | LCC = 0.8062 | SRCC = 0.8042 ] +INFO:root:Average loss=0.09889094326167534 +INFO:root: +[Valid][23][UTT][ MSE = 0.3022 | LCC = 0.7999 | SRCC = 0.7951 ] +INFO:root:Average loss=0.10024719924116746 +INFO:root: +[Valid][24][UTT][ MSE = 0.2944 | LCC = 0.8045 | SRCC = 0.8032 ] +INFO:root:Average loss=0.09855504515461433 +INFO:root: +[Valid][25][UTT][ MSE = 0.3377 | LCC = 0.8006 | SRCC = 0.7987 ] +INFO:root:Average loss=0.0897445416985414 +INFO:root: +[Valid][26][UTT][ MSE = 0.2945 | LCC = 0.8053 | SRCC = 0.8027 ] +INFO:root:Average loss=0.09385144428755993 +INFO:root: +[Valid][27][UTT][ MSE = 0.2901 | LCC = 0.8105 | SRCC = 0.8083 ] +INFO:root:Average loss=0.08767436697888069 +INFO:root: +[Valid][28][UTT][ MSE = 0.3362 | LCC = 0.8008 | SRCC = 0.7999 ] +INFO:root:Average loss=0.08080437120336753 +INFO:root: +[Valid][29][UTT][ MSE = 0.2995 | LCC = 0.8042 | SRCC = 0.8023 ] +INFO:root:Average loss=0.07892066178222497 +INFO:root: +[Valid][30][UTT][ MSE = 0.2960 | LCC = 0.8047 | SRCC = 0.8022 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3090 | LCC = 0.8106 | SRCC = 0.8118 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run0/config.gin b/results_bvcc/w2v2_large_asr_de/layer18/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run0/model.pt b/results_bvcc/w2v2_large_asr_de/layer18/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ef71babc86e7ccba790ed04aa9f2f6f1f758f197 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8430844275156807144c26716ffde1ffcde1ff741784999dbd4991136908a8d2 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run0/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer18/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..11616ac2c762c9e5d65da5b07f797309284e13b0 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:94387e37c830d044100b79f61df32473986a106b182a828081b8806dde86e3b9 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run0/train.log b/results_bvcc/w2v2_large_asr_de/layer18/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a289cfba36d3e277a8b28b2b1b2c993d3c54500a --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/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=0.5430552053910035 +INFO:root: +[Valid][0][UTT][ MSE = 0.4448 | LCC = 0.7331 | SRCC = 0.7358 ] +INFO:root:Average loss=0.3959625854324072 +INFO:root: +[Valid][1][UTT][ MSE = 0.4089 | LCC = 0.7667 | SRCC = 0.7698 ] +INFO:root:Average loss=0.34559528548748064 +INFO:root: +[Valid][2][UTT][ MSE = 0.3589 | LCC = 0.7819 | SRCC = 0.7822 ] +INFO:root:Average loss=0.3211754023646697 +INFO:root: +[Valid][3][UTT][ MSE = 0.3553 | LCC = 0.7900 | SRCC = 0.7891 ] +INFO:root:Average loss=0.286390663530582 +INFO:root: +[Valid][4][UTT][ MSE = 0.3446 | LCC = 0.7871 | SRCC = 0.7866 ] +INFO:root:Average loss=0.30325997200531835 +INFO:root: +[Valid][5][UTT][ MSE = 0.3443 | LCC = 0.7920 | SRCC = 0.7908 ] +INFO:root:Average loss=0.28401236694592696 +INFO:root: +[Valid][6][UTT][ MSE = 0.3843 | LCC = 0.7907 | SRCC = 0.7955 ] +INFO:root:Average loss=0.2742432932823132 +INFO:root: +[Valid][7][UTT][ MSE = 0.3356 | LCC = 0.7949 | SRCC = 0.7924 ] +INFO:root:Average loss=0.248327687000617 +INFO:root: +[Valid][8][UTT][ MSE = 0.3155 | LCC = 0.8008 | SRCC = 0.8016 ] +INFO:root:Average loss=0.23902654552306885 +INFO:root: +[Valid][9][UTT][ MSE = 0.3373 | LCC = 0.7919 | SRCC = 0.7926 ] +INFO:root:Average loss=0.2301817952822416 +INFO:root: +[Valid][10][UTT][ MSE = 0.3751 | LCC = 0.7993 | SRCC = 0.8006 ] +INFO:root:Average loss=0.21343464155991873 +INFO:root: +[Valid][11][UTT][ MSE = 0.3483 | LCC = 0.7982 | SRCC = 0.7962 ] +INFO:root:Average loss=0.2058701044282852 +INFO:root: +[Valid][12][UTT][ MSE = 0.3246 | LCC = 0.7931 | SRCC = 0.7907 ] +INFO:root:Average loss=0.1963260690562236 +INFO:root: +[Valid][13][UTT][ MSE = 0.3244 | LCC = 0.7970 | SRCC = 0.7968 ] +INFO:root:Average loss=0.20117889917813814 +INFO:root: +[Valid][14][UTT][ MSE = 0.3111 | LCC = 0.7988 | SRCC = 0.7986 ] +INFO:root:Average loss=0.1948957870212885 +INFO:root: +[Valid][15][UTT][ MSE = 0.3399 | LCC = 0.7941 | SRCC = 0.7952 ] +INFO:root:Average loss=0.17682396362607294 +INFO:root: +[Valid][16][UTT][ MSE = 0.3198 | LCC = 0.7955 | SRCC = 0.7951 ] +INFO:root:Average loss=0.1694297457161622 +INFO:root: +[Valid][17][UTT][ MSE = 0.3513 | LCC = 0.7971 | SRCC = 0.7975 ] +INFO:root:Average loss=0.16127249799095666 +INFO:root: +[Valid][18][UTT][ MSE = 0.3005 | LCC = 0.8020 | SRCC = 0.8027 ] +INFO:root:Average loss=0.16327807479179823 +INFO:root: +[Valid][19][UTT][ MSE = 0.3278 | LCC = 0.8013 | SRCC = 0.8016 ] +INFO:root:Average loss=0.15003115989458868 +INFO:root: +[Valid][20][UTT][ MSE = 0.3276 | LCC = 0.7916 | SRCC = 0.7915 ] +INFO:root:Average loss=0.14197518112949836 +INFO:root: +[Valid][21][UTT][ MSE = 0.3326 | LCC = 0.7833 | SRCC = 0.7808 ] +INFO:root:Average loss=0.1747162717466171 +INFO:root: +[Valid][22][UTT][ MSE = 0.3051 | LCC = 0.7982 | SRCC = 0.7977 ] +INFO:root:Average loss=0.13249131970298597 +INFO:root: +[Valid][23][UTT][ MSE = 0.3315 | LCC = 0.7826 | SRCC = 0.7809 ] +INFO:root:Average loss=0.14645838785247925 +INFO:root: +[Valid][24][UTT][ MSE = 0.3233 | LCC = 0.7946 | SRCC = 0.7943 ] +INFO:root:Average loss=0.12575224643716446 +INFO:root: +[Valid][25][UTT][ MSE = 0.3129 | LCC = 0.7916 | SRCC = 0.7898 ] +INFO:root:Average loss=0.1271457908054193 +INFO:root: +[Valid][26][UTT][ MSE = 0.3220 | LCC = 0.7901 | SRCC = 0.7883 ] +INFO:root:Average loss=0.1351425593289045 +INFO:root: +[Valid][27][UTT][ MSE = 0.3095 | LCC = 0.7930 | SRCC = 0.7917 ] +INFO:root:Average loss=0.11435517363059215 +INFO:root: +[Valid][28][UTT][ MSE = 0.3139 | LCC = 0.7941 | SRCC = 0.7933 ] +INFO:root:Average loss=0.11060014099646838 +INFO:root: +[Valid][29][UTT][ MSE = 0.3175 | LCC = 0.7900 | SRCC = 0.7889 ] +INFO:root:Average loss=0.10544308103047885 +INFO:root: +[Valid][30][UTT][ MSE = 0.3304 | LCC = 0.7881 | SRCC = 0.7885 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3075 | LCC = 0.7996 | SRCC = 0.8004 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run1/config.gin b/results_bvcc/w2v2_large_asr_de/layer18/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run1/model.pt b/results_bvcc/w2v2_large_asr_de/layer18/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d7014babf81fe0f81cf7fb736124ebbf529ed668 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:66e1cf46b27cda74330c72ced77bf0be75391926351088eeaba864ede6294984 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run1/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer18/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d5f059c96d38c4a33f7277ae1ad697ded992dd13 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ea48d8eb0423e0e44e8243ebce5f90508f5e8ed3bbd5c9c8b63b5bad7ba975ab +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run1/train.log b/results_bvcc/w2v2_large_asr_de/layer18/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..86613d322f2f2f1690b90ae701598dbc99a18bdc --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/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=0.5655203094849219 +INFO:root: +[Valid][0][UTT][ MSE = 0.4447 | LCC = 0.7226 | SRCC = 0.7260 ] +INFO:root:Average loss=0.39102101478821194 +INFO:root: +[Valid][1][UTT][ MSE = 0.4000 | LCC = 0.7592 | SRCC = 0.7597 ] +INFO:root:Average loss=0.3458898342572726 +INFO:root: +[Valid][2][UTT][ MSE = 0.3552 | LCC = 0.7824 | SRCC = 0.7805 ] +INFO:root:Average loss=0.3028152048205718 +INFO:root: +[Valid][3][UTT][ MSE = 0.3709 | LCC = 0.7864 | SRCC = 0.7853 ] +INFO:root:Average loss=0.2978914892062163 +INFO:root: +[Valid][4][UTT][ MSE = 0.3384 | LCC = 0.7928 | SRCC = 0.7913 ] +INFO:root:Average loss=0.2772844993533232 +INFO:root: +[Valid][5][UTT][ MSE = 0.3416 | LCC = 0.7987 | SRCC = 0.7967 ] +INFO:root:Average loss=0.28181845541947925 +INFO:root: +[Valid][6][UTT][ MSE = 0.3099 | LCC = 0.7948 | SRCC = 0.7930 ] +INFO:root:Average loss=0.25110779301478314 +INFO:root: +[Valid][7][UTT][ MSE = 0.3147 | LCC = 0.7978 | SRCC = 0.7968 ] +INFO:root:Average loss=0.23605375469495088 +INFO:root: +[Valid][8][UTT][ MSE = 0.3278 | LCC = 0.7931 | SRCC = 0.7922 ] +INFO:root:Average loss=0.23314139686333826 +INFO:root: +[Valid][9][UTT][ MSE = 0.3226 | LCC = 0.7936 | SRCC = 0.7930 ] +INFO:root:Average loss=0.23043293276658425 +INFO:root: +[Valid][10][UTT][ MSE = 0.3300 | LCC = 0.7959 | SRCC = 0.7943 ] +INFO:root:Average loss=0.21144506058249718 +INFO:root: +[Valid][11][UTT][ MSE = 0.3198 | LCC = 0.7900 | SRCC = 0.7866 ] +INFO:root:Average loss=0.2154373091000777 +INFO:root: +[Valid][12][UTT][ MSE = 0.3091 | LCC = 0.7967 | SRCC = 0.7950 ] +INFO:root:Average loss=0.2032951328616876 +INFO:root: +[Valid][13][UTT][ MSE = 0.3111 | LCC = 0.8010 | SRCC = 0.7989 ] +INFO:root:Average loss=0.19073782460047647 +INFO:root: +[Valid][14][UTT][ MSE = 0.3134 | LCC = 0.8004 | SRCC = 0.7989 ] +INFO:root:Average loss=0.1784702934897863 +INFO:root: +[Valid][15][UTT][ MSE = 0.3376 | LCC = 0.7915 | SRCC = 0.7880 ] +INFO:root:Average loss=0.17549477231044036 +INFO:root: +[Valid][16][UTT][ MSE = 0.3036 | LCC = 0.7966 | SRCC = 0.7944 ] +INFO:root:Average loss=0.19362997225461862 +INFO:root: +[Valid][17][UTT][ MSE = 0.3241 | LCC = 0.7926 | SRCC = 0.7906 ] +INFO:root:Average loss=0.15950619658598533 +INFO:root: +[Valid][18][UTT][ MSE = 0.3224 | LCC = 0.7882 | SRCC = 0.7851 ] +INFO:root:Average loss=0.16380978299257082 +INFO:root: +[Valid][19][UTT][ MSE = 0.3075 | LCC = 0.7973 | SRCC = 0.7952 ] +INFO:root:Average loss=0.15366996203859648 +INFO:root: +[Valid][20][UTT][ MSE = 0.3202 | LCC = 0.7901 | SRCC = 0.7881 ] +INFO:root:Average loss=0.16021833062554017 +INFO:root: +[Valid][21][UTT][ MSE = 0.3245 | LCC = 0.7901 | SRCC = 0.7884 ] +INFO:root:Average loss=0.14059047458263543 +INFO:root: +[Valid][22][UTT][ MSE = 0.3365 | LCC = 0.7912 | SRCC = 0.7899 ] +INFO:root:Average loss=0.14815025365887544 +INFO:root: +[Valid][23][UTT][ MSE = 0.3254 | LCC = 0.7922 | SRCC = 0.7899 ] +INFO:root:Average loss=0.14857764981496027 +INFO:root: +[Valid][24][UTT][ MSE = 0.3104 | LCC = 0.7942 | SRCC = 0.7921 ] +INFO:root:Average loss=0.12374840657680462 +INFO:root: +[Valid][25][UTT][ MSE = 0.3029 | LCC = 0.7977 | SRCC = 0.7967 ] +INFO:root:Average loss=0.13833304752524084 +INFO:root: +[Valid][26][UTT][ MSE = 0.3036 | LCC = 0.7985 | SRCC = 0.7976 ] +INFO:root:Average loss=0.11772080491750668 +INFO:root: +[Valid][27][UTT][ MSE = 0.3225 | LCC = 0.7898 | SRCC = 0.7864 ] +INFO:root:Average loss=0.11279832800993553 +INFO:root: +[Valid][28][UTT][ MSE = 0.3038 | LCC = 0.7971 | SRCC = 0.7964 ] +INFO:root:Average loss=0.11533730457990597 +INFO:root: +[Valid][29][UTT][ MSE = 0.3195 | LCC = 0.7931 | SRCC = 0.7910 ] +INFO:root:Average loss=0.11719377716191304 +INFO:root: +[Valid][30][UTT][ MSE = 0.3153 | LCC = 0.7902 | SRCC = 0.7878 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3136 | LCC = 0.8073 | SRCC = 0.8078 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run2/config.gin b/results_bvcc/w2v2_large_asr_de/layer18/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run2/model.pt b/results_bvcc/w2v2_large_asr_de/layer18/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..097cae65e26584ff08072423fa09882a039b640d --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:97b29571b7c0d3622121a06864c12838cf04dd3fe94a73668879b9c6f17b3e8f +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run2/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer18/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b15824a0c359d42edceb61947db35b899309c6e4 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe7818236e0a3be9a9763085cfa6f396ab25db4e704eaa045a035a9cddf92a11 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run2/train.log b/results_bvcc/w2v2_large_asr_de/layer18/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..837a064b60a4aded8a5f26720342ac02e8ebcc83 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/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=0.6075752862753012 +INFO:root: +[Valid][0][UTT][ MSE = 0.4418 | LCC = 0.7115 | SRCC = 0.7134 ] +INFO:root:Average loss=0.406880745329918 +INFO:root: +[Valid][1][UTT][ MSE = 0.3980 | LCC = 0.7559 | SRCC = 0.7588 ] +INFO:root:Average loss=0.3636105465583312 +INFO:root: +[Valid][2][UTT][ MSE = 0.4282 | LCC = 0.7677 | SRCC = 0.7737 ] +INFO:root:Average loss=0.3325467186096387 +INFO:root: +[Valid][3][UTT][ MSE = 0.4216 | LCC = 0.7740 | SRCC = 0.7767 ] +INFO:root:Average loss=0.29714417190123826 +INFO:root: +[Valid][4][UTT][ MSE = 0.3709 | LCC = 0.7853 | SRCC = 0.7844 ] +INFO:root:Average loss=0.2760877691400357 +INFO:root: +[Valid][5][UTT][ MSE = 0.3945 | LCC = 0.7846 | SRCC = 0.7855 ] +INFO:root:Average loss=0.26689000313098615 +INFO:root: +[Valid][6][UTT][ MSE = 0.3146 | LCC = 0.7959 | SRCC = 0.7957 ] +INFO:root:Average loss=0.26079726906923145 +INFO:root: +[Valid][7][UTT][ MSE = 0.3306 | LCC = 0.7922 | SRCC = 0.7902 ] +INFO:root:Average loss=0.2487953796218603 +INFO:root: +[Valid][8][UTT][ MSE = 0.3401 | LCC = 0.7967 | SRCC = 0.7955 ] +INFO:root:Average loss=0.22680648817465857 +INFO:root: +[Valid][9][UTT][ MSE = 0.3418 | LCC = 0.7930 | SRCC = 0.7916 ] +INFO:root:Average loss=0.2503543851467279 +INFO:root: +[Valid][10][UTT][ MSE = 0.4732 | LCC = 0.7812 | SRCC = 0.7885 ] +INFO:root:Average loss=0.21726452692961082 +INFO:root: +[Valid][11][UTT][ MSE = 0.3207 | LCC = 0.7920 | SRCC = 0.7900 ] +INFO:root:Average loss=0.20752034718409562 +INFO:root: +[Valid][12][UTT][ MSE = 0.3711 | LCC = 0.7972 | SRCC = 0.7985 ] +INFO:root:Average loss=0.19374217073886824 +INFO:root: +[Valid][13][UTT][ MSE = 0.3096 | LCC = 0.7936 | SRCC = 0.7924 ] +INFO:root:Average loss=0.18798339634369582 +INFO:root: +[Valid][14][UTT][ MSE = 0.3131 | LCC = 0.7986 | SRCC = 0.7983 ] +INFO:root:Average loss=0.1768167058053689 +INFO:root: +[Valid][15][UTT][ MSE = 0.3123 | LCC = 0.7980 | SRCC = 0.7960 ] +INFO:root:Average loss=0.18033930783470473 +INFO:root: +[Valid][16][UTT][ MSE = 0.3702 | LCC = 0.7973 | SRCC = 0.7973 ] +INFO:root:Average loss=0.1696244661624615 +INFO:root: +[Valid][17][UTT][ MSE = 0.3249 | LCC = 0.7919 | SRCC = 0.7904 ] +INFO:root:Average loss=0.16574703615445358 +INFO:root: +[Valid][18][UTT][ MSE = 0.3012 | LCC = 0.8004 | SRCC = 0.7999 ] +INFO:root:Average loss=0.16833562661822027 +INFO:root: +[Valid][19][UTT][ MSE = 0.3172 | LCC = 0.8015 | SRCC = 0.8028 ] +INFO:root:Average loss=0.16285048148188835 +INFO:root: +[Valid][20][UTT][ MSE = 0.3325 | LCC = 0.7996 | SRCC = 0.7989 ] +INFO:root:Average loss=0.13990158415757692 +INFO:root: +[Valid][21][UTT][ MSE = 0.3149 | LCC = 0.7970 | SRCC = 0.7962 ] +INFO:root:Average loss=0.1529277238326195 +INFO:root: +[Valid][22][UTT][ MSE = 0.3087 | LCC = 0.7977 | SRCC = 0.7967 ] +INFO:root:Average loss=0.14422086034065637 +INFO:root: +[Valid][23][UTT][ MSE = 0.3599 | LCC = 0.7966 | SRCC = 0.7974 ] +INFO:root:Average loss=0.13070155154818144 +INFO:root: +[Valid][24][UTT][ MSE = 0.3028 | LCC = 0.7973 | SRCC = 0.7959 ] +INFO:root:Average loss=0.13449823760833496 +INFO:root: +[Valid][25][UTT][ MSE = 0.3064 | LCC = 0.7970 | SRCC = 0.7956 ] +INFO:root:Average loss=0.1274630280259328 +INFO:root: +[Valid][26][UTT][ MSE = 0.3249 | LCC = 0.7932 | SRCC = 0.7930 ] +INFO:root:Average loss=0.11600212604762652 +INFO:root: +[Valid][27][UTT][ MSE = 0.3258 | LCC = 0.7931 | SRCC = 0.7927 ] +INFO:root:Average loss=0.1060332117172388 +INFO:root: +[Valid][28][UTT][ MSE = 0.3145 | LCC = 0.7975 | SRCC = 0.7972 ] +INFO:root:Average loss=0.11476653623275268 +INFO:root: +[Valid][29][UTT][ MSE = 0.3027 | LCC = 0.8017 | SRCC = 0.8009 ] +INFO:root:Average loss=0.11048942235990977 +INFO:root: +[Valid][30][UTT][ MSE = 0.3193 | LCC = 0.7934 | SRCC = 0.7917 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3274 | LCC = 0.7937 | SRCC = 0.7936 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run3/config.gin b/results_bvcc/w2v2_large_asr_de/layer18/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run3/model.pt b/results_bvcc/w2v2_large_asr_de/layer18/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d16cbe3be8b2da56d1c989b73c34cb8062da7543 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:00c9280ce98f119d9acfe5bb740c7487eaff1cb930fc451432375b0fac5641ac +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run3/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer18/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..80dba671e33d69bb28418a773ad1048d11ca4dc5 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ee81cde36225a862b169244e9a7886f506d0c512e0d11c0758f052b2f589d332 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run3/train.log b/results_bvcc/w2v2_large_asr_de/layer18/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..17b1ec3c725c1352a73846b3b77ffa38b07d211e --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/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=0.5661212775187615 +INFO:root: +[Valid][0][UTT][ MSE = 0.4567 | LCC = 0.7258 | SRCC = 0.7243 ] +INFO:root:Average loss=0.40549363348728573 +INFO:root: +[Valid][1][UTT][ MSE = 0.3849 | LCC = 0.7597 | SRCC = 0.7604 ] +INFO:root:Average loss=0.34565304124202484 +INFO:root: +[Valid][2][UTT][ MSE = 0.3718 | LCC = 0.7800 | SRCC = 0.7786 ] +INFO:root:Average loss=0.31272299625934696 +INFO:root: +[Valid][3][UTT][ MSE = 0.3424 | LCC = 0.7921 | SRCC = 0.7886 ] +INFO:root:Average loss=0.2946911991024629 +INFO:root: +[Valid][4][UTT][ MSE = 0.3266 | LCC = 0.8008 | SRCC = 0.7988 ] +INFO:root:Average loss=0.28509330119077975 +INFO:root: +[Valid][5][UTT][ MSE = 0.3232 | LCC = 0.8008 | SRCC = 0.7987 ] +INFO:root:Average loss=0.27759254838411623 +INFO:root: +[Valid][6][UTT][ MSE = 0.3637 | LCC = 0.7992 | SRCC = 0.7985 ] +INFO:root:Average loss=0.25433166325092316 +INFO:root: +[Valid][7][UTT][ MSE = 0.3239 | LCC = 0.7994 | SRCC = 0.7955 ] +INFO:root:Average loss=0.23547575317132166 +INFO:root: +[Valid][8][UTT][ MSE = 0.3190 | LCC = 0.8036 | SRCC = 0.8015 ] +INFO:root:Average loss=0.2328138510004068 +INFO:root: +[Valid][9][UTT][ MSE = 0.3201 | LCC = 0.7996 | SRCC = 0.7963 ] +INFO:root:Average loss=0.2236831369690406 +INFO:root: +[Valid][10][UTT][ MSE = 0.3451 | LCC = 0.8000 | SRCC = 0.7984 ] +INFO:root:Average loss=0.23330557365448046 +INFO:root: +[Valid][11][UTT][ MSE = 0.3212 | LCC = 0.7987 | SRCC = 0.7944 ] +INFO:root:Average loss=0.20195535226510122 +INFO:root: +[Valid][12][UTT][ MSE = 0.3295 | LCC = 0.7935 | SRCC = 0.7897 ] +INFO:root:Average loss=0.1966242830340679 +INFO:root: +[Valid][13][UTT][ MSE = 0.3198 | LCC = 0.8058 | SRCC = 0.8037 ] +INFO:root:Average loss=0.18679254062664816 +INFO:root: +[Valid][14][UTT][ MSE = 0.3254 | LCC = 0.8057 | SRCC = 0.8041 ] +INFO:root:Average loss=0.20721092695991197 +INFO:root: +[Valid][15][UTT][ MSE = 0.3038 | LCC = 0.8119 | SRCC = 0.8088 ] +INFO:root:Average loss=0.1725694858110868 +INFO:root: +[Valid][16][UTT][ MSE = 0.3029 | LCC = 0.8043 | SRCC = 0.8009 ] +INFO:root:Average loss=0.16814578859469828 +INFO:root: +[Valid][17][UTT][ MSE = 0.3433 | LCC = 0.7963 | SRCC = 0.7945 ] +INFO:root:Average loss=0.17410614914619005 +INFO:root: +[Valid][18][UTT][ MSE = 0.3283 | LCC = 0.7990 | SRCC = 0.7975 ] +INFO:root:Average loss=0.15427666081067842 +INFO:root: +[Valid][19][UTT][ MSE = 0.3417 | LCC = 0.7971 | SRCC = 0.7950 ] +INFO:root:Average loss=0.14966926800134855 +INFO:root: +[Valid][20][UTT][ MSE = 0.3688 | LCC = 0.7993 | SRCC = 0.8001 ] +INFO:root:Average loss=0.1374995227998648 +INFO:root: +[Valid][21][UTT][ MSE = 0.3102 | LCC = 0.7987 | SRCC = 0.7947 ] +INFO:root:Average loss=0.13689952248182052 +INFO:root: +[Valid][22][UTT][ MSE = 0.3057 | LCC = 0.8025 | SRCC = 0.7996 ] +INFO:root:Average loss=0.13205707493500832 +INFO:root: +[Valid][23][UTT][ MSE = 0.3117 | LCC = 0.7989 | SRCC = 0.7959 ] +INFO:root:Average loss=0.14463827224114004 +INFO:root: +[Valid][24][UTT][ MSE = 0.3089 | LCC = 0.7970 | SRCC = 0.7935 ] +INFO:root:Average loss=0.13222027503145048 +INFO:root: +[Valid][25][UTT][ MSE = 0.2967 | LCC = 0.8059 | SRCC = 0.8026 ] +INFO:root:Average loss=0.121078071781458 +INFO:root: +[Valid][26][UTT][ MSE = 0.2991 | LCC = 0.8034 | SRCC = 0.8002 ] +INFO:root:Average loss=0.12199988311681992 +INFO:root: +[Valid][27][UTT][ MSE = 0.2943 | LCC = 0.8068 | SRCC = 0.8040 ] +INFO:root:Average loss=0.1055014926271561 +INFO:root: +[Valid][28][UTT][ MSE = 0.3132 | LCC = 0.7989 | SRCC = 0.7967 ] +INFO:root:Average loss=0.11440011515067174 +INFO:root: +[Valid][29][UTT][ MSE = 0.3015 | LCC = 0.8034 | SRCC = 0.7999 ] +INFO:root:Average loss=0.10570373620169285 +INFO:root: +[Valid][30][UTT][ MSE = 0.3395 | LCC = 0.8037 | SRCC = 0.8053 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3098 | LCC = 0.8054 | SRCC = 0.8028 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run4/config.gin b/results_bvcc/w2v2_large_asr_de/layer18/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run4/model.pt b/results_bvcc/w2v2_large_asr_de/layer18/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5f6d31988474d7aa79358f12376a8fd2c3c28396 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7a55894f47a09623ecdec463e652b0cdb9f79c9a66fbe8ee1989c80903992d22 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run4/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer18/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..fe9ee3ccce1ddb2110565cfbcdbfd6219a44d00f --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:74b83e456c71bf76bad28605376f578eec9bbceeecb1e9d22dc07ca014bf8d94 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer18/run4/train.log b/results_bvcc/w2v2_large_asr_de/layer18/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1272903d909551ef923d0e7f05885d5aec2a1dbf --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer18/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=0.5453994163335898 +INFO:root: +[Valid][0][UTT][ MSE = 0.4681 | LCC = 0.7257 | SRCC = 0.7268 ] +INFO:root:Average loss=0.3963668669263522 +INFO:root: +[Valid][1][UTT][ MSE = 0.3836 | LCC = 0.7636 | SRCC = 0.7644 ] +INFO:root:Average loss=0.3508294356557039 +INFO:root: +[Valid][2][UTT][ MSE = 0.5026 | LCC = 0.7703 | SRCC = 0.7775 ] +INFO:root:Average loss=0.3616045445967943 +INFO:root: +[Valid][3][UTT][ MSE = 0.3615 | LCC = 0.7809 | SRCC = 0.7797 ] +INFO:root:Average loss=0.2959902964723416 +INFO:root: +[Valid][4][UTT][ MSE = 0.3409 | LCC = 0.7866 | SRCC = 0.7853 ] +INFO:root:Average loss=0.28907874360298497 +INFO:root: +[Valid][5][UTT][ MSE = 0.3344 | LCC = 0.7935 | SRCC = 0.7922 ] +INFO:root:Average loss=0.27400187173715007 +INFO:root: +[Valid][6][UTT][ MSE = 0.3574 | LCC = 0.7902 | SRCC = 0.7890 ] +INFO:root:Average loss=0.2568397562091167 +INFO:root: +[Valid][7][UTT][ MSE = 0.3304 | LCC = 0.7952 | SRCC = 0.7931 ] +INFO:root:Average loss=0.25641035861693895 +INFO:root: +[Valid][8][UTT][ MSE = 0.3292 | LCC = 0.7917 | SRCC = 0.7906 ] +INFO:root:Average loss=0.23417788782180884 +INFO:root: +[Valid][9][UTT][ MSE = 0.3578 | LCC = 0.7875 | SRCC = 0.7851 ] +INFO:root:Average loss=0.239524357402936 +INFO:root: +[Valid][10][UTT][ MSE = 0.3365 | LCC = 0.7914 | SRCC = 0.7915 ] +INFO:root:Average loss=0.2229439260867926 +INFO:root: +[Valid][11][UTT][ MSE = 0.3880 | LCC = 0.7903 | SRCC = 0.7912 ] +INFO:root:Average loss=0.20671668514991418 +INFO:root: +[Valid][12][UTT][ MSE = 0.3511 | LCC = 0.7903 | SRCC = 0.7890 ] +INFO:root:Average loss=0.20019145930806795 +INFO:root: +[Valid][13][UTT][ MSE = 0.3201 | LCC = 0.7918 | SRCC = 0.7912 ] +INFO:root:Average loss=0.20359207154848638 +INFO:root: +[Valid][14][UTT][ MSE = 0.3313 | LCC = 0.7884 | SRCC = 0.7876 ] +INFO:root:Average loss=0.1766721477302221 +INFO:root: +[Valid][15][UTT][ MSE = 0.3285 | LCC = 0.7858 | SRCC = 0.7830 ] +INFO:root:Average loss=0.17226534938583007 +INFO:root: +[Valid][16][UTT][ MSE = 0.3512 | LCC = 0.7859 | SRCC = 0.7836 ] +INFO:root:Average loss=0.17437449460610366 +INFO:root: +[Valid][17][UTT][ MSE = 0.3154 | LCC = 0.7933 | SRCC = 0.7910 ] +INFO:root:Average loss=0.15867527688925082 +INFO:root: +[Valid][18][UTT][ MSE = 0.3280 | LCC = 0.7893 | SRCC = 0.7879 ] +INFO:root:Average loss=0.15337857785515296 +INFO:root: +[Valid][19][UTT][ MSE = 0.3450 | LCC = 0.7814 | SRCC = 0.7800 ] +INFO:root:Average loss=0.15320467022367013 +INFO:root: +[Valid][20][UTT][ MSE = 0.3613 | LCC = 0.7872 | SRCC = 0.7838 ] +INFO:root:Average loss=0.14304988630689108 +INFO:root: +[Valid][21][UTT][ MSE = 0.3368 | LCC = 0.7872 | SRCC = 0.7850 ] +INFO:root:Average loss=0.1405020456474561 +INFO:root: +[Valid][22][UTT][ MSE = 0.3489 | LCC = 0.7796 | SRCC = 0.7777 ] +INFO:root:Average loss=0.1314182865122954 +INFO:root: +[Valid][23][UTT][ MSE = 0.3454 | LCC = 0.7860 | SRCC = 0.7828 ] +INFO:root:Average loss=0.13691006725033125 +INFO:root: +[Valid][24][UTT][ MSE = 0.3164 | LCC = 0.7910 | SRCC = 0.7897 ] +INFO:root:Average loss=0.12765391323810968 +INFO:root: +[Valid][25][UTT][ MSE = 0.3378 | LCC = 0.7926 | SRCC = 0.7931 ] +INFO:root:Average loss=0.12290564513741395 +INFO:root: +[Valid][26][UTT][ MSE = 0.3396 | LCC = 0.7895 | SRCC = 0.7879 ] +INFO:root:Average loss=0.11194706517152297 +INFO:root: +[Valid][27][UTT][ MSE = 0.3173 | LCC = 0.7866 | SRCC = 0.7857 ] +INFO:root:Average loss=0.1084236877086835 +INFO:root: +[Valid][28][UTT][ MSE = 0.3255 | LCC = 0.7850 | SRCC = 0.7828 ] +INFO:root:Average loss=0.11978345039563301 +INFO:root: +[Valid][29][UTT][ MSE = 0.3272 | LCC = 0.7922 | SRCC = 0.7900 ] +INFO:root:Average loss=0.10966202951012513 +INFO:root: +[Valid][30][UTT][ MSE = 0.3308 | LCC = 0.7878 | SRCC = 0.7854 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3203 | LCC = 0.8060 | SRCC = 0.8087 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run0/config.gin b/results_bvcc/w2v2_large_asr_de/layer2/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run0/model.pt b/results_bvcc/w2v2_large_asr_de/layer2/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f1098e5fe9605f246bb8ed75039125a5a4d2fdd2 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:907a35155b0cd2255106972b80576467185408959e2105b14791a20125eba326 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run0/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer2/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..dfad78bbe6b72e8c01bcfcb7ae1ea19b903ce917 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4e8306a9c38fd826cc3b55a101983394050574bee6b09a16d18ea9df8f988fe3 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run0/train.log b/results_bvcc/w2v2_large_asr_de/layer2/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d3c47638a77e3991f973fdb95e36d7917d5de000 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/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=0.548113057628656 +INFO:root: +[Valid][0][UTT][ MSE = 0.4203 | LCC = 0.7344 | SRCC = 0.7401 ] +INFO:root:Average loss=0.34246856298966283 +INFO:root: +[Valid][1][UTT][ MSE = 0.4081 | LCC = 0.7586 | SRCC = 0.7702 ] +INFO:root:Average loss=0.3105777981571662 +INFO:root: +[Valid][2][UTT][ MSE = 0.3422 | LCC = 0.7995 | SRCC = 0.8013 ] +INFO:root:Average loss=0.27119896025993884 +INFO:root: +[Valid][3][UTT][ MSE = 0.3151 | LCC = 0.8085 | SRCC = 0.8135 ] +INFO:root:Average loss=0.24930145133000153 +INFO:root: +[Valid][4][UTT][ MSE = 0.3103 | LCC = 0.8209 | SRCC = 0.8201 ] +INFO:root:Average loss=0.23523401125119284 +INFO:root: +[Valid][5][UTT][ MSE = 0.2822 | LCC = 0.8364 | SRCC = 0.8384 ] +INFO:root:Average loss=0.22160010230846894 +INFO:root: +[Valid][6][UTT][ MSE = 0.2726 | LCC = 0.8336 | SRCC = 0.8323 ] +INFO:root:Average loss=0.21799776282830116 +INFO:root: +[Valid][7][UTT][ MSE = 0.2634 | LCC = 0.8473 | SRCC = 0.8452 ] +INFO:root:Average loss=0.1988833982210893 +INFO:root: +[Valid][8][UTT][ MSE = 0.3050 | LCC = 0.8348 | SRCC = 0.8416 ] +INFO:root:Average loss=0.21393229038669512 +INFO:root: +[Valid][9][UTT][ MSE = 0.2730 | LCC = 0.8427 | SRCC = 0.8445 ] +INFO:root:Average loss=0.1962253999633667 +INFO:root: +[Valid][10][UTT][ MSE = 0.3051 | LCC = 0.8497 | SRCC = 0.8506 ] +INFO:root:Average loss=0.20858129658378088 +INFO:root: +[Valid][11][UTT][ MSE = 0.2791 | LCC = 0.8514 | SRCC = 0.8520 ] +INFO:root:Average loss=0.197235756004468 +INFO:root: +[Valid][12][UTT][ MSE = 0.2708 | LCC = 0.8514 | SRCC = 0.8516 ] +INFO:root:Average loss=0.17608246532006142 +INFO:root: +[Valid][13][UTT][ MSE = 0.2443 | LCC = 0.8518 | SRCC = 0.8531 ] +INFO:root:Average loss=0.15593822672963142 +INFO:root: +[Valid][14][UTT][ MSE = 0.2307 | LCC = 0.8570 | SRCC = 0.8579 ] +INFO:root:Average loss=0.15132799563117516 +INFO:root: +[Valid][15][UTT][ MSE = 0.2396 | LCC = 0.8518 | SRCC = 0.8525 ] +INFO:root:Average loss=0.14045366291434336 +INFO:root: +[Valid][16][UTT][ MSE = 0.2661 | LCC = 0.8606 | SRCC = 0.8627 ] +INFO:root:Average loss=0.14808576964797118 +INFO:root: +[Valid][17][UTT][ MSE = 0.2278 | LCC = 0.8592 | SRCC = 0.8612 ] +INFO:root:Average loss=0.1392933088235366 +INFO:root: +[Valid][18][UTT][ MSE = 0.2694 | LCC = 0.8544 | SRCC = 0.8564 ] +INFO:root:Average loss=0.13159012718078417 +INFO:root: +[Valid][19][UTT][ MSE = 0.3105 | LCC = 0.8587 | SRCC = 0.8595 ] +INFO:root:Average loss=0.13295962900305405 +INFO:root: +[Valid][20][UTT][ MSE = 0.2274 | LCC = 0.8611 | SRCC = 0.8617 ] +INFO:root:Average loss=0.12764376755326223 +INFO:root: +[Valid][21][UTT][ MSE = 0.2335 | LCC = 0.8595 | SRCC = 0.8608 ] +INFO:root:Average loss=0.12635419269402823 +INFO:root: +[Valid][22][UTT][ MSE = 0.2333 | LCC = 0.8606 | SRCC = 0.8601 ] +INFO:root:Average loss=0.11318914277049211 +INFO:root: +[Valid][23][UTT][ MSE = 0.2252 | LCC = 0.8567 | SRCC = 0.8561 ] +INFO:root:Average loss=0.12558926766117415 +INFO:root: +[Valid][24][UTT][ MSE = 0.2350 | LCC = 0.8628 | SRCC = 0.8622 ] +INFO:root:Average loss=0.1026107934422982 +INFO:root: +[Valid][25][UTT][ MSE = 0.2163 | LCC = 0.8624 | SRCC = 0.8624 ] +INFO:root:Average loss=0.11287763437781578 +INFO:root: +[Valid][26][UTT][ MSE = 0.2143 | LCC = 0.8623 | SRCC = 0.8636 ] +INFO:root:Average loss=0.09706527080673438 +INFO:root: +[Valid][27][UTT][ MSE = 0.2265 | LCC = 0.8596 | SRCC = 0.8611 ] +INFO:root:Average loss=0.09635041883358589 +INFO:root: +[Valid][28][UTT][ MSE = 0.2503 | LCC = 0.8557 | SRCC = 0.8594 ] +INFO:root:Average loss=0.1047548873302264 +INFO:root: +[Valid][29][UTT][ MSE = 0.2152 | LCC = 0.8623 | SRCC = 0.8634 ] +INFO:root:Average loss=0.08722311119811657 +INFO:root: +[Valid][30][UTT][ MSE = 0.2307 | LCC = 0.8624 | SRCC = 0.8635 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2721 | LCC = 0.8579 | SRCC = 0.8582 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run1/config.gin b/results_bvcc/w2v2_large_asr_de/layer2/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run1/model.pt b/results_bvcc/w2v2_large_asr_de/layer2/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..275d1be339f3ff96d8e72573e32c89ac5291805a --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0a60b83e334a57285a0a9fb9eeaf5eb5d88496f237879d573c2f085cb383efd5 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run1/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer2/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..09bc15d47ed2a5a33934b30dc3cba39623d01dc2 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:84dbb83b192240e54a6d1c4df1c78ab1c2226b4e78403dbb092be43613cfd13b +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run1/train.log b/results_bvcc/w2v2_large_asr_de/layer2/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..925f87d5aac2a8335dbd10fc688a6a74f999790e --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/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=0.6112437777412243 +INFO:root: +[Valid][0][UTT][ MSE = 0.4003 | LCC = 0.7351 | SRCC = 0.7394 ] +INFO:root:Average loss=0.3375915839121892 +INFO:root: +[Valid][1][UTT][ MSE = 0.3462 | LCC = 0.7737 | SRCC = 0.7761 ] +INFO:root:Average loss=0.29629112226076615 +INFO:root: +[Valid][2][UTT][ MSE = 0.3112 | LCC = 0.7932 | SRCC = 0.7961 ] +INFO:root:Average loss=0.2670996722120505 +INFO:root: +[Valid][3][UTT][ MSE = 0.2999 | LCC = 0.8110 | SRCC = 0.8130 ] +INFO:root:Average loss=0.2563386523188689 +INFO:root: +[Valid][4][UTT][ MSE = 0.2991 | LCC = 0.8294 | SRCC = 0.8262 ] +INFO:root:Average loss=0.22754407960634965 +INFO:root: +[Valid][5][UTT][ MSE = 0.2803 | LCC = 0.8334 | SRCC = 0.8333 ] +INFO:root:Average loss=0.21456225350117072 +INFO:root: +[Valid][6][UTT][ MSE = 0.2578 | LCC = 0.8446 | SRCC = 0.8432 ] +INFO:root:Average loss=0.20503633240094551 +INFO:root: +[Valid][7][UTT][ MSE = 0.2629 | LCC = 0.8412 | SRCC = 0.8412 ] +INFO:root:Average loss=0.19269676965016586 +INFO:root: +[Valid][8][UTT][ MSE = 0.2522 | LCC = 0.8538 | SRCC = 0.8517 ] +INFO:root:Average loss=0.1821637144073462 +INFO:root: +[Valid][9][UTT][ MSE = 0.2397 | LCC = 0.8520 | SRCC = 0.8518 ] +INFO:root:Average loss=0.2051240164691057 +INFO:root: +[Valid][10][UTT][ MSE = 0.2425 | LCC = 0.8502 | SRCC = 0.8516 ] +INFO:root:Average loss=0.17734508502941865 +INFO:root: +[Valid][11][UTT][ MSE = 0.2370 | LCC = 0.8599 | SRCC = 0.8592 ] +INFO:root:Average loss=0.16730584051364508 +INFO:root: +[Valid][12][UTT][ MSE = 0.2452 | LCC = 0.8551 | SRCC = 0.8553 ] +INFO:root:Average loss=0.17508319383248305 +INFO:root: +[Valid][13][UTT][ MSE = 0.2401 | LCC = 0.8477 | SRCC = 0.8478 ] +INFO:root:Average loss=0.15524471264619094 +INFO:root: +[Valid][14][UTT][ MSE = 0.2252 | LCC = 0.8583 | SRCC = 0.8573 ] +INFO:root:Average loss=0.15463698416566238 +INFO:root: +[Valid][15][UTT][ MSE = 0.2526 | LCC = 0.8526 | SRCC = 0.8531 ] +INFO:root:Average loss=0.13953731142175504 +INFO:root: +[Valid][16][UTT][ MSE = 0.2304 | LCC = 0.8617 | SRCC = 0.8615 ] +INFO:root:Average loss=0.13501567183396754 +INFO:root: +[Valid][17][UTT][ MSE = 0.2358 | LCC = 0.8589 | SRCC = 0.8586 ] +INFO:root:Average loss=0.13073231805211458 +INFO:root: +[Valid][18][UTT][ MSE = 0.2880 | LCC = 0.8430 | SRCC = 0.8545 ] +INFO:root:Average loss=0.13554595859769064 +INFO:root: +[Valid][19][UTT][ MSE = 0.2492 | LCC = 0.8555 | SRCC = 0.8569 ] +INFO:root:Average loss=0.14163002772973135 +INFO:root: +[Valid][20][UTT][ MSE = 0.2409 | LCC = 0.8606 | SRCC = 0.8611 ] +INFO:root:Average loss=0.1211335889230936 +INFO:root: +[Valid][21][UTT][ MSE = 0.2446 | LCC = 0.8618 | SRCC = 0.8599 ] +INFO:root:Average loss=0.11841987331326191 +INFO:root: +[Valid][22][UTT][ MSE = 0.2402 | LCC = 0.8624 | SRCC = 0.8622 ] +INFO:root:Average loss=0.1099918538179153 +INFO:root: +[Valid][23][UTT][ MSE = 0.2386 | LCC = 0.8589 | SRCC = 0.8585 ] +INFO:root:Average loss=0.10371735940376918 +INFO:root: +[Valid][24][UTT][ MSE = 0.2223 | LCC = 0.8641 | SRCC = 0.8649 ] +INFO:root:Average loss=0.10878177273731965 +INFO:root: +[Valid][25][UTT][ MSE = 0.2636 | LCC = 0.8625 | SRCC = 0.8621 ] +INFO:root:Average loss=0.0954450133901376 +INFO:root: +[Valid][26][UTT][ MSE = 0.2312 | LCC = 0.8603 | SRCC = 0.8590 ] +INFO:root:Average loss=0.09495519541012935 +INFO:root: +[Valid][27][UTT][ MSE = 0.2166 | LCC = 0.8619 | SRCC = 0.8613 ] +INFO:root:Average loss=0.09316457855777863 +INFO:root: +[Valid][28][UTT][ MSE = 0.2410 | LCC = 0.8590 | SRCC = 0.8588 ] +INFO:root:Average loss=0.08822518969193482 +INFO:root: +[Valid][29][UTT][ MSE = 0.2251 | LCC = 0.8604 | SRCC = 0.8593 ] +INFO:root:Average loss=0.09218468507513022 +INFO:root: +[Valid][30][UTT][ MSE = 0.2658 | LCC = 0.8555 | SRCC = 0.8575 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2223 | LCC = 0.8639 | SRCC = 0.8625 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run2/config.gin b/results_bvcc/w2v2_large_asr_de/layer2/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run2/model.pt b/results_bvcc/w2v2_large_asr_de/layer2/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..123ea2cbd98b4cbf067e82d9257872a9d7b832a6 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8e33a1c2d7fe292599ec24d98741a513ecc0b45130f9544c191794b886842e75 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run2/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer2/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f2c969ec3de60314af2bcc464a2da43a892b2e7f --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2ef4e7c0c5519acb4518841daaf391477a5654b7b3a8978f14a065b7c081baa8 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run2/train.log b/results_bvcc/w2v2_large_asr_de/layer2/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2bf7080921907ff8c4a261d9e4102b5425a65703 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/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=0.5482789098452299 +INFO:root: +[Valid][0][UTT][ MSE = 0.4643 | LCC = 0.7283 | SRCC = 0.7416 ] +INFO:root:Average loss=0.35055500249832106 +INFO:root: +[Valid][1][UTT][ MSE = 0.3755 | LCC = 0.7779 | SRCC = 0.7794 ] +INFO:root:Average loss=0.2966714810866576 +INFO:root: +[Valid][2][UTT][ MSE = 0.3629 | LCC = 0.8019 | SRCC = 0.7974 ] +INFO:root:Average loss=0.27083189422503495 +INFO:root: +[Valid][3][UTT][ MSE = 0.3156 | LCC = 0.8136 | SRCC = 0.8085 ] +INFO:root:Average loss=0.2652506969678096 +INFO:root: +[Valid][4][UTT][ MSE = 0.2905 | LCC = 0.8231 | SRCC = 0.8239 ] +INFO:root:Average loss=0.2526898619074088 +INFO:root: +[Valid][5][UTT][ MSE = 0.3191 | LCC = 0.8364 | SRCC = 0.8331 ] +INFO:root:Average loss=0.227549305328956 +INFO:root: +[Valid][6][UTT][ MSE = 0.2803 | LCC = 0.8332 | SRCC = 0.8322 ] +INFO:root:Average loss=0.2111321414510409 +INFO:root: +[Valid][7][UTT][ MSE = 0.2882 | LCC = 0.8445 | SRCC = 0.8426 ] +INFO:root:Average loss=0.19920112144870636 +INFO:root: +[Valid][8][UTT][ MSE = 0.2842 | LCC = 0.8494 | SRCC = 0.8458 ] +INFO:root:Average loss=0.191711969100512 +INFO:root: +[Valid][9][UTT][ MSE = 0.2621 | LCC = 0.8455 | SRCC = 0.8418 ] +INFO:root:Average loss=0.1987774840150124 +INFO:root: +[Valid][10][UTT][ MSE = 0.2607 | LCC = 0.8494 | SRCC = 0.8466 ] +INFO:root:Average loss=0.1743836770646083 +INFO:root: +[Valid][11][UTT][ MSE = 0.2641 | LCC = 0.8498 | SRCC = 0.8488 ] +INFO:root:Average loss=0.18180253222966805 +INFO:root: +[Valid][12][UTT][ MSE = 0.2523 | LCC = 0.8488 | SRCC = 0.8502 ] +INFO:root:Average loss=0.1704141547282537 +INFO:root: +[Valid][13][UTT][ MSE = 0.2618 | LCC = 0.8535 | SRCC = 0.8534 ] +INFO:root:Average loss=0.15645631851676184 +INFO:root: +[Valid][14][UTT][ MSE = 0.2463 | LCC = 0.8561 | SRCC = 0.8549 ] +INFO:root:Average loss=0.16635118090571502 +INFO:root: +[Valid][15][UTT][ MSE = 0.2367 | LCC = 0.8554 | SRCC = 0.8548 ] +INFO:root:Average loss=0.14473832522829375 +INFO:root: +[Valid][16][UTT][ MSE = 0.2323 | LCC = 0.8588 | SRCC = 0.8580 ] +INFO:root:Average loss=0.14037980378056183 +INFO:root: +[Valid][17][UTT][ MSE = 0.2812 | LCC = 0.8487 | SRCC = 0.8483 ] +INFO:root:Average loss=0.14536816607683134 +INFO:root: +[Valid][18][UTT][ MSE = 0.2759 | LCC = 0.8495 | SRCC = 0.8498 ] +INFO:root:Average loss=0.13427245120207468 +INFO:root: +[Valid][19][UTT][ MSE = 0.2450 | LCC = 0.8552 | SRCC = 0.8552 ] +INFO:root:Average loss=0.12971923920588616 +INFO:root: +[Valid][20][UTT][ MSE = 0.2429 | LCC = 0.8569 | SRCC = 0.8564 ] +INFO:root:Average loss=0.12021593243265763 +INFO:root: +[Valid][21][UTT][ MSE = 0.2564 | LCC = 0.8549 | SRCC = 0.8565 ] +INFO:root:Average loss=0.12786446960690695 +INFO:root: +[Valid][22][UTT][ MSE = 0.2281 | LCC = 0.8613 | SRCC = 0.8610 ] +INFO:root:Average loss=0.11484828510154517 +INFO:root: +[Valid][23][UTT][ MSE = 0.2294 | LCC = 0.8599 | SRCC = 0.8592 ] +INFO:root:Average loss=0.12100163493783046 +INFO:root: +[Valid][24][UTT][ MSE = 0.2601 | LCC = 0.8608 | SRCC = 0.8608 ] +INFO:root:Average loss=0.10454752148152927 +INFO:root: +[Valid][25][UTT][ MSE = 0.2559 | LCC = 0.8574 | SRCC = 0.8581 ] +INFO:root:Average loss=0.12035586837774669 +INFO:root: +[Valid][26][UTT][ MSE = 0.2550 | LCC = 0.8545 | SRCC = 0.8563 ] +INFO:root:Average loss=0.09899546664494735 +INFO:root: +[Valid][27][UTT][ MSE = 0.2318 | LCC = 0.8556 | SRCC = 0.8589 ] +INFO:root:Average loss=0.10469627461563318 +INFO:root: +[Valid][28][UTT][ MSE = 0.2512 | LCC = 0.8538 | SRCC = 0.8572 ] +INFO:root:Average loss=0.09518068083203755 +INFO:root: +[Valid][29][UTT][ MSE = 0.2291 | LCC = 0.8572 | SRCC = 0.8556 ] +INFO:root:Average loss=0.09254194729221173 +INFO:root: +[Valid][30][UTT][ MSE = 0.2268 | LCC = 0.8605 | SRCC = 0.8607 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2358 | LCC = 0.8634 | SRCC = 0.8626 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run3/config.gin b/results_bvcc/w2v2_large_asr_de/layer2/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run3/model.pt b/results_bvcc/w2v2_large_asr_de/layer2/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..21bf30f5b67345858c5d479fed7014baa5e246ca --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7c61da1c01b7bd59b6fea817586153660e7fdbcd8ec6bd4a70b1edfa41996e39 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run3/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer2/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..516d163303870bf5376fbad1e528a5293ae02bd1 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:80978ba565703af31c567614ce401a99365e7ee03bc71359d1db66ab0013891b +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run3/train.log b/results_bvcc/w2v2_large_asr_de/layer2/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..807ff36285340972d42c29230b9b5ade8feceff1 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/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=0.5448490327749497 +INFO:root: +[Valid][0][UTT][ MSE = 0.4218 | LCC = 0.7346 | SRCC = 0.7333 ] +INFO:root:Average loss=0.32487037090154797 +INFO:root: +[Valid][1][UTT][ MSE = 0.3428 | LCC = 0.7870 | SRCC = 0.7867 ] +INFO:root:Average loss=0.2837297748296689 +INFO:root: +[Valid][2][UTT][ MSE = 0.3275 | LCC = 0.8167 | SRCC = 0.8144 ] +INFO:root:Average loss=0.25860144866582674 +INFO:root: +[Valid][3][UTT][ MSE = 0.3322 | LCC = 0.8167 | SRCC = 0.8153 ] +INFO:root:Average loss=0.2519337152823424 +INFO:root: +[Valid][4][UTT][ MSE = 0.2887 | LCC = 0.8280 | SRCC = 0.8275 ] +INFO:root:Average loss=0.23333994958263177 +INFO:root: +[Valid][5][UTT][ MSE = 0.2724 | LCC = 0.8347 | SRCC = 0.8389 ] +INFO:root:Average loss=0.2143934204792365 +INFO:root: +[Valid][6][UTT][ MSE = 0.2973 | LCC = 0.8365 | SRCC = 0.8413 ] +INFO:root:Average loss=0.21769253737651384 +INFO:root: +[Valid][7][UTT][ MSE = 0.2763 | LCC = 0.8540 | SRCC = 0.8505 ] +INFO:root:Average loss=0.20216148146069968 +INFO:root: +[Valid][8][UTT][ MSE = 0.2655 | LCC = 0.8547 | SRCC = 0.8522 ] +INFO:root:Average loss=0.19956443334619203 +INFO:root: +[Valid][9][UTT][ MSE = 0.2578 | LCC = 0.8484 | SRCC = 0.8517 ] +INFO:root:Average loss=0.1848296715089908 +INFO:root: +[Valid][10][UTT][ MSE = 0.2525 | LCC = 0.8545 | SRCC = 0.8540 ] +INFO:root:Average loss=0.17429533084997764 +INFO:root: +[Valid][11][UTT][ MSE = 0.2512 | LCC = 0.8538 | SRCC = 0.8533 ] +INFO:root:Average loss=0.16668966909249625 +INFO:root: +[Valid][12][UTT][ MSE = 0.2417 | LCC = 0.8563 | SRCC = 0.8575 ] +INFO:root:Average loss=0.17004263047606516 +INFO:root: +[Valid][13][UTT][ MSE = 0.2330 | LCC = 0.8561 | SRCC = 0.8582 ] +INFO:root:Average loss=0.1584037926334601 +INFO:root: +[Valid][14][UTT][ MSE = 0.2554 | LCC = 0.8536 | SRCC = 0.8603 ] +INFO:root:Average loss=0.15641571218386674 +INFO:root: +[Valid][15][UTT][ MSE = 0.2563 | LCC = 0.8553 | SRCC = 0.8588 ] +INFO:root:Average loss=0.14314731191366148 +INFO:root: +[Valid][16][UTT][ MSE = 0.2340 | LCC = 0.8633 | SRCC = 0.8624 ] +INFO:root:Average loss=0.13652758500896967 +INFO:root: +[Valid][17][UTT][ MSE = 0.2347 | LCC = 0.8593 | SRCC = 0.8601 ] +INFO:root:Average loss=0.13099603603283563 +INFO:root: +[Valid][18][UTT][ MSE = 0.2313 | LCC = 0.8591 | SRCC = 0.8596 ] +INFO:root:Average loss=0.13969334721183166 +INFO:root: +[Valid][19][UTT][ MSE = 0.2258 | LCC = 0.8572 | SRCC = 0.8593 ] +INFO:root:Average loss=0.1256908549903295 +INFO:root: +[Valid][20][UTT][ MSE = 0.2421 | LCC = 0.8608 | SRCC = 0.8621 ] +INFO:root:Average loss=0.12383253308825004 +INFO:root: +[Valid][21][UTT][ MSE = 0.2470 | LCC = 0.8554 | SRCC = 0.8579 ] +INFO:root:Average loss=0.1149594865930386 +INFO:root: +[Valid][22][UTT][ MSE = 0.2505 | LCC = 0.8636 | SRCC = 0.8638 ] +INFO:root:Average loss=0.11768039296834897 +INFO:root: +[Valid][23][UTT][ MSE = 0.2171 | LCC = 0.8633 | SRCC = 0.8627 ] +INFO:root:Average loss=0.11405898917179841 +INFO:root: +[Valid][24][UTT][ MSE = 0.2295 | LCC = 0.8653 | SRCC = 0.8647 ] +INFO:root:Average loss=0.10979265366227199 +INFO:root: +[Valid][25][UTT][ MSE = 0.2273 | LCC = 0.8649 | SRCC = 0.8650 ] +INFO:root:Average loss=0.10227161502608886 +INFO:root: +[Valid][26][UTT][ MSE = 0.2360 | LCC = 0.8633 | SRCC = 0.8622 ] +INFO:root:Average loss=0.10476467978113736 +INFO:root: +[Valid][27][UTT][ MSE = 0.2661 | LCC = 0.8649 | SRCC = 0.8662 ] +INFO:root:Average loss=0.10508967649478179 +INFO:root: +[Valid][28][UTT][ MSE = 0.2898 | LCC = 0.8572 | SRCC = 0.8609 ] +INFO:root:Average loss=0.09683587000920223 +INFO:root: +[Valid][29][UTT][ MSE = 0.2311 | LCC = 0.8657 | SRCC = 0.8659 ] +INFO:root:Average loss=0.09390103965042493 +INFO:root: +[Valid][30][UTT][ MSE = 0.2954 | LCC = 0.8559 | SRCC = 0.8605 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2253 | LCC = 0.8611 | SRCC = 0.8588 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run4/config.gin b/results_bvcc/w2v2_large_asr_de/layer2/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run4/model.pt b/results_bvcc/w2v2_large_asr_de/layer2/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f71f9c9fe2670ea5e1a16a05da6c13472e785018 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9dd8b8325854c18f0e82e7e4b317f52e183fa6ce13979f66f2228bbaf4425d6f +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run4/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer2/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b46982314d66b9a03585ec6917246c5264a14340 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:630699736a11c060aef6fd0a3e10ad694b56610c8da746a23df99671fbcad032 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer2/run4/train.log b/results_bvcc/w2v2_large_asr_de/layer2/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..48ce0b99115258800742bf931519ed0dc29a16c4 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer2/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=0.5116566922038029 +INFO:root: +[Valid][0][UTT][ MSE = 0.3876 | LCC = 0.7447 | SRCC = 0.7408 ] +INFO:root:Average loss=0.3483809278561519 +INFO:root: +[Valid][1][UTT][ MSE = 0.3400 | LCC = 0.7823 | SRCC = 0.7826 ] +INFO:root:Average loss=0.3181225080520679 +INFO:root: +[Valid][2][UTT][ MSE = 0.3529 | LCC = 0.8009 | SRCC = 0.7994 ] +INFO:root:Average loss=0.2610351557915027 +INFO:root: +[Valid][3][UTT][ MSE = 0.3197 | LCC = 0.8150 | SRCC = 0.8160 ] +INFO:root:Average loss=0.2533854591922882 +INFO:root: +[Valid][4][UTT][ MSE = 0.2806 | LCC = 0.8274 | SRCC = 0.8265 ] +INFO:root:Average loss=0.2263547482016759 +INFO:root: +[Valid][5][UTT][ MSE = 0.2648 | LCC = 0.8356 | SRCC = 0.8345 ] +INFO:root:Average loss=0.21251326111646798 +INFO:root: +[Valid][6][UTT][ MSE = 0.2540 | LCC = 0.8445 | SRCC = 0.8456 ] +INFO:root:Average loss=0.22034921879187608 +INFO:root: +[Valid][7][UTT][ MSE = 0.2711 | LCC = 0.8465 | SRCC = 0.8465 ] +INFO:root:Average loss=0.19705814868211746 +INFO:root: +[Valid][8][UTT][ MSE = 0.3074 | LCC = 0.8547 | SRCC = 0.8536 ] +INFO:root:Average loss=0.19121994307407966 +INFO:root: +[Valid][9][UTT][ MSE = 0.2511 | LCC = 0.8544 | SRCC = 0.8534 ] +INFO:root:Average loss=0.1859973518130107 +INFO:root: +[Valid][10][UTT][ MSE = 0.2655 | LCC = 0.8569 | SRCC = 0.8567 ] +INFO:root:Average loss=0.16995908520542657 +INFO:root: +[Valid][11][UTT][ MSE = 0.2387 | LCC = 0.8573 | SRCC = 0.8610 ] +INFO:root:Average loss=0.17403989132398215 +INFO:root: +[Valid][12][UTT][ MSE = 0.2654 | LCC = 0.8594 | SRCC = 0.8601 ] +INFO:root:Average loss=0.17114968282672074 +INFO:root: +[Valid][13][UTT][ MSE = 0.2270 | LCC = 0.8610 | SRCC = 0.8642 ] +INFO:root:Average loss=0.15591157094026223 +INFO:root: +[Valid][14][UTT][ MSE = 0.2252 | LCC = 0.8591 | SRCC = 0.8615 ] +INFO:root:Average loss=0.15680620742914003 +INFO:root: +[Valid][15][UTT][ MSE = 0.2381 | LCC = 0.8636 | SRCC = 0.8631 ] +INFO:root:Average loss=0.15586914809850547 +INFO:root: +[Valid][16][UTT][ MSE = 0.2250 | LCC = 0.8619 | SRCC = 0.8632 ] +INFO:root:Average loss=0.14794194535949293 +INFO:root: +[Valid][17][UTT][ MSE = 0.2442 | LCC = 0.8698 | SRCC = 0.8689 ] +INFO:root:Average loss=0.13725440290111762 +INFO:root: +[Valid][18][UTT][ MSE = 0.2200 | LCC = 0.8615 | SRCC = 0.8633 ] +INFO:root:Average loss=0.13382499817854318 +INFO:root: +[Valid][19][UTT][ MSE = 0.2334 | LCC = 0.8648 | SRCC = 0.8651 ] +INFO:root:Average loss=0.1343163199340686 +INFO:root: +[Valid][20][UTT][ MSE = 0.2260 | LCC = 0.8651 | SRCC = 0.8648 ] +INFO:root:Average loss=0.12998584954020304 +INFO:root: +[Valid][21][UTT][ MSE = 0.2099 | LCC = 0.8680 | SRCC = 0.8690 ] +INFO:root:Average loss=0.12684337278971305 +INFO:root: +[Valid][22][UTT][ MSE = 0.2193 | LCC = 0.8664 | SRCC = 0.8676 ] +INFO:root:Average loss=0.11329671253378575 +INFO:root: +[Valid][23][UTT][ MSE = 0.2758 | LCC = 0.8642 | SRCC = 0.8662 ] +INFO:root:Average loss=0.11241409918054557 +INFO:root: +[Valid][24][UTT][ MSE = 0.2145 | LCC = 0.8688 | SRCC = 0.8689 ] +INFO:root:Average loss=0.10632362078206661 +INFO:root: +[Valid][25][UTT][ MSE = 0.2469 | LCC = 0.8683 | SRCC = 0.8707 ] +INFO:root:Average loss=0.10806590834489235 +INFO:root: +[Valid][26][UTT][ MSE = 0.2253 | LCC = 0.8651 | SRCC = 0.8651 ] +INFO:root:Average loss=0.09851715102409705 +INFO:root: +[Valid][27][UTT][ MSE = 0.2133 | LCC = 0.8683 | SRCC = 0.8689 ] +INFO:root:Average loss=0.0917419141683823 +INFO:root: +[Valid][28][UTT][ MSE = 0.2147 | LCC = 0.8701 | SRCC = 0.8707 ] +INFO:root:Average loss=0.09004453784571244 +INFO:root: +[Valid][29][UTT][ MSE = 0.2205 | LCC = 0.8690 | SRCC = 0.8682 ] +INFO:root:Average loss=0.09134517247096086 +INFO:root: +[Valid][30][UTT][ MSE = 0.2605 | LCC = 0.8681 | SRCC = 0.8704 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2514 | LCC = 0.8642 | SRCC = 0.8634 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run0/config.gin b/results_bvcc/w2v2_large_asr_de/layer20/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run0/model.pt b/results_bvcc/w2v2_large_asr_de/layer20/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1ea0433106dba9bc914daedd36a11bf3c3bddce8 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c33591a312588f5f7235d2868b4b7345656953fe67651e5991cf662c8cc09f5f +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run0/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer20/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0f7c7870958356ecb9f9656c2d0e6e5d0ce1a881 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8846011f3a18defebaede5e90900c2d0e561749fc80e37a92b24ef6f5c3323cf +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run0/train.log b/results_bvcc/w2v2_large_asr_de/layer20/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ae92a4ce56ca18f2f3aba21a79a06ef1dd0d8185 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/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=0.5683864534665377 +INFO:root: +[Valid][0][UTT][ MSE = 0.5021 | LCC = 0.6898 | SRCC = 0.6890 ] +INFO:root:Average loss=0.423326245867289 +INFO:root: +[Valid][1][UTT][ MSE = 0.4474 | LCC = 0.7442 | SRCC = 0.7442 ] +INFO:root:Average loss=0.3813898716217432 +INFO:root: +[Valid][2][UTT][ MSE = 0.3860 | LCC = 0.7600 | SRCC = 0.7617 ] +INFO:root:Average loss=0.35968766514307415 +INFO:root: +[Valid][3][UTT][ MSE = 0.4517 | LCC = 0.7624 | SRCC = 0.7628 ] +INFO:root:Average loss=0.33716041843096417 +INFO:root: +[Valid][4][UTT][ MSE = 0.4039 | LCC = 0.7721 | SRCC = 0.7709 ] +INFO:root:Average loss=0.30175758057679886 +INFO:root: +[Valid][5][UTT][ MSE = 0.3600 | LCC = 0.7711 | SRCC = 0.7676 ] +INFO:root:Average loss=0.30850065575960356 +INFO:root: +[Valid][6][UTT][ MSE = 0.4026 | LCC = 0.7718 | SRCC = 0.7711 ] +INFO:root:Average loss=0.2916584381690392 +INFO:root: +[Valid][7][UTT][ MSE = 0.3567 | LCC = 0.7825 | SRCC = 0.7817 ] +INFO:root:Average loss=0.2749557865735812 +INFO:root: +[Valid][8][UTT][ MSE = 0.3492 | LCC = 0.7813 | SRCC = 0.7808 ] +INFO:root:Average loss=0.25837496629892254 +INFO:root: +[Valid][9][UTT][ MSE = 0.3582 | LCC = 0.7713 | SRCC = 0.7710 ] +INFO:root:Average loss=0.26359968727979904 +INFO:root: +[Valid][10][UTT][ MSE = 0.4262 | LCC = 0.7829 | SRCC = 0.7840 ] +INFO:root:Average loss=0.2656311081388058 +INFO:root: +[Valid][11][UTT][ MSE = 0.3520 | LCC = 0.7798 | SRCC = 0.7786 ] +INFO:root:Average loss=0.24452504286399254 +INFO:root: +[Valid][12][UTT][ MSE = 0.3631 | LCC = 0.7833 | SRCC = 0.7839 ] +INFO:root:Average loss=0.26724999321576876 +INFO:root: +[Valid][13][UTT][ MSE = 0.3510 | LCC = 0.7851 | SRCC = 0.7839 ] +INFO:root:Average loss=0.2518748314334796 +INFO:root: +[Valid][14][UTT][ MSE = 0.3355 | LCC = 0.7891 | SRCC = 0.7872 ] +INFO:root:Average loss=0.222800819728619 +INFO:root: +[Valid][15][UTT][ MSE = 0.3803 | LCC = 0.7827 | SRCC = 0.7824 ] +INFO:root:Average loss=0.21595465124417573 +INFO:root: +[Valid][16][UTT][ MSE = 0.3627 | LCC = 0.7793 | SRCC = 0.7788 ] +INFO:root:Average loss=0.22538860772664732 +INFO:root: +[Valid][17][UTT][ MSE = 0.3456 | LCC = 0.7813 | SRCC = 0.7824 ] +INFO:root:Average loss=0.2198225511954381 +INFO:root: +[Valid][18][UTT][ MSE = 0.3463 | LCC = 0.7859 | SRCC = 0.7865 ] +INFO:root:Average loss=0.19279495951456901 +INFO:root: +[Valid][19][UTT][ MSE = 0.3367 | LCC = 0.7812 | SRCC = 0.7804 ] +INFO:root:Average loss=0.20333971560765535 +INFO:root: +[Valid][20][UTT][ MSE = 0.4165 | LCC = 0.7789 | SRCC = 0.7830 ] +INFO:root:Average loss=0.21543654474692467 +INFO:root: +[Valid][21][UTT][ MSE = 0.3686 | LCC = 0.7851 | SRCC = 0.7864 ] +INFO:root:Average loss=0.18810375961355674 +INFO:root: +[Valid][22][UTT][ MSE = 0.3377 | LCC = 0.7820 | SRCC = 0.7813 ] +INFO:root:Average loss=0.17111399311285752 +INFO:root: +[Valid][23][UTT][ MSE = 0.3497 | LCC = 0.7842 | SRCC = 0.7835 ] +INFO:root:Average loss=0.1780255731099691 +INFO:root: +[Valid][24][UTT][ MSE = 0.3533 | LCC = 0.7801 | SRCC = 0.7808 ] +INFO:root:Average loss=0.18571152194188192 +INFO:root: +[Valid][25][UTT][ MSE = 0.3312 | LCC = 0.7804 | SRCC = 0.7814 ] +INFO:root:Average loss=0.1672484393303211 +INFO:root: +[Valid][26][UTT][ MSE = 0.4327 | LCC = 0.7841 | SRCC = 0.7864 ] +INFO:root:Average loss=0.16604980205496153 +INFO:root: +[Valid][27][UTT][ MSE = 0.3811 | LCC = 0.7737 | SRCC = 0.7740 ] +INFO:root:Average loss=0.15708938069068468 +INFO:root: +[Valid][28][UTT][ MSE = 0.3675 | LCC = 0.7851 | SRCC = 0.7872 ] +INFO:root:Average loss=0.14725997251195785 +INFO:root: +[Valid][29][UTT][ MSE = 0.3379 | LCC = 0.7776 | SRCC = 0.7776 ] +INFO:root:Average loss=0.1507171748731381 +INFO:root: +[Valid][30][UTT][ MSE = 0.3797 | LCC = 0.7811 | SRCC = 0.7818 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3407 | LCC = 0.7887 | SRCC = 0.7900 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run1/config.gin b/results_bvcc/w2v2_large_asr_de/layer20/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run1/model.pt b/results_bvcc/w2v2_large_asr_de/layer20/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..77c2e3531c54438ea9688dc0a4f1ba1a5b97d24f --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:841e5d3537a55af83b5ac8262aa4d187654a1032f4c419d3effee3d49becd879 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run1/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer20/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..3c7d27a07cb3e1b5cfc977b5cd3a8559412aa6de --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:91a017beb70144acacf87dc0374bf462ad1b0bd56be0d32e6ba8f0b328cec48c +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run1/train.log b/results_bvcc/w2v2_large_asr_de/layer20/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6a133fe8c0c819623d81c69d6b30dc5096b8508f --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/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=0.5541676351657281 +INFO:root: +[Valid][0][UTT][ MSE = 0.4906 | LCC = 0.6876 | SRCC = 0.6876 ] +INFO:root:Average loss=0.4146558974797909 +INFO:root: +[Valid][1][UTT][ MSE = 0.4187 | LCC = 0.7390 | SRCC = 0.7395 ] +INFO:root:Average loss=0.3688035806020101 +INFO:root: +[Valid][2][UTT][ MSE = 0.3846 | LCC = 0.7566 | SRCC = 0.7566 ] +INFO:root:Average loss=0.3405746336166675 +INFO:root: +[Valid][3][UTT][ MSE = 0.4208 | LCC = 0.7647 | SRCC = 0.7660 ] +INFO:root:Average loss=0.33557097155314225 +INFO:root: +[Valid][4][UTT][ MSE = 0.3695 | LCC = 0.7699 | SRCC = 0.7692 ] +INFO:root:Average loss=0.2999085678886145 +INFO:root: +[Valid][5][UTT][ MSE = 0.3693 | LCC = 0.7660 | SRCC = 0.7654 ] +INFO:root:Average loss=0.30256646489485717 +INFO:root: +[Valid][6][UTT][ MSE = 0.3505 | LCC = 0.7831 | SRCC = 0.7821 ] +INFO:root:Average loss=0.2817976029637532 +INFO:root: +[Valid][7][UTT][ MSE = 0.3706 | LCC = 0.7800 | SRCC = 0.7819 ] +INFO:root:Average loss=0.26987659510893697 +INFO:root: +[Valid][8][UTT][ MSE = 0.3429 | LCC = 0.7755 | SRCC = 0.7737 ] +INFO:root:Average loss=0.2557452556032401 +INFO:root: +[Valid][9][UTT][ MSE = 0.3480 | LCC = 0.7784 | SRCC = 0.7783 ] +INFO:root:Average loss=0.25810061662625045 +INFO:root: +[Valid][10][UTT][ MSE = 0.3368 | LCC = 0.7821 | SRCC = 0.7820 ] +INFO:root:Average loss=0.24652921637663475 +INFO:root: +[Valid][11][UTT][ MSE = 0.3525 | LCC = 0.7863 | SRCC = 0.7865 ] +INFO:root:Average loss=0.2526751309633255 +INFO:root: +[Valid][12][UTT][ MSE = 0.3243 | LCC = 0.7854 | SRCC = 0.7837 ] +INFO:root:Average loss=0.22913659784274223 +INFO:root: +[Valid][13][UTT][ MSE = 0.3313 | LCC = 0.7893 | SRCC = 0.7887 ] +INFO:root:Average loss=0.21953123655074683 +INFO:root: +[Valid][14][UTT][ MSE = 0.3718 | LCC = 0.7604 | SRCC = 0.7594 ] +INFO:root:Average loss=0.21134107063213983 +INFO:root: +[Valid][15][UTT][ MSE = 0.3528 | LCC = 0.7877 | SRCC = 0.7863 ] +INFO:root:Average loss=0.205657503925837 +INFO:root: +[Valid][16][UTT][ MSE = 0.3936 | LCC = 0.7827 | SRCC = 0.7848 ] +INFO:root:Average loss=0.2034198874846483 +INFO:root: +[Valid][17][UTT][ MSE = 0.3624 | LCC = 0.7798 | SRCC = 0.7817 ] +INFO:root:Average loss=0.20850020398696265 +INFO:root: +[Valid][18][UTT][ MSE = 0.3273 | LCC = 0.7834 | SRCC = 0.7833 ] +INFO:root:Average loss=0.1810304538752788 +INFO:root: +[Valid][19][UTT][ MSE = 0.3403 | LCC = 0.7844 | SRCC = 0.7843 ] +INFO:root:Average loss=0.18830585078551218 +INFO:root: +[Valid][20][UTT][ MSE = 0.3468 | LCC = 0.7822 | SRCC = 0.7839 ] +INFO:root:Average loss=0.18320850569468278 +INFO:root: +[Valid][21][UTT][ MSE = 0.3451 | LCC = 0.7759 | SRCC = 0.7765 ] +INFO:root:Average loss=0.1948315155429718 +INFO:root: +[Valid][22][UTT][ MSE = 0.3564 | LCC = 0.7795 | SRCC = 0.7783 ] +INFO:root:Average loss=0.18885007481544447 +INFO:root: +[Valid][23][UTT][ MSE = 0.3278 | LCC = 0.7868 | SRCC = 0.7849 ] +INFO:root:Average loss=0.18032923684670374 +INFO:root: +[Valid][24][UTT][ MSE = 0.3605 | LCC = 0.7753 | SRCC = 0.7734 ] +INFO:root:Average loss=0.17498889412635413 +INFO:root: +[Valid][25][UTT][ MSE = 0.3467 | LCC = 0.7746 | SRCC = 0.7741 ] +INFO:root:Average loss=0.16165667695876879 +INFO:root: +[Valid][26][UTT][ MSE = 0.3325 | LCC = 0.7881 | SRCC = 0.7884 ] +INFO:root:Average loss=0.1645113604191022 +INFO:root: +[Valid][27][UTT][ MSE = 0.3751 | LCC = 0.7827 | SRCC = 0.7837 ] +INFO:root:Average loss=0.15958445404584592 +INFO:root: +[Valid][28][UTT][ MSE = 0.3187 | LCC = 0.7879 | SRCC = 0.7872 ] +INFO:root:Average loss=0.14812324243860367 +INFO:root: +[Valid][29][UTT][ MSE = 0.3300 | LCC = 0.7826 | SRCC = 0.7822 ] +INFO:root:Average loss=0.140303568675732 +INFO:root: +[Valid][30][UTT][ MSE = 0.3488 | LCC = 0.7755 | SRCC = 0.7769 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3639 | LCC = 0.7783 | SRCC = 0.7788 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run2/config.gin b/results_bvcc/w2v2_large_asr_de/layer20/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run2/model.pt b/results_bvcc/w2v2_large_asr_de/layer20/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..477c319600e43f72385d613cc2f93ed7b307efa6 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a69589f61913e254bf595d30b693f45a39419eeec49006695d23fd3dda5e050e +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run2/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer20/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..408164061d9cdcc7aefcb24be1852d2a27e46d40 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:15797de50b13bf957d2c418494634e19e7e185894779b1dde9e7646ed31b0afb +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run2/train.log b/results_bvcc/w2v2_large_asr_de/layer20/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..74db21144c14664a9be53dc06807072c4429c09e --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/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=0.5515873997639387 +INFO:root: +[Valid][0][UTT][ MSE = 0.4725 | LCC = 0.6903 | SRCC = 0.6936 ] +INFO:root:Average loss=0.4313161701728136 +INFO:root: +[Valid][1][UTT][ MSE = 0.4310 | LCC = 0.7342 | SRCC = 0.7359 ] +INFO:root:Average loss=0.37136984979495025 +INFO:root: +[Valid][2][UTT][ MSE = 0.3798 | LCC = 0.7531 | SRCC = 0.7539 ] +INFO:root:Average loss=0.3458013351146991 +INFO:root: +[Valid][3][UTT][ MSE = 0.4544 | LCC = 0.7594 | SRCC = 0.7616 ] +INFO:root:Average loss=0.3425026788161351 +INFO:root: +[Valid][4][UTT][ MSE = 0.4007 | LCC = 0.7700 | SRCC = 0.7707 ] +INFO:root:Average loss=0.3210572783763592 +INFO:root: +[Valid][5][UTT][ MSE = 0.3908 | LCC = 0.7731 | SRCC = 0.7739 ] +INFO:root:Average loss=0.3153138498847301 +INFO:root: +[Valid][6][UTT][ MSE = 0.3678 | LCC = 0.7777 | SRCC = 0.7769 ] +INFO:root:Average loss=0.3005429964799147 +INFO:root: +[Valid][7][UTT][ MSE = 0.3745 | LCC = 0.7706 | SRCC = 0.7702 ] +INFO:root:Average loss=0.28351867084319776 +INFO:root: +[Valid][8][UTT][ MSE = 0.3583 | LCC = 0.7705 | SRCC = 0.7698 ] +INFO:root:Average loss=0.2680609390521661 +INFO:root: +[Valid][9][UTT][ MSE = 0.3594 | LCC = 0.7701 | SRCC = 0.7681 ] +INFO:root:Average loss=0.25856799746935183 +INFO:root: +[Valid][10][UTT][ MSE = 0.3400 | LCC = 0.7870 | SRCC = 0.7861 ] +INFO:root:Average loss=0.2648068895706764 +INFO:root: +[Valid][11][UTT][ MSE = 0.4014 | LCC = 0.7794 | SRCC = 0.7796 ] +INFO:root:Average loss=0.2615197728841733 +INFO:root: +[Valid][12][UTT][ MSE = 0.3438 | LCC = 0.7862 | SRCC = 0.7852 ] +INFO:root:Average loss=0.25173969242053157 +INFO:root: +[Valid][13][UTT][ MSE = 0.3381 | LCC = 0.7876 | SRCC = 0.7870 ] +INFO:root:Average loss=0.231049905029627 +INFO:root: +[Valid][14][UTT][ MSE = 0.3233 | LCC = 0.7862 | SRCC = 0.7841 ] +INFO:root:Average loss=0.223273893006337 +INFO:root: +[Valid][15][UTT][ MSE = 0.3717 | LCC = 0.7716 | SRCC = 0.7735 ] +INFO:root:Average loss=0.21714731210317367 +INFO:root: +[Valid][16][UTT][ MSE = 0.3943 | LCC = 0.7793 | SRCC = 0.7804 ] +INFO:root:Average loss=0.20055236438146004 +INFO:root: +[Valid][17][UTT][ MSE = 0.3409 | LCC = 0.7818 | SRCC = 0.7813 ] +INFO:root:Average loss=0.2047856589540457 +INFO:root: +[Valid][18][UTT][ MSE = 0.3345 | LCC = 0.7833 | SRCC = 0.7819 ] +INFO:root:Average loss=0.21016321904384172 +INFO:root: +[Valid][19][UTT][ MSE = 0.4205 | LCC = 0.7689 | SRCC = 0.7717 ] +INFO:root:Average loss=0.18487905462582907 +INFO:root: +[Valid][20][UTT][ MSE = 0.3277 | LCC = 0.7840 | SRCC = 0.7819 ] +INFO:root:Average loss=0.18618680412570635 +INFO:root: +[Valid][21][UTT][ MSE = 0.3379 | LCC = 0.7936 | SRCC = 0.7924 ] +INFO:root:Average loss=0.1829277295141648 +INFO:root: +[Valid][22][UTT][ MSE = 0.3567 | LCC = 0.7706 | SRCC = 0.7685 ] +INFO:root:Average loss=0.18770042510750967 +INFO:root: +[Valid][23][UTT][ MSE = 0.3715 | LCC = 0.7775 | SRCC = 0.7757 ] +INFO:root:Average loss=0.16954501307545564 +INFO:root: +[Valid][24][UTT][ MSE = 0.3650 | LCC = 0.7765 | SRCC = 0.7772 ] +INFO:root:Average loss=0.16863412954486334 +INFO:root: +[Valid][25][UTT][ MSE = 0.3287 | LCC = 0.7842 | SRCC = 0.7839 ] +INFO:root:Average loss=0.16011087663280657 +INFO:root: +[Valid][26][UTT][ MSE = 0.3694 | LCC = 0.7738 | SRCC = 0.7742 ] +INFO:root:Average loss=0.1591313065817723 +INFO:root: +[Valid][27][UTT][ MSE = 0.3352 | LCC = 0.7829 | SRCC = 0.7808 ] +INFO:root:Average loss=0.1665274428251462 +INFO:root: +[Valid][28][UTT][ MSE = 0.3482 | LCC = 0.7774 | SRCC = 0.7750 ] +INFO:root:Average loss=0.16743082744188797 +INFO:root: +[Valid][29][UTT][ MSE = 0.3470 | LCC = 0.7809 | SRCC = 0.7783 ] +INFO:root:Average loss=0.15942291160806632 +INFO:root: +[Valid][30][UTT][ MSE = 0.3595 | LCC = 0.7800 | SRCC = 0.7794 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3368 | LCC = 0.7823 | SRCC = 0.7774 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run3/config.gin b/results_bvcc/w2v2_large_asr_de/layer20/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run3/model.pt b/results_bvcc/w2v2_large_asr_de/layer20/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d0bb486d5f8b2b3089b99b03b12a2535e80c1674 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:45ea28c2918c4b9e831e2c9384582b5f5d390085d2863188d159174f391a987b +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run3/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer20/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..eed0d71987de0b7c6f042b8b254ca449bc725af8 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:811585926c2ea9908e8fb37ef48ddf3092c9dc5959c3256b299ef901287e0214 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run3/train.log b/results_bvcc/w2v2_large_asr_de/layer20/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9ec5b8e37d7f1c59eb7270b4578b1864ddb82b0f --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/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=0.5872592062522204 +INFO:root: +[Valid][0][UTT][ MSE = 0.4770 | LCC = 0.6784 | SRCC = 0.6794 ] +INFO:root:Average loss=0.41934598925021976 +INFO:root: +[Valid][1][UTT][ MSE = 0.4356 | LCC = 0.7307 | SRCC = 0.7334 ] +INFO:root:Average loss=0.3704422586239301 +INFO:root: +[Valid][2][UTT][ MSE = 0.3904 | LCC = 0.7494 | SRCC = 0.7500 ] +INFO:root:Average loss=0.38580377170672786 +INFO:root: +[Valid][3][UTT][ MSE = 0.4651 | LCC = 0.7629 | SRCC = 0.7645 ] +INFO:root:Average loss=0.32755785569166523 +INFO:root: +[Valid][4][UTT][ MSE = 0.3503 | LCC = 0.7745 | SRCC = 0.7754 ] +INFO:root:Average loss=0.3151144392979451 +INFO:root: +[Valid][5][UTT][ MSE = 0.3600 | LCC = 0.7776 | SRCC = 0.7781 ] +INFO:root:Average loss=0.3119034060300925 +INFO:root: +[Valid][6][UTT][ MSE = 0.3636 | LCC = 0.7797 | SRCC = 0.7808 ] +INFO:root:Average loss=0.28812577288884383 +INFO:root: +[Valid][7][UTT][ MSE = 0.3732 | LCC = 0.7747 | SRCC = 0.7760 ] +INFO:root:Average loss=0.2806721241810383 +INFO:root: +[Valid][8][UTT][ MSE = 0.3629 | LCC = 0.7859 | SRCC = 0.7853 ] +INFO:root:Average loss=0.27063492513619936 +INFO:root: +[Valid][9][UTT][ MSE = 0.3780 | LCC = 0.7690 | SRCC = 0.7681 ] +INFO:root:Average loss=0.26613675439969087 +INFO:root: +[Valid][10][UTT][ MSE = 0.4122 | LCC = 0.7838 | SRCC = 0.7853 ] +INFO:root:Average loss=0.2439862908079074 +INFO:root: +[Valid][11][UTT][ MSE = 0.3494 | LCC = 0.7753 | SRCC = 0.7742 ] +INFO:root:Average loss=0.25421227495639753 +INFO:root: +[Valid][12][UTT][ MSE = 0.3401 | LCC = 0.7874 | SRCC = 0.7855 ] +INFO:root:Average loss=0.2281128955193055 +INFO:root: +[Valid][13][UTT][ MSE = 0.3450 | LCC = 0.7787 | SRCC = 0.7774 ] +INFO:root:Average loss=0.23399756046441886 +INFO:root: +[Valid][14][UTT][ MSE = 0.3415 | LCC = 0.7909 | SRCC = 0.7898 ] +INFO:root:Average loss=0.25211275770114017 +INFO:root: +[Valid][15][UTT][ MSE = 0.3720 | LCC = 0.7829 | SRCC = 0.7830 ] +INFO:root:Average loss=0.22387508016366225 +INFO:root: +[Valid][16][UTT][ MSE = 0.3170 | LCC = 0.7901 | SRCC = 0.7882 ] +INFO:root:Average loss=0.21548241395980883 +INFO:root: +[Valid][17][UTT][ MSE = 0.3304 | LCC = 0.7866 | SRCC = 0.7841 ] +INFO:root:Average loss=0.20527617308573845 +INFO:root: +[Valid][18][UTT][ MSE = 0.3171 | LCC = 0.7924 | SRCC = 0.7899 ] +INFO:root:Average loss=0.2064129821000955 +INFO:root: +[Valid][19][UTT][ MSE = 0.3196 | LCC = 0.7902 | SRCC = 0.7891 ] +INFO:root:Average loss=0.19210461297860512 +INFO:root: +[Valid][20][UTT][ MSE = 0.3406 | LCC = 0.7904 | SRCC = 0.7907 ] +INFO:root:Average loss=0.2002338782334939 +INFO:root: +[Valid][21][UTT][ MSE = 0.3260 | LCC = 0.7881 | SRCC = 0.7879 ] +INFO:root:Average loss=0.18546423698082948 +INFO:root: +[Valid][22][UTT][ MSE = 0.3115 | LCC = 0.7912 | SRCC = 0.7897 ] +INFO:root:Average loss=0.17719071138745698 +INFO:root: +[Valid][23][UTT][ MSE = 0.3252 | LCC = 0.7897 | SRCC = 0.7877 ] +INFO:root:Average loss=0.20052328428778893 +INFO:root: +[Valid][24][UTT][ MSE = 0.3236 | LCC = 0.7871 | SRCC = 0.7867 ] +INFO:root:Average loss=0.1662503746457589 +INFO:root: +[Valid][25][UTT][ MSE = 0.3161 | LCC = 0.7876 | SRCC = 0.7864 ] +INFO:root:Average loss=0.16517594122351745 +INFO:root: +[Valid][26][UTT][ MSE = 0.3168 | LCC = 0.7906 | SRCC = 0.7906 ] +INFO:root:Average loss=0.1621734702434295 +INFO:root: +[Valid][27][UTT][ MSE = 0.3455 | LCC = 0.7785 | SRCC = 0.7802 ] +INFO:root:Average loss=0.1553742304826394 +INFO:root: +[Valid][28][UTT][ MSE = 0.3227 | LCC = 0.7848 | SRCC = 0.7853 ] +INFO:root:Average loss=0.157403105057967 +INFO:root: +[Valid][29][UTT][ MSE = 0.3240 | LCC = 0.7852 | SRCC = 0.7853 ] +INFO:root:Average loss=0.14923795245778868 +INFO:root: +[Valid][30][UTT][ MSE = 0.3294 | LCC = 0.7851 | SRCC = 0.7848 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3355 | LCC = 0.7827 | SRCC = 0.7802 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run4/config.gin b/results_bvcc/w2v2_large_asr_de/layer20/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run4/model.pt b/results_bvcc/w2v2_large_asr_de/layer20/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6252521edce0da80fbdb03332089df2d5b2ab0dc --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1257e0009ab4930fc1c1f86f8d910931f0f0352908ef28cbbf680bcd64bc9222 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run4/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer20/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6e05fb1a23e935e2bccded39830c07510538a8ea --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1441220c931d37d0392bc994465bbe9e97e1cdbb139d3111c7bb5413314ad268 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer20/run4/train.log b/results_bvcc/w2v2_large_asr_de/layer20/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f80b9dd583129c0d13f765f32f46492bc99636c8 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer20/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=0.5724242432759359 +INFO:root: +[Valid][0][UTT][ MSE = 0.4607 | LCC = 0.6822 | SRCC = 0.6820 ] +INFO:root:Average loss=0.44734807083239925 +INFO:root: +[Valid][1][UTT][ MSE = 0.4978 | LCC = 0.7216 | SRCC = 0.7259 ] +INFO:root:Average loss=0.3892896670179489 +INFO:root: +[Valid][2][UTT][ MSE = 0.3900 | LCC = 0.7529 | SRCC = 0.7543 ] +INFO:root:Average loss=0.35825058072805405 +INFO:root: +[Valid][3][UTT][ MSE = 0.3876 | LCC = 0.7609 | SRCC = 0.7605 ] +INFO:root:Average loss=0.3266534763268935 +INFO:root: +[Valid][4][UTT][ MSE = 0.3655 | LCC = 0.7737 | SRCC = 0.7729 ] +INFO:root:Average loss=0.32422913649143315 +INFO:root: +[Valid][5][UTT][ MSE = 0.3872 | LCC = 0.7811 | SRCC = 0.7811 ] +INFO:root:Average loss=0.3078129965907488 +INFO:root: +[Valid][6][UTT][ MSE = 0.3625 | LCC = 0.7817 | SRCC = 0.7808 ] +INFO:root:Average loss=0.2883992980305965 +INFO:root: +[Valid][7][UTT][ MSE = 0.3511 | LCC = 0.7815 | SRCC = 0.7803 ] +INFO:root:Average loss=0.2804428923588533 +INFO:root: +[Valid][8][UTT][ MSE = 0.3714 | LCC = 0.7881 | SRCC = 0.7856 ] +INFO:root:Average loss=0.27943608527764296 +INFO:root: +[Valid][9][UTT][ MSE = 0.3807 | LCC = 0.7824 | SRCC = 0.7803 ] +INFO:root:Average loss=0.2669598339841916 +INFO:root: +[Valid][10][UTT][ MSE = 0.3832 | LCC = 0.7922 | SRCC = 0.7904 ] +INFO:root:Average loss=0.2555019670189955 +INFO:root: +[Valid][11][UTT][ MSE = 0.3462 | LCC = 0.7821 | SRCC = 0.7796 ] +INFO:root:Average loss=0.23950779189666113 +INFO:root: +[Valid][12][UTT][ MSE = 0.3735 | LCC = 0.7876 | SRCC = 0.7852 ] +INFO:root:Average loss=0.24281750351954728 +INFO:root: +[Valid][13][UTT][ MSE = 0.3354 | LCC = 0.7907 | SRCC = 0.7899 ] +INFO:root:Average loss=0.2366249450506308 +INFO:root: +[Valid][14][UTT][ MSE = 0.3678 | LCC = 0.7852 | SRCC = 0.7832 ] +INFO:root:Average loss=0.22602146176191476 +INFO:root: +[Valid][15][UTT][ MSE = 0.3587 | LCC = 0.7830 | SRCC = 0.7809 ] +INFO:root:Average loss=0.21014790294262078 +INFO:root: +[Valid][16][UTT][ MSE = 0.3331 | LCC = 0.7879 | SRCC = 0.7870 ] +INFO:root:Average loss=0.22574606327674326 +INFO:root: +[Valid][17][UTT][ MSE = 0.3205 | LCC = 0.7946 | SRCC = 0.7933 ] +INFO:root:Average loss=0.2283030948959864 +INFO:root: +[Valid][18][UTT][ MSE = 0.3597 | LCC = 0.7864 | SRCC = 0.7898 ] +INFO:root:Average loss=0.22691160469101027 +INFO:root: +[Valid][19][UTT][ MSE = 0.3382 | LCC = 0.7921 | SRCC = 0.7927 ] +INFO:root:Average loss=0.18740512039034796 +INFO:root: +[Valid][20][UTT][ MSE = 0.3451 | LCC = 0.7891 | SRCC = 0.7884 ] +INFO:root:Average loss=0.20240340668421525 +INFO:root: +[Valid][21][UTT][ MSE = 0.3624 | LCC = 0.7837 | SRCC = 0.7848 ] +INFO:root:Average loss=0.1783308156598837 +INFO:root: +[Valid][22][UTT][ MSE = 0.3326 | LCC = 0.7888 | SRCC = 0.7877 ] +INFO:root:Average loss=0.17346766333167368 +INFO:root: +[Valid][23][UTT][ MSE = 0.3357 | LCC = 0.7913 | SRCC = 0.7908 ] +INFO:root:Average loss=0.1616421841467038 +INFO:root: +[Valid][24][UTT][ MSE = 0.3422 | LCC = 0.7967 | SRCC = 0.7957 ] +INFO:root:Average loss=0.17094266920899734 +INFO:root: +[Valid][25][UTT][ MSE = 0.3445 | LCC = 0.7968 | SRCC = 0.7971 ] +INFO:root:Average loss=0.1723877753202732 +INFO:root: +[Valid][26][UTT][ MSE = 0.3031 | LCC = 0.8013 | SRCC = 0.8004 ] +INFO:root:Average loss=0.16098496785912758 +INFO:root: +[Valid][27][UTT][ MSE = 0.3207 | LCC = 0.7917 | SRCC = 0.7911 ] +INFO:root:Average loss=0.17517973425296637 +INFO:root: +[Valid][28][UTT][ MSE = 0.3108 | LCC = 0.7945 | SRCC = 0.7934 ] +INFO:root:Average loss=0.14782263376773933 +INFO:root: +[Valid][29][UTT][ MSE = 0.3589 | LCC = 0.7784 | SRCC = 0.7781 ] +INFO:root:Average loss=0.1468600924962606 +INFO:root: +[Valid][30][UTT][ MSE = 0.3351 | LCC = 0.7948 | SRCC = 0.7960 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3319 | LCC = 0.7846 | SRCC = 0.7818 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run0/config.gin b/results_bvcc/w2v2_large_asr_de/layer22/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run0/model.pt b/results_bvcc/w2v2_large_asr_de/layer22/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9d41fb706a86d21a574bb95a8671247c77254043 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cb66f0d87b6430749f50b3ffe17ed13813bc5b1c2768f851d9b45337be61dff3 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run0/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer22/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9fd07302a56e4660b66818facd398b52a16bfa66 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9c86023ca2191521cf5a24078e3c3fba30eb923482c3db96c850c0905dae80c7 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run0/train.log b/results_bvcc/w2v2_large_asr_de/layer22/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a45c766207bdc3903159715dbd48424c162b739d --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/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=0.5792290682976062 +INFO:root: +[Valid][0][UTT][ MSE = 0.4905 | LCC = 0.6565 | SRCC = 0.6550 ] +INFO:root:Average loss=0.4584319782562745 +INFO:root: +[Valid][1][UTT][ MSE = 0.4607 | LCC = 0.7001 | SRCC = 0.6996 ] +INFO:root:Average loss=0.42125041477191144 +INFO:root: +[Valid][2][UTT][ MSE = 0.4660 | LCC = 0.7150 | SRCC = 0.7196 ] +INFO:root:Average loss=0.3945807323623926 +INFO:root: +[Valid][3][UTT][ MSE = 0.4307 | LCC = 0.7284 | SRCC = 0.7302 ] +INFO:root:Average loss=0.3669429526497156 +INFO:root: +[Valid][4][UTT][ MSE = 0.4078 | LCC = 0.7427 | SRCC = 0.7424 ] +INFO:root:Average loss=0.37442680964103114 +INFO:root: +[Valid][5][UTT][ MSE = 0.3834 | LCC = 0.7435 | SRCC = 0.7421 ] +INFO:root:Average loss=0.35649848270874757 +INFO:root: +[Valid][6][UTT][ MSE = 0.3981 | LCC = 0.7471 | SRCC = 0.7466 ] +INFO:root:Average loss=0.3336093960664211 +INFO:root: +[Valid][7][UTT][ MSE = 0.4439 | LCC = 0.7479 | SRCC = 0.7486 ] +INFO:root:Average loss=0.3343349353243143 +INFO:root: +[Valid][8][UTT][ MSE = 0.3881 | LCC = 0.7562 | SRCC = 0.7550 ] +INFO:root:Average loss=0.30672918126369136 +INFO:root: +[Valid][9][UTT][ MSE = 0.3857 | LCC = 0.7608 | SRCC = 0.7588 ] +INFO:root:Average loss=0.29054739364446736 +INFO:root: +[Valid][10][UTT][ MSE = 0.4016 | LCC = 0.7556 | SRCC = 0.7537 ] +INFO:root:Average loss=0.2878680813771028 +INFO:root: +[Valid][11][UTT][ MSE = 0.3587 | LCC = 0.7612 | SRCC = 0.7582 ] +INFO:root:Average loss=0.2896010249088972 +INFO:root: +[Valid][12][UTT][ MSE = 0.3585 | LCC = 0.7632 | SRCC = 0.7601 ] +INFO:root:Average loss=0.29137011483693737 +INFO:root: +[Valid][13][UTT][ MSE = 0.4287 | LCC = 0.7604 | SRCC = 0.7597 ] +INFO:root:Average loss=0.2862882102147127 +INFO:root: +[Valid][14][UTT][ MSE = 0.3692 | LCC = 0.7671 | SRCC = 0.7664 ] +INFO:root:Average loss=0.28091970028785557 +INFO:root: +[Valid][15][UTT][ MSE = 0.3652 | LCC = 0.7660 | SRCC = 0.7642 ] +INFO:root:Average loss=0.2673268887477043 +INFO:root: +[Valid][16][UTT][ MSE = 0.3431 | LCC = 0.7721 | SRCC = 0.7690 ] +INFO:root:Average loss=0.2448500070052269 +INFO:root: +[Valid][17][UTT][ MSE = 0.3618 | LCC = 0.7662 | SRCC = 0.7663 ] +INFO:root:Average loss=0.2611735858596288 +INFO:root: +[Valid][18][UTT][ MSE = 0.3539 | LCC = 0.7672 | SRCC = 0.7655 ] +INFO:root:Average loss=0.2396852204050773 +INFO:root: +[Valid][19][UTT][ MSE = 0.3511 | LCC = 0.7666 | SRCC = 0.7642 ] +INFO:root:Average loss=0.24582500068040994 +INFO:root: +[Valid][20][UTT][ MSE = 0.3550 | LCC = 0.7653 | SRCC = 0.7637 ] +INFO:root:Average loss=0.24060687518272644 +INFO:root: +[Valid][21][UTT][ MSE = 0.3463 | LCC = 0.7691 | SRCC = 0.7676 ] +INFO:root:Average loss=0.22196515458516586 +INFO:root: +[Valid][22][UTT][ MSE = 0.3349 | LCC = 0.7733 | SRCC = 0.7716 ] +INFO:root:Average loss=0.21024397359444544 +INFO:root: +[Valid][23][UTT][ MSE = 0.3633 | LCC = 0.7711 | SRCC = 0.7685 ] +INFO:root:Average loss=0.2186942152105845 +INFO:root: +[Valid][24][UTT][ MSE = 0.3594 | LCC = 0.7664 | SRCC = 0.7646 ] +INFO:root:Average loss=0.2446222051213949 +INFO:root: +[Valid][25][UTT][ MSE = 0.3909 | LCC = 0.7677 | SRCC = 0.7684 ] +INFO:root:Average loss=0.21219192875119355 +INFO:root: +[Valid][26][UTT][ MSE = 0.3587 | LCC = 0.7615 | SRCC = 0.7584 ] +INFO:root:Average loss=0.19598958373833925 +INFO:root: +[Valid][27][UTT][ MSE = 0.3644 | LCC = 0.7641 | SRCC = 0.7628 ] +INFO:root:Average loss=0.20797854088819945 +INFO:root: +[Valid][28][UTT][ MSE = 0.4136 | LCC = 0.7695 | SRCC = 0.7721 ] +INFO:root:Average loss=0.2132497019110582 +INFO:root: +[Valid][29][UTT][ MSE = 0.3467 | LCC = 0.7711 | SRCC = 0.7683 ] +INFO:root:Average loss=0.19824822208820245 +INFO:root: +[Valid][30][UTT][ MSE = 0.3469 | LCC = 0.7687 | SRCC = 0.7688 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3559 | LCC = 0.7681 | SRCC = 0.7665 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run1/config.gin b/results_bvcc/w2v2_large_asr_de/layer22/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run1/model.pt b/results_bvcc/w2v2_large_asr_de/layer22/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..34c0ae9772888d49d1f640086eaaeab64112974b --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6cdf31a4ff54500947f88adfe145b79c32733f9129a638567f8344bbf93da87 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run1/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer22/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..bd3ee46e9f9c12ccf16d3e34d8054eac4ad8f971 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:357e98ca6fb6a6340520ab3181b055b44835de06511aec19a6a5be6eee7acbff +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run1/train.log b/results_bvcc/w2v2_large_asr_de/layer22/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..51c1e00e0b0b32dcf3c506c4a22c1de1c8345c75 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/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=0.5853146199996655 +INFO:root: +[Valid][0][UTT][ MSE = 0.4928 | LCC = 0.6621 | SRCC = 0.6632 ] +INFO:root:Average loss=0.44606865904270077 +INFO:root: +[Valid][1][UTT][ MSE = 0.4894 | LCC = 0.7064 | SRCC = 0.7113 ] +INFO:root:Average loss=0.40875081488719356 +INFO:root: +[Valid][2][UTT][ MSE = 0.4439 | LCC = 0.7201 | SRCC = 0.7195 ] +INFO:root:Average loss=0.39612890760868025 +INFO:root: +[Valid][3][UTT][ MSE = 0.4180 | LCC = 0.7376 | SRCC = 0.7380 ] +INFO:root:Average loss=0.3645577663794542 +INFO:root: +[Valid][4][UTT][ MSE = 0.4293 | LCC = 0.7422 | SRCC = 0.7411 ] +INFO:root:Average loss=0.3432926730467723 +INFO:root: +[Valid][5][UTT][ MSE = 0.4126 | LCC = 0.7377 | SRCC = 0.7364 ] +INFO:root:Average loss=0.3595488491730812 +INFO:root: +[Valid][6][UTT][ MSE = 0.3998 | LCC = 0.7574 | SRCC = 0.7550 ] +INFO:root:Average loss=0.3343846091093161 +INFO:root: +[Valid][7][UTT][ MSE = 0.4645 | LCC = 0.7439 | SRCC = 0.7473 ] +INFO:root:Average loss=0.32019886661034364 +INFO:root: +[Valid][8][UTT][ MSE = 0.3691 | LCC = 0.7739 | SRCC = 0.7719 ] +INFO:root:Average loss=0.30519142823341566 +INFO:root: +[Valid][9][UTT][ MSE = 0.3721 | LCC = 0.7621 | SRCC = 0.7604 ] +INFO:root:Average loss=0.3031376004219055 +INFO:root: +[Valid][10][UTT][ MSE = 0.3739 | LCC = 0.7759 | SRCC = 0.7752 ] +INFO:root:Average loss=0.29748608133731747 +INFO:root: +[Valid][11][UTT][ MSE = 0.3715 | LCC = 0.7798 | SRCC = 0.7794 ] +INFO:root:Average loss=0.27357009530831605 +INFO:root: +[Valid][12][UTT][ MSE = 0.4592 | LCC = 0.7631 | SRCC = 0.7669 ] +INFO:root:Average loss=0.30581162831722164 +INFO:root: +[Valid][13][UTT][ MSE = 0.3500 | LCC = 0.7764 | SRCC = 0.7755 ] +INFO:root:Average loss=0.26309832281027085 +INFO:root: +[Valid][14][UTT][ MSE = 0.3869 | LCC = 0.7641 | SRCC = 0.7631 ] +INFO:root:Average loss=0.2675344787347011 +INFO:root: +[Valid][15][UTT][ MSE = 0.3601 | LCC = 0.7671 | SRCC = 0.7672 ] +INFO:root:Average loss=0.2625078988763002 +INFO:root: +[Valid][16][UTT][ MSE = 0.3927 | LCC = 0.7697 | SRCC = 0.7694 ] +INFO:root:Average loss=0.2927737688789001 +INFO:root: +[Valid][17][UTT][ MSE = 0.3551 | LCC = 0.7751 | SRCC = 0.7732 ] +INFO:root:Average loss=0.24485217206753218 +INFO:root: +[Valid][18][UTT][ MSE = 0.3619 | LCC = 0.7714 | SRCC = 0.7688 ] +INFO:root:Average loss=0.25146780277674013 +INFO:root: +[Valid][19][UTT][ MSE = 0.3757 | LCC = 0.7766 | SRCC = 0.7761 ] +INFO:root:Average loss=0.2380062026473192 +INFO:root: +[Valid][20][UTT][ MSE = 0.3544 | LCC = 0.7702 | SRCC = 0.7693 ] +INFO:root:Average loss=0.23726026904888642 +INFO:root: +[Valid][21][UTT][ MSE = 0.3420 | LCC = 0.7765 | SRCC = 0.7769 ] +INFO:root:Average loss=0.22351605827227616 +INFO:root: +[Valid][22][UTT][ MSE = 0.4479 | LCC = 0.7531 | SRCC = 0.7552 ] +INFO:root:Average loss=0.22745815989298698 +INFO:root: +[Valid][23][UTT][ MSE = 0.3580 | LCC = 0.7702 | SRCC = 0.7691 ] +INFO:root:Average loss=0.21453384023446304 +INFO:root: +[Valid][24][UTT][ MSE = 0.3774 | LCC = 0.7749 | SRCC = 0.7731 ] +INFO:root:Average loss=0.23458622319576067 +INFO:root: +[Valid][25][UTT][ MSE = 0.3611 | LCC = 0.7804 | SRCC = 0.7783 ] +INFO:root:Average loss=0.19845185161401063 +INFO:root: +[Valid][26][UTT][ MSE = 0.3454 | LCC = 0.7753 | SRCC = 0.7741 ] +INFO:root:Average loss=0.19813320613824403 +INFO:root: +[Valid][27][UTT][ MSE = 0.3609 | LCC = 0.7758 | SRCC = 0.7766 ] +INFO:root:Average loss=0.19208286014887002 +INFO:root: +[Valid][28][UTT][ MSE = 0.3455 | LCC = 0.7783 | SRCC = 0.7781 ] +INFO:root:Average loss=0.19563940482643935 +INFO:root: +[Valid][29][UTT][ MSE = 0.3575 | LCC = 0.7632 | SRCC = 0.7641 ] +INFO:root:Average loss=0.19130707684999856 +INFO:root: +[Valid][30][UTT][ MSE = 0.3429 | LCC = 0.7770 | SRCC = 0.7780 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3891 | LCC = 0.7608 | SRCC = 0.7667 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run2/config.gin b/results_bvcc/w2v2_large_asr_de/layer22/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run2/model.pt b/results_bvcc/w2v2_large_asr_de/layer22/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5db32d0559f4f12800efb03d4a5ca798c711ee9a --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2c988342c5b7cc36156873a027865d258e77cbeed7c25445b43c9ad79ee4d0e4 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run2/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer22/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0d221ea49a2d255f132159f30c917345614b808e --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a9a6d4366939317e0246cc97bb5c62e880386ba073983128ecddc304c057f337 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run2/train.log b/results_bvcc/w2v2_large_asr_de/layer22/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..669b831a6f36804351a8bd4393cf784dc563881a --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/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=0.5456754527030847 +INFO:root: +[Valid][0][UTT][ MSE = 0.4789 | LCC = 0.6805 | SRCC = 0.6801 ] +INFO:root:Average loss=0.44970252345769834 +INFO:root: +[Valid][1][UTT][ MSE = 0.5230 | LCC = 0.7039 | SRCC = 0.7086 ] +INFO:root:Average loss=0.4104968657096227 +INFO:root: +[Valid][2][UTT][ MSE = 0.5525 | LCC = 0.7118 | SRCC = 0.7249 ] +INFO:root:Average loss=0.38849637791132313 +INFO:root: +[Valid][3][UTT][ MSE = 0.4021 | LCC = 0.7378 | SRCC = 0.7368 ] +INFO:root:Average loss=0.37756811808317137 +INFO:root: +[Valid][4][UTT][ MSE = 0.4026 | LCC = 0.7448 | SRCC = 0.7451 ] +INFO:root:Average loss=0.35721465945243835 +INFO:root: +[Valid][5][UTT][ MSE = 0.3963 | LCC = 0.7489 | SRCC = 0.7471 ] +INFO:root:Average loss=0.35570508318069655 +INFO:root: +[Valid][6][UTT][ MSE = 0.3762 | LCC = 0.7582 | SRCC = 0.7584 ] +INFO:root:Average loss=0.32077016624120563 +INFO:root: +[Valid][7][UTT][ MSE = 0.4880 | LCC = 0.7455 | SRCC = 0.7521 ] +INFO:root:Average loss=0.3137362782771771 +INFO:root: +[Valid][8][UTT][ MSE = 0.5065 | LCC = 0.7464 | SRCC = 0.7565 ] +INFO:root:Average loss=0.3060911637850297 +INFO:root: +[Valid][9][UTT][ MSE = 0.3937 | LCC = 0.7581 | SRCC = 0.7574 ] +INFO:root:Average loss=0.27998596009535665 +INFO:root: +[Valid][10][UTT][ MSE = 0.3705 | LCC = 0.7618 | SRCC = 0.7604 ] +INFO:root:Average loss=0.28198423026463926 +INFO:root: +[Valid][11][UTT][ MSE = 0.3944 | LCC = 0.7647 | SRCC = 0.7646 ] +INFO:root:Average loss=0.27239271979301405 +INFO:root: +[Valid][12][UTT][ MSE = 0.5818 | LCC = 0.7523 | SRCC = 0.7611 ] +INFO:root:Average loss=0.28155004042081344 +INFO:root: +[Valid][13][UTT][ MSE = 0.3976 | LCC = 0.7638 | SRCC = 0.7625 ] +INFO:root:Average loss=0.2662438731162976 +INFO:root: +[Valid][14][UTT][ MSE = 0.3570 | LCC = 0.7733 | SRCC = 0.7714 ] +INFO:root:Average loss=0.24810907244682312 +INFO:root: +[Valid][15][UTT][ MSE = 0.3698 | LCC = 0.7694 | SRCC = 0.7698 ] +INFO:root:Average loss=0.26899700802870286 +INFO:root: +[Valid][16][UTT][ MSE = 0.3543 | LCC = 0.7753 | SRCC = 0.7760 ] +INFO:root:Average loss=0.2495295911645278 +INFO:root: +[Valid][17][UTT][ MSE = 0.3508 | LCC = 0.7754 | SRCC = 0.7746 ] +INFO:root:Average loss=0.2378157526254654 +INFO:root: +[Valid][18][UTT][ MSE = 0.3559 | LCC = 0.7722 | SRCC = 0.7715 ] +INFO:root:Average loss=0.23781044380022928 +INFO:root: +[Valid][19][UTT][ MSE = 0.3557 | LCC = 0.7783 | SRCC = 0.7791 ] +INFO:root:Average loss=0.22678483048310646 +INFO:root: +[Valid][20][UTT][ MSE = 0.3673 | LCC = 0.7712 | SRCC = 0.7717 ] +INFO:root:Average loss=0.21948133122462493 +INFO:root: +[Valid][21][UTT][ MSE = 0.3505 | LCC = 0.7780 | SRCC = 0.7784 ] +INFO:root:Average loss=0.2177375153853343 +INFO:root: +[Valid][22][UTT][ MSE = 0.4044 | LCC = 0.7713 | SRCC = 0.7720 ] +INFO:root:Average loss=0.21665954551635644 +INFO:root: +[Valid][23][UTT][ MSE = 0.3789 | LCC = 0.7652 | SRCC = 0.7631 ] +INFO:root:Average loss=0.19266627738491082 +INFO:root: +[Valid][24][UTT][ MSE = 0.4246 | LCC = 0.7687 | SRCC = 0.7709 ] +INFO:root:Average loss=0.20767171107805693 +INFO:root: +[Valid][25][UTT][ MSE = 0.3561 | LCC = 0.7641 | SRCC = 0.7632 ] +INFO:root:Average loss=0.192017110876548 +INFO:root: +[Valid][26][UTT][ MSE = 0.3581 | LCC = 0.7765 | SRCC = 0.7748 ] +INFO:root:Average loss=0.19899392892152834 +INFO:root: +[Valid][27][UTT][ MSE = 0.3616 | LCC = 0.7625 | SRCC = 0.7614 ] +INFO:root:Average loss=0.18404542463712203 +INFO:root: +[Valid][28][UTT][ MSE = 0.3716 | LCC = 0.7794 | SRCC = 0.7790 ] +INFO:root:Average loss=0.17905035156470078 +INFO:root: +[Valid][29][UTT][ MSE = 0.3423 | LCC = 0.7757 | SRCC = 0.7763 ] +INFO:root:Average loss=0.17955231045683226 +INFO:root: +[Valid][30][UTT][ MSE = 0.4737 | LCC = 0.7653 | SRCC = 0.7709 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3624 | LCC = 0.7685 | SRCC = 0.7653 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run3/config.gin b/results_bvcc/w2v2_large_asr_de/layer22/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run3/model.pt b/results_bvcc/w2v2_large_asr_de/layer22/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..355ef2420734755980f30b42b9f5e94b0f3c354f --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:54bfadcf330888ae614b097d242409b59e365acda3d2899ba7e5dc9c5b7d1843 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run3/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer22/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..94ba62c2c915f6447d56e2d9368e0c03b9f994f4 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:818ac4cbfc751b5111a98d67e58ca5fa3deb1f93ccc19abf1f67f99f34637edc +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run3/train.log b/results_bvcc/w2v2_large_asr_de/layer22/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d9a4d437ab3e1add15d07cf86d82481cad9f28cc --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/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=0.5510232314849511 +INFO:root: +[Valid][0][UTT][ MSE = 0.5950 | LCC = 0.6621 | SRCC = 0.6687 ] +INFO:root:Average loss=0.44584523217800337 +INFO:root: +[Valid][1][UTT][ MSE = 0.4429 | LCC = 0.7059 | SRCC = 0.7066 ] +INFO:root:Average loss=0.41765538278298503 +INFO:root: +[Valid][2][UTT][ MSE = 0.4310 | LCC = 0.7284 | SRCC = 0.7275 ] +INFO:root:Average loss=0.39741105949267364 +INFO:root: +[Valid][3][UTT][ MSE = 0.5300 | LCC = 0.7346 | SRCC = 0.7347 ] +INFO:root:Average loss=0.3609136117574496 +INFO:root: +[Valid][4][UTT][ MSE = 0.3992 | LCC = 0.7362 | SRCC = 0.7367 ] +INFO:root:Average loss=0.3406305725757892 +INFO:root: +[Valid][5][UTT][ MSE = 0.4034 | LCC = 0.7450 | SRCC = 0.7468 ] +INFO:root:Average loss=0.3265012856095265 +INFO:root: +[Valid][6][UTT][ MSE = 0.4294 | LCC = 0.7573 | SRCC = 0.7551 ] +INFO:root:Average loss=0.35044773324177814 +INFO:root: +[Valid][7][UTT][ MSE = 0.5108 | LCC = 0.7548 | SRCC = 0.7540 ] +INFO:root:Average loss=0.3335664526392252 +INFO:root: +[Valid][8][UTT][ MSE = 0.3839 | LCC = 0.7584 | SRCC = 0.7570 ] +INFO:root:Average loss=0.30366115730542403 +INFO:root: +[Valid][9][UTT][ MSE = 0.3902 | LCC = 0.7568 | SRCC = 0.7575 ] +INFO:root:Average loss=0.29366346888053113 +INFO:root: +[Valid][10][UTT][ MSE = 0.6005 | LCC = 0.7674 | SRCC = 0.7684 ] +INFO:root:Average loss=0.2865483413139979 +INFO:root: +[Valid][11][UTT][ MSE = 0.3399 | LCC = 0.7742 | SRCC = 0.7747 ] +INFO:root:Average loss=0.2872888921545102 +INFO:root: +[Valid][12][UTT][ MSE = 0.3748 | LCC = 0.7679 | SRCC = 0.7684 ] +INFO:root:Average loss=0.2758646536713991 +INFO:root: +[Valid][13][UTT][ MSE = 0.4738 | LCC = 0.7612 | SRCC = 0.7607 ] +INFO:root:Average loss=0.26391406452808625 +INFO:root: +[Valid][14][UTT][ MSE = 0.3740 | LCC = 0.7627 | SRCC = 0.7622 ] +INFO:root:Average loss=0.2705120725127367 +INFO:root: +[Valid][15][UTT][ MSE = 0.3762 | LCC = 0.7745 | SRCC = 0.7726 ] +INFO:root:Average loss=0.24836584830131286 +INFO:root: +[Valid][16][UTT][ MSE = 0.3929 | LCC = 0.7668 | SRCC = 0.7655 ] +INFO:root:Average loss=0.2473823775847753 +INFO:root: +[Valid][17][UTT][ MSE = 0.3553 | LCC = 0.7682 | SRCC = 0.7662 ] +INFO:root:Average loss=0.25715669492880505 +INFO:root: +[Valid][18][UTT][ MSE = 0.3644 | LCC = 0.7745 | SRCC = 0.7728 ] +INFO:root:Average loss=0.23262663242908624 +INFO:root: +[Valid][19][UTT][ MSE = 0.3374 | LCC = 0.7782 | SRCC = 0.7772 ] +INFO:root:Average loss=0.24765249895743835 +INFO:root: +[Valid][20][UTT][ MSE = 0.4031 | LCC = 0.7673 | SRCC = 0.7679 ] +INFO:root:Average loss=0.22521591434876123 +INFO:root: +[Valid][21][UTT][ MSE = 0.3719 | LCC = 0.7760 | SRCC = 0.7762 ] +INFO:root:Average loss=0.22240510907692787 +INFO:root: +[Valid][22][UTT][ MSE = 0.3893 | LCC = 0.7842 | SRCC = 0.7845 ] +INFO:root:Average loss=0.2480293189485868 +INFO:root: +[Valid][23][UTT][ MSE = 0.3411 | LCC = 0.7773 | SRCC = 0.7784 ] +INFO:root:Average loss=0.22777515248610422 +INFO:root: +[Valid][24][UTT][ MSE = 0.3551 | LCC = 0.7785 | SRCC = 0.7785 ] +INFO:root:Average loss=0.21895074997192773 +INFO:root: +[Valid][25][UTT][ MSE = 0.3813 | LCC = 0.7694 | SRCC = 0.7703 ] +INFO:root:Average loss=0.200142891647724 +INFO:root: +[Valid][26][UTT][ MSE = 0.3523 | LCC = 0.7800 | SRCC = 0.7799 ] +INFO:root:Average loss=0.1959588861045165 +INFO:root: +[Valid][27][UTT][ MSE = 0.3808 | LCC = 0.7766 | SRCC = 0.7753 ] +INFO:root:Average loss=0.19289402377146941 +INFO:root: +[Valid][28][UTT][ MSE = 0.3737 | LCC = 0.7693 | SRCC = 0.7687 ] +INFO:root:Average loss=0.18623356960522822 +INFO:root: +[Valid][29][UTT][ MSE = 0.3385 | LCC = 0.7809 | SRCC = 0.7812 ] +INFO:root:Average loss=0.19244918399132216 +INFO:root: +[Valid][30][UTT][ MSE = 0.3522 | LCC = 0.7824 | SRCC = 0.7823 ] +INFO:root: +[Test][31][UTT][ MSE = 0.4454 | LCC = 0.7698 | SRCC = 0.7762 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run4/config.gin b/results_bvcc/w2v2_large_asr_de/layer22/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/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 diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run4/model.pt b/results_bvcc/w2v2_large_asr_de/layer22/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5f5bb5d76706f26811acd8c533a6d2da6839d757 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af286ce4dbf77ff103a5a8c65e6fe4bcbec1c21bef86ab41a63901a4a6a5ca74 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run4/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer22/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..527695cce58a1661c65235a2805e2de5ac5dc9ed --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd59d8c5baf8f53c758e706f601a5ec8a961d65325aaee84ec853b24db781ab2 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer22/run4/train.log b/results_bvcc/w2v2_large_asr_de/layer22/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9cd9e73da8b5578dfbc60f4e2c90ec660d897d35 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer22/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=0.6243808842622317 +INFO:root: +[Valid][0][UTT][ MSE = 0.5420 | LCC = 0.6531 | SRCC = 0.6526 ] +INFO:root:Average loss=0.45909213102780855 +INFO:root: +[Valid][1][UTT][ MSE = 0.5195 | LCC = 0.6938 | SRCC = 0.6989 ] +INFO:root:Average loss=0.4252075615983743 +INFO:root: +[Valid][2][UTT][ MSE = 0.5106 | LCC = 0.7112 | SRCC = 0.7139 ] +INFO:root:Average loss=0.3934865861366957 +INFO:root: +[Valid][3][UTT][ MSE = 0.4318 | LCC = 0.7324 | SRCC = 0.7338 ] +INFO:root:Average loss=0.36113608246430373 +INFO:root: +[Valid][4][UTT][ MSE = 0.4070 | LCC = 0.7494 | SRCC = 0.7477 ] +INFO:root:Average loss=0.3517799853132321 +INFO:root: +[Valid][5][UTT][ MSE = 0.3887 | LCC = 0.7542 | SRCC = 0.7532 ] +INFO:root:Average loss=0.36408549891068387 +INFO:root: +[Valid][6][UTT][ MSE = 0.4067 | LCC = 0.7574 | SRCC = 0.7581 ] +INFO:root:Average loss=0.3276085301469534 +INFO:root: +[Valid][7][UTT][ MSE = 0.5248 | LCC = 0.7491 | SRCC = 0.7547 ] +INFO:root:Average loss=0.3422556391511208 +INFO:root: +[Valid][8][UTT][ MSE = 0.3712 | LCC = 0.7616 | SRCC = 0.7619 ] +INFO:root:Average loss=0.3202034781376521 +INFO:root: +[Valid][9][UTT][ MSE = 0.4800 | LCC = 0.7642 | SRCC = 0.7690 ] +INFO:root:Average loss=0.29792583733797073 +INFO:root: +[Valid][10][UTT][ MSE = 0.3721 | LCC = 0.7626 | SRCC = 0.7622 ] +INFO:root:Average loss=0.2914535750945409 +INFO:root: +[Valid][11][UTT][ MSE = 0.3708 | LCC = 0.7706 | SRCC = 0.7712 ] +INFO:root:Average loss=0.2918268829966203 +INFO:root: +[Valid][12][UTT][ MSE = 0.3772 | LCC = 0.7672 | SRCC = 0.7668 ] +INFO:root:Average loss=0.27223777637267726 +INFO:root: +[Valid][13][UTT][ MSE = 0.3608 | LCC = 0.7666 | SRCC = 0.7647 ] +INFO:root:Average loss=0.2644152332765934 +INFO:root: +[Valid][14][UTT][ MSE = 0.4145 | LCC = 0.7655 | SRCC = 0.7647 ] +INFO:root:Average loss=0.2649791601758737 +INFO:root: +[Valid][15][UTT][ MSE = 0.3584 | LCC = 0.7728 | SRCC = 0.7736 ] +INFO:root:Average loss=0.2483828142285347 +INFO:root: +[Valid][16][UTT][ MSE = 0.3794 | LCC = 0.7638 | SRCC = 0.7636 ] +INFO:root:Average loss=0.2548061581567312 +INFO:root: +[Valid][17][UTT][ MSE = 0.3629 | LCC = 0.7785 | SRCC = 0.7800 ] +INFO:root:Average loss=0.24788548759160897 +INFO:root: +[Valid][18][UTT][ MSE = 0.3625 | LCC = 0.7785 | SRCC = 0.7784 ] +INFO:root:Average loss=0.23751192387097922 +INFO:root: +[Valid][19][UTT][ MSE = 0.3624 | LCC = 0.7743 | SRCC = 0.7746 ] +INFO:root:Average loss=0.2294952284831267 +INFO:root: +[Valid][20][UTT][ MSE = 0.3509 | LCC = 0.7785 | SRCC = 0.7770 ] +INFO:root:Average loss=0.2142828221504505 +INFO:root: +[Valid][21][UTT][ MSE = 0.3378 | LCC = 0.7786 | SRCC = 0.7779 ] +INFO:root:Average loss=0.2157575256931476 +INFO:root: +[Valid][22][UTT][ MSE = 0.3362 | LCC = 0.7795 | SRCC = 0.7771 ] +INFO:root:Average loss=0.23282648642093706 +INFO:root: +[Valid][23][UTT][ MSE = 0.3716 | LCC = 0.7753 | SRCC = 0.7741 ] +INFO:root:Average loss=0.21186279008785883 +INFO:root: +[Valid][24][UTT][ MSE = 0.3422 | LCC = 0.7729 | SRCC = 0.7717 ] +INFO:root:Average loss=0.21016967334808448 +INFO:root: +[Valid][25][UTT][ MSE = 0.3492 | LCC = 0.7688 | SRCC = 0.7660 ] +INFO:root:Average loss=0.21919134221015832 +INFO:root: +[Valid][26][UTT][ MSE = 0.3518 | LCC = 0.7693 | SRCC = 0.7687 ] +INFO:root:Average loss=0.20105618429489625 +INFO:root: +[Valid][27][UTT][ MSE = 0.3856 | LCC = 0.7666 | SRCC = 0.7675 ] +INFO:root:Average loss=0.19819808923281157 +INFO:root: +[Valid][28][UTT][ MSE = 0.4072 | LCC = 0.7636 | SRCC = 0.7638 ] +INFO:root:Average loss=0.193429924452152 +INFO:root: +[Valid][29][UTT][ MSE = 0.3572 | LCC = 0.7666 | SRCC = 0.7648 ] +INFO:root:Average loss=0.1852926557453779 +INFO:root: +[Valid][30][UTT][ MSE = 0.3469 | LCC = 0.7675 | SRCC = 0.7658 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3662 | LCC = 0.7626 | SRCC = 0.7656 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run0/config.gin b/results_bvcc/w2v2_large_asr_de/layer4/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run0/model.pt b/results_bvcc/w2v2_large_asr_de/layer4/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..64a811fe67c46a37f45d538f9fc7c24d90852f11 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4beed4f906e17f5a38f2feadd95ba4de166cb07633f65b57583692bb0f584d59 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run0/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer4/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6d065c29334592dd9d354b9f69fa54f360e3c087 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dc8a0361cc7c97ecdba942d21e5c5d12434d8d12f971b02fb435d7892bc1c333 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run0/train.log b/results_bvcc/w2v2_large_asr_de/layer4/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9320f7e35ab69b9ef73762336d469d7853a1f14d --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/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=0.5580237389375002 +INFO:root: +[Valid][0][UTT][ MSE = 0.3777 | LCC = 0.7666 | SRCC = 0.7709 ] +INFO:root:Average loss=0.3056341113570409 +INFO:root: +[Valid][1][UTT][ MSE = 0.3158 | LCC = 0.8149 | SRCC = 0.8140 ] +INFO:root:Average loss=0.2463596136524127 +INFO:root: +[Valid][2][UTT][ MSE = 0.2827 | LCC = 0.8366 | SRCC = 0.8362 ] +INFO:root:Average loss=0.2355591994829667 +INFO:root: +[Valid][3][UTT][ MSE = 0.2675 | LCC = 0.8404 | SRCC = 0.8393 ] +INFO:root:Average loss=0.22042755706188005 +INFO:root: +[Valid][4][UTT][ MSE = 0.2748 | LCC = 0.8492 | SRCC = 0.8474 ] +INFO:root:Average loss=0.20112144268858126 +INFO:root: +[Valid][5][UTT][ MSE = 0.2451 | LCC = 0.8499 | SRCC = 0.8517 ] +INFO:root:Average loss=0.19269712164233893 +INFO:root: +[Valid][6][UTT][ MSE = 0.2637 | LCC = 0.8506 | SRCC = 0.8494 ] +INFO:root:Average loss=0.19174267227451006 +INFO:root: +[Valid][7][UTT][ MSE = 0.2701 | LCC = 0.8504 | SRCC = 0.8526 ] +INFO:root:Average loss=0.16778331555617162 +INFO:root: +[Valid][8][UTT][ MSE = 0.2511 | LCC = 0.8583 | SRCC = 0.8580 ] +INFO:root:Average loss=0.16867167875170708 +INFO:root: +[Valid][9][UTT][ MSE = 0.2480 | LCC = 0.8563 | SRCC = 0.8589 ] +INFO:root:Average loss=0.14720906374546197 +INFO:root: +[Valid][10][UTT][ MSE = 0.2477 | LCC = 0.8567 | SRCC = 0.8549 ] +INFO:root:Average loss=0.15013035682913584 +INFO:root: +[Valid][11][UTT][ MSE = 0.2410 | LCC = 0.8563 | SRCC = 0.8552 ] +INFO:root:Average loss=0.14142377082353982 +INFO:root: +[Valid][12][UTT][ MSE = 0.2311 | LCC = 0.8583 | SRCC = 0.8577 ] +INFO:root:Average loss=0.14680529242524734 +INFO:root: +[Valid][13][UTT][ MSE = 0.2359 | LCC = 0.8593 | SRCC = 0.8583 ] +INFO:root:Average loss=0.12614997992148766 +INFO:root: +[Valid][14][UTT][ MSE = 0.2430 | LCC = 0.8557 | SRCC = 0.8536 ] +INFO:root:Average loss=0.11913803124274963 +INFO:root: +[Valid][15][UTT][ MSE = 0.2320 | LCC = 0.8560 | SRCC = 0.8552 ] +INFO:root:Average loss=0.1143606774126872 +INFO:root: +[Valid][16][UTT][ MSE = 0.2296 | LCC = 0.8576 | SRCC = 0.8558 ] +INFO:root:Average loss=0.11451524056685276 +INFO:root: +[Valid][17][UTT][ MSE = 0.3010 | LCC = 0.8499 | SRCC = 0.8544 ] +INFO:root:Average loss=0.10386477453777423 +INFO:root: +[Valid][18][UTT][ MSE = 0.2384 | LCC = 0.8566 | SRCC = 0.8553 ] +INFO:root:Average loss=0.10874902437894772 +INFO:root: +[Valid][19][UTT][ MSE = 0.2453 | LCC = 0.8593 | SRCC = 0.8576 ] +INFO:root:Average loss=0.09954571914978516 +INFO:root: +[Valid][20][UTT][ MSE = 0.2352 | LCC = 0.8576 | SRCC = 0.8562 ] +INFO:root:Average loss=0.09369729468837762 +INFO:root: +[Valid][21][UTT][ MSE = 0.2365 | LCC = 0.8575 | SRCC = 0.8571 ] +INFO:root:Average loss=0.0882659310904833 +INFO:root: +[Valid][22][UTT][ MSE = 0.2277 | LCC = 0.8576 | SRCC = 0.8579 ] +INFO:root:Average loss=0.08953975475369355 +INFO:root: +[Valid][23][UTT][ MSE = 0.2416 | LCC = 0.8565 | SRCC = 0.8551 ] +INFO:root:Average loss=0.08286652054924232 +INFO:root: +[Valid][24][UTT][ MSE = 0.2304 | LCC = 0.8557 | SRCC = 0.8551 ] +INFO:root:Average loss=0.08264925908774902 +INFO:root: +[Valid][25][UTT][ MSE = 0.2459 | LCC = 0.8535 | SRCC = 0.8549 ] +INFO:root:Average loss=0.08276403084015235 +INFO:root: +[Valid][26][UTT][ MSE = 0.2287 | LCC = 0.8566 | SRCC = 0.8548 ] +INFO:root:Average loss=0.08127261464221355 +INFO:root: +[Valid][27][UTT][ MSE = 0.2240 | LCC = 0.8590 | SRCC = 0.8575 ] +INFO:root:Average loss=0.07184853337896176 +INFO:root: +[Valid][28][UTT][ MSE = 0.2279 | LCC = 0.8593 | SRCC = 0.8602 ] +INFO:root:Average loss=0.06921057298015325 +INFO:root: +[Valid][29][UTT][ MSE = 0.2261 | LCC = 0.8605 | SRCC = 0.8600 ] +INFO:root:Average loss=0.06300882658419701 +INFO:root: +[Valid][30][UTT][ MSE = 0.2273 | LCC = 0.8586 | SRCC = 0.8567 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2464 | LCC = 0.8529 | SRCC = 0.8521 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run1/config.gin b/results_bvcc/w2v2_large_asr_de/layer4/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run1/model.pt b/results_bvcc/w2v2_large_asr_de/layer4/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1e7d1b6ef1675cde4a4f6f180f8b1aa5f671e2b4 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e25c437b1db4a0318a81b355950c75f62b7cd0c0fe882c498f972f33250a9c6 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run1/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer4/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c80996d24e2c31e9ba4fc27c8e8e75cabe3a26ef --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1d05e3ba4e239d69cc29016c53ea5adc610b32e4a3d04f0219b621c58e3fdae8 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run1/train.log b/results_bvcc/w2v2_large_asr_de/layer4/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e60c6cd87d01ea46e90f19753d1fb585095b3a47 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/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=0.5127991479940903 +INFO:root: +[Valid][0][UTT][ MSE = 0.3751 | LCC = 0.7801 | SRCC = 0.7797 ] +INFO:root:Average loss=0.28971375391269344 +INFO:root: +[Valid][1][UTT][ MSE = 0.3466 | LCC = 0.8296 | SRCC = 0.8250 ] +INFO:root:Average loss=0.25305446218221617 +INFO:root: +[Valid][2][UTT][ MSE = 0.2739 | LCC = 0.8377 | SRCC = 0.8369 ] +INFO:root:Average loss=0.24685031882463357 +INFO:root: +[Valid][3][UTT][ MSE = 0.2960 | LCC = 0.8417 | SRCC = 0.8413 ] +INFO:root:Average loss=0.21797957519690195 +INFO:root: +[Valid][4][UTT][ MSE = 0.3040 | LCC = 0.8522 | SRCC = 0.8521 ] +INFO:root:Average loss=0.20899939708984816 +INFO:root: +[Valid][5][UTT][ MSE = 0.2697 | LCC = 0.8539 | SRCC = 0.8544 ] +INFO:root:Average loss=0.19169311196758196 +INFO:root: +[Valid][6][UTT][ MSE = 0.2567 | LCC = 0.8609 | SRCC = 0.8583 ] +INFO:root:Average loss=0.17672472982070384 +INFO:root: +[Valid][7][UTT][ MSE = 0.2650 | LCC = 0.8590 | SRCC = 0.8582 ] +INFO:root:Average loss=0.17761680799034926 +INFO:root: +[Valid][8][UTT][ MSE = 0.2528 | LCC = 0.8635 | SRCC = 0.8626 ] +INFO:root:Average loss=0.15940566179461968 +INFO:root: +[Valid][9][UTT][ MSE = 0.2511 | LCC = 0.8646 | SRCC = 0.8630 ] +INFO:root:Average loss=0.15750550717497483 +INFO:root: +[Valid][10][UTT][ MSE = 0.2436 | LCC = 0.8674 | SRCC = 0.8651 ] +INFO:root:Average loss=0.15080616212426087 +INFO:root: +[Valid][11][UTT][ MSE = 0.2370 | LCC = 0.8653 | SRCC = 0.8640 ] +INFO:root:Average loss=0.13601527286645693 +INFO:root: +[Valid][12][UTT][ MSE = 0.2603 | LCC = 0.8653 | SRCC = 0.8641 ] +INFO:root:Average loss=0.13317892155968225 +INFO:root: +[Valid][13][UTT][ MSE = 0.2401 | LCC = 0.8610 | SRCC = 0.8605 ] +INFO:root:Average loss=0.12270703405523911 +INFO:root: +[Valid][14][UTT][ MSE = 0.2604 | LCC = 0.8626 | SRCC = 0.8626 ] +INFO:root:Average loss=0.1256831228159941 +INFO:root: +[Valid][15][UTT][ MSE = 0.2299 | LCC = 0.8655 | SRCC = 0.8652 ] +INFO:root:Average loss=0.13100502296135977 +INFO:root: +[Valid][16][UTT][ MSE = 0.2512 | LCC = 0.8635 | SRCC = 0.8627 ] +INFO:root:Average loss=0.11303746108061229 +INFO:root: +[Valid][17][UTT][ MSE = 0.2268 | LCC = 0.8615 | SRCC = 0.8608 ] +INFO:root:Average loss=0.11394697141188842 +INFO:root: +[Valid][18][UTT][ MSE = 0.2665 | LCC = 0.8607 | SRCC = 0.8596 ] +INFO:root:Average loss=0.10206059151544021 +INFO:root: +[Valid][19][UTT][ MSE = 0.2274 | LCC = 0.8619 | SRCC = 0.8609 ] +INFO:root:Average loss=0.10008729105958572 +INFO:root: +[Valid][20][UTT][ MSE = 0.2443 | LCC = 0.8608 | SRCC = 0.8594 ] +INFO:root:Average loss=0.10550717059045266 +INFO:root: +[Valid][21][UTT][ MSE = 0.2311 | LCC = 0.8643 | SRCC = 0.8630 ] +INFO:root:Average loss=0.09551701283989808 +INFO:root: +[Valid][22][UTT][ MSE = 0.2461 | LCC = 0.8595 | SRCC = 0.8580 ] +INFO:root:Average loss=0.08311456480087379 +INFO:root: +[Valid][23][UTT][ MSE = 0.2429 | LCC = 0.8576 | SRCC = 0.8556 ] +INFO:root:Average loss=0.08199721584335351 +INFO:root: +[Valid][24][UTT][ MSE = 0.2308 | LCC = 0.8586 | SRCC = 0.8579 ] +INFO:root:Average loss=0.10146005652271785 +INFO:root: +[Valid][25][UTT][ MSE = 0.2273 | LCC = 0.8631 | SRCC = 0.8630 ] +INFO:root:Average loss=0.07818890033433071 +INFO:root: +[Valid][26][UTT][ MSE = 0.2387 | LCC = 0.8612 | SRCC = 0.8600 ] +INFO:root:Average loss=0.0708662130129643 +INFO:root: +[Valid][27][UTT][ MSE = 0.2275 | LCC = 0.8608 | SRCC = 0.8600 ] +INFO:root:Average loss=0.07027764475116363 +INFO:root: +[Valid][28][UTT][ MSE = 0.2274 | LCC = 0.8589 | SRCC = 0.8578 ] +INFO:root:Average loss=0.07392879838171677 +INFO:root: +[Valid][29][UTT][ MSE = 0.2511 | LCC = 0.8593 | SRCC = 0.8603 ] +INFO:root:Average loss=0.0692031529182807 +INFO:root: +[Valid][30][UTT][ MSE = 0.2507 | LCC = 0.8579 | SRCC = 0.8581 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2390 | LCC = 0.8604 | SRCC = 0.8605 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run2/config.gin b/results_bvcc/w2v2_large_asr_de/layer4/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run2/model.pt b/results_bvcc/w2v2_large_asr_de/layer4/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b20b2f7233a92e1dcae95cab3f368737de78b482 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3ddea4f27bcfb296a2ca0f2b94e13604581110e0ecb729bcccd0645d576c8721 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run2/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer4/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c380e165d8d8d819a9a7e675e8b1c7e5ada29156 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f46374ab988120e90a179d5c2b73dc9f9e9857cb1c63866367025f7726c5942d +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run2/train.log b/results_bvcc/w2v2_large_asr_de/layer4/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e78d8547ba3c725af6c8abce8ec2e7f273cb782b --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/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=0.616841754470116 +INFO:root: +[Valid][0][UTT][ MSE = 0.4310 | LCC = 0.7731 | SRCC = 0.7804 ] +INFO:root:Average loss=0.3125424828284826 +INFO:root: +[Valid][1][UTT][ MSE = 0.3090 | LCC = 0.8170 | SRCC = 0.8214 ] +INFO:root:Average loss=0.2647538397174615 +INFO:root: +[Valid][2][UTT][ MSE = 0.3109 | LCC = 0.8388 | SRCC = 0.8380 ] +INFO:root:Average loss=0.2285197162284301 +INFO:root: +[Valid][3][UTT][ MSE = 0.3055 | LCC = 0.8441 | SRCC = 0.8454 ] +INFO:root:Average loss=0.2171405482177551 +INFO:root: +[Valid][4][UTT][ MSE = 0.2849 | LCC = 0.8557 | SRCC = 0.8532 ] +INFO:root:Average loss=0.20065485361294869 +INFO:root: +[Valid][5][UTT][ MSE = 0.2466 | LCC = 0.8580 | SRCC = 0.8576 ] +INFO:root:Average loss=0.19194704867326295 +INFO:root: +[Valid][6][UTT][ MSE = 0.2406 | LCC = 0.8641 | SRCC = 0.8621 ] +INFO:root:Average loss=0.18496850486367178 +INFO:root: +[Valid][7][UTT][ MSE = 0.2626 | LCC = 0.8648 | SRCC = 0.8632 ] +INFO:root:Average loss=0.1702098537905094 +INFO:root: +[Valid][8][UTT][ MSE = 0.2523 | LCC = 0.8659 | SRCC = 0.8672 ] +INFO:root:Average loss=0.15917719623599297 +INFO:root: +[Valid][9][UTT][ MSE = 0.2458 | LCC = 0.8648 | SRCC = 0.8623 ] +INFO:root:Average loss=0.150637316207091 +INFO:root: +[Valid][10][UTT][ MSE = 0.2454 | LCC = 0.8704 | SRCC = 0.8691 ] +INFO:root:Average loss=0.13819419392026389 +INFO:root: +[Valid][11][UTT][ MSE = 0.2634 | LCC = 0.8676 | SRCC = 0.8658 ] +INFO:root:Average loss=0.13823755314716926 +INFO:root: +[Valid][12][UTT][ MSE = 0.2095 | LCC = 0.8724 | SRCC = 0.8712 ] +INFO:root:Average loss=0.12990060181189805 +INFO:root: +[Valid][13][UTT][ MSE = 0.2612 | LCC = 0.8689 | SRCC = 0.8679 ] +INFO:root:Average loss=0.12344401912429394 +INFO:root: +[Valid][14][UTT][ MSE = 0.2158 | LCC = 0.8734 | SRCC = 0.8712 ] +INFO:root:Average loss=0.1144090480147264 +INFO:root: +[Valid][15][UTT][ MSE = 0.2890 | LCC = 0.8638 | SRCC = 0.8621 ] +INFO:root:Average loss=0.11585515259932248 +INFO:root: +[Valid][16][UTT][ MSE = 0.2794 | LCC = 0.8662 | SRCC = 0.8680 ] +INFO:root:Average loss=0.11246082626092128 +INFO:root: +[Valid][17][UTT][ MSE = 0.2183 | LCC = 0.8727 | SRCC = 0.8710 ] +INFO:root:Average loss=0.10793260141060902 +INFO:root: +[Valid][18][UTT][ MSE = 0.2148 | LCC = 0.8687 | SRCC = 0.8669 ] +INFO:root:Average loss=0.10290850832676277 +INFO:root: +[Valid][19][UTT][ MSE = 0.2206 | LCC = 0.8714 | SRCC = 0.8694 ] +INFO:root:Average loss=0.09442719845817639 +INFO:root: +[Valid][20][UTT][ MSE = 0.2240 | LCC = 0.8703 | SRCC = 0.8680 ] +INFO:root:Average loss=0.08771189674735069 +INFO:root: +[Valid][21][UTT][ MSE = 0.2250 | LCC = 0.8695 | SRCC = 0.8682 ] +INFO:root:Average loss=0.0868216194690038 +INFO:root: +[Valid][22][UTT][ MSE = 0.2454 | LCC = 0.8685 | SRCC = 0.8665 ] +INFO:root:Average loss=0.08816164273482102 +INFO:root: +[Valid][23][UTT][ MSE = 0.2574 | LCC = 0.8690 | SRCC = 0.8688 ] +INFO:root:Average loss=0.09451509114259328 +INFO:root: +[Valid][24][UTT][ MSE = 0.2278 | LCC = 0.8678 | SRCC = 0.8686 ] +INFO:root:Average loss=0.07754345715810092 +INFO:root: +[Valid][25][UTT][ MSE = 0.2251 | LCC = 0.8711 | SRCC = 0.8706 ] +INFO:root:Average loss=0.07637572212096973 +INFO:root: +[Valid][26][UTT][ MSE = 0.2560 | LCC = 0.8652 | SRCC = 0.8667 ] +INFO:root:Average loss=0.07038003492813844 +INFO:root: +[Valid][27][UTT][ MSE = 0.2215 | LCC = 0.8681 | SRCC = 0.8668 ] +INFO:root:Average loss=0.06609911414293143 +INFO:root: +[Valid][28][UTT][ MSE = 0.2333 | LCC = 0.8693 | SRCC = 0.8694 ] +INFO:root:Average loss=0.06202237112208819 +INFO:root: +[Valid][29][UTT][ MSE = 0.2340 | LCC = 0.8644 | SRCC = 0.8640 ] +INFO:root:Average loss=0.07139545726852539 +INFO:root: +[Valid][30][UTT][ MSE = 0.2160 | LCC = 0.8682 | SRCC = 0.8673 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2581 | LCC = 0.8546 | SRCC = 0.8562 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run3/config.gin b/results_bvcc/w2v2_large_asr_de/layer4/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run3/model.pt b/results_bvcc/w2v2_large_asr_de/layer4/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..fb8e3749821689d71d6cdeff75bd6b314537ad11 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b34537161b1a4fde6e419a13de9bebc016d96e68912e3393d172272619575e48 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run3/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer4/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..23b3b315479c92108b1dbc60d7b8d50d43717a95 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4355772e8c7ccd9e3c8286ae79ccd1094ca530e308f9566d9727fd6543f7ce2 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run3/train.log b/results_bvcc/w2v2_large_asr_de/layer4/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..fc23397a1028a55011a0fe47e1ebd13d121dd117 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/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=0.6225571731726328 +INFO:root: +[Valid][0][UTT][ MSE = 0.3667 | LCC = 0.7688 | SRCC = 0.7731 ] +INFO:root:Average loss=0.31198874574441177 +INFO:root: +[Valid][1][UTT][ MSE = 0.3289 | LCC = 0.8199 | SRCC = 0.8193 ] +INFO:root:Average loss=0.25627981508389497 +INFO:root: +[Valid][2][UTT][ MSE = 0.3199 | LCC = 0.8269 | SRCC = 0.8320 ] +INFO:root:Average loss=0.23597500702509513 +INFO:root: +[Valid][3][UTT][ MSE = 0.2683 | LCC = 0.8448 | SRCC = 0.8453 ] +INFO:root:Average loss=0.2206867732680761 +INFO:root: +[Valid][4][UTT][ MSE = 0.2725 | LCC = 0.8530 | SRCC = 0.8511 ] +INFO:root:Average loss=0.2103391341291941 +INFO:root: +[Valid][5][UTT][ MSE = 0.2797 | LCC = 0.8474 | SRCC = 0.8458 ] +INFO:root:Average loss=0.19731196445914415 +INFO:root: +[Valid][6][UTT][ MSE = 0.2629 | LCC = 0.8575 | SRCC = 0.8547 ] +INFO:root:Average loss=0.1936232348283132 +INFO:root: +[Valid][7][UTT][ MSE = 0.2551 | LCC = 0.8614 | SRCC = 0.8583 ] +INFO:root:Average loss=0.17756999790286407 +INFO:root: +[Valid][8][UTT][ MSE = 0.2521 | LCC = 0.8618 | SRCC = 0.8625 ] +INFO:root:Average loss=0.16174995822784227 +INFO:root: +[Valid][9][UTT][ MSE = 0.2429 | LCC = 0.8585 | SRCC = 0.8553 ] +INFO:root:Average loss=0.1682454681931398 +INFO:root: +[Valid][10][UTT][ MSE = 0.2443 | LCC = 0.8636 | SRCC = 0.8600 ] +INFO:root:Average loss=0.15461547519916144 +INFO:root: +[Valid][11][UTT][ MSE = 0.2300 | LCC = 0.8592 | SRCC = 0.8571 ] +INFO:root:Average loss=0.14483314189009178 +INFO:root: +[Valid][12][UTT][ MSE = 0.2349 | LCC = 0.8631 | SRCC = 0.8610 ] +INFO:root:Average loss=0.14358067063566965 +INFO:root: +[Valid][13][UTT][ MSE = 0.2363 | LCC = 0.8609 | SRCC = 0.8581 ] +INFO:root:Average loss=0.12380184940038583 +INFO:root: +[Valid][14][UTT][ MSE = 0.2255 | LCC = 0.8619 | SRCC = 0.8603 ] +INFO:root:Average loss=0.12490720416490848 +INFO:root: +[Valid][15][UTT][ MSE = 0.2442 | LCC = 0.8571 | SRCC = 0.8515 ] +INFO:root:Average loss=0.12211977136440766 +INFO:root: +[Valid][16][UTT][ MSE = 0.3072 | LCC = 0.8478 | SRCC = 0.8498 ] +INFO:root:Average loss=0.12170843474375896 +INFO:root: +[Valid][17][UTT][ MSE = 0.2478 | LCC = 0.8551 | SRCC = 0.8505 ] +INFO:root:Average loss=0.1068577708151096 +INFO:root: +[Valid][18][UTT][ MSE = 0.2769 | LCC = 0.8524 | SRCC = 0.8529 ] +INFO:root:Average loss=0.10677047121601227 +INFO:root: +[Valid][19][UTT][ MSE = 0.2502 | LCC = 0.8542 | SRCC = 0.8523 ] +INFO:root:Average loss=0.09969621619735008 +INFO:root: +[Valid][20][UTT][ MSE = 0.2398 | LCC = 0.8596 | SRCC = 0.8564 ] +INFO:root:Average loss=0.10133528556579198 +INFO:root: +[Valid][21][UTT][ MSE = 0.2312 | LCC = 0.8571 | SRCC = 0.8532 ] +INFO:root:Average loss=0.08819223543963371 +INFO:root: +[Valid][22][UTT][ MSE = 0.2294 | LCC = 0.8556 | SRCC = 0.8514 ] +INFO:root:Average loss=0.08481480181217194 +INFO:root: +[Valid][23][UTT][ MSE = 0.2389 | LCC = 0.8583 | SRCC = 0.8561 ] +INFO:root:Average loss=0.08410877524278103 +INFO:root: +[Valid][24][UTT][ MSE = 0.2293 | LCC = 0.8572 | SRCC = 0.8550 ] +INFO:root:Average loss=0.07517849157253902 +INFO:root: +[Valid][25][UTT][ MSE = 0.2455 | LCC = 0.8574 | SRCC = 0.8542 ] +INFO:root:Average loss=0.08439035856953034 +INFO:root: +[Valid][26][UTT][ MSE = 0.2292 | LCC = 0.8588 | SRCC = 0.8553 ] +INFO:root:Average loss=0.07472107483026309 +INFO:root: +[Valid][27][UTT][ MSE = 0.2478 | LCC = 0.8549 | SRCC = 0.8525 ] +INFO:root:Average loss=0.07040325466256875 +INFO:root: +[Valid][28][UTT][ MSE = 0.2269 | LCC = 0.8565 | SRCC = 0.8532 ] +INFO:root:Average loss=0.07085569370060395 +INFO:root: +[Valid][29][UTT][ MSE = 0.2413 | LCC = 0.8543 | SRCC = 0.8503 ] +INFO:root:Average loss=0.06392568053725438 +INFO:root: +[Valid][30][UTT][ MSE = 0.2242 | LCC = 0.8580 | SRCC = 0.8546 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2576 | LCC = 0.8615 | SRCC = 0.8618 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run4/config.gin b/results_bvcc/w2v2_large_asr_de/layer4/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run4/model.pt b/results_bvcc/w2v2_large_asr_de/layer4/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7621406711e22c3abd30d3827cf018d675904f3a --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3b32313fc7615b62cff86839ff403d3cd3277c73c378022ca8ce4a8672defc28 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run4/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer4/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..58bd398604f1cbfba479b78d16b30fa3726714ed --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f204006473fcb936da30c43c2fe76ca1177bfe1e77bc691940a54df18bf948c3 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer4/run4/train.log b/results_bvcc/w2v2_large_asr_de/layer4/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ac97f4c7406dd2ae980b83aad76d9aa3cf235155 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer4/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=0.5155468235413233 +INFO:root: +[Valid][0][UTT][ MSE = 0.3732 | LCC = 0.7833 | SRCC = 0.7869 ] +INFO:root:Average loss=0.2968836228052775 +INFO:root: +[Valid][1][UTT][ MSE = 0.3455 | LCC = 0.8270 | SRCC = 0.8234 ] +INFO:root:Average loss=0.2558045832392497 +INFO:root: +[Valid][2][UTT][ MSE = 0.4778 | LCC = 0.8285 | SRCC = 0.8357 ] +INFO:root:Average loss=0.23792723814646402 +INFO:root: +[Valid][3][UTT][ MSE = 0.2660 | LCC = 0.8544 | SRCC = 0.8513 ] +INFO:root:Average loss=0.21639272226737097 +INFO:root: +[Valid][4][UTT][ MSE = 0.2666 | LCC = 0.8531 | SRCC = 0.8519 ] +INFO:root:Average loss=0.20608235800113434 +INFO:root: +[Valid][5][UTT][ MSE = 0.2648 | LCC = 0.8595 | SRCC = 0.8569 ] +INFO:root:Average loss=0.19388451675573984 +INFO:root: +[Valid][6][UTT][ MSE = 0.2559 | LCC = 0.8596 | SRCC = 0.8549 ] +INFO:root:Average loss=0.17879478346842986 +INFO:root: +[Valid][7][UTT][ MSE = 0.2403 | LCC = 0.8658 | SRCC = 0.8623 ] +INFO:root:Average loss=0.17804831572068044 +INFO:root: +[Valid][8][UTT][ MSE = 0.2553 | LCC = 0.8660 | SRCC = 0.8632 ] +INFO:root:Average loss=0.16805251382100275 +INFO:root: +[Valid][9][UTT][ MSE = 0.2340 | LCC = 0.8668 | SRCC = 0.8643 ] +INFO:root:Average loss=0.16658295805637652 +INFO:root: +[Valid][10][UTT][ MSE = 0.3284 | LCC = 0.8537 | SRCC = 0.8574 ] +INFO:root:Average loss=0.15036824068579918 +INFO:root: +[Valid][11][UTT][ MSE = 0.2361 | LCC = 0.8681 | SRCC = 0.8655 ] +INFO:root:Average loss=0.14117895850004294 +INFO:root: +[Valid][12][UTT][ MSE = 0.2208 | LCC = 0.8659 | SRCC = 0.8640 ] +INFO:root:Average loss=0.13453584833022875 +INFO:root: +[Valid][13][UTT][ MSE = 0.2805 | LCC = 0.8587 | SRCC = 0.8556 ] +INFO:root:Average loss=0.125928152543612 +INFO:root: +[Valid][14][UTT][ MSE = 0.2289 | LCC = 0.8652 | SRCC = 0.8615 ] +INFO:root:Average loss=0.12187153282455909 +INFO:root: +[Valid][15][UTT][ MSE = 0.2337 | LCC = 0.8640 | SRCC = 0.8594 ] +INFO:root:Average loss=0.12164712429810794 +INFO:root: +[Valid][16][UTT][ MSE = 0.2500 | LCC = 0.8625 | SRCC = 0.8583 ] +INFO:root:Average loss=0.11255148282417884 +INFO:root: +[Valid][17][UTT][ MSE = 0.2218 | LCC = 0.8645 | SRCC = 0.8610 ] +INFO:root:Average loss=0.111123343786368 +INFO:root: +[Valid][18][UTT][ MSE = 0.2393 | LCC = 0.8614 | SRCC = 0.8555 ] +INFO:root:Average loss=0.11031947117776443 +INFO:root: +[Valid][19][UTT][ MSE = 0.2190 | LCC = 0.8663 | SRCC = 0.8643 ] +INFO:root:Average loss=0.09898281551133363 +INFO:root: +[Valid][20][UTT][ MSE = 0.2711 | LCC = 0.8615 | SRCC = 0.8597 ] +INFO:root:Average loss=0.09159928832489711 +INFO:root: +[Valid][21][UTT][ MSE = 0.2489 | LCC = 0.8617 | SRCC = 0.8592 ] +INFO:root:Average loss=0.09895973609617123 +INFO:root: +[Valid][22][UTT][ MSE = 0.2209 | LCC = 0.8655 | SRCC = 0.8650 ] +INFO:root:Average loss=0.08349029118051896 +INFO:root: +[Valid][23][UTT][ MSE = 0.2412 | LCC = 0.8635 | SRCC = 0.8604 ] +INFO:root:Average loss=0.08397028872217888 +INFO:root: +[Valid][24][UTT][ MSE = 0.2400 | LCC = 0.8606 | SRCC = 0.8597 ] +INFO:root:Average loss=0.08310194838887607 +INFO:root: +[Valid][25][UTT][ MSE = 0.2408 | LCC = 0.8597 | SRCC = 0.8560 ] +INFO:root:Average loss=0.0769566883547948 +INFO:root: +[Valid][26][UTT][ MSE = 0.2385 | LCC = 0.8598 | SRCC = 0.8591 ] +INFO:root:Average loss=0.07586740960295384 +INFO:root: +[Valid][27][UTT][ MSE = 0.2547 | LCC = 0.8594 | SRCC = 0.8570 ] +INFO:root:Average loss=0.0801882372261622 +INFO:root: +[Valid][28][UTT][ MSE = 0.2397 | LCC = 0.8602 | SRCC = 0.8567 ] +INFO:root:Average loss=0.07340575305697246 +INFO:root: +[Valid][29][UTT][ MSE = 0.2238 | LCC = 0.8635 | SRCC = 0.8616 ] +INFO:root:Average loss=0.06356236782784645 +INFO:root: +[Valid][30][UTT][ MSE = 0.2226 | LCC = 0.8639 | SRCC = 0.8622 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2386 | LCC = 0.8619 | SRCC = 0.8592 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run0/config.gin b/results_bvcc/w2v2_large_asr_de/layer6/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run0/model.pt b/results_bvcc/w2v2_large_asr_de/layer6/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..313d5a148bd2a887f5865f74300afa8b8dd36bf6 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f2d8a973cfac5c6ee0eac4ffdc7e0c9d2423bed4a1a81dae6a1482a0a85e4876 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run0/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer6/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..38002d4436bd58f21a6e1324142c79f3385d0ff8 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c58687a30d58564c8cbe09d0e6e04eeb3a3280e9000b4392880818faad281b32 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run0/train.log b/results_bvcc/w2v2_large_asr_de/layer6/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5888b4f58bcffb4c3c0fe1b9fc834571992c9270 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/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=0.4851124875056438 +INFO:root: +[Valid][0][UTT][ MSE = 0.3224 | LCC = 0.8063 | SRCC = 0.8092 ] +INFO:root:Average loss=0.282545852737549 +INFO:root: +[Valid][1][UTT][ MSE = 0.2853 | LCC = 0.8228 | SRCC = 0.8281 ] +INFO:root:Average loss=0.24569753614755777 +INFO:root: +[Valid][2][UTT][ MSE = 0.3409 | LCC = 0.8449 | SRCC = 0.8428 ] +INFO:root:Average loss=0.2232936571041743 +INFO:root: +[Valid][3][UTT][ MSE = 0.2663 | LCC = 0.8490 | SRCC = 0.8464 ] +INFO:root:Average loss=0.20718000065057707 +INFO:root: +[Valid][4][UTT][ MSE = 0.2523 | LCC = 0.8494 | SRCC = 0.8484 ] +INFO:root:Average loss=0.19272399521791017 +INFO:root: +[Valid][5][UTT][ MSE = 0.2545 | LCC = 0.8485 | SRCC = 0.8449 ] +INFO:root:Average loss=0.17888484083307096 +INFO:root: +[Valid][6][UTT][ MSE = 0.2399 | LCC = 0.8509 | SRCC = 0.8496 ] +INFO:root:Average loss=0.1886102360410568 +INFO:root: +[Valid][7][UTT][ MSE = 0.2558 | LCC = 0.8502 | SRCC = 0.8474 ] +INFO:root:Average loss=0.1804630206181453 +INFO:root: +[Valid][8][UTT][ MSE = 0.3159 | LCC = 0.8545 | SRCC = 0.8499 ] +INFO:root:Average loss=0.1559490650319136 +INFO:root: +[Valid][9][UTT][ MSE = 0.2418 | LCC = 0.8494 | SRCC = 0.8455 ] +INFO:root:Average loss=0.15068939624306482 +INFO:root: +[Valid][10][UTT][ MSE = 0.2464 | LCC = 0.8489 | SRCC = 0.8452 ] +INFO:root:Average loss=0.14161127137067991 +INFO:root: +[Valid][11][UTT][ MSE = 0.2498 | LCC = 0.8426 | SRCC = 0.8459 ] +INFO:root:Average loss=0.12913298807465112 +INFO:root: +[Valid][12][UTT][ MSE = 0.2544 | LCC = 0.8442 | SRCC = 0.8399 ] +INFO:root:Average loss=0.12957147900492716 +INFO:root: +[Valid][13][UTT][ MSE = 0.2410 | LCC = 0.8503 | SRCC = 0.8485 ] +INFO:root:Average loss=0.12410328174248719 +INFO:root: +[Valid][14][UTT][ MSE = 0.2369 | LCC = 0.8521 | SRCC = 0.8500 ] +INFO:root:Average loss=0.11830321814005192 +INFO:root: +[Valid][15][UTT][ MSE = 0.2425 | LCC = 0.8505 | SRCC = 0.8475 ] +INFO:root:Average loss=0.11146146020828149 +INFO:root: +[Valid][16][UTT][ MSE = 0.2386 | LCC = 0.8481 | SRCC = 0.8468 ] +INFO:root:Average loss=0.10208720374756898 +INFO:root: +[Valid][17][UTT][ MSE = 0.2436 | LCC = 0.8474 | SRCC = 0.8458 ] +INFO:root:Average loss=0.09429574203796875 +INFO:root: +[Valid][18][UTT][ MSE = 0.2642 | LCC = 0.8451 | SRCC = 0.8429 ] +INFO:root:Average loss=0.09456843524598159 +INFO:root: +[Valid][19][UTT][ MSE = 0.2493 | LCC = 0.8471 | SRCC = 0.8447 ] +INFO:root:Average loss=0.08562681427559792 +INFO:root: +[Valid][20][UTT][ MSE = 0.2421 | LCC = 0.8450 | SRCC = 0.8449 ] +INFO:root:Average loss=0.09301977270306686 +INFO:root: +[Valid][21][UTT][ MSE = 0.2864 | LCC = 0.8512 | SRCC = 0.8484 ] +INFO:root:Average loss=0.08017809498004425 +INFO:root: +[Valid][22][UTT][ MSE = 0.2441 | LCC = 0.8502 | SRCC = 0.8469 ] +INFO:root:Average loss=0.08345720195808472 +INFO:root: +[Valid][23][UTT][ MSE = 0.2393 | LCC = 0.8525 | SRCC = 0.8499 ] +INFO:root:Average loss=0.07333140963545212 +INFO:root: +[Valid][24][UTT][ MSE = 0.2518 | LCC = 0.8483 | SRCC = 0.8456 ] +INFO:root:Average loss=0.06879252983400455 +INFO:root: +[Valid][25][UTT][ MSE = 0.2379 | LCC = 0.8484 | SRCC = 0.8464 ] +INFO:root:Average loss=0.07284772749512623 +INFO:root: +[Valid][26][UTT][ MSE = 0.2650 | LCC = 0.8486 | SRCC = 0.8449 ] +INFO:root:Average loss=0.06586927423874538 +INFO:root: +[Valid][27][UTT][ MSE = 0.2410 | LCC = 0.8503 | SRCC = 0.8496 ] +INFO:root:Average loss=0.07285987103405671 +INFO:root: +[Valid][28][UTT][ MSE = 0.2379 | LCC = 0.8477 | SRCC = 0.8471 ] +INFO:root:Average loss=0.06167302562449223 +INFO:root: +[Valid][29][UTT][ MSE = 0.2441 | LCC = 0.8492 | SRCC = 0.8466 ] +INFO:root:Average loss=0.06019928057988485 +INFO:root: +[Valid][30][UTT][ MSE = 0.2374 | LCC = 0.8464 | SRCC = 0.8448 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3393 | LCC = 0.8477 | SRCC = 0.8515 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run1/config.gin b/results_bvcc/w2v2_large_asr_de/layer6/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run1/model.pt b/results_bvcc/w2v2_large_asr_de/layer6/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..77f9e1819272ac5b947999c5159082ff24ac25a0 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:db7a94710c6a6f9404de15c9f6877f60e7c089bfd57f7784dca01c32954740d6 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run1/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer6/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..cf2913eb4a8d27faf7d3752f681571768f7b470c --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f49e9c7472930d7acd4572c8a2fb90d637a84699464c9774c6e3d3e58be5be6f +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run1/train.log b/results_bvcc/w2v2_large_asr_de/layer6/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3bc2e24b4b1698d7df4b6258cb5fe811ea50cf14 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/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=0.4691153877438643 +INFO:root: +[Valid][0][UTT][ MSE = 0.3278 | LCC = 0.8073 | SRCC = 0.8145 ] +INFO:root:Average loss=0.2773654050169847 +INFO:root: +[Valid][1][UTT][ MSE = 0.2820 | LCC = 0.8361 | SRCC = 0.8375 ] +INFO:root:Average loss=0.25229763239622116 +INFO:root: +[Valid][2][UTT][ MSE = 0.2722 | LCC = 0.8444 | SRCC = 0.8434 ] +INFO:root:Average loss=0.22193399320046106 +INFO:root: +[Valid][3][UTT][ MSE = 0.2558 | LCC = 0.8461 | SRCC = 0.8448 ] +INFO:root:Average loss=0.20566482899280694 +INFO:root: +[Valid][4][UTT][ MSE = 0.2778 | LCC = 0.8464 | SRCC = 0.8434 ] +INFO:root:Average loss=0.19304026376742583 +INFO:root: +[Valid][5][UTT][ MSE = 0.2561 | LCC = 0.8529 | SRCC = 0.8503 ] +INFO:root:Average loss=0.18358176488142747 +INFO:root: +[Valid][6][UTT][ MSE = 0.2696 | LCC = 0.8520 | SRCC = 0.8475 ] +INFO:root:Average loss=0.1795508731634189 +INFO:root: +[Valid][7][UTT][ MSE = 0.2476 | LCC = 0.8524 | SRCC = 0.8495 ] +INFO:root:Average loss=0.15777918877892005 +INFO:root: +[Valid][8][UTT][ MSE = 0.2408 | LCC = 0.8560 | SRCC = 0.8534 ] +INFO:root:Average loss=0.15199635922908783 +INFO:root: +[Valid][9][UTT][ MSE = 0.2332 | LCC = 0.8571 | SRCC = 0.8558 ] +INFO:root:Average loss=0.1499962512499247 +INFO:root: +[Valid][10][UTT][ MSE = 0.2406 | LCC = 0.8545 | SRCC = 0.8506 ] +INFO:root:Average loss=0.1389270588182486 +INFO:root: +[Valid][11][UTT][ MSE = 0.2682 | LCC = 0.8486 | SRCC = 0.8486 ] +INFO:root:Average loss=0.1301760145295889 +INFO:root: +[Valid][12][UTT][ MSE = 0.2444 | LCC = 0.8500 | SRCC = 0.8477 ] +INFO:root:Average loss=0.11634074543149044 +INFO:root: +[Valid][13][UTT][ MSE = 0.2426 | LCC = 0.8529 | SRCC = 0.8494 ] +INFO:root:Average loss=0.11558912885494721 +INFO:root: +[Valid][14][UTT][ MSE = 0.2425 | LCC = 0.8501 | SRCC = 0.8467 ] +INFO:root:Average loss=0.1131007201396502 +INFO:root: +[Valid][15][UTT][ MSE = 0.2546 | LCC = 0.8503 | SRCC = 0.8477 ] +INFO:root:Average loss=0.11518981107152425 +INFO:root: +[Valid][16][UTT][ MSE = 0.2738 | LCC = 0.8552 | SRCC = 0.8538 ] +INFO:root:Average loss=0.10807967529847072 +INFO:root: +[Valid][17][UTT][ MSE = 0.3053 | LCC = 0.8481 | SRCC = 0.8469 ] +INFO:root:Average loss=0.09635583903544988 +INFO:root: +[Valid][18][UTT][ MSE = 0.2452 | LCC = 0.8503 | SRCC = 0.8477 ] +INFO:root:Average loss=0.09523682921933822 +INFO:root: +[Valid][19][UTT][ MSE = 0.2363 | LCC = 0.8514 | SRCC = 0.8479 ] +INFO:root:Average loss=0.08733623599012692 +INFO:root: +[Valid][20][UTT][ MSE = 0.2473 | LCC = 0.8495 | SRCC = 0.8469 ] +INFO:root:Average loss=0.08005838783887717 +INFO:root: +[Valid][21][UTT][ MSE = 0.2498 | LCC = 0.8487 | SRCC = 0.8462 ] +INFO:root:Average loss=0.07905549245576064 +INFO:root: +[Valid][22][UTT][ MSE = 0.2441 | LCC = 0.8482 | SRCC = 0.8457 ] +INFO:root:Average loss=0.07980574146868327 +INFO:root: +[Valid][23][UTT][ MSE = 0.2380 | LCC = 0.8501 | SRCC = 0.8472 ] +INFO:root:Average loss=0.07476850517858298 +INFO:root: +[Valid][24][UTT][ MSE = 0.2356 | LCC = 0.8521 | SRCC = 0.8501 ] +INFO:root:Average loss=0.06619256247694676 +INFO:root: +[Valid][25][UTT][ MSE = 0.2437 | LCC = 0.8447 | SRCC = 0.8416 ] +INFO:root:Average loss=0.06776151978052579 +INFO:root: +[Valid][26][UTT][ MSE = 0.2411 | LCC = 0.8479 | SRCC = 0.8454 ] +INFO:root:Average loss=0.06246754928277089 +INFO:root: +[Valid][27][UTT][ MSE = 0.2471 | LCC = 0.8488 | SRCC = 0.8454 ] +INFO:root:Average loss=0.059922690383898906 +INFO:root: +[Valid][28][UTT][ MSE = 0.2394 | LCC = 0.8492 | SRCC = 0.8466 ] +INFO:root:Average loss=0.06172641070607381 +INFO:root: +[Valid][29][UTT][ MSE = 0.2421 | LCC = 0.8515 | SRCC = 0.8489 ] +INFO:root:Average loss=0.06563095557384002 +INFO:root: +[Valid][30][UTT][ MSE = 0.2408 | LCC = 0.8497 | SRCC = 0.8468 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2506 | LCC = 0.8573 | SRCC = 0.8577 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run2/config.gin b/results_bvcc/w2v2_large_asr_de/layer6/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run2/model.pt b/results_bvcc/w2v2_large_asr_de/layer6/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..70a00dad91574700fce2e4bc2e68c85367c06553 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:15fffd0e84b0c052df48e2ffa9fc0b1e3579c1800172d8ddb217268c92a65dcd +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run2/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer6/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..8745ec07f8e63ae80fa08ff6a7fba7e1378a57bd --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3bc54f63175092412f18e243cd657a311c58e443888b83b910cce9e752287974 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run2/train.log b/results_bvcc/w2v2_large_asr_de/layer6/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..dbfadd57f53c5bd1769b933e9d988b673d4c2a25 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/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=0.5148605028024087 +INFO:root: +[Valid][0][UTT][ MSE = 0.3460 | LCC = 0.7998 | SRCC = 0.7992 ] +INFO:root:Average loss=0.29832356938949 +INFO:root: +[Valid][1][UTT][ MSE = 0.3273 | LCC = 0.8297 | SRCC = 0.8281 ] +INFO:root:Average loss=0.2495599490327713 +INFO:root: +[Valid][2][UTT][ MSE = 0.2865 | LCC = 0.8392 | SRCC = 0.8391 ] +INFO:root:Average loss=0.23161582763378435 +INFO:root: +[Valid][3][UTT][ MSE = 0.2631 | LCC = 0.8466 | SRCC = 0.8449 ] +INFO:root:Average loss=0.2099841540822616 +INFO:root: +[Valid][4][UTT][ MSE = 0.2642 | LCC = 0.8466 | SRCC = 0.8444 ] +INFO:root:Average loss=0.197149151410812 +INFO:root: +[Valid][5][UTT][ MSE = 0.2606 | LCC = 0.8511 | SRCC = 0.8493 ] +INFO:root:Average loss=0.19821535213253436 +INFO:root: +[Valid][6][UTT][ MSE = 0.2832 | LCC = 0.8419 | SRCC = 0.8447 ] +INFO:root:Average loss=0.17679694801186904 +INFO:root: +[Valid][7][UTT][ MSE = 0.2689 | LCC = 0.8512 | SRCC = 0.8490 ] +INFO:root:Average loss=0.18550017839058852 +INFO:root: +[Valid][8][UTT][ MSE = 0.2702 | LCC = 0.8512 | SRCC = 0.8481 ] +INFO:root:Average loss=0.16548010821525866 +INFO:root: +[Valid][9][UTT][ MSE = 0.2714 | LCC = 0.8528 | SRCC = 0.8487 ] +INFO:root:Average loss=0.15330139060433096 +INFO:root: +[Valid][10][UTT][ MSE = 0.2545 | LCC = 0.8515 | SRCC = 0.8500 ] +INFO:root:Average loss=0.14909918271960357 +INFO:root: +[Valid][11][UTT][ MSE = 0.2747 | LCC = 0.8544 | SRCC = 0.8509 ] +INFO:root:Average loss=0.13933497304335618 +INFO:root: +[Valid][12][UTT][ MSE = 0.2433 | LCC = 0.8508 | SRCC = 0.8489 ] +INFO:root:Average loss=0.1282439964512984 +INFO:root: +[Valid][13][UTT][ MSE = 0.2620 | LCC = 0.8482 | SRCC = 0.8465 ] +INFO:root:Average loss=0.131112394711146 +INFO:root: +[Valid][14][UTT][ MSE = 0.2539 | LCC = 0.8523 | SRCC = 0.8504 ] +INFO:root:Average loss=0.10972104364862809 +INFO:root: +[Valid][15][UTT][ MSE = 0.2580 | LCC = 0.8464 | SRCC = 0.8448 ] +INFO:root:Average loss=0.13875230258473983 +INFO:root: +[Valid][16][UTT][ MSE = 0.2450 | LCC = 0.8526 | SRCC = 0.8501 ] +INFO:root:Average loss=0.10047180272447757 +INFO:root: +[Valid][17][UTT][ MSE = 0.2439 | LCC = 0.8518 | SRCC = 0.8528 ] +INFO:root:Average loss=0.1003046518143935 +INFO:root: +[Valid][18][UTT][ MSE = 0.2493 | LCC = 0.8494 | SRCC = 0.8499 ] +INFO:root:Average loss=0.09454582483531573 +INFO:root: +[Valid][19][UTT][ MSE = 0.2590 | LCC = 0.8499 | SRCC = 0.8488 ] +INFO:root:Average loss=0.09199815147962326 +INFO:root: +[Valid][20][UTT][ MSE = 0.2453 | LCC = 0.8481 | SRCC = 0.8457 ] +INFO:root:Average loss=0.08730425790716441 +INFO:root: +[Valid][21][UTT][ MSE = 0.2399 | LCC = 0.8507 | SRCC = 0.8481 ] +INFO:root:Average loss=0.08677109249700339 +INFO:root: +[Valid][22][UTT][ MSE = 0.2369 | LCC = 0.8512 | SRCC = 0.8504 ] +INFO:root:Average loss=0.07890521171383369 +INFO:root: +[Valid][23][UTT][ MSE = 0.2453 | LCC = 0.8502 | SRCC = 0.8472 ] +INFO:root:Average loss=0.08184381560064279 +INFO:root: +[Valid][24][UTT][ MSE = 0.2448 | LCC = 0.8510 | SRCC = 0.8482 ] +INFO:root:Average loss=0.0774308110658939 +INFO:root: +[Valid][25][UTT][ MSE = 0.2381 | LCC = 0.8505 | SRCC = 0.8487 ] +INFO:root:Average loss=0.07170180761470245 +INFO:root: +[Valid][26][UTT][ MSE = 0.2523 | LCC = 0.8518 | SRCC = 0.8507 ] +INFO:root:Average loss=0.06643479274442562 +INFO:root: +[Valid][27][UTT][ MSE = 0.2568 | LCC = 0.8512 | SRCC = 0.8509 ] +INFO:root:Average loss=0.06351401064640437 +INFO:root: +[Valid][28][UTT][ MSE = 0.2507 | LCC = 0.8496 | SRCC = 0.8476 ] +INFO:root:Average loss=0.06333479801049599 +INFO:root: +[Valid][29][UTT][ MSE = 0.2568 | LCC = 0.8511 | SRCC = 0.8501 ] +INFO:root:Average loss=0.059796842197195076 +INFO:root: +[Valid][30][UTT][ MSE = 0.2555 | LCC = 0.8472 | SRCC = 0.8450 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2971 | LCC = 0.8499 | SRCC = 0.8507 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run3/config.gin b/results_bvcc/w2v2_large_asr_de/layer6/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run3/model.pt b/results_bvcc/w2v2_large_asr_de/layer6/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6c99e159ea33592c48ad3aab0389b205b1045cda --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:557f186b6e4dc767f639abb75c1cb612340dae9bfff7da55ba2eba83129fecd1 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run3/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer6/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..57e8f94b0325bccd8ebba2e12c781856154a2fe9 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:99086fe523f2534888a4fc2ce348544e8590af6b590db570016e3ef4b68221c2 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run3/train.log b/results_bvcc/w2v2_large_asr_de/layer6/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..4e7554920f42bccc6c4b12927ec5a3f30ade93bc --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/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=0.5805381056972039 +INFO:root: +[Valid][0][UTT][ MSE = 0.3664 | LCC = 0.7954 | SRCC = 0.7951 ] +INFO:root:Average loss=0.29023660241793364 +INFO:root: +[Valid][1][UTT][ MSE = 0.3358 | LCC = 0.8288 | SRCC = 0.8278 ] +INFO:root:Average loss=0.2508332937573775 +INFO:root: +[Valid][2][UTT][ MSE = 0.3156 | LCC = 0.8396 | SRCC = 0.8367 ] +INFO:root:Average loss=0.23633572573845202 +INFO:root: +[Valid][3][UTT][ MSE = 0.2856 | LCC = 0.8447 | SRCC = 0.8406 ] +INFO:root:Average loss=0.2076333789871289 +INFO:root: +[Valid][4][UTT][ MSE = 0.2677 | LCC = 0.8504 | SRCC = 0.8464 ] +INFO:root:Average loss=0.20537175983190536 +INFO:root: +[Valid][5][UTT][ MSE = 0.2764 | LCC = 0.8444 | SRCC = 0.8429 ] +INFO:root:Average loss=0.18295703799678728 +INFO:root: +[Valid][6][UTT][ MSE = 0.2548 | LCC = 0.8512 | SRCC = 0.8474 ] +INFO:root:Average loss=0.17478852403851655 +INFO:root: +[Valid][7][UTT][ MSE = 0.2690 | LCC = 0.8536 | SRCC = 0.8494 ] +INFO:root:Average loss=0.17801964798798928 +INFO:root: +[Valid][8][UTT][ MSE = 0.2386 | LCC = 0.8551 | SRCC = 0.8506 ] +INFO:root:Average loss=0.15627208896554434 +INFO:root: +[Valid][9][UTT][ MSE = 0.2840 | LCC = 0.8490 | SRCC = 0.8486 ] +INFO:root:Average loss=0.15751713609848267 +INFO:root: +[Valid][10][UTT][ MSE = 0.2627 | LCC = 0.8542 | SRCC = 0.8520 ] +INFO:root:Average loss=0.14940467906685975 +INFO:root: +[Valid][11][UTT][ MSE = 0.2707 | LCC = 0.8514 | SRCC = 0.8473 ] +INFO:root:Average loss=0.13372727999320397 +INFO:root: +[Valid][12][UTT][ MSE = 0.2774 | LCC = 0.8523 | SRCC = 0.8479 ] +INFO:root:Average loss=0.13222944383055735 +INFO:root: +[Valid][13][UTT][ MSE = 0.3110 | LCC = 0.8447 | SRCC = 0.8415 ] +INFO:root:Average loss=0.12721199036026612 +INFO:root: +[Valid][14][UTT][ MSE = 0.2432 | LCC = 0.8469 | SRCC = 0.8437 ] +INFO:root:Average loss=0.10832684544416574 +INFO:root: +[Valid][15][UTT][ MSE = 0.2379 | LCC = 0.8504 | SRCC = 0.8467 ] +INFO:root:Average loss=0.11711834657650727 +INFO:root: +[Valid][16][UTT][ MSE = 0.2445 | LCC = 0.8500 | SRCC = 0.8476 ] +INFO:root:Average loss=0.10983231090582334 +INFO:root: +[Valid][17][UTT][ MSE = 0.2493 | LCC = 0.8468 | SRCC = 0.8422 ] +INFO:root:Average loss=0.09948356569004364 +INFO:root: +[Valid][18][UTT][ MSE = 0.2641 | LCC = 0.8411 | SRCC = 0.8361 ] +INFO:root:Average loss=0.09346691877222978 +INFO:root: +[Valid][19][UTT][ MSE = 0.2466 | LCC = 0.8468 | SRCC = 0.8430 ] +INFO:root:Average loss=0.090185754144421 +INFO:root: +[Valid][20][UTT][ MSE = 0.2539 | LCC = 0.8477 | SRCC = 0.8458 ] +INFO:root:Average loss=0.08390556457333076 +INFO:root: +[Valid][21][UTT][ MSE = 0.2988 | LCC = 0.8456 | SRCC = 0.8443 ] +INFO:root:Average loss=0.07686354816915134 +INFO:root: +[Valid][22][UTT][ MSE = 0.2748 | LCC = 0.8460 | SRCC = 0.8422 ] +INFO:root:Average loss=0.0765274167060852 +INFO:root: +[Valid][23][UTT][ MSE = 0.2420 | LCC = 0.8468 | SRCC = 0.8424 ] +INFO:root:Average loss=0.0816068019813452 +INFO:root: +[Valid][24][UTT][ MSE = 0.2478 | LCC = 0.8474 | SRCC = 0.8437 ] +INFO:root:Average loss=0.07331543950698315 +INFO:root: +[Valid][25][UTT][ MSE = 0.2723 | LCC = 0.8456 | SRCC = 0.8430 ] +INFO:root:Average loss=0.07519310292525169 +INFO:root: +[Valid][26][UTT][ MSE = 0.2442 | LCC = 0.8485 | SRCC = 0.8454 ] +INFO:root:Average loss=0.06584445325036843 +INFO:root: +[Valid][27][UTT][ MSE = 0.3240 | LCC = 0.8402 | SRCC = 0.8398 ] +INFO:root:Average loss=0.06627158529292314 +INFO:root: +[Valid][28][UTT][ MSE = 0.2441 | LCC = 0.8468 | SRCC = 0.8436 ] +INFO:root:Average loss=0.05965265244818651 +INFO:root: +[Valid][29][UTT][ MSE = 0.2544 | LCC = 0.8494 | SRCC = 0.8479 ] +INFO:root:Average loss=0.05821765930606769 +INFO:root: +[Valid][30][UTT][ MSE = 0.2504 | LCC = 0.8435 | SRCC = 0.8405 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2391 | LCC = 0.8496 | SRCC = 0.8503 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run4/config.gin b/results_bvcc/w2v2_large_asr_de/layer6/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run4/model.pt b/results_bvcc/w2v2_large_asr_de/layer6/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8bff7b8236a257dd20245e3334dc27c2e5c05dfa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b0ea31f68d3e310a57c216551fbba7a87ea60873f47b3e8cde67cadfc5470c58 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run4/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer6/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..fc18e9a1406d976eeda40c1ed278dc8b193dbc1c --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:060dcab78f604a1c9460aa404d380c8bfe393096bafaea75494d06e6cce74913 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer6/run4/train.log b/results_bvcc/w2v2_large_asr_de/layer6/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f0f3f3d6f7fcf26c4a9232e48403b10a2a0b12af --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer6/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=0.5404636947772442 +INFO:root: +[Valid][0][UTT][ MSE = 0.3445 | LCC = 0.7937 | SRCC = 0.7982 ] +INFO:root:Average loss=0.2861730309251027 +INFO:root: +[Valid][1][UTT][ MSE = 0.2965 | LCC = 0.8309 | SRCC = 0.8338 ] +INFO:root:Average loss=0.24168789635101953 +INFO:root: +[Valid][2][UTT][ MSE = 0.3037 | LCC = 0.8396 | SRCC = 0.8393 ] +INFO:root:Average loss=0.21672601405626687 +INFO:root: +[Valid][3][UTT][ MSE = 0.2880 | LCC = 0.8406 | SRCC = 0.8429 ] +INFO:root:Average loss=0.2070119648407667 +INFO:root: +[Valid][4][UTT][ MSE = 0.2508 | LCC = 0.8521 | SRCC = 0.8485 ] +INFO:root:Average loss=0.20764400848211387 +INFO:root: +[Valid][5][UTT][ MSE = 0.2747 | LCC = 0.8464 | SRCC = 0.8449 ] +INFO:root:Average loss=0.18346196489456373 +INFO:root: +[Valid][6][UTT][ MSE = 0.2430 | LCC = 0.8472 | SRCC = 0.8474 ] +INFO:root:Average loss=0.17066695436071128 +INFO:root: +[Valid][7][UTT][ MSE = 0.2479 | LCC = 0.8516 | SRCC = 0.8488 ] +INFO:root:Average loss=0.15734374513610816 +INFO:root: +[Valid][8][UTT][ MSE = 0.2919 | LCC = 0.8429 | SRCC = 0.8438 ] +INFO:root:Average loss=0.15169957308815077 +INFO:root: +[Valid][9][UTT][ MSE = 0.2539 | LCC = 0.8500 | SRCC = 0.8476 ] +INFO:root:Average loss=0.13793161005163804 +INFO:root: +[Valid][10][UTT][ MSE = 0.2564 | LCC = 0.8494 | SRCC = 0.8459 ] +INFO:root:Average loss=0.1341448836028576 +INFO:root: +[Valid][11][UTT][ MSE = 0.2557 | LCC = 0.8434 | SRCC = 0.8408 ] +INFO:root:Average loss=0.12239093190202346 +INFO:root: +[Valid][12][UTT][ MSE = 0.2428 | LCC = 0.8479 | SRCC = 0.8462 ] +INFO:root:Average loss=0.1260394559074671 +INFO:root: +[Valid][13][UTT][ MSE = 0.2597 | LCC = 0.8435 | SRCC = 0.8457 ] +INFO:root:Average loss=0.11665791359085304 +INFO:root: +[Valid][14][UTT][ MSE = 0.2413 | LCC = 0.8506 | SRCC = 0.8474 ] +INFO:root:Average loss=0.10678075082026996 +INFO:root: +[Valid][15][UTT][ MSE = 0.2496 | LCC = 0.8428 | SRCC = 0.8392 ] +INFO:root:Average loss=0.10061027314991523 +INFO:root: +[Valid][16][UTT][ MSE = 0.2538 | LCC = 0.8468 | SRCC = 0.8447 ] +INFO:root:Average loss=0.09895167170235744 +INFO:root: +[Valid][17][UTT][ MSE = 0.2715 | LCC = 0.8377 | SRCC = 0.8377 ] +INFO:root:Average loss=0.10242887705755539 +INFO:root: +[Valid][18][UTT][ MSE = 0.2553 | LCC = 0.8444 | SRCC = 0.8424 ] +INFO:root:Average loss=0.09260057166027717 +INFO:root: +[Valid][19][UTT][ MSE = 0.2683 | LCC = 0.8415 | SRCC = 0.8397 ] +INFO:root:Average loss=0.09050310895037957 +INFO:root: +[Valid][20][UTT][ MSE = 0.2925 | LCC = 0.8352 | SRCC = 0.8371 ] +INFO:root:Average loss=0.08721686145051932 +INFO:root: +[Valid][21][UTT][ MSE = 0.2560 | LCC = 0.8407 | SRCC = 0.8382 ] +INFO:root:Average loss=0.08687347436371522 +INFO:root: +[Valid][22][UTT][ MSE = 0.2447 | LCC = 0.8432 | SRCC = 0.8420 ] +INFO:root:Average loss=0.07730966147321922 +INFO:root: +[Valid][23][UTT][ MSE = 0.2550 | LCC = 0.8416 | SRCC = 0.8413 ] +INFO:root:Average loss=0.07705544059475262 +INFO:root: +[Valid][24][UTT][ MSE = 0.2439 | LCC = 0.8454 | SRCC = 0.8454 ] +INFO:root:Average loss=0.07137835804277505 +INFO:root: +[Valid][25][UTT][ MSE = 0.2505 | LCC = 0.8402 | SRCC = 0.8383 ] +INFO:root:Average loss=0.06853805950436836 +INFO:root: +[Valid][26][UTT][ MSE = 0.2478 | LCC = 0.8390 | SRCC = 0.8388 ] +INFO:root:Average loss=0.06409692057432273 +INFO:root: +[Valid][27][UTT][ MSE = 0.2466 | LCC = 0.8444 | SRCC = 0.8429 ] +INFO:root:Average loss=0.06798157387245925 +INFO:root: +[Valid][28][UTT][ MSE = 0.2582 | LCC = 0.8417 | SRCC = 0.8409 ] +INFO:root:Average loss=0.06409000204159664 +INFO:root: +[Valid][29][UTT][ MSE = 0.2649 | LCC = 0.8446 | SRCC = 0.8430 ] +INFO:root:Average loss=0.0618122833279463 +INFO:root: +[Valid][30][UTT][ MSE = 0.2545 | LCC = 0.8422 | SRCC = 0.8397 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2552 | LCC = 0.8526 | SRCC = 0.8522 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run0/config.gin b/results_bvcc/w2v2_large_asr_de/layer8/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run0/model.pt b/results_bvcc/w2v2_large_asr_de/layer8/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c21352e1b91b81d10e4392a1c15b97ce544a7fef --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a4a8aa56f7df2783ffc342f4428cfa0fb581bd0d42ebea0cd4694841e1f9b9ea +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run0/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer8/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7f0251f0a3be76f0c9b149395ba0d4f7e6324ca2 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0f7d0226ee00b00f084caebe727aa122206cca4759c7d7c34c944240802a1d38 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run0/train.log b/results_bvcc/w2v2_large_asr_de/layer8/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0998a4fbfd05a1a84daa0c9747fe45bb94c7124c --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/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=0.5716966673349723 +INFO:root: +[Valid][0][UTT][ MSE = 0.3765 | LCC = 0.7808 | SRCC = 0.7888 ] +INFO:root:Average loss=0.31315111063229734 +INFO:root: +[Valid][1][UTT][ MSE = 0.3133 | LCC = 0.8225 | SRCC = 0.8222 ] +INFO:root:Average loss=0.2770265524203961 +INFO:root: +[Valid][2][UTT][ MSE = 0.3565 | LCC = 0.8281 | SRCC = 0.8318 ] +INFO:root:Average loss=0.251781263985695 +INFO:root: +[Valid][3][UTT][ MSE = 0.2883 | LCC = 0.8387 | SRCC = 0.8367 ] +INFO:root:Average loss=0.22513115138579637 +INFO:root: +[Valid][4][UTT][ MSE = 0.3480 | LCC = 0.8288 | SRCC = 0.8358 ] +INFO:root:Average loss=0.23150963546373907 +INFO:root: +[Valid][5][UTT][ MSE = 0.2913 | LCC = 0.8298 | SRCC = 0.8360 ] +INFO:root:Average loss=0.1950228639329091 +INFO:root: +[Valid][6][UTT][ MSE = 0.2926 | LCC = 0.8371 | SRCC = 0.8397 ] +INFO:root:Average loss=0.18693824570912582 +INFO:root: +[Valid][7][UTT][ MSE = 0.2725 | LCC = 0.8429 | SRCC = 0.8396 ] +INFO:root:Average loss=0.17707666066976693 +INFO:root: +[Valid][8][UTT][ MSE = 0.2689 | LCC = 0.8431 | SRCC = 0.8389 ] +INFO:root:Average loss=0.16341754354727575 +INFO:root: +[Valid][9][UTT][ MSE = 0.2791 | LCC = 0.8387 | SRCC = 0.8368 ] +INFO:root:Average loss=0.15229401450890762 +INFO:root: +[Valid][10][UTT][ MSE = 0.2845 | LCC = 0.8424 | SRCC = 0.8384 ] +INFO:root:Average loss=0.13786920102743003 +INFO:root: +[Valid][11][UTT][ MSE = 0.2553 | LCC = 0.8432 | SRCC = 0.8406 ] +INFO:root:Average loss=0.15621309775190476 +INFO:root: +[Valid][12][UTT][ MSE = 0.3016 | LCC = 0.8326 | SRCC = 0.8370 ] +INFO:root:Average loss=0.13474881534393018 +INFO:root: +[Valid][13][UTT][ MSE = 0.2832 | LCC = 0.8350 | SRCC = 0.8354 ] +INFO:root:Average loss=0.11936371716169211 +INFO:root: +[Valid][14][UTT][ MSE = 0.2683 | LCC = 0.8359 | SRCC = 0.8334 ] +INFO:root:Average loss=0.11998812166544107 +INFO:root: +[Valid][15][UTT][ MSE = 0.2676 | LCC = 0.8373 | SRCC = 0.8355 ] +INFO:root:Average loss=0.10905070822590436 +INFO:root: +[Valid][16][UTT][ MSE = 0.2526 | LCC = 0.8403 | SRCC = 0.8381 ] +INFO:root:Average loss=0.09926971229605186 +INFO:root: +[Valid][17][UTT][ MSE = 0.3238 | LCC = 0.8348 | SRCC = 0.8366 ] +INFO:root:Average loss=0.10214508530230094 +INFO:root: +[Valid][18][UTT][ MSE = 0.2708 | LCC = 0.8369 | SRCC = 0.8351 ] +INFO:root:Average loss=0.09242792312915508 +INFO:root: +[Valid][19][UTT][ MSE = 0.2550 | LCC = 0.8368 | SRCC = 0.8352 ] +INFO:root:Average loss=0.09115389428841762 +INFO:root: +[Valid][20][UTT][ MSE = 0.2686 | LCC = 0.8384 | SRCC = 0.8380 ] +INFO:root:Average loss=0.08902440181909463 +INFO:root: +[Valid][21][UTT][ MSE = 0.3103 | LCC = 0.8392 | SRCC = 0.8361 ] +INFO:root:Average loss=0.09285792383628014 +INFO:root: +[Valid][22][UTT][ MSE = 0.2557 | LCC = 0.8381 | SRCC = 0.8393 ] +INFO:root:Average loss=0.07139346237557057 +INFO:root: +[Valid][23][UTT][ MSE = 0.2635 | LCC = 0.8353 | SRCC = 0.8344 ] +INFO:root:Average loss=0.07292765441040198 +INFO:root: +[Valid][24][UTT][ MSE = 0.2698 | LCC = 0.8330 | SRCC = 0.8319 ] +INFO:root:Average loss=0.0655657706829982 +INFO:root: +[Valid][25][UTT][ MSE = 0.2698 | LCC = 0.8351 | SRCC = 0.8334 ] +INFO:root:Average loss=0.0674530798330521 +INFO:root: +[Valid][26][UTT][ MSE = 0.2585 | LCC = 0.8352 | SRCC = 0.8331 ] +INFO:root:Average loss=0.06280125090135978 +INFO:root: +[Valid][27][UTT][ MSE = 0.2628 | LCC = 0.8371 | SRCC = 0.8373 ] +INFO:root:Average loss=0.06886500655076443 +INFO:root: +[Valid][28][UTT][ MSE = 0.2550 | LCC = 0.8364 | SRCC = 0.8346 ] +INFO:root:Average loss=0.05485389362543057 +INFO:root: +[Valid][29][UTT][ MSE = 0.2716 | LCC = 0.8333 | SRCC = 0.8328 ] +INFO:root:Average loss=0.06133549493283798 +INFO:root: +[Valid][30][UTT][ MSE = 0.2718 | LCC = 0.8316 | SRCC = 0.8286 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2952 | LCC = 0.8368 | SRCC = 0.8427 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run1/config.gin b/results_bvcc/w2v2_large_asr_de/layer8/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run1/model.pt b/results_bvcc/w2v2_large_asr_de/layer8/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..94e2554193c8e36fc881fae5c394935097bf6b06 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a755324133e78f69c350129d313d52c1d28ee8cf85230627962368b8c4eae7a5 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run1/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer8/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..63ee3c53c431eb0d6570b4f458d1f9a368459dc3 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:86ea0e0755f112675ce426c50eda93a5b2236f9a767291c9300fccc07b003697 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run1/train.log b/results_bvcc/w2v2_large_asr_de/layer8/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0d06848acb2594b3c4c681ea9831beecdbcea345 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/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=0.6572162700005066 +INFO:root: +[Valid][0][UTT][ MSE = 0.3854 | LCC = 0.7701 | SRCC = 0.7834 ] +INFO:root:Average loss=0.31923054922849703 +INFO:root: +[Valid][1][UTT][ MSE = 0.3202 | LCC = 0.8159 | SRCC = 0.8175 ] +INFO:root:Average loss=0.27132683591200757 +INFO:root: +[Valid][2][UTT][ MSE = 0.3212 | LCC = 0.8326 | SRCC = 0.8325 ] +INFO:root:Average loss=0.26904770970726627 +INFO:root: +[Valid][3][UTT][ MSE = 0.2935 | LCC = 0.8380 | SRCC = 0.8369 ] +INFO:root:Average loss=0.2421479385632735 +INFO:root: +[Valid][4][UTT][ MSE = 0.2838 | LCC = 0.8396 | SRCC = 0.8380 ] +INFO:root:Average loss=0.21302312994614625 +INFO:root: +[Valid][5][UTT][ MSE = 0.2874 | LCC = 0.8417 | SRCC = 0.8383 ] +INFO:root:Average loss=0.2024380820683944 +INFO:root: +[Valid][6][UTT][ MSE = 0.2796 | LCC = 0.8434 | SRCC = 0.8414 ] +INFO:root:Average loss=0.1833040150694358 +INFO:root: +[Valid][7][UTT][ MSE = 0.2672 | LCC = 0.8392 | SRCC = 0.8345 ] +INFO:root:Average loss=0.18486988983857325 +INFO:root: +[Valid][8][UTT][ MSE = 0.2557 | LCC = 0.8473 | SRCC = 0.8450 ] +INFO:root:Average loss=0.1659092740752758 +INFO:root: +[Valid][9][UTT][ MSE = 0.2738 | LCC = 0.8461 | SRCC = 0.8416 ] +INFO:root:Average loss=0.15783800585911825 +INFO:root: +[Valid][10][UTT][ MSE = 0.2612 | LCC = 0.8443 | SRCC = 0.8380 ] +INFO:root:Average loss=0.14855503778044993 +INFO:root: +[Valid][11][UTT][ MSE = 0.2596 | LCC = 0.8430 | SRCC = 0.8373 ] +INFO:root:Average loss=0.1381389071735052 +INFO:root: +[Valid][12][UTT][ MSE = 0.2551 | LCC = 0.8411 | SRCC = 0.8368 ] +INFO:root:Average loss=0.12864852572480837 +INFO:root: +[Valid][13][UTT][ MSE = 0.2586 | LCC = 0.8427 | SRCC = 0.8395 ] +INFO:root:Average loss=0.12513785952558884 +INFO:root: +[Valid][14][UTT][ MSE = 0.2941 | LCC = 0.8338 | SRCC = 0.8280 ] +INFO:root:Average loss=0.10969179410200852 +INFO:root: +[Valid][15][UTT][ MSE = 0.2597 | LCC = 0.8384 | SRCC = 0.8338 ] +INFO:root:Average loss=0.1207789876140081 +INFO:root: +[Valid][16][UTT][ MSE = 0.2484 | LCC = 0.8398 | SRCC = 0.8349 ] +INFO:root:Average loss=0.10360670175689918 +INFO:root: +[Valid][17][UTT][ MSE = 0.2594 | LCC = 0.8391 | SRCC = 0.8346 ] +INFO:root:Average loss=0.08912871085489407 +INFO:root: +[Valid][18][UTT][ MSE = 0.2543 | LCC = 0.8382 | SRCC = 0.8353 ] +INFO:root:Average loss=0.08782098886485283 +INFO:root: +[Valid][19][UTT][ MSE = 0.2540 | LCC = 0.8358 | SRCC = 0.8325 ] +INFO:root:Average loss=0.09089187312966739 +INFO:root: +[Valid][20][UTT][ MSE = 0.2760 | LCC = 0.8404 | SRCC = 0.8379 ] +INFO:root:Average loss=0.0798626928948439 +INFO:root: +[Valid][21][UTT][ MSE = 0.2575 | LCC = 0.8396 | SRCC = 0.8368 ] +INFO:root:Average loss=0.07561933191922995 +INFO:root: +[Valid][22][UTT][ MSE = 0.2602 | LCC = 0.8394 | SRCC = 0.8356 ] +INFO:root:Average loss=0.08136393583546846 +INFO:root: +[Valid][23][UTT][ MSE = 0.2607 | LCC = 0.8394 | SRCC = 0.8372 ] +INFO:root:Average loss=0.0758805665641259 +INFO:root: +[Valid][24][UTT][ MSE = 0.2573 | LCC = 0.8379 | SRCC = 0.8347 ] +INFO:root:Average loss=0.07196037437862311 +INFO:root: +[Valid][25][UTT][ MSE = 0.2492 | LCC = 0.8439 | SRCC = 0.8407 ] +INFO:root:Average loss=0.06516033014616905 +INFO:root: +[Valid][26][UTT][ MSE = 0.2525 | LCC = 0.8408 | SRCC = 0.8384 ] +INFO:root:Average loss=0.06069709742680574 +INFO:root: +[Valid][27][UTT][ MSE = 0.2607 | LCC = 0.8388 | SRCC = 0.8357 ] +INFO:root:Average loss=0.06736749663757972 +INFO:root: +[Valid][28][UTT][ MSE = 0.2531 | LCC = 0.8409 | SRCC = 0.8380 ] +INFO:root:Average loss=0.06218698893028956 +INFO:root: +[Valid][29][UTT][ MSE = 0.2593 | LCC = 0.8418 | SRCC = 0.8379 ] +INFO:root:Average loss=0.059050663971366026 +INFO:root: +[Valid][30][UTT][ MSE = 0.2663 | LCC = 0.8428 | SRCC = 0.8394 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2641 | LCC = 0.8461 | SRCC = 0.8486 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run2/config.gin b/results_bvcc/w2v2_large_asr_de/layer8/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run2/model.pt b/results_bvcc/w2v2_large_asr_de/layer8/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ea974b170b6883de53125f90f1700a0cb507395d --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2cf776cd53308369990b6d95870f8707728350181eb2f7f2f7dd81869260d78c +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run2/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer8/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..49ee404028271bd43129194b3489ad0891a70de3 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6fba3f11676ab625a49f3d25c70162156f0aaf892ea1dfb9dd0ef25951540668 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run2/train.log b/results_bvcc/w2v2_large_asr_de/layer8/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f611da02aa0cfc6bb016a91c48970e55661c2ec5 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/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=0.6045333936046331 +INFO:root: +[Valid][0][UTT][ MSE = 0.3690 | LCC = 0.7787 | SRCC = 0.7855 ] +INFO:root:Average loss=0.3138218310016852 +INFO:root: +[Valid][1][UTT][ MSE = 0.3357 | LCC = 0.8206 | SRCC = 0.8196 ] +INFO:root:Average loss=0.26709911628411365 +INFO:root: +[Valid][2][UTT][ MSE = 0.2903 | LCC = 0.8295 | SRCC = 0.8279 ] +INFO:root:Average loss=0.2531397747687804 +INFO:root: +[Valid][3][UTT][ MSE = 0.3169 | LCC = 0.8256 | SRCC = 0.8282 ] +INFO:root:Average loss=0.22493994102264062 +INFO:root: +[Valid][4][UTT][ MSE = 0.2953 | LCC = 0.8317 | SRCC = 0.8301 ] +INFO:root:Average loss=0.20617208515222257 +INFO:root: +[Valid][5][UTT][ MSE = 0.2953 | LCC = 0.8325 | SRCC = 0.8312 ] +INFO:root:Average loss=0.19473297874896955 +INFO:root: +[Valid][6][UTT][ MSE = 0.2711 | LCC = 0.8347 | SRCC = 0.8320 ] +INFO:root:Average loss=0.1812149649247145 +INFO:root: +[Valid][7][UTT][ MSE = 0.2747 | LCC = 0.8424 | SRCC = 0.8389 ] +INFO:root:Average loss=0.17290363890620378 +INFO:root: +[Valid][8][UTT][ MSE = 0.2692 | LCC = 0.8300 | SRCC = 0.8272 ] +INFO:root:Average loss=0.15704932092474058 +INFO:root: +[Valid][9][UTT][ MSE = 0.3060 | LCC = 0.8322 | SRCC = 0.8312 ] +INFO:root:Average loss=0.15558133360285026 +INFO:root: +[Valid][10][UTT][ MSE = 0.2604 | LCC = 0.8378 | SRCC = 0.8335 ] +INFO:root:Average loss=0.14433260023211822 +INFO:root: +[Valid][11][UTT][ MSE = 0.2686 | LCC = 0.8301 | SRCC = 0.8262 ] +INFO:root:Average loss=0.13249553004518533 +INFO:root: +[Valid][12][UTT][ MSE = 0.2674 | LCC = 0.8285 | SRCC = 0.8225 ] +INFO:root:Average loss=0.13983512660249686 +INFO:root: +[Valid][13][UTT][ MSE = 0.2718 | LCC = 0.8275 | SRCC = 0.8265 ] +INFO:root:Average loss=0.129372095498137 +INFO:root: +[Valid][14][UTT][ MSE = 0.2696 | LCC = 0.8308 | SRCC = 0.8289 ] +INFO:root:Average loss=0.11842990819460307 +INFO:root: +[Valid][15][UTT][ MSE = 0.2622 | LCC = 0.8354 | SRCC = 0.8340 ] +INFO:root:Average loss=0.1077669046055048 +INFO:root: +[Valid][16][UTT][ MSE = 0.2691 | LCC = 0.8288 | SRCC = 0.8257 ] +INFO:root:Average loss=0.10049529664982589 +INFO:root: +[Valid][17][UTT][ MSE = 0.2733 | LCC = 0.8222 | SRCC = 0.8184 ] +INFO:root:Average loss=0.10322827931780082 +INFO:root: +[Valid][18][UTT][ MSE = 0.2907 | LCC = 0.8232 | SRCC = 0.8224 ] +INFO:root:Average loss=0.10211465364465347 +INFO:root: +[Valid][19][UTT][ MSE = 0.2701 | LCC = 0.8297 | SRCC = 0.8284 ] +INFO:root:Average loss=0.08497489802539349 +INFO:root: +[Valid][20][UTT][ MSE = 0.2774 | LCC = 0.8263 | SRCC = 0.8258 ] +INFO:root:Average loss=0.08334865421056747 +INFO:root: +[Valid][21][UTT][ MSE = 0.2660 | LCC = 0.8262 | SRCC = 0.8230 ] +INFO:root:Average loss=0.08554304973819317 +INFO:root: +[Valid][22][UTT][ MSE = 0.2724 | LCC = 0.8242 | SRCC = 0.8208 ] +INFO:root:Average loss=0.07696745573328091 +INFO:root: +[Valid][23][UTT][ MSE = 0.2705 | LCC = 0.8229 | SRCC = 0.8191 ] +INFO:root:Average loss=0.08564412775330055 +INFO:root: +[Valid][24][UTT][ MSE = 0.2817 | LCC = 0.8245 | SRCC = 0.8222 ] +INFO:root:Average loss=0.07249013644953568 +INFO:root: +[Valid][25][UTT][ MSE = 0.2793 | LCC = 0.8190 | SRCC = 0.8163 ] +INFO:root:Average loss=0.06884744409949352 +INFO:root: +[Valid][26][UTT][ MSE = 0.2714 | LCC = 0.8258 | SRCC = 0.8231 ] +INFO:root:Average loss=0.065908669326932 +INFO:root: +[Valid][27][UTT][ MSE = 0.2908 | LCC = 0.8166 | SRCC = 0.8128 ] +INFO:root:Average loss=0.06760630928553067 +INFO:root: +[Valid][28][UTT][ MSE = 0.2898 | LCC = 0.8207 | SRCC = 0.8180 ] +INFO:root:Average loss=0.06813164351460262 +INFO:root: +[Valid][29][UTT][ MSE = 0.3319 | LCC = 0.8190 | SRCC = 0.8209 ] +INFO:root:Average loss=0.06185921656493193 +INFO:root: +[Valid][30][UTT][ MSE = 0.2824 | LCC = 0.8233 | SRCC = 0.8213 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2859 | LCC = 0.8451 | SRCC = 0.8496 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run3/config.gin b/results_bvcc/w2v2_large_asr_de/layer8/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run3/model.pt b/results_bvcc/w2v2_large_asr_de/layer8/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a648994c9232f3b50719e054ea4fdb520d56e292 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:425b11167af3111cc2687eb719f0c13cb9ff9275504e3643b3c609cf022a8bda +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run3/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer8/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..31f818e31525d337872ab57eb137b54c273a326c --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e69a09df14203039bda38b662aa94461bd3d5869c5ba63b07d22a455ab976f00 +size 3405062 diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run3/train.log b/results_bvcc/w2v2_large_asr_de/layer8/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..84050dd9992952e0cb55843bdd826114893f8a5d --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/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=0.5463600846437308 +INFO:root: +[Valid][0][UTT][ MSE = 0.3869 | LCC = 0.7819 | SRCC = 0.7864 ] +INFO:root:Average loss=0.3283559120236299 +INFO:root: +[Valid][1][UTT][ MSE = 0.4415 | LCC = 0.8107 | SRCC = 0.8152 ] +INFO:root:Average loss=0.2740174090633026 +INFO:root: +[Valid][2][UTT][ MSE = 0.2821 | LCC = 0.8257 | SRCC = 0.8268 ] +INFO:root:Average loss=0.24875514553143427 +INFO:root: +[Valid][3][UTT][ MSE = 0.2754 | LCC = 0.8386 | SRCC = 0.8354 ] +INFO:root:Average loss=0.24005443564592263 +INFO:root: +[Valid][4][UTT][ MSE = 0.2625 | LCC = 0.8376 | SRCC = 0.8378 ] +INFO:root:Average loss=0.21682054549455643 +INFO:root: +[Valid][5][UTT][ MSE = 0.2753 | LCC = 0.8330 | SRCC = 0.8326 ] +INFO:root:Average loss=0.19749988787449324 +INFO:root: +[Valid][6][UTT][ MSE = 0.3160 | LCC = 0.8390 | SRCC = 0.8368 ] +INFO:root:Average loss=0.1874988331244542 +INFO:root: +[Valid][7][UTT][ MSE = 0.2754 | LCC = 0.8356 | SRCC = 0.8336 ] +INFO:root:Average loss=0.18493723640075097 +INFO:root: +[Valid][8][UTT][ MSE = 0.2550 | LCC = 0.8366 | SRCC = 0.8374 ] +INFO:root:Average loss=0.16190989477894244 +INFO:root: +[Valid][9][UTT][ MSE = 0.2678 | LCC = 0.8348 | SRCC = 0.8352 ] +INFO:root:Average loss=0.15557648738225302 +INFO:root: +[Valid][10][UTT][ MSE = 0.2637 | LCC = 0.8371 | SRCC = 0.8340 ] +INFO:root:Average loss=0.1633995507772152 +INFO:root: +[Valid][11][UTT][ MSE = 0.2570 | LCC = 0.8386 | SRCC = 0.8352 ] +INFO:root:Average loss=0.14853680181579712 +INFO:root: +[Valid][12][UTT][ MSE = 0.2706 | LCC = 0.8385 | SRCC = 0.8351 ] +INFO:root:Average loss=0.1324801525244346 +INFO:root: +[Valid][13][UTT][ MSE = 0.2681 | LCC = 0.8335 | SRCC = 0.8300 ] +INFO:root:Average loss=0.12465629688440225 +INFO:root: +[Valid][14][UTT][ MSE = 0.2746 | LCC = 0.8316 | SRCC = 0.8287 ] +INFO:root:Average loss=0.11847279846477203 +INFO:root: +[Valid][15][UTT][ MSE = 0.2659 | LCC = 0.8303 | SRCC = 0.8289 ] +INFO:root:Average loss=0.11023164722017753 +INFO:root: +[Valid][16][UTT][ MSE = 0.2649 | LCC = 0.8316 | SRCC = 0.8284 ] +INFO:root:Average loss=0.101711272333677 +INFO:root: +[Valid][17][UTT][ MSE = 0.2717 | LCC = 0.8312 | SRCC = 0.8287 ] +INFO:root:Average loss=0.09478000212365235 +INFO:root: +[Valid][18][UTT][ MSE = 0.2775 | LCC = 0.8272 | SRCC = 0.8250 ] +INFO:root:Average loss=0.0990070337190842 +INFO:root: +[Valid][19][UTT][ MSE = 0.2657 | LCC = 0.8267 | SRCC = 0.8233 ] +INFO:root:Average loss=0.09079471679452138 +INFO:root: +[Valid][20][UTT][ MSE = 0.2793 | LCC = 0.8271 | SRCC = 0.8248 ] +INFO:root:Average loss=0.09852380935962383 +INFO:root: +[Valid][21][UTT][ MSE = 0.2733 | LCC = 0.8280 | SRCC = 0.8251 ] +INFO:root:Average loss=0.08086565538094594 +INFO:root: +[Valid][22][UTT][ MSE = 0.2712 | LCC = 0.8267 | SRCC = 0.8254 ] +INFO:root:Average loss=0.0789323408061113 +INFO:root: +[Valid][23][UTT][ MSE = 0.2877 | LCC = 0.8236 | SRCC = 0.8205 ] +INFO:root:Average loss=0.07567452534269063 +INFO:root: +[Valid][24][UTT][ MSE = 0.2673 | LCC = 0.8270 | SRCC = 0.8251 ] +INFO:root:Average loss=0.06810210521022479 +INFO:root: +[Valid][25][UTT][ MSE = 0.2602 | LCC = 0.8281 | SRCC = 0.8272 ] +INFO:root:Average loss=0.06576790302418746 +INFO:root: +[Valid][26][UTT][ MSE = 0.2670 | LCC = 0.8266 | SRCC = 0.8243 ] +INFO:root:Average loss=0.0700220838189125 +INFO:root: +[Valid][27][UTT][ MSE = 0.2713 | LCC = 0.8232 | SRCC = 0.8200 ] +INFO:root:Average loss=0.06676567532122135 +INFO:root: +[Valid][28][UTT][ MSE = 0.2682 | LCC = 0.8293 | SRCC = 0.8279 ] +INFO:root:Average loss=0.05733174768586954 +INFO:root: +[Valid][29][UTT][ MSE = 0.2656 | LCC = 0.8290 | SRCC = 0.8274 ] +INFO:root:Average loss=0.055259755884225555 +INFO:root: +[Valid][30][UTT][ MSE = 0.2714 | LCC = 0.8332 | SRCC = 0.8295 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3565 | LCC = 0.8374 | SRCC = 0.8441 ] diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run4/config.gin b/results_bvcc/w2v2_large_asr_de/layer8/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run4/model.pt b/results_bvcc/w2v2_large_asr_de/layer8/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..427ef39dcece554a82bfb887f9117b3236aa3b77 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e3edcea120e2bf4b45f42a2f73acdce7b7bd2e7193a76103456f2607ae6b7b8 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run4/model_best.pt b/results_bvcc/w2v2_large_asr_de/layer8/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..71245493182faf0a2c52cbdf13e47cd91f562c18 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:685bbd3b72e3cd14482a5ce1e0e66a8c454d36361d9fe74ee1f8cbb024e20345 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_de/layer8/run4/train.log b/results_bvcc/w2v2_large_asr_de/layer8/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..027824597104d0a3e7f12ec56ee5f17dd45c4d74 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_de/layer8/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=0.5458873155025336 +INFO:root: +[Valid][0][UTT][ MSE = 0.3602 | LCC = 0.7790 | SRCC = 0.7811 ] +INFO:root:Average loss=0.3193378719763878 +INFO:root: +[Valid][1][UTT][ MSE = 0.3182 | LCC = 0.8117 | SRCC = 0.8114 ] +INFO:root:Average loss=0.27328855716265166 +INFO:root: +[Valid][2][UTT][ MSE = 0.3601 | LCC = 0.8157 | SRCC = 0.8218 ] +INFO:root:Average loss=0.25363257622871643 +INFO:root: +[Valid][3][UTT][ MSE = 0.2808 | LCC = 0.8318 | SRCC = 0.8287 ] +INFO:root:Average loss=0.23393167517124078 +INFO:root: +[Valid][4][UTT][ MSE = 0.2693 | LCC = 0.8352 | SRCC = 0.8336 ] +INFO:root:Average loss=0.2170824796343461 +INFO:root: +[Valid][5][UTT][ MSE = 0.2672 | LCC = 0.8402 | SRCC = 0.8378 ] +INFO:root:Average loss=0.19444437582905477 +INFO:root: +[Valid][6][UTT][ MSE = 0.2584 | LCC = 0.8441 | SRCC = 0.8409 ] +INFO:root:Average loss=0.20071922806211007 +INFO:root: +[Valid][7][UTT][ MSE = 0.2572 | LCC = 0.8420 | SRCC = 0.8392 ] +INFO:root:Average loss=0.16833894126690352 +INFO:root: +[Valid][8][UTT][ MSE = 0.2687 | LCC = 0.8358 | SRCC = 0.8357 ] +INFO:root:Average loss=0.17435249562064806 +INFO:root: +[Valid][9][UTT][ MSE = 0.2874 | LCC = 0.8408 | SRCC = 0.8406 ] +INFO:root:Average loss=0.15821839658877787 +INFO:root: +[Valid][10][UTT][ MSE = 0.2520 | LCC = 0.8424 | SRCC = 0.8395 ] +INFO:root:Average loss=0.15733649256901863 +INFO:root: +[Valid][11][UTT][ MSE = 0.2514 | LCC = 0.8431 | SRCC = 0.8402 ] +INFO:root:Average loss=0.15727679116221574 +INFO:root: +[Valid][12][UTT][ MSE = 0.2621 | LCC = 0.8413 | SRCC = 0.8391 ] +INFO:root:Average loss=0.13184545513911125 +INFO:root: +[Valid][13][UTT][ MSE = 0.2589 | LCC = 0.8382 | SRCC = 0.8369 ] +INFO:root:Average loss=0.12417368944256733 +INFO:root: +[Valid][14][UTT][ MSE = 0.2771 | LCC = 0.8410 | SRCC = 0.8387 ] +INFO:root:Average loss=0.13200494933586854 +INFO:root: +[Valid][15][UTT][ MSE = 0.2589 | LCC = 0.8415 | SRCC = 0.8396 ] +INFO:root:Average loss=0.1106750107346437 +INFO:root: +[Valid][16][UTT][ MSE = 0.2670 | LCC = 0.8421 | SRCC = 0.8394 ] +INFO:root:Average loss=0.11341068922327115 +INFO:root: +[Valid][17][UTT][ MSE = 0.2628 | LCC = 0.8399 | SRCC = 0.8368 ] +INFO:root:Average loss=0.10661594349986468 +INFO:root: +[Valid][18][UTT][ MSE = 0.2910 | LCC = 0.8404 | SRCC = 0.8375 ] +INFO:root:Average loss=0.10067294342204547 +INFO:root: +[Valid][19][UTT][ MSE = 0.2869 | LCC = 0.8367 | SRCC = 0.8328 ] +INFO:root:Average loss=0.10303519379634124 +INFO:root: +[Valid][20][UTT][ MSE = 0.2582 | LCC = 0.8412 | SRCC = 0.8379 ] +INFO:root:Average loss=0.09666331188801007 +INFO:root: +[Valid][21][UTT][ MSE = 0.2625 | LCC = 0.8405 | SRCC = 0.8364 ] +INFO:root:Average loss=0.08022609668282363 +INFO:root: +[Valid][22][UTT][ MSE = 0.2720 | LCC = 0.8366 | SRCC = 0.8347 ] +INFO:root:Average loss=0.07767546769136037 +INFO:root: +[Valid][23][UTT][ MSE = 0.2502 | LCC = 0.8389 | SRCC = 0.8357 ] +INFO:root:Average loss=0.07895386056640209 +INFO:root: +[Valid][24][UTT][ MSE = 0.2649 | LCC = 0.8402 | SRCC = 0.8372 ] +INFO:root:Average loss=0.08242684822434033 +INFO:root: +[Valid][25][UTT][ MSE = 0.2517 | LCC = 0.8403 | SRCC = 0.8371 ] +INFO:root:Average loss=0.07004150624076526 +INFO:root: +[Valid][26][UTT][ MSE = 0.2492 | LCC = 0.8385 | SRCC = 0.8361 ] +INFO:root:Average loss=0.07162482568468803 +INFO:root: +[Valid][27][UTT][ MSE = 0.2599 | LCC = 0.8401 | SRCC = 0.8376 ] +INFO:root:Average loss=0.06521176857252915 +INFO:root: +[Valid][28][UTT][ MSE = 0.2537 | LCC = 0.8372 | SRCC = 0.8344 ] +INFO:root:Average loss=0.06177563444735148 +INFO:root: +[Valid][29][UTT][ MSE = 0.2525 | LCC = 0.8366 | SRCC = 0.8342 ] +INFO:root:Average loss=0.06119767552576004 +INFO:root: +[Valid][30][UTT][ MSE = 0.2535 | LCC = 0.8372 | SRCC = 0.8339 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2566 | LCC = 0.8457 | SRCC = 0.8478 ] diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run0/config.gin b/results_bvcc/w2v2_large_asr_en/layer0/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run0/model.pt b/results_bvcc/w2v2_large_asr_en/layer0/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..22748e798d6f910a0b29bc5fde587a042396af09 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:87372cc3eca1dd4d339e3c0b902f9607133692b47725f79328a573ad650117ed +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run0/model_best.pt b/results_bvcc/w2v2_large_asr_en/layer0/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b90f507263f0bc37b0e31bed42d51bd97fcb130c --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8165ba3d5fab9889403b38f430cce14137bb50b55ed660afe9b5747b4d4856d0 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run0/train.log b/results_bvcc/w2v2_large_asr_en/layer0/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a5516cec07890e970a710279ec5d3addc228099a --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/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=0.5701359747312008 +INFO:root: +[Valid][0][UTT][ MSE = 0.4666 | LCC = 0.6932 | SRCC = 0.7022 ] +INFO:root:Average loss=0.3851214758096597 +INFO:root: +[Valid][1][UTT][ MSE = 0.4164 | LCC = 0.7349 | SRCC = 0.7372 ] +INFO:root:Average loss=0.3661072231256045 +INFO:root: +[Valid][2][UTT][ MSE = 0.4355 | LCC = 0.7412 | SRCC = 0.7434 ] +INFO:root:Average loss=0.35030015997397596 +INFO:root: +[Valid][3][UTT][ MSE = 0.3643 | LCC = 0.7604 | SRCC = 0.7569 ] +INFO:root:Average loss=0.33809973815312755 +INFO:root: +[Valid][4][UTT][ MSE = 0.3700 | LCC = 0.7702 | SRCC = 0.7650 ] +INFO:root:Average loss=0.31561735883737224 +INFO:root: +[Valid][5][UTT][ MSE = 0.3925 | LCC = 0.7667 | SRCC = 0.7605 ] +INFO:root:Average loss=0.3026770769785612 +INFO:root: +[Valid][6][UTT][ MSE = 0.3906 | LCC = 0.7773 | SRCC = 0.7763 ] +INFO:root:Average loss=0.29805986277568036 +INFO:root: +[Valid][7][UTT][ MSE = 0.3291 | LCC = 0.7817 | SRCC = 0.7771 ] +INFO:root:Average loss=0.2911683082198485 +INFO:root: +[Valid][8][UTT][ MSE = 0.3307 | LCC = 0.7919 | SRCC = 0.7895 ] +INFO:root:Average loss=0.27025299824965304 +INFO:root: +[Valid][9][UTT][ MSE = 0.3326 | LCC = 0.7971 | SRCC = 0.7950 ] +INFO:root:Average loss=0.2615054569756373 +INFO:root: +[Valid][10][UTT][ MSE = 0.3130 | LCC = 0.7933 | SRCC = 0.7919 ] +INFO:root:Average loss=0.25685598529302156 +INFO:root: +[Valid][11][UTT][ MSE = 0.3263 | LCC = 0.7983 | SRCC = 0.7962 ] +INFO:root:Average loss=0.2490006379592113 +INFO:root: +[Valid][12][UTT][ MSE = 0.3580 | LCC = 0.7941 | SRCC = 0.7887 ] +INFO:root:Average loss=0.2652061382929484 +INFO:root: +[Valid][13][UTT][ MSE = 0.3209 | LCC = 0.7988 | SRCC = 0.7952 ] +INFO:root:Average loss=0.23463100653428298 +INFO:root: +[Valid][14][UTT][ MSE = 0.3149 | LCC = 0.8032 | SRCC = 0.8012 ] +INFO:root:Average loss=0.23945921659469604 +INFO:root: +[Valid][15][UTT][ MSE = 0.3348 | LCC = 0.7993 | SRCC = 0.8001 ] +INFO:root:Average loss=0.22678197891666338 +INFO:root: +[Valid][16][UTT][ MSE = 0.3203 | LCC = 0.7938 | SRCC = 0.7911 ] +INFO:root:Average loss=0.23481771560051504 +INFO:root: +[Valid][17][UTT][ MSE = 0.3920 | LCC = 0.8032 | SRCC = 0.8030 ] +INFO:root:Average loss=0.20895064870516458 +INFO:root: +[Valid][18][UTT][ MSE = 0.3131 | LCC = 0.8035 | SRCC = 0.8039 ] +INFO:root:Average loss=0.20637951925014839 +INFO:root: +[Valid][19][UTT][ MSE = 0.3156 | LCC = 0.8053 | SRCC = 0.8062 ] +INFO:root:Average loss=0.20322561378662402 +INFO:root: +[Valid][20][UTT][ MSE = 0.3014 | LCC = 0.8124 | SRCC = 0.8123 ] +INFO:root:Average loss=0.22685109365444917 +INFO:root: +[Valid][21][UTT][ MSE = 0.3058 | LCC = 0.8096 | SRCC = 0.8099 ] +INFO:root:Average loss=0.2004397957561872 +INFO:root: +[Valid][22][UTT][ MSE = 0.2871 | LCC = 0.8147 | SRCC = 0.8143 ] +INFO:root:Average loss=0.17885761679365084 +INFO:root: +[Valid][23][UTT][ MSE = 0.3067 | LCC = 0.8097 | SRCC = 0.8078 ] +INFO:root:Average loss=0.1855978698302538 +INFO:root: +[Valid][24][UTT][ MSE = 0.2909 | LCC = 0.8124 | SRCC = 0.8098 ] +INFO:root:Average loss=0.17833677297219253 +INFO:root: +[Valid][25][UTT][ MSE = 0.2920 | LCC = 0.8111 | SRCC = 0.8115 ] +INFO:root:Average loss=0.17209506455140236 +INFO:root: +[Valid][26][UTT][ MSE = 0.2951 | LCC = 0.8084 | SRCC = 0.8083 ] +INFO:root:Average loss=0.16995569576437658 +INFO:root: +[Valid][27][UTT][ MSE = 0.2848 | LCC = 0.8155 | SRCC = 0.8146 ] +INFO:root:Average loss=0.16199631043351614 +INFO:root: +[Valid][28][UTT][ MSE = 0.2883 | LCC = 0.8142 | SRCC = 0.8142 ] +INFO:root:Average loss=0.18460609286259383 +INFO:root: +[Valid][29][UTT][ MSE = 0.2950 | LCC = 0.8133 | SRCC = 0.8150 ] +INFO:root:Average loss=0.15693960720912004 +INFO:root: +[Valid][30][UTT][ MSE = 0.3578 | LCC = 0.8106 | SRCC = 0.8149 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2729 | LCC = 0.8274 | SRCC = 0.8242 ] diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run1/config.gin b/results_bvcc/w2v2_large_asr_en/layer0/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run1/model.pt b/results_bvcc/w2v2_large_asr_en/layer0/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1fb1989a6fa63fa81ac1d27dc25417717e135b1d --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a432fed50f789795a94a1c848476669c33dc966a0c6261bc33199039016752f8 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run1/model_best.pt b/results_bvcc/w2v2_large_asr_en/layer0/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a538bf636c7705f533916a63d35857966bb75a18 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:856b585fd18fd1b1fe2dbfda98ac737c5e0a34273c0a313183315cdc7ee25e15 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run1/train.log b/results_bvcc/w2v2_large_asr_en/layer0/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..92b495fa20af34688f6a2a17a2742634c4dd2862 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/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=0.587884838382403 +INFO:root: +[Valid][0][UTT][ MSE = 0.4488 | LCC = 0.6829 | SRCC = 0.6931 ] +INFO:root:Average loss=0.38921721336933285 +INFO:root: +[Valid][1][UTT][ MSE = 0.4062 | LCC = 0.7192 | SRCC = 0.7246 ] +INFO:root:Average loss=0.3519178338539906 +INFO:root: +[Valid][2][UTT][ MSE = 0.3764 | LCC = 0.7465 | SRCC = 0.7476 ] +INFO:root:Average loss=0.33950758362427735 +INFO:root: +[Valid][3][UTT][ MSE = 0.3658 | LCC = 0.7567 | SRCC = 0.7532 ] +INFO:root:Average loss=0.3276895617063229 +INFO:root: +[Valid][4][UTT][ MSE = 0.3587 | LCC = 0.7716 | SRCC = 0.7675 ] +INFO:root:Average loss=0.33319396067124146 +INFO:root: +[Valid][5][UTT][ MSE = 0.3370 | LCC = 0.7697 | SRCC = 0.7676 ] +INFO:root:Average loss=0.2939891242063962 +INFO:root: +[Valid][6][UTT][ MSE = 0.3713 | LCC = 0.7773 | SRCC = 0.7768 ] +INFO:root:Average loss=0.2935104752198244 +INFO:root: +[Valid][7][UTT][ MSE = 0.3618 | LCC = 0.7780 | SRCC = 0.7759 ] +INFO:root:Average loss=0.3188480681333786 +INFO:root: +[Valid][8][UTT][ MSE = 0.3360 | LCC = 0.7820 | SRCC = 0.7766 ] +INFO:root:Average loss=0.2827664600350918 +INFO:root: +[Valid][9][UTT][ MSE = 0.4153 | LCC = 0.7889 | SRCC = 0.7883 ] +INFO:root:Average loss=0.2902277058515793 +INFO:root: +[Valid][10][UTT][ MSE = 0.3382 | LCC = 0.7918 | SRCC = 0.7894 ] +INFO:root:Average loss=0.25259092373725694 +INFO:root: +[Valid][11][UTT][ MSE = 0.3470 | LCC = 0.7846 | SRCC = 0.7816 ] +INFO:root:Average loss=0.256659423884673 +INFO:root: +[Valid][12][UTT][ MSE = 0.3215 | LCC = 0.7919 | SRCC = 0.7889 ] +INFO:root:Average loss=0.243104049601616 +INFO:root: +[Valid][13][UTT][ MSE = 0.3234 | LCC = 0.7934 | SRCC = 0.7930 ] +INFO:root:Average loss=0.2401633681013034 +INFO:root: +[Valid][14][UTT][ MSE = 0.3259 | LCC = 0.7916 | SRCC = 0.7894 ] +INFO:root:Average loss=0.23865927755832672 +INFO:root: +[Valid][15][UTT][ MSE = 0.3231 | LCC = 0.7894 | SRCC = 0.7911 ] +INFO:root:Average loss=0.21809533181098792 +INFO:root: +[Valid][16][UTT][ MSE = 0.3238 | LCC = 0.7931 | SRCC = 0.7962 ] +INFO:root:Average loss=0.21170130333839318 +INFO:root: +[Valid][17][UTT][ MSE = 0.3279 | LCC = 0.7956 | SRCC = 0.7943 ] +INFO:root:Average loss=0.2108817452039474 +INFO:root: +[Valid][18][UTT][ MSE = 0.3221 | LCC = 0.7951 | SRCC = 0.7910 ] +INFO:root:Average loss=0.21993152052164078 +INFO:root: +[Valid][19][UTT][ MSE = 0.3106 | LCC = 0.7977 | SRCC = 0.7986 ] +INFO:root:Average loss=0.20573172737390566 +INFO:root: +[Valid][20][UTT][ MSE = 0.3286 | LCC = 0.8001 | SRCC = 0.8016 ] +INFO:root:Average loss=0.19180094813689208 +INFO:root: +[Valid][21][UTT][ MSE = 0.3416 | LCC = 0.8056 | SRCC = 0.8062 ] +INFO:root:Average loss=0.18347540736580506 +INFO:root: +[Valid][22][UTT][ MSE = 0.3037 | LCC = 0.8028 | SRCC = 0.8024 ] +INFO:root:Average loss=0.19479754242377403 +INFO:root: +[Valid][23][UTT][ MSE = 0.3132 | LCC = 0.8022 | SRCC = 0.8025 ] +INFO:root:Average loss=0.18759273546628463 +INFO:root: +[Valid][24][UTT][ MSE = 0.3117 | LCC = 0.7993 | SRCC = 0.8001 ] +INFO:root:Average loss=0.1709384159781994 +INFO:root: +[Valid][25][UTT][ MSE = 0.3159 | LCC = 0.8031 | SRCC = 0.8040 ] +INFO:root:Average loss=0.1689623165398072 +INFO:root: +[Valid][26][UTT][ MSE = 0.2905 | LCC = 0.8095 | SRCC = 0.8072 ] +INFO:root:Average loss=0.17681544149915376 +INFO:root: +[Valid][27][UTT][ MSE = 0.2985 | LCC = 0.8121 | SRCC = 0.8116 ] +INFO:root:Average loss=0.16622647929650086 +INFO:root: +[Valid][28][UTT][ MSE = 0.2921 | LCC = 0.8104 | SRCC = 0.8104 ] +INFO:root:Average loss=0.1715819803185952 +INFO:root: +[Valid][29][UTT][ MSE = 0.2972 | LCC = 0.8061 | SRCC = 0.8055 ] +INFO:root:Average loss=0.16649696584313345 +INFO:root: +[Valid][30][UTT][ MSE = 0.2958 | LCC = 0.8056 | SRCC = 0.8055 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3191 | LCC = 0.8252 | SRCC = 0.8275 ] diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run2/config.gin b/results_bvcc/w2v2_large_asr_en/layer0/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run2/model.pt b/results_bvcc/w2v2_large_asr_en/layer0/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8afb715df52ea971aa7f33e7571d227f95488b4d --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e18fc393a2d935fd6a0d2758b7ec165d810702b800107c5d808791a114cc42ad +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run2/model_best.pt b/results_bvcc/w2v2_large_asr_en/layer0/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9466f3f8a404f8ebab32607fec3d2c0e0b74b57a --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:42a6dbd153d7c826bfd9098c8d72bcc398f490bc52e63f7e1128c54c9f38b37a +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run2/train.log b/results_bvcc/w2v2_large_asr_en/layer0/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..74ec2965eed3c23474e8aec1a2ae2a0b0a1a8b2f --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/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=0.5225937381768838 +INFO:root: +[Valid][0][UTT][ MSE = 0.4888 | LCC = 0.6784 | SRCC = 0.6840 ] +INFO:root:Average loss=0.396029049769426 +INFO:root: +[Valid][1][UTT][ MSE = 0.4607 | LCC = 0.7239 | SRCC = 0.7248 ] +INFO:root:Average loss=0.38408729911614686 +INFO:root: +[Valid][2][UTT][ MSE = 0.4136 | LCC = 0.7349 | SRCC = 0.7355 ] +INFO:root:Average loss=0.3511576805359278 +INFO:root: +[Valid][3][UTT][ MSE = 0.3921 | LCC = 0.7444 | SRCC = 0.7449 ] +INFO:root:Average loss=0.33846559738501525 +INFO:root: +[Valid][4][UTT][ MSE = 0.3754 | LCC = 0.7539 | SRCC = 0.7516 ] +INFO:root:Average loss=0.33097163473184293 +INFO:root: +[Valid][5][UTT][ MSE = 0.3799 | LCC = 0.7598 | SRCC = 0.7577 ] +INFO:root:Average loss=0.31762313059507274 +INFO:root: +[Valid][6][UTT][ MSE = 0.3834 | LCC = 0.7585 | SRCC = 0.7544 ] +INFO:root:Average loss=0.2962230497445816 +INFO:root: +[Valid][7][UTT][ MSE = 0.3574 | LCC = 0.7656 | SRCC = 0.7663 ] +INFO:root:Average loss=0.29471448885324675 +INFO:root: +[Valid][8][UTT][ MSE = 0.3472 | LCC = 0.7817 | SRCC = 0.7801 ] +INFO:root:Average loss=0.28297718939108724 +INFO:root: +[Valid][9][UTT][ MSE = 0.3428 | LCC = 0.7831 | SRCC = 0.7805 ] +INFO:root:Average loss=0.26894119190864074 +INFO:root: +[Valid][10][UTT][ MSE = 0.3701 | LCC = 0.7775 | SRCC = 0.7768 ] +INFO:root:Average loss=0.2620463789655612 +INFO:root: +[Valid][11][UTT][ MSE = 0.3751 | LCC = 0.7864 | SRCC = 0.7853 ] +INFO:root:Average loss=0.25835611862249863 +INFO:root: +[Valid][12][UTT][ MSE = 0.3292 | LCC = 0.7893 | SRCC = 0.7899 ] +INFO:root:Average loss=0.27435895571341884 +INFO:root: +[Valid][13][UTT][ MSE = 0.3381 | LCC = 0.7861 | SRCC = 0.7869 ] +INFO:root:Average loss=0.2431433596290075 +INFO:root: +[Valid][14][UTT][ MSE = 0.3188 | LCC = 0.7932 | SRCC = 0.7949 ] +INFO:root:Average loss=0.23855512933089182 +INFO:root: +[Valid][15][UTT][ MSE = 0.3354 | LCC = 0.7871 | SRCC = 0.7902 ] +INFO:root:Average loss=0.23819734824773592 +INFO:root: +[Valid][16][UTT][ MSE = 0.3168 | LCC = 0.7920 | SRCC = 0.7940 ] +INFO:root:Average loss=0.2175974374016126 +INFO:root: +[Valid][17][UTT][ MSE = 0.3018 | LCC = 0.8037 | SRCC = 0.8031 ] +INFO:root:Average loss=0.21742309983342123 +INFO:root: +[Valid][18][UTT][ MSE = 0.3010 | LCC = 0.8028 | SRCC = 0.8015 ] +INFO:root:Average loss=0.21841386839365348 +INFO:root: +[Valid][19][UTT][ MSE = 0.2998 | LCC = 0.8017 | SRCC = 0.8042 ] +INFO:root:Average loss=0.20245016270723099 +INFO:root: +[Valid][20][UTT][ MSE = 0.3180 | LCC = 0.8078 | SRCC = 0.8088 ] +INFO:root:Average loss=0.2162011910516482 +INFO:root: +[Valid][21][UTT][ MSE = 0.3185 | LCC = 0.7996 | SRCC = 0.8016 ] +INFO:root:Average loss=0.19734553897227997 +INFO:root: +[Valid][22][UTT][ MSE = 0.3082 | LCC = 0.7983 | SRCC = 0.8020 ] +INFO:root:Average loss=0.19028087361500814 +INFO:root: +[Valid][23][UTT][ MSE = 0.3157 | LCC = 0.8133 | SRCC = 0.8109 ] +INFO:root:Average loss=0.18345951384458786 +INFO:root: +[Valid][24][UTT][ MSE = 0.2954 | LCC = 0.8076 | SRCC = 0.8048 ] +INFO:root:Average loss=0.18578795152596939 +INFO:root: +[Valid][25][UTT][ MSE = 0.2917 | LCC = 0.8057 | SRCC = 0.8052 ] +INFO:root:Average loss=0.1845464595617392 +INFO:root: +[Valid][26][UTT][ MSE = 0.2918 | LCC = 0.8070 | SRCC = 0.8084 ] +INFO:root:Average loss=0.18337567704610336 +INFO:root: +[Valid][27][UTT][ MSE = 0.2933 | LCC = 0.8053 | SRCC = 0.8065 ] +INFO:root:Average loss=0.1665862864599778 +INFO:root: +[Valid][28][UTT][ MSE = 0.2984 | LCC = 0.8066 | SRCC = 0.8048 ] +INFO:root:Average loss=0.1627491615139521 +INFO:root: +[Valid][29][UTT][ MSE = 0.3059 | LCC = 0.8065 | SRCC = 0.8063 ] +INFO:root:Average loss=0.1600438393652439 +INFO:root: +[Valid][30][UTT][ MSE = 0.2988 | LCC = 0.8037 | SRCC = 0.8048 ] +INFO:root: +[Test][31][UTT][ MSE = 0.3339 | LCC = 0.8275 | SRCC = 0.8280 ] diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run3/config.gin b/results_bvcc/w2v2_large_asr_en/layer0/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run3/model.pt b/results_bvcc/w2v2_large_asr_en/layer0/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3009f3aed672ec1aacce3c8ab99e271d5a3f805a --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4cc985343a846741529393a1a0b9466ec6fef8bf27389eb9ea8e288d31a67eb8 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run3/model_best.pt b/results_bvcc/w2v2_large_asr_en/layer0/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9143b8213bc9ae2ba13f14f249bf9697f10e99a8 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1c3879721fdd6c49bef45d4c3aa5237c1130a96143b34261c04cc4c476466d77 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run3/train.log b/results_bvcc/w2v2_large_asr_en/layer0/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..213919dd9fbb494452799601990b36c9752cf3b9 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/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=0.6071857962852869 +INFO:root: +[Valid][0][UTT][ MSE = 0.4543 | LCC = 0.7029 | SRCC = 0.7023 ] +INFO:root:Average loss=0.39591913880446017 +INFO:root: +[Valid][1][UTT][ MSE = 0.4714 | LCC = 0.7151 | SRCC = 0.7123 ] +INFO:root:Average loss=0.36418489194833314 +INFO:root: +[Valid][2][UTT][ MSE = 0.4126 | LCC = 0.7438 | SRCC = 0.7390 ] +INFO:root:Average loss=0.3760197447278561 +INFO:root: +[Valid][3][UTT][ MSE = 0.4852 | LCC = 0.7255 | SRCC = 0.7272 ] +INFO:root:Average loss=0.3516629785299301 +INFO:root: +[Valid][4][UTT][ MSE = 0.3785 | LCC = 0.7574 | SRCC = 0.7561 ] +INFO:root:Average loss=0.3333591910508963 +INFO:root: +[Valid][5][UTT][ MSE = 0.3597 | LCC = 0.7656 | SRCC = 0.7614 ] +INFO:root:Average loss=0.319829975374234 +INFO:root: +[Valid][6][UTT][ MSE = 0.3746 | LCC = 0.7702 | SRCC = 0.7672 ] +INFO:root:Average loss=0.30363919872504014 +INFO:root: +[Valid][7][UTT][ MSE = 0.3830 | LCC = 0.7805 | SRCC = 0.7739 ] +INFO:root:Average loss=0.30854350271133274 +INFO:root: +[Valid][8][UTT][ MSE = 0.3531 | LCC = 0.7844 | SRCC = 0.7802 ] +INFO:root:Average loss=0.30288173525761336 +INFO:root: +[Valid][9][UTT][ MSE = 0.3289 | LCC = 0.7829 | SRCC = 0.7824 ] +INFO:root:Average loss=0.27822946107540375 +INFO:root: +[Valid][10][UTT][ MSE = 0.3461 | LCC = 0.7928 | SRCC = 0.7922 ] +INFO:root:Average loss=0.2768056115660912 +INFO:root: +[Valid][11][UTT][ MSE = 0.3398 | LCC = 0.7965 | SRCC = 0.7933 ] +INFO:root:Average loss=0.25503508803936153 +INFO:root: +[Valid][12][UTT][ MSE = 0.4221 | LCC = 0.7889 | SRCC = 0.7942 ] +INFO:root:Average loss=0.26114143144625884 +INFO:root: +[Valid][13][UTT][ MSE = 0.3600 | LCC = 0.7918 | SRCC = 0.7925 ] +INFO:root:Average loss=0.2619390196334093 +INFO:root: +[Valid][14][UTT][ MSE = 0.3724 | LCC = 0.7918 | SRCC = 0.7933 ] +INFO:root:Average loss=0.2511514759598634 +INFO:root: +[Valid][15][UTT][ MSE = 0.3376 | LCC = 0.8030 | SRCC = 0.8009 ] +INFO:root:Average loss=0.23227168906193513 +INFO:root: +[Valid][16][UTT][ MSE = 0.3648 | LCC = 0.8020 | SRCC = 0.8033 ] +INFO:root:Average loss=0.23541842974149263 +INFO:root: +[Valid][17][UTT][ MSE = 0.3103 | LCC = 0.8053 | SRCC = 0.8045 ] +INFO:root:Average loss=0.23131166532253608 +INFO:root: +[Valid][18][UTT][ MSE = 0.3174 | LCC = 0.8096 | SRCC = 0.8076 ] +INFO:root:Average loss=0.22092737276584673 +INFO:root: +[Valid][19][UTT][ MSE = 0.3119 | LCC = 0.8081 | SRCC = 0.8074 ] +INFO:root:Average loss=0.21470993470687133 +INFO:root: +[Valid][20][UTT][ MSE = 0.3142 | LCC = 0.8154 | SRCC = 0.8134 ] +INFO:root:Average loss=0.20683150213116255 +INFO:root: +[Valid][21][UTT][ MSE = 0.2958 | LCC = 0.8084 | SRCC = 0.8097 ] +INFO:root:Average loss=0.2068842853873204 +INFO:root: +[Valid][22][UTT][ MSE = 0.3132 | LCC = 0.8092 | SRCC = 0.8058 ] +INFO:root:Average loss=0.19372016516251442 +INFO:root: +[Valid][23][UTT][ MSE = 0.3034 | LCC = 0.8087 | SRCC = 0.8070 ] +INFO:root:Average loss=0.20426618021268111 +INFO:root: +[Valid][24][UTT][ MSE = 0.3654 | LCC = 0.8097 | SRCC = 0.8120 ] +INFO:root:Average loss=0.1868165736206067 +INFO:root: +[Valid][25][UTT][ MSE = 0.3245 | LCC = 0.8116 | SRCC = 0.8113 ] +INFO:root:Average loss=0.18301460051383728 +INFO:root: +[Valid][26][UTT][ MSE = 0.3052 | LCC = 0.8068 | SRCC = 0.8066 ] +INFO:root:Average loss=0.1875750587727779 +INFO:root: +[Valid][27][UTT][ MSE = 0.3133 | LCC = 0.8109 | SRCC = 0.8107 ] +INFO:root:Average loss=0.1788502690883783 +INFO:root: +[Valid][28][UTT][ MSE = 0.3348 | LCC = 0.8138 | SRCC = 0.8137 ] +INFO:root:Average loss=0.17097312269302514 +INFO:root: +[Valid][29][UTT][ MSE = 0.3062 | LCC = 0.8103 | SRCC = 0.8087 ] +INFO:root:Average loss=0.17076765554837692 +INFO:root: +[Valid][30][UTT][ MSE = 0.3428 | LCC = 0.8027 | SRCC = 0.7999 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2719 | LCC = 0.8382 | SRCC = 0.8378 ] diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run4/config.gin b/results_bvcc/w2v2_large_asr_en/layer0/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on BVCC dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +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 diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run4/model.pt b/results_bvcc/w2v2_large_asr_en/layer0/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bb455785d66a3c33c98e2a8f851a9ad877308c02 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:591f006efef4c6e56fe2ebfeb3685755e4004388c05753317a921177dc98607d +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run4/model_best.pt b/results_bvcc/w2v2_large_asr_en/layer0/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..02898204e257efd5820ada8aad554d86872522ba --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer0/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:57bfe46e5683f44ad90c9ac5578808883fb682e318002df17ba15baa330b89d6 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_en/layer0/run4/train.log b/results_bvcc/w2v2_large_asr_en/layer0/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..4e2da1b96c9308d99330d7783d152225916a9dd2 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/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=0.5526194045176873 +INFO:root: +[Valid][0][UTT][ MSE = 0.4378 | LCC = 0.6938 | SRCC = 0.6926 ] +INFO:root:Average loss=0.39802595533621615 +INFO:root: +[Valid][1][UTT][ MSE = 0.4338 | LCC = 0.7280 | SRCC = 0.7323 ] +INFO:root:Average loss=0.3685478503123308 +INFO:root: +[Valid][2][UTT][ MSE = 0.3943 | LCC = 0.7433 | SRCC = 0.7454 ] +INFO:root:Average loss=0.34969803881950867 +INFO:root: +[Valid][3][UTT][ MSE = 0.3769 | LCC = 0.7464 | SRCC = 0.7463 ] +INFO:root:Average loss=0.3349000402750113 +INFO:root: +[Valid][4][UTT][ MSE = 0.3565 | LCC = 0.7581 | SRCC = 0.7601 ] +INFO:root:Average loss=0.3195168538353382 +INFO:root: +[Valid][5][UTT][ MSE = 0.3653 | LCC = 0.7717 | SRCC = 0.7698 ] +INFO:root:Average loss=0.2976507823436688 +INFO:root: +[Valid][6][UTT][ MSE = 0.3518 | LCC = 0.7791 | SRCC = 0.7753 ] +INFO:root:Average loss=0.29990381174362624 +INFO:root: +[Valid][7][UTT][ MSE = 0.3660 | LCC = 0.7798 | SRCC = 0.7784 ] +INFO:root:Average loss=0.2730198397468298 +INFO:root: +[Valid][8][UTT][ MSE = 0.3226 | LCC = 0.7870 | SRCC = 0.7858 ] +INFO:root:Average loss=0.2725668379511589 +INFO:root: +[Valid][9][UTT][ MSE = 0.3231 | LCC = 0.7915 | SRCC = 0.7903 ] +INFO:root:Average loss=0.26630570013553667 +INFO:root: +[Valid][10][UTT][ MSE = 0.3281 | LCC = 0.7902 | SRCC = 0.7891 ] +INFO:root:Average loss=0.26178804211891615 +INFO:root: +[Valid][11][UTT][ MSE = 0.3089 | LCC = 0.7975 | SRCC = 0.7937 ] +INFO:root:Average loss=0.2448428051593976 +INFO:root: +[Valid][12][UTT][ MSE = 0.3054 | LCC = 0.8002 | SRCC = 0.8002 ] +INFO:root:Average loss=0.25383346183941913 +INFO:root: +[Valid][13][UTT][ MSE = 0.3384 | LCC = 0.8016 | SRCC = 0.7990 ] +INFO:root:Average loss=0.24660900082343665 +INFO:root: +[Valid][14][UTT][ MSE = 0.3003 | LCC = 0.8036 | SRCC = 0.8029 ] +INFO:root:Average loss=0.24557857177196404 +INFO:root: +[Valid][15][UTT][ MSE = 0.3020 | LCC = 0.8027 | SRCC = 0.8009 ] +INFO:root:Average loss=0.22149616861954713 +INFO:root: +[Valid][16][UTT][ MSE = 0.3133 | LCC = 0.8111 | SRCC = 0.8076 ] +INFO:root:Average loss=0.21785555856350142 +INFO:root: +[Valid][17][UTT][ MSE = 0.3502 | LCC = 0.8023 | SRCC = 0.8013 ] +INFO:root:Average loss=0.21072998814857924 +INFO:root: +[Valid][18][UTT][ MSE = 0.3065 | LCC = 0.8094 | SRCC = 0.8061 ] +INFO:root:Average loss=0.20408149025379083 +INFO:root: +[Valid][19][UTT][ MSE = 0.2922 | LCC = 0.8155 | SRCC = 0.8117 ] +INFO:root:Average loss=0.2061957061672822 +INFO:root: +[Valid][20][UTT][ MSE = 0.3031 | LCC = 0.8129 | SRCC = 0.8094 ] +INFO:root:Average loss=0.1998237119271205 +INFO:root: +[Valid][21][UTT][ MSE = 0.2921 | LCC = 0.8114 | SRCC = 0.8101 ] +INFO:root:Average loss=0.20242196856400904 +INFO:root: +[Valid][22][UTT][ MSE = 0.2882 | LCC = 0.8108 | SRCC = 0.8095 ] +INFO:root:Average loss=0.20007255491920006 +INFO:root: +[Valid][23][UTT][ MSE = 0.2984 | LCC = 0.8141 | SRCC = 0.8115 ] +INFO:root:Average loss=0.18236741566887268 +INFO:root: +[Valid][24][UTT][ MSE = 0.2851 | LCC = 0.8156 | SRCC = 0.8114 ] +INFO:root:Average loss=0.1795405632792375 +INFO:root: +[Valid][25][UTT][ MSE = 0.2797 | LCC = 0.8201 | SRCC = 0.8172 ] +INFO:root:Average loss=0.17148591969639826 +INFO:root: +[Valid][26][UTT][ MSE = 0.3007 | LCC = 0.8197 | SRCC = 0.8195 ] +INFO:root:Average loss=0.18045368016912386 +INFO:root: +[Valid][27][UTT][ MSE = 0.2894 | LCC = 0.8164 | SRCC = 0.8134 ] +INFO:root:Average loss=0.18569092815503097 +INFO:root: +[Valid][28][UTT][ MSE = 0.3031 | LCC = 0.8104 | SRCC = 0.8096 ] +INFO:root:Average loss=0.17143913329793856 +INFO:root: +[Valid][29][UTT][ MSE = 0.3194 | LCC = 0.8131 | SRCC = 0.8118 ] +INFO:root:Average loss=0.20602074953225943 +INFO:root: +[Valid][30][UTT][ MSE = 0.2925 | LCC = 0.8143 | SRCC = 0.8136 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2752 | LCC = 0.8349 | SRCC = 0.8318 ] diff --git a/results_bvcc/w2v2_large_asr_en/layer10/run0/config.gin b/results_bvcc/w2v2_large_asr_en/layer10/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/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 diff --git a/results_bvcc/w2v2_large_asr_en/layer10/run0/model.pt b/results_bvcc/w2v2_large_asr_en/layer10/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f1286e67c6db83c3d9295e3e23a83e2fb9ae5b8e --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer10/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0c7c4ace4220d31baa4090a8fb207459546f21830f12dcfab05c21487523e297 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_en/layer10/run0/model_best.pt b/results_bvcc/w2v2_large_asr_en/layer10/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4c80f8fda63cf35d0e1d4ae5c22480fae9ce329b --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer10/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:054d5f2ebbaaa21fe8fdf36cd23814cffe209c202e0a88cb558cff8655d766b2 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_en/layer10/run0/train.log b/results_bvcc/w2v2_large_asr_en/layer10/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..271155123a15f8e3be287dab0b1a8b5540867fb5 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/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=0.6028603418515279 +INFO:root: +[Valid][0][UTT][ MSE = 0.3968 | LCC = 0.7559 | SRCC = 0.7594 ] +INFO:root:Average loss=0.3520315127112927 +INFO:root: +[Valid][1][UTT][ MSE = 0.3346 | LCC = 0.7939 | SRCC = 0.7947 ] +INFO:root:Average loss=0.28907438291188997 +INFO:root: +[Valid][2][UTT][ MSE = 0.3208 | LCC = 0.8110 | SRCC = 0.8095 ] +INFO:root:Average loss=0.2698086630075406 +INFO:root: +[Valid][3][UTT][ MSE = 0.3140 | LCC = 0.8136 | SRCC = 0.8119 ] +INFO:root:Average loss=0.23878259441027275 +INFO:root: +[Valid][4][UTT][ MSE = 0.3152 | LCC = 0.8166 | SRCC = 0.8131 ] +INFO:root:Average loss=0.22693349497440535 +INFO:root: +[Valid][5][UTT][ MSE = 0.2787 | LCC = 0.8203 | SRCC = 0.8181 ] +INFO:root:Average loss=0.21102316352801445 +INFO:root: +[Valid][6][UTT][ MSE = 0.4050 | LCC = 0.8105 | SRCC = 0.8159 ] +INFO:root:Average loss=0.1996055112626308 +INFO:root: +[Valid][7][UTT][ MSE = 0.3164 | LCC = 0.8180 | SRCC = 0.8150 ] +INFO:root:Average loss=0.18182895189294448 +INFO:root: +[Valid][8][UTT][ MSE = 0.3002 | LCC = 0.8153 | SRCC = 0.8125 ] +INFO:root:Average loss=0.17122940957928315 +INFO:root: +[Valid][9][UTT][ MSE = 0.3077 | LCC = 0.8088 | SRCC = 0.8055 ] +INFO:root:Average loss=0.1579790480243854 +INFO:root: +[Valid][10][UTT][ MSE = 0.2871 | LCC = 0.8194 | SRCC = 0.8155 ] +INFO:root:Average loss=0.15095182087941048 +INFO:root: +[Valid][11][UTT][ MSE = 0.2885 | LCC = 0.8109 | SRCC = 0.8081 ] +INFO:root:Average loss=0.13404934901075485 +INFO:root: +[Valid][12][UTT][ MSE = 0.2836 | LCC = 0.8129 | SRCC = 0.8110 ] +INFO:root:Average loss=0.132908073755411 +INFO:root: +[Valid][13][UTT][ MSE = 0.2856 | LCC = 0.8126 | SRCC = 0.8102 ] +INFO:root:Average loss=0.12902581706069982 +INFO:root: +[Valid][14][UTT][ MSE = 0.3192 | LCC = 0.8134 | SRCC = 0.8113 ] +INFO:root:Average loss=0.11546467149104828 +INFO:root: +[Valid][15][UTT][ MSE = 0.2979 | LCC = 0.8055 | SRCC = 0.8042 ] +INFO:root:Average loss=0.10495919949160172 +INFO:root: +[Valid][16][UTT][ MSE = 0.3032 | LCC = 0.8011 | SRCC = 0.8011 ] +INFO:root:Average loss=0.10205904088723354 +INFO:root: +[Valid][17][UTT][ MSE = 0.2981 | LCC = 0.8026 | SRCC = 0.8006 ] +INFO:root:Average loss=0.09480657547903366 +INFO:root: +[Valid][18][UTT][ MSE = 0.3003 | LCC = 0.8030 | SRCC = 0.8005 ] +INFO:root:Average loss=0.08873219606585991 +INFO:root: +[Valid][19][UTT][ MSE = 0.3022 | LCC = 0.8093 | SRCC = 0.8052 ] +INFO:root:Average loss=0.09181962744929852 +INFO:root: +[Valid][20][UTT][ MSE = 0.2981 | LCC = 0.8027 | SRCC = 0.8011 ] +INFO:root:Average loss=0.08021729410840915 +INFO:root: +[Valid][21][UTT][ MSE = 0.3361 | LCC = 0.8015 | SRCC = 0.8004 ] +INFO:root:Average loss=0.07694346849352886 +INFO:root: +[Valid][22][UTT][ MSE = 0.3387 | LCC = 0.8026 | SRCC = 0.8014 ] +INFO:root:Average loss=0.07925359384180644 +INFO:root: +[Valid][23][UTT][ MSE = 0.3031 | LCC = 0.8011 | SRCC = 0.7982 ] +INFO:root:Average loss=0.07065363443241669 +INFO:root: +[Valid][24][UTT][ MSE = 0.3008 | LCC = 0.8010 | SRCC = 0.7986 ] +INFO:root:Average loss=0.06920202501500264 +INFO:root: +[Valid][25][UTT][ MSE = 0.3116 | LCC = 0.7961 | SRCC = 0.7938 ] +INFO:root:Average loss=0.06347758394594376 +INFO:root: +[Valid][26][UTT][ MSE = 0.3019 | LCC = 0.8024 | SRCC = 0.8004 ] +INFO:root:Average loss=0.06204031694393892 +INFO:root: +[Valid][27][UTT][ MSE = 0.2956 | LCC = 0.8056 | SRCC = 0.8025 ] +INFO:root:Average loss=0.06325010759517169 +INFO:root: +[Valid][28][UTT][ MSE = 0.3050 | LCC = 0.7984 | SRCC = 0.7954 ] +INFO:root:Average loss=0.061374314630833954 +INFO:root: +[Valid][29][UTT][ MSE = 0.3021 | LCC = 0.8046 | SRCC = 0.8014 ] +INFO:root:Average loss=0.054543173943574615 +INFO:root: +[Valid][30][UTT][ MSE = 0.3030 | LCC = 0.7991 | SRCC = 0.7964 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2802 | LCC = 0.8260 | SRCC = 0.8300 ] diff --git a/results_bvcc/w2v2_large_asr_en/layer10/run1/config.gin b/results_bvcc/w2v2_large_asr_en/layer10/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/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 diff --git a/results_bvcc/w2v2_large_asr_en/layer10/run1/model.pt b/results_bvcc/w2v2_large_asr_en/layer10/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3539f043a082bc8f2ad585b56e6477285f86f353 --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer10/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a596737427dfefc9a1139e8e26e18fe9032ba27c574a6d7ba4a8f50a49c10453 +size 3403339 diff --git a/results_bvcc/w2v2_large_asr_en/layer10/run1/model_best.pt b/results_bvcc/w2v2_large_asr_en/layer10/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..df3e812286615127e4922fccc1e79b97700c277c --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/layer10/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f222e029e53b261dee435fba3bbe43a6d8a6432fc043f4f39c8e5ae6d52fcc77 +size 3404998 diff --git a/results_bvcc/w2v2_large_asr_en/layer10/run1/train.log b/results_bvcc/w2v2_large_asr_en/layer10/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..62276c4ffa6af24fb1ef1335f8904d7eade71a8d --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/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=0.657606031650152 +INFO:root: +[Valid][0][UTT][ MSE = 0.4171 | LCC = 0.7364 | SRCC = 0.7391 ] +INFO:root:Average loss=0.3388948971644426 +INFO:root: +[Valid][1][UTT][ MSE = 0.3730 | LCC = 0.7996 | SRCC = 0.7979 ] +INFO:root:Average loss=0.2880323133789576 +INFO:root: +[Valid][2][UTT][ MSE = 0.3440 | LCC = 0.8121 | SRCC = 0.8122 ] +INFO:root:Average loss=0.25945473443239164 +INFO:root: +[Valid][3][UTT][ MSE = 0.3366 | LCC = 0.8157 | SRCC = 0.8149 ] +INFO:root:Average loss=0.23618142899030295 +INFO:root: +[Valid][4][UTT][ MSE = 0.3230 | LCC = 0.8191 | SRCC = 0.8183 ] +INFO:root:Average loss=0.21676802042967233 +INFO:root: +[Valid][5][UTT][ MSE = 0.3234 | LCC = 0.8158 | SRCC = 0.8138 ] +INFO:root:Average loss=0.20414516539909902 +INFO:root: +[Valid][6][UTT][ MSE = 0.3748 | LCC = 0.8093 | SRCC = 0.8135 ] +INFO:root:Average loss=0.1912420070133148 +INFO:root: +[Valid][7][UTT][ MSE = 0.3084 | LCC = 0.8133 | SRCC = 0.8129 ] +INFO:root:Average loss=0.17680748322835335 +INFO:root: +[Valid][8][UTT][ MSE = 0.2942 | LCC = 0.8131 | SRCC = 0.8115 ] +INFO:root:Average loss=0.17743936228828552 +INFO:root: +[Valid][9][UTT][ MSE = 0.2981 | LCC = 0.8143 | SRCC = 0.8139 ] +INFO:root:Average loss=0.16456265413226226 +INFO:root: +[Valid][10][UTT][ MSE = 0.3066 | LCC = 0.8096 | SRCC = 0.8069 ] +INFO:root:Average loss=0.15084278870087403 +INFO:root: +[Valid][11][UTT][ MSE = 0.3924 | LCC = 0.8060 | SRCC = 0.8120 ] +INFO:root:Average loss=0.14443998143841058 +INFO:root: +[Valid][12][UTT][ MSE = 0.3106 | LCC = 0.8101 | SRCC = 0.8098 ] +INFO:root:Average loss=0.1236291761772755 +INFO:root: +[Valid][13][UTT][ MSE = 0.3083 | LCC = 0.8057 | SRCC = 0.8050 ] +INFO:root:Average loss=0.13151927946660763 +INFO:root: +[Valid][14][UTT][ MSE = 0.2922 | LCC = 0.8080 | SRCC = 0.8060 ] +INFO:root:Average loss=0.1170132760054026 +INFO:root: +[Valid][15][UTT][ MSE = 0.2890 | LCC = 0.8123 | SRCC = 0.8096 ] +INFO:root:Average loss=0.10568319920163888 +INFO:root: +[Valid][16][UTT][ MSE = 0.3093 | LCC = 0.8012 | SRCC = 0.7989 ] +INFO:root:Average loss=0.10586166090499131 +INFO:root: +[Valid][17][UTT][ MSE = 0.2963 | LCC = 0.8055 | SRCC = 0.8034 ] +INFO:root:Average loss=0.10226922167035249 +INFO:root: +[Valid][18][UTT][ MSE = 0.2845 | LCC = 0.8135 | SRCC = 0.8120 ] +INFO:root:Average loss=0.09451565241966492 +INFO:root: +[Valid][19][UTT][ MSE = 0.3123 | LCC = 0.8015 | SRCC = 0.8003 ] +INFO:root:Average loss=0.09655640145333913 +INFO:root: +[Valid][20][UTT][ MSE = 0.3118 | LCC = 0.8045 | SRCC = 0.8037 ] +INFO:root:Average loss=0.09106795590084332 +INFO:root: +[Valid][21][UTT][ MSE = 0.4069 | LCC = 0.7980 | SRCC = 0.8028 ] +INFO:root:Average loss=0.07974360396082585 +INFO:root: +[Valid][22][UTT][ MSE = 0.3310 | LCC = 0.8022 | SRCC = 0.8030 ] +INFO:root:Average loss=0.07698451082866925 +INFO:root: +[Valid][23][UTT][ MSE = 0.3163 | LCC = 0.7987 | SRCC = 0.7972 ] +INFO:root:Average loss=0.07039723970378056 +INFO:root: +[Valid][24][UTT][ MSE = 0.3070 | LCC = 0.8012 | SRCC = 0.7988 ] +INFO:root:Average loss=0.072444185662346 +INFO:root: +[Valid][25][UTT][ MSE = 0.3002 | LCC = 0.8030 | SRCC = 0.8007 ] +INFO:root:Average loss=0.06731770006127846 +INFO:root: +[Valid][26][UTT][ MSE = 0.3050 | LCC = 0.8017 | SRCC = 0.8002 ] +INFO:root:Average loss=0.06315959760775933 +INFO:root: +[Valid][27][UTT][ MSE = 0.3034 | LCC = 0.8104 | SRCC = 0.8103 ] +INFO:root:Average loss=0.062450845630314104 +INFO:root: +[Valid][28][UTT][ MSE = 0.3214 | LCC = 0.8017 | SRCC = 0.8012 ] +INFO:root:Average loss=0.06026032361655663 +INFO:root: +[Valid][29][UTT][ MSE = 0.3820 | LCC = 0.7959 | SRCC = 0.7977 ] +INFO:root:Average loss=0.06848886102820054 +INFO:root: +[Valid][30][UTT][ MSE = 0.3366 | LCC = 0.7990 | SRCC = 0.7991 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2831 | LCC = 0.8307 | SRCC = 0.8312 ] diff --git a/results_bvcc/w2v2_large_asr_en/layer10/run2/config.gin b/results_bvcc/w2v2_large_asr_en/layer10/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..d36a6964ab640c490fd7442e2d46a6fbe5fc84aa --- /dev/null +++ b/results_bvcc/w2v2_large_asr_en/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