diff --git a/results_tencent/w2v2_large_asr_en/layer0/run0/model.pt b/results_tencent/w2v2_large_asr_en/layer0/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..63aa0e0ff8bb06fab636d594f4022e2ccd0bd7fc --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f44dee60f861cfe5dde45b2487f4a9158d7d0225ce4f8e6a280877f6f168acf3 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer0/run0/model_best.pt b/results_tencent/w2v2_large_asr_en/layer0/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..211c00375cb872b6430ba555f0d9d37049f3be59 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:58a78ba6675856841641c2e13a55ee3fb0acf2a01be27ebdf6e0bc75c85e9155 +size 3404998 diff --git a/results_tencent/w2v2_large_asr_en/layer0/run0/train.log b/results_tencent/w2v2_large_asr_en/layer0/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..000bd3ad806d46659a92b9ce14e7f45c22bc704f --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6649895780086518 +INFO:root: +[Valid][0][UTT][ MSE = 0.5021 | LCC = 0.8662 | SRCC = 0.8684 ] +INFO:root:Average loss=0.36423822593688965 +INFO:root: +[Valid][1][UTT][ MSE = 0.3618 | LCC = 0.8960 | SRCC = 0.8962 ] +INFO:root:Average loss=0.34296851766109465 +INFO:root: +[Valid][2][UTT][ MSE = 0.3343 | LCC = 0.9080 | SRCC = 0.9090 ] +INFO:root:Average loss=0.2726272573471069 +INFO:root: +[Valid][3][UTT][ MSE = 0.3904 | LCC = 0.9064 | SRCC = 0.9085 ] +INFO:root:Average loss=0.2669645261764526 +INFO:root: +[Valid][4][UTT][ MSE = 0.3021 | LCC = 0.9187 | SRCC = 0.9204 ] +INFO:root:Average loss=0.2269453490972519 +INFO:root: +[Valid][5][UTT][ MSE = 0.2834 | LCC = 0.9185 | SRCC = 0.9199 ] +INFO:root:Average loss=0.21508347779512405 +INFO:root: +[Valid][6][UTT][ MSE = 0.3365 | LCC = 0.9255 | SRCC = 0.9266 ] +INFO:root:Average loss=0.207648442029953 +INFO:root: +[Valid][7][UTT][ MSE = 0.2773 | LCC = 0.9282 | SRCC = 0.9301 ] +INFO:root:Average loss=0.18569231790304183 +INFO:root: +[Valid][8][UTT][ MSE = 0.2999 | LCC = 0.9301 | SRCC = 0.9312 ] +INFO:root:Average loss=0.17414556181430815 +INFO:root: +[Valid][9][UTT][ MSE = 0.2337 | LCC = 0.9339 | SRCC = 0.9356 ] +INFO:root:Average loss=0.1659562236070633 +INFO:root: +[Valid][10][UTT][ MSE = 0.2474 | LCC = 0.9327 | SRCC = 0.9336 ] +INFO:root:Average loss=0.158993832051754 +INFO:root: +[Valid][11][UTT][ MSE = 0.2262 | LCC = 0.9356 | SRCC = 0.9357 ] +INFO:root:Average loss=0.15157693582773207 +INFO:root: +[Valid][12][UTT][ MSE = 0.2717 | LCC = 0.9352 | SRCC = 0.9362 ] +INFO:root:Average loss=0.1491793937087059 +INFO:root: +[Valid][13][UTT][ MSE = 0.2489 | LCC = 0.9392 | SRCC = 0.9401 ] +INFO:root:Average loss=0.1365955706834793 +INFO:root: +[Valid][14][UTT][ MSE = 0.2534 | LCC = 0.9349 | SRCC = 0.9373 ] +INFO:root:Average loss=0.13709089648723602 +INFO:root: +[Valid][15][UTT][ MSE = 0.2557 | LCC = 0.9383 | SRCC = 0.9394 ] +INFO:root:Average loss=0.12561885887384414 +INFO:root: +[Valid][16][UTT][ MSE = 0.2210 | LCC = 0.9391 | SRCC = 0.9391 ] +INFO:root:Average loss=0.12346147406101227 +INFO:root: +[Valid][17][UTT][ MSE = 0.2727 | LCC = 0.9422 | SRCC = 0.9433 ] +INFO:root:Average loss=0.1230960151553154 +INFO:root: +[Valid][18][UTT][ MSE = 0.2270 | LCC = 0.9409 | SRCC = 0.9413 ] +INFO:root:Average loss=0.11931769216060638 +INFO:root: +[Valid][19][UTT][ MSE = 0.2474 | LCC = 0.9398 | SRCC = 0.9401 ] +INFO:root:Average loss=0.11049323427677155 +INFO:root: +[Valid][20][UTT][ MSE = 0.2099 | LCC = 0.9368 | SRCC = 0.9388 ] +INFO:root:Average loss=0.10819276887178421 +INFO:root: +[Valid][21][UTT][ MSE = 0.2211 | LCC = 0.9432 | SRCC = 0.9435 ] +INFO:root:Average loss=0.1075104551911354 +INFO:root: +[Valid][22][UTT][ MSE = 0.2665 | LCC = 0.9410 | SRCC = 0.9428 ] +INFO:root:Average loss=0.10115887171030044 +INFO:root: +[Valid][23][UTT][ MSE = 0.2205 | LCC = 0.9448 | SRCC = 0.9444 ] +INFO:root:Average loss=0.09877768251299858 +INFO:root: +[Valid][24][UTT][ MSE = 0.2133 | LCC = 0.9466 | SRCC = 0.9471 ] +INFO:root:Average loss=0.0910660448372364 +INFO:root: +[Valid][25][UTT][ MSE = 0.2269 | LCC = 0.9449 | SRCC = 0.9449 ] +INFO:root:Average loss=0.08806130051612854 +INFO:root: +[Valid][26][UTT][ MSE = 0.2242 | LCC = 0.9458 | SRCC = 0.9458 ] +INFO:root:Average loss=0.08847611147165299 +INFO:root: +[Valid][27][UTT][ MSE = 0.2219 | LCC = 0.9423 | SRCC = 0.9422 ] +INFO:root:Average loss=0.08456234943866729 +INFO:root: +[Valid][28][UTT][ MSE = 0.2153 | LCC = 0.9445 | SRCC = 0.9442 ] +INFO:root:Average loss=0.08081513851881027 +INFO:root: +[Valid][29][UTT][ MSE = 0.2277 | LCC = 0.9467 | SRCC = 0.9470 ] +INFO:root:Average loss=0.08008565375208855 +INFO:root: +[Valid][30][UTT][ MSE = 0.2023 | LCC = 0.9461 | SRCC = 0.9467 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2323 | LCC = 0.9470 | SRCC = 0.9481 ] diff --git a/results_tencent/w2v2_large_asr_en/layer0/run1/config.gin b/results_tencent/w2v2_large_asr_en/layer0/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer0/run1/model.pt b/results_tencent/w2v2_large_asr_en/layer0/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2a04063bb0a59d2fb8cfa2f526df8b11f43c272e --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:621abede2a5b23eb12d984e3d8d5ed7f6c63e4b51f8732f2562cc3b51c33b469 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer0/run1/model_best.pt b/results_tencent/w2v2_large_asr_en/layer0/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c55f9251b7590ad738313a3334b2630bc88fb90d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e72a903b536b993be1ef9291259bb1d2c32406d78f44746ab26cf5d6e5b6f116 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer0/run1/train.log b/results_tencent/w2v2_large_asr_en/layer0/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..dad832789bc015331fa374184b81be34b0beb24d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5740894581079483 +INFO:root: +[Valid][0][UTT][ MSE = 0.3895 | LCC = 0.8815 | SRCC = 0.8794 ] +INFO:root:Average loss=0.33942470848560335 +INFO:root: +[Valid][1][UTT][ MSE = 0.3404 | LCC = 0.9064 | SRCC = 0.9070 ] +INFO:root:Average loss=0.3118090524673462 +INFO:root: +[Valid][2][UTT][ MSE = 0.3112 | LCC = 0.9094 | SRCC = 0.9135 ] +INFO:root:Average loss=0.2815272874832153 +INFO:root: +[Valid][3][UTT][ MSE = 0.3010 | LCC = 0.9181 | SRCC = 0.9195 ] +INFO:root:Average loss=0.24935106205940247 +INFO:root: +[Valid][4][UTT][ MSE = 0.2845 | LCC = 0.9242 | SRCC = 0.9261 ] +INFO:root:Average loss=0.21995245516300202 +INFO:root: +[Valid][5][UTT][ MSE = 0.2458 | LCC = 0.9232 | SRCC = 0.9250 ] +INFO:root:Average loss=0.2059981990456581 +INFO:root: +[Valid][6][UTT][ MSE = 0.2909 | LCC = 0.9264 | SRCC = 0.9283 ] +INFO:root:Average loss=0.188407022356987 +INFO:root: +[Valid][7][UTT][ MSE = 0.2773 | LCC = 0.9291 | SRCC = 0.9319 ] +INFO:root:Average loss=0.18127622067928315 +INFO:root: +[Valid][8][UTT][ MSE = 0.2528 | LCC = 0.9309 | SRCC = 0.9322 ] +INFO:root:Average loss=0.1760287303328514 +INFO:root: +[Valid][9][UTT][ MSE = 0.2519 | LCC = 0.9310 | SRCC = 0.9342 ] +INFO:root:Average loss=0.17421934288740157 +INFO:root: +[Valid][10][UTT][ MSE = 0.2777 | LCC = 0.9329 | SRCC = 0.9346 ] +INFO:root:Average loss=0.16470972645282744 +INFO:root: +[Valid][11][UTT][ MSE = 0.2320 | LCC = 0.9374 | SRCC = 0.9389 ] +INFO:root:Average loss=0.15133613926172257 +INFO:root: +[Valid][12][UTT][ MSE = 0.2863 | LCC = 0.9390 | SRCC = 0.9406 ] +INFO:root:Average loss=0.14636920660734176 +INFO:root: +[Valid][13][UTT][ MSE = 0.2580 | LCC = 0.9366 | SRCC = 0.9379 ] +INFO:root:Average loss=0.13522404074668884 +INFO:root: +[Valid][14][UTT][ MSE = 0.2360 | LCC = 0.9392 | SRCC = 0.9405 ] +INFO:root:Average loss=0.1288481270670891 +INFO:root: +[Valid][15][UTT][ MSE = 0.2471 | LCC = 0.9382 | SRCC = 0.9410 ] +INFO:root:Average loss=0.12814946675300598 +INFO:root: +[Valid][16][UTT][ MSE = 0.2587 | LCC = 0.9423 | SRCC = 0.9434 ] +INFO:root:Average loss=0.12289551103115082 +INFO:root: +[Valid][17][UTT][ MSE = 0.2572 | LCC = 0.9385 | SRCC = 0.9405 ] +INFO:root:Average loss=0.11703359311819077 +INFO:root: +[Valid][18][UTT][ MSE = 0.2786 | LCC = 0.9360 | SRCC = 0.9370 ] +INFO:root:Average loss=0.11796789079904556 +INFO:root: +[Valid][19][UTT][ MSE = 0.2052 | LCC = 0.9372 | SRCC = 0.9407 ] +INFO:root:Average loss=0.11379064375162125 +INFO:root: +[Valid][20][UTT][ MSE = 0.2206 | LCC = 0.9382 | SRCC = 0.9385 ] +INFO:root:Average loss=0.1046984760761261 +INFO:root: +[Valid][21][UTT][ MSE = 0.2565 | LCC = 0.9438 | SRCC = 0.9440 ] +INFO:root:Average loss=0.10058193561434746 +INFO:root: +[Valid][22][UTT][ MSE = 0.2175 | LCC = 0.9445 | SRCC = 0.9452 ] +INFO:root:Average loss=0.09745386728644372 +INFO:root: +[Valid][23][UTT][ MSE = 0.2305 | LCC = 0.9428 | SRCC = 0.9440 ] +INFO:root:Average loss=0.09204903349280358 +INFO:root: +[Valid][24][UTT][ MSE = 0.2314 | LCC = 0.9466 | SRCC = 0.9471 ] +INFO:root:Average loss=0.09020958796143531 +INFO:root: +[Valid][25][UTT][ MSE = 0.2185 | LCC = 0.9457 | SRCC = 0.9468 ] +INFO:root:Average loss=0.08709004393219948 +INFO:root: +[Valid][26][UTT][ MSE = 0.2193 | LCC = 0.9459 | SRCC = 0.9458 ] +INFO:root:Average loss=0.08658751103281975 +INFO:root: +[Valid][27][UTT][ MSE = 0.2141 | LCC = 0.9461 | SRCC = 0.9466 ] +INFO:root:Average loss=0.08099481245875359 +INFO:root: +[Valid][28][UTT][ MSE = 0.2104 | LCC = 0.9446 | SRCC = 0.9444 ] +INFO:root:Average loss=0.07930216109752655 +INFO:root: +[Valid][29][UTT][ MSE = 0.2238 | LCC = 0.9449 | SRCC = 0.9455 ] +INFO:root:Average loss=0.07841939279437066 +INFO:root: +[Valid][30][UTT][ MSE = 0.2343 | LCC = 0.9450 | SRCC = 0.9454 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2431 | LCC = 0.9452 | SRCC = 0.9466 ] diff --git a/results_tencent/w2v2_large_asr_en/layer0/run2/config.gin b/results_tencent/w2v2_large_asr_en/layer0/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer0/run2/model.pt b/results_tencent/w2v2_large_asr_en/layer0/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..63b20307849c48f580c6161f705db77662d3ec62 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5b955b930d0ac4b0f88820c999998baac16d6ddb8928503c3702f5b0a1274903 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer0/run2/model_best.pt b/results_tencent/w2v2_large_asr_en/layer0/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..8fdfe95eea4702bdccfa97f4b5774df02036db5a --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4ca61e766ae81e8f00c63c2bb5293bd7f2725ec3cf8356b5292527dc04b888e8 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer0/run2/train.log b/results_tencent/w2v2_large_asr_en/layer0/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0f8e5d507b31bdcb8c257e7606a34f04d393fe74 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.709221761226654 +INFO:root: +[Valid][0][UTT][ MSE = 0.4366 | LCC = 0.8583 | SRCC = 0.8577 ] +INFO:root:Average loss=0.3729499524831772 +INFO:root: +[Valid][1][UTT][ MSE = 0.4500 | LCC = 0.8869 | SRCC = 0.8869 ] +INFO:root:Average loss=0.325014719247818 +INFO:root: +[Valid][2][UTT][ MSE = 0.3388 | LCC = 0.9091 | SRCC = 0.9101 ] +INFO:root:Average loss=0.2722231197357178 +INFO:root: +[Valid][3][UTT][ MSE = 0.3134 | LCC = 0.9185 | SRCC = 0.9205 ] +INFO:root:Average loss=0.2466454672217369 +INFO:root: +[Valid][4][UTT][ MSE = 0.2878 | LCC = 0.9220 | SRCC = 0.9232 ] +INFO:root:Average loss=0.2227160667181015 +INFO:root: +[Valid][5][UTT][ MSE = 0.2816 | LCC = 0.9283 | SRCC = 0.9294 ] +INFO:root:Average loss=0.19683788174390793 +INFO:root: +[Valid][6][UTT][ MSE = 0.2595 | LCC = 0.9282 | SRCC = 0.9294 ] +INFO:root:Average loss=0.18636493563652037 +INFO:root: +[Valid][7][UTT][ MSE = 0.2931 | LCC = 0.9303 | SRCC = 0.9313 ] +INFO:root:Average loss=0.17900047713518144 +INFO:root: +[Valid][8][UTT][ MSE = 0.2693 | LCC = 0.9297 | SRCC = 0.9300 ] +INFO:root:Average loss=0.16832722479104995 +INFO:root: +[Valid][9][UTT][ MSE = 0.2727 | LCC = 0.9351 | SRCC = 0.9361 ] +INFO:root:Average loss=0.16136809945106506 +INFO:root: +[Valid][10][UTT][ MSE = 0.2506 | LCC = 0.9382 | SRCC = 0.9386 ] +INFO:root:Average loss=0.1498782126903534 +INFO:root: +[Valid][11][UTT][ MSE = 0.2281 | LCC = 0.9387 | SRCC = 0.9392 ] +INFO:root:Average loss=0.13958915346860887 +INFO:root: +[Valid][12][UTT][ MSE = 0.2393 | LCC = 0.9368 | SRCC = 0.9372 ] +INFO:root:Average loss=0.13631629371643067 +INFO:root: +[Valid][13][UTT][ MSE = 0.2315 | LCC = 0.9338 | SRCC = 0.9358 ] +INFO:root:Average loss=0.13064119237661362 +INFO:root: +[Valid][14][UTT][ MSE = 0.2396 | LCC = 0.9422 | SRCC = 0.9424 ] +INFO:root:Average loss=0.1272465716600418 +INFO:root: +[Valid][15][UTT][ MSE = 0.2452 | LCC = 0.9417 | SRCC = 0.9423 ] +INFO:root:Average loss=0.12118836542963982 +INFO:root: +[Valid][16][UTT][ MSE = 0.2440 | LCC = 0.9349 | SRCC = 0.9366 ] +INFO:root:Average loss=0.11433009663224221 +INFO:root: +[Valid][17][UTT][ MSE = 0.2635 | LCC = 0.9389 | SRCC = 0.9398 ] +INFO:root:Average loss=0.11710317665338517 +INFO:root: +[Valid][18][UTT][ MSE = 0.2165 | LCC = 0.9407 | SRCC = 0.9409 ] +INFO:root:Average loss=0.10485217961668968 +INFO:root: +[Valid][19][UTT][ MSE = 0.2331 | LCC = 0.9427 | SRCC = 0.9429 ] +INFO:root:Average loss=0.1080205430984497 +INFO:root: +[Valid][20][UTT][ MSE = 0.2403 | LCC = 0.9436 | SRCC = 0.9431 ] +INFO:root:Average loss=0.10018497738242149 +INFO:root: +[Valid][21][UTT][ MSE = 0.2411 | LCC = 0.9449 | SRCC = 0.9452 ] +INFO:root:Average loss=0.09613898831605912 +INFO:root: +[Valid][22][UTT][ MSE = 0.2187 | LCC = 0.9426 | SRCC = 0.9430 ] +INFO:root:Average loss=0.09334326723217964 +INFO:root: +[Valid][23][UTT][ MSE = 0.2130 | LCC = 0.9405 | SRCC = 0.9414 ] +INFO:root:Average loss=0.09391682940721512 +INFO:root: +[Valid][24][UTT][ MSE = 0.2603 | LCC = 0.9421 | SRCC = 0.9423 ] +INFO:root:Average loss=0.08849663078784943 +INFO:root: +[Valid][25][UTT][ MSE = 0.2087 | LCC = 0.9442 | SRCC = 0.9436 ] +INFO:root:Average loss=0.08483491101861 +INFO:root: +[Valid][26][UTT][ MSE = 0.2210 | LCC = 0.9464 | SRCC = 0.9460 ] +INFO:root:Average loss=0.08490307739377022 +INFO:root: +[Valid][27][UTT][ MSE = 0.2096 | LCC = 0.9458 | SRCC = 0.9456 ] +INFO:root:Average loss=0.07878244167566299 +INFO:root: +[Valid][28][UTT][ MSE = 0.2280 | LCC = 0.9454 | SRCC = 0.9449 ] +INFO:root:Average loss=0.08274864104390145 +INFO:root: +[Valid][29][UTT][ MSE = 0.2225 | LCC = 0.9455 | SRCC = 0.9447 ] +INFO:root:Average loss=0.07389444380998611 +INFO:root: +[Valid][30][UTT][ MSE = 0.2241 | LCC = 0.9453 | SRCC = 0.9458 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2283 | LCC = 0.9462 | SRCC = 0.9462 ] diff --git a/results_tencent/w2v2_large_asr_en/layer0/run3/config.gin b/results_tencent/w2v2_large_asr_en/layer0/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer0/run3/model.pt b/results_tencent/w2v2_large_asr_en/layer0/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..70084f75968e0fd10bc764cf46d5cd9cda265c59 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cfe4b7e7983d56b8b8372e210b298f660ea2da87d54f68abeb9a82b17460e939 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer0/run3/model_best.pt b/results_tencent/w2v2_large_asr_en/layer0/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..248336a7ee9b4b0f5a9654b58c34750f7d03ef63 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:467a0b3bd794b8d3f9f30addb265f8eba0d9eefe981c70e4bf1fa0890444f6d5 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer0/run3/train.log b/results_tencent/w2v2_large_asr_en/layer0/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..89c4b19f1a3be89cd498af6edf5a4a2c82c3f0c6 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.618389678478241 +INFO:root: +[Valid][0][UTT][ MSE = 0.4107 | LCC = 0.8755 | SRCC = 0.8745 ] +INFO:root:Average loss=0.35121494472026826 +INFO:root: +[Valid][1][UTT][ MSE = 0.3838 | LCC = 0.8972 | SRCC = 0.9007 ] +INFO:root:Average loss=0.2892568988800049 +INFO:root: +[Valid][2][UTT][ MSE = 0.3334 | LCC = 0.9132 | SRCC = 0.9137 ] +INFO:root:Average loss=0.2573729250431061 +INFO:root: +[Valid][3][UTT][ MSE = 0.3814 | LCC = 0.9068 | SRCC = 0.9069 ] +INFO:root:Average loss=0.23396009427309036 +INFO:root: +[Valid][4][UTT][ MSE = 0.2855 | LCC = 0.9237 | SRCC = 0.9248 ] +INFO:root:Average loss=0.2149351971745491 +INFO:root: +[Valid][5][UTT][ MSE = 0.3038 | LCC = 0.9195 | SRCC = 0.9217 ] +INFO:root:Average loss=0.20843857795000076 +INFO:root: +[Valid][6][UTT][ MSE = 0.3273 | LCC = 0.9234 | SRCC = 0.9271 ] +INFO:root:Average loss=0.19983563405275345 +INFO:root: +[Valid][7][UTT][ MSE = 0.2876 | LCC = 0.9341 | SRCC = 0.9357 ] +INFO:root:Average loss=0.18420539343357087 +INFO:root: +[Valid][8][UTT][ MSE = 0.3130 | LCC = 0.9320 | SRCC = 0.9338 ] +INFO:root:Average loss=0.17544530963897706 +INFO:root: +[Valid][9][UTT][ MSE = 0.2366 | LCC = 0.9367 | SRCC = 0.9381 ] +INFO:root:Average loss=0.15911000841856002 +INFO:root: +[Valid][10][UTT][ MSE = 0.2164 | LCC = 0.9380 | SRCC = 0.9393 ] +INFO:root:Average loss=0.14822474956512452 +INFO:root: +[Valid][11][UTT][ MSE = 0.2618 | LCC = 0.9359 | SRCC = 0.9368 ] +INFO:root:Average loss=0.15723540180921555 +INFO:root: +[Valid][12][UTT][ MSE = 0.2825 | LCC = 0.9366 | SRCC = 0.9376 ] +INFO:root:Average loss=0.13719176137447356 +INFO:root: +[Valid][13][UTT][ MSE = 0.2436 | LCC = 0.9390 | SRCC = 0.9391 ] +INFO:root:Average loss=0.13555410206317903 +INFO:root: +[Valid][14][UTT][ MSE = 0.2749 | LCC = 0.9406 | SRCC = 0.9417 ] +INFO:root:Average loss=0.12607394224405288 +INFO:root: +[Valid][15][UTT][ MSE = 0.2703 | LCC = 0.9366 | SRCC = 0.9365 ] +INFO:root:Average loss=0.1240776018500328 +INFO:root: +[Valid][16][UTT][ MSE = 0.2650 | LCC = 0.9378 | SRCC = 0.9390 ] +INFO:root:Average loss=0.11930628028512001 +INFO:root: +[Valid][17][UTT][ MSE = 0.2459 | LCC = 0.9448 | SRCC = 0.9454 ] +INFO:root:Average loss=0.117901809155941 +INFO:root: +[Valid][18][UTT][ MSE = 0.2413 | LCC = 0.9395 | SRCC = 0.9396 ] +INFO:root:Average loss=0.11178152811527252 +INFO:root: +[Valid][19][UTT][ MSE = 0.2299 | LCC = 0.9432 | SRCC = 0.9433 ] +INFO:root:Average loss=0.10334135204553604 +INFO:root: +[Valid][20][UTT][ MSE = 0.2176 | LCC = 0.9450 | SRCC = 0.9454 ] +INFO:root:Average loss=0.10154286581277848 +INFO:root: +[Valid][21][UTT][ MSE = 0.2054 | LCC = 0.9439 | SRCC = 0.9445 ] +INFO:root:Average loss=0.10280697202682496 +INFO:root: +[Valid][22][UTT][ MSE = 0.2558 | LCC = 0.9424 | SRCC = 0.9435 ] +INFO:root:Average loss=0.10026279300451278 +INFO:root: +[Valid][23][UTT][ MSE = 0.2804 | LCC = 0.9442 | SRCC = 0.9451 ] +INFO:root:Average loss=0.09706531736254692 +INFO:root: +[Valid][24][UTT][ MSE = 0.2455 | LCC = 0.9395 | SRCC = 0.9400 ] +INFO:root:Average loss=0.09057048386335373 +INFO:root: +[Valid][25][UTT][ MSE = 0.2213 | LCC = 0.9452 | SRCC = 0.9451 ] +INFO:root:Average loss=0.08575144201517106 +INFO:root: +[Valid][26][UTT][ MSE = 0.2087 | LCC = 0.9467 | SRCC = 0.9462 ] +INFO:root:Average loss=0.08538841089606285 +INFO:root: +[Valid][27][UTT][ MSE = 0.2330 | LCC = 0.9386 | SRCC = 0.9425 ] +INFO:root:Average loss=0.08338552322983742 +INFO:root: +[Valid][28][UTT][ MSE = 0.1994 | LCC = 0.9471 | SRCC = 0.9473 ] +INFO:root:Average loss=0.07937182173132896 +INFO:root: +[Valid][29][UTT][ MSE = 0.2331 | LCC = 0.9433 | SRCC = 0.9441 ] +INFO:root:Average loss=0.07781142038106918 +INFO:root: +[Valid][30][UTT][ MSE = 0.2223 | LCC = 0.9439 | SRCC = 0.9439 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2097 | LCC = 0.9457 | SRCC = 0.9458 ] diff --git a/results_tencent/w2v2_large_asr_en/layer0/run4/config.gin b/results_tencent/w2v2_large_asr_en/layer0/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer0/run4/model.pt b/results_tencent/w2v2_large_asr_en/layer0/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4e441f7e4b552f1a3b0056dc987c267e41fb53b8 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:195ac7072f097542df6bce24e23182196ac367d448be61714f0a2e31c7a1a97c +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer0/run4/model_best.pt b/results_tencent/w2v2_large_asr_en/layer0/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..91abc81740a83a4b3f76e51e9cdb3bad9e8055cd --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:61ba9166506e4ea57d609855098bb36dfcd5d15f33d37650c212af79c0e446b9 +size 3404998 diff --git a/results_tencent/w2v2_large_asr_en/layer0/run4/train.log b/results_tencent/w2v2_large_asr_en/layer0/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e9747dc400d39a596e80c164178a3c9ce6e9cf90 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer0/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5631957949399948 +INFO:root: +[Valid][0][UTT][ MSE = 0.4494 | LCC = 0.8740 | SRCC = 0.8730 ] +INFO:root:Average loss=0.3411657792329788 +INFO:root: +[Valid][1][UTT][ MSE = 0.3488 | LCC = 0.8964 | SRCC = 0.8966 ] +INFO:root:Average loss=0.2976254588365555 +INFO:root: +[Valid][2][UTT][ MSE = 0.3333 | LCC = 0.9059 | SRCC = 0.9046 ] +INFO:root:Average loss=0.2650909913778305 +INFO:root: +[Valid][3][UTT][ MSE = 0.3201 | LCC = 0.9120 | SRCC = 0.9129 ] +INFO:root:Average loss=0.2358146229982376 +INFO:root: +[Valid][4][UTT][ MSE = 0.3306 | LCC = 0.9116 | SRCC = 0.9145 ] +INFO:root:Average loss=0.22154437685012818 +INFO:root: +[Valid][5][UTT][ MSE = 0.2832 | LCC = 0.9230 | SRCC = 0.9233 ] +INFO:root:Average loss=0.20374881803989411 +INFO:root: +[Valid][6][UTT][ MSE = 0.2576 | LCC = 0.9301 | SRCC = 0.9307 ] +INFO:root:Average loss=0.19040218222141267 +INFO:root: +[Valid][7][UTT][ MSE = 0.2612 | LCC = 0.9320 | SRCC = 0.9325 ] +INFO:root:Average loss=0.17888089042901992 +INFO:root: +[Valid][8][UTT][ MSE = 0.2740 | LCC = 0.9285 | SRCC = 0.9289 ] +INFO:root:Average loss=0.1687595579624176 +INFO:root: +[Valid][9][UTT][ MSE = 0.2560 | LCC = 0.9289 | SRCC = 0.9297 ] +INFO:root:Average loss=0.16323179107904434 +INFO:root: +[Valid][10][UTT][ MSE = 0.2372 | LCC = 0.9362 | SRCC = 0.9374 ] +INFO:root:Average loss=0.14992485547065734 +INFO:root: +[Valid][11][UTT][ MSE = 0.2628 | LCC = 0.9367 | SRCC = 0.9375 ] +INFO:root:Average loss=0.14807665467262268 +INFO:root: +[Valid][12][UTT][ MSE = 0.2566 | LCC = 0.9340 | SRCC = 0.9343 ] +INFO:root:Average loss=0.144786834359169 +INFO:root: +[Valid][13][UTT][ MSE = 0.2536 | LCC = 0.9360 | SRCC = 0.9367 ] +INFO:root:Average loss=0.13513494366407394 +INFO:root: +[Valid][14][UTT][ MSE = 0.2272 | LCC = 0.9365 | SRCC = 0.9370 ] +INFO:root:Average loss=0.1273358070254326 +INFO:root: +[Valid][15][UTT][ MSE = 0.2196 | LCC = 0.9397 | SRCC = 0.9410 ] +INFO:root:Average loss=0.12330739915370942 +INFO:root: +[Valid][16][UTT][ MSE = 0.2180 | LCC = 0.9398 | SRCC = 0.9407 ] +INFO:root:Average loss=0.11635331761837006 +INFO:root: +[Valid][17][UTT][ MSE = 0.2562 | LCC = 0.9406 | SRCC = 0.9415 ] +INFO:root:Average loss=0.11635982328653335 +INFO:root: +[Valid][18][UTT][ MSE = 0.2257 | LCC = 0.9436 | SRCC = 0.9445 ] +INFO:root:Average loss=0.11165623599290848 +INFO:root: +[Valid][19][UTT][ MSE = 0.2017 | LCC = 0.9418 | SRCC = 0.9426 ] +INFO:root:Average loss=0.1059285136461258 +INFO:root: +[Valid][20][UTT][ MSE = 0.2069 | LCC = 0.9419 | SRCC = 0.9425 ] +INFO:root:Average loss=0.10608018425107002 +INFO:root: +[Valid][21][UTT][ MSE = 0.2171 | LCC = 0.9429 | SRCC = 0.9436 ] +INFO:root:Average loss=0.10404559701681138 +INFO:root: +[Valid][22][UTT][ MSE = 0.2162 | LCC = 0.9451 | SRCC = 0.9457 ] +INFO:root:Average loss=0.09825759163498879 +INFO:root: +[Valid][23][UTT][ MSE = 0.2109 | LCC = 0.9450 | SRCC = 0.9455 ] +INFO:root:Average loss=0.09583988213539124 +INFO:root: +[Valid][24][UTT][ MSE = 0.2023 | LCC = 0.9409 | SRCC = 0.9427 ] +INFO:root:Average loss=0.0902070954144001 +INFO:root: +[Valid][25][UTT][ MSE = 0.2103 | LCC = 0.9449 | SRCC = 0.9450 ] +INFO:root:Average loss=0.09075407809019088 +INFO:root: +[Valid][26][UTT][ MSE = 0.2025 | LCC = 0.9472 | SRCC = 0.9478 ] +INFO:root:Average loss=0.08497983774542808 +INFO:root: +[Valid][27][UTT][ MSE = 0.2004 | LCC = 0.9453 | SRCC = 0.9455 ] +INFO:root:Average loss=0.08363067823648453 +INFO:root: +[Valid][28][UTT][ MSE = 0.2146 | LCC = 0.9453 | SRCC = 0.9453 ] +INFO:root:Average loss=0.08124600887298584 +INFO:root: +[Valid][29][UTT][ MSE = 0.2302 | LCC = 0.9455 | SRCC = 0.9453 ] +INFO:root:Average loss=0.08334646984934807 +INFO:root: +[Valid][30][UTT][ MSE = 0.1921 | LCC = 0.9465 | SRCC = 0.9462 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2072 | LCC = 0.9479 | SRCC = 0.9486 ] diff --git a/results_tencent/w2v2_large_asr_en/layer10/run0/config.gin b/results_tencent/w2v2_large_asr_en/layer10/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer10/run0/model.pt b/results_tencent/w2v2_large_asr_en/layer10/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a9df5741b0118fe212065a5a1f49087438f7fc07 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8dacd0896b3b6f8f490a4c27412fbe80d584c6a7f473778a120d3611ec75d410 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer10/run0/model_best.pt b/results_tencent/w2v2_large_asr_en/layer10/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e56490d02d71a736e99829d7be5017529a075098 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:09756ea7c63f4192036f30c43e302f9dda1faf63ac1aa8f66461a96c5243fd5d +size 3404998 diff --git a/results_tencent/w2v2_large_asr_en/layer10/run0/train.log b/results_tencent/w2v2_large_asr_en/layer10/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..25cb458cb887d5b64f972f3bae011e0c90d67601 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5010222245454788 +INFO:root: +[Valid][0][UTT][ MSE = 0.3356 | LCC = 0.9079 | SRCC = 0.9067 ] +INFO:root:Average loss=0.22929055041074753 +INFO:root: +[Valid][1][UTT][ MSE = 0.2595 | LCC = 0.9292 | SRCC = 0.9272 ] +INFO:root:Average loss=0.18144736593961716 +INFO:root: +[Valid][2][UTT][ MSE = 0.2243 | LCC = 0.9390 | SRCC = 0.9375 ] +INFO:root:Average loss=0.15326694977283478 +INFO:root: +[Valid][3][UTT][ MSE = 0.2462 | LCC = 0.9455 | SRCC = 0.9449 ] +INFO:root:Average loss=0.13040729987621308 +INFO:root: +[Valid][4][UTT][ MSE = 0.1936 | LCC = 0.9489 | SRCC = 0.9490 ] +INFO:root:Average loss=0.11421453300118446 +INFO:root: +[Valid][5][UTT][ MSE = 0.2101 | LCC = 0.9508 | SRCC = 0.9506 ] +INFO:root:Average loss=0.10502200680971145 +INFO:root: +[Valid][6][UTT][ MSE = 0.2050 | LCC = 0.9525 | SRCC = 0.9523 ] +INFO:root:Average loss=0.10170439013838768 +INFO:root: +[Valid][7][UTT][ MSE = 0.1755 | LCC = 0.9533 | SRCC = 0.9527 ] +INFO:root:Average loss=0.100304323554039 +INFO:root: +[Valid][8][UTT][ MSE = 0.1908 | LCC = 0.9543 | SRCC = 0.9541 ] +INFO:root:Average loss=0.08370878720283509 +INFO:root: +[Valid][9][UTT][ MSE = 0.1980 | LCC = 0.9562 | SRCC = 0.9555 ] +INFO:root:Average loss=0.07891840344667435 +INFO:root: +[Valid][10][UTT][ MSE = 0.2028 | LCC = 0.9559 | SRCC = 0.9552 ] +INFO:root:Average loss=0.07109805217385293 +INFO:root: +[Valid][11][UTT][ MSE = 0.1710 | LCC = 0.9553 | SRCC = 0.9546 ] +INFO:root:Average loss=0.07400908750295639 +INFO:root: +[Valid][12][UTT][ MSE = 0.2114 | LCC = 0.9551 | SRCC = 0.9550 ] +INFO:root:Average loss=0.06603842422366142 +INFO:root: +[Valid][13][UTT][ MSE = 0.2193 | LCC = 0.9560 | SRCC = 0.9554 ] +INFO:root:Average loss=0.06004586634039879 +INFO:root: +[Valid][14][UTT][ MSE = 0.1790 | LCC = 0.9560 | SRCC = 0.9555 ] +INFO:root:Average loss=0.061918595537543294 +INFO:root: +[Valid][15][UTT][ MSE = 0.2059 | LCC = 0.9556 | SRCC = 0.9553 ] +INFO:root:Average loss=0.05104477721452713 +INFO:root: +[Valid][16][UTT][ MSE = 0.1862 | LCC = 0.9563 | SRCC = 0.9552 ] +INFO:root:Average loss=0.04904086169600487 +INFO:root: +[Valid][17][UTT][ MSE = 0.1856 | LCC = 0.9561 | SRCC = 0.9550 ] +INFO:root:Average loss=0.04913599762320518 +INFO:root: +[Valid][18][UTT][ MSE = 0.2109 | LCC = 0.9555 | SRCC = 0.9548 ] +INFO:root:Average loss=0.04368538872897625 +INFO:root: +[Valid][19][UTT][ MSE = 0.1904 | LCC = 0.9568 | SRCC = 0.9551 ] +INFO:root:Average loss=0.04146124598383903 +INFO:root: +[Valid][20][UTT][ MSE = 0.1693 | LCC = 0.9564 | SRCC = 0.9548 ] +INFO:root:Average loss=0.04018029852211475 +INFO:root: +[Valid][21][UTT][ MSE = 0.1828 | LCC = 0.9564 | SRCC = 0.9550 ] +INFO:root:Average loss=0.03878097864985466 +INFO:root: +[Valid][22][UTT][ MSE = 0.1932 | LCC = 0.9556 | SRCC = 0.9541 ] +INFO:root:Average loss=0.03755505308508873 +INFO:root: +[Valid][23][UTT][ MSE = 0.1820 | LCC = 0.9572 | SRCC = 0.9559 ] +INFO:root:Average loss=0.03594144396483898 +INFO:root: +[Valid][24][UTT][ MSE = 0.1679 | LCC = 0.9556 | SRCC = 0.9544 ] +INFO:root:Average loss=0.03358615885674954 +INFO:root: +[Valid][25][UTT][ MSE = 0.1846 | LCC = 0.9563 | SRCC = 0.9552 ] +INFO:root:Average loss=0.031770835474133494 +INFO:root: +[Valid][26][UTT][ MSE = 0.1805 | LCC = 0.9578 | SRCC = 0.9562 ] +INFO:root:Average loss=0.030430072009563446 +INFO:root: +[Valid][27][UTT][ MSE = 0.1853 | LCC = 0.9567 | SRCC = 0.9553 ] +INFO:root:Average loss=0.028494312539696694 +INFO:root: +[Valid][28][UTT][ MSE = 0.1808 | LCC = 0.9569 | SRCC = 0.9555 ] +INFO:root:Average loss=0.02948863671720028 +INFO:root: +[Valid][29][UTT][ MSE = 0.1704 | LCC = 0.9568 | SRCC = 0.9554 ] +INFO:root:Average loss=0.027734526216983794 +INFO:root: +[Valid][30][UTT][ MSE = 0.1943 | LCC = 0.9565 | SRCC = 0.9553 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1827 | LCC = 0.9590 | SRCC = 0.9593 ] diff --git a/results_tencent/w2v2_large_asr_en/layer10/run1/config.gin b/results_tencent/w2v2_large_asr_en/layer10/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer10/run1/model.pt b/results_tencent/w2v2_large_asr_en/layer10/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7a4461c530e23df9ea53591381059cd65844d249 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2b8c0529f25555f4e3e5dc2f51ea5aaf5d0c12e07f5d98269b201f7327ae20ac +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer10/run1/model_best.pt b/results_tencent/w2v2_large_asr_en/layer10/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d3abbe5f91e120e2349e5b7c7e7dd7967828d5f4 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:01ef34f7d3abfae5e3448480f86dead58af5f7b54b983043b4c53021e0d5e6f4 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer10/run1/train.log b/results_tencent/w2v2_large_asr_en/layer10/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b06b2193990b23c5a469a29825fe8e2c8f7cc77c --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5603787733316421 +INFO:root: +[Valid][0][UTT][ MSE = 0.3758 | LCC = 0.8986 | SRCC = 0.9004 ] +INFO:root:Average loss=0.2266037258505821 +INFO:root: +[Valid][1][UTT][ MSE = 0.3077 | LCC = 0.9323 | SRCC = 0.9312 ] +INFO:root:Average loss=0.17661147236824035 +INFO:root: +[Valid][2][UTT][ MSE = 0.2388 | LCC = 0.9411 | SRCC = 0.9413 ] +INFO:root:Average loss=0.16322550082206727 +INFO:root: +[Valid][3][UTT][ MSE = 0.2356 | LCC = 0.9469 | SRCC = 0.9468 ] +INFO:root:Average loss=0.146523366689682 +INFO:root: +[Valid][4][UTT][ MSE = 0.2567 | LCC = 0.9492 | SRCC = 0.9496 ] +INFO:root:Average loss=0.12082720899581909 +INFO:root: +[Valid][5][UTT][ MSE = 0.2178 | LCC = 0.9527 | SRCC = 0.9522 ] +INFO:root:Average loss=0.10759995889663697 +INFO:root: +[Valid][6][UTT][ MSE = 0.2165 | LCC = 0.9545 | SRCC = 0.9543 ] +INFO:root:Average loss=0.0963390297293663 +INFO:root: +[Valid][7][UTT][ MSE = 0.2414 | LCC = 0.9537 | SRCC = 0.9538 ] +INFO:root:Average loss=0.0931014162003994 +INFO:root: +[Valid][8][UTT][ MSE = 0.2232 | LCC = 0.9546 | SRCC = 0.9547 ] +INFO:root:Average loss=0.08377202424407006 +INFO:root: +[Valid][9][UTT][ MSE = 0.2424 | LCC = 0.9552 | SRCC = 0.9549 ] +INFO:root:Average loss=0.0782404323220253 +INFO:root: +[Valid][10][UTT][ MSE = 0.2017 | LCC = 0.9566 | SRCC = 0.9560 ] +INFO:root:Average loss=0.07125890260934829 +INFO:root: +[Valid][11][UTT][ MSE = 0.2306 | LCC = 0.9543 | SRCC = 0.9542 ] +INFO:root:Average loss=0.06637917655706406 +INFO:root: +[Valid][12][UTT][ MSE = 0.2030 | LCC = 0.9560 | SRCC = 0.9553 ] +INFO:root:Average loss=0.06531040970981121 +INFO:root: +[Valid][13][UTT][ MSE = 0.2353 | LCC = 0.9560 | SRCC = 0.9559 ] +INFO:root:Average loss=0.05905053097009659 +INFO:root: +[Valid][14][UTT][ MSE = 0.2284 | LCC = 0.9565 | SRCC = 0.9567 ] +INFO:root:Average loss=0.05572305777668953 +INFO:root: +[Valid][15][UTT][ MSE = 0.2050 | LCC = 0.9569 | SRCC = 0.9565 ] +INFO:root:Average loss=0.049080815181136134 +INFO:root: +[Valid][16][UTT][ MSE = 0.2475 | LCC = 0.9548 | SRCC = 0.9554 ] +INFO:root:Average loss=0.049781987801194194 +INFO:root: +[Valid][17][UTT][ MSE = 0.2452 | LCC = 0.9548 | SRCC = 0.9554 ] +INFO:root:Average loss=0.04713604381680488 +INFO:root: +[Valid][18][UTT][ MSE = 0.2155 | LCC = 0.9533 | SRCC = 0.9526 ] +INFO:root:Average loss=0.04531936103105545 +INFO:root: +[Valid][19][UTT][ MSE = 0.2101 | LCC = 0.9552 | SRCC = 0.9546 ] +INFO:root:Average loss=0.04381908024847508 +INFO:root: +[Valid][20][UTT][ MSE = 0.2090 | LCC = 0.9558 | SRCC = 0.9552 ] +INFO:root:Average loss=0.03895128692686558 +INFO:root: +[Valid][21][UTT][ MSE = 0.1920 | LCC = 0.9547 | SRCC = 0.9539 ] +INFO:root:Average loss=0.03864530585706234 +INFO:root: +[Valid][22][UTT][ MSE = 0.1911 | LCC = 0.9552 | SRCC = 0.9540 ] +INFO:root:Average loss=0.03801547911763191 +INFO:root: +[Valid][23][UTT][ MSE = 0.2114 | LCC = 0.9566 | SRCC = 0.9558 ] +INFO:root:Average loss=0.03450804188847542 +INFO:root: +[Valid][24][UTT][ MSE = 0.2195 | LCC = 0.9553 | SRCC = 0.9553 ] +INFO:root:Average loss=0.03415686471760273 +INFO:root: +[Valid][25][UTT][ MSE = 0.2021 | LCC = 0.9558 | SRCC = 0.9552 ] +INFO:root:Average loss=0.031128952831029892 +INFO:root: +[Valid][26][UTT][ MSE = 0.2291 | LCC = 0.9552 | SRCC = 0.9555 ] +INFO:root:Average loss=0.03010747820138931 +INFO:root: +[Valid][27][UTT][ MSE = 0.2379 | LCC = 0.9548 | SRCC = 0.9549 ] +INFO:root:Average loss=0.02996332250535488 +INFO:root: +[Valid][28][UTT][ MSE = 0.2181 | LCC = 0.9561 | SRCC = 0.9556 ] +INFO:root:Average loss=0.028923279009759426 +INFO:root: +[Valid][29][UTT][ MSE = 0.2238 | LCC = 0.9542 | SRCC = 0.9536 ] +INFO:root:Average loss=0.027694006398320196 +INFO:root: +[Valid][30][UTT][ MSE = 0.2069 | LCC = 0.9549 | SRCC = 0.9540 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2182 | LCC = 0.9548 | SRCC = 0.9564 ] diff --git a/results_tencent/w2v2_large_asr_en/layer10/run2/config.gin b/results_tencent/w2v2_large_asr_en/layer10/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer10/run2/model.pt b/results_tencent/w2v2_large_asr_en/layer10/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d949abac2c0cb395067ed6d14c6c47720b0cffeb --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a315d1cd6b47fc559bd25de3646cf62c9727503bb12777188169988a2611c87a +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer10/run2/model_best.pt b/results_tencent/w2v2_large_asr_en/layer10/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..688f3b52f9e810f61f7775129afbd89c4c7376ea --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:786764fac91eeb10e5844efd1460952eb5771e8190986ef1c5a29511412e9d83 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer10/run2/train.log b/results_tencent/w2v2_large_asr_en/layer10/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..03484899f0615b918f0118ff7a3e37ec184d610a --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5990666986703873 +INFO:root: +[Valid][0][UTT][ MSE = 0.3596 | LCC = 0.8934 | SRCC = 0.8964 ] +INFO:root:Average loss=0.23174694460630416 +INFO:root: +[Valid][1][UTT][ MSE = 0.2969 | LCC = 0.9247 | SRCC = 0.9234 ] +INFO:root:Average loss=0.19153335118293763 +INFO:root: +[Valid][2][UTT][ MSE = 0.2587 | LCC = 0.9346 | SRCC = 0.9355 ] +INFO:root:Average loss=0.16517070412635804 +INFO:root: +[Valid][3][UTT][ MSE = 0.2289 | LCC = 0.9430 | SRCC = 0.9434 ] +INFO:root:Average loss=0.13948925870656967 +INFO:root: +[Valid][4][UTT][ MSE = 0.2573 | LCC = 0.9450 | SRCC = 0.9447 ] +INFO:root:Average loss=0.13110521811246872 +INFO:root: +[Valid][5][UTT][ MSE = 0.2340 | LCC = 0.9502 | SRCC = 0.9499 ] +INFO:root:Average loss=0.11405591505765915 +INFO:root: +[Valid][6][UTT][ MSE = 0.2594 | LCC = 0.9498 | SRCC = 0.9502 ] +INFO:root:Average loss=0.10888171818852424 +INFO:root: +[Valid][7][UTT][ MSE = 0.2659 | LCC = 0.9497 | SRCC = 0.9500 ] +INFO:root:Average loss=0.09537407386302948 +INFO:root: +[Valid][8][UTT][ MSE = 0.2092 | LCC = 0.9542 | SRCC = 0.9540 ] +INFO:root:Average loss=0.08707308340072632 +INFO:root: +[Valid][9][UTT][ MSE = 0.1970 | LCC = 0.9549 | SRCC = 0.9543 ] +INFO:root:Average loss=0.08054021772742272 +INFO:root: +[Valid][10][UTT][ MSE = 0.1781 | LCC = 0.9552 | SRCC = 0.9555 ] +INFO:root:Average loss=0.07421033498644829 +INFO:root: +[Valid][11][UTT][ MSE = 0.2133 | LCC = 0.9555 | SRCC = 0.9553 ] +INFO:root:Average loss=0.06622179347276688 +INFO:root: +[Valid][12][UTT][ MSE = 0.2521 | LCC = 0.9535 | SRCC = 0.9537 ] +INFO:root:Average loss=0.07030860358476639 +INFO:root: +[Valid][13][UTT][ MSE = 0.1998 | LCC = 0.9570 | SRCC = 0.9557 ] +INFO:root:Average loss=0.05837066516280174 +INFO:root: +[Valid][14][UTT][ MSE = 0.1780 | LCC = 0.9545 | SRCC = 0.9544 ] +INFO:root:Average loss=0.05676400099694729 +INFO:root: +[Valid][15][UTT][ MSE = 0.2248 | LCC = 0.9560 | SRCC = 0.9555 ] +INFO:root:Average loss=0.0513066251128912 +INFO:root: +[Valid][16][UTT][ MSE = 0.1889 | LCC = 0.9573 | SRCC = 0.9563 ] +INFO:root:Average loss=0.04992554013431072 +INFO:root: +[Valid][17][UTT][ MSE = 0.1918 | LCC = 0.9578 | SRCC = 0.9568 ] +INFO:root:Average loss=0.04731315691769123 +INFO:root: +[Valid][18][UTT][ MSE = 0.2096 | LCC = 0.9567 | SRCC = 0.9549 ] +INFO:root:Average loss=0.04341712997853756 +INFO:root: +[Valid][19][UTT][ MSE = 0.2004 | LCC = 0.9579 | SRCC = 0.9564 ] +INFO:root:Average loss=0.04224585399031639 +INFO:root: +[Valid][20][UTT][ MSE = 0.1944 | LCC = 0.9563 | SRCC = 0.9545 ] +INFO:root:Average loss=0.040629407063126566 +INFO:root: +[Valid][21][UTT][ MSE = 0.2280 | LCC = 0.9552 | SRCC = 0.9545 ] +INFO:root:Average loss=0.039427185893058776 +INFO:root: +[Valid][22][UTT][ MSE = 0.1903 | LCC = 0.9572 | SRCC = 0.9561 ] +INFO:root:Average loss=0.03544106629490852 +INFO:root: +[Valid][23][UTT][ MSE = 0.2200 | LCC = 0.9564 | SRCC = 0.9553 ] +INFO:root:Average loss=0.03559505261480808 +INFO:root: +[Valid][24][UTT][ MSE = 0.2227 | LCC = 0.9568 | SRCC = 0.9558 ] +INFO:root:Average loss=0.03225844207406044 +INFO:root: +[Valid][25][UTT][ MSE = 0.1825 | LCC = 0.9565 | SRCC = 0.9557 ] +INFO:root:Average loss=0.03080828219652176 +INFO:root: +[Valid][26][UTT][ MSE = 0.2094 | LCC = 0.9573 | SRCC = 0.9561 ] +INFO:root:Average loss=0.029174505949020386 +INFO:root: +[Valid][27][UTT][ MSE = 0.2377 | LCC = 0.9557 | SRCC = 0.9553 ] +INFO:root:Average loss=0.027514540180563925 +INFO:root: +[Valid][28][UTT][ MSE = 0.2186 | LCC = 0.9571 | SRCC = 0.9561 ] +INFO:root:Average loss=0.027990632250905038 +INFO:root: +[Valid][29][UTT][ MSE = 0.1997 | LCC = 0.9578 | SRCC = 0.9565 ] +INFO:root:Average loss=0.02687372672557831 +INFO:root: +[Valid][30][UTT][ MSE = 0.2045 | LCC = 0.9567 | SRCC = 0.9554 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2121 | LCC = 0.9568 | SRCC = 0.9573 ] diff --git a/results_tencent/w2v2_large_asr_en/layer10/run3/config.gin b/results_tencent/w2v2_large_asr_en/layer10/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer10/run3/model.pt b/results_tencent/w2v2_large_asr_en/layer10/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..306eb4457fb843f7a32e39e4871c73ac46792c23 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c5c62df18135890ce298a17c9a644675cb85598a8942b349728aabcc94887a24 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer10/run3/model_best.pt b/results_tencent/w2v2_large_asr_en/layer10/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0262f0d212e2c16a33899d198622efc254c28c25 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e729122522432f0835838f26d2cad3fa76b7925d6f28f007f20fe56e3c1ea3e4 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer10/run3/train.log b/results_tencent/w2v2_large_asr_en/layer10/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..fc706047b4cef7c94b1aa8b6da18796454077750 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5285066267251969 +INFO:root: +[Valid][0][UTT][ MSE = 0.3495 | LCC = 0.9046 | SRCC = 0.9025 ] +INFO:root:Average loss=0.22524885666370392 +INFO:root: +[Valid][1][UTT][ MSE = 0.2786 | LCC = 0.9305 | SRCC = 0.9289 ] +INFO:root:Average loss=0.1791326133608818 +INFO:root: +[Valid][2][UTT][ MSE = 0.2578 | LCC = 0.9395 | SRCC = 0.9389 ] +INFO:root:Average loss=0.15077698367834091 +INFO:root: +[Valid][3][UTT][ MSE = 0.2336 | LCC = 0.9419 | SRCC = 0.9421 ] +INFO:root:Average loss=0.12690383332967758 +INFO:root: +[Valid][4][UTT][ MSE = 0.1791 | LCC = 0.9497 | SRCC = 0.9504 ] +INFO:root:Average loss=0.11462375044822692 +INFO:root: +[Valid][5][UTT][ MSE = 0.1806 | LCC = 0.9521 | SRCC = 0.9532 ] +INFO:root:Average loss=0.10862916389107705 +INFO:root: +[Valid][6][UTT][ MSE = 0.2203 | LCC = 0.9544 | SRCC = 0.9539 ] +INFO:root:Average loss=0.10128800648450852 +INFO:root: +[Valid][7][UTT][ MSE = 0.2116 | LCC = 0.9529 | SRCC = 0.9530 ] +INFO:root:Average loss=0.0923602003455162 +INFO:root: +[Valid][8][UTT][ MSE = 0.1858 | LCC = 0.9569 | SRCC = 0.9568 ] +INFO:root:Average loss=0.08683293190598487 +INFO:root: +[Valid][9][UTT][ MSE = 0.2051 | LCC = 0.9565 | SRCC = 0.9563 ] +INFO:root:Average loss=0.07982811743021011 +INFO:root: +[Valid][10][UTT][ MSE = 0.2036 | LCC = 0.9551 | SRCC = 0.9543 ] +INFO:root:Average loss=0.07268274888396263 +INFO:root: +[Valid][11][UTT][ MSE = 0.1692 | LCC = 0.9566 | SRCC = 0.9562 ] +INFO:root:Average loss=0.06928026416897774 +INFO:root: +[Valid][12][UTT][ MSE = 0.1884 | LCC = 0.9563 | SRCC = 0.9557 ] +INFO:root:Average loss=0.06979656729102135 +INFO:root: +[Valid][13][UTT][ MSE = 0.2117 | LCC = 0.9575 | SRCC = 0.9568 ] +INFO:root:Average loss=0.062166826754808426 +INFO:root: +[Valid][14][UTT][ MSE = 0.1802 | LCC = 0.9578 | SRCC = 0.9569 ] +INFO:root:Average loss=0.05576982533931732 +INFO:root: +[Valid][15][UTT][ MSE = 0.1588 | LCC = 0.9577 | SRCC = 0.9569 ] +INFO:root:Average loss=0.05502957463264466 +INFO:root: +[Valid][16][UTT][ MSE = 0.1907 | LCC = 0.9581 | SRCC = 0.9568 ] +INFO:root:Average loss=0.05059075519442558 +INFO:root: +[Valid][17][UTT][ MSE = 0.1882 | LCC = 0.9579 | SRCC = 0.9565 ] +INFO:root:Average loss=0.04845566868782043 +INFO:root: +[Valid][18][UTT][ MSE = 0.1956 | LCC = 0.9557 | SRCC = 0.9551 ] +INFO:root:Average loss=0.04369436579942703 +INFO:root: +[Valid][19][UTT][ MSE = 0.1805 | LCC = 0.9541 | SRCC = 0.9543 ] +INFO:root:Average loss=0.04274707092344761 +INFO:root: +[Valid][20][UTT][ MSE = 0.1729 | LCC = 0.9543 | SRCC = 0.9549 ] +INFO:root:Average loss=0.04194863259792328 +INFO:root: +[Valid][21][UTT][ MSE = 0.1857 | LCC = 0.9552 | SRCC = 0.9548 ] +INFO:root:Average loss=0.03765071985125542 +INFO:root: +[Valid][22][UTT][ MSE = 0.1634 | LCC = 0.9571 | SRCC = 0.9564 ] +INFO:root:Average loss=0.036161151051521304 +INFO:root: +[Valid][23][UTT][ MSE = 0.1778 | LCC = 0.9563 | SRCC = 0.9553 ] +INFO:root:Average loss=0.03502218876779079 +INFO:root: +[Valid][24][UTT][ MSE = 0.1776 | LCC = 0.9573 | SRCC = 0.9566 ] +INFO:root:Average loss=0.03258236289024353 +INFO:root: +[Valid][25][UTT][ MSE = 0.1839 | LCC = 0.9563 | SRCC = 0.9556 ] +INFO:root:Average loss=0.0320270359814167 +INFO:root: +[Valid][26][UTT][ MSE = 0.1739 | LCC = 0.9572 | SRCC = 0.9563 ] +INFO:root:Average loss=0.03372133381664753 +INFO:root: +[Valid][27][UTT][ MSE = 0.1898 | LCC = 0.9562 | SRCC = 0.9556 ] +INFO:root:Average loss=0.029788642928004264 +INFO:root: +[Valid][28][UTT][ MSE = 0.1704 | LCC = 0.9580 | SRCC = 0.9566 ] +INFO:root:Average loss=0.027675464630126954 +INFO:root: +[Valid][29][UTT][ MSE = 0.1706 | LCC = 0.9574 | SRCC = 0.9562 ] +INFO:root:Average loss=0.02931845810264349 +INFO:root: +[Valid][30][UTT][ MSE = 0.1761 | LCC = 0.9568 | SRCC = 0.9559 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2010 | LCC = 0.9559 | SRCC = 0.9561 ] diff --git a/results_tencent/w2v2_large_asr_en/layer10/run4/config.gin b/results_tencent/w2v2_large_asr_en/layer10/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer10/run4/model.pt b/results_tencent/w2v2_large_asr_en/layer10/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d4011738d8c804b5d67a3cbd889a589cb843e2d1 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7d89fb5b7ee874d8f2b2329bfdaf00206a69a2c6d50fcffb28fcc7018a85f686 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer10/run4/model_best.pt b/results_tencent/w2v2_large_asr_en/layer10/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..222e0c14ecd83cc1aee1140d4465d97714f65761 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b29ccc4dd14e6d74d110e0d069399158ed91be6cbc2d0a38471482ba971c2f0b +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer10/run4/train.log b/results_tencent/w2v2_large_asr_en/layer10/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6f89ff8517694a6f207572914fd03eee03b1d009 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer10/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44579941058158873 +INFO:root: +[Valid][0][UTT][ MSE = 0.3057 | LCC = 0.9118 | SRCC = 0.9096 ] +INFO:root:Average loss=0.21829465001821519 +INFO:root: +[Valid][1][UTT][ MSE = 0.2358 | LCC = 0.9341 | SRCC = 0.9331 ] +INFO:root:Average loss=0.17866651034355163 +INFO:root: +[Valid][2][UTT][ MSE = 0.2394 | LCC = 0.9425 | SRCC = 0.9419 ] +INFO:root:Average loss=0.14484544485807418 +INFO:root: +[Valid][3][UTT][ MSE = 0.2758 | LCC = 0.9447 | SRCC = 0.9464 ] +INFO:root:Average loss=0.12815326994657517 +INFO:root: +[Valid][4][UTT][ MSE = 0.2093 | LCC = 0.9523 | SRCC = 0.9521 ] +INFO:root:Average loss=0.12252556091547012 +INFO:root: +[Valid][5][UTT][ MSE = 0.2464 | LCC = 0.9509 | SRCC = 0.9519 ] +INFO:root:Average loss=0.10439216762781144 +INFO:root: +[Valid][6][UTT][ MSE = 0.2176 | LCC = 0.9536 | SRCC = 0.9537 ] +INFO:root:Average loss=0.09497732418775559 +INFO:root: +[Valid][7][UTT][ MSE = 0.2263 | LCC = 0.9553 | SRCC = 0.9552 ] +INFO:root:Average loss=0.08847856071591377 +INFO:root: +[Valid][8][UTT][ MSE = 0.2399 | LCC = 0.9537 | SRCC = 0.9539 ] +INFO:root:Average loss=0.08257314336299897 +INFO:root: +[Valid][9][UTT][ MSE = 0.1794 | LCC = 0.9571 | SRCC = 0.9569 ] +INFO:root:Average loss=0.07822251987457275 +INFO:root: +[Valid][10][UTT][ MSE = 0.1992 | LCC = 0.9567 | SRCC = 0.9561 ] +INFO:root:Average loss=0.06896637162566185 +INFO:root: +[Valid][11][UTT][ MSE = 0.2219 | LCC = 0.9544 | SRCC = 0.9556 ] +INFO:root:Average loss=0.06737528908252716 +INFO:root: +[Valid][12][UTT][ MSE = 0.2053 | LCC = 0.9568 | SRCC = 0.9565 ] +INFO:root:Average loss=0.060266792118549344 +INFO:root: +[Valid][13][UTT][ MSE = 0.1941 | LCC = 0.9572 | SRCC = 0.9562 ] +INFO:root:Average loss=0.05566529126465321 +INFO:root: +[Valid][14][UTT][ MSE = 0.2073 | LCC = 0.9576 | SRCC = 0.9571 ] +INFO:root:Average loss=0.052824598386883734 +INFO:root: +[Valid][15][UTT][ MSE = 0.1887 | LCC = 0.9592 | SRCC = 0.9582 ] +INFO:root:Average loss=0.049190616264939306 +INFO:root: +[Valid][16][UTT][ MSE = 0.2054 | LCC = 0.9580 | SRCC = 0.9580 ] +INFO:root:Average loss=0.04600373490154743 +INFO:root: +[Valid][17][UTT][ MSE = 0.1944 | LCC = 0.9581 | SRCC = 0.9573 ] +INFO:root:Average loss=0.04593441557884216 +INFO:root: +[Valid][18][UTT][ MSE = 0.1909 | LCC = 0.9585 | SRCC = 0.9576 ] +INFO:root:Average loss=0.04096190817654133 +INFO:root: +[Valid][19][UTT][ MSE = 0.2361 | LCC = 0.9555 | SRCC = 0.9565 ] +INFO:root:Average loss=0.0421585553586483 +INFO:root: +[Valid][20][UTT][ MSE = 0.2020 | LCC = 0.9573 | SRCC = 0.9567 ] +INFO:root:Average loss=0.039209578692913055 +INFO:root: +[Valid][21][UTT][ MSE = 0.1755 | LCC = 0.9584 | SRCC = 0.9572 ] +INFO:root:Average loss=0.035189966961741447 +INFO:root: +[Valid][22][UTT][ MSE = 0.2024 | LCC = 0.9577 | SRCC = 0.9573 ] +INFO:root:Average loss=0.035884251922369007 +INFO:root: +[Valid][23][UTT][ MSE = 0.2137 | LCC = 0.9573 | SRCC = 0.9571 ] +INFO:root:Average loss=0.03409900195896626 +INFO:root: +[Valid][24][UTT][ MSE = 0.1867 | LCC = 0.9573 | SRCC = 0.9562 ] +INFO:root:Average loss=0.03412789665162563 +INFO:root: +[Valid][25][UTT][ MSE = 0.2245 | LCC = 0.9564 | SRCC = 0.9563 ] +INFO:root:Average loss=0.030063528805971147 +INFO:root: +[Valid][26][UTT][ MSE = 0.1909 | LCC = 0.9580 | SRCC = 0.9576 ] +INFO:root:Average loss=0.02850383374094963 +INFO:root: +[Valid][27][UTT][ MSE = 0.2043 | LCC = 0.9570 | SRCC = 0.9568 ] +INFO:root:Average loss=0.026539348259568215 +INFO:root: +[Valid][28][UTT][ MSE = 0.1923 | LCC = 0.9572 | SRCC = 0.9564 ] +INFO:root:Average loss=0.027131718397140504 +INFO:root: +[Valid][29][UTT][ MSE = 0.2032 | LCC = 0.9571 | SRCC = 0.9573 ] +INFO:root:Average loss=0.02667437396198511 +INFO:root: +[Valid][30][UTT][ MSE = 0.1941 | LCC = 0.9578 | SRCC = 0.9576 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1959 | LCC = 0.9581 | SRCC = 0.9585 ] diff --git a/results_tencent/w2v2_large_asr_en/layer12/run0/config.gin b/results_tencent/w2v2_large_asr_en/layer12/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer12/run0/model.pt b/results_tencent/w2v2_large_asr_en/layer12/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..843fe9bead4f3bbf6100bf0ce78b816abc9cc3ed --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8f5490d3050280dd1d0dcd42aa4f25880b59d9ef1a4683b60a8b09daaf1744b2 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer12/run0/model_best.pt b/results_tencent/w2v2_large_asr_en/layer12/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..5cb632e38600159a4e41bd8985b95b9d2741804e --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:37d30204eb352a8f128f1e245eaf9dccf89799781a279e4dbbbc3b2439a1437e +size 3404998 diff --git a/results_tencent/w2v2_large_asr_en/layer12/run0/train.log b/results_tencent/w2v2_large_asr_en/layer12/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a92b41abe6a3831609f09553f30cbd6f403e47ad --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.45763952028751376 +INFO:root: +[Valid][0][UTT][ MSE = 0.3437 | LCC = 0.9045 | SRCC = 0.9028 ] +INFO:root:Average loss=0.24739523291587828 +INFO:root: +[Valid][1][UTT][ MSE = 0.3047 | LCC = 0.9260 | SRCC = 0.9258 ] +INFO:root:Average loss=0.18281020152568817 +INFO:root: +[Valid][2][UTT][ MSE = 0.2355 | LCC = 0.9383 | SRCC = 0.9379 ] +INFO:root:Average loss=0.15608748096227645 +INFO:root: +[Valid][3][UTT][ MSE = 0.2260 | LCC = 0.9424 | SRCC = 0.9421 ] +INFO:root:Average loss=0.1361794403195381 +INFO:root: +[Valid][4][UTT][ MSE = 0.2345 | LCC = 0.9475 | SRCC = 0.9471 ] +INFO:root:Average loss=0.12376651817560196 +INFO:root: +[Valid][5][UTT][ MSE = 0.2251 | LCC = 0.9479 | SRCC = 0.9481 ] +INFO:root:Average loss=0.11315364986658097 +INFO:root: +[Valid][6][UTT][ MSE = 0.2690 | LCC = 0.9493 | SRCC = 0.9497 ] +INFO:root:Average loss=0.1021771268248558 +INFO:root: +[Valid][7][UTT][ MSE = 0.2839 | LCC = 0.9500 | SRCC = 0.9500 ] +INFO:root:Average loss=0.09653241819143295 +INFO:root: +[Valid][8][UTT][ MSE = 0.2269 | LCC = 0.9508 | SRCC = 0.9509 ] +INFO:root:Average loss=0.0871372711956501 +INFO:root: +[Valid][9][UTT][ MSE = 0.2007 | LCC = 0.9517 | SRCC = 0.9512 ] +INFO:root:Average loss=0.08229440996050835 +INFO:root: +[Valid][10][UTT][ MSE = 0.1970 | LCC = 0.9513 | SRCC = 0.9499 ] +INFO:root:Average loss=0.07431881454586983 +INFO:root: +[Valid][11][UTT][ MSE = 0.1926 | LCC = 0.9520 | SRCC = 0.9512 ] +INFO:root:Average loss=0.07261369603872299 +INFO:root: +[Valid][12][UTT][ MSE = 0.2162 | LCC = 0.9520 | SRCC = 0.9511 ] +INFO:root:Average loss=0.06527467966079711 +INFO:root: +[Valid][13][UTT][ MSE = 0.2009 | LCC = 0.9535 | SRCC = 0.9523 ] +INFO:root:Average loss=0.05860449504852295 +INFO:root: +[Valid][14][UTT][ MSE = 0.2104 | LCC = 0.9516 | SRCC = 0.9515 ] +INFO:root:Average loss=0.057510033637285234 +INFO:root: +[Valid][15][UTT][ MSE = 0.2019 | LCC = 0.9529 | SRCC = 0.9523 ] +INFO:root:Average loss=0.055362048029899594 +INFO:root: +[Valid][16][UTT][ MSE = 0.1898 | LCC = 0.9530 | SRCC = 0.9514 ] +INFO:root:Average loss=0.05099905371665955 +INFO:root: +[Valid][17][UTT][ MSE = 0.1937 | LCC = 0.9522 | SRCC = 0.9506 ] +INFO:root:Average loss=0.04922864617407322 +INFO:root: +[Valid][18][UTT][ MSE = 0.2055 | LCC = 0.9525 | SRCC = 0.9510 ] +INFO:root:Average loss=0.04761053936183453 +INFO:root: +[Valid][19][UTT][ MSE = 0.1911 | LCC = 0.9522 | SRCC = 0.9506 ] +INFO:root:Average loss=0.043699118316173555 +INFO:root: +[Valid][20][UTT][ MSE = 0.2161 | LCC = 0.9512 | SRCC = 0.9502 ] +INFO:root:Average loss=0.04208632671833038 +INFO:root: +[Valid][21][UTT][ MSE = 0.2412 | LCC = 0.9494 | SRCC = 0.9483 ] +INFO:root:Average loss=0.038514207944273945 +INFO:root: +[Valid][22][UTT][ MSE = 0.2358 | LCC = 0.9523 | SRCC = 0.9519 ] +INFO:root:Average loss=0.036645598217844966 +INFO:root: +[Valid][23][UTT][ MSE = 0.2005 | LCC = 0.9529 | SRCC = 0.9517 ] +INFO:root:Average loss=0.03616778099536896 +INFO:root: +[Valid][24][UTT][ MSE = 0.1920 | LCC = 0.9535 | SRCC = 0.9520 ] +INFO:root:Average loss=0.03495833389461041 +INFO:root: +[Valid][25][UTT][ MSE = 0.1988 | LCC = 0.9514 | SRCC = 0.9511 ] +INFO:root:Average loss=0.03206825006008148 +INFO:root: +[Valid][26][UTT][ MSE = 0.1873 | LCC = 0.9535 | SRCC = 0.9519 ] +INFO:root:Average loss=0.030375116378068924 +INFO:root: +[Valid][27][UTT][ MSE = 0.1942 | LCC = 0.9518 | SRCC = 0.9505 ] +INFO:root:Average loss=0.030753920525312423 +INFO:root: +[Valid][28][UTT][ MSE = 0.1980 | LCC = 0.9509 | SRCC = 0.9502 ] +INFO:root:Average loss=0.028391597509384154 +INFO:root: +[Valid][29][UTT][ MSE = 0.1949 | LCC = 0.9528 | SRCC = 0.9513 ] +INFO:root:Average loss=0.026957260347902776 +INFO:root: +[Valid][30][UTT][ MSE = 0.1927 | LCC = 0.9515 | SRCC = 0.9500 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2032 | LCC = 0.9542 | SRCC = 0.9545 ] diff --git a/results_tencent/w2v2_large_asr_en/layer12/run1/config.gin b/results_tencent/w2v2_large_asr_en/layer12/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer12/run1/model.pt b/results_tencent/w2v2_large_asr_en/layer12/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..47b445b2a0a3e60a36cb674f9b106b86921db9f2 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:068f827b4e26008dac4f4ad6a7df63ec4066478904352478d90efbd4c561646f +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer12/run1/model_best.pt b/results_tencent/w2v2_large_asr_en/layer12/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e7bce1f22178823e094acdfca5194f90411d0908 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:abf5f71f20517f37cd1f136856396f91a6a0c6e2061cf47e29efbd942b4c05df +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer12/run1/train.log b/results_tencent/w2v2_large_asr_en/layer12/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ef6444d6d6b78c3f2748ecb0b2c109e30304b891 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5242924990653992 +INFO:root: +[Valid][0][UTT][ MSE = 0.3788 | LCC = 0.8942 | SRCC = 0.8931 ] +INFO:root:Average loss=0.24546742510795594 +INFO:root: +[Valid][1][UTT][ MSE = 0.3027 | LCC = 0.9209 | SRCC = 0.9223 ] +INFO:root:Average loss=0.18577863663434982 +INFO:root: +[Valid][2][UTT][ MSE = 0.2813 | LCC = 0.9356 | SRCC = 0.9340 ] +INFO:root:Average loss=0.16184018909931183 +INFO:root: +[Valid][3][UTT][ MSE = 0.2496 | LCC = 0.9412 | SRCC = 0.9415 ] +INFO:root:Average loss=0.13380484807491302 +INFO:root: +[Valid][4][UTT][ MSE = 0.2195 | LCC = 0.9447 | SRCC = 0.9466 ] +INFO:root:Average loss=0.12330664277076721 +INFO:root: +[Valid][5][UTT][ MSE = 0.2310 | LCC = 0.9502 | SRCC = 0.9500 ] +INFO:root:Average loss=0.10776367819309235 +INFO:root: +[Valid][6][UTT][ MSE = 0.2348 | LCC = 0.9504 | SRCC = 0.9501 ] +INFO:root:Average loss=0.10152350184321403 +INFO:root: +[Valid][7][UTT][ MSE = 0.1826 | LCC = 0.9526 | SRCC = 0.9525 ] +INFO:root:Average loss=0.09220429676771164 +INFO:root: +[Valid][8][UTT][ MSE = 0.2323 | LCC = 0.9520 | SRCC = 0.9519 ] +INFO:root:Average loss=0.08813998591899871 +INFO:root: +[Valid][9][UTT][ MSE = 0.2221 | LCC = 0.9508 | SRCC = 0.9503 ] +INFO:root:Average loss=0.08187764972448348 +INFO:root: +[Valid][10][UTT][ MSE = 0.1963 | LCC = 0.9533 | SRCC = 0.9525 ] +INFO:root:Average loss=0.07334526661038399 +INFO:root: +[Valid][11][UTT][ MSE = 0.1770 | LCC = 0.9530 | SRCC = 0.9522 ] +INFO:root:Average loss=0.06764051532745362 +INFO:root: +[Valid][12][UTT][ MSE = 0.1863 | LCC = 0.9535 | SRCC = 0.9526 ] +INFO:root:Average loss=0.06379980775713921 +INFO:root: +[Valid][13][UTT][ MSE = 0.1914 | LCC = 0.9536 | SRCC = 0.9529 ] +INFO:root:Average loss=0.057914644062519074 +INFO:root: +[Valid][14][UTT][ MSE = 0.2259 | LCC = 0.9540 | SRCC = 0.9534 ] +INFO:root:Average loss=0.05512319526076317 +INFO:root: +[Valid][15][UTT][ MSE = 0.2205 | LCC = 0.9510 | SRCC = 0.9500 ] +INFO:root:Average loss=0.05528501798212528 +INFO:root: +[Valid][16][UTT][ MSE = 0.2199 | LCC = 0.9544 | SRCC = 0.9541 ] +INFO:root:Average loss=0.049439133182168006 +INFO:root: +[Valid][17][UTT][ MSE = 0.2171 | LCC = 0.9530 | SRCC = 0.9525 ] +INFO:root:Average loss=0.049794646456837656 +INFO:root: +[Valid][18][UTT][ MSE = 0.2210 | LCC = 0.9545 | SRCC = 0.9539 ] +INFO:root:Average loss=0.045176528945565224 +INFO:root: +[Valid][19][UTT][ MSE = 0.2093 | LCC = 0.9544 | SRCC = 0.9537 ] +INFO:root:Average loss=0.040274559304118156 +INFO:root: +[Valid][20][UTT][ MSE = 0.2322 | LCC = 0.9528 | SRCC = 0.9519 ] +INFO:root:Average loss=0.037963664829730986 +INFO:root: +[Valid][21][UTT][ MSE = 0.1920 | LCC = 0.9537 | SRCC = 0.9527 ] +INFO:root:Average loss=0.03684345783293247 +INFO:root: +[Valid][22][UTT][ MSE = 0.1989 | LCC = 0.9549 | SRCC = 0.9538 ] +INFO:root:Average loss=0.037093067303299906 +INFO:root: +[Valid][23][UTT][ MSE = 0.2083 | LCC = 0.9552 | SRCC = 0.9541 ] +INFO:root:Average loss=0.03685079918801785 +INFO:root: +[Valid][24][UTT][ MSE = 0.2049 | LCC = 0.9553 | SRCC = 0.9544 ] +INFO:root:Average loss=0.032861173123121264 +INFO:root: +[Valid][25][UTT][ MSE = 0.1961 | LCC = 0.9542 | SRCC = 0.9538 ] +INFO:root:Average loss=0.030662636265158652 +INFO:root: +[Valid][26][UTT][ MSE = 0.1996 | LCC = 0.9541 | SRCC = 0.9537 ] +INFO:root:Average loss=0.029129821971058845 +INFO:root: +[Valid][27][UTT][ MSE = 0.2035 | LCC = 0.9540 | SRCC = 0.9536 ] +INFO:root:Average loss=0.029114353954792022 +INFO:root: +[Valid][28][UTT][ MSE = 0.2031 | LCC = 0.9548 | SRCC = 0.9543 ] +INFO:root:Average loss=0.0274615458548069 +INFO:root: +[Valid][29][UTT][ MSE = 0.2411 | LCC = 0.9537 | SRCC = 0.9531 ] +INFO:root:Average loss=0.026666629649698733 +INFO:root: +[Valid][30][UTT][ MSE = 0.2095 | LCC = 0.9546 | SRCC = 0.9535 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2065 | LCC = 0.9542 | SRCC = 0.9548 ] diff --git a/results_tencent/w2v2_large_asr_en/layer12/run2/config.gin b/results_tencent/w2v2_large_asr_en/layer12/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer12/run2/model.pt b/results_tencent/w2v2_large_asr_en/layer12/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ddbd39584955345b73c9deed27738beebfe279a0 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6e4ac779f0ae2adf79ab1e5c41fedf75f8cdec951fbba9079c3e19c2aa006334 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer12/run2/model_best.pt b/results_tencent/w2v2_large_asr_en/layer12/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0f14ebd94590b14aade41be1bf4e6f4f710e5106 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c18347eb9d88284d63e91c24739e6e1a6703c195440327d404f064c14e959731 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer12/run2/train.log b/results_tencent/w2v2_large_asr_en/layer12/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3da7402179c1dcb894e5e2e18b699c1dc30a79da --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6757287484407425 +INFO:root: +[Valid][0][UTT][ MSE = 0.3738 | LCC = 0.8893 | SRCC = 0.8886 ] +INFO:root:Average loss=0.2679739565849304 +INFO:root: +[Valid][1][UTT][ MSE = 0.3158 | LCC = 0.9248 | SRCC = 0.9232 ] +INFO:root:Average loss=0.19419372761249543 +INFO:root: +[Valid][2][UTT][ MSE = 0.3382 | LCC = 0.9328 | SRCC = 0.9344 ] +INFO:root:Average loss=0.15869603860378265 +INFO:root: +[Valid][3][UTT][ MSE = 0.2362 | LCC = 0.9432 | SRCC = 0.9440 ] +INFO:root:Average loss=0.1375147550702095 +INFO:root: +[Valid][4][UTT][ MSE = 0.2261 | LCC = 0.9471 | SRCC = 0.9472 ] +INFO:root:Average loss=0.12159028851985931 +INFO:root: +[Valid][5][UTT][ MSE = 0.2012 | LCC = 0.9508 | SRCC = 0.9508 ] +INFO:root:Average loss=0.11006107926368713 +INFO:root: +[Valid][6][UTT][ MSE = 0.2302 | LCC = 0.9523 | SRCC = 0.9515 ] +INFO:root:Average loss=0.09858286887407303 +INFO:root: +[Valid][7][UTT][ MSE = 0.1960 | LCC = 0.9506 | SRCC = 0.9507 ] +INFO:root:Average loss=0.09057428655028343 +INFO:root: +[Valid][8][UTT][ MSE = 0.2456 | LCC = 0.9522 | SRCC = 0.9519 ] +INFO:root:Average loss=0.08236310881376266 +INFO:root: +[Valid][9][UTT][ MSE = 0.1770 | LCC = 0.9548 | SRCC = 0.9539 ] +INFO:root:Average loss=0.07861484640836716 +INFO:root: +[Valid][10][UTT][ MSE = 0.2087 | LCC = 0.9526 | SRCC = 0.9523 ] +INFO:root:Average loss=0.07109885635972023 +INFO:root: +[Valid][11][UTT][ MSE = 0.1994 | LCC = 0.9546 | SRCC = 0.9539 ] +INFO:root:Average loss=0.06370282331109046 +INFO:root: +[Valid][12][UTT][ MSE = 0.2028 | LCC = 0.9544 | SRCC = 0.9538 ] +INFO:root:Average loss=0.0634592150747776 +INFO:root: +[Valid][13][UTT][ MSE = 0.2498 | LCC = 0.9549 | SRCC = 0.9549 ] +INFO:root:Average loss=0.05790201726555824 +INFO:root: +[Valid][14][UTT][ MSE = 0.1918 | LCC = 0.9546 | SRCC = 0.9541 ] +INFO:root:Average loss=0.05352707424759865 +INFO:root: +[Valid][15][UTT][ MSE = 0.1993 | LCC = 0.9553 | SRCC = 0.9543 ] +INFO:root:Average loss=0.0517303583920002 +INFO:root: +[Valid][16][UTT][ MSE = 0.2238 | LCC = 0.9555 | SRCC = 0.9544 ] +INFO:root:Average loss=0.04881771016120911 +INFO:root: +[Valid][17][UTT][ MSE = 0.2170 | LCC = 0.9544 | SRCC = 0.9535 ] +INFO:root:Average loss=0.044669029638171195 +INFO:root: +[Valid][18][UTT][ MSE = 0.1793 | LCC = 0.9558 | SRCC = 0.9545 ] +INFO:root:Average loss=0.04257909871637821 +INFO:root: +[Valid][19][UTT][ MSE = 0.1812 | LCC = 0.9558 | SRCC = 0.9549 ] +INFO:root:Average loss=0.042038551509380344 +INFO:root: +[Valid][20][UTT][ MSE = 0.1912 | LCC = 0.9558 | SRCC = 0.9545 ] +INFO:root:Average loss=0.03811486278474331 +INFO:root: +[Valid][21][UTT][ MSE = 0.1957 | LCC = 0.9551 | SRCC = 0.9541 ] +INFO:root:Average loss=0.04279070121049881 +INFO:root: +[Valid][22][UTT][ MSE = 0.1890 | LCC = 0.9557 | SRCC = 0.9547 ] +INFO:root:Average loss=0.03414521931111813 +INFO:root: +[Valid][23][UTT][ MSE = 0.1868 | LCC = 0.9548 | SRCC = 0.9544 ] +INFO:root:Average loss=0.03581663563847542 +INFO:root: +[Valid][24][UTT][ MSE = 0.1795 | LCC = 0.9557 | SRCC = 0.9546 ] +INFO:root:Average loss=0.03218266187608242 +INFO:root: +[Valid][25][UTT][ MSE = 0.1991 | LCC = 0.9543 | SRCC = 0.9532 ] +INFO:root:Average loss=0.0318458870947361 +INFO:root: +[Valid][26][UTT][ MSE = 0.2148 | LCC = 0.9555 | SRCC = 0.9547 ] +INFO:root:Average loss=0.02922616572678089 +INFO:root: +[Valid][27][UTT][ MSE = 0.1823 | LCC = 0.9561 | SRCC = 0.9553 ] +INFO:root:Average loss=0.027773556597530842 +INFO:root: +[Valid][28][UTT][ MSE = 0.1828 | LCC = 0.9573 | SRCC = 0.9562 ] +INFO:root:Average loss=0.027801938980817793 +INFO:root: +[Valid][29][UTT][ MSE = 0.1861 | LCC = 0.9577 | SRCC = 0.9564 ] +INFO:root:Average loss=0.026241359099745752 +INFO:root: +[Valid][30][UTT][ MSE = 0.1822 | LCC = 0.9569 | SRCC = 0.9559 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1953 | LCC = 0.9544 | SRCC = 0.9549 ] diff --git a/results_tencent/w2v2_large_asr_en/layer12/run3/config.gin b/results_tencent/w2v2_large_asr_en/layer12/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer12/run3/model.pt b/results_tencent/w2v2_large_asr_en/layer12/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9bb8f2497a5b9d94758aba66c418cb0750863177 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e9e85295d8d046ca4cf1035341ce4758cf3723b76dea8db0a204a2a934094d69 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer12/run3/model_best.pt b/results_tencent/w2v2_large_asr_en/layer12/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..933727280bf6081f2f13b512e68c4aa51f832d97 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f7d5960ee466bf238d57881dfcb4a6235dfcd5268957c7f491838321e2517e5c +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer12/run3/train.log b/results_tencent/w2v2_large_asr_en/layer12/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3ad99d95216e611d39a55a35de98aad788b0bfef --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5375166082382202 +INFO:root: +[Valid][0][UTT][ MSE = 0.3422 | LCC = 0.8941 | SRCC = 0.8942 ] +INFO:root:Average loss=0.250018660068512 +INFO:root: +[Valid][1][UTT][ MSE = 0.2841 | LCC = 0.9231 | SRCC = 0.9223 ] +INFO:root:Average loss=0.19470773559808732 +INFO:root: +[Valid][2][UTT][ MSE = 0.2563 | LCC = 0.9360 | SRCC = 0.9343 ] +INFO:root:Average loss=0.16482936161756515 +INFO:root: +[Valid][3][UTT][ MSE = 0.2459 | LCC = 0.9436 | SRCC = 0.9427 ] +INFO:root:Average loss=0.14472211700677873 +INFO:root: +[Valid][4][UTT][ MSE = 0.2389 | LCC = 0.9463 | SRCC = 0.9452 ] +INFO:root:Average loss=0.12842424088716506 +INFO:root: +[Valid][5][UTT][ MSE = 0.2006 | LCC = 0.9477 | SRCC = 0.9467 ] +INFO:root:Average loss=0.11677552860975265 +INFO:root: +[Valid][6][UTT][ MSE = 0.2147 | LCC = 0.9510 | SRCC = 0.9504 ] +INFO:root:Average loss=0.10746226644515991 +INFO:root: +[Valid][7][UTT][ MSE = 0.2142 | LCC = 0.9515 | SRCC = 0.9511 ] +INFO:root:Average loss=0.09361090683937073 +INFO:root: +[Valid][8][UTT][ MSE = 0.1963 | LCC = 0.9514 | SRCC = 0.9504 ] +INFO:root:Average loss=0.08734988901019096 +INFO:root: +[Valid][9][UTT][ MSE = 0.1770 | LCC = 0.9525 | SRCC = 0.9518 ] +INFO:root:Average loss=0.08280417311191558 +INFO:root: +[Valid][10][UTT][ MSE = 0.2417 | LCC = 0.9509 | SRCC = 0.9508 ] +INFO:root:Average loss=0.07240606063604355 +INFO:root: +[Valid][11][UTT][ MSE = 0.1959 | LCC = 0.9531 | SRCC = 0.9521 ] +INFO:root:Average loss=0.07137446177005768 +INFO:root: +[Valid][12][UTT][ MSE = 0.2172 | LCC = 0.9487 | SRCC = 0.9481 ] +INFO:root:Average loss=0.06360782867670059 +INFO:root: +[Valid][13][UTT][ MSE = 0.1830 | LCC = 0.9539 | SRCC = 0.9528 ] +INFO:root:Average loss=0.059254483669996263 +INFO:root: +[Valid][14][UTT][ MSE = 0.1848 | LCC = 0.9542 | SRCC = 0.9532 ] +INFO:root:Average loss=0.05762724828720093 +INFO:root: +[Valid][15][UTT][ MSE = 0.2116 | LCC = 0.9553 | SRCC = 0.9539 ] +INFO:root:Average loss=0.05284594468772411 +INFO:root: +[Valid][16][UTT][ MSE = 0.1867 | LCC = 0.9538 | SRCC = 0.9526 ] +INFO:root:Average loss=0.048749615862965584 +INFO:root: +[Valid][17][UTT][ MSE = 0.1930 | LCC = 0.9541 | SRCC = 0.9526 ] +INFO:root:Average loss=0.046826241075992585 +INFO:root: +[Valid][18][UTT][ MSE = 0.1815 | LCC = 0.9546 | SRCC = 0.9532 ] +INFO:root:Average loss=0.04577606678009033 +INFO:root: +[Valid][19][UTT][ MSE = 0.1729 | LCC = 0.9531 | SRCC = 0.9522 ] +INFO:root:Average loss=0.04313128617405891 +INFO:root: +[Valid][20][UTT][ MSE = 0.1877 | LCC = 0.9535 | SRCC = 0.9518 ] +INFO:root:Average loss=0.040350855335593225 +INFO:root: +[Valid][21][UTT][ MSE = 0.1891 | LCC = 0.9537 | SRCC = 0.9534 ] +INFO:root:Average loss=0.04337225806713104 +INFO:root: +[Valid][22][UTT][ MSE = 0.1884 | LCC = 0.9541 | SRCC = 0.9531 ] +INFO:root:Average loss=0.038789474800229075 +INFO:root: +[Valid][23][UTT][ MSE = 0.1656 | LCC = 0.9532 | SRCC = 0.9530 ] +INFO:root:Average loss=0.036198144808411596 +INFO:root: +[Valid][24][UTT][ MSE = 0.1907 | LCC = 0.9546 | SRCC = 0.9537 ] +INFO:root:Average loss=0.034880789548158646 +INFO:root: +[Valid][25][UTT][ MSE = 0.1824 | LCC = 0.9549 | SRCC = 0.9541 ] +INFO:root:Average loss=0.032877663999795914 +INFO:root: +[Valid][26][UTT][ MSE = 0.1796 | LCC = 0.9548 | SRCC = 0.9536 ] +INFO:root:Average loss=0.03280086897313595 +INFO:root: +[Valid][27][UTT][ MSE = 0.1929 | LCC = 0.9551 | SRCC = 0.9539 ] +INFO:root:Average loss=0.030508009284734725 +INFO:root: +[Valid][28][UTT][ MSE = 0.2112 | LCC = 0.9538 | SRCC = 0.9535 ] +INFO:root:Average loss=0.03114625346660614 +INFO:root: +[Valid][29][UTT][ MSE = 0.2103 | LCC = 0.9551 | SRCC = 0.9543 ] +INFO:root:Average loss=0.028319912195205688 +INFO:root: +[Valid][30][UTT][ MSE = 0.1825 | LCC = 0.9548 | SRCC = 0.9529 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2198 | LCC = 0.9535 | SRCC = 0.9541 ] diff --git a/results_tencent/w2v2_large_asr_en/layer12/run4/config.gin b/results_tencent/w2v2_large_asr_en/layer12/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer12/run4/model.pt b/results_tencent/w2v2_large_asr_en/layer12/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5eb66e14bb56941bc55326f51a721aa7bf3126a9 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f443cb5381d8d4b019b7db880487bbf2df2a7f99270284032e87668c4f523183 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer12/run4/model_best.pt b/results_tencent/w2v2_large_asr_en/layer12/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9825e590817c0cdca766766886201cd9147afd1e --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ac5b83a86b1f1d81936a4a3525997d00768195c62326beb89c25ef7354eda4c +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer12/run4/train.log b/results_tencent/w2v2_large_asr_en/layer12/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..08af7e28c14e0da79a9ace4bc6f7ed0b807cb158 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer12/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.48969306945800783 +INFO:root: +[Valid][0][UTT][ MSE = 0.3890 | LCC = 0.8956 | SRCC = 0.8985 ] +INFO:root:Average loss=0.24037487196922303 +INFO:root: +[Valid][1][UTT][ MSE = 0.2633 | LCC = 0.9270 | SRCC = 0.9250 ] +INFO:root:Average loss=0.18700029945373536 +INFO:root: +[Valid][2][UTT][ MSE = 0.2667 | LCC = 0.9352 | SRCC = 0.9341 ] +INFO:root:Average loss=0.15695950227975847 +INFO:root: +[Valid][3][UTT][ MSE = 0.2067 | LCC = 0.9428 | SRCC = 0.9419 ] +INFO:root:Average loss=0.13714287787675858 +INFO:root: +[Valid][4][UTT][ MSE = 0.2393 | LCC = 0.9458 | SRCC = 0.9460 ] +INFO:root:Average loss=0.12443991750478745 +INFO:root: +[Valid][5][UTT][ MSE = 0.1958 | LCC = 0.9499 | SRCC = 0.9491 ] +INFO:root:Average loss=0.11410837212204933 +INFO:root: +[Valid][6][UTT][ MSE = 0.2178 | LCC = 0.9508 | SRCC = 0.9503 ] +INFO:root:Average loss=0.09912937197089196 +INFO:root: +[Valid][7][UTT][ MSE = 0.2318 | LCC = 0.9529 | SRCC = 0.9523 ] +INFO:root:Average loss=0.09005245691537857 +INFO:root: +[Valid][8][UTT][ MSE = 0.1924 | LCC = 0.9530 | SRCC = 0.9523 ] +INFO:root:Average loss=0.08485665133595467 +INFO:root: +[Valid][9][UTT][ MSE = 0.1970 | LCC = 0.9528 | SRCC = 0.9523 ] +INFO:root:Average loss=0.07679563370347023 +INFO:root: +[Valid][10][UTT][ MSE = 0.2059 | LCC = 0.9538 | SRCC = 0.9532 ] +INFO:root:Average loss=0.07210501632094383 +INFO:root: +[Valid][11][UTT][ MSE = 0.2231 | LCC = 0.9541 | SRCC = 0.9539 ] +INFO:root:Average loss=0.07221907383203506 +INFO:root: +[Valid][12][UTT][ MSE = 0.2187 | LCC = 0.9537 | SRCC = 0.9528 ] +INFO:root:Average loss=0.06421428018808364 +INFO:root: +[Valid][13][UTT][ MSE = 0.2001 | LCC = 0.9552 | SRCC = 0.9546 ] +INFO:root:Average loss=0.05703340637683869 +INFO:root: +[Valid][14][UTT][ MSE = 0.1932 | LCC = 0.9551 | SRCC = 0.9537 ] +INFO:root:Average loss=0.052466608464717866 +INFO:root: +[Valid][15][UTT][ MSE = 0.1937 | LCC = 0.9548 | SRCC = 0.9538 ] +INFO:root:Average loss=0.04959605272114277 +INFO:root: +[Valid][16][UTT][ MSE = 0.1861 | LCC = 0.9546 | SRCC = 0.9536 ] +INFO:root:Average loss=0.04710390955209732 +INFO:root: +[Valid][17][UTT][ MSE = 0.2035 | LCC = 0.9539 | SRCC = 0.9533 ] +INFO:root:Average loss=0.04568047751486302 +INFO:root: +[Valid][18][UTT][ MSE = 0.1879 | LCC = 0.9539 | SRCC = 0.9528 ] +INFO:root:Average loss=0.04698190812766552 +INFO:root: +[Valid][19][UTT][ MSE = 0.1953 | LCC = 0.9539 | SRCC = 0.9531 ] +INFO:root:Average loss=0.03946910098195076 +INFO:root: +[Valid][20][UTT][ MSE = 0.2207 | LCC = 0.9544 | SRCC = 0.9543 ] +INFO:root:Average loss=0.039535019978880884 +INFO:root: +[Valid][21][UTT][ MSE = 0.1963 | LCC = 0.9546 | SRCC = 0.9534 ] +INFO:root:Average loss=0.03632773269712925 +INFO:root: +[Valid][22][UTT][ MSE = 0.2244 | LCC = 0.9541 | SRCC = 0.9534 ] +INFO:root:Average loss=0.036970660865306856 +INFO:root: +[Valid][23][UTT][ MSE = 0.2280 | LCC = 0.9540 | SRCC = 0.9530 ] +INFO:root:Average loss=0.03327654682099819 +INFO:root: +[Valid][24][UTT][ MSE = 0.1846 | LCC = 0.9544 | SRCC = 0.9532 ] +INFO:root:Average loss=0.031694126591086384 +INFO:root: +[Valid][25][UTT][ MSE = 0.2027 | LCC = 0.9542 | SRCC = 0.9532 ] +INFO:root:Average loss=0.02941347177326679 +INFO:root: +[Valid][26][UTT][ MSE = 0.2034 | LCC = 0.9535 | SRCC = 0.9526 ] +INFO:root:Average loss=0.02983308358490467 +INFO:root: +[Valid][27][UTT][ MSE = 0.1994 | LCC = 0.9540 | SRCC = 0.9530 ] +INFO:root:Average loss=0.027994363069534303 +INFO:root: +[Valid][28][UTT][ MSE = 0.1997 | LCC = 0.9522 | SRCC = 0.9511 ] +INFO:root:Average loss=0.027168861135840414 +INFO:root: +[Valid][29][UTT][ MSE = 0.1853 | LCC = 0.9537 | SRCC = 0.9527 ] +INFO:root:Average loss=0.02756318011879921 +INFO:root: +[Valid][30][UTT][ MSE = 0.2046 | LCC = 0.9544 | SRCC = 0.9535 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2029 | LCC = 0.9547 | SRCC = 0.9552 ] diff --git a/results_tencent/w2v2_large_asr_en/layer14/run0/config.gin b/results_tencent/w2v2_large_asr_en/layer14/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer14/run0/model.pt b/results_tencent/w2v2_large_asr_en/layer14/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0e350e4bd91d5f0a6b46b3ffce80d2bdda412b47 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:89202e9ddb40e91bb675c6d5abb698d9c3cd1df3e5f20be9e93fd288db4f0dd6 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer14/run0/model_best.pt b/results_tencent/w2v2_large_asr_en/layer14/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9d4d3de20192cb15f7366ede4e3b4f72bc15cb9e --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:65276c74e5efb22ee8c2c9d6d26d87d02b58bedc981724e9efef8650626cf4ff +size 3404998 diff --git a/results_tencent/w2v2_large_asr_en/layer14/run0/train.log b/results_tencent/w2v2_large_asr_en/layer14/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c09cc763e586f7fcd2109c4bd40f1fc37bf245ef --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.54922159075737 +INFO:root: +[Valid][0][UTT][ MSE = 0.3538 | LCC = 0.8956 | SRCC = 0.8958 ] +INFO:root:Average loss=0.24895685875415802 +INFO:root: +[Valid][1][UTT][ MSE = 0.2994 | LCC = 0.9130 | SRCC = 0.9137 ] +INFO:root:Average loss=0.20395837098360062 +INFO:root: +[Valid][2][UTT][ MSE = 0.2614 | LCC = 0.9296 | SRCC = 0.9290 ] +INFO:root:Average loss=0.17405326241254807 +INFO:root: +[Valid][3][UTT][ MSE = 0.2404 | LCC = 0.9355 | SRCC = 0.9360 ] +INFO:root:Average loss=0.1544607428908348 +INFO:root: +[Valid][4][UTT][ MSE = 0.2435 | LCC = 0.9401 | SRCC = 0.9395 ] +INFO:root:Average loss=0.13882305181026458 +INFO:root: +[Valid][5][UTT][ MSE = 0.2167 | LCC = 0.9440 | SRCC = 0.9432 ] +INFO:root:Average loss=0.12702199053764343 +INFO:root: +[Valid][6][UTT][ MSE = 0.2366 | LCC = 0.9464 | SRCC = 0.9462 ] +INFO:root:Average loss=0.11459969568252563 +INFO:root: +[Valid][7][UTT][ MSE = 0.2025 | LCC = 0.9490 | SRCC = 0.9481 ] +INFO:root:Average loss=0.1041225359737873 +INFO:root: +[Valid][8][UTT][ MSE = 0.1991 | LCC = 0.9448 | SRCC = 0.9459 ] +INFO:root:Average loss=0.09842336928844451 +INFO:root: +[Valid][9][UTT][ MSE = 0.2124 | LCC = 0.9488 | SRCC = 0.9480 ] +INFO:root:Average loss=0.0926875322163105 +INFO:root: +[Valid][10][UTT][ MSE = 0.2318 | LCC = 0.9357 | SRCC = 0.9392 ] +INFO:root:Average loss=0.0847308597266674 +INFO:root: +[Valid][11][UTT][ MSE = 0.1877 | LCC = 0.9516 | SRCC = 0.9506 ] +INFO:root:Average loss=0.0778160948753357 +INFO:root: +[Valid][12][UTT][ MSE = 0.1966 | LCC = 0.9504 | SRCC = 0.9493 ] +INFO:root:Average loss=0.073416448533535 +INFO:root: +[Valid][13][UTT][ MSE = 0.2164 | LCC = 0.9506 | SRCC = 0.9489 ] +INFO:root:Average loss=0.06983406618237495 +INFO:root: +[Valid][14][UTT][ MSE = 0.2145 | LCC = 0.9515 | SRCC = 0.9506 ] +INFO:root:Average loss=0.06513086065649987 +INFO:root: +[Valid][15][UTT][ MSE = 0.2095 | LCC = 0.9498 | SRCC = 0.9495 ] +INFO:root:Average loss=0.05859164160490036 +INFO:root: +[Valid][16][UTT][ MSE = 0.2065 | LCC = 0.9512 | SRCC = 0.9501 ] +INFO:root:Average loss=0.06008460786938667 +INFO:root: +[Valid][17][UTT][ MSE = 0.1808 | LCC = 0.9522 | SRCC = 0.9511 ] +INFO:root:Average loss=0.054771614700555804 +INFO:root: +[Valid][18][UTT][ MSE = 0.1818 | LCC = 0.9498 | SRCC = 0.9494 ] +INFO:root:Average loss=0.054267239555716516 +INFO:root: +[Valid][19][UTT][ MSE = 0.1732 | LCC = 0.9476 | SRCC = 0.9486 ] +INFO:root:Average loss=0.05048366080224514 +INFO:root: +[Valid][20][UTT][ MSE = 0.1770 | LCC = 0.9519 | SRCC = 0.9513 ] +INFO:root:Average loss=0.04854577974975109 +INFO:root: +[Valid][21][UTT][ MSE = 0.1798 | LCC = 0.9532 | SRCC = 0.9525 ] +INFO:root:Average loss=0.04798381893336773 +INFO:root: +[Valid][22][UTT][ MSE = 0.1771 | LCC = 0.9532 | SRCC = 0.9524 ] +INFO:root:Average loss=0.041685299664735796 +INFO:root: +[Valid][23][UTT][ MSE = 0.1902 | LCC = 0.9505 | SRCC = 0.9506 ] +INFO:root:Average loss=0.040749021753668783 +INFO:root: +[Valid][24][UTT][ MSE = 0.2230 | LCC = 0.9435 | SRCC = 0.9460 ] +INFO:root:Average loss=0.04313231506943703 +INFO:root: +[Valid][25][UTT][ MSE = 0.1910 | LCC = 0.9528 | SRCC = 0.9522 ] +INFO:root:Average loss=0.04009281532466412 +INFO:root: +[Valid][26][UTT][ MSE = 0.2202 | LCC = 0.9442 | SRCC = 0.9431 ] +INFO:root:Average loss=0.04247490791976452 +INFO:root: +[Valid][27][UTT][ MSE = 0.1746 | LCC = 0.9517 | SRCC = 0.9511 ] +INFO:root:Average loss=0.03650574244558811 +INFO:root: +[Valid][28][UTT][ MSE = 0.1845 | LCC = 0.9534 | SRCC = 0.9525 ] +INFO:root:Average loss=0.033169330283999446 +INFO:root: +[Valid][29][UTT][ MSE = 0.1852 | LCC = 0.9530 | SRCC = 0.9521 ] +INFO:root:Average loss=0.03196757324039936 +INFO:root: +[Valid][30][UTT][ MSE = 0.1958 | LCC = 0.9521 | SRCC = 0.9506 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1878 | LCC = 0.9536 | SRCC = 0.9539 ] diff --git a/results_tencent/w2v2_large_asr_en/layer14/run1/config.gin b/results_tencent/w2v2_large_asr_en/layer14/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer14/run1/model.pt b/results_tencent/w2v2_large_asr_en/layer14/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9dca665ffdf7635fa9b166be41e2c51d7ba99659 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a08a22319baee791454557f62311f81c8482e0240e78d3e5645807b4c188576d +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer14/run1/model_best.pt b/results_tencent/w2v2_large_asr_en/layer14/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..04878ce7e33f4ff793fdf70b444cc4c9706b2c46 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:30cbd7c09fa9483e5f27ebf22df89cb0b4d0843503e7d62ef1fd96cd29a58797 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer14/run1/train.log b/results_tencent/w2v2_large_asr_en/layer14/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0586e172c95e75db8a3aa03485e98a585050dd7d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4892630400657654 +INFO:root: +[Valid][0][UTT][ MSE = 0.3949 | LCC = 0.8961 | SRCC = 0.8920 ] +INFO:root:Average loss=0.2552416575551033 +INFO:root: +[Valid][1][UTT][ MSE = 0.3408 | LCC = 0.9192 | SRCC = 0.9194 ] +INFO:root:Average loss=0.2022055692076683 +INFO:root: +[Valid][2][UTT][ MSE = 0.2760 | LCC = 0.9333 | SRCC = 0.9331 ] +INFO:root:Average loss=0.16919863033294677 +INFO:root: +[Valid][3][UTT][ MSE = 0.2403 | LCC = 0.9406 | SRCC = 0.9396 ] +INFO:root:Average loss=0.1473761703968048 +INFO:root: +[Valid][4][UTT][ MSE = 0.2478 | LCC = 0.9447 | SRCC = 0.9445 ] +INFO:root:Average loss=0.13543677371740342 +INFO:root: +[Valid][5][UTT][ MSE = 0.2462 | LCC = 0.9487 | SRCC = 0.9484 ] +INFO:root:Average loss=0.11909966176748275 +INFO:root: +[Valid][6][UTT][ MSE = 0.2601 | LCC = 0.9486 | SRCC = 0.9491 ] +INFO:root:Average loss=0.10725821456313134 +INFO:root: +[Valid][7][UTT][ MSE = 0.2390 | LCC = 0.9508 | SRCC = 0.9507 ] +INFO:root:Average loss=0.10090193718671799 +INFO:root: +[Valid][8][UTT][ MSE = 0.1901 | LCC = 0.9521 | SRCC = 0.9513 ] +INFO:root:Average loss=0.09426473501324653 +INFO:root: +[Valid][9][UTT][ MSE = 0.1994 | LCC = 0.9521 | SRCC = 0.9510 ] +INFO:root:Average loss=0.0894008238017559 +INFO:root: +[Valid][10][UTT][ MSE = 0.1942 | LCC = 0.9522 | SRCC = 0.9516 ] +INFO:root:Average loss=0.08128375428915023 +INFO:root: +[Valid][11][UTT][ MSE = 0.2173 | LCC = 0.9514 | SRCC = 0.9502 ] +INFO:root:Average loss=0.07984818494319916 +INFO:root: +[Valid][12][UTT][ MSE = 0.2025 | LCC = 0.9519 | SRCC = 0.9516 ] +INFO:root:Average loss=0.06906987172365188 +INFO:root: +[Valid][13][UTT][ MSE = 0.2226 | LCC = 0.9526 | SRCC = 0.9514 ] +INFO:root:Average loss=0.0687119662463665 +INFO:root: +[Valid][14][UTT][ MSE = 0.2018 | LCC = 0.9528 | SRCC = 0.9519 ] +INFO:root:Average loss=0.06076328006386757 +INFO:root: +[Valid][15][UTT][ MSE = 0.1690 | LCC = 0.9519 | SRCC = 0.9510 ] +INFO:root:Average loss=0.06552400729060173 +INFO:root: +[Valid][16][UTT][ MSE = 0.1794 | LCC = 0.9529 | SRCC = 0.9519 ] +INFO:root:Average loss=0.06277139380574226 +INFO:root: +[Valid][17][UTT][ MSE = 0.2122 | LCC = 0.9526 | SRCC = 0.9517 ] +INFO:root:Average loss=0.05474069005250931 +INFO:root: +[Valid][18][UTT][ MSE = 0.2134 | LCC = 0.9536 | SRCC = 0.9526 ] +INFO:root:Average loss=0.05104421249032021 +INFO:root: +[Valid][19][UTT][ MSE = 0.2091 | LCC = 0.9535 | SRCC = 0.9524 ] +INFO:root:Average loss=0.0476639733761549 +INFO:root: +[Valid][20][UTT][ MSE = 0.2007 | LCC = 0.9535 | SRCC = 0.9530 ] +INFO:root:Average loss=0.049712723419070244 +INFO:root: +[Valid][21][UTT][ MSE = 0.1900 | LCC = 0.9537 | SRCC = 0.9523 ] +INFO:root:Average loss=0.04106781105697155 +INFO:root: +[Valid][22][UTT][ MSE = 0.2154 | LCC = 0.9533 | SRCC = 0.9526 ] +INFO:root:Average loss=0.04340207900106907 +INFO:root: +[Valid][23][UTT][ MSE = 0.2198 | LCC = 0.9538 | SRCC = 0.9530 ] +INFO:root:Average loss=0.04059512905776501 +INFO:root: +[Valid][24][UTT][ MSE = 0.2182 | LCC = 0.9512 | SRCC = 0.9511 ] +INFO:root:Average loss=0.03948132736980915 +INFO:root: +[Valid][25][UTT][ MSE = 0.2061 | LCC = 0.9533 | SRCC = 0.9527 ] +INFO:root:Average loss=0.0373004230260849 +INFO:root: +[Valid][26][UTT][ MSE = 0.2138 | LCC = 0.9537 | SRCC = 0.9526 ] +INFO:root:Average loss=0.03571428829431534 +INFO:root: +[Valid][27][UTT][ MSE = 0.2245 | LCC = 0.9511 | SRCC = 0.9507 ] +INFO:root:Average loss=0.034137490101158616 +INFO:root: +[Valid][28][UTT][ MSE = 0.2242 | LCC = 0.9520 | SRCC = 0.9510 ] +INFO:root:Average loss=0.03123912301659584 +INFO:root: +[Valid][29][UTT][ MSE = 0.2162 | LCC = 0.9527 | SRCC = 0.9516 ] +INFO:root:Average loss=0.03136922983825207 +INFO:root: +[Valid][30][UTT][ MSE = 0.1989 | LCC = 0.9515 | SRCC = 0.9510 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2283 | LCC = 0.9526 | SRCC = 0.9537 ] diff --git a/results_tencent/w2v2_large_asr_en/layer14/run2/config.gin b/results_tencent/w2v2_large_asr_en/layer14/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer14/run2/model.pt b/results_tencent/w2v2_large_asr_en/layer14/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5197dd412cc529e0d51f9ae47046329b28446511 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:753199599585ef2acac18be8587ee4b4880069e68f1239ada33457e8ee595ebf +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer14/run2/model_best.pt b/results_tencent/w2v2_large_asr_en/layer14/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7784f133525e2a6f7ddb322bcfa0b8ed97358f3f --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:022a7fbf0809f27eabde5ce6402e85ccec9e0f510682eb3f48285869a974db7c +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer14/run2/train.log b/results_tencent/w2v2_large_asr_en/layer14/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2eeeb3a8c5029013501211dd170e45c91d59467e --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5335993529558182 +INFO:root: +[Valid][0][UTT][ MSE = 0.4134 | LCC = 0.8901 | SRCC = 0.8893 ] +INFO:root:Average loss=0.25513332426548 +INFO:root: +[Valid][1][UTT][ MSE = 0.2883 | LCC = 0.9192 | SRCC = 0.9185 ] +INFO:root:Average loss=0.20066920971870422 +INFO:root: +[Valid][2][UTT][ MSE = 0.2601 | LCC = 0.9348 | SRCC = 0.9339 ] +INFO:root:Average loss=0.1747853227853775 +INFO:root: +[Valid][3][UTT][ MSE = 0.2194 | LCC = 0.9412 | SRCC = 0.9404 ] +INFO:root:Average loss=0.15377047395706175 +INFO:root: +[Valid][4][UTT][ MSE = 0.2819 | LCC = 0.9443 | SRCC = 0.9443 ] +INFO:root:Average loss=0.13315432491898538 +INFO:root: +[Valid][5][UTT][ MSE = 0.2543 | LCC = 0.9451 | SRCC = 0.9446 ] +INFO:root:Average loss=0.1216096328496933 +INFO:root: +[Valid][6][UTT][ MSE = 0.2386 | LCC = 0.9471 | SRCC = 0.9469 ] +INFO:root:Average loss=0.11663855969905854 +INFO:root: +[Valid][7][UTT][ MSE = 0.2185 | LCC = 0.9486 | SRCC = 0.9481 ] +INFO:root:Average loss=0.10525313559174537 +INFO:root: +[Valid][8][UTT][ MSE = 0.2415 | LCC = 0.9461 | SRCC = 0.9470 ] +INFO:root:Average loss=0.0956497251689434 +INFO:root: +[Valid][9][UTT][ MSE = 0.2104 | LCC = 0.9518 | SRCC = 0.9504 ] +INFO:root:Average loss=0.09435685795545579 +INFO:root: +[Valid][10][UTT][ MSE = 0.2181 | LCC = 0.9477 | SRCC = 0.9472 ] +INFO:root:Average loss=0.08948610943555832 +INFO:root: +[Valid][11][UTT][ MSE = 0.2263 | LCC = 0.9506 | SRCC = 0.9494 ] +INFO:root:Average loss=0.08323833549022674 +INFO:root: +[Valid][12][UTT][ MSE = 0.2337 | LCC = 0.9488 | SRCC = 0.9476 ] +INFO:root:Average loss=0.07929729914665222 +INFO:root: +[Valid][13][UTT][ MSE = 0.2049 | LCC = 0.9511 | SRCC = 0.9507 ] +INFO:root:Average loss=0.07074439945816993 +INFO:root: +[Valid][14][UTT][ MSE = 0.2183 | LCC = 0.9499 | SRCC = 0.9490 ] +INFO:root:Average loss=0.06689348131418228 +INFO:root: +[Valid][15][UTT][ MSE = 0.2081 | LCC = 0.9485 | SRCC = 0.9468 ] +INFO:root:Average loss=0.06398344498872757 +INFO:root: +[Valid][16][UTT][ MSE = 0.1942 | LCC = 0.9521 | SRCC = 0.9514 ] +INFO:root:Average loss=0.062459114164113996 +INFO:root: +[Valid][17][UTT][ MSE = 0.2258 | LCC = 0.9495 | SRCC = 0.9483 ] +INFO:root:Average loss=0.055662234842777256 +INFO:root: +[Valid][18][UTT][ MSE = 0.1838 | LCC = 0.9504 | SRCC = 0.9494 ] +INFO:root:Average loss=0.05354380936920643 +INFO:root: +[Valid][19][UTT][ MSE = 0.2097 | LCC = 0.9505 | SRCC = 0.9491 ] +INFO:root:Average loss=0.05058123806118965 +INFO:root: +[Valid][20][UTT][ MSE = 0.2157 | LCC = 0.9526 | SRCC = 0.9511 ] +INFO:root:Average loss=0.04927424217760563 +INFO:root: +[Valid][21][UTT][ MSE = 0.2145 | LCC = 0.9524 | SRCC = 0.9510 ] +INFO:root:Average loss=0.045387397915124895 +INFO:root: +[Valid][22][UTT][ MSE = 0.2180 | LCC = 0.9533 | SRCC = 0.9519 ] +INFO:root:Average loss=0.042084883823990824 +INFO:root: +[Valid][23][UTT][ MSE = 0.2301 | LCC = 0.9526 | SRCC = 0.9513 ] +INFO:root:Average loss=0.042660769000649454 +INFO:root: +[Valid][24][UTT][ MSE = 0.2148 | LCC = 0.9519 | SRCC = 0.9509 ] +INFO:root:Average loss=0.039923586949706075 +INFO:root: +[Valid][25][UTT][ MSE = 0.2000 | LCC = 0.9513 | SRCC = 0.9501 ] +INFO:root:Average loss=0.03871237862110138 +INFO:root: +[Valid][26][UTT][ MSE = 0.2006 | LCC = 0.9534 | SRCC = 0.9522 ] +INFO:root:Average loss=0.03500842563807964 +INFO:root: +[Valid][27][UTT][ MSE = 0.1977 | LCC = 0.9531 | SRCC = 0.9519 ] +INFO:root:Average loss=0.037313402816653254 +INFO:root: +[Valid][28][UTT][ MSE = 0.2261 | LCC = 0.9517 | SRCC = 0.9509 ] +INFO:root:Average loss=0.03377894531190395 +INFO:root: +[Valid][29][UTT][ MSE = 0.1966 | LCC = 0.9529 | SRCC = 0.9514 ] +INFO:root:Average loss=0.033855651259422304 +INFO:root: +[Valid][30][UTT][ MSE = 0.2170 | LCC = 0.9550 | SRCC = 0.9540 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2244 | LCC = 0.9521 | SRCC = 0.9531 ] diff --git a/results_tencent/w2v2_large_asr_en/layer14/run3/config.gin b/results_tencent/w2v2_large_asr_en/layer14/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer14/run3/model.pt b/results_tencent/w2v2_large_asr_en/layer14/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..06cbea0b3a97c06f20be1e16ad642535d6a7a48d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c5f69eae754a09d868fbbf9240998922ea08adeef9f54f6e8099f1d28645a8bd +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer14/run3/model_best.pt b/results_tencent/w2v2_large_asr_en/layer14/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4088a709b8bb1b083212d5b5495275d9e7a7613a --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d459f3d26bbd30082c0bc9ff05052470699feb051069ec26704c305f1f514867 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer14/run3/train.log b/results_tencent/w2v2_large_asr_en/layer14/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b98707303f0f3f30a542a87f9983332a10f55f95 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5030220943689346 +INFO:root: +[Valid][0][UTT][ MSE = 0.3585 | LCC = 0.8921 | SRCC = 0.8929 ] +INFO:root:Average loss=0.2734427508115768 +INFO:root: +[Valid][1][UTT][ MSE = 0.2979 | LCC = 0.9163 | SRCC = 0.9149 ] +INFO:root:Average loss=0.21440149617195128 +INFO:root: +[Valid][2][UTT][ MSE = 0.2302 | LCC = 0.9299 | SRCC = 0.9283 ] +INFO:root:Average loss=0.18008034533262252 +INFO:root: +[Valid][3][UTT][ MSE = 0.2188 | LCC = 0.9363 | SRCC = 0.9345 ] +INFO:root:Average loss=0.15625116723775864 +INFO:root: +[Valid][4][UTT][ MSE = 0.2463 | LCC = 0.9412 | SRCC = 0.9400 ] +INFO:root:Average loss=0.1373051186800003 +INFO:root: +[Valid][5][UTT][ MSE = 0.2127 | LCC = 0.9459 | SRCC = 0.9456 ] +INFO:root:Average loss=0.12357754111289979 +INFO:root: +[Valid][6][UTT][ MSE = 0.2107 | LCC = 0.9484 | SRCC = 0.9476 ] +INFO:root:Average loss=0.11394237226247787 +INFO:root: +[Valid][7][UTT][ MSE = 0.2416 | LCC = 0.9477 | SRCC = 0.9475 ] +INFO:root:Average loss=0.10838579767942429 +INFO:root: +[Valid][8][UTT][ MSE = 0.2241 | LCC = 0.9487 | SRCC = 0.9488 ] +INFO:root:Average loss=0.09907291066646576 +INFO:root: +[Valid][9][UTT][ MSE = 0.2013 | LCC = 0.9506 | SRCC = 0.9500 ] +INFO:root:Average loss=0.0920728110074997 +INFO:root: +[Valid][10][UTT][ MSE = 0.2178 | LCC = 0.9509 | SRCC = 0.9503 ] +INFO:root:Average loss=0.0838188137114048 +INFO:root: +[Valid][11][UTT][ MSE = 0.2064 | LCC = 0.9514 | SRCC = 0.9513 ] +INFO:root:Average loss=0.07698396545648575 +INFO:root: +[Valid][12][UTT][ MSE = 0.2225 | LCC = 0.9529 | SRCC = 0.9524 ] +INFO:root:Average loss=0.07351504614949227 +INFO:root: +[Valid][13][UTT][ MSE = 0.2418 | LCC = 0.9514 | SRCC = 0.9504 ] +INFO:root:Average loss=0.07513643348217011 +INFO:root: +[Valid][14][UTT][ MSE = 0.2883 | LCC = 0.9510 | SRCC = 0.9499 ] +INFO:root:Average loss=0.06850581678748131 +INFO:root: +[Valid][15][UTT][ MSE = 0.1792 | LCC = 0.9525 | SRCC = 0.9514 ] +INFO:root:Average loss=0.06044243866205216 +INFO:root: +[Valid][16][UTT][ MSE = 0.2308 | LCC = 0.9523 | SRCC = 0.9510 ] +INFO:root:Average loss=0.0579308627396822 +INFO:root: +[Valid][17][UTT][ MSE = 0.1953 | LCC = 0.9497 | SRCC = 0.9493 ] +INFO:root:Average loss=0.05658239084482193 +INFO:root: +[Valid][18][UTT][ MSE = 0.2237 | LCC = 0.9525 | SRCC = 0.9509 ] +INFO:root:Average loss=0.05409075902402401 +INFO:root: +[Valid][19][UTT][ MSE = 0.1938 | LCC = 0.9534 | SRCC = 0.9518 ] +INFO:root:Average loss=0.04678757391870022 +INFO:root: +[Valid][20][UTT][ MSE = 0.1910 | LCC = 0.9523 | SRCC = 0.9510 ] +INFO:root:Average loss=0.045708898454904556 +INFO:root: +[Valid][21][UTT][ MSE = 0.2140 | LCC = 0.9528 | SRCC = 0.9519 ] +INFO:root:Average loss=0.04404704988002777 +INFO:root: +[Valid][22][UTT][ MSE = 0.2242 | LCC = 0.9530 | SRCC = 0.9524 ] +INFO:root:Average loss=0.04372848255932331 +INFO:root: +[Valid][23][UTT][ MSE = 0.1947 | LCC = 0.9531 | SRCC = 0.9531 ] +INFO:root:Average loss=0.042109160885214804 +INFO:root: +[Valid][24][UTT][ MSE = 0.2114 | LCC = 0.9539 | SRCC = 0.9531 ] +INFO:root:Average loss=0.0407519250959158 +INFO:root: +[Valid][25][UTT][ MSE = 0.1961 | LCC = 0.9526 | SRCC = 0.9520 ] +INFO:root:Average loss=0.03566263890266418 +INFO:root: +[Valid][26][UTT][ MSE = 0.2007 | LCC = 0.9550 | SRCC = 0.9540 ] +INFO:root:Average loss=0.03587048900127411 +INFO:root: +[Valid][27][UTT][ MSE = 0.1902 | LCC = 0.9532 | SRCC = 0.9525 ] +INFO:root:Average loss=0.037027692168951035 +INFO:root: +[Valid][28][UTT][ MSE = 0.2107 | LCC = 0.9537 | SRCC = 0.9527 ] +INFO:root:Average loss=0.035558348372578624 +INFO:root: +[Valid][29][UTT][ MSE = 0.2135 | LCC = 0.9540 | SRCC = 0.9534 ] +INFO:root:Average loss=0.03260030683875084 +INFO:root: +[Valid][30][UTT][ MSE = 0.2120 | LCC = 0.9532 | SRCC = 0.9521 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2032 | LCC = 0.9532 | SRCC = 0.9534 ] diff --git a/results_tencent/w2v2_large_asr_en/layer14/run4/config.gin b/results_tencent/w2v2_large_asr_en/layer14/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer14/run4/model.pt b/results_tencent/w2v2_large_asr_en/layer14/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8ccf2be3a53b984ee070b5261c7a11646c34b0c3 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2db38ccf6111b29b68df5e880c0c437b6727ab72ef652361cb8290a075e807aa +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer14/run4/model_best.pt b/results_tencent/w2v2_large_asr_en/layer14/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d994473abe1dee37fb22390091e68dc760432387 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6fec27a85b63e5f8d8f7e91208a2d154aa2a75b4159c4b9c08676b0f58c34268 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer14/run4/train.log b/results_tencent/w2v2_large_asr_en/layer14/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9b44b057a7e0fb89a873cb17f46cf81bebe1618a --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer14/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5194311436414719 +INFO:root: +[Valid][0][UTT][ MSE = 0.3585 | LCC = 0.9035 | SRCC = 0.9019 ] +INFO:root:Average loss=0.24773919320106505 +INFO:root: +[Valid][1][UTT][ MSE = 0.2884 | LCC = 0.9233 | SRCC = 0.9209 ] +INFO:root:Average loss=0.1876129841208458 +INFO:root: +[Valid][2][UTT][ MSE = 0.2409 | LCC = 0.9365 | SRCC = 0.9375 ] +INFO:root:Average loss=0.1698625009059906 +INFO:root: +[Valid][3][UTT][ MSE = 0.2763 | LCC = 0.9425 | SRCC = 0.9423 ] +INFO:root:Average loss=0.1491332585811615 +INFO:root: +[Valid][4][UTT][ MSE = 0.2527 | LCC = 0.9454 | SRCC = 0.9453 ] +INFO:root:Average loss=0.1308723664879799 +INFO:root: +[Valid][5][UTT][ MSE = 0.2228 | LCC = 0.9462 | SRCC = 0.9460 ] +INFO:root:Average loss=0.12514855486154555 +INFO:root: +[Valid][6][UTT][ MSE = 0.2578 | LCC = 0.9483 | SRCC = 0.9492 ] +INFO:root:Average loss=0.11618740290403366 +INFO:root: +[Valid][7][UTT][ MSE = 0.2331 | LCC = 0.9500 | SRCC = 0.9500 ] +INFO:root:Average loss=0.10654223608970642 +INFO:root: +[Valid][8][UTT][ MSE = 0.2304 | LCC = 0.9505 | SRCC = 0.9504 ] +INFO:root:Average loss=0.09331984621286392 +INFO:root: +[Valid][9][UTT][ MSE = 0.2014 | LCC = 0.9518 | SRCC = 0.9514 ] +INFO:root:Average loss=0.08936907330155372 +INFO:root: +[Valid][10][UTT][ MSE = 0.2187 | LCC = 0.9503 | SRCC = 0.9500 ] +INFO:root:Average loss=0.08158381953835488 +INFO:root: +[Valid][11][UTT][ MSE = 0.1795 | LCC = 0.9500 | SRCC = 0.9511 ] +INFO:root:Average loss=0.08113741999864578 +INFO:root: +[Valid][12][UTT][ MSE = 0.2030 | LCC = 0.9526 | SRCC = 0.9520 ] +INFO:root:Average loss=0.07112480947375298 +INFO:root: +[Valid][13][UTT][ MSE = 0.2308 | LCC = 0.9532 | SRCC = 0.9526 ] +INFO:root:Average loss=0.06777209952473641 +INFO:root: +[Valid][14][UTT][ MSE = 0.1730 | LCC = 0.9506 | SRCC = 0.9510 ] +INFO:root:Average loss=0.06744507113099098 +INFO:root: +[Valid][15][UTT][ MSE = 0.2170 | LCC = 0.9537 | SRCC = 0.9531 ] +INFO:root:Average loss=0.06003505051136017 +INFO:root: +[Valid][16][UTT][ MSE = 0.2230 | LCC = 0.9526 | SRCC = 0.9524 ] +INFO:root:Average loss=0.05706766611337662 +INFO:root: +[Valid][17][UTT][ MSE = 0.2038 | LCC = 0.9532 | SRCC = 0.9524 ] +INFO:root:Average loss=0.051421189904212954 +INFO:root: +[Valid][18][UTT][ MSE = 0.2132 | LCC = 0.9540 | SRCC = 0.9534 ] +INFO:root:Average loss=0.051378836557269096 +INFO:root: +[Valid][19][UTT][ MSE = 0.2248 | LCC = 0.9523 | SRCC = 0.9518 ] +INFO:root:Average loss=0.04762920397520065 +INFO:root: +[Valid][20][UTT][ MSE = 0.2405 | LCC = 0.9516 | SRCC = 0.9519 ] +INFO:root:Average loss=0.04501514711976051 +INFO:root: +[Valid][21][UTT][ MSE = 0.2180 | LCC = 0.9520 | SRCC = 0.9518 ] +INFO:root:Average loss=0.042463918805122375 +INFO:root: +[Valid][22][UTT][ MSE = 0.2312 | LCC = 0.9535 | SRCC = 0.9537 ] +INFO:root:Average loss=0.04002577959001064 +INFO:root: +[Valid][23][UTT][ MSE = 0.2221 | LCC = 0.9533 | SRCC = 0.9529 ] +INFO:root:Average loss=0.03882360310852528 +INFO:root: +[Valid][24][UTT][ MSE = 0.2024 | LCC = 0.9527 | SRCC = 0.9520 ] +INFO:root:Average loss=0.04058428762853145 +INFO:root: +[Valid][25][UTT][ MSE = 0.2179 | LCC = 0.9529 | SRCC = 0.9524 ] +INFO:root:Average loss=0.03908217154443264 +INFO:root: +[Valid][26][UTT][ MSE = 0.2068 | LCC = 0.9525 | SRCC = 0.9521 ] +INFO:root:Average loss=0.03701003135740757 +INFO:root: +[Valid][27][UTT][ MSE = 0.2046 | LCC = 0.9535 | SRCC = 0.9537 ] +INFO:root:Average loss=0.03364495411515236 +INFO:root: +[Valid][28][UTT][ MSE = 0.2013 | LCC = 0.9534 | SRCC = 0.9533 ] +INFO:root:Average loss=0.031445912450551984 +INFO:root: +[Valid][29][UTT][ MSE = 0.2189 | LCC = 0.9522 | SRCC = 0.9526 ] +INFO:root:Average loss=0.029662458896636962 +INFO:root: +[Valid][30][UTT][ MSE = 0.2180 | LCC = 0.9534 | SRCC = 0.9532 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2189 | LCC = 0.9525 | SRCC = 0.9533 ] diff --git a/results_tencent/w2v2_large_asr_en/layer16/run0/config.gin b/results_tencent/w2v2_large_asr_en/layer16/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer16/run0/model.pt b/results_tencent/w2v2_large_asr_en/layer16/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..315db26289cb31e12f53a66080612eae51a0c3c2 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bee9f78dffbd37db1b676c6c78571bccd66150d3a4fdad9491369667a218da84 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer16/run0/model_best.pt b/results_tencent/w2v2_large_asr_en/layer16/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b03ca3a16768302b56d14fdc464780d3add8de24 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b57ed1128474e184bbcd5a791da1b178bd0cecb0671a0436a1e4b3961f34354e +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer16/run0/train.log b/results_tencent/w2v2_large_asr_en/layer16/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..00fa37451bc6d8d5cd4b20331d2559219ebe1ed7 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4856996750831604 +INFO:root: +[Valid][0][UTT][ MSE = 0.3897 | LCC = 0.8944 | SRCC = 0.8932 ] +INFO:root:Average loss=0.2687903637886047 +INFO:root: +[Valid][1][UTT][ MSE = 0.2988 | LCC = 0.9127 | SRCC = 0.9120 ] +INFO:root:Average loss=0.24928828531503677 +INFO:root: +[Valid][2][UTT][ MSE = 0.3244 | LCC = 0.9247 | SRCC = 0.9235 ] +INFO:root:Average loss=0.1958979539871216 +INFO:root: +[Valid][3][UTT][ MSE = 0.3415 | LCC = 0.9324 | SRCC = 0.9306 ] +INFO:root:Average loss=0.1681976116299629 +INFO:root: +[Valid][4][UTT][ MSE = 0.3006 | LCC = 0.9360 | SRCC = 0.9347 ] +INFO:root:Average loss=0.14829363787174224 +INFO:root: +[Valid][5][UTT][ MSE = 0.3118 | LCC = 0.9301 | SRCC = 0.9275 ] +INFO:root:Average loss=0.14454983311891556 +INFO:root: +[Valid][6][UTT][ MSE = 0.2444 | LCC = 0.9431 | SRCC = 0.9425 ] +INFO:root:Average loss=0.13627413523197174 +INFO:root: +[Valid][7][UTT][ MSE = 0.1931 | LCC = 0.9457 | SRCC = 0.9450 ] +INFO:root:Average loss=0.1195155050754547 +INFO:root: +[Valid][8][UTT][ MSE = 0.2258 | LCC = 0.9460 | SRCC = 0.9451 ] +INFO:root:Average loss=0.10929182446002961 +INFO:root: +[Valid][9][UTT][ MSE = 0.2355 | LCC = 0.9457 | SRCC = 0.9451 ] +INFO:root:Average loss=0.10495617824792862 +INFO:root: +[Valid][10][UTT][ MSE = 0.2214 | LCC = 0.9478 | SRCC = 0.9475 ] +INFO:root:Average loss=0.0993091215789318 +INFO:root: +[Valid][11][UTT][ MSE = 0.2142 | LCC = 0.9488 | SRCC = 0.9475 ] +INFO:root:Average loss=0.09612851756811142 +INFO:root: +[Valid][12][UTT][ MSE = 0.2157 | LCC = 0.9469 | SRCC = 0.9467 ] +INFO:root:Average loss=0.08993917205929756 +INFO:root: +[Valid][13][UTT][ MSE = 0.2143 | LCC = 0.9498 | SRCC = 0.9489 ] +INFO:root:Average loss=0.08163232299685479 +INFO:root: +[Valid][14][UTT][ MSE = 0.2114 | LCC = 0.9492 | SRCC = 0.9483 ] +INFO:root:Average loss=0.07615359380841255 +INFO:root: +[Valid][15][UTT][ MSE = 0.1953 | LCC = 0.9501 | SRCC = 0.9493 ] +INFO:root:Average loss=0.07652289292216301 +INFO:root: +[Valid][16][UTT][ MSE = 0.2056 | LCC = 0.9515 | SRCC = 0.9504 ] +INFO:root:Average loss=0.0698668512403965 +INFO:root: +[Valid][17][UTT][ MSE = 0.1831 | LCC = 0.9502 | SRCC = 0.9498 ] +INFO:root:Average loss=0.06737595874071121 +INFO:root: +[Valid][18][UTT][ MSE = 0.2047 | LCC = 0.9500 | SRCC = 0.9490 ] +INFO:root:Average loss=0.06558417639136314 +INFO:root: +[Valid][19][UTT][ MSE = 0.1876 | LCC = 0.9513 | SRCC = 0.9503 ] +INFO:root:Average loss=0.06445445641875266 +INFO:root: +[Valid][20][UTT][ MSE = 0.1901 | LCC = 0.9484 | SRCC = 0.9483 ] +INFO:root:Average loss=0.05680218589305878 +INFO:root: +[Valid][21][UTT][ MSE = 0.1904 | LCC = 0.9520 | SRCC = 0.9509 ] +INFO:root:Average loss=0.05801394355297089 +INFO:root: +[Valid][22][UTT][ MSE = 0.2034 | LCC = 0.9524 | SRCC = 0.9516 ] +INFO:root:Average loss=0.05728647993505001 +INFO:root: +[Valid][23][UTT][ MSE = 0.2036 | LCC = 0.9514 | SRCC = 0.9511 ] +INFO:root:Average loss=0.053349253743886944 +INFO:root: +[Valid][24][UTT][ MSE = 0.2032 | LCC = 0.9508 | SRCC = 0.9501 ] +INFO:root:Average loss=0.04985411167144775 +INFO:root: +[Valid][25][UTT][ MSE = 0.2063 | LCC = 0.9495 | SRCC = 0.9486 ] +INFO:root:Average loss=0.051593284025788305 +INFO:root: +[Valid][26][UTT][ MSE = 0.1797 | LCC = 0.9530 | SRCC = 0.9521 ] +INFO:root:Average loss=0.04573911036550999 +INFO:root: +[Valid][27][UTT][ MSE = 0.1814 | LCC = 0.9525 | SRCC = 0.9515 ] +INFO:root:Average loss=0.044345170095562936 +INFO:root: +[Valid][28][UTT][ MSE = 0.2003 | LCC = 0.9520 | SRCC = 0.9507 ] +INFO:root:Average loss=0.04318983408808708 +INFO:root: +[Valid][29][UTT][ MSE = 0.2118 | LCC = 0.9517 | SRCC = 0.9504 ] +INFO:root:Average loss=0.04197598119080067 +INFO:root: +[Valid][30][UTT][ MSE = 0.2144 | LCC = 0.9523 | SRCC = 0.9512 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1834 | LCC = 0.9526 | SRCC = 0.9527 ] diff --git a/results_tencent/w2v2_large_asr_en/layer16/run1/config.gin b/results_tencent/w2v2_large_asr_en/layer16/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer16/run1/model.pt b/results_tencent/w2v2_large_asr_en/layer16/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7a6b416af0d8a2b3b2554cd6e3adb41f0c6349bd --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d484019ce419602196f0514cb51eff7c4c6348a40ecdc67b869b2ddb4d139f8 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer16/run1/model_best.pt b/results_tencent/w2v2_large_asr_en/layer16/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..cb431f24901872d6c6f7e850d8b112e90503491e --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fea3784c4c4e02d7a4d35d59a731f64dba7717a52e6d3639a18425f4ea38343c +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer16/run1/train.log b/results_tencent/w2v2_large_asr_en/layer16/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b29c5ee3ee4e1ea90b4d04912022550a2100a9d9 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5458868503570556 +INFO:root: +[Valid][0][UTT][ MSE = 0.4073 | LCC = 0.8888 | SRCC = 0.8876 ] +INFO:root:Average loss=0.2760826225280762 +INFO:root: +[Valid][1][UTT][ MSE = 0.3106 | LCC = 0.9091 | SRCC = 0.9093 ] +INFO:root:Average loss=0.2273536355495453 +INFO:root: +[Valid][2][UTT][ MSE = 0.2639 | LCC = 0.9256 | SRCC = 0.9234 ] +INFO:root:Average loss=0.18860364365577698 +INFO:root: +[Valid][3][UTT][ MSE = 0.2745 | LCC = 0.9344 | SRCC = 0.9337 ] +INFO:root:Average loss=0.17320758956670762 +INFO:root: +[Valid][4][UTT][ MSE = 0.2262 | LCC = 0.9389 | SRCC = 0.9375 ] +INFO:root:Average loss=0.15458383971452713 +INFO:root: +[Valid][5][UTT][ MSE = 0.2228 | LCC = 0.9411 | SRCC = 0.9410 ] +INFO:root:Average loss=0.1447045069336891 +INFO:root: +[Valid][6][UTT][ MSE = 0.2207 | LCC = 0.9433 | SRCC = 0.9436 ] +INFO:root:Average loss=0.12571005934476853 +INFO:root: +[Valid][7][UTT][ MSE = 0.2414 | LCC = 0.9465 | SRCC = 0.9458 ] +INFO:root:Average loss=0.11477209287881851 +INFO:root: +[Valid][8][UTT][ MSE = 0.2077 | LCC = 0.9456 | SRCC = 0.9458 ] +INFO:root:Average loss=0.11383606514334679 +INFO:root: +[Valid][9][UTT][ MSE = 0.2012 | LCC = 0.9481 | SRCC = 0.9475 ] +INFO:root:Average loss=0.10367852398753166 +INFO:root: +[Valid][10][UTT][ MSE = 0.2029 | LCC = 0.9458 | SRCC = 0.9446 ] +INFO:root:Average loss=0.09496787494421005 +INFO:root: +[Valid][11][UTT][ MSE = 0.2394 | LCC = 0.9473 | SRCC = 0.9473 ] +INFO:root:Average loss=0.09414773553609848 +INFO:root: +[Valid][12][UTT][ MSE = 0.1840 | LCC = 0.9492 | SRCC = 0.9480 ] +INFO:root:Average loss=0.09253448370099068 +INFO:root: +[Valid][13][UTT][ MSE = 0.2087 | LCC = 0.9478 | SRCC = 0.9478 ] +INFO:root:Average loss=0.08535980659723282 +INFO:root: +[Valid][14][UTT][ MSE = 0.1940 | LCC = 0.9481 | SRCC = 0.9475 ] +INFO:root:Average loss=0.08346982005238533 +INFO:root: +[Valid][15][UTT][ MSE = 0.1988 | LCC = 0.9501 | SRCC = 0.9490 ] +INFO:root:Average loss=0.0743149499297142 +INFO:root: +[Valid][16][UTT][ MSE = 0.2197 | LCC = 0.9498 | SRCC = 0.9490 ] +INFO:root:Average loss=0.075631289601326 +INFO:root: +[Valid][17][UTT][ MSE = 0.2107 | LCC = 0.9494 | SRCC = 0.9487 ] +INFO:root:Average loss=0.06675165581703187 +INFO:root: +[Valid][18][UTT][ MSE = 0.1997 | LCC = 0.9492 | SRCC = 0.9481 ] +INFO:root:Average loss=0.06697707077860832 +INFO:root: +[Valid][19][UTT][ MSE = 0.1896 | LCC = 0.9492 | SRCC = 0.9481 ] +INFO:root:Average loss=0.06380356812477112 +INFO:root: +[Valid][20][UTT][ MSE = 0.2116 | LCC = 0.9505 | SRCC = 0.9496 ] +INFO:root:Average loss=0.06067245882749558 +INFO:root: +[Valid][21][UTT][ MSE = 0.1934 | LCC = 0.9501 | SRCC = 0.9495 ] +INFO:root:Average loss=0.05718036398291588 +INFO:root: +[Valid][22][UTT][ MSE = 0.1979 | LCC = 0.9500 | SRCC = 0.9494 ] +INFO:root:Average loss=0.054801963567733765 +INFO:root: +[Valid][23][UTT][ MSE = 0.2106 | LCC = 0.9506 | SRCC = 0.9503 ] +INFO:root:Average loss=0.05251650530099869 +INFO:root: +[Valid][24][UTT][ MSE = 0.2136 | LCC = 0.9507 | SRCC = 0.9494 ] +INFO:root:Average loss=0.04990819938480854 +INFO:root: +[Valid][25][UTT][ MSE = 0.1741 | LCC = 0.9487 | SRCC = 0.9478 ] +INFO:root:Average loss=0.04907350182533264 +INFO:root: +[Valid][26][UTT][ MSE = 0.2120 | LCC = 0.9498 | SRCC = 0.9491 ] +INFO:root:Average loss=0.04741374255716801 +INFO:root: +[Valid][27][UTT][ MSE = 0.2051 | LCC = 0.9501 | SRCC = 0.9492 ] +INFO:root:Average loss=0.044617123305797576 +INFO:root: +[Valid][28][UTT][ MSE = 0.2145 | LCC = 0.9495 | SRCC = 0.9485 ] +INFO:root:Average loss=0.04169784553349018 +INFO:root: +[Valid][29][UTT][ MSE = 0.2068 | LCC = 0.9512 | SRCC = 0.9502 ] +INFO:root:Average loss=0.04045932227373123 +INFO:root: +[Valid][30][UTT][ MSE = 0.1938 | LCC = 0.9521 | SRCC = 0.9511 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2007 | LCC = 0.9506 | SRCC = 0.9522 ] diff --git a/results_tencent/w2v2_large_asr_en/layer16/run2/config.gin b/results_tencent/w2v2_large_asr_en/layer16/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer16/run2/model.pt b/results_tencent/w2v2_large_asr_en/layer16/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2fadcb47ace6426b7f4f12c2b866cb9a877bfbf6 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1cd8e96ea94f0fea17683ffba744d595fb90d92503fe116f0bbd0462881ee625 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer16/run2/model_best.pt b/results_tencent/w2v2_large_asr_en/layer16/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..051445eb903fd22f22e32669cbe0453b4cfd3a36 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b263889e466452437a6ba9bd3e9cf439f61caa1323832017e3562e58da9853d6 +size 3404998 diff --git a/results_tencent/w2v2_large_asr_en/layer16/run2/train.log b/results_tencent/w2v2_large_asr_en/layer16/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6b91995921b8874e0424a68a969c41743199f4a6 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5347247335910797 +INFO:root: +[Valid][0][UTT][ MSE = 0.4544 | LCC = 0.8874 | SRCC = 0.8883 ] +INFO:root:Average loss=0.2753719846010208 +INFO:root: +[Valid][1][UTT][ MSE = 0.3068 | LCC = 0.9129 | SRCC = 0.9103 ] +INFO:root:Average loss=0.22109946298599242 +INFO:root: +[Valid][2][UTT][ MSE = 0.2858 | LCC = 0.9236 | SRCC = 0.9223 ] +INFO:root:Average loss=0.19482133144140243 +INFO:root: +[Valid][3][UTT][ MSE = 0.2866 | LCC = 0.9281 | SRCC = 0.9292 ] +INFO:root:Average loss=0.17028854817152023 +INFO:root: +[Valid][4][UTT][ MSE = 0.2625 | LCC = 0.9371 | SRCC = 0.9364 ] +INFO:root:Average loss=0.1525483143925667 +INFO:root: +[Valid][5][UTT][ MSE = 0.2639 | LCC = 0.9374 | SRCC = 0.9372 ] +INFO:root:Average loss=0.14123824733495713 +INFO:root: +[Valid][6][UTT][ MSE = 0.2511 | LCC = 0.9398 | SRCC = 0.9380 ] +INFO:root:Average loss=0.12820467615127565 +INFO:root: +[Valid][7][UTT][ MSE = 0.2619 | LCC = 0.9438 | SRCC = 0.9423 ] +INFO:root:Average loss=0.1245275257229805 +INFO:root: +[Valid][8][UTT][ MSE = 0.2734 | LCC = 0.9458 | SRCC = 0.9451 ] +INFO:root:Average loss=0.1139382506608963 +INFO:root: +[Valid][9][UTT][ MSE = 0.2453 | LCC = 0.9430 | SRCC = 0.9415 ] +INFO:root:Average loss=0.11146407839655877 +INFO:root: +[Valid][10][UTT][ MSE = 0.1989 | LCC = 0.9466 | SRCC = 0.9451 ] +INFO:root:Average loss=0.10303527820110321 +INFO:root: +[Valid][11][UTT][ MSE = 0.3070 | LCC = 0.9438 | SRCC = 0.9442 ] +INFO:root:Average loss=0.09001839971542358 +INFO:root: +[Valid][12][UTT][ MSE = 0.2456 | LCC = 0.9477 | SRCC = 0.9463 ] +INFO:root:Average loss=0.08680304434895515 +INFO:root: +[Valid][13][UTT][ MSE = 0.2162 | LCC = 0.9477 | SRCC = 0.9462 ] +INFO:root:Average loss=0.08153292635083198 +INFO:root: +[Valid][14][UTT][ MSE = 0.2286 | LCC = 0.9474 | SRCC = 0.9464 ] +INFO:root:Average loss=0.07738163673877715 +INFO:root: +[Valid][15][UTT][ MSE = 0.2114 | LCC = 0.9481 | SRCC = 0.9468 ] +INFO:root:Average loss=0.07530498865246772 +INFO:root: +[Valid][16][UTT][ MSE = 0.2520 | LCC = 0.9496 | SRCC = 0.9484 ] +INFO:root:Average loss=0.06915085238218308 +INFO:root: +[Valid][17][UTT][ MSE = 0.2323 | LCC = 0.9494 | SRCC = 0.9491 ] +INFO:root:Average loss=0.06463987547159195 +INFO:root: +[Valid][18][UTT][ MSE = 0.2144 | LCC = 0.9494 | SRCC = 0.9487 ] +INFO:root:Average loss=0.06146396514773369 +INFO:root: +[Valid][19][UTT][ MSE = 0.2317 | LCC = 0.9451 | SRCC = 0.9434 ] +INFO:root:Average loss=0.05877405516803265 +INFO:root: +[Valid][20][UTT][ MSE = 0.1957 | LCC = 0.9506 | SRCC = 0.9501 ] +INFO:root:Average loss=0.05805317428708077 +INFO:root: +[Valid][21][UTT][ MSE = 0.2193 | LCC = 0.9507 | SRCC = 0.9497 ] +INFO:root:Average loss=0.056335014045238496 +INFO:root: +[Valid][22][UTT][ MSE = 0.2278 | LCC = 0.9482 | SRCC = 0.9485 ] +INFO:root:Average loss=0.05232980588078499 +INFO:root: +[Valid][23][UTT][ MSE = 0.2318 | LCC = 0.9503 | SRCC = 0.9490 ] +INFO:root:Average loss=0.04991289395093918 +INFO:root: +[Valid][24][UTT][ MSE = 0.2277 | LCC = 0.9461 | SRCC = 0.9466 ] +INFO:root:Average loss=0.050190911859273914 +INFO:root: +[Valid][25][UTT][ MSE = 0.2370 | LCC = 0.9495 | SRCC = 0.9483 ] +INFO:root:Average loss=0.04577631182968617 +INFO:root: +[Valid][26][UTT][ MSE = 0.2377 | LCC = 0.9491 | SRCC = 0.9484 ] +INFO:root:Average loss=0.043554035112261774 +INFO:root: +[Valid][27][UTT][ MSE = 0.2384 | LCC = 0.9502 | SRCC = 0.9491 ] +INFO:root:Average loss=0.04279230819642544 +INFO:root: +[Valid][28][UTT][ MSE = 0.1952 | LCC = 0.9510 | SRCC = 0.9499 ] +INFO:root:Average loss=0.04190757989883423 +INFO:root: +[Valid][29][UTT][ MSE = 0.2112 | LCC = 0.9501 | SRCC = 0.9494 ] +INFO:root:Average loss=0.039433083042502405 +INFO:root: +[Valid][30][UTT][ MSE = 0.2094 | LCC = 0.9501 | SRCC = 0.9485 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1902 | LCC = 0.9547 | SRCC = 0.9549 ] diff --git a/results_tencent/w2v2_large_asr_en/layer16/run3/config.gin b/results_tencent/w2v2_large_asr_en/layer16/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer16/run3/model.pt b/results_tencent/w2v2_large_asr_en/layer16/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d3a2d37db4019f117ff69b0076944fb73b739e77 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d64023b5f6848e8f8a8cd817d2a6192ab509a8eb94d6e6e7db2a66ed9a154764 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer16/run3/model_best.pt b/results_tencent/w2v2_large_asr_en/layer16/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9491682cf52324de8e786af397d638c7d989f6d6 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:913be8f4bad1f63f1d608e2abc833b68010eb34c21921bf2de921c59a950c933 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer16/run3/train.log b/results_tencent/w2v2_large_asr_en/layer16/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6ca48f2885ab26b1cab9f507781683f01408b38f --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5707780562639236 +INFO:root: +[Valid][0][UTT][ MSE = 0.4153 | LCC = 0.8804 | SRCC = 0.8840 ] +INFO:root:Average loss=0.27540026092529296 +INFO:root: +[Valid][1][UTT][ MSE = 0.3016 | LCC = 0.9115 | SRCC = 0.9097 ] +INFO:root:Average loss=0.23207801258563995 +INFO:root: +[Valid][2][UTT][ MSE = 0.3124 | LCC = 0.9223 | SRCC = 0.9224 ] +INFO:root:Average loss=0.18727122783660888 +INFO:root: +[Valid][3][UTT][ MSE = 0.3323 | LCC = 0.9278 | SRCC = 0.9269 ] +INFO:root:Average loss=0.17848377716541292 +INFO:root: +[Valid][4][UTT][ MSE = 0.2709 | LCC = 0.9380 | SRCC = 0.9374 ] +INFO:root:Average loss=0.1523270342350006 +INFO:root: +[Valid][5][UTT][ MSE = 0.2529 | LCC = 0.9372 | SRCC = 0.9369 ] +INFO:root:Average loss=0.13979186964035034 +INFO:root: +[Valid][6][UTT][ MSE = 0.2357 | LCC = 0.9445 | SRCC = 0.9439 ] +INFO:root:Average loss=0.13034211897850037 +INFO:root: +[Valid][7][UTT][ MSE = 0.2170 | LCC = 0.9447 | SRCC = 0.9451 ] +INFO:root:Average loss=0.12120924234390258 +INFO:root: +[Valid][8][UTT][ MSE = 0.2169 | LCC = 0.9453 | SRCC = 0.9451 ] +INFO:root:Average loss=0.1143731240928173 +INFO:root: +[Valid][9][UTT][ MSE = 0.2220 | LCC = 0.9445 | SRCC = 0.9438 ] +INFO:root:Average loss=0.10393361258506775 +INFO:root: +[Valid][10][UTT][ MSE = 0.2577 | LCC = 0.9465 | SRCC = 0.9465 ] +INFO:root:Average loss=0.09845035737752915 +INFO:root: +[Valid][11][UTT][ MSE = 0.2391 | LCC = 0.9486 | SRCC = 0.9488 ] +INFO:root:Average loss=0.09416294637322425 +INFO:root: +[Valid][12][UTT][ MSE = 0.2052 | LCC = 0.9480 | SRCC = 0.9484 ] +INFO:root:Average loss=0.0808141233921051 +INFO:root: +[Valid][13][UTT][ MSE = 0.2190 | LCC = 0.9503 | SRCC = 0.9496 ] +INFO:root:Average loss=0.08411891347169877 +INFO:root: +[Valid][14][UTT][ MSE = 0.2264 | LCC = 0.9522 | SRCC = 0.9516 ] +INFO:root:Average loss=0.07658110591769218 +INFO:root: +[Valid][15][UTT][ MSE = 0.1945 | LCC = 0.9519 | SRCC = 0.9513 ] +INFO:root:Average loss=0.07183029028773308 +INFO:root: +[Valid][16][UTT][ MSE = 0.2107 | LCC = 0.9525 | SRCC = 0.9516 ] +INFO:root:Average loss=0.06861397105455398 +INFO:root: +[Valid][17][UTT][ MSE = 0.2161 | LCC = 0.9509 | SRCC = 0.9501 ] +INFO:root:Average loss=0.06570045959949493 +INFO:root: +[Valid][18][UTT][ MSE = 0.2109 | LCC = 0.9529 | SRCC = 0.9523 ] +INFO:root:Average loss=0.06008033022284508 +INFO:root: +[Valid][19][UTT][ MSE = 0.2064 | LCC = 0.9527 | SRCC = 0.9514 ] +INFO:root:Average loss=0.06137162485718727 +INFO:root: +[Valid][20][UTT][ MSE = 0.2515 | LCC = 0.9508 | SRCC = 0.9506 ] +INFO:root:Average loss=0.056522438168525695 +INFO:root: +[Valid][21][UTT][ MSE = 0.2158 | LCC = 0.9517 | SRCC = 0.9510 ] +INFO:root:Average loss=0.05350951996445656 +INFO:root: +[Valid][22][UTT][ MSE = 0.2290 | LCC = 0.9522 | SRCC = 0.9519 ] +INFO:root:Average loss=0.053528031766414644 +INFO:root: +[Valid][23][UTT][ MSE = 0.2203 | LCC = 0.9503 | SRCC = 0.9487 ] +INFO:root:Average loss=0.0488593961596489 +INFO:root: +[Valid][24][UTT][ MSE = 0.2088 | LCC = 0.9527 | SRCC = 0.9521 ] +INFO:root:Average loss=0.048269747406244276 +INFO:root: +[Valid][25][UTT][ MSE = 0.2123 | LCC = 0.9511 | SRCC = 0.9503 ] +INFO:root:Average loss=0.045784059062600135 +INFO:root: +[Valid][26][UTT][ MSE = 0.2087 | LCC = 0.9516 | SRCC = 0.9505 ] +INFO:root:Average loss=0.04461502702534199 +INFO:root: +[Valid][27][UTT][ MSE = 0.2318 | LCC = 0.9512 | SRCC = 0.9514 ] +INFO:root:Average loss=0.04358299878239632 +INFO:root: +[Valid][28][UTT][ MSE = 0.1917 | LCC = 0.9533 | SRCC = 0.9522 ] +INFO:root:Average loss=0.041087168648839 +INFO:root: +[Valid][29][UTT][ MSE = 0.2085 | LCC = 0.9516 | SRCC = 0.9509 ] +INFO:root:Average loss=0.03818717433512211 +INFO:root: +[Valid][30][UTT][ MSE = 0.2031 | LCC = 0.9539 | SRCC = 0.9526 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2108 | LCC = 0.9512 | SRCC = 0.9509 ] diff --git a/results_tencent/w2v2_large_asr_en/layer16/run4/config.gin b/results_tencent/w2v2_large_asr_en/layer16/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer16/run4/model_best.pt b/results_tencent/w2v2_large_asr_en/layer16/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6bddecb1bb47f4b8d951eb94e304d918ed1679be --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5609b47644321ebf7f59e7d517e76fd5e1d8378d00126b385af8024a2ac60570 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer16/run4/train.log b/results_tencent/w2v2_large_asr_en/layer16/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..52515212d7912e34c73a3c9703eb7d63c5f06698 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer16/run4/train.log @@ -0,0 +1,22 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5266109042167664 +INFO:root: +[Valid][0][UTT][ MSE = 0.4241 | LCC = 0.8817 | SRCC = 0.8864 ] +INFO:root:Average loss=0.2642705200314522 +INFO:root: +[Valid][1][UTT][ MSE = 0.3274 | LCC = 0.9117 | SRCC = 0.9116 ] +INFO:root:Average loss=0.2289748174548149 +INFO:root: +[Valid][2][UTT][ MSE = 0.3550 | LCC = 0.9194 | SRCC = 0.9215 ] +INFO:root:Average loss=0.19244337528944017 +INFO:root: +[Valid][3][UTT][ MSE = 0.3044 | LCC = 0.9306 | SRCC = 0.9324 ] +INFO:root:Average loss=0.16494584101438522 +INFO:root: +[Valid][4][UTT][ MSE = 0.2922 | LCC = 0.9367 | SRCC = 0.9362 ] +INFO:root:Average loss=0.14753428238630295 +INFO:root: +[Valid][5][UTT][ MSE = 0.2887 | LCC = 0.9412 | SRCC = 0.9408 ] diff --git a/results_tencent/w2v2_large_asr_en/layer18/run0/config.gin b/results_tencent/w2v2_large_asr_en/layer18/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer18/run0/model.pt b/results_tencent/w2v2_large_asr_en/layer18/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d61e54636f58c76f804bb7d97a0d9f723746877c --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7204cb401f72e4285b670d0009948a082d188e41544293731ad0abc471b2f102 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer18/run0/model_best.pt b/results_tencent/w2v2_large_asr_en/layer18/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..59761dfb6a855251c0838c4121b81cb6415a5556 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b93a0758b6242a5ab7d275491b18f5efb52b90fe50c151c312d6b3f4272e838c +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer18/run0/train.log b/results_tencent/w2v2_large_asr_en/layer18/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..45c0379fe6a8ca4c26056553b1f13514258388ec --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5191819825172425 +INFO:root: +[Valid][0][UTT][ MSE = 0.4255 | LCC = 0.8794 | SRCC = 0.8756 ] +INFO:root:Average loss=0.2814740070104599 +INFO:root: +[Valid][1][UTT][ MSE = 0.3735 | LCC = 0.9021 | SRCC = 0.9040 ] +INFO:root:Average loss=0.25265787935256956 +INFO:root: +[Valid][2][UTT][ MSE = 0.3053 | LCC = 0.9169 | SRCC = 0.9148 ] +INFO:root:Average loss=0.22710462087392808 +INFO:root: +[Valid][3][UTT][ MSE = 0.2956 | LCC = 0.9219 | SRCC = 0.9205 ] +INFO:root:Average loss=0.22209204894304274 +INFO:root: +[Valid][4][UTT][ MSE = 0.2514 | LCC = 0.9256 | SRCC = 0.9254 ] +INFO:root:Average loss=0.19580471497774124 +INFO:root: +[Valid][5][UTT][ MSE = 0.2926 | LCC = 0.9303 | SRCC = 0.9303 ] +INFO:root:Average loss=0.1726304923892021 +INFO:root: +[Valid][6][UTT][ MSE = 0.2388 | LCC = 0.9371 | SRCC = 0.9355 ] +INFO:root:Average loss=0.15439590495824815 +INFO:root: +[Valid][7][UTT][ MSE = 0.2676 | LCC = 0.9365 | SRCC = 0.9352 ] +INFO:root:Average loss=0.15143473768234253 +INFO:root: +[Valid][8][UTT][ MSE = 0.2716 | LCC = 0.9407 | SRCC = 0.9396 ] +INFO:root:Average loss=0.14548587226867676 +INFO:root: +[Valid][9][UTT][ MSE = 0.2488 | LCC = 0.9416 | SRCC = 0.9418 ] +INFO:root:Average loss=0.12815136831998825 +INFO:root: +[Valid][10][UTT][ MSE = 0.2343 | LCC = 0.9414 | SRCC = 0.9408 ] +INFO:root:Average loss=0.1274440271258354 +INFO:root: +[Valid][11][UTT][ MSE = 0.2559 | LCC = 0.9442 | SRCC = 0.9441 ] +INFO:root:Average loss=0.12289266926050187 +INFO:root: +[Valid][12][UTT][ MSE = 0.2335 | LCC = 0.9450 | SRCC = 0.9437 ] +INFO:root:Average loss=0.1151851966381073 +INFO:root: +[Valid][13][UTT][ MSE = 0.2167 | LCC = 0.9427 | SRCC = 0.9415 ] +INFO:root:Average loss=0.10751168459653854 +INFO:root: +[Valid][14][UTT][ MSE = 0.2285 | LCC = 0.9449 | SRCC = 0.9443 ] +INFO:root:Average loss=0.10217434924840928 +INFO:root: +[Valid][15][UTT][ MSE = 0.2190 | LCC = 0.9442 | SRCC = 0.9436 ] +INFO:root:Average loss=0.10290952506661415 +INFO:root: +[Valid][16][UTT][ MSE = 0.1906 | LCC = 0.9461 | SRCC = 0.9443 ] +INFO:root:Average loss=0.09182910132408142 +INFO:root: +[Valid][17][UTT][ MSE = 0.2377 | LCC = 0.9444 | SRCC = 0.9426 ] +INFO:root:Average loss=0.08860397362709045 +INFO:root: +[Valid][18][UTT][ MSE = 0.2035 | LCC = 0.9464 | SRCC = 0.9445 ] +INFO:root:Average loss=0.08524392607808114 +INFO:root: +[Valid][19][UTT][ MSE = 0.2664 | LCC = 0.9398 | SRCC = 0.9398 ] +INFO:root:Average loss=0.08280681645870208 +INFO:root: +[Valid][20][UTT][ MSE = 0.2168 | LCC = 0.9459 | SRCC = 0.9451 ] +INFO:root:Average loss=0.08550848859548568 +INFO:root: +[Valid][21][UTT][ MSE = 0.2136 | LCC = 0.9465 | SRCC = 0.9455 ] +INFO:root:Average loss=0.07445637160539627 +INFO:root: +[Valid][22][UTT][ MSE = 0.2115 | LCC = 0.9465 | SRCC = 0.9449 ] +INFO:root:Average loss=0.0724333301782608 +INFO:root: +[Valid][23][UTT][ MSE = 0.2063 | LCC = 0.9475 | SRCC = 0.9454 ] +INFO:root:Average loss=0.07047219654917716 +INFO:root: +[Valid][24][UTT][ MSE = 0.2162 | LCC = 0.9476 | SRCC = 0.9462 ] +INFO:root:Average loss=0.06995917093753815 +INFO:root: +[Valid][25][UTT][ MSE = 0.2229 | LCC = 0.9492 | SRCC = 0.9477 ] +INFO:root:Average loss=0.06618736419081687 +INFO:root: +[Valid][26][UTT][ MSE = 0.2057 | LCC = 0.9479 | SRCC = 0.9461 ] +INFO:root:Average loss=0.06381599044799804 +INFO:root: +[Valid][27][UTT][ MSE = 0.2147 | LCC = 0.9427 | SRCC = 0.9415 ] +INFO:root:Average loss=0.06893604889512062 +INFO:root: +[Valid][28][UTT][ MSE = 0.2027 | LCC = 0.9496 | SRCC = 0.9479 ] +INFO:root:Average loss=0.06064879259467125 +INFO:root: +[Valid][29][UTT][ MSE = 0.2094 | LCC = 0.9454 | SRCC = 0.9455 ] +INFO:root:Average loss=0.055672909051179886 +INFO:root: +[Valid][30][UTT][ MSE = 0.2238 | LCC = 0.9495 | SRCC = 0.9482 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2088 | LCC = 0.9468 | SRCC = 0.9466 ] diff --git a/results_tencent/w2v2_large_asr_en/layer18/run1/config.gin b/results_tencent/w2v2_large_asr_en/layer18/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer18/run1/model.pt b/results_tencent/w2v2_large_asr_en/layer18/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0f8a183515cba41b5eddd398fc2fb61038648243 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d8442424afa129d4edd38b8286d92f9dabed14787cd8342b41c9b937c5a818be +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer18/run1/model_best.pt b/results_tencent/w2v2_large_asr_en/layer18/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..566699ebb3e2ae22bcc267e875475268cb04a9ee --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:910965450c269583bb2195ffc8001ef35cfedfdcd0107dd4e57d6e4ce5d169a6 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer18/run1/train.log b/results_tencent/w2v2_large_asr_en/layer18/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f8312e77b2a133a431c9b76ec44c14e86dc43bd5 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5081580939292908 +INFO:root: +[Valid][0][UTT][ MSE = 0.3853 | LCC = 0.8747 | SRCC = 0.8781 ] +INFO:root:Average loss=0.2937455267906189 +INFO:root: +[Valid][1][UTT][ MSE = 0.3708 | LCC = 0.9054 | SRCC = 0.9039 ] +INFO:root:Average loss=0.2515104023218155 +INFO:root: +[Valid][2][UTT][ MSE = 0.3805 | LCC = 0.9056 | SRCC = 0.9103 ] +INFO:root:Average loss=0.22284016132354736 +INFO:root: +[Valid][3][UTT][ MSE = 0.3297 | LCC = 0.9190 | SRCC = 0.9203 ] +INFO:root:Average loss=0.21586076068878174 +INFO:root: +[Valid][4][UTT][ MSE = 0.3601 | LCC = 0.9208 | SRCC = 0.9210 ] +INFO:root:Average loss=0.19687352967262267 +INFO:root: +[Valid][5][UTT][ MSE = 0.2625 | LCC = 0.9282 | SRCC = 0.9263 ] +INFO:root:Average loss=0.18185841822624207 +INFO:root: +[Valid][6][UTT][ MSE = 0.2723 | LCC = 0.9342 | SRCC = 0.9331 ] +INFO:root:Average loss=0.15717426216602326 +INFO:root: +[Valid][7][UTT][ MSE = 0.2541 | LCC = 0.9374 | SRCC = 0.9355 ] +INFO:root:Average loss=0.15049373912811279 +INFO:root: +[Valid][8][UTT][ MSE = 0.2617 | LCC = 0.9408 | SRCC = 0.9396 ] +INFO:root:Average loss=0.1464448833465576 +INFO:root: +[Valid][9][UTT][ MSE = 0.2731 | LCC = 0.9404 | SRCC = 0.9396 ] +INFO:root:Average loss=0.14671941685676576 +INFO:root: +[Valid][10][UTT][ MSE = 0.2379 | LCC = 0.9415 | SRCC = 0.9414 ] +INFO:root:Average loss=0.13935771191120147 +INFO:root: +[Valid][11][UTT][ MSE = 0.2641 | LCC = 0.9442 | SRCC = 0.9432 ] +INFO:root:Average loss=0.1254286516904831 +INFO:root: +[Valid][12][UTT][ MSE = 0.2309 | LCC = 0.9426 | SRCC = 0.9418 ] +INFO:root:Average loss=0.12126022878289222 +INFO:root: +[Valid][13][UTT][ MSE = 0.2440 | LCC = 0.9420 | SRCC = 0.9400 ] +INFO:root:Average loss=0.1152936897277832 +INFO:root: +[Valid][14][UTT][ MSE = 0.2840 | LCC = 0.9417 | SRCC = 0.9413 ] +INFO:root:Average loss=0.11443478974699973 +INFO:root: +[Valid][15][UTT][ MSE = 0.2411 | LCC = 0.9441 | SRCC = 0.9423 ] +INFO:root:Average loss=0.10304613476991653 +INFO:root: +[Valid][16][UTT][ MSE = 0.2248 | LCC = 0.9466 | SRCC = 0.9451 ] +INFO:root:Average loss=0.10165491300821304 +INFO:root: +[Valid][17][UTT][ MSE = 0.2591 | LCC = 0.9458 | SRCC = 0.9445 ] +INFO:root:Average loss=0.10094668155908584 +INFO:root: +[Valid][18][UTT][ MSE = 0.2345 | LCC = 0.9475 | SRCC = 0.9460 ] +INFO:root:Average loss=0.08804670834541321 +INFO:root: +[Valid][19][UTT][ MSE = 0.2321 | LCC = 0.9478 | SRCC = 0.9466 ] +INFO:root:Average loss=0.08359804236888886 +INFO:root: +[Valid][20][UTT][ MSE = 0.2269 | LCC = 0.9477 | SRCC = 0.9464 ] +INFO:root:Average loss=0.08243797951936722 +INFO:root: +[Valid][21][UTT][ MSE = 0.2140 | LCC = 0.9478 | SRCC = 0.9465 ] +INFO:root:Average loss=0.08120449927449226 +INFO:root: +[Valid][22][UTT][ MSE = 0.2608 | LCC = 0.9478 | SRCC = 0.9476 ] +INFO:root:Average loss=0.0771249811053276 +INFO:root: +[Valid][23][UTT][ MSE = 0.2048 | LCC = 0.9474 | SRCC = 0.9459 ] +INFO:root:Average loss=0.07294185990095138 +INFO:root: +[Valid][24][UTT][ MSE = 0.2022 | LCC = 0.9453 | SRCC = 0.9449 ] +INFO:root:Average loss=0.06867296913266183 +INFO:root: +[Valid][25][UTT][ MSE = 0.2378 | LCC = 0.9469 | SRCC = 0.9459 ] +INFO:root:Average loss=0.06729265555739403 +INFO:root: +[Valid][26][UTT][ MSE = 0.1985 | LCC = 0.9469 | SRCC = 0.9458 ] +INFO:root:Average loss=0.0637848660349846 +INFO:root: +[Valid][27][UTT][ MSE = 0.2084 | LCC = 0.9496 | SRCC = 0.9485 ] +INFO:root:Average loss=0.06761202466487884 +INFO:root: +[Valid][28][UTT][ MSE = 0.2038 | LCC = 0.9460 | SRCC = 0.9452 ] +INFO:root:Average loss=0.06356001135706901 +INFO:root: +[Valid][29][UTT][ MSE = 0.2447 | LCC = 0.9499 | SRCC = 0.9488 ] +INFO:root:Average loss=0.06443742766976357 +INFO:root: +[Valid][30][UTT][ MSE = 0.2124 | LCC = 0.9470 | SRCC = 0.9461 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2552 | LCC = 0.9477 | SRCC = 0.9475 ] diff --git a/results_tencent/w2v2_large_asr_en/layer18/run2/config.gin b/results_tencent/w2v2_large_asr_en/layer18/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer18/run2/model.pt b/results_tencent/w2v2_large_asr_en/layer18/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ced7180755aecba36871e30f05c1f923f5af1dcc --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39f9214a1129c657f365575280ee9b779f7d8aa12c09903651c260f82debffc9 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer18/run2/model_best.pt b/results_tencent/w2v2_large_asr_en/layer18/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..25732d34396601b46d1b68a5f9f821936d42404e --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:965fb474ccde20140fc27f8ca134717573eaba2527a9b652a23a6367d36cb552 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer18/run2/train.log b/results_tencent/w2v2_large_asr_en/layer18/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..7398a452d793784e42d44604ed7dccba88868424 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.646148072719574 +INFO:root: +[Valid][0][UTT][ MSE = 0.3948 | LCC = 0.8700 | SRCC = 0.8739 ] +INFO:root:Average loss=0.307578421831131 +INFO:root: +[Valid][1][UTT][ MSE = 0.4022 | LCC = 0.8914 | SRCC = 0.8948 ] +INFO:root:Average loss=0.26367814481258395 +INFO:root: +[Valid][2][UTT][ MSE = 0.2886 | LCC = 0.9144 | SRCC = 0.9130 ] +INFO:root:Average loss=0.2378634649515152 +INFO:root: +[Valid][3][UTT][ MSE = 0.3767 | LCC = 0.9222 | SRCC = 0.9206 ] +INFO:root:Average loss=0.19844043457508087 +INFO:root: +[Valid][4][UTT][ MSE = 0.2993 | LCC = 0.9212 | SRCC = 0.9193 ] +INFO:root:Average loss=0.1981033689379692 +INFO:root: +[Valid][5][UTT][ MSE = 0.2810 | LCC = 0.9306 | SRCC = 0.9289 ] +INFO:root:Average loss=0.17900265091657638 +INFO:root: +[Valid][6][UTT][ MSE = 0.2533 | LCC = 0.9329 | SRCC = 0.9324 ] +INFO:root:Average loss=0.16254315000772476 +INFO:root: +[Valid][7][UTT][ MSE = 0.2450 | LCC = 0.9341 | SRCC = 0.9341 ] +INFO:root:Average loss=0.1523592206835747 +INFO:root: +[Valid][8][UTT][ MSE = 0.2604 | LCC = 0.9399 | SRCC = 0.9395 ] +INFO:root:Average loss=0.13664497703313827 +INFO:root: +[Valid][9][UTT][ MSE = 0.2582 | LCC = 0.9401 | SRCC = 0.9391 ] +INFO:root:Average loss=0.12877360647916794 +INFO:root: +[Valid][10][UTT][ MSE = 0.2658 | LCC = 0.9417 | SRCC = 0.9415 ] +INFO:root:Average loss=0.12502879625558852 +INFO:root: +[Valid][11][UTT][ MSE = 0.3057 | LCC = 0.9368 | SRCC = 0.9369 ] +INFO:root:Average loss=0.11848401811718941 +INFO:root: +[Valid][12][UTT][ MSE = 0.2504 | LCC = 0.9438 | SRCC = 0.9431 ] +INFO:root:Average loss=0.11262818413972854 +INFO:root: +[Valid][13][UTT][ MSE = 0.2512 | LCC = 0.9414 | SRCC = 0.9417 ] +INFO:root:Average loss=0.10764080381393433 +INFO:root: +[Valid][14][UTT][ MSE = 0.2419 | LCC = 0.9455 | SRCC = 0.9445 ] +INFO:root:Average loss=0.09808597365021705 +INFO:root: +[Valid][15][UTT][ MSE = 0.2315 | LCC = 0.9448 | SRCC = 0.9436 ] +INFO:root:Average loss=0.09829512909054756 +INFO:root: +[Valid][16][UTT][ MSE = 0.2416 | LCC = 0.9468 | SRCC = 0.9458 ] +INFO:root:Average loss=0.09237351730465888 +INFO:root: +[Valid][17][UTT][ MSE = 0.2750 | LCC = 0.9435 | SRCC = 0.9434 ] +INFO:root:Average loss=0.08774328246712684 +INFO:root: +[Valid][18][UTT][ MSE = 0.2317 | LCC = 0.9474 | SRCC = 0.9470 ] +INFO:root:Average loss=0.08789881801605225 +INFO:root: +[Valid][19][UTT][ MSE = 0.2216 | LCC = 0.9480 | SRCC = 0.9474 ] +INFO:root:Average loss=0.08136333683133125 +INFO:root: +[Valid][20][UTT][ MSE = 0.2343 | LCC = 0.9455 | SRCC = 0.9452 ] +INFO:root:Average loss=0.07996691846847534 +INFO:root: +[Valid][21][UTT][ MSE = 0.2463 | LCC = 0.9485 | SRCC = 0.9479 ] +INFO:root:Average loss=0.07310264709591865 +INFO:root: +[Valid][22][UTT][ MSE = 0.2112 | LCC = 0.9474 | SRCC = 0.9458 ] +INFO:root:Average loss=0.07425879642367363 +INFO:root: +[Valid][23][UTT][ MSE = 0.2546 | LCC = 0.9468 | SRCC = 0.9466 ] +INFO:root:Average loss=0.06643794727325439 +INFO:root: +[Valid][24][UTT][ MSE = 0.2017 | LCC = 0.9486 | SRCC = 0.9476 ] +INFO:root:Average loss=0.06780462202429771 +INFO:root: +[Valid][25][UTT][ MSE = 0.2338 | LCC = 0.9494 | SRCC = 0.9483 ] +INFO:root:Average loss=0.06650401580333709 +INFO:root: +[Valid][26][UTT][ MSE = 0.1868 | LCC = 0.9490 | SRCC = 0.9472 ] +INFO:root:Average loss=0.06535463839769363 +INFO:root: +[Valid][27][UTT][ MSE = 0.2226 | LCC = 0.9483 | SRCC = 0.9469 ] +INFO:root:Average loss=0.05992889207601547 +INFO:root: +[Valid][28][UTT][ MSE = 0.2389 | LCC = 0.9486 | SRCC = 0.9483 ] +INFO:root:Average loss=0.05986798514425755 +INFO:root: +[Valid][29][UTT][ MSE = 0.1965 | LCC = 0.9474 | SRCC = 0.9470 ] +INFO:root:Average loss=0.059553347408771516 +INFO:root: +[Valid][30][UTT][ MSE = 0.2428 | LCC = 0.9481 | SRCC = 0.9474 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2392 | LCC = 0.9489 | SRCC = 0.9492 ] diff --git a/results_tencent/w2v2_large_asr_en/layer18/run3/config.gin b/results_tencent/w2v2_large_asr_en/layer18/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer18/run3/model.pt b/results_tencent/w2v2_large_asr_en/layer18/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..482c267e9ad0c062d8fae32e5848a3ad28650857 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b73059beddd5049d143a2fea838ed0ab905434b6fee3c4e24340a0298ec3b29d +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer18/run3/model_best.pt b/results_tencent/w2v2_large_asr_en/layer18/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..dcedf6cbade514d5c4d0bc13952861c62cc54872 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df0c23f89cbb281b0eb2292b34aab4a20d9a8f7a8341d7e557f7b79e1fd7419d +size 3404998 diff --git a/results_tencent/w2v2_large_asr_en/layer18/run3/train.log b/results_tencent/w2v2_large_asr_en/layer18/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f7220f3e1a32e49e11402408d58cfb249b635430 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.542840999007225 +INFO:root: +[Valid][0][UTT][ MSE = 0.4522 | LCC = 0.8686 | SRCC = 0.8643 ] +INFO:root:Average loss=0.303243665933609 +INFO:root: +[Valid][1][UTT][ MSE = 0.3971 | LCC = 0.9019 | SRCC = 0.9008 ] +INFO:root:Average loss=0.25667809027433397 +INFO:root: +[Valid][2][UTT][ MSE = 0.3353 | LCC = 0.9160 | SRCC = 0.9142 ] +INFO:root:Average loss=0.22553687220811844 +INFO:root: +[Valid][3][UTT][ MSE = 0.3260 | LCC = 0.9194 | SRCC = 0.9175 ] +INFO:root:Average loss=0.20512797152996062 +INFO:root: +[Valid][4][UTT][ MSE = 0.3104 | LCC = 0.9238 | SRCC = 0.9239 ] +INFO:root:Average loss=0.18597009319067 +INFO:root: +[Valid][5][UTT][ MSE = 0.3170 | LCC = 0.9315 | SRCC = 0.9308 ] +INFO:root:Average loss=0.17957467126846313 +INFO:root: +[Valid][6][UTT][ MSE = 0.2949 | LCC = 0.9338 | SRCC = 0.9321 ] +INFO:root:Average loss=0.15729636657238005 +INFO:root: +[Valid][7][UTT][ MSE = 0.3258 | LCC = 0.9367 | SRCC = 0.9363 ] +INFO:root:Average loss=0.14899472576379777 +INFO:root: +[Valid][8][UTT][ MSE = 0.2971 | LCC = 0.9268 | SRCC = 0.9277 ] +INFO:root:Average loss=0.13205405294895173 +INFO:root: +[Valid][9][UTT][ MSE = 0.2655 | LCC = 0.9373 | SRCC = 0.9355 ] +INFO:root:Average loss=0.13576966911554336 +INFO:root: +[Valid][10][UTT][ MSE = 0.2252 | LCC = 0.9428 | SRCC = 0.9416 ] +INFO:root:Average loss=0.13276039987802504 +INFO:root: +[Valid][11][UTT][ MSE = 0.2598 | LCC = 0.9377 | SRCC = 0.9376 ] +INFO:root:Average loss=0.11596429860591888 +INFO:root: +[Valid][12][UTT][ MSE = 0.2295 | LCC = 0.9410 | SRCC = 0.9393 ] +INFO:root:Average loss=0.10739486885070801 +INFO:root: +[Valid][13][UTT][ MSE = 0.2303 | LCC = 0.9434 | SRCC = 0.9420 ] +INFO:root:Average loss=0.10439918795228005 +INFO:root: +[Valid][14][UTT][ MSE = 0.2885 | LCC = 0.9426 | SRCC = 0.9427 ] +INFO:root:Average loss=0.09738990345597268 +INFO:root: +[Valid][15][UTT][ MSE = 0.2258 | LCC = 0.9441 | SRCC = 0.9432 ] +INFO:root:Average loss=0.09340444296598434 +INFO:root: +[Valid][16][UTT][ MSE = 0.2814 | LCC = 0.9436 | SRCC = 0.9435 ] +INFO:root:Average loss=0.0902908126115799 +INFO:root: +[Valid][17][UTT][ MSE = 0.2370 | LCC = 0.9425 | SRCC = 0.9437 ] +INFO:root:Average loss=0.09312630727887154 +INFO:root: +[Valid][18][UTT][ MSE = 0.2607 | LCC = 0.9442 | SRCC = 0.9428 ] +INFO:root:Average loss=0.08534973469376564 +INFO:root: +[Valid][19][UTT][ MSE = 0.2348 | LCC = 0.9449 | SRCC = 0.9428 ] +INFO:root:Average loss=0.08205999481678009 +INFO:root: +[Valid][20][UTT][ MSE = 0.2555 | LCC = 0.9455 | SRCC = 0.9444 ] +INFO:root:Average loss=0.07901296889781952 +INFO:root: +[Valid][21][UTT][ MSE = 0.2285 | LCC = 0.9444 | SRCC = 0.9443 ] +INFO:root:Average loss=0.08428770443797111 +INFO:root: +[Valid][22][UTT][ MSE = 0.2647 | LCC = 0.9452 | SRCC = 0.9456 ] +INFO:root:Average loss=0.07852569377422333 +INFO:root: +[Valid][23][UTT][ MSE = 0.2516 | LCC = 0.9455 | SRCC = 0.9448 ] +INFO:root:Average loss=0.06840201875567437 +INFO:root: +[Valid][24][UTT][ MSE = 0.2073 | LCC = 0.9446 | SRCC = 0.9437 ] +INFO:root:Average loss=0.06443069812655448 +INFO:root: +[Valid][25][UTT][ MSE = 0.2137 | LCC = 0.9454 | SRCC = 0.9439 ] +INFO:root:Average loss=0.06541255724430084 +INFO:root: +[Valid][26][UTT][ MSE = 0.2044 | LCC = 0.9456 | SRCC = 0.9448 ] +INFO:root:Average loss=0.06097881990671158 +INFO:root: +[Valid][27][UTT][ MSE = 0.2436 | LCC = 0.9471 | SRCC = 0.9461 ] +INFO:root:Average loss=0.05917632919549942 +INFO:root: +[Valid][28][UTT][ MSE = 0.2305 | LCC = 0.9460 | SRCC = 0.9448 ] +INFO:root:Average loss=0.056854727953672406 +INFO:root: +[Valid][29][UTT][ MSE = 0.2643 | LCC = 0.9447 | SRCC = 0.9441 ] +INFO:root:Average loss=0.05559663450717926 +INFO:root: +[Valid][30][UTT][ MSE = 0.2252 | LCC = 0.9468 | SRCC = 0.9449 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2442 | LCC = 0.9487 | SRCC = 0.9485 ] diff --git a/results_tencent/w2v2_large_asr_en/layer18/run4/config.gin b/results_tencent/w2v2_large_asr_en/layer18/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer18/run4/model.pt b/results_tencent/w2v2_large_asr_en/layer18/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e3719b6f551a1fd99ca269c1c1fb5bcdf95e63ab --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3158fef1c70428d3c2807afc7d265ee68c5083953dbd6376d0759722cca5c642 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer18/run4/model_best.pt b/results_tencent/w2v2_large_asr_en/layer18/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..960a4dee98ecf9f863861b902b1f9ea300535c90 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2442810196c32004780c9268d91da4be104fb52f82c7c7a268d3eff8f83cfbc7 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer18/run4/train.log b/results_tencent/w2v2_large_asr_en/layer18/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d7950629c67ec3936e7fc596843e02bba9a524b6 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer18/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6102126265764236 +INFO:root: +[Valid][0][UTT][ MSE = 0.4418 | LCC = 0.8626 | SRCC = 0.8630 ] +INFO:root:Average loss=0.30551685619354246 +INFO:root: +[Valid][1][UTT][ MSE = 0.3453 | LCC = 0.8939 | SRCC = 0.8952 ] +INFO:root:Average loss=0.26828211069107055 +INFO:root: +[Valid][2][UTT][ MSE = 0.3601 | LCC = 0.9104 | SRCC = 0.9087 ] +INFO:root:Average loss=0.2332269948720932 +INFO:root: +[Valid][3][UTT][ MSE = 0.3603 | LCC = 0.9161 | SRCC = 0.9157 ] +INFO:root:Average loss=0.20093673610687257 +INFO:root: +[Valid][4][UTT][ MSE = 0.2872 | LCC = 0.9272 | SRCC = 0.9250 ] +INFO:root:Average loss=0.19345596462488174 +INFO:root: +[Valid][5][UTT][ MSE = 0.3616 | LCC = 0.9246 | SRCC = 0.9249 ] +INFO:root:Average loss=0.17888455140590667 +INFO:root: +[Valid][6][UTT][ MSE = 0.3858 | LCC = 0.9295 | SRCC = 0.9294 ] +INFO:root:Average loss=0.16333846986293793 +INFO:root: +[Valid][7][UTT][ MSE = 0.2587 | LCC = 0.9370 | SRCC = 0.9349 ] +INFO:root:Average loss=0.14902453243732453 +INFO:root: +[Valid][8][UTT][ MSE = 0.2638 | LCC = 0.9395 | SRCC = 0.9383 ] +INFO:root:Average loss=0.15124420261383056 +INFO:root: +[Valid][9][UTT][ MSE = 0.2579 | LCC = 0.9397 | SRCC = 0.9391 ] +INFO:root:Average loss=0.13568703180551528 +INFO:root: +[Valid][10][UTT][ MSE = 0.2613 | LCC = 0.9415 | SRCC = 0.9405 ] +INFO:root:Average loss=0.1272286741733551 +INFO:root: +[Valid][11][UTT][ MSE = 0.2717 | LCC = 0.9438 | SRCC = 0.9426 ] +INFO:root:Average loss=0.12224530172348022 +INFO:root: +[Valid][12][UTT][ MSE = 0.2429 | LCC = 0.9443 | SRCC = 0.9432 ] +INFO:root:Average loss=0.11902501362562179 +INFO:root: +[Valid][13][UTT][ MSE = 0.2860 | LCC = 0.9416 | SRCC = 0.9428 ] +INFO:root:Average loss=0.11196209427714347 +INFO:root: +[Valid][14][UTT][ MSE = 0.2327 | LCC = 0.9468 | SRCC = 0.9460 ] +INFO:root:Average loss=0.10526347625255585 +INFO:root: +[Valid][15][UTT][ MSE = 0.2293 | LCC = 0.9447 | SRCC = 0.9444 ] +INFO:root:Average loss=0.10039496326446533 +INFO:root: +[Valid][16][UTT][ MSE = 0.2708 | LCC = 0.9476 | SRCC = 0.9472 ] +INFO:root:Average loss=0.08976116329431534 +INFO:root: +[Valid][17][UTT][ MSE = 0.2537 | LCC = 0.9482 | SRCC = 0.9472 ] +INFO:root:Average loss=0.08886381298303604 +INFO:root: +[Valid][18][UTT][ MSE = 0.2225 | LCC = 0.9469 | SRCC = 0.9462 ] +INFO:root:Average loss=0.08594983112812042 +INFO:root: +[Valid][19][UTT][ MSE = 0.1874 | LCC = 0.9474 | SRCC = 0.9463 ] +INFO:root:Average loss=0.0835591955780983 +INFO:root: +[Valid][20][UTT][ MSE = 0.2643 | LCC = 0.9456 | SRCC = 0.9452 ] +INFO:root:Average loss=0.0778877203464508 +INFO:root: +[Valid][21][UTT][ MSE = 0.2852 | LCC = 0.9459 | SRCC = 0.9464 ] +INFO:root:Average loss=0.07563944989442825 +INFO:root: +[Valid][22][UTT][ MSE = 0.2259 | LCC = 0.9484 | SRCC = 0.9476 ] +INFO:root:Average loss=0.07159450531005859 +INFO:root: +[Valid][23][UTT][ MSE = 0.2624 | LCC = 0.9474 | SRCC = 0.9464 ] +INFO:root:Average loss=0.07086541756987572 +INFO:root: +[Valid][24][UTT][ MSE = 0.2311 | LCC = 0.9488 | SRCC = 0.9473 ] +INFO:root:Average loss=0.07061072859168052 +INFO:root: +[Valid][25][UTT][ MSE = 0.2258 | LCC = 0.9488 | SRCC = 0.9477 ] +INFO:root:Average loss=0.0662876361310482 +INFO:root: +[Valid][26][UTT][ MSE = 0.2661 | LCC = 0.9476 | SRCC = 0.9479 ] +INFO:root:Average loss=0.06478247329592705 +INFO:root: +[Valid][27][UTT][ MSE = 0.3038 | LCC = 0.9438 | SRCC = 0.9458 ] +INFO:root:Average loss=0.06044148802757263 +INFO:root: +[Valid][28][UTT][ MSE = 0.2340 | LCC = 0.9485 | SRCC = 0.9472 ] +INFO:root:Average loss=0.05938212922215462 +INFO:root: +[Valid][29][UTT][ MSE = 0.2006 | LCC = 0.9480 | SRCC = 0.9474 ] +INFO:root:Average loss=0.059635157078504565 +INFO:root: +[Valid][30][UTT][ MSE = 0.2685 | LCC = 0.9449 | SRCC = 0.9457 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2334 | LCC = 0.9475 | SRCC = 0.9471 ] diff --git a/results_tencent/w2v2_large_asr_en/layer2/run0/config.gin b/results_tencent/w2v2_large_asr_en/layer2/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer2/run0/model.pt b/results_tencent/w2v2_large_asr_en/layer2/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..eff006a7f2841c38799600fba8a44de16ea857e5 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6cbd8539977de14f7732b3d094e17b30e2e058f7376b137451c511acf21a36cb +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer2/run0/model_best.pt b/results_tencent/w2v2_large_asr_en/layer2/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e7009ada0be3510f442798b23d3e4cce17599245 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5c3c331cc46834b6971708a3f1edcef0c942cfe98360281b431a979a4f4c2c01 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer2/run0/train.log b/results_tencent/w2v2_large_asr_en/layer2/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5f50d33fa2fd88aee62a3d2f62a752718e1460cd --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.41756377959251406 +INFO:root: +[Valid][0][UTT][ MSE = 0.3429 | LCC = 0.9123 | SRCC = 0.9150 ] +INFO:root:Average loss=0.22392227405309678 +INFO:root: +[Valid][1][UTT][ MSE = 0.2536 | LCC = 0.9346 | SRCC = 0.9349 ] +INFO:root:Average loss=0.17686618876457214 +INFO:root: +[Valid][2][UTT][ MSE = 0.2823 | LCC = 0.9395 | SRCC = 0.9402 ] +INFO:root:Average loss=0.15490979105234146 +INFO:root: +[Valid][3][UTT][ MSE = 0.2594 | LCC = 0.9484 | SRCC = 0.9485 ] +INFO:root:Average loss=0.14069553565979004 +INFO:root: +[Valid][4][UTT][ MSE = 0.2730 | LCC = 0.9456 | SRCC = 0.9474 ] +INFO:root:Average loss=0.13692270827293396 +INFO:root: +[Valid][5][UTT][ MSE = 0.2356 | LCC = 0.9520 | SRCC = 0.9530 ] +INFO:root:Average loss=0.12145705634355546 +INFO:root: +[Valid][6][UTT][ MSE = 0.2445 | LCC = 0.9523 | SRCC = 0.9542 ] +INFO:root:Average loss=0.1164344362616539 +INFO:root: +[Valid][7][UTT][ MSE = 0.2749 | LCC = 0.9544 | SRCC = 0.9556 ] +INFO:root:Average loss=0.10552947291731835 +INFO:root: +[Valid][8][UTT][ MSE = 0.2282 | LCC = 0.9545 | SRCC = 0.9551 ] +INFO:root:Average loss=0.09850871229171752 +INFO:root: +[Valid][9][UTT][ MSE = 0.1950 | LCC = 0.9562 | SRCC = 0.9572 ] +INFO:root:Average loss=0.08938854837417602 +INFO:root: +[Valid][10][UTT][ MSE = 0.1844 | LCC = 0.9537 | SRCC = 0.9550 ] +INFO:root:Average loss=0.08770890617370605 +INFO:root: +[Valid][11][UTT][ MSE = 0.1843 | LCC = 0.9558 | SRCC = 0.9572 ] +INFO:root:Average loss=0.08409557941555977 +INFO:root: +[Valid][12][UTT][ MSE = 0.2226 | LCC = 0.9583 | SRCC = 0.9591 ] +INFO:root:Average loss=0.07956362891197205 +INFO:root: +[Valid][13][UTT][ MSE = 0.1944 | LCC = 0.9603 | SRCC = 0.9612 ] +INFO:root:Average loss=0.07134847077727317 +INFO:root: +[Valid][14][UTT][ MSE = 0.2219 | LCC = 0.9587 | SRCC = 0.9600 ] +INFO:root:Average loss=0.07405072623491288 +INFO:root: +[Valid][15][UTT][ MSE = 0.1914 | LCC = 0.9594 | SRCC = 0.9599 ] +INFO:root:Average loss=0.07079020871222019 +INFO:root: +[Valid][16][UTT][ MSE = 0.1981 | LCC = 0.9603 | SRCC = 0.9604 ] +INFO:root:Average loss=0.06041300314664841 +INFO:root: +[Valid][17][UTT][ MSE = 0.2008 | LCC = 0.9588 | SRCC = 0.9594 ] +INFO:root:Average loss=0.06530716925859452 +INFO:root: +[Valid][18][UTT][ MSE = 0.1777 | LCC = 0.9600 | SRCC = 0.9600 ] +INFO:root:Average loss=0.05848041471838951 +INFO:root: +[Valid][19][UTT][ MSE = 0.2153 | LCC = 0.9593 | SRCC = 0.9597 ] +INFO:root:Average loss=0.05877746561169624 +INFO:root: +[Valid][20][UTT][ MSE = 0.2145 | LCC = 0.9607 | SRCC = 0.9610 ] +INFO:root:Average loss=0.05124252784252167 +INFO:root: +[Valid][21][UTT][ MSE = 0.1699 | LCC = 0.9610 | SRCC = 0.9614 ] +INFO:root:Average loss=0.05127454434335232 +INFO:root: +[Valid][22][UTT][ MSE = 0.2210 | LCC = 0.9593 | SRCC = 0.9593 ] +INFO:root:Average loss=0.049063825607299806 +INFO:root: +[Valid][23][UTT][ MSE = 0.2056 | LCC = 0.9607 | SRCC = 0.9603 ] +INFO:root:Average loss=0.04651565146446228 +INFO:root: +[Valid][24][UTT][ MSE = 0.1686 | LCC = 0.9598 | SRCC = 0.9607 ] +INFO:root:Average loss=0.047851205214858056 +INFO:root: +[Valid][25][UTT][ MSE = 0.2041 | LCC = 0.9608 | SRCC = 0.9607 ] +INFO:root:Average loss=0.04880768731236458 +INFO:root: +[Valid][26][UTT][ MSE = 0.1898 | LCC = 0.9625 | SRCC = 0.9626 ] +INFO:root:Average loss=0.04123785138130188 +INFO:root: +[Valid][27][UTT][ MSE = 0.1999 | LCC = 0.9603 | SRCC = 0.9601 ] +INFO:root:Average loss=0.041690359011292456 +INFO:root: +[Valid][28][UTT][ MSE = 0.2076 | LCC = 0.9607 | SRCC = 0.9605 ] +INFO:root:Average loss=0.04121166342496872 +INFO:root: +[Valid][29][UTT][ MSE = 0.1910 | LCC = 0.9581 | SRCC = 0.9578 ] +INFO:root:Average loss=0.038247710466384886 +INFO:root: +[Valid][30][UTT][ MSE = 0.1852 | LCC = 0.9602 | SRCC = 0.9600 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1921 | LCC = 0.9615 | SRCC = 0.9626 ] diff --git a/results_tencent/w2v2_large_asr_en/layer2/run1/config.gin b/results_tencent/w2v2_large_asr_en/layer2/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer2/run1/model.pt b/results_tencent/w2v2_large_asr_en/layer2/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a86f4b15c8fd1057b45a9c2721a3717b36a14216 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:59230aa9e1d1f9a51e7632a1de6107c1d05607ce7faba4d6c2383fb55a443638 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer2/run1/model_best.pt b/results_tencent/w2v2_large_asr_en/layer2/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..26a7c2cee5d9718c8b893723c0fb1018b7d7a6f6 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bd6d77138ed458b723164e8e69aece4dc8b5c2cba88176154953a96b3f22b182 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer2/run1/train.log b/results_tencent/w2v2_large_asr_en/layer2/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e2bd5a277bd0dbfae10b9ccb532bfb9a470def8c --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.48212439227104187 +INFO:root: +[Valid][0][UTT][ MSE = 0.3624 | LCC = 0.9142 | SRCC = 0.9161 ] +INFO:root:Average loss=0.22311221241950988 +INFO:root: +[Valid][1][UTT][ MSE = 0.2971 | LCC = 0.9345 | SRCC = 0.9343 ] +INFO:root:Average loss=0.18076727372407914 +INFO:root: +[Valid][2][UTT][ MSE = 0.2566 | LCC = 0.9446 | SRCC = 0.9446 ] +INFO:root:Average loss=0.1601833249926567 +INFO:root: +[Valid][3][UTT][ MSE = 0.2574 | LCC = 0.9475 | SRCC = 0.9478 ] +INFO:root:Average loss=0.1526476390361786 +INFO:root: +[Valid][4][UTT][ MSE = 0.3151 | LCC = 0.9457 | SRCC = 0.9460 ] +INFO:root:Average loss=0.14028702473640442 +INFO:root: +[Valid][5][UTT][ MSE = 0.3771 | LCC = 0.9462 | SRCC = 0.9492 ] +INFO:root:Average loss=0.12393110507726669 +INFO:root: +[Valid][6][UTT][ MSE = 0.2318 | LCC = 0.9508 | SRCC = 0.9517 ] +INFO:root:Average loss=0.1150194766819477 +INFO:root: +[Valid][7][UTT][ MSE = 0.2262 | LCC = 0.9568 | SRCC = 0.9578 ] +INFO:root:Average loss=0.10698247209191322 +INFO:root: +[Valid][8][UTT][ MSE = 0.2142 | LCC = 0.9550 | SRCC = 0.9547 ] +INFO:root:Average loss=0.09729893431067467 +INFO:root: +[Valid][9][UTT][ MSE = 0.1907 | LCC = 0.9560 | SRCC = 0.9561 ] +INFO:root:Average loss=0.0920704645216465 +INFO:root: +[Valid][10][UTT][ MSE = 0.2373 | LCC = 0.9556 | SRCC = 0.9567 ] +INFO:root:Average loss=0.09182034489512443 +INFO:root: +[Valid][11][UTT][ MSE = 0.1960 | LCC = 0.9596 | SRCC = 0.9595 ] +INFO:root:Average loss=0.08446029952168464 +INFO:root: +[Valid][12][UTT][ MSE = 0.2212 | LCC = 0.9591 | SRCC = 0.9595 ] +INFO:root:Average loss=0.0811423495709896 +INFO:root: +[Valid][13][UTT][ MSE = 0.1733 | LCC = 0.9609 | SRCC = 0.9612 ] +INFO:root:Average loss=0.07341405662894249 +INFO:root: +[Valid][14][UTT][ MSE = 0.2048 | LCC = 0.9580 | SRCC = 0.9579 ] +INFO:root:Average loss=0.0704014017879963 +INFO:root: +[Valid][15][UTT][ MSE = 0.1915 | LCC = 0.9614 | SRCC = 0.9607 ] +INFO:root:Average loss=0.07042464593052863 +INFO:root: +[Valid][16][UTT][ MSE = 0.2263 | LCC = 0.9582 | SRCC = 0.9587 ] +INFO:root:Average loss=0.06585309052467346 +INFO:root: +[Valid][17][UTT][ MSE = 0.1768 | LCC = 0.9618 | SRCC = 0.9615 ] +INFO:root:Average loss=0.0597386679649353 +INFO:root: +[Valid][18][UTT][ MSE = 0.2023 | LCC = 0.9628 | SRCC = 0.9628 ] +INFO:root:Average loss=0.06594741824269294 +INFO:root: +[Valid][19][UTT][ MSE = 0.2128 | LCC = 0.9605 | SRCC = 0.9616 ] +INFO:root:Average loss=0.05767556753754616 +INFO:root: +[Valid][20][UTT][ MSE = 0.1809 | LCC = 0.9622 | SRCC = 0.9621 ] +INFO:root:Average loss=0.053921602010726925 +INFO:root: +[Valid][21][UTT][ MSE = 0.1923 | LCC = 0.9629 | SRCC = 0.9631 ] +INFO:root:Average loss=0.054010206028819084 +INFO:root: +[Valid][22][UTT][ MSE = 0.1769 | LCC = 0.9642 | SRCC = 0.9638 ] +INFO:root:Average loss=0.053976046711206434 +INFO:root: +[Valid][23][UTT][ MSE = 0.1993 | LCC = 0.9617 | SRCC = 0.9612 ] +INFO:root:Average loss=0.05031919214129448 +INFO:root: +[Valid][24][UTT][ MSE = 0.1938 | LCC = 0.9635 | SRCC = 0.9636 ] +INFO:root:Average loss=0.04507946875691414 +INFO:root: +[Valid][25][UTT][ MSE = 0.2001 | LCC = 0.9599 | SRCC = 0.9596 ] +INFO:root:Average loss=0.04238516403734684 +INFO:root: +[Valid][26][UTT][ MSE = 0.2086 | LCC = 0.9629 | SRCC = 0.9626 ] +INFO:root:Average loss=0.04357936298847199 +INFO:root: +[Valid][27][UTT][ MSE = 0.1692 | LCC = 0.9633 | SRCC = 0.9624 ] +INFO:root:Average loss=0.04148554413020611 +INFO:root: +[Valid][28][UTT][ MSE = 0.2083 | LCC = 0.9634 | SRCC = 0.9631 ] +INFO:root:Average loss=0.03983789491653442 +INFO:root: +[Valid][29][UTT][ MSE = 0.2207 | LCC = 0.9598 | SRCC = 0.9599 ] +INFO:root:Average loss=0.03876499502360821 +INFO:root: +[Valid][30][UTT][ MSE = 0.1820 | LCC = 0.9629 | SRCC = 0.9626 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1859 | LCC = 0.9610 | SRCC = 0.9618 ] diff --git a/results_tencent/w2v2_large_asr_en/layer2/run2/config.gin b/results_tencent/w2v2_large_asr_en/layer2/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer2/run2/model.pt b/results_tencent/w2v2_large_asr_en/layer2/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0849981c4d91fea5fee684ca4ad3d248e2beac61 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2bb47841db65bc2761ae89f2b172aba3b1b8a67d9b34239eac32db52fbeb1270 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer2/run2/model_best.pt b/results_tencent/w2v2_large_asr_en/layer2/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f7a2ac70c444d11553fa86a61496a28036b2f191 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0d6a5094a6f77e89d0e104f45e22b208d7093272136b392ea4a773c2718a1c5 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer2/run2/train.log b/results_tencent/w2v2_large_asr_en/layer2/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f12755d29c103c811c1198912d9151236559cd36 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5559580850601197 +INFO:root: +[Valid][0][UTT][ MSE = 0.3797 | LCC = 0.9132 | SRCC = 0.9143 ] +INFO:root:Average loss=0.21602416539192199 +INFO:root: +[Valid][1][UTT][ MSE = 0.2875 | LCC = 0.9375 | SRCC = 0.9382 ] +INFO:root:Average loss=0.1765017756819725 +INFO:root: +[Valid][2][UTT][ MSE = 0.2419 | LCC = 0.9427 | SRCC = 0.9444 ] +INFO:root:Average loss=0.1652507044672966 +INFO:root: +[Valid][3][UTT][ MSE = 0.2355 | LCC = 0.9459 | SRCC = 0.9478 ] +INFO:root:Average loss=0.14446209466457366 +INFO:root: +[Valid][4][UTT][ MSE = 0.2659 | LCC = 0.9497 | SRCC = 0.9502 ] +INFO:root:Average loss=0.13016019350290298 +INFO:root: +[Valid][5][UTT][ MSE = 0.2371 | LCC = 0.9533 | SRCC = 0.9537 ] +INFO:root:Average loss=0.12232233434915543 +INFO:root: +[Valid][6][UTT][ MSE = 0.2422 | LCC = 0.9532 | SRCC = 0.9547 ] +INFO:root:Average loss=0.10957212167978286 +INFO:root: +[Valid][7][UTT][ MSE = 0.2599 | LCC = 0.9545 | SRCC = 0.9562 ] +INFO:root:Average loss=0.10280134725570679 +INFO:root: +[Valid][8][UTT][ MSE = 0.2578 | LCC = 0.9559 | SRCC = 0.9572 ] +INFO:root:Average loss=0.09154695600271225 +INFO:root: +[Valid][9][UTT][ MSE = 0.2491 | LCC = 0.9566 | SRCC = 0.9568 ] +INFO:root:Average loss=0.0935698593556881 +INFO:root: +[Valid][10][UTT][ MSE = 0.1958 | LCC = 0.9610 | SRCC = 0.9612 ] +INFO:root:Average loss=0.08651286816596984 +INFO:root: +[Valid][11][UTT][ MSE = 0.2227 | LCC = 0.9601 | SRCC = 0.9608 ] +INFO:root:Average loss=0.08179792726039886 +INFO:root: +[Valid][12][UTT][ MSE = 0.1942 | LCC = 0.9558 | SRCC = 0.9561 ] +INFO:root:Average loss=0.07698289376497269 +INFO:root: +[Valid][13][UTT][ MSE = 0.2143 | LCC = 0.9588 | SRCC = 0.9599 ] +INFO:root:Average loss=0.07036341574788094 +INFO:root: +[Valid][14][UTT][ MSE = 0.2148 | LCC = 0.9618 | SRCC = 0.9620 ] +INFO:root:Average loss=0.07680560085177422 +INFO:root: +[Valid][15][UTT][ MSE = 0.2061 | LCC = 0.9600 | SRCC = 0.9607 ] +INFO:root:Average loss=0.06405531406402588 +INFO:root: +[Valid][16][UTT][ MSE = 0.1928 | LCC = 0.9597 | SRCC = 0.9601 ] +INFO:root:Average loss=0.06217843180894852 +INFO:root: +[Valid][17][UTT][ MSE = 0.1935 | LCC = 0.9630 | SRCC = 0.9636 ] +INFO:root:Average loss=0.05763460858166218 +INFO:root: +[Valid][18][UTT][ MSE = 0.2119 | LCC = 0.9610 | SRCC = 0.9610 ] +INFO:root:Average loss=0.05820499211549759 +INFO:root: +[Valid][19][UTT][ MSE = 0.2267 | LCC = 0.9614 | SRCC = 0.9626 ] +INFO:root:Average loss=0.05538643550872803 +INFO:root: +[Valid][20][UTT][ MSE = 0.2207 | LCC = 0.9606 | SRCC = 0.9613 ] +INFO:root:Average loss=0.053446696013212204 +INFO:root: +[Valid][21][UTT][ MSE = 0.2012 | LCC = 0.9616 | SRCC = 0.9621 ] +INFO:root:Average loss=0.049023329704999924 +INFO:root: +[Valid][22][UTT][ MSE = 0.2453 | LCC = 0.9621 | SRCC = 0.9636 ] +INFO:root:Average loss=0.04697053681313992 +INFO:root: +[Valid][23][UTT][ MSE = 0.1986 | LCC = 0.9604 | SRCC = 0.9614 ] +INFO:root:Average loss=0.04506545740365982 +INFO:root: +[Valid][24][UTT][ MSE = 0.2310 | LCC = 0.9625 | SRCC = 0.9634 ] +INFO:root:Average loss=0.046177057549357414 +INFO:root: +[Valid][25][UTT][ MSE = 0.1786 | LCC = 0.9618 | SRCC = 0.9618 ] +INFO:root:Average loss=0.04341256372630596 +INFO:root: +[Valid][26][UTT][ MSE = 0.2111 | LCC = 0.9632 | SRCC = 0.9635 ] +INFO:root:Average loss=0.043877051994204525 +INFO:root: +[Valid][27][UTT][ MSE = 0.2080 | LCC = 0.9631 | SRCC = 0.9634 ] +INFO:root:Average loss=0.040699367865920065 +INFO:root: +[Valid][28][UTT][ MSE = 0.1978 | LCC = 0.9635 | SRCC = 0.9639 ] +INFO:root:Average loss=0.03825125788152218 +INFO:root: +[Valid][29][UTT][ MSE = 0.1722 | LCC = 0.9627 | SRCC = 0.9632 ] +INFO:root:Average loss=0.03784282621741295 +INFO:root: +[Valid][30][UTT][ MSE = 0.2034 | LCC = 0.9632 | SRCC = 0.9638 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2105 | LCC = 0.9601 | SRCC = 0.9607 ] diff --git a/results_tencent/w2v2_large_asr_en/layer2/run3/config.gin b/results_tencent/w2v2_large_asr_en/layer2/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer2/run3/model.pt b/results_tencent/w2v2_large_asr_en/layer2/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7d439027c9724aef8ad32d1018e992100727f28c --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d3c8eebfffc229b6dd2a735a6703ceb3d8de906eaccbc5fdc86a12ed9506b474 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer2/run3/model_best.pt b/results_tencent/w2v2_large_asr_en/layer2/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..95645044b53a41dc6fd4111d9b441729701a62c7 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cadb74c80b461ad979b6f6ffe08bce1d2dd2ae51df6182741ea09b70f7bd36a8 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer2/run3/train.log b/results_tencent/w2v2_large_asr_en/layer2/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c1679af697a9308b5621237b086f33f50f8f2c37 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5355130237340927 +INFO:root: +[Valid][0][UTT][ MSE = 0.3735 | LCC = 0.9098 | SRCC = 0.9098 ] +INFO:root:Average loss=0.225749487221241 +INFO:root: +[Valid][1][UTT][ MSE = 0.3031 | LCC = 0.9301 | SRCC = 0.9302 ] +INFO:root:Average loss=0.1854689300060272 +INFO:root: +[Valid][2][UTT][ MSE = 0.3377 | LCC = 0.9386 | SRCC = 0.9402 ] +INFO:root:Average loss=0.16843498545885086 +INFO:root: +[Valid][3][UTT][ MSE = 0.2633 | LCC = 0.9428 | SRCC = 0.9431 ] +INFO:root:Average loss=0.14374856144189835 +INFO:root: +[Valid][4][UTT][ MSE = 0.2416 | LCC = 0.9470 | SRCC = 0.9473 ] +INFO:root:Average loss=0.1330959338247776 +INFO:root: +[Valid][5][UTT][ MSE = 0.2201 | LCC = 0.9443 | SRCC = 0.9462 ] +INFO:root:Average loss=0.1269034831225872 +INFO:root: +[Valid][6][UTT][ MSE = 0.2106 | LCC = 0.9508 | SRCC = 0.9518 ] +INFO:root:Average loss=0.11339101564884185 +INFO:root: +[Valid][7][UTT][ MSE = 0.2506 | LCC = 0.9534 | SRCC = 0.9543 ] +INFO:root:Average loss=0.11085647475719453 +INFO:root: +[Valid][8][UTT][ MSE = 0.2260 | LCC = 0.9543 | SRCC = 0.9546 ] +INFO:root:Average loss=0.1002938235104084 +INFO:root: +[Valid][9][UTT][ MSE = 0.2210 | LCC = 0.9568 | SRCC = 0.9567 ] +INFO:root:Average loss=0.0961852189898491 +INFO:root: +[Valid][10][UTT][ MSE = 0.1996 | LCC = 0.9571 | SRCC = 0.9576 ] +INFO:root:Average loss=0.08836196330189705 +INFO:root: +[Valid][11][UTT][ MSE = 0.1857 | LCC = 0.9577 | SRCC = 0.9578 ] +INFO:root:Average loss=0.08094926622509957 +INFO:root: +[Valid][12][UTT][ MSE = 0.2185 | LCC = 0.9563 | SRCC = 0.9568 ] +INFO:root:Average loss=0.0790971406698227 +INFO:root: +[Valid][13][UTT][ MSE = 0.1957 | LCC = 0.9587 | SRCC = 0.9591 ] +INFO:root:Average loss=0.07730079644918442 +INFO:root: +[Valid][14][UTT][ MSE = 0.1814 | LCC = 0.9578 | SRCC = 0.9587 ] +INFO:root:Average loss=0.07160906445980073 +INFO:root: +[Valid][15][UTT][ MSE = 0.2065 | LCC = 0.9558 | SRCC = 0.9559 ] +INFO:root:Average loss=0.06808837223052978 +INFO:root: +[Valid][16][UTT][ MSE = 0.1852 | LCC = 0.9594 | SRCC = 0.9591 ] +INFO:root:Average loss=0.06572064319252968 +INFO:root: +[Valid][17][UTT][ MSE = 0.1863 | LCC = 0.9568 | SRCC = 0.9586 ] +INFO:root:Average loss=0.0650870690047741 +INFO:root: +[Valid][18][UTT][ MSE = 0.1883 | LCC = 0.9602 | SRCC = 0.9607 ] +INFO:root:Average loss=0.05874285012483597 +INFO:root: +[Valid][19][UTT][ MSE = 0.2180 | LCC = 0.9585 | SRCC = 0.9590 ] +INFO:root:Average loss=0.059842857897281644 +INFO:root: +[Valid][20][UTT][ MSE = 0.2096 | LCC = 0.9603 | SRCC = 0.9606 ] +INFO:root:Average loss=0.053929484397172926 +INFO:root: +[Valid][21][UTT][ MSE = 0.1986 | LCC = 0.9593 | SRCC = 0.9597 ] +INFO:root:Average loss=0.05089017528295517 +INFO:root: +[Valid][22][UTT][ MSE = 0.2114 | LCC = 0.9588 | SRCC = 0.9590 ] +INFO:root:Average loss=0.05127060253918171 +INFO:root: +[Valid][23][UTT][ MSE = 0.1868 | LCC = 0.9601 | SRCC = 0.9600 ] +INFO:root:Average loss=0.04751098343729973 +INFO:root: +[Valid][24][UTT][ MSE = 0.1988 | LCC = 0.9572 | SRCC = 0.9578 ] +INFO:root:Average loss=0.04515090717375279 +INFO:root: +[Valid][25][UTT][ MSE = 0.2112 | LCC = 0.9585 | SRCC = 0.9584 ] +INFO:root:Average loss=0.04424972467124462 +INFO:root: +[Valid][26][UTT][ MSE = 0.2159 | LCC = 0.9597 | SRCC = 0.9601 ] +INFO:root:Average loss=0.04354153883457184 +INFO:root: +[Valid][27][UTT][ MSE = 0.1987 | LCC = 0.9591 | SRCC = 0.9592 ] +INFO:root:Average loss=0.040391827270388606 +INFO:root: +[Valid][28][UTT][ MSE = 0.1975 | LCC = 0.9610 | SRCC = 0.9608 ] +INFO:root:Average loss=0.04200786478817463 +INFO:root: +[Valid][29][UTT][ MSE = 0.1845 | LCC = 0.9596 | SRCC = 0.9590 ] +INFO:root:Average loss=0.03738595090806484 +INFO:root: +[Valid][30][UTT][ MSE = 0.1980 | LCC = 0.9596 | SRCC = 0.9594 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2035 | LCC = 0.9600 | SRCC = 0.9604 ] diff --git a/results_tencent/w2v2_large_asr_en/layer2/run4/config.gin b/results_tencent/w2v2_large_asr_en/layer2/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer2/run4/model.pt b/results_tencent/w2v2_large_asr_en/layer2/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1a12aa9ea0e08d3693d0d56c4c7e1715144ca8cd --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c8367ad80d731de389193a72f2c2b81d507c17b04e1584aee20f42d2750a2f7 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer2/run4/model_best.pt b/results_tencent/w2v2_large_asr_en/layer2/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..89279f25b552da0ec7f75e840cbbb83f0e5f8eb6 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ad1fab0c0e9aac233d2cc3514f16592ba391771639cb739d5f58388907dc5c0a +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer2/run4/train.log b/results_tencent/w2v2_large_asr_en/layer2/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9e151f655dff339571c8188bca524b2014403ff5 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer2/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4701287450790405 +INFO:root: +[Valid][0][UTT][ MSE = 0.2962 | LCC = 0.9191 | SRCC = 0.9184 ] +INFO:root:Average loss=0.21427657234668732 +INFO:root: +[Valid][1][UTT][ MSE = 0.2636 | LCC = 0.9335 | SRCC = 0.9335 ] +INFO:root:Average loss=0.17918231403827667 +INFO:root: +[Valid][2][UTT][ MSE = 0.2553 | LCC = 0.9451 | SRCC = 0.9452 ] +INFO:root:Average loss=0.15987289774417876 +INFO:root: +[Valid][3][UTT][ MSE = 0.2605 | LCC = 0.9466 | SRCC = 0.9476 ] +INFO:root:Average loss=0.1457455381155014 +INFO:root: +[Valid][4][UTT][ MSE = 0.2238 | LCC = 0.9488 | SRCC = 0.9500 ] +INFO:root:Average loss=0.13014294576644897 +INFO:root: +[Valid][5][UTT][ MSE = 0.1948 | LCC = 0.9524 | SRCC = 0.9534 ] +INFO:root:Average loss=0.11642742443084717 +INFO:root: +[Valid][6][UTT][ MSE = 0.2278 | LCC = 0.9536 | SRCC = 0.9544 ] +INFO:root:Average loss=0.11127803713083267 +INFO:root: +[Valid][7][UTT][ MSE = 0.1993 | LCC = 0.9575 | SRCC = 0.9580 ] +INFO:root:Average loss=0.10273869544267654 +INFO:root: +[Valid][8][UTT][ MSE = 0.2184 | LCC = 0.9575 | SRCC = 0.9585 ] +INFO:root:Average loss=0.09885690921545029 +INFO:root: +[Valid][9][UTT][ MSE = 0.2563 | LCC = 0.9580 | SRCC = 0.9588 ] +INFO:root:Average loss=0.09123285222053527 +INFO:root: +[Valid][10][UTT][ MSE = 0.2023 | LCC = 0.9596 | SRCC = 0.9602 ] +INFO:root:Average loss=0.08367112568020821 +INFO:root: +[Valid][11][UTT][ MSE = 0.2311 | LCC = 0.9608 | SRCC = 0.9622 ] +INFO:root:Average loss=0.07931741985678673 +INFO:root: +[Valid][12][UTT][ MSE = 0.1941 | LCC = 0.9602 | SRCC = 0.9606 ] +INFO:root:Average loss=0.0781889889240265 +INFO:root: +[Valid][13][UTT][ MSE = 0.1867 | LCC = 0.9613 | SRCC = 0.9617 ] +INFO:root:Average loss=0.07406434205174446 +INFO:root: +[Valid][14][UTT][ MSE = 0.2105 | LCC = 0.9612 | SRCC = 0.9611 ] +INFO:root:Average loss=0.06704205816984177 +INFO:root: +[Valid][15][UTT][ MSE = 0.1734 | LCC = 0.9624 | SRCC = 0.9626 ] +INFO:root:Average loss=0.06646271282434464 +INFO:root: +[Valid][16][UTT][ MSE = 0.2183 | LCC = 0.9604 | SRCC = 0.9616 ] +INFO:root:Average loss=0.06429656702280044 +INFO:root: +[Valid][17][UTT][ MSE = 0.1933 | LCC = 0.9628 | SRCC = 0.9629 ] +INFO:root:Average loss=0.06152344512939453 +INFO:root: +[Valid][18][UTT][ MSE = 0.2286 | LCC = 0.9601 | SRCC = 0.9606 ] +INFO:root:Average loss=0.06263234862685203 +INFO:root: +[Valid][19][UTT][ MSE = 0.1999 | LCC = 0.9622 | SRCC = 0.9626 ] +INFO:root:Average loss=0.05284657308459282 +INFO:root: +[Valid][20][UTT][ MSE = 0.1871 | LCC = 0.9614 | SRCC = 0.9615 ] +INFO:root:Average loss=0.052460508272051815 +INFO:root: +[Valid][21][UTT][ MSE = 0.1704 | LCC = 0.9640 | SRCC = 0.9641 ] +INFO:root:Average loss=0.05159590576589108 +INFO:root: +[Valid][22][UTT][ MSE = 0.2300 | LCC = 0.9610 | SRCC = 0.9617 ] +INFO:root:Average loss=0.049756208464503286 +INFO:root: +[Valid][23][UTT][ MSE = 0.2072 | LCC = 0.9624 | SRCC = 0.9636 ] +INFO:root:Average loss=0.046578710570931436 +INFO:root: +[Valid][24][UTT][ MSE = 0.1785 | LCC = 0.9608 | SRCC = 0.9606 ] +INFO:root:Average loss=0.04717000886797905 +INFO:root: +[Valid][25][UTT][ MSE = 0.1842 | LCC = 0.9622 | SRCC = 0.9622 ] +INFO:root:Average loss=0.04494200803339481 +INFO:root: +[Valid][26][UTT][ MSE = 0.1825 | LCC = 0.9636 | SRCC = 0.9638 ] +INFO:root:Average loss=0.042171341359615325 +INFO:root: +[Valid][27][UTT][ MSE = 0.2172 | LCC = 0.9622 | SRCC = 0.9632 ] +INFO:root:Average loss=0.040654755756258966 +INFO:root: +[Valid][28][UTT][ MSE = 0.2196 | LCC = 0.9609 | SRCC = 0.9625 ] +INFO:root:Average loss=0.039393205121159555 +INFO:root: +[Valid][29][UTT][ MSE = 0.2071 | LCC = 0.9606 | SRCC = 0.9612 ] +INFO:root:Average loss=0.03564129790663719 +INFO:root: +[Valid][30][UTT][ MSE = 0.1829 | LCC = 0.9622 | SRCC = 0.9623 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1751 | LCC = 0.9625 | SRCC = 0.9629 ] diff --git a/results_tencent/w2v2_large_asr_en/layer20/run0/config.gin b/results_tencent/w2v2_large_asr_en/layer20/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer20/run0/model.pt b/results_tencent/w2v2_large_asr_en/layer20/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7d8704d426382417f94f8463be19c9d9e0344826 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:357e44ad8b7da5fb7055c5796d1ec8e64ba4e35f0d3cafc6bc0bfd8bb868345f +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer20/run0/model_best.pt b/results_tencent/w2v2_large_asr_en/layer20/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2d3917139eae18f6ee68584ee07abbbe076df37c --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e75c80b00a350744cfa68a58e8733b96e9f2b0813867624b071a3b54f88b383c +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer20/run0/train.log b/results_tencent/w2v2_large_asr_en/layer20/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..940ed191d0f0cf8eb4ad828d604cd3b17756c48d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.570779758810997 +INFO:root: +[Valid][0][UTT][ MSE = 0.4253 | LCC = 0.8617 | SRCC = 0.8575 ] +INFO:root:Average loss=0.3494913934469223 +INFO:root: +[Valid][1][UTT][ MSE = 0.3704 | LCC = 0.8885 | SRCC = 0.8876 ] +INFO:root:Average loss=0.2926796270608902 +INFO:root: +[Valid][2][UTT][ MSE = 0.3669 | LCC = 0.9007 | SRCC = 0.8996 ] +INFO:root:Average loss=0.25853764390945433 +INFO:root: +[Valid][3][UTT][ MSE = 0.3161 | LCC = 0.9092 | SRCC = 0.9080 ] +INFO:root:Average loss=0.24539841747283936 +INFO:root: +[Valid][4][UTT][ MSE = 0.3407 | LCC = 0.9160 | SRCC = 0.9137 ] +INFO:root:Average loss=0.22299443578720093 +INFO:root: +[Valid][5][UTT][ MSE = 0.3756 | LCC = 0.9104 | SRCC = 0.9093 ] +INFO:root:Average loss=0.20048441052436827 +INFO:root: +[Valid][6][UTT][ MSE = 0.2974 | LCC = 0.9258 | SRCC = 0.9251 ] +INFO:root:Average loss=0.19592836117744447 +INFO:root: +[Valid][7][UTT][ MSE = 0.2921 | LCC = 0.9280 | SRCC = 0.9263 ] +INFO:root:Average loss=0.19494073390960692 +INFO:root: +[Valid][8][UTT][ MSE = 0.2962 | LCC = 0.9289 | SRCC = 0.9279 ] +INFO:root:Average loss=0.16749224799871446 +INFO:root: +[Valid][9][UTT][ MSE = 0.3161 | LCC = 0.9315 | SRCC = 0.9310 ] +INFO:root:Average loss=0.16167177772521973 +INFO:root: +[Valid][10][UTT][ MSE = 0.2292 | LCC = 0.9332 | SRCC = 0.9315 ] +INFO:root:Average loss=0.1554455338716507 +INFO:root: +[Valid][11][UTT][ MSE = 0.3011 | LCC = 0.9354 | SRCC = 0.9333 ] +INFO:root:Average loss=0.1701450901031494 +INFO:root: +[Valid][12][UTT][ MSE = 0.2572 | LCC = 0.9323 | SRCC = 0.9310 ] +INFO:root:Average loss=0.15511913317441942 +INFO:root: +[Valid][13][UTT][ MSE = 0.2846 | LCC = 0.9377 | SRCC = 0.9368 ] +INFO:root:Average loss=0.14330658954381942 +INFO:root: +[Valid][14][UTT][ MSE = 0.2166 | LCC = 0.9345 | SRCC = 0.9336 ] +INFO:root:Average loss=0.13432050901651382 +INFO:root: +[Valid][15][UTT][ MSE = 0.2823 | LCC = 0.9345 | SRCC = 0.9340 ] +INFO:root:Average loss=0.1253918132185936 +INFO:root: +[Valid][16][UTT][ MSE = 0.2246 | LCC = 0.9391 | SRCC = 0.9382 ] +INFO:root:Average loss=0.12618280059099196 +INFO:root: +[Valid][17][UTT][ MSE = 0.2648 | LCC = 0.9395 | SRCC = 0.9389 ] +INFO:root:Average loss=0.1344930454492569 +INFO:root: +[Valid][18][UTT][ MSE = 0.2558 | LCC = 0.9408 | SRCC = 0.9388 ] +INFO:root:Average loss=0.11676693201065064 +INFO:root: +[Valid][19][UTT][ MSE = 0.2806 | LCC = 0.9385 | SRCC = 0.9378 ] +INFO:root:Average loss=0.1155834636092186 +INFO:root: +[Valid][20][UTT][ MSE = 0.2412 | LCC = 0.9346 | SRCC = 0.9339 ] +INFO:root:Average loss=0.11302318227291107 +INFO:root: +[Valid][21][UTT][ MSE = 0.2512 | LCC = 0.9420 | SRCC = 0.9413 ] +INFO:root:Average loss=0.11091441577672959 +INFO:root: +[Valid][22][UTT][ MSE = 0.2748 | LCC = 0.9420 | SRCC = 0.9410 ] +INFO:root:Average loss=0.10649399250745774 +INFO:root: +[Valid][23][UTT][ MSE = 0.2271 | LCC = 0.9410 | SRCC = 0.9399 ] +INFO:root:Average loss=0.10088192504644394 +INFO:root: +[Valid][24][UTT][ MSE = 0.2215 | LCC = 0.9419 | SRCC = 0.9419 ] +INFO:root:Average loss=0.09999546825885773 +INFO:root: +[Valid][25][UTT][ MSE = 0.2324 | LCC = 0.9416 | SRCC = 0.9419 ] +INFO:root:Average loss=0.09269825503230095 +INFO:root: +[Valid][26][UTT][ MSE = 0.2207 | LCC = 0.9384 | SRCC = 0.9377 ] +INFO:root:Average loss=0.08645941925048828 +INFO:root: +[Valid][27][UTT][ MSE = 0.2380 | LCC = 0.9421 | SRCC = 0.9411 ] +INFO:root:Average loss=0.08632152605056763 +INFO:root: +[Valid][28][UTT][ MSE = 0.2273 | LCC = 0.9397 | SRCC = 0.9404 ] +INFO:root:Average loss=0.0828598543703556 +INFO:root: +[Valid][29][UTT][ MSE = 0.2778 | LCC = 0.9397 | SRCC = 0.9391 ] +INFO:root:Average loss=0.08570702037215233 +INFO:root: +[Valid][30][UTT][ MSE = 0.2271 | LCC = 0.9407 | SRCC = 0.9396 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2466 | LCC = 0.9399 | SRCC = 0.9393 ] diff --git a/results_tencent/w2v2_large_asr_en/layer20/run1/config.gin b/results_tencent/w2v2_large_asr_en/layer20/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer20/run1/model.pt b/results_tencent/w2v2_large_asr_en/layer20/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6ed400f4a1ddb93d4c0a483ac1af92a84d8613a2 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ba71f392b52d54f22cd957f3cd192fee1d9f4a0994722bed192719bfb5cde6f0 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer20/run1/model_best.pt b/results_tencent/w2v2_large_asr_en/layer20/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..79b9edacc37cf5610dc369b120a8e8f6615c66a2 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d72e18a6059b53faba77ee48b7ecc5fe21d5066d379d46d7dc1ce6cda56c9bdf +size 3404998 diff --git a/results_tencent/w2v2_large_asr_en/layer20/run1/train.log b/results_tencent/w2v2_large_asr_en/layer20/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..99dca9f0a2cd02b4c8ea4452c75a731b10c08785 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5571152033805847 +INFO:root: +[Valid][0][UTT][ MSE = 0.4972 | LCC = 0.8626 | SRCC = 0.8635 ] +INFO:root:Average loss=0.34834359467029574 +INFO:root: +[Valid][1][UTT][ MSE = 0.4351 | LCC = 0.8714 | SRCC = 0.8754 ] +INFO:root:Average loss=0.292546440243721 +INFO:root: +[Valid][2][UTT][ MSE = 0.4441 | LCC = 0.8972 | SRCC = 0.8972 ] +INFO:root:Average loss=0.2613844178915024 +INFO:root: +[Valid][3][UTT][ MSE = 0.3235 | LCC = 0.9084 | SRCC = 0.9075 ] +INFO:root:Average loss=0.23459394431114197 +INFO:root: +[Valid][4][UTT][ MSE = 0.3786 | LCC = 0.9148 | SRCC = 0.9134 ] +INFO:root:Average loss=0.23906402009725572 +INFO:root: +[Valid][5][UTT][ MSE = 0.2861 | LCC = 0.9157 | SRCC = 0.9154 ] +INFO:root:Average loss=0.21357218664884567 +INFO:root: +[Valid][6][UTT][ MSE = 0.3002 | LCC = 0.9182 | SRCC = 0.9170 ] +INFO:root:Average loss=0.19441990971565246 +INFO:root: +[Valid][7][UTT][ MSE = 0.3148 | LCC = 0.9237 | SRCC = 0.9226 ] +INFO:root:Average loss=0.18962210386991502 +INFO:root: +[Valid][8][UTT][ MSE = 0.2930 | LCC = 0.9270 | SRCC = 0.9255 ] +INFO:root:Average loss=0.16526164036989213 +INFO:root: +[Valid][9][UTT][ MSE = 0.3321 | LCC = 0.9302 | SRCC = 0.9291 ] +INFO:root:Average loss=0.1678971756696701 +INFO:root: +[Valid][10][UTT][ MSE = 0.2911 | LCC = 0.9283 | SRCC = 0.9272 ] +INFO:root:Average loss=0.16354320269823075 +INFO:root: +[Valid][11][UTT][ MSE = 0.3139 | LCC = 0.9326 | SRCC = 0.9312 ] +INFO:root:Average loss=0.15752978736162185 +INFO:root: +[Valid][12][UTT][ MSE = 0.2843 | LCC = 0.9307 | SRCC = 0.9304 ] +INFO:root:Average loss=0.17048921871185302 +INFO:root: +[Valid][13][UTT][ MSE = 0.2809 | LCC = 0.9320 | SRCC = 0.9301 ] +INFO:root:Average loss=0.13673573225736618 +INFO:root: +[Valid][14][UTT][ MSE = 0.3697 | LCC = 0.9312 | SRCC = 0.9325 ] +INFO:root:Average loss=0.13268520599603653 +INFO:root: +[Valid][15][UTT][ MSE = 0.2947 | LCC = 0.9322 | SRCC = 0.9316 ] +INFO:root:Average loss=0.12350560849905014 +INFO:root: +[Valid][16][UTT][ MSE = 0.2817 | LCC = 0.9372 | SRCC = 0.9359 ] +INFO:root:Average loss=0.12151035779714585 +INFO:root: +[Valid][17][UTT][ MSE = 0.2311 | LCC = 0.9366 | SRCC = 0.9358 ] +INFO:root:Average loss=0.11156313249468804 +INFO:root: +[Valid][18][UTT][ MSE = 0.2770 | LCC = 0.9361 | SRCC = 0.9364 ] +INFO:root:Average loss=0.10825905752182007 +INFO:root: +[Valid][19][UTT][ MSE = 0.2460 | LCC = 0.9354 | SRCC = 0.9345 ] +INFO:root:Average loss=0.11008946883678436 +INFO:root: +[Valid][20][UTT][ MSE = 0.2670 | LCC = 0.9380 | SRCC = 0.9370 ] +INFO:root:Average loss=0.10142210942506791 +INFO:root: +[Valid][21][UTT][ MSE = 0.2739 | LCC = 0.9384 | SRCC = 0.9383 ] +INFO:root:Average loss=0.10279419937729836 +INFO:root: +[Valid][22][UTT][ MSE = 0.2601 | LCC = 0.9402 | SRCC = 0.9396 ] +INFO:root:Average loss=0.09752039620280266 +INFO:root: +[Valid][23][UTT][ MSE = 0.2578 | LCC = 0.9395 | SRCC = 0.9392 ] +INFO:root:Average loss=0.09405639925599098 +INFO:root: +[Valid][24][UTT][ MSE = 0.2664 | LCC = 0.9378 | SRCC = 0.9373 ] +INFO:root:Average loss=0.0884711058139801 +INFO:root: +[Valid][25][UTT][ MSE = 0.2682 | LCC = 0.9379 | SRCC = 0.9374 ] +INFO:root:Average loss=0.09479101666808129 +INFO:root: +[Valid][26][UTT][ MSE = 0.3011 | LCC = 0.9377 | SRCC = 0.9378 ] +INFO:root:Average loss=0.08854658791422844 +INFO:root: +[Valid][27][UTT][ MSE = 0.2751 | LCC = 0.9379 | SRCC = 0.9384 ] +INFO:root:Average loss=0.08757530090212821 +INFO:root: +[Valid][28][UTT][ MSE = 0.2759 | LCC = 0.9390 | SRCC = 0.9390 ] +INFO:root:Average loss=0.08046538078784943 +INFO:root: +[Valid][29][UTT][ MSE = 0.2865 | LCC = 0.9383 | SRCC = 0.9380 ] +INFO:root:Average loss=0.07816117364168167 +INFO:root: +[Valid][30][UTT][ MSE = 0.2460 | LCC = 0.9412 | SRCC = 0.9404 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2423 | LCC = 0.9433 | SRCC = 0.9432 ] diff --git a/results_tencent/w2v2_large_asr_en/layer20/run2/config.gin b/results_tencent/w2v2_large_asr_en/layer20/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer20/run2/model.pt b/results_tencent/w2v2_large_asr_en/layer20/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..fbee42518fa307bf4b3cb8d08d50d9e1132e757a --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bbec04adbded1889c61fd4955c0ae1dbc62891bd4a1fafd770f7b7227e28fd36 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer20/run2/model_best.pt b/results_tencent/w2v2_large_asr_en/layer20/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..08a3552c908a67f55ed0e42feb2fd4e8f270ae37 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8b7e615bb10949091156ecce8a73cc6808cacc6599f11b307c1ed7f8a3763086 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer20/run2/train.log b/results_tencent/w2v2_large_asr_en/layer20/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..922ea0c0146882eadfafffe18d631152e0ab50cf --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.571088136792183 +INFO:root: +[Valid][0][UTT][ MSE = 0.4718 | LCC = 0.8473 | SRCC = 0.8492 ] +INFO:root:Average loss=0.35511121010780333 +INFO:root: +[Valid][1][UTT][ MSE = 0.4642 | LCC = 0.8826 | SRCC = 0.8823 ] +INFO:root:Average loss=0.2941438942551613 +INFO:root: +[Valid][2][UTT][ MSE = 0.3639 | LCC = 0.8993 | SRCC = 0.8980 ] +INFO:root:Average loss=0.2644513244628906 +INFO:root: +[Valid][3][UTT][ MSE = 0.3294 | LCC = 0.9036 | SRCC = 0.8997 ] +INFO:root:Average loss=0.2521021190881729 +INFO:root: +[Valid][4][UTT][ MSE = 0.3875 | LCC = 0.9099 | SRCC = 0.9107 ] +INFO:root:Average loss=0.21689273142814636 +INFO:root: +[Valid][5][UTT][ MSE = 0.2699 | LCC = 0.9214 | SRCC = 0.9198 ] +INFO:root:Average loss=0.20683700484037398 +INFO:root: +[Valid][6][UTT][ MSE = 0.3305 | LCC = 0.9234 | SRCC = 0.9227 ] +INFO:root:Average loss=0.19525082862377166 +INFO:root: +[Valid][7][UTT][ MSE = 0.3306 | LCC = 0.9284 | SRCC = 0.9272 ] +INFO:root:Average loss=0.18021054089069366 +INFO:root: +[Valid][8][UTT][ MSE = 0.3125 | LCC = 0.9258 | SRCC = 0.9258 ] +INFO:root:Average loss=0.18121760833263398 +INFO:root: +[Valid][9][UTT][ MSE = 0.3108 | LCC = 0.9332 | SRCC = 0.9328 ] +INFO:root:Average loss=0.16053584814071656 +INFO:root: +[Valid][10][UTT][ MSE = 0.2683 | LCC = 0.9305 | SRCC = 0.9311 ] +INFO:root:Average loss=0.16361240446567535 +INFO:root: +[Valid][11][UTT][ MSE = 0.2480 | LCC = 0.9339 | SRCC = 0.9345 ] +INFO:root:Average loss=0.15053082722425462 +INFO:root: +[Valid][12][UTT][ MSE = 0.3033 | LCC = 0.9301 | SRCC = 0.9315 ] +INFO:root:Average loss=0.14923762083053588 +INFO:root: +[Valid][13][UTT][ MSE = 0.2585 | LCC = 0.9373 | SRCC = 0.9361 ] +INFO:root:Average loss=0.1387394533753395 +INFO:root: +[Valid][14][UTT][ MSE = 0.2299 | LCC = 0.9361 | SRCC = 0.9349 ] +INFO:root:Average loss=0.128150108397007 +INFO:root: +[Valid][15][UTT][ MSE = 0.2603 | LCC = 0.9390 | SRCC = 0.9382 ] +INFO:root:Average loss=0.12639884805679322 +INFO:root: +[Valid][16][UTT][ MSE = 0.2455 | LCC = 0.9410 | SRCC = 0.9398 ] +INFO:root:Average loss=0.1221097796857357 +INFO:root: +[Valid][17][UTT][ MSE = 0.2456 | LCC = 0.9375 | SRCC = 0.9378 ] +INFO:root:Average loss=0.1129828531742096 +INFO:root: +[Valid][18][UTT][ MSE = 0.2364 | LCC = 0.9412 | SRCC = 0.9404 ] +INFO:root:Average loss=0.11129274326562881 +INFO:root: +[Valid][19][UTT][ MSE = 0.2432 | LCC = 0.9405 | SRCC = 0.9401 ] +INFO:root:Average loss=0.1054775173664093 +INFO:root: +[Valid][20][UTT][ MSE = 0.2422 | LCC = 0.9417 | SRCC = 0.9417 ] +INFO:root:Average loss=0.10347302135825157 +INFO:root: +[Valid][21][UTT][ MSE = 0.2550 | LCC = 0.9302 | SRCC = 0.9297 ] +INFO:root:Average loss=0.10697641825675964 +INFO:root: +[Valid][22][UTT][ MSE = 0.2294 | LCC = 0.9417 | SRCC = 0.9404 ] +INFO:root:Average loss=0.10059783667325974 +INFO:root: +[Valid][23][UTT][ MSE = 0.2512 | LCC = 0.9426 | SRCC = 0.9420 ] +INFO:root:Average loss=0.09167015621066094 +INFO:root: +[Valid][24][UTT][ MSE = 0.2549 | LCC = 0.9415 | SRCC = 0.9410 ] +INFO:root:Average loss=0.0923120894730091 +INFO:root: +[Valid][25][UTT][ MSE = 0.2465 | LCC = 0.9439 | SRCC = 0.9434 ] +INFO:root:Average loss=0.09026724582910538 +INFO:root: +[Valid][26][UTT][ MSE = 0.2169 | LCC = 0.9421 | SRCC = 0.9413 ] +INFO:root:Average loss=0.08687295967340469 +INFO:root: +[Valid][27][UTT][ MSE = 0.2461 | LCC = 0.9416 | SRCC = 0.9418 ] +INFO:root:Average loss=0.09182200747728347 +INFO:root: +[Valid][28][UTT][ MSE = 0.2291 | LCC = 0.9416 | SRCC = 0.9403 ] +INFO:root:Average loss=0.08074097180366516 +INFO:root: +[Valid][29][UTT][ MSE = 0.2162 | LCC = 0.9429 | SRCC = 0.9424 ] +INFO:root:Average loss=0.07773874500393868 +INFO:root: +[Valid][30][UTT][ MSE = 0.2320 | LCC = 0.9429 | SRCC = 0.9419 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2551 | LCC = 0.9418 | SRCC = 0.9412 ] diff --git a/results_tencent/w2v2_large_asr_en/layer20/run3/config.gin b/results_tencent/w2v2_large_asr_en/layer20/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer20/run3/model.pt b/results_tencent/w2v2_large_asr_en/layer20/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6556b1f5811682cded7fa71074a42c3f25c20b70 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e8d3b05976578f95c9322a0597c5a078016d9e9aeae1acba0b2ce8ad658bec22 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer20/run3/model_best.pt b/results_tencent/w2v2_large_asr_en/layer20/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7000e6530f20fa4d50f12eacf272feca615db356 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b02e02e35ee5299bac20e5c0c76da32932c4248b5497d7b4fc70cdc44caa0671 +size 3404998 diff --git a/results_tencent/w2v2_large_asr_en/layer20/run3/train.log b/results_tencent/w2v2_large_asr_en/layer20/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9da8fd02cab980c56761a70b88907463e5f43625 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5722440329790115 +INFO:root: +[Valid][0][UTT][ MSE = 0.5105 | LCC = 0.8517 | SRCC = 0.8538 ] +INFO:root:Average loss=0.3347939331531525 +INFO:root: +[Valid][1][UTT][ MSE = 0.3616 | LCC = 0.8799 | SRCC = 0.8832 ] +INFO:root:Average loss=0.2908579462766647 +INFO:root: +[Valid][2][UTT][ MSE = 0.3919 | LCC = 0.8984 | SRCC = 0.8975 ] +INFO:root:Average loss=0.26793857955932615 +INFO:root: +[Valid][3][UTT][ MSE = 0.3345 | LCC = 0.9057 | SRCC = 0.9046 ] +INFO:root:Average loss=0.2537023549079895 +INFO:root: +[Valid][4][UTT][ MSE = 0.3804 | LCC = 0.9097 | SRCC = 0.9079 ] +INFO:root:Average loss=0.2197005895972252 +INFO:root: +[Valid][5][UTT][ MSE = 0.2900 | LCC = 0.9149 | SRCC = 0.9128 ] +INFO:root:Average loss=0.22090983068943024 +INFO:root: +[Valid][6][UTT][ MSE = 0.2770 | LCC = 0.9193 | SRCC = 0.9176 ] +INFO:root:Average loss=0.1961937210559845 +INFO:root: +[Valid][7][UTT][ MSE = 0.3082 | LCC = 0.9263 | SRCC = 0.9245 ] +INFO:root:Average loss=0.18366835451126098 +INFO:root: +[Valid][8][UTT][ MSE = 0.2930 | LCC = 0.9274 | SRCC = 0.9262 ] +INFO:root:Average loss=0.1728554278612137 +INFO:root: +[Valid][9][UTT][ MSE = 0.2752 | LCC = 0.9319 | SRCC = 0.9304 ] +INFO:root:Average loss=0.16572149616479873 +INFO:root: +[Valid][10][UTT][ MSE = 0.3044 | LCC = 0.9302 | SRCC = 0.9296 ] +INFO:root:Average loss=0.15364411902427674 +INFO:root: +[Valid][11][UTT][ MSE = 0.2903 | LCC = 0.9250 | SRCC = 0.9231 ] +INFO:root:Average loss=0.15129924839735032 +INFO:root: +[Valid][12][UTT][ MSE = 0.3218 | LCC = 0.9368 | SRCC = 0.9347 ] +INFO:root:Average loss=0.1408944683074951 +INFO:root: +[Valid][13][UTT][ MSE = 0.3347 | LCC = 0.9335 | SRCC = 0.9340 ] +INFO:root:Average loss=0.14077809715270997 +INFO:root: +[Valid][14][UTT][ MSE = 0.3272 | LCC = 0.9365 | SRCC = 0.9364 ] +INFO:root:Average loss=0.13530577498674393 +INFO:root: +[Valid][15][UTT][ MSE = 0.2453 | LCC = 0.9399 | SRCC = 0.9387 ] +INFO:root:Average loss=0.13013106179237366 +INFO:root: +[Valid][16][UTT][ MSE = 0.2653 | LCC = 0.9380 | SRCC = 0.9380 ] +INFO:root:Average loss=0.12081256997585296 +INFO:root: +[Valid][17][UTT][ MSE = 0.3447 | LCC = 0.9332 | SRCC = 0.9351 ] +INFO:root:Average loss=0.11114966195821763 +INFO:root: +[Valid][18][UTT][ MSE = 0.2461 | LCC = 0.9417 | SRCC = 0.9402 ] +INFO:root:Average loss=0.10860725057125091 +INFO:root: +[Valid][19][UTT][ MSE = 0.2604 | LCC = 0.9390 | SRCC = 0.9379 ] +INFO:root:Average loss=0.1111840363740921 +INFO:root: +[Valid][20][UTT][ MSE = 0.2903 | LCC = 0.9409 | SRCC = 0.9404 ] +INFO:root:Average loss=0.10958324140310287 +INFO:root: +[Valid][21][UTT][ MSE = 0.2730 | LCC = 0.9421 | SRCC = 0.9414 ] +INFO:root:Average loss=0.1076541559100151 +INFO:root: +[Valid][22][UTT][ MSE = 0.2570 | LCC = 0.9420 | SRCC = 0.9414 ] +INFO:root:Average loss=0.09758102804422379 +INFO:root: +[Valid][23][UTT][ MSE = 0.2522 | LCC = 0.9405 | SRCC = 0.9395 ] +INFO:root:Average loss=0.09464734324812889 +INFO:root: +[Valid][24][UTT][ MSE = 0.2345 | LCC = 0.9373 | SRCC = 0.9369 ] +INFO:root:Average loss=0.09177326047420502 +INFO:root: +[Valid][25][UTT][ MSE = 0.2614 | LCC = 0.9417 | SRCC = 0.9413 ] +INFO:root:Average loss=0.09086647346615791 +INFO:root: +[Valid][26][UTT][ MSE = 0.2508 | LCC = 0.9427 | SRCC = 0.9417 ] +INFO:root:Average loss=0.08426251396536827 +INFO:root: +[Valid][27][UTT][ MSE = 0.2719 | LCC = 0.9387 | SRCC = 0.9380 ] +INFO:root:Average loss=0.08258171924948693 +INFO:root: +[Valid][28][UTT][ MSE = 0.2207 | LCC = 0.9413 | SRCC = 0.9410 ] +INFO:root:Average loss=0.0832513025701046 +INFO:root: +[Valid][29][UTT][ MSE = 0.2625 | LCC = 0.9258 | SRCC = 0.9259 ] +INFO:root:Average loss=0.08361900335550308 +INFO:root: +[Valid][30][UTT][ MSE = 0.2426 | LCC = 0.9391 | SRCC = 0.9377 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2529 | LCC = 0.9428 | SRCC = 0.9422 ] diff --git a/results_tencent/w2v2_large_asr_en/layer20/run4/config.gin b/results_tencent/w2v2_large_asr_en/layer20/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer20/run4/model.pt b/results_tencent/w2v2_large_asr_en/layer20/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e7d8369c68943a6da53abe282f24955fe9f7d077 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3fe5652a0b2333c8f0ebe4f6be0c816aee32b82a0751de9e05a833a0eb844cea +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer20/run4/model_best.pt b/results_tencent/w2v2_large_asr_en/layer20/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c5bf68a3a48b26b00a514f18af9e2cdc8698729e --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:461228a357255428c5cb5d8d2092452607696070ad76a964fabc3434d1454f00 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer20/run4/train.log b/results_tencent/w2v2_large_asr_en/layer20/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..51338fbe1518096cef911fe242ab848b19e1a1ab --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer20/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5419193053245545 +INFO:root: +[Valid][0][UTT][ MSE = 0.5423 | LCC = 0.8608 | SRCC = 0.8585 ] +INFO:root:Average loss=0.33697918224334716 +INFO:root: +[Valid][1][UTT][ MSE = 0.3423 | LCC = 0.8876 | SRCC = 0.8864 ] +INFO:root:Average loss=0.28328376889228823 +INFO:root: +[Valid][2][UTT][ MSE = 0.3927 | LCC = 0.8993 | SRCC = 0.9007 ] +INFO:root:Average loss=0.2589107985496521 +INFO:root: +[Valid][3][UTT][ MSE = 0.2986 | LCC = 0.9112 | SRCC = 0.9098 ] +INFO:root:Average loss=0.23122643423080444 +INFO:root: +[Valid][4][UTT][ MSE = 0.2800 | LCC = 0.9186 | SRCC = 0.9168 ] +INFO:root:Average loss=0.20708195036649704 +INFO:root: +[Valid][5][UTT][ MSE = 0.2753 | LCC = 0.9255 | SRCC = 0.9236 ] +INFO:root:Average loss=0.20166318565607072 +INFO:root: +[Valid][6][UTT][ MSE = 0.3024 | LCC = 0.9205 | SRCC = 0.9188 ] +INFO:root:Average loss=0.19140678173303605 +INFO:root: +[Valid][7][UTT][ MSE = 0.3325 | LCC = 0.9214 | SRCC = 0.9208 ] +INFO:root:Average loss=0.17162500363588334 +INFO:root: +[Valid][8][UTT][ MSE = 0.2562 | LCC = 0.9331 | SRCC = 0.9314 ] +INFO:root:Average loss=0.16744403046369552 +INFO:root: +[Valid][9][UTT][ MSE = 0.3277 | LCC = 0.9309 | SRCC = 0.9302 ] +INFO:root:Average loss=0.15973969453573228 +INFO:root: +[Valid][10][UTT][ MSE = 0.2663 | LCC = 0.9350 | SRCC = 0.9351 ] +INFO:root:Average loss=0.15098946517705916 +INFO:root: +[Valid][11][UTT][ MSE = 0.2685 | LCC = 0.9377 | SRCC = 0.9371 ] +INFO:root:Average loss=0.14883477675914764 +INFO:root: +[Valid][12][UTT][ MSE = 0.2502 | LCC = 0.9390 | SRCC = 0.9381 ] +INFO:root:Average loss=0.13945196884870528 +INFO:root: +[Valid][13][UTT][ MSE = 0.3094 | LCC = 0.9375 | SRCC = 0.9373 ] +INFO:root:Average loss=0.1308291465640068 +INFO:root: +[Valid][14][UTT][ MSE = 0.2305 | LCC = 0.9389 | SRCC = 0.9383 ] +INFO:root:Average loss=0.12584595078229904 +INFO:root: +[Valid][15][UTT][ MSE = 0.2823 | LCC = 0.9409 | SRCC = 0.9394 ] +INFO:root:Average loss=0.12010245007276535 +INFO:root: +[Valid][16][UTT][ MSE = 0.2516 | LCC = 0.9393 | SRCC = 0.9386 ] +INFO:root:Average loss=0.11554812204837799 +INFO:root: +[Valid][17][UTT][ MSE = 0.2164 | LCC = 0.9408 | SRCC = 0.9390 ] +INFO:root:Average loss=0.11108415016531945 +INFO:root: +[Valid][18][UTT][ MSE = 0.2095 | LCC = 0.9386 | SRCC = 0.9373 ] +INFO:root:Average loss=0.11036321365833282 +INFO:root: +[Valid][19][UTT][ MSE = 0.2243 | LCC = 0.9421 | SRCC = 0.9410 ] +INFO:root:Average loss=0.10317478519678115 +INFO:root: +[Valid][20][UTT][ MSE = 0.2381 | LCC = 0.9414 | SRCC = 0.9405 ] +INFO:root:Average loss=0.10167122545838356 +INFO:root: +[Valid][21][UTT][ MSE = 0.2128 | LCC = 0.9443 | SRCC = 0.9431 ] +INFO:root:Average loss=0.10094096013903618 +INFO:root: +[Valid][22][UTT][ MSE = 0.2031 | LCC = 0.9429 | SRCC = 0.9421 ] +INFO:root:Average loss=0.09427850982546807 +INFO:root: +[Valid][23][UTT][ MSE = 0.2076 | LCC = 0.9457 | SRCC = 0.9446 ] +INFO:root:Average loss=0.09470171630382537 +INFO:root: +[Valid][24][UTT][ MSE = 0.2544 | LCC = 0.9416 | SRCC = 0.9421 ] +INFO:root:Average loss=0.09058746084570884 +INFO:root: +[Valid][25][UTT][ MSE = 0.1933 | LCC = 0.9390 | SRCC = 0.9376 ] +INFO:root:Average loss=0.08354965156316757 +INFO:root: +[Valid][26][UTT][ MSE = 0.2332 | LCC = 0.9417 | SRCC = 0.9410 ] +INFO:root:Average loss=0.08319823387265206 +INFO:root: +[Valid][27][UTT][ MSE = 0.2291 | LCC = 0.9443 | SRCC = 0.9428 ] +INFO:root:Average loss=0.08096489143371582 +INFO:root: +[Valid][28][UTT][ MSE = 0.2287 | LCC = 0.9448 | SRCC = 0.9439 ] +INFO:root:Average loss=0.07893598031997681 +INFO:root: +[Valid][29][UTT][ MSE = 0.2447 | LCC = 0.9422 | SRCC = 0.9417 ] +INFO:root:Average loss=0.07452070236206054 +INFO:root: +[Valid][30][UTT][ MSE = 0.2150 | LCC = 0.9433 | SRCC = 0.9412 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2111 | LCC = 0.9435 | SRCC = 0.9435 ] diff --git a/results_tencent/w2v2_large_asr_en/layer22/run0/config.gin b/results_tencent/w2v2_large_asr_en/layer22/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer22/run0/model.pt b/results_tencent/w2v2_large_asr_en/layer22/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1f3bf6674e29b7a7e1ed83e120793f69f347d264 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1ed5afee09ec164bd2f95cf4190451430756bf2eb5efd86850a1f1d2e249e9e9 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer22/run0/model_best.pt b/results_tencent/w2v2_large_asr_en/layer22/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..5126be88e8c9b773d3a2309a79d95fcf826f56dc --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:86e7f29eaed13efd4b144d6192b42f1d3a6554cf4130978c64e30780f051f2f4 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer22/run0/train.log b/results_tencent/w2v2_large_asr_en/layer22/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..fda033856189eb9a74b0039e6921cc39fe056b34 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5897784457206726 +INFO:root: +[Valid][0][UTT][ MSE = 0.5391 | LCC = 0.8345 | SRCC = 0.8297 ] +INFO:root:Average loss=0.3781484271287918 +INFO:root: +[Valid][1][UTT][ MSE = 0.4468 | LCC = 0.8663 | SRCC = 0.8627 ] +INFO:root:Average loss=0.3418886270523071 +INFO:root: +[Valid][2][UTT][ MSE = 0.3796 | LCC = 0.8844 | SRCC = 0.8833 ] +INFO:root:Average loss=0.2951031597852707 +INFO:root: +[Valid][3][UTT][ MSE = 0.3921 | LCC = 0.8945 | SRCC = 0.8910 ] +INFO:root:Average loss=0.2799289700984955 +INFO:root: +[Valid][4][UTT][ MSE = 0.3610 | LCC = 0.9013 | SRCC = 0.8986 ] +INFO:root:Average loss=0.2461425881385803 +INFO:root: +[Valid][5][UTT][ MSE = 0.3452 | LCC = 0.9058 | SRCC = 0.9028 ] +INFO:root:Average loss=0.23520653855800627 +INFO:root: +[Valid][6][UTT][ MSE = 0.2973 | LCC = 0.9109 | SRCC = 0.9086 ] +INFO:root:Average loss=0.22713188815116883 +INFO:root: +[Valid][7][UTT][ MSE = 0.3221 | LCC = 0.9161 | SRCC = 0.9139 ] +INFO:root:Average loss=0.2186810907125473 +INFO:root: +[Valid][8][UTT][ MSE = 0.3596 | LCC = 0.9093 | SRCC = 0.9084 ] +INFO:root:Average loss=0.20723852449655533 +INFO:root: +[Valid][9][UTT][ MSE = 0.3826 | LCC = 0.9100 | SRCC = 0.9106 ] +INFO:root:Average loss=0.1909127761721611 +INFO:root: +[Valid][10][UTT][ MSE = 0.2692 | LCC = 0.9216 | SRCC = 0.9192 ] +INFO:root:Average loss=0.17775832867622376 +INFO:root: +[Valid][11][UTT][ MSE = 0.3333 | LCC = 0.9214 | SRCC = 0.9193 ] +INFO:root:Average loss=0.18024291449785232 +INFO:root: +[Valid][12][UTT][ MSE = 0.3103 | LCC = 0.9256 | SRCC = 0.9223 ] +INFO:root:Average loss=0.16286313331127167 +INFO:root: +[Valid][13][UTT][ MSE = 0.3054 | LCC = 0.9276 | SRCC = 0.9267 ] +INFO:root:Average loss=0.1565928136706352 +INFO:root: +[Valid][14][UTT][ MSE = 0.2416 | LCC = 0.9266 | SRCC = 0.9249 ] +INFO:root:Average loss=0.14793560898303987 +INFO:root: +[Valid][15][UTT][ MSE = 0.3021 | LCC = 0.9279 | SRCC = 0.9269 ] +INFO:root:Average loss=0.15011053258180618 +INFO:root: +[Valid][16][UTT][ MSE = 0.2475 | LCC = 0.9309 | SRCC = 0.9293 ] +INFO:root:Average loss=0.13999230098724366 +INFO:root: +[Valid][17][UTT][ MSE = 0.2682 | LCC = 0.9320 | SRCC = 0.9311 ] +INFO:root:Average loss=0.13385246235132217 +INFO:root: +[Valid][18][UTT][ MSE = 0.2373 | LCC = 0.9334 | SRCC = 0.9323 ] +INFO:root:Average loss=0.13556723701953888 +INFO:root: +[Valid][19][UTT][ MSE = 0.2479 | LCC = 0.9349 | SRCC = 0.9336 ] +INFO:root:Average loss=0.13290116786956788 +INFO:root: +[Valid][20][UTT][ MSE = 0.3094 | LCC = 0.9295 | SRCC = 0.9294 ] +INFO:root:Average loss=0.1286758942604065 +INFO:root: +[Valid][21][UTT][ MSE = 0.2859 | LCC = 0.9339 | SRCC = 0.9325 ] +INFO:root:Average loss=0.11760492032766343 +INFO:root: +[Valid][22][UTT][ MSE = 0.2835 | LCC = 0.9316 | SRCC = 0.9303 ] +INFO:root:Average loss=0.11317218169569969 +INFO:root: +[Valid][23][UTT][ MSE = 0.2746 | LCC = 0.9347 | SRCC = 0.9342 ] +INFO:root:Average loss=0.11258434686064721 +INFO:root: +[Valid][24][UTT][ MSE = 0.2424 | LCC = 0.9352 | SRCC = 0.9344 ] +INFO:root:Average loss=0.11277378112077713 +INFO:root: +[Valid][25][UTT][ MSE = 0.2343 | LCC = 0.9339 | SRCC = 0.9339 ] +INFO:root:Average loss=0.1058045932650566 +INFO:root: +[Valid][26][UTT][ MSE = 0.2541 | LCC = 0.9370 | SRCC = 0.9359 ] +INFO:root:Average loss=0.11175987100601197 +INFO:root: +[Valid][27][UTT][ MSE = 0.2537 | LCC = 0.9375 | SRCC = 0.9360 ] +INFO:root:Average loss=0.099593052983284 +INFO:root: +[Valid][28][UTT][ MSE = 0.2250 | LCC = 0.9365 | SRCC = 0.9358 ] +INFO:root:Average loss=0.10468415582180023 +INFO:root: +[Valid][29][UTT][ MSE = 0.2759 | LCC = 0.9368 | SRCC = 0.9351 ] +INFO:root:Average loss=0.09926666769385338 +INFO:root: +[Valid][30][UTT][ MSE = 0.2677 | LCC = 0.9380 | SRCC = 0.9370 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2663 | LCC = 0.9380 | SRCC = 0.9379 ] diff --git a/results_tencent/w2v2_large_asr_en/layer22/run1/config.gin b/results_tencent/w2v2_large_asr_en/layer22/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer22/run1/model.pt b/results_tencent/w2v2_large_asr_en/layer22/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..87bde019c7a52b44fea9e90589be24a5e8b567b6 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0de0d9925356753024268a4dd88af4f13d18c5d254522653023ebc53e501797 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer22/run1/model_best.pt b/results_tencent/w2v2_large_asr_en/layer22/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..fd3c31194aa4b086cceee73f7baa20b97f33dd9f --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f700c2324cf0fbf962b5293ebd80762a6c1a16352238451dc07690e92e15382a +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer22/run1/train.log b/results_tencent/w2v2_large_asr_en/layer22/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..4039b2ac6b892fc1860d6c1ef68c29c0270d5c3a --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5636141817569733 +INFO:root: +[Valid][0][UTT][ MSE = 0.4810 | LCC = 0.8528 | SRCC = 0.8467 ] +INFO:root:Average loss=0.3525434806346893 +INFO:root: +[Valid][1][UTT][ MSE = 0.4094 | LCC = 0.8780 | SRCC = 0.8769 ] +INFO:root:Average loss=0.3120790466070175 +INFO:root: +[Valid][2][UTT][ MSE = 0.3873 | LCC = 0.8912 | SRCC = 0.8897 ] +INFO:root:Average loss=0.28926258683204653 +INFO:root: +[Valid][3][UTT][ MSE = 0.3380 | LCC = 0.8982 | SRCC = 0.8963 ] +INFO:root:Average loss=0.28702483999729156 +INFO:root: +[Valid][4][UTT][ MSE = 0.3723 | LCC = 0.9040 | SRCC = 0.9015 ] +INFO:root:Average loss=0.2542877168655395 +INFO:root: +[Valid][5][UTT][ MSE = 0.3639 | LCC = 0.9032 | SRCC = 0.9012 ] +INFO:root:Average loss=0.23932238882780074 +INFO:root: +[Valid][6][UTT][ MSE = 0.3470 | LCC = 0.9171 | SRCC = 0.9145 ] +INFO:root:Average loss=0.23754289090633393 +INFO:root: +[Valid][7][UTT][ MSE = 0.4409 | LCC = 0.9072 | SRCC = 0.9121 ] +INFO:root:Average loss=0.20829445093870164 +INFO:root: +[Valid][8][UTT][ MSE = 0.3493 | LCC = 0.9226 | SRCC = 0.9202 ] +INFO:root:Average loss=0.20296432161331177 +INFO:root: +[Valid][9][UTT][ MSE = 0.3370 | LCC = 0.9174 | SRCC = 0.9159 ] +INFO:root:Average loss=0.19190106600522994 +INFO:root: +[Valid][10][UTT][ MSE = 0.2978 | LCC = 0.9196 | SRCC = 0.9172 ] +INFO:root:Average loss=0.17637180823087692 +INFO:root: +[Valid][11][UTT][ MSE = 0.2944 | LCC = 0.9277 | SRCC = 0.9253 ] +INFO:root:Average loss=0.16775718069076537 +INFO:root: +[Valid][12][UTT][ MSE = 0.2308 | LCC = 0.9319 | SRCC = 0.9302 ] +INFO:root:Average loss=0.15804749399423598 +INFO:root: +[Valid][13][UTT][ MSE = 0.3158 | LCC = 0.9316 | SRCC = 0.9303 ] +INFO:root:Average loss=0.16334939861297607 +INFO:root: +[Valid][14][UTT][ MSE = 0.3174 | LCC = 0.9307 | SRCC = 0.9302 ] +INFO:root:Average loss=0.15430401158332824 +INFO:root: +[Valid][15][UTT][ MSE = 0.2672 | LCC = 0.9313 | SRCC = 0.9293 ] +INFO:root:Average loss=0.14568600630760192 +INFO:root: +[Valid][16][UTT][ MSE = 0.2665 | LCC = 0.9342 | SRCC = 0.9333 ] +INFO:root:Average loss=0.14912316519021987 +INFO:root: +[Valid][17][UTT][ MSE = 0.2698 | LCC = 0.9339 | SRCC = 0.9328 ] +INFO:root:Average loss=0.15149657678604125 +INFO:root: +[Valid][18][UTT][ MSE = 0.2831 | LCC = 0.9337 | SRCC = 0.9326 ] +INFO:root:Average loss=0.13359137481451033 +INFO:root: +[Valid][19][UTT][ MSE = 0.2697 | LCC = 0.9318 | SRCC = 0.9300 ] +INFO:root:Average loss=0.1291338313817978 +INFO:root: +[Valid][20][UTT][ MSE = 0.2599 | LCC = 0.9352 | SRCC = 0.9340 ] +INFO:root:Average loss=0.12270363795757294 +INFO:root: +[Valid][21][UTT][ MSE = 0.2973 | LCC = 0.9340 | SRCC = 0.9336 ] +INFO:root:Average loss=0.1333803173303604 +INFO:root: +[Valid][22][UTT][ MSE = 0.2526 | LCC = 0.9366 | SRCC = 0.9346 ] +INFO:root:Average loss=0.11533284217119216 +INFO:root: +[Valid][23][UTT][ MSE = 0.2603 | LCC = 0.9379 | SRCC = 0.9366 ] +INFO:root:Average loss=0.11870836952328682 +INFO:root: +[Valid][24][UTT][ MSE = 0.2531 | LCC = 0.9377 | SRCC = 0.9365 ] +INFO:root:Average loss=0.11011113935709 +INFO:root: +[Valid][25][UTT][ MSE = 0.2259 | LCC = 0.9341 | SRCC = 0.9314 ] +INFO:root:Average loss=0.10318287140130997 +INFO:root: +[Valid][26][UTT][ MSE = 0.2385 | LCC = 0.9385 | SRCC = 0.9366 ] +INFO:root:Average loss=0.10209246149659157 +INFO:root: +[Valid][27][UTT][ MSE = 0.2346 | LCC = 0.9357 | SRCC = 0.9329 ] +INFO:root:Average loss=0.10068185597658158 +INFO:root: +[Valid][28][UTT][ MSE = 0.2467 | LCC = 0.9362 | SRCC = 0.9336 ] +INFO:root:Average loss=0.09796376246213913 +INFO:root: +[Valid][29][UTT][ MSE = 0.2442 | LCC = 0.9401 | SRCC = 0.9380 ] +INFO:root:Average loss=0.0945982860326767 +INFO:root: +[Valid][30][UTT][ MSE = 0.2739 | LCC = 0.9385 | SRCC = 0.9379 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2563 | LCC = 0.9351 | SRCC = 0.9347 ] diff --git a/results_tencent/w2v2_large_asr_en/layer22/run2/config.gin b/results_tencent/w2v2_large_asr_en/layer22/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer22/run2/model.pt b/results_tencent/w2v2_large_asr_en/layer22/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..701bd6e21ae4a382341036ae443f4bea6de7b9ed --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a4d35e8e157646ed6acf87b6e7c04a767e86d24978eda1ca729e0e003b3bba99 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer22/run2/model_best.pt b/results_tencent/w2v2_large_asr_en/layer22/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..fca686c2b5d2a9031f92004e210037a61c2dcd33 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:abec2df0105b268d56c064099e89a752e063d8f7722d10c8392d94f679a0aeda +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer22/run2/train.log b/results_tencent/w2v2_large_asr_en/layer22/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..37f44bd01a4bdb657362a1823a0a5765afa7d7e6 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5972103016376495 +INFO:root: +[Valid][0][UTT][ MSE = 0.4432 | LCC = 0.8449 | SRCC = 0.8410 ] +INFO:root:Average loss=0.38135843122005464 +INFO:root: +[Valid][1][UTT][ MSE = 0.4626 | LCC = 0.8770 | SRCC = 0.8746 ] +INFO:root:Average loss=0.32819598948955536 +INFO:root: +[Valid][2][UTT][ MSE = 0.4235 | LCC = 0.8860 | SRCC = 0.8874 ] +INFO:root:Average loss=0.29219688594341275 +INFO:root: +[Valid][3][UTT][ MSE = 0.3630 | LCC = 0.8984 | SRCC = 0.8971 ] +INFO:root:Average loss=0.2655452103614807 +INFO:root: +[Valid][4][UTT][ MSE = 0.4143 | LCC = 0.9041 | SRCC = 0.9023 ] +INFO:root:Average loss=0.24415083211660385 +INFO:root: +[Valid][5][UTT][ MSE = 0.3411 | LCC = 0.9139 | SRCC = 0.9114 ] +INFO:root:Average loss=0.2350845286846161 +INFO:root: +[Valid][6][UTT][ MSE = 0.3438 | LCC = 0.9170 | SRCC = 0.9140 ] +INFO:root:Average loss=0.21783815705776213 +INFO:root: +[Valid][7][UTT][ MSE = 0.2883 | LCC = 0.9222 | SRCC = 0.9194 ] +INFO:root:Average loss=0.20829196321964263 +INFO:root: +[Valid][8][UTT][ MSE = 0.2518 | LCC = 0.9235 | SRCC = 0.9218 ] +INFO:root:Average loss=0.19831743532419205 +INFO:root: +[Valid][9][UTT][ MSE = 0.3334 | LCC = 0.9202 | SRCC = 0.9216 ] +INFO:root:Average loss=0.19493862271308898 +INFO:root: +[Valid][10][UTT][ MSE = 0.2706 | LCC = 0.9274 | SRCC = 0.9246 ] +INFO:root:Average loss=0.18110518580675125 +INFO:root: +[Valid][11][UTT][ MSE = 0.3084 | LCC = 0.9279 | SRCC = 0.9267 ] +INFO:root:Average loss=0.1626242477297783 +INFO:root: +[Valid][12][UTT][ MSE = 0.3018 | LCC = 0.9310 | SRCC = 0.9301 ] +INFO:root:Average loss=0.15853124541044236 +INFO:root: +[Valid][13][UTT][ MSE = 0.2897 | LCC = 0.9305 | SRCC = 0.9297 ] +INFO:root:Average loss=0.16345707684755326 +INFO:root: +[Valid][14][UTT][ MSE = 0.2522 | LCC = 0.9329 | SRCC = 0.9312 ] +INFO:root:Average loss=0.14529302048683165 +INFO:root: +[Valid][15][UTT][ MSE = 0.2395 | LCC = 0.9346 | SRCC = 0.9327 ] +INFO:root:Average loss=0.14286678257584573 +INFO:root: +[Valid][16][UTT][ MSE = 0.2681 | LCC = 0.9336 | SRCC = 0.9325 ] +INFO:root:Average loss=0.13916210317611694 +INFO:root: +[Valid][17][UTT][ MSE = 0.3124 | LCC = 0.9329 | SRCC = 0.9338 ] +INFO:root:Average loss=0.13229643988609313 +INFO:root: +[Valid][18][UTT][ MSE = 0.2459 | LCC = 0.9354 | SRCC = 0.9349 ] +INFO:root:Average loss=0.12886137121915817 +INFO:root: +[Valid][19][UTT][ MSE = 0.2740 | LCC = 0.9372 | SRCC = 0.9359 ] +INFO:root:Average loss=0.13239783662557603 +INFO:root: +[Valid][20][UTT][ MSE = 0.2566 | LCC = 0.9357 | SRCC = 0.9343 ] +INFO:root:Average loss=0.12378757873177529 +INFO:root: +[Valid][21][UTT][ MSE = 0.3094 | LCC = 0.9358 | SRCC = 0.9356 ] +INFO:root:Average loss=0.11822506153583527 +INFO:root: +[Valid][22][UTT][ MSE = 0.2564 | LCC = 0.9385 | SRCC = 0.9377 ] +INFO:root:Average loss=0.11610961270332336 +INFO:root: +[Valid][23][UTT][ MSE = 0.2908 | LCC = 0.9363 | SRCC = 0.9360 ] +INFO:root:Average loss=0.10929083874821663 +INFO:root: +[Valid][24][UTT][ MSE = 0.2527 | LCC = 0.9377 | SRCC = 0.9365 ] +INFO:root:Average loss=0.10222520646452904 +INFO:root: +[Valid][25][UTT][ MSE = 0.2551 | LCC = 0.9397 | SRCC = 0.9388 ] +INFO:root:Average loss=0.10175182008743286 +INFO:root: +[Valid][26][UTT][ MSE = 0.2206 | LCC = 0.9365 | SRCC = 0.9352 ] +INFO:root:Average loss=0.0991397641003132 +INFO:root: +[Valid][27][UTT][ MSE = 0.2698 | LCC = 0.9374 | SRCC = 0.9368 ] +INFO:root:Average loss=0.0992351357638836 +INFO:root: +[Valid][28][UTT][ MSE = 0.2820 | LCC = 0.9395 | SRCC = 0.9386 ] +INFO:root:Average loss=0.09569387155771256 +INFO:root: +[Valid][29][UTT][ MSE = 0.2860 | LCC = 0.9382 | SRCC = 0.9378 ] +INFO:root:Average loss=0.09200032013654709 +INFO:root: +[Valid][30][UTT][ MSE = 0.2598 | LCC = 0.9393 | SRCC = 0.9382 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2594 | LCC = 0.9385 | SRCC = 0.9384 ] diff --git a/results_tencent/w2v2_large_asr_en/layer22/run3/config.gin b/results_tencent/w2v2_large_asr_en/layer22/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer22/run3/model.pt b/results_tencent/w2v2_large_asr_en/layer22/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9fd516ea0744d85e49fe6b5663059967d6589302 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c306e7686d3e90bd44d8dfc6319e118efd06ec432cbabadcaead1b74f3090638 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer22/run3/model_best.pt b/results_tencent/w2v2_large_asr_en/layer22/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b9e3e1034d105b3aeaf522acb76fd249fca4d1db --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:26e4d289edc03c5345e273920adbc3500aa49a96d6b41c115515572ad1894a81 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer22/run3/train.log b/results_tencent/w2v2_large_asr_en/layer22/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b46cfffd080c26fbe541d7fd0d17e84b82bd132e --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5355237860679627 +INFO:root: +[Valid][0][UTT][ MSE = 0.4737 | LCC = 0.8519 | SRCC = 0.8449 ] +INFO:root:Average loss=0.3520196957588196 +INFO:root: +[Valid][1][UTT][ MSE = 0.4217 | LCC = 0.8753 | SRCC = 0.8743 ] +INFO:root:Average loss=0.33181757426261904 +INFO:root: +[Valid][2][UTT][ MSE = 0.4001 | LCC = 0.8866 | SRCC = 0.8863 ] +INFO:root:Average loss=0.2830224075317383 +INFO:root: +[Valid][3][UTT][ MSE = 0.3940 | LCC = 0.8933 | SRCC = 0.8918 ] +INFO:root:Average loss=0.2656959124207497 +INFO:root: +[Valid][4][UTT][ MSE = 0.3553 | LCC = 0.9019 | SRCC = 0.9008 ] +INFO:root:Average loss=0.24916751372814178 +INFO:root: +[Valid][5][UTT][ MSE = 0.3355 | LCC = 0.9056 | SRCC = 0.9045 ] +INFO:root:Average loss=0.24123628318309784 +INFO:root: +[Valid][6][UTT][ MSE = 0.3523 | LCC = 0.9047 | SRCC = 0.9064 ] +INFO:root:Average loss=0.23377276062965394 +INFO:root: +[Valid][7][UTT][ MSE = 0.3476 | LCC = 0.9096 | SRCC = 0.9080 ] +INFO:root:Average loss=0.21065986955165864 +INFO:root: +[Valid][8][UTT][ MSE = 0.3485 | LCC = 0.9158 | SRCC = 0.9146 ] +INFO:root:Average loss=0.19938995331525802 +INFO:root: +[Valid][9][UTT][ MSE = 0.3011 | LCC = 0.9145 | SRCC = 0.9138 ] +INFO:root:Average loss=0.19274595504999162 +INFO:root: +[Valid][10][UTT][ MSE = 0.3177 | LCC = 0.9171 | SRCC = 0.9165 ] +INFO:root:Average loss=0.18669966411590577 +INFO:root: +[Valid][11][UTT][ MSE = 0.3120 | LCC = 0.9215 | SRCC = 0.9200 ] +INFO:root:Average loss=0.17444315534830093 +INFO:root: +[Valid][12][UTT][ MSE = 0.3121 | LCC = 0.9237 | SRCC = 0.9216 ] +INFO:root:Average loss=0.17254026353359223 +INFO:root: +[Valid][13][UTT][ MSE = 0.2739 | LCC = 0.9231 | SRCC = 0.9207 ] +INFO:root:Average loss=0.17114194631576538 +INFO:root: +[Valid][14][UTT][ MSE = 0.2736 | LCC = 0.9276 | SRCC = 0.9265 ] +INFO:root:Average loss=0.1542787786126137 +INFO:root: +[Valid][15][UTT][ MSE = 0.2849 | LCC = 0.9298 | SRCC = 0.9294 ] +INFO:root:Average loss=0.14990125453472136 +INFO:root: +[Valid][16][UTT][ MSE = 0.2476 | LCC = 0.9297 | SRCC = 0.9281 ] +INFO:root:Average loss=0.1539065936803818 +INFO:root: +[Valid][17][UTT][ MSE = 0.3311 | LCC = 0.9286 | SRCC = 0.9280 ] +INFO:root:Average loss=0.14429236876964568 +INFO:root: +[Valid][18][UTT][ MSE = 0.2925 | LCC = 0.9299 | SRCC = 0.9296 ] +INFO:root:Average loss=0.13910675573349 +INFO:root: +[Valid][19][UTT][ MSE = 0.2447 | LCC = 0.9329 | SRCC = 0.9319 ] +INFO:root:Average loss=0.13263670057058335 +INFO:root: +[Valid][20][UTT][ MSE = 0.3572 | LCC = 0.9267 | SRCC = 0.9282 ] +INFO:root:Average loss=0.1253247890472412 +INFO:root: +[Valid][21][UTT][ MSE = 0.2339 | LCC = 0.9334 | SRCC = 0.9330 ] +INFO:root:Average loss=0.12017880952358245 +INFO:root: +[Valid][22][UTT][ MSE = 0.2293 | LCC = 0.9322 | SRCC = 0.9320 ] +INFO:root:Average loss=0.11371466875076294 +INFO:root: +[Valid][23][UTT][ MSE = 0.2535 | LCC = 0.9353 | SRCC = 0.9344 ] +INFO:root:Average loss=0.1177149579524994 +INFO:root: +[Valid][24][UTT][ MSE = 0.2294 | LCC = 0.9370 | SRCC = 0.9356 ] +INFO:root:Average loss=0.11350199592113495 +INFO:root: +[Valid][25][UTT][ MSE = 0.2851 | LCC = 0.9315 | SRCC = 0.9313 ] +INFO:root:Average loss=0.11047359806299209 +INFO:root: +[Valid][26][UTT][ MSE = 0.2290 | LCC = 0.9346 | SRCC = 0.9353 ] +INFO:root:Average loss=0.11222968560457229 +INFO:root: +[Valid][27][UTT][ MSE = 0.2617 | LCC = 0.9350 | SRCC = 0.9346 ] +INFO:root:Average loss=0.1037919754087925 +INFO:root: +[Valid][28][UTT][ MSE = 0.2377 | LCC = 0.9327 | SRCC = 0.9328 ] +INFO:root:Average loss=0.102411281645298 +INFO:root: +[Valid][29][UTT][ MSE = 0.2036 | LCC = 0.9360 | SRCC = 0.9349 ] +INFO:root:Average loss=0.09455926218628884 +INFO:root: +[Valid][30][UTT][ MSE = 0.2192 | LCC = 0.9360 | SRCC = 0.9342 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2341 | LCC = 0.9353 | SRCC = 0.9342 ] diff --git a/results_tencent/w2v2_large_asr_en/layer22/run4/config.gin b/results_tencent/w2v2_large_asr_en/layer22/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer22/run4/model.pt b/results_tencent/w2v2_large_asr_en/layer22/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..aba5ab1c4b3626b745a797240483f87c5c205e14 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39831ef75116ef26a6d2f5e0a51268502b9c2377d429b5c27ca16a5d2dd61292 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer22/run4/model_best.pt b/results_tencent/w2v2_large_asr_en/layer22/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c695d23a9dab529b03e7bac95e54f676e108b03b --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:466e11e5ca4515753309db1b7188a51d3032e3676e8ad10e2d0f73ea5469d489 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer22/run4/train.log b/results_tencent/w2v2_large_asr_en/layer22/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1b82e7140557eadbdf0240d5acf3e68ee585b73c --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer22/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5661868907213211 +INFO:root: +[Valid][0][UTT][ MSE = 0.4986 | LCC = 0.8494 | SRCC = 0.8426 ] +INFO:root:Average loss=0.3582417205572128 +INFO:root: +[Valid][1][UTT][ MSE = 0.3893 | LCC = 0.8803 | SRCC = 0.8772 ] +INFO:root:Average loss=0.3286102602481842 +INFO:root: +[Valid][2][UTT][ MSE = 0.4145 | LCC = 0.8926 | SRCC = 0.8896 ] +INFO:root:Average loss=0.2858977131843567 +INFO:root: +[Valid][3][UTT][ MSE = 0.3763 | LCC = 0.9003 | SRCC = 0.8993 ] +INFO:root:Average loss=0.25885898238420485 +INFO:root: +[Valid][4][UTT][ MSE = 0.3333 | LCC = 0.9083 | SRCC = 0.9064 ] +INFO:root:Average loss=0.24351658415794372 +INFO:root: +[Valid][5][UTT][ MSE = 0.2998 | LCC = 0.9117 | SRCC = 0.9099 ] +INFO:root:Average loss=0.26245529341697693 +INFO:root: +[Valid][6][UTT][ MSE = 0.3879 | LCC = 0.9048 | SRCC = 0.9072 ] +INFO:root:Average loss=0.22474016177654266 +INFO:root: +[Valid][7][UTT][ MSE = 0.3527 | LCC = 0.9125 | SRCC = 0.9115 ] +INFO:root:Average loss=0.21396267783641815 +INFO:root: +[Valid][8][UTT][ MSE = 0.3371 | LCC = 0.9108 | SRCC = 0.9108 ] +INFO:root:Average loss=0.20578602755069733 +INFO:root: +[Valid][9][UTT][ MSE = 0.2840 | LCC = 0.9213 | SRCC = 0.9193 ] +INFO:root:Average loss=0.1909430514574051 +INFO:root: +[Valid][10][UTT][ MSE = 0.3747 | LCC = 0.9211 | SRCC = 0.9202 ] +INFO:root:Average loss=0.18498300433158874 +INFO:root: +[Valid][11][UTT][ MSE = 0.3081 | LCC = 0.9278 | SRCC = 0.9261 ] +INFO:root:Average loss=0.1770166419148445 +INFO:root: +[Valid][12][UTT][ MSE = 0.2904 | LCC = 0.9252 | SRCC = 0.9236 ] +INFO:root:Average loss=0.16475183916091918 +INFO:root: +[Valid][13][UTT][ MSE = 0.3136 | LCC = 0.9291 | SRCC = 0.9284 ] +INFO:root:Average loss=0.1540866850614548 +INFO:root: +[Valid][14][UTT][ MSE = 0.3092 | LCC = 0.9271 | SRCC = 0.9266 ] +INFO:root:Average loss=0.1558628081679344 +INFO:root: +[Valid][15][UTT][ MSE = 0.2719 | LCC = 0.9320 | SRCC = 0.9304 ] +INFO:root:Average loss=0.14496686369180678 +INFO:root: +[Valid][16][UTT][ MSE = 0.2577 | LCC = 0.9336 | SRCC = 0.9310 ] +INFO:root:Average loss=0.1425081650018692 +INFO:root: +[Valid][17][UTT][ MSE = 0.3165 | LCC = 0.9260 | SRCC = 0.9269 ] +INFO:root:Average loss=0.13353432619571687 +INFO:root: +[Valid][18][UTT][ MSE = 0.2763 | LCC = 0.9337 | SRCC = 0.9329 ] +INFO:root:Average loss=0.13360411828756333 +INFO:root: +[Valid][19][UTT][ MSE = 0.2576 | LCC = 0.9320 | SRCC = 0.9313 ] +INFO:root:Average loss=0.12478964805603027 +INFO:root: +[Valid][20][UTT][ MSE = 0.2972 | LCC = 0.9316 | SRCC = 0.9309 ] +INFO:root:Average loss=0.12268084764480591 +INFO:root: +[Valid][21][UTT][ MSE = 0.2851 | LCC = 0.9354 | SRCC = 0.9345 ] +INFO:root:Average loss=0.11843598854541779 +INFO:root: +[Valid][22][UTT][ MSE = 0.2227 | LCC = 0.9353 | SRCC = 0.9343 ] +INFO:root:Average loss=0.11258886539936065 +INFO:root: +[Valid][23][UTT][ MSE = 0.2567 | LCC = 0.9355 | SRCC = 0.9343 ] +INFO:root:Average loss=0.1114313525557518 +INFO:root: +[Valid][24][UTT][ MSE = 0.2929 | LCC = 0.9311 | SRCC = 0.9302 ] +INFO:root:Average loss=0.11002680915594101 +INFO:root: +[Valid][25][UTT][ MSE = 0.2662 | LCC = 0.9353 | SRCC = 0.9341 ] +INFO:root:Average loss=0.10526309671998024 +INFO:root: +[Valid][26][UTT][ MSE = 0.2698 | LCC = 0.9354 | SRCC = 0.9339 ] +INFO:root:Average loss=0.11212263989448547 +INFO:root: +[Valid][27][UTT][ MSE = 0.2739 | LCC = 0.9374 | SRCC = 0.9360 ] +INFO:root:Average loss=0.09882259565591812 +INFO:root: +[Valid][28][UTT][ MSE = 0.2332 | LCC = 0.9377 | SRCC = 0.9357 ] +INFO:root:Average loss=0.09759674626588821 +INFO:root: +[Valid][29][UTT][ MSE = 0.2361 | LCC = 0.9375 | SRCC = 0.9357 ] +INFO:root:Average loss=0.09214141401648522 +INFO:root: +[Valid][30][UTT][ MSE = 0.2709 | LCC = 0.9318 | SRCC = 0.9306 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2349 | LCC = 0.9358 | SRCC = 0.9350 ] diff --git a/results_tencent/w2v2_large_asr_en/layer4/run0/config.gin b/results_tencent/w2v2_large_asr_en/layer4/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer4/run0/model.pt b/results_tencent/w2v2_large_asr_en/layer4/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c54591c720b18e7d1bc3cee6ab717a183acb00fe --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c11e78b53f3e871cff12100ec9cdee6d1676f21bc5955a1258fc9c214a048161 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer4/run0/model_best.pt b/results_tencent/w2v2_large_asr_en/layer4/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..5f4e7e6ad881e39a981bbef83e84971a5a280b60 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b426707508e0c01d159ca9e4709cb939fab4ff1d71d1502199025f993c6bf9a7 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer4/run0/train.log b/results_tencent/w2v2_large_asr_en/layer4/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..05f505a0cd7d27e984d760525997305334d759e1 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.38543831753730773 +INFO:root: +[Valid][0][UTT][ MSE = 0.2859 | LCC = 0.9302 | SRCC = 0.9295 ] +INFO:root:Average loss=0.18594727826118468 +INFO:root: +[Valid][1][UTT][ MSE = 0.2437 | LCC = 0.9460 | SRCC = 0.9463 ] +INFO:root:Average loss=0.1473394020795822 +INFO:root: +[Valid][2][UTT][ MSE = 0.2101 | LCC = 0.9520 | SRCC = 0.9533 ] +INFO:root:Average loss=0.13442426055669784 +INFO:root: +[Valid][3][UTT][ MSE = 0.1982 | LCC = 0.9578 | SRCC = 0.9581 ] +INFO:root:Average loss=0.1170977224111557 +INFO:root: +[Valid][4][UTT][ MSE = 0.2009 | LCC = 0.9588 | SRCC = 0.9598 ] +INFO:root:Average loss=0.10732782822847366 +INFO:root: +[Valid][5][UTT][ MSE = 0.2378 | LCC = 0.9574 | SRCC = 0.9605 ] +INFO:root:Average loss=0.10041946876049042 +INFO:root: +[Valid][6][UTT][ MSE = 0.1771 | LCC = 0.9607 | SRCC = 0.9610 ] +INFO:root:Average loss=0.09066307288408279 +INFO:root: +[Valid][7][UTT][ MSE = 0.1808 | LCC = 0.9618 | SRCC = 0.9618 ] +INFO:root:Average loss=0.0853277530670166 +INFO:root: +[Valid][8][UTT][ MSE = 0.1885 | LCC = 0.9648 | SRCC = 0.9652 ] +INFO:root:Average loss=0.0809361031651497 +INFO:root: +[Valid][9][UTT][ MSE = 0.2287 | LCC = 0.9647 | SRCC = 0.9645 ] +INFO:root:Average loss=0.07680627897381782 +INFO:root: +[Valid][10][UTT][ MSE = 0.1790 | LCC = 0.9650 | SRCC = 0.9650 ] +INFO:root:Average loss=0.07549533191323281 +INFO:root: +[Valid][11][UTT][ MSE = 0.2020 | LCC = 0.9650 | SRCC = 0.9659 ] +INFO:root:Average loss=0.06750786671042443 +INFO:root: +[Valid][12][UTT][ MSE = 0.2099 | LCC = 0.9639 | SRCC = 0.9656 ] +INFO:root:Average loss=0.06278260758519173 +INFO:root: +[Valid][13][UTT][ MSE = 0.2044 | LCC = 0.9637 | SRCC = 0.9649 ] +INFO:root:Average loss=0.06384580457210541 +INFO:root: +[Valid][14][UTT][ MSE = 0.1912 | LCC = 0.9650 | SRCC = 0.9660 ] +INFO:root:Average loss=0.05696350744366646 +INFO:root: +[Valid][15][UTT][ MSE = 0.1747 | LCC = 0.9662 | SRCC = 0.9660 ] +INFO:root:Average loss=0.05214034582674503 +INFO:root: +[Valid][16][UTT][ MSE = 0.1747 | LCC = 0.9656 | SRCC = 0.9661 ] +INFO:root:Average loss=0.05210281509160995 +INFO:root: +[Valid][17][UTT][ MSE = 0.1677 | LCC = 0.9660 | SRCC = 0.9662 ] +INFO:root:Average loss=0.052911043286323546 +INFO:root: +[Valid][18][UTT][ MSE = 0.1577 | LCC = 0.9662 | SRCC = 0.9662 ] +INFO:root:Average loss=0.04813958428800106 +INFO:root: +[Valid][19][UTT][ MSE = 0.1452 | LCC = 0.9668 | SRCC = 0.9667 ] +INFO:root:Average loss=0.04627388000488281 +INFO:root: +[Valid][20][UTT][ MSE = 0.1774 | LCC = 0.9657 | SRCC = 0.9663 ] +INFO:root:Average loss=0.04289150613546371 +INFO:root: +[Valid][21][UTT][ MSE = 0.1748 | LCC = 0.9670 | SRCC = 0.9671 ] +INFO:root:Average loss=0.03914232270419598 +INFO:root: +[Valid][22][UTT][ MSE = 0.1835 | LCC = 0.9653 | SRCC = 0.9658 ] +INFO:root:Average loss=0.039453628838062284 +INFO:root: +[Valid][23][UTT][ MSE = 0.1885 | LCC = 0.9656 | SRCC = 0.9665 ] +INFO:root:Average loss=0.03746335950493813 +INFO:root: +[Valid][24][UTT][ MSE = 0.1783 | LCC = 0.9665 | SRCC = 0.9670 ] +INFO:root:Average loss=0.03476564422249794 +INFO:root: +[Valid][25][UTT][ MSE = 0.1542 | LCC = 0.9675 | SRCC = 0.9673 ] +INFO:root:Average loss=0.03711455811560154 +INFO:root: +[Valid][26][UTT][ MSE = 0.1628 | LCC = 0.9662 | SRCC = 0.9666 ] +INFO:root:Average loss=0.0333039049655199 +INFO:root: +[Valid][27][UTT][ MSE = 0.1849 | LCC = 0.9662 | SRCC = 0.9667 ] +INFO:root:Average loss=0.03292758139967918 +INFO:root: +[Valid][28][UTT][ MSE = 0.1573 | LCC = 0.9675 | SRCC = 0.9674 ] +INFO:root:Average loss=0.030571515962481498 +INFO:root: +[Valid][29][UTT][ MSE = 0.1677 | LCC = 0.9670 | SRCC = 0.9671 ] +INFO:root:Average loss=0.02844385711848736 +INFO:root: +[Valid][30][UTT][ MSE = 0.1617 | LCC = 0.9668 | SRCC = 0.9668 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1685 | LCC = 0.9638 | SRCC = 0.9641 ] diff --git a/results_tencent/w2v2_large_asr_en/layer4/run1/config.gin b/results_tencent/w2v2_large_asr_en/layer4/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer4/run1/model.pt b/results_tencent/w2v2_large_asr_en/layer4/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6fc174c3c57e901e16bddc2ecb7acf74e3f72a4a --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1452116dbf3e018d092d1ce19c9d3aa29b0b80a72e34284794495fc707878a5d +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer4/run1/model_best.pt b/results_tencent/w2v2_large_asr_en/layer4/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f9a50c7176428fbfa0c7b3f7fec1c438262b6d7c --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4de8d77dc7660c176e91adfeafcd5992ab411f8e6daa0adca280c23b03d284a5 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer4/run1/train.log b/results_tencent/w2v2_large_asr_en/layer4/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e7e10a0a23615d24d5f5cb6f75dc9e482e25b765 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4221428810358048 +INFO:root: +[Valid][0][UTT][ MSE = 0.3162 | LCC = 0.9266 | SRCC = 0.9269 ] +INFO:root:Average loss=0.18647263097763062 +INFO:root: +[Valid][1][UTT][ MSE = 0.2441 | LCC = 0.9463 | SRCC = 0.9474 ] +INFO:root:Average loss=0.15019476032257081 +INFO:root: +[Valid][2][UTT][ MSE = 0.2205 | LCC = 0.9513 | SRCC = 0.9526 ] +INFO:root:Average loss=0.13305391591787338 +INFO:root: +[Valid][3][UTT][ MSE = 0.2446 | LCC = 0.9553 | SRCC = 0.9561 ] +INFO:root:Average loss=0.12185597741603851 +INFO:root: +[Valid][4][UTT][ MSE = 0.1843 | LCC = 0.9591 | SRCC = 0.9597 ] +INFO:root:Average loss=0.10876754301786423 +INFO:root: +[Valid][5][UTT][ MSE = 0.2320 | LCC = 0.9585 | SRCC = 0.9603 ] +INFO:root:Average loss=0.09892539513111115 +INFO:root: +[Valid][6][UTT][ MSE = 0.2229 | LCC = 0.9563 | SRCC = 0.9597 ] +INFO:root:Average loss=0.0941687039732933 +INFO:root: +[Valid][7][UTT][ MSE = 0.1755 | LCC = 0.9620 | SRCC = 0.9626 ] +INFO:root:Average loss=0.08669672784209251 +INFO:root: +[Valid][8][UTT][ MSE = 0.1715 | LCC = 0.9634 | SRCC = 0.9635 ] +INFO:root:Average loss=0.08094768327474594 +INFO:root: +[Valid][9][UTT][ MSE = 0.2163 | LCC = 0.9609 | SRCC = 0.9628 ] +INFO:root:Average loss=0.07578179091215134 +INFO:root: +[Valid][10][UTT][ MSE = 0.2272 | LCC = 0.9623 | SRCC = 0.9631 ] +INFO:root:Average loss=0.07201830938458442 +INFO:root: +[Valid][11][UTT][ MSE = 0.1888 | LCC = 0.9639 | SRCC = 0.9639 ] +INFO:root:Average loss=0.06536325800418853 +INFO:root: +[Valid][12][UTT][ MSE = 0.1724 | LCC = 0.9629 | SRCC = 0.9630 ] +INFO:root:Average loss=0.06439981538057327 +INFO:root: +[Valid][13][UTT][ MSE = 0.2063 | LCC = 0.9637 | SRCC = 0.9634 ] +INFO:root:Average loss=0.05954594445228577 +INFO:root: +[Valid][14][UTT][ MSE = 0.1695 | LCC = 0.9655 | SRCC = 0.9652 ] +INFO:root:Average loss=0.05546100081503391 +INFO:root: +[Valid][15][UTT][ MSE = 0.2039 | LCC = 0.9650 | SRCC = 0.9655 ] +INFO:root:Average loss=0.05252594877779484 +INFO:root: +[Valid][16][UTT][ MSE = 0.1991 | LCC = 0.9632 | SRCC = 0.9637 ] +INFO:root:Average loss=0.05029292277991772 +INFO:root: +[Valid][17][UTT][ MSE = 0.2174 | LCC = 0.9634 | SRCC = 0.9630 ] +INFO:root:Average loss=0.04876022115349769 +INFO:root: +[Valid][18][UTT][ MSE = 0.1911 | LCC = 0.9652 | SRCC = 0.9649 ] +INFO:root:Average loss=0.044976863116025924 +INFO:root: +[Valid][19][UTT][ MSE = 0.1864 | LCC = 0.9644 | SRCC = 0.9648 ] +INFO:root:Average loss=0.0445248910933733 +INFO:root: +[Valid][20][UTT][ MSE = 0.1719 | LCC = 0.9655 | SRCC = 0.9652 ] +INFO:root:Average loss=0.041578134208917615 +INFO:root: +[Valid][21][UTT][ MSE = 0.1778 | LCC = 0.9650 | SRCC = 0.9641 ] +INFO:root:Average loss=0.03981306117773056 +INFO:root: +[Valid][22][UTT][ MSE = 0.2010 | LCC = 0.9637 | SRCC = 0.9635 ] +INFO:root:Average loss=0.03760789909958839 +INFO:root: +[Valid][23][UTT][ MSE = 0.1892 | LCC = 0.9652 | SRCC = 0.9653 ] +INFO:root:Average loss=0.03843505191802978 +INFO:root: +[Valid][24][UTT][ MSE = 0.1628 | LCC = 0.9631 | SRCC = 0.9627 ] +INFO:root:Average loss=0.03996465885639191 +INFO:root: +[Valid][25][UTT][ MSE = 0.1762 | LCC = 0.9661 | SRCC = 0.9656 ] +INFO:root:Average loss=0.033921227879822255 +INFO:root: +[Valid][26][UTT][ MSE = 0.1750 | LCC = 0.9653 | SRCC = 0.9648 ] +INFO:root:Average loss=0.031718314930796625 +INFO:root: +[Valid][27][UTT][ MSE = 0.1850 | LCC = 0.9638 | SRCC = 0.9635 ] +INFO:root:Average loss=0.03207292023301125 +INFO:root: +[Valid][28][UTT][ MSE = 0.1822 | LCC = 0.9639 | SRCC = 0.9634 ] +INFO:root:Average loss=0.0315369284003973 +INFO:root: +[Valid][29][UTT][ MSE = 0.2005 | LCC = 0.9636 | SRCC = 0.9643 ] +INFO:root:Average loss=0.029694451227784156 +INFO:root: +[Valid][30][UTT][ MSE = 0.1946 | LCC = 0.9651 | SRCC = 0.9649 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1798 | LCC = 0.9652 | SRCC = 0.9653 ] diff --git a/results_tencent/w2v2_large_asr_en/layer4/run2/config.gin b/results_tencent/w2v2_large_asr_en/layer4/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer4/run2/model.pt b/results_tencent/w2v2_large_asr_en/layer4/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9aac9cde6fb69cb74c6a9e372a259c1acd6933b8 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd7d9e0d88e081663735ca79e0e081d4d7d72fccda5358dd6003cc2a019e95aa +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer4/run2/model_best.pt b/results_tencent/w2v2_large_asr_en/layer4/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d1231af88361b1d8d4d607457b15b49c0e2aef40 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3455fdedccd6e1dd8864371313743e678c475d433cacf642b38fed4ea7a6192e +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer4/run2/train.log b/results_tencent/w2v2_large_asr_en/layer4/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2c5d3c6489a7f271057a158898d97ee9de8fdb77 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42106850039958954 +INFO:root: +[Valid][0][UTT][ MSE = 0.2909 | LCC = 0.9294 | SRCC = 0.9287 ] +INFO:root:Average loss=0.1869671178460121 +INFO:root: +[Valid][1][UTT][ MSE = 0.2536 | LCC = 0.9457 | SRCC = 0.9461 ] +INFO:root:Average loss=0.15901684057712556 +INFO:root: +[Valid][2][UTT][ MSE = 0.2279 | LCC = 0.9521 | SRCC = 0.9526 ] +INFO:root:Average loss=0.1375205750465393 +INFO:root: +[Valid][3][UTT][ MSE = 0.2151 | LCC = 0.9563 | SRCC = 0.9565 ] +INFO:root:Average loss=0.1207219153046608 +INFO:root: +[Valid][4][UTT][ MSE = 0.2124 | LCC = 0.9587 | SRCC = 0.9590 ] +INFO:root:Average loss=0.11309421837329864 +INFO:root: +[Valid][5][UTT][ MSE = 0.1824 | LCC = 0.9581 | SRCC = 0.9611 ] +INFO:root:Average loss=0.10162322482466697 +INFO:root: +[Valid][6][UTT][ MSE = 0.2089 | LCC = 0.9609 | SRCC = 0.9623 ] +INFO:root:Average loss=0.095776064068079 +INFO:root: +[Valid][7][UTT][ MSE = 0.2193 | LCC = 0.9613 | SRCC = 0.9627 ] +INFO:root:Average loss=0.0872385855615139 +INFO:root: +[Valid][8][UTT][ MSE = 0.1709 | LCC = 0.9629 | SRCC = 0.9634 ] +INFO:root:Average loss=0.0818317459821701 +INFO:root: +[Valid][9][UTT][ MSE = 0.2051 | LCC = 0.9640 | SRCC = 0.9644 ] +INFO:root:Average loss=0.07956311586499214 +INFO:root: +[Valid][10][UTT][ MSE = 0.1776 | LCC = 0.9649 | SRCC = 0.9649 ] +INFO:root:Average loss=0.0739378282725811 +INFO:root: +[Valid][11][UTT][ MSE = 0.2146 | LCC = 0.9639 | SRCC = 0.9644 ] +INFO:root:Average loss=0.06780030649900437 +INFO:root: +[Valid][12][UTT][ MSE = 0.1886 | LCC = 0.9635 | SRCC = 0.9638 ] +INFO:root:Average loss=0.06595878946781158 +INFO:root: +[Valid][13][UTT][ MSE = 0.1855 | LCC = 0.9660 | SRCC = 0.9662 ] +INFO:root:Average loss=0.059750383317470554 +INFO:root: +[Valid][14][UTT][ MSE = 0.1812 | LCC = 0.9658 | SRCC = 0.9660 ] +INFO:root:Average loss=0.05707553282380104 +INFO:root: +[Valid][15][UTT][ MSE = 0.1774 | LCC = 0.9653 | SRCC = 0.9653 ] +INFO:root:Average loss=0.05549422565102577 +INFO:root: +[Valid][16][UTT][ MSE = 0.1703 | LCC = 0.9678 | SRCC = 0.9673 ] +INFO:root:Average loss=0.05105836789309978 +INFO:root: +[Valid][17][UTT][ MSE = 0.1873 | LCC = 0.9651 | SRCC = 0.9661 ] +INFO:root:Average loss=0.049749864667654034 +INFO:root: +[Valid][18][UTT][ MSE = 0.1931 | LCC = 0.9670 | SRCC = 0.9677 ] +INFO:root:Average loss=0.049917013317346576 +INFO:root: +[Valid][19][UTT][ MSE = 0.1587 | LCC = 0.9664 | SRCC = 0.9670 ] +INFO:root:Average loss=0.04544611720740795 +INFO:root: +[Valid][20][UTT][ MSE = 0.1706 | LCC = 0.9668 | SRCC = 0.9669 ] +INFO:root:Average loss=0.04163267767429352 +INFO:root: +[Valid][21][UTT][ MSE = 0.1658 | LCC = 0.9674 | SRCC = 0.9667 ] +INFO:root:Average loss=0.040656992107629776 +INFO:root: +[Valid][22][UTT][ MSE = 0.1990 | LCC = 0.9663 | SRCC = 0.9671 ] +INFO:root:Average loss=0.03983272996544838 +INFO:root: +[Valid][23][UTT][ MSE = 0.1803 | LCC = 0.9665 | SRCC = 0.9671 ] +INFO:root:Average loss=0.03759927083551884 +INFO:root: +[Valid][24][UTT][ MSE = 0.1645 | LCC = 0.9667 | SRCC = 0.9667 ] +INFO:root:Average loss=0.036310606628656386 +INFO:root: +[Valid][25][UTT][ MSE = 0.1622 | LCC = 0.9673 | SRCC = 0.9668 ] +INFO:root:Average loss=0.03476523225009441 +INFO:root: +[Valid][26][UTT][ MSE = 0.1885 | LCC = 0.9667 | SRCC = 0.9669 ] +INFO:root:Average loss=0.03208407415449619 +INFO:root: +[Valid][27][UTT][ MSE = 0.1632 | LCC = 0.9662 | SRCC = 0.9661 ] +INFO:root:Average loss=0.03180764265358448 +INFO:root: +[Valid][28][UTT][ MSE = 0.1604 | LCC = 0.9671 | SRCC = 0.9669 ] +INFO:root:Average loss=0.02994294311106205 +INFO:root: +[Valid][29][UTT][ MSE = 0.1714 | LCC = 0.9674 | SRCC = 0.9672 ] +INFO:root:Average loss=0.031327447310090066 +INFO:root: +[Valid][30][UTT][ MSE = 0.1746 | LCC = 0.9665 | SRCC = 0.9661 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1801 | LCC = 0.9644 | SRCC = 0.9649 ] diff --git a/results_tencent/w2v2_large_asr_en/layer4/run3/config.gin b/results_tencent/w2v2_large_asr_en/layer4/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer4/run3/model.pt b/results_tencent/w2v2_large_asr_en/layer4/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..02959ee6ffefb342d882846e804a67f34b2dfd8f --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eabf88427683de4b69f3c59a56760ee906f8f438b00cc46e017994e26f76706e +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer4/run3/model_best.pt b/results_tencent/w2v2_large_asr_en/layer4/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..51089039b4bee207c9e3d0b5fb0512097cc3884a --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1bb0e876e76e0d7439ea2d74b4a36d60e57411cc7697b6ead3c7f2550bc677fe +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer4/run3/train.log b/results_tencent/w2v2_large_asr_en/layer4/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..629eb598b2ffa62dc9004bca6531fd6da9c7bb0f --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4625466632843018 +INFO:root: +[Valid][0][UTT][ MSE = 0.2681 | LCC = 0.9242 | SRCC = 0.9237 ] +INFO:root:Average loss=0.20443135231733323 +INFO:root: +[Valid][1][UTT][ MSE = 0.2358 | LCC = 0.9405 | SRCC = 0.9401 ] +INFO:root:Average loss=0.16048751401901246 +INFO:root: +[Valid][2][UTT][ MSE = 0.2215 | LCC = 0.9521 | SRCC = 0.9524 ] +INFO:root:Average loss=0.13031393909454345 +INFO:root: +[Valid][3][UTT][ MSE = 0.2505 | LCC = 0.9538 | SRCC = 0.9534 ] +INFO:root:Average loss=0.12098752784729004 +INFO:root: +[Valid][4][UTT][ MSE = 0.1848 | LCC = 0.9587 | SRCC = 0.9590 ] +INFO:root:Average loss=0.10543209207057953 +INFO:root: +[Valid][5][UTT][ MSE = 0.1930 | LCC = 0.9590 | SRCC = 0.9596 ] +INFO:root:Average loss=0.09937078219652176 +INFO:root: +[Valid][6][UTT][ MSE = 0.2026 | LCC = 0.9587 | SRCC = 0.9598 ] +INFO:root:Average loss=0.09152632164955139 +INFO:root: +[Valid][7][UTT][ MSE = 0.1995 | LCC = 0.9628 | SRCC = 0.9626 ] +INFO:root:Average loss=0.08673545944690704 +INFO:root: +[Valid][8][UTT][ MSE = 0.1784 | LCC = 0.9640 | SRCC = 0.9640 ] +INFO:root:Average loss=0.07737213972210884 +INFO:root: +[Valid][9][UTT][ MSE = 0.1841 | LCC = 0.9636 | SRCC = 0.9643 ] +INFO:root:Average loss=0.07431969049572945 +INFO:root: +[Valid][10][UTT][ MSE = 0.2100 | LCC = 0.9626 | SRCC = 0.9632 ] +INFO:root:Average loss=0.06926242816448212 +INFO:root: +[Valid][11][UTT][ MSE = 0.1577 | LCC = 0.9632 | SRCC = 0.9634 ] +INFO:root:Average loss=0.06397379651665687 +INFO:root: +[Valid][12][UTT][ MSE = 0.1797 | LCC = 0.9638 | SRCC = 0.9646 ] +INFO:root:Average loss=0.06586832955479621 +INFO:root: +[Valid][13][UTT][ MSE = 0.1808 | LCC = 0.9656 | SRCC = 0.9657 ] +INFO:root:Average loss=0.05707314360141754 +INFO:root: +[Valid][14][UTT][ MSE = 0.1622 | LCC = 0.9664 | SRCC = 0.9658 ] +INFO:root:Average loss=0.055626183167099955 +INFO:root: +[Valid][15][UTT][ MSE = 0.1568 | LCC = 0.9661 | SRCC = 0.9657 ] +INFO:root:Average loss=0.05175528921186924 +INFO:root: +[Valid][16][UTT][ MSE = 0.1805 | LCC = 0.9655 | SRCC = 0.9661 ] +INFO:root:Average loss=0.050628695473074914 +INFO:root: +[Valid][17][UTT][ MSE = 0.1680 | LCC = 0.9662 | SRCC = 0.9656 ] +INFO:root:Average loss=0.050106913343071935 +INFO:root: +[Valid][18][UTT][ MSE = 0.1783 | LCC = 0.9654 | SRCC = 0.9654 ] +INFO:root:Average loss=0.04441190633177757 +INFO:root: +[Valid][19][UTT][ MSE = 0.1738 | LCC = 0.9650 | SRCC = 0.9645 ] +INFO:root:Average loss=0.04534296888113022 +INFO:root: +[Valid][20][UTT][ MSE = 0.1776 | LCC = 0.9660 | SRCC = 0.9658 ] +INFO:root:Average loss=0.04121384508907795 +INFO:root: +[Valid][21][UTT][ MSE = 0.1928 | LCC = 0.9649 | SRCC = 0.9639 ] +INFO:root:Average loss=0.04190830288827419 +INFO:root: +[Valid][22][UTT][ MSE = 0.1580 | LCC = 0.9668 | SRCC = 0.9665 ] +INFO:root:Average loss=0.03816135135293007 +INFO:root: +[Valid][23][UTT][ MSE = 0.1653 | LCC = 0.9659 | SRCC = 0.9656 ] +INFO:root:Average loss=0.03760126775503159 +INFO:root: +[Valid][24][UTT][ MSE = 0.1715 | LCC = 0.9657 | SRCC = 0.9652 ] +INFO:root:Average loss=0.03499039854109287 +INFO:root: +[Valid][25][UTT][ MSE = 0.2002 | LCC = 0.9659 | SRCC = 0.9651 ] +INFO:root:Average loss=0.03509222161769867 +INFO:root: +[Valid][26][UTT][ MSE = 0.1736 | LCC = 0.9654 | SRCC = 0.9651 ] +INFO:root:Average loss=0.03238797178864479 +INFO:root: +[Valid][27][UTT][ MSE = 0.1790 | LCC = 0.9664 | SRCC = 0.9661 ] +INFO:root:Average loss=0.030899213775992394 +INFO:root: +[Valid][28][UTT][ MSE = 0.1968 | LCC = 0.9665 | SRCC = 0.9663 ] +INFO:root:Average loss=0.030679566979408265 +INFO:root: +[Valid][29][UTT][ MSE = 0.1636 | LCC = 0.9662 | SRCC = 0.9658 ] +INFO:root:Average loss=0.03014762182533741 +INFO:root: +[Valid][30][UTT][ MSE = 0.1843 | LCC = 0.9669 | SRCC = 0.9664 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1899 | LCC = 0.9645 | SRCC = 0.9647 ] diff --git a/results_tencent/w2v2_large_asr_en/layer4/run4/config.gin b/results_tencent/w2v2_large_asr_en/layer4/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer4/run4/model.pt b/results_tencent/w2v2_large_asr_en/layer4/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5069d3fdaad900b28557061a92baf37dc1e9e929 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4ab175b42616da20045a4eca8dcaf21d21ac1d22b4a9a9f4b80be3d33efc2e43 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer4/run4/model_best.pt b/results_tencent/w2v2_large_asr_en/layer4/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..aee3decd480152042ee62efd74442af39e814cd5 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6c6171dcf1d18dfdc40c83afa19427decd27c10da60db562b20f986642dc273 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer4/run4/train.log b/results_tencent/w2v2_large_asr_en/layer4/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..09e929155b0d0959367440889724627fb2e16dfd --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer4/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.47332065188884737 +INFO:root: +[Valid][0][UTT][ MSE = 0.2744 | LCC = 0.9235 | SRCC = 0.9241 ] +INFO:root:Average loss=0.1934062297940254 +INFO:root: +[Valid][1][UTT][ MSE = 0.2973 | LCC = 0.9383 | SRCC = 0.9408 ] +INFO:root:Average loss=0.17091308742761613 +INFO:root: +[Valid][2][UTT][ MSE = 0.2620 | LCC = 0.9486 | SRCC = 0.9493 ] +INFO:root:Average loss=0.14382527601718903 +INFO:root: +[Valid][3][UTT][ MSE = 0.2434 | LCC = 0.9521 | SRCC = 0.9533 ] +INFO:root:Average loss=0.12815906918048858 +INFO:root: +[Valid][4][UTT][ MSE = 0.2129 | LCC = 0.9551 | SRCC = 0.9560 ] +INFO:root:Average loss=0.11142356884479522 +INFO:root: +[Valid][5][UTT][ MSE = 0.2147 | LCC = 0.9577 | SRCC = 0.9584 ] +INFO:root:Average loss=0.10380018672347069 +INFO:root: +[Valid][6][UTT][ MSE = 0.2169 | LCC = 0.9591 | SRCC = 0.9595 ] +INFO:root:Average loss=0.09746662122011185 +INFO:root: +[Valid][7][UTT][ MSE = 0.2223 | LCC = 0.9586 | SRCC = 0.9589 ] +INFO:root:Average loss=0.09305656066536903 +INFO:root: +[Valid][8][UTT][ MSE = 0.2218 | LCC = 0.9606 | SRCC = 0.9618 ] +INFO:root:Average loss=0.08224126616120339 +INFO:root: +[Valid][9][UTT][ MSE = 0.2014 | LCC = 0.9616 | SRCC = 0.9626 ] +INFO:root:Average loss=0.07674979028105736 +INFO:root: +[Valid][10][UTT][ MSE = 0.1937 | LCC = 0.9626 | SRCC = 0.9629 ] +INFO:root:Average loss=0.07292879965901375 +INFO:root: +[Valid][11][UTT][ MSE = 0.1925 | LCC = 0.9602 | SRCC = 0.9613 ] +INFO:root:Average loss=0.06755660444498063 +INFO:root: +[Valid][12][UTT][ MSE = 0.2041 | LCC = 0.9627 | SRCC = 0.9631 ] +INFO:root:Average loss=0.0665672362446785 +INFO:root: +[Valid][13][UTT][ MSE = 0.1802 | LCC = 0.9640 | SRCC = 0.9640 ] +INFO:root:Average loss=0.06432863119244575 +INFO:root: +[Valid][14][UTT][ MSE = 0.1612 | LCC = 0.9629 | SRCC = 0.9630 ] +INFO:root:Average loss=0.05658770504593849 +INFO:root: +[Valid][15][UTT][ MSE = 0.1659 | LCC = 0.9632 | SRCC = 0.9628 ] +INFO:root:Average loss=0.05657769027352333 +INFO:root: +[Valid][16][UTT][ MSE = 0.1839 | LCC = 0.9639 | SRCC = 0.9639 ] +INFO:root:Average loss=0.05088978552818298 +INFO:root: +[Valid][17][UTT][ MSE = 0.1883 | LCC = 0.9639 | SRCC = 0.9632 ] +INFO:root:Average loss=0.05079701602458954 +INFO:root: +[Valid][18][UTT][ MSE = 0.1948 | LCC = 0.9635 | SRCC = 0.9635 ] +INFO:root:Average loss=0.04895923227071762 +INFO:root: +[Valid][19][UTT][ MSE = 0.1982 | LCC = 0.9628 | SRCC = 0.9629 ] +INFO:root:Average loss=0.0441070858836174 +INFO:root: +[Valid][20][UTT][ MSE = 0.1908 | LCC = 0.9631 | SRCC = 0.9625 ] +INFO:root:Average loss=0.043180518954992295 +INFO:root: +[Valid][21][UTT][ MSE = 0.1861 | LCC = 0.9631 | SRCC = 0.9630 ] +INFO:root:Average loss=0.041120040893554685 +INFO:root: +[Valid][22][UTT][ MSE = 0.1863 | LCC = 0.9638 | SRCC = 0.9638 ] +INFO:root:Average loss=0.04366115440428257 +INFO:root: +[Valid][23][UTT][ MSE = 0.1821 | LCC = 0.9646 | SRCC = 0.9644 ] +INFO:root:Average loss=0.038061881348490714 +INFO:root: +[Valid][24][UTT][ MSE = 0.1857 | LCC = 0.9636 | SRCC = 0.9639 ] +INFO:root:Average loss=0.03775146892666817 +INFO:root: +[Valid][25][UTT][ MSE = 0.1698 | LCC = 0.9642 | SRCC = 0.9640 ] +INFO:root:Average loss=0.037654535308480265 +INFO:root: +[Valid][26][UTT][ MSE = 0.1690 | LCC = 0.9646 | SRCC = 0.9643 ] +INFO:root:Average loss=0.033078351005911825 +INFO:root: +[Valid][27][UTT][ MSE = 0.1709 | LCC = 0.9640 | SRCC = 0.9637 ] +INFO:root:Average loss=0.033345479875803 +INFO:root: +[Valid][28][UTT][ MSE = 0.1925 | LCC = 0.9631 | SRCC = 0.9636 ] +INFO:root:Average loss=0.031342560842633246 +INFO:root: +[Valid][29][UTT][ MSE = 0.2133 | LCC = 0.9629 | SRCC = 0.9642 ] +INFO:root:Average loss=0.030500440150499342 +INFO:root: +[Valid][30][UTT][ MSE = 0.1682 | LCC = 0.9634 | SRCC = 0.9634 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1765 | LCC = 0.9632 | SRCC = 0.9637 ] diff --git a/results_tencent/w2v2_large_asr_en/layer6/run0/config.gin b/results_tencent/w2v2_large_asr_en/layer6/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer6/run0/model.pt b/results_tencent/w2v2_large_asr_en/layer6/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5e1f99fde3feacd9d0ae7b6b4352a44bf9e9c079 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6beecca4141f0ff1844676a0bbb2dfb9441c6b4b1ad874365662dabcd4418987 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer6/run0/model_best.pt b/results_tencent/w2v2_large_asr_en/layer6/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b4dcaae61f7572b17e2ba42b65d33ee0650e0431 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:364ad6fd4b80c29ceb788341834faffe9e2cc70f5ae86a0e04f4ae879ce493b0 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer6/run0/train.log b/results_tencent/w2v2_large_asr_en/layer6/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2b1ff883e36b5bb681c418d40cca1eee1c5a425f --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.49821924889087676 +INFO:root: +[Valid][0][UTT][ MSE = 0.3504 | LCC = 0.9232 | SRCC = 0.9253 ] +INFO:root:Average loss=0.2020507807135582 +INFO:root: +[Valid][1][UTT][ MSE = 0.2580 | LCC = 0.9417 | SRCC = 0.9415 ] +INFO:root:Average loss=0.15317598736286164 +INFO:root: +[Valid][2][UTT][ MSE = 0.2508 | LCC = 0.9427 | SRCC = 0.9437 ] +INFO:root:Average loss=0.13105828726291657 +INFO:root: +[Valid][3][UTT][ MSE = 0.2342 | LCC = 0.9543 | SRCC = 0.9542 ] +INFO:root:Average loss=0.11684628883004189 +INFO:root: +[Valid][4][UTT][ MSE = 0.1889 | LCC = 0.9579 | SRCC = 0.9576 ] +INFO:root:Average loss=0.10536563330888749 +INFO:root: +[Valid][5][UTT][ MSE = 0.1904 | LCC = 0.9590 | SRCC = 0.9587 ] +INFO:root:Average loss=0.0970396974682808 +INFO:root: +[Valid][6][UTT][ MSE = 0.2070 | LCC = 0.9594 | SRCC = 0.9587 ] +INFO:root:Average loss=0.09159786292910575 +INFO:root: +[Valid][7][UTT][ MSE = 0.1919 | LCC = 0.9608 | SRCC = 0.9611 ] +INFO:root:Average loss=0.08395052048563957 +INFO:root: +[Valid][8][UTT][ MSE = 0.1700 | LCC = 0.9620 | SRCC = 0.9624 ] +INFO:root:Average loss=0.0752015329003334 +INFO:root: +[Valid][9][UTT][ MSE = 0.1913 | LCC = 0.9616 | SRCC = 0.9620 ] +INFO:root:Average loss=0.07156747201085091 +INFO:root: +[Valid][10][UTT][ MSE = 0.1696 | LCC = 0.9631 | SRCC = 0.9631 ] +INFO:root:Average loss=0.0682992978990078 +INFO:root: +[Valid][11][UTT][ MSE = 0.1765 | LCC = 0.9641 | SRCC = 0.9638 ] +INFO:root:Average loss=0.06375686024129391 +INFO:root: +[Valid][12][UTT][ MSE = 0.1696 | LCC = 0.9640 | SRCC = 0.9643 ] +INFO:root:Average loss=0.061536416739225386 +INFO:root: +[Valid][13][UTT][ MSE = 0.2014 | LCC = 0.9641 | SRCC = 0.9640 ] +INFO:root:Average loss=0.056850548446178435 +INFO:root: +[Valid][14][UTT][ MSE = 0.1919 | LCC = 0.9635 | SRCC = 0.9645 ] +INFO:root:Average loss=0.0539630369246006 +INFO:root: +[Valid][15][UTT][ MSE = 0.1833 | LCC = 0.9642 | SRCC = 0.9644 ] +INFO:root:Average loss=0.05036824378371239 +INFO:root: +[Valid][16][UTT][ MSE = 0.1919 | LCC = 0.9649 | SRCC = 0.9649 ] +INFO:root:Average loss=0.04965345057845116 +INFO:root: +[Valid][17][UTT][ MSE = 0.1862 | LCC = 0.9646 | SRCC = 0.9644 ] +INFO:root:Average loss=0.04673938292264938 +INFO:root: +[Valid][18][UTT][ MSE = 0.1640 | LCC = 0.9645 | SRCC = 0.9642 ] +INFO:root:Average loss=0.04417206409573555 +INFO:root: +[Valid][19][UTT][ MSE = 0.1937 | LCC = 0.9638 | SRCC = 0.9639 ] +INFO:root:Average loss=0.04233111310005188 +INFO:root: +[Valid][20][UTT][ MSE = 0.1590 | LCC = 0.9649 | SRCC = 0.9648 ] +INFO:root:Average loss=0.040343502819538114 +INFO:root: +[Valid][21][UTT][ MSE = 0.1864 | LCC = 0.9655 | SRCC = 0.9656 ] +INFO:root:Average loss=0.03533714334666729 +INFO:root: +[Valid][22][UTT][ MSE = 0.1771 | LCC = 0.9657 | SRCC = 0.9652 ] +INFO:root:Average loss=0.035094133049249646 +INFO:root: +[Valid][23][UTT][ MSE = 0.1899 | LCC = 0.9637 | SRCC = 0.9640 ] +INFO:root:Average loss=0.035229577034711836 +INFO:root: +[Valid][24][UTT][ MSE = 0.1952 | LCC = 0.9645 | SRCC = 0.9645 ] +INFO:root:Average loss=0.03256025655567646 +INFO:root: +[Valid][25][UTT][ MSE = 0.1797 | LCC = 0.9654 | SRCC = 0.9652 ] +INFO:root:Average loss=0.03191466352343559 +INFO:root: +[Valid][26][UTT][ MSE = 0.1757 | LCC = 0.9650 | SRCC = 0.9651 ] +INFO:root:Average loss=0.03062227690219879 +INFO:root: +[Valid][27][UTT][ MSE = 0.1736 | LCC = 0.9647 | SRCC = 0.9648 ] +INFO:root:Average loss=0.02829884570837021 +INFO:root: +[Valid][28][UTT][ MSE = 0.1821 | LCC = 0.9650 | SRCC = 0.9651 ] +INFO:root:Average loss=0.02769964273273945 +INFO:root: +[Valid][29][UTT][ MSE = 0.1732 | LCC = 0.9649 | SRCC = 0.9650 ] +INFO:root:Average loss=0.02722964106500149 +INFO:root: +[Valid][30][UTT][ MSE = 0.1904 | LCC = 0.9648 | SRCC = 0.9644 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1791 | LCC = 0.9641 | SRCC = 0.9650 ] diff --git a/results_tencent/w2v2_large_asr_en/layer6/run1/config.gin b/results_tencent/w2v2_large_asr_en/layer6/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer6/run1/model.pt b/results_tencent/w2v2_large_asr_en/layer6/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..479af926e4d79004d0a7a31917dafd65b2c26cd2 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6feae22f4c842343d0f9794e8321988bff159cb97d5b79c393d7441a17a7dd47 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer6/run1/model_best.pt b/results_tencent/w2v2_large_asr_en/layer6/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..dc8f8352c98efe80791f9b80c67eb31222b7d65f --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:981d55e504e7cc6f0232d447181178f2b00a304f7c3008c5f930910945003506 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer6/run1/train.log b/results_tencent/w2v2_large_asr_en/layer6/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..40e34123c232e456344352969d291468adf98bd5 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42447751605510714 +INFO:root: +[Valid][0][UTT][ MSE = 0.2731 | LCC = 0.9316 | SRCC = 0.9289 ] +INFO:root:Average loss=0.18279764115810393 +INFO:root: +[Valid][1][UTT][ MSE = 0.3093 | LCC = 0.9447 | SRCC = 0.9438 ] +INFO:root:Average loss=0.14753565150499343 +INFO:root: +[Valid][2][UTT][ MSE = 0.2545 | LCC = 0.9490 | SRCC = 0.9491 ] +INFO:root:Average loss=0.13355881077051163 +INFO:root: +[Valid][3][UTT][ MSE = 0.2078 | LCC = 0.9562 | SRCC = 0.9566 ] +INFO:root:Average loss=0.11350516453385354 +INFO:root: +[Valid][4][UTT][ MSE = 0.2135 | LCC = 0.9576 | SRCC = 0.9583 ] +INFO:root:Average loss=0.10156402748823166 +INFO:root: +[Valid][5][UTT][ MSE = 0.1968 | LCC = 0.9616 | SRCC = 0.9615 ] +INFO:root:Average loss=0.09118670496344566 +INFO:root: +[Valid][6][UTT][ MSE = 0.2359 | LCC = 0.9608 | SRCC = 0.9614 ] +INFO:root:Average loss=0.08307325392961502 +INFO:root: +[Valid][7][UTT][ MSE = 0.2297 | LCC = 0.9608 | SRCC = 0.9607 ] +INFO:root:Average loss=0.07877889469265938 +INFO:root: +[Valid][8][UTT][ MSE = 0.1894 | LCC = 0.9625 | SRCC = 0.9629 ] +INFO:root:Average loss=0.07251003202795982 +INFO:root: +[Valid][9][UTT][ MSE = 0.1775 | LCC = 0.9635 | SRCC = 0.9633 ] +INFO:root:Average loss=0.06861440896987915 +INFO:root: +[Valid][10][UTT][ MSE = 0.2035 | LCC = 0.9644 | SRCC = 0.9643 ] +INFO:root:Average loss=0.06131222492456436 +INFO:root: +[Valid][11][UTT][ MSE = 0.2169 | LCC = 0.9633 | SRCC = 0.9636 ] +INFO:root:Average loss=0.061755076825618746 +INFO:root: +[Valid][12][UTT][ MSE = 0.1725 | LCC = 0.9640 | SRCC = 0.9634 ] +INFO:root:Average loss=0.05968631401658058 +INFO:root: +[Valid][13][UTT][ MSE = 0.1841 | LCC = 0.9652 | SRCC = 0.9651 ] +INFO:root:Average loss=0.052356165125966074 +INFO:root: +[Valid][14][UTT][ MSE = 0.2353 | LCC = 0.9613 | SRCC = 0.9638 ] +INFO:root:Average loss=0.04982278223335743 +INFO:root: +[Valid][15][UTT][ MSE = 0.1737 | LCC = 0.9651 | SRCC = 0.9650 ] +INFO:root:Average loss=0.04992925311625004 +INFO:root: +[Valid][16][UTT][ MSE = 0.2179 | LCC = 0.9647 | SRCC = 0.9642 ] +INFO:root:Average loss=0.048539483606815335 +INFO:root: +[Valid][17][UTT][ MSE = 0.1781 | LCC = 0.9637 | SRCC = 0.9639 ] +INFO:root:Average loss=0.04177609333395958 +INFO:root: +[Valid][18][UTT][ MSE = 0.1723 | LCC = 0.9635 | SRCC = 0.9635 ] +INFO:root:Average loss=0.040251337870955466 +INFO:root: +[Valid][19][UTT][ MSE = 0.1693 | LCC = 0.9648 | SRCC = 0.9643 ] +INFO:root:Average loss=0.03877095347642898 +INFO:root: +[Valid][20][UTT][ MSE = 0.1894 | LCC = 0.9653 | SRCC = 0.9652 ] +INFO:root:Average loss=0.039864749521017075 +INFO:root: +[Valid][21][UTT][ MSE = 0.2129 | LCC = 0.9636 | SRCC = 0.9637 ] +INFO:root:Average loss=0.03607527555525303 +INFO:root: +[Valid][22][UTT][ MSE = 0.1967 | LCC = 0.9647 | SRCC = 0.9646 ] +INFO:root:Average loss=0.03448130993545055 +INFO:root: +[Valid][23][UTT][ MSE = 0.1945 | LCC = 0.9646 | SRCC = 0.9645 ] +INFO:root:Average loss=0.03228059484064579 +INFO:root: +[Valid][24][UTT][ MSE = 0.1856 | LCC = 0.9652 | SRCC = 0.9651 ] +INFO:root:Average loss=0.03076499131321907 +INFO:root: +[Valid][25][UTT][ MSE = 0.1754 | LCC = 0.9651 | SRCC = 0.9648 ] +INFO:root:Average loss=0.03126687938719988 +INFO:root: +[Valid][26][UTT][ MSE = 0.1868 | LCC = 0.9637 | SRCC = 0.9639 ] +INFO:root:Average loss=0.02958538392186165 +INFO:root: +[Valid][27][UTT][ MSE = 0.2082 | LCC = 0.9649 | SRCC = 0.9645 ] +INFO:root:Average loss=0.027809459142386913 +INFO:root: +[Valid][28][UTT][ MSE = 0.1807 | LCC = 0.9654 | SRCC = 0.9649 ] +INFO:root:Average loss=0.02599536257237196 +INFO:root: +[Valid][29][UTT][ MSE = 0.1866 | LCC = 0.9657 | SRCC = 0.9649 ] +INFO:root:Average loss=0.024923365235328673 +INFO:root: +[Valid][30][UTT][ MSE = 0.2000 | LCC = 0.9641 | SRCC = 0.9641 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1896 | LCC = 0.9641 | SRCC = 0.9646 ] diff --git a/results_tencent/w2v2_large_asr_en/layer6/run2/config.gin b/results_tencent/w2v2_large_asr_en/layer6/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer6/run2/model.pt b/results_tencent/w2v2_large_asr_en/layer6/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d242d9045be5303f0c527200fd4d2523935802c3 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:476fe21e28b55549c9815c66941fa7928c7c84a9588c87088041876b85cb0dc8 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer6/run2/model_best.pt b/results_tencent/w2v2_large_asr_en/layer6/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..50576b632fd75a41edc3e679e984ffd01b28b869 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e9c47070c1b75338e6630e4a4addda2cf390df1608ae4c034e810fe891f2ed57 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer6/run2/train.log b/results_tencent/w2v2_large_asr_en/layer6/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..de04d5dcd92cfd130372c9bc63709978df0c2f54 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4026720604300499 +INFO:root: +[Valid][0][UTT][ MSE = 0.2543 | LCC = 0.9267 | SRCC = 0.9254 ] +INFO:root:Average loss=0.17499379348754882 +INFO:root: +[Valid][1][UTT][ MSE = 0.2110 | LCC = 0.9456 | SRCC = 0.9451 ] +INFO:root:Average loss=0.14593066322803497 +INFO:root: +[Valid][2][UTT][ MSE = 0.1943 | LCC = 0.9522 | SRCC = 0.9526 ] +INFO:root:Average loss=0.12585309529304503 +INFO:root: +[Valid][3][UTT][ MSE = 0.2346 | LCC = 0.9564 | SRCC = 0.9574 ] +INFO:root:Average loss=0.10916799759864806 +INFO:root: +[Valid][4][UTT][ MSE = 0.2113 | LCC = 0.9606 | SRCC = 0.9607 ] +INFO:root:Average loss=0.09883858269453048 +INFO:root: +[Valid][5][UTT][ MSE = 0.2298 | LCC = 0.9595 | SRCC = 0.9605 ] +INFO:root:Average loss=0.09931996673345565 +INFO:root: +[Valid][6][UTT][ MSE = 0.2230 | LCC = 0.9601 | SRCC = 0.9604 ] +INFO:root:Average loss=0.0839550718665123 +INFO:root: +[Valid][7][UTT][ MSE = 0.1928 | LCC = 0.9624 | SRCC = 0.9629 ] +INFO:root:Average loss=0.08092363166809081 +INFO:root: +[Valid][8][UTT][ MSE = 0.2123 | LCC = 0.9609 | SRCC = 0.9621 ] +INFO:root:Average loss=0.07474247792363167 +INFO:root: +[Valid][9][UTT][ MSE = 0.1855 | LCC = 0.9627 | SRCC = 0.9631 ] +INFO:root:Average loss=0.06855841401219367 +INFO:root: +[Valid][10][UTT][ MSE = 0.2094 | LCC = 0.9631 | SRCC = 0.9639 ] +INFO:root:Average loss=0.06343987423181534 +INFO:root: +[Valid][11][UTT][ MSE = 0.1939 | LCC = 0.9649 | SRCC = 0.9652 ] +INFO:root:Average loss=0.06029405921697616 +INFO:root: +[Valid][12][UTT][ MSE = 0.1742 | LCC = 0.9610 | SRCC = 0.9619 ] +INFO:root:Average loss=0.0588157517015934 +INFO:root: +[Valid][13][UTT][ MSE = 0.1919 | LCC = 0.9646 | SRCC = 0.9649 ] +INFO:root:Average loss=0.05372024603188038 +INFO:root: +[Valid][14][UTT][ MSE = 0.1975 | LCC = 0.9640 | SRCC = 0.9645 ] +INFO:root:Average loss=0.050123248130083085 +INFO:root: +[Valid][15][UTT][ MSE = 0.1834 | LCC = 0.9651 | SRCC = 0.9647 ] +INFO:root:Average loss=0.050257614374160764 +INFO:root: +[Valid][16][UTT][ MSE = 0.1782 | LCC = 0.9640 | SRCC = 0.9642 ] +INFO:root:Average loss=0.04834371602535248 +INFO:root: +[Valid][17][UTT][ MSE = 0.1829 | LCC = 0.9663 | SRCC = 0.9661 ] +INFO:root:Average loss=0.04217220379412174 +INFO:root: +[Valid][18][UTT][ MSE = 0.1849 | LCC = 0.9653 | SRCC = 0.9651 ] +INFO:root:Average loss=0.043084576025605205 +INFO:root: +[Valid][19][UTT][ MSE = 0.1894 | LCC = 0.9660 | SRCC = 0.9658 ] +INFO:root:Average loss=0.04070400682091713 +INFO:root: +[Valid][20][UTT][ MSE = 0.1836 | LCC = 0.9659 | SRCC = 0.9656 ] +INFO:root:Average loss=0.04018664167821407 +INFO:root: +[Valid][21][UTT][ MSE = 0.1723 | LCC = 0.9662 | SRCC = 0.9662 ] +INFO:root:Average loss=0.03729250138998032 +INFO:root: +[Valid][22][UTT][ MSE = 0.1648 | LCC = 0.9664 | SRCC = 0.9662 ] +INFO:root:Average loss=0.03522733652591705 +INFO:root: +[Valid][23][UTT][ MSE = 0.1836 | LCC = 0.9670 | SRCC = 0.9665 ] +INFO:root:Average loss=0.03163534162938595 +INFO:root: +[Valid][24][UTT][ MSE = 0.1957 | LCC = 0.9653 | SRCC = 0.9656 ] +INFO:root:Average loss=0.0319235420525074 +INFO:root: +[Valid][25][UTT][ MSE = 0.1630 | LCC = 0.9659 | SRCC = 0.9655 ] +INFO:root:Average loss=0.03096470533311367 +INFO:root: +[Valid][26][UTT][ MSE = 0.1756 | LCC = 0.9654 | SRCC = 0.9655 ] +INFO:root:Average loss=0.03196676729619503 +INFO:root: +[Valid][27][UTT][ MSE = 0.1607 | LCC = 0.9666 | SRCC = 0.9661 ] +INFO:root:Average loss=0.028632036671042443 +INFO:root: +[Valid][28][UTT][ MSE = 0.1771 | LCC = 0.9668 | SRCC = 0.9662 ] +INFO:root:Average loss=0.027178023338317873 +INFO:root: +[Valid][29][UTT][ MSE = 0.1698 | LCC = 0.9665 | SRCC = 0.9659 ] +INFO:root:Average loss=0.026800749525427818 +INFO:root: +[Valid][30][UTT][ MSE = 0.1773 | LCC = 0.9664 | SRCC = 0.9655 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1867 | LCC = 0.9647 | SRCC = 0.9649 ] diff --git a/results_tencent/w2v2_large_asr_en/layer6/run3/config.gin b/results_tencent/w2v2_large_asr_en/layer6/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer6/run3/model.pt b/results_tencent/w2v2_large_asr_en/layer6/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0ec1d1a2c9b0ba292f9c2145af73dec73737dc89 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:56dd01d6e52ad5457cbe7e755fcaba9e7f32eecffe3df85b46d95f73fe54e5ca +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer6/run3/model_best.pt b/results_tencent/w2v2_large_asr_en/layer6/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a2c5f85b300cc9be5a0c842e36d0f8a36b436f3b --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:51c11d3c795ca55c48fac23799adb4ba0e99d92f90bc475e1836a005122f172b +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer6/run3/train.log b/results_tencent/w2v2_large_asr_en/layer6/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a64d1864bc95e96f72bfd3dcf7f4a897ee40ade6 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.37453754484653473 +INFO:root: +[Valid][0][UTT][ MSE = 0.3191 | LCC = 0.9280 | SRCC = 0.9275 ] +INFO:root:Average loss=0.18199093616008757 +INFO:root: +[Valid][1][UTT][ MSE = 0.2774 | LCC = 0.9438 | SRCC = 0.9442 ] +INFO:root:Average loss=0.14833564203977584 +INFO:root: +[Valid][2][UTT][ MSE = 0.1990 | LCC = 0.9527 | SRCC = 0.9533 ] +INFO:root:Average loss=0.12577738893032073 +INFO:root: +[Valid][3][UTT][ MSE = 0.2286 | LCC = 0.9552 | SRCC = 0.9543 ] +INFO:root:Average loss=0.11198640909790993 +INFO:root: +[Valid][4][UTT][ MSE = 0.2137 | LCC = 0.9587 | SRCC = 0.9583 ] +INFO:root:Average loss=0.10165587201714515 +INFO:root: +[Valid][5][UTT][ MSE = 0.1823 | LCC = 0.9610 | SRCC = 0.9606 ] +INFO:root:Average loss=0.09581315931677818 +INFO:root: +[Valid][6][UTT][ MSE = 0.2207 | LCC = 0.9604 | SRCC = 0.9613 ] +INFO:root:Average loss=0.09011014640331268 +INFO:root: +[Valid][7][UTT][ MSE = 0.1980 | LCC = 0.9614 | SRCC = 0.9617 ] +INFO:root:Average loss=0.08566073617339134 +INFO:root: +[Valid][8][UTT][ MSE = 0.1856 | LCC = 0.9628 | SRCC = 0.9629 ] +INFO:root:Average loss=0.08193867206573487 +INFO:root: +[Valid][9][UTT][ MSE = 0.1755 | LCC = 0.9641 | SRCC = 0.9634 ] +INFO:root:Average loss=0.07125110045075417 +INFO:root: +[Valid][10][UTT][ MSE = 0.1938 | LCC = 0.9631 | SRCC = 0.9632 ] +INFO:root:Average loss=0.0674230415225029 +INFO:root: +[Valid][11][UTT][ MSE = 0.2059 | LCC = 0.9616 | SRCC = 0.9623 ] +INFO:root:Average loss=0.06378869372606277 +INFO:root: +[Valid][12][UTT][ MSE = 0.2723 | LCC = 0.9640 | SRCC = 0.9641 ] +INFO:root:Average loss=0.06031427025794983 +INFO:root: +[Valid][13][UTT][ MSE = 0.2000 | LCC = 0.9606 | SRCC = 0.9618 ] +INFO:root:Average loss=0.05627043832838535 +INFO:root: +[Valid][14][UTT][ MSE = 0.1752 | LCC = 0.9650 | SRCC = 0.9648 ] +INFO:root:Average loss=0.05195636102557182 +INFO:root: +[Valid][15][UTT][ MSE = 0.1717 | LCC = 0.9646 | SRCC = 0.9645 ] +INFO:root:Average loss=0.05016370278596878 +INFO:root: +[Valid][16][UTT][ MSE = 0.1957 | LCC = 0.9635 | SRCC = 0.9637 ] +INFO:root:Average loss=0.047644722059369086 +INFO:root: +[Valid][17][UTT][ MSE = 0.1753 | LCC = 0.9641 | SRCC = 0.9638 ] +INFO:root:Average loss=0.04586529275774956 +INFO:root: +[Valid][18][UTT][ MSE = 0.1884 | LCC = 0.9645 | SRCC = 0.9647 ] +INFO:root:Average loss=0.0443224790841341 +INFO:root: +[Valid][19][UTT][ MSE = 0.1743 | LCC = 0.9641 | SRCC = 0.9638 ] +INFO:root:Average loss=0.040755538299679754 +INFO:root: +[Valid][20][UTT][ MSE = 0.1823 | LCC = 0.9643 | SRCC = 0.9644 ] +INFO:root:Average loss=0.039417760640382764 +INFO:root: +[Valid][21][UTT][ MSE = 0.1975 | LCC = 0.9628 | SRCC = 0.9634 ] +INFO:root:Average loss=0.03525868116319179 +INFO:root: +[Valid][22][UTT][ MSE = 0.1931 | LCC = 0.9637 | SRCC = 0.9637 ] +INFO:root:Average loss=0.0366339406222105 +INFO:root: +[Valid][23][UTT][ MSE = 0.1707 | LCC = 0.9653 | SRCC = 0.9649 ] +INFO:root:Average loss=0.03493885861337185 +INFO:root: +[Valid][24][UTT][ MSE = 0.2078 | LCC = 0.9642 | SRCC = 0.9644 ] +INFO:root:Average loss=0.0324242639541626 +INFO:root: +[Valid][25][UTT][ MSE = 0.1751 | LCC = 0.9641 | SRCC = 0.9642 ] +INFO:root:Average loss=0.030466555878520013 +INFO:root: +[Valid][26][UTT][ MSE = 0.1687 | LCC = 0.9651 | SRCC = 0.9645 ] +INFO:root:Average loss=0.029990160375833512 +INFO:root: +[Valid][27][UTT][ MSE = 0.1825 | LCC = 0.9649 | SRCC = 0.9648 ] +INFO:root:Average loss=0.028777973636984824 +INFO:root: +[Valid][28][UTT][ MSE = 0.1985 | LCC = 0.9643 | SRCC = 0.9643 ] +INFO:root:Average loss=0.027329094044864177 +INFO:root: +[Valid][29][UTT][ MSE = 0.1913 | LCC = 0.9645 | SRCC = 0.9644 ] +INFO:root:Average loss=0.026284653186798096 +INFO:root: +[Valid][30][UTT][ MSE = 0.1806 | LCC = 0.9649 | SRCC = 0.9644 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1762 | LCC = 0.9640 | SRCC = 0.9643 ] diff --git a/results_tencent/w2v2_large_asr_en/layer6/run4/config.gin b/results_tencent/w2v2_large_asr_en/layer6/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer6/run4/model.pt b/results_tencent/w2v2_large_asr_en/layer6/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..61b37f831f5cd57c2a577d96fc4a024a39280f98 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8f199313302cd3b0aed52b2699260c302be6a81df3eee70898892c26ebcea36b +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer6/run4/model_best.pt b/results_tencent/w2v2_large_asr_en/layer6/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0e60dd633845c65c17b9e3dd33e09e211eaa3561 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:00665af8ac7b9740e9ed7d1709edf067060f183e823ff19ce493f38400a8b8b5 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer6/run4/train.log b/results_tencent/w2v2_large_asr_en/layer6/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d7d1ef2163610afc0e01a4839e9dc2707938e50a --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer6/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3992909665107727 +INFO:root: +[Valid][0][UTT][ MSE = 0.3285 | LCC = 0.9262 | SRCC = 0.9254 ] +INFO:root:Average loss=0.18409973430633544 +INFO:root: +[Valid][1][UTT][ MSE = 0.2463 | LCC = 0.9437 | SRCC = 0.9426 ] +INFO:root:Average loss=0.14651582491397858 +INFO:root: +[Valid][2][UTT][ MSE = 0.2300 | LCC = 0.9496 | SRCC = 0.9502 ] +INFO:root:Average loss=0.12670693224668503 +INFO:root: +[Valid][3][UTT][ MSE = 0.2224 | LCC = 0.9561 | SRCC = 0.9558 ] +INFO:root:Average loss=0.1105999658703804 +INFO:root: +[Valid][4][UTT][ MSE = 0.1966 | LCC = 0.9562 | SRCC = 0.9572 ] +INFO:root:Average loss=0.09805730831623077 +INFO:root: +[Valid][5][UTT][ MSE = 0.1694 | LCC = 0.9580 | SRCC = 0.9586 ] +INFO:root:Average loss=0.08906917968392372 +INFO:root: +[Valid][6][UTT][ MSE = 0.2110 | LCC = 0.9597 | SRCC = 0.9599 ] +INFO:root:Average loss=0.08325223037600517 +INFO:root: +[Valid][7][UTT][ MSE = 0.1889 | LCC = 0.9595 | SRCC = 0.9594 ] +INFO:root:Average loss=0.07808695766329765 +INFO:root: +[Valid][8][UTT][ MSE = 0.2013 | LCC = 0.9616 | SRCC = 0.9615 ] +INFO:root:Average loss=0.08338794350624085 +INFO:root: +[Valid][9][UTT][ MSE = 0.2154 | LCC = 0.9627 | SRCC = 0.9627 ] +INFO:root:Average loss=0.07035858383774757 +INFO:root: +[Valid][10][UTT][ MSE = 0.2195 | LCC = 0.9633 | SRCC = 0.9631 ] +INFO:root:Average loss=0.06618773087859153 +INFO:root: +[Valid][11][UTT][ MSE = 0.1887 | LCC = 0.9626 | SRCC = 0.9629 ] +INFO:root:Average loss=0.06074905022978783 +INFO:root: +[Valid][12][UTT][ MSE = 0.1871 | LCC = 0.9628 | SRCC = 0.9630 ] +INFO:root:Average loss=0.06142364051938057 +INFO:root: +[Valid][13][UTT][ MSE = 0.1803 | LCC = 0.9641 | SRCC = 0.9639 ] +INFO:root:Average loss=0.053983022540807724 +INFO:root: +[Valid][14][UTT][ MSE = 0.1684 | LCC = 0.9623 | SRCC = 0.9623 ] +INFO:root:Average loss=0.05070575216412544 +INFO:root: +[Valid][15][UTT][ MSE = 0.1822 | LCC = 0.9650 | SRCC = 0.9646 ] +INFO:root:Average loss=0.047639902129769325 +INFO:root: +[Valid][16][UTT][ MSE = 0.1577 | LCC = 0.9648 | SRCC = 0.9645 ] +INFO:root:Average loss=0.046851357012987135 +INFO:root: +[Valid][17][UTT][ MSE = 0.2104 | LCC = 0.9647 | SRCC = 0.9642 ] +INFO:root:Average loss=0.042352829709649084 +INFO:root: +[Valid][18][UTT][ MSE = 0.1693 | LCC = 0.9649 | SRCC = 0.9647 ] +INFO:root:Average loss=0.040883607879281046 +INFO:root: +[Valid][19][UTT][ MSE = 0.1867 | LCC = 0.9654 | SRCC = 0.9653 ] +INFO:root:Average loss=0.03886447755992412 +INFO:root: +[Valid][20][UTT][ MSE = 0.1895 | LCC = 0.9657 | SRCC = 0.9652 ] +INFO:root:Average loss=0.0377595407217741 +INFO:root: +[Valid][21][UTT][ MSE = 0.1721 | LCC = 0.9651 | SRCC = 0.9644 ] +INFO:root:Average loss=0.037344728127121925 +INFO:root: +[Valid][22][UTT][ MSE = 0.1800 | LCC = 0.9656 | SRCC = 0.9653 ] +INFO:root:Average loss=0.034021318323910234 +INFO:root: +[Valid][23][UTT][ MSE = 0.1770 | LCC = 0.9656 | SRCC = 0.9655 ] +INFO:root:Average loss=0.03281845131516457 +INFO:root: +[Valid][24][UTT][ MSE = 0.1795 | LCC = 0.9657 | SRCC = 0.9649 ] +INFO:root:Average loss=0.032963941067457196 +INFO:root: +[Valid][25][UTT][ MSE = 0.1715 | LCC = 0.9653 | SRCC = 0.9649 ] +INFO:root:Average loss=0.030723959386348726 +INFO:root: +[Valid][26][UTT][ MSE = 0.1911 | LCC = 0.9662 | SRCC = 0.9658 ] +INFO:root:Average loss=0.030006975173950196 +INFO:root: +[Valid][27][UTT][ MSE = 0.1893 | LCC = 0.9655 | SRCC = 0.9651 ] +INFO:root:Average loss=0.028687345340847967 +INFO:root: +[Valid][28][UTT][ MSE = 0.2013 | LCC = 0.9650 | SRCC = 0.9649 ] +INFO:root:Average loss=0.027873841017484666 +INFO:root: +[Valid][29][UTT][ MSE = 0.1742 | LCC = 0.9637 | SRCC = 0.9635 ] +INFO:root:Average loss=0.0278079771772027 +INFO:root: +[Valid][30][UTT][ MSE = 0.1606 | LCC = 0.9667 | SRCC = 0.9660 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1680 | LCC = 0.9639 | SRCC = 0.9648 ] diff --git a/results_tencent/w2v2_large_asr_en/layer8/run0/config.gin b/results_tencent/w2v2_large_asr_en/layer8/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer8/run0/model.pt b/results_tencent/w2v2_large_asr_en/layer8/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6102248b32225710131b65fadb33fd7bc7ee1c91 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0eb9cd460e485c36c14c576a91c3a013813918c6a9b12a0a85ef7b5198b2c08e +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer8/run0/model_best.pt b/results_tencent/w2v2_large_asr_en/layer8/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6d1488a39729145c00e075d3d971e9656192242a --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d7df123d26aef21185ae08a50ea920765225f2b40e50c74199ad9eb04fbc57d +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer8/run0/train.log b/results_tencent/w2v2_large_asr_en/layer8/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..00e4279df9ffb0518c778b248863e98fba3cd32e --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4969356902837753 +INFO:root: +[Valid][0][UTT][ MSE = 0.3538 | LCC = 0.9117 | SRCC = 0.9088 ] +INFO:root:Average loss=0.20660232681035995 +INFO:root: +[Valid][1][UTT][ MSE = 0.3145 | LCC = 0.9353 | SRCC = 0.9349 ] +INFO:root:Average loss=0.1727935298681259 +INFO:root: +[Valid][2][UTT][ MSE = 0.2195 | LCC = 0.9428 | SRCC = 0.9430 ] +INFO:root:Average loss=0.14739169657230378 +INFO:root: +[Valid][3][UTT][ MSE = 0.2256 | LCC = 0.9490 | SRCC = 0.9479 ] +INFO:root:Average loss=0.13124897795915605 +INFO:root: +[Valid][4][UTT][ MSE = 0.2780 | LCC = 0.9534 | SRCC = 0.9543 ] +INFO:root:Average loss=0.11531874236464501 +INFO:root: +[Valid][5][UTT][ MSE = 0.1792 | LCC = 0.9541 | SRCC = 0.9548 ] +INFO:root:Average loss=0.10934632009267807 +INFO:root: +[Valid][6][UTT][ MSE = 0.1938 | LCC = 0.9567 | SRCC = 0.9568 ] +INFO:root:Average loss=0.09554993796348572 +INFO:root: +[Valid][7][UTT][ MSE = 0.1765 | LCC = 0.9573 | SRCC = 0.9589 ] +INFO:root:Average loss=0.08614400812983512 +INFO:root: +[Valid][8][UTT][ MSE = 0.2014 | LCC = 0.9601 | SRCC = 0.9600 ] +INFO:root:Average loss=0.07937271857261657 +INFO:root: +[Valid][9][UTT][ MSE = 0.1842 | LCC = 0.9602 | SRCC = 0.9601 ] +INFO:root:Average loss=0.07782872533798217 +INFO:root: +[Valid][10][UTT][ MSE = 0.2149 | LCC = 0.9595 | SRCC = 0.9596 ] +INFO:root:Average loss=0.07077567836642265 +INFO:root: +[Valid][11][UTT][ MSE = 0.1831 | LCC = 0.9608 | SRCC = 0.9610 ] +INFO:root:Average loss=0.06685224029421806 +INFO:root: +[Valid][12][UTT][ MSE = 0.2009 | LCC = 0.9593 | SRCC = 0.9599 ] +INFO:root:Average loss=0.06139763075113296 +INFO:root: +[Valid][13][UTT][ MSE = 0.2190 | LCC = 0.9592 | SRCC = 0.9602 ] +INFO:root:Average loss=0.05808876045048237 +INFO:root: +[Valid][14][UTT][ MSE = 0.1779 | LCC = 0.9605 | SRCC = 0.9609 ] +INFO:root:Average loss=0.055856756210327145 +INFO:root: +[Valid][15][UTT][ MSE = 0.1766 | LCC = 0.9609 | SRCC = 0.9609 ] +INFO:root:Average loss=0.05319724370539188 +INFO:root: +[Valid][16][UTT][ MSE = 0.1531 | LCC = 0.9606 | SRCC = 0.9610 ] +INFO:root:Average loss=0.05008779241144657 +INFO:root: +[Valid][17][UTT][ MSE = 0.1996 | LCC = 0.9615 | SRCC = 0.9621 ] +INFO:root:Average loss=0.046126644283533094 +INFO:root: +[Valid][18][UTT][ MSE = 0.1971 | LCC = 0.9619 | SRCC = 0.9624 ] +INFO:root:Average loss=0.045175865828990934 +INFO:root: +[Valid][19][UTT][ MSE = 0.2149 | LCC = 0.9609 | SRCC = 0.9616 ] +INFO:root:Average loss=0.04304197768867016 +INFO:root: +[Valid][20][UTT][ MSE = 0.1981 | LCC = 0.9630 | SRCC = 0.9631 ] +INFO:root:Average loss=0.04003041100502014 +INFO:root: +[Valid][21][UTT][ MSE = 0.2019 | LCC = 0.9621 | SRCC = 0.9624 ] +INFO:root:Average loss=0.03754772785305977 +INFO:root: +[Valid][22][UTT][ MSE = 0.1831 | LCC = 0.9625 | SRCC = 0.9629 ] +INFO:root:Average loss=0.037801571279764176 +INFO:root: +[Valid][23][UTT][ MSE = 0.1755 | LCC = 0.9623 | SRCC = 0.9623 ] +INFO:root:Average loss=0.03516244818270207 +INFO:root: +[Valid][24][UTT][ MSE = 0.1810 | LCC = 0.9633 | SRCC = 0.9634 ] +INFO:root:Average loss=0.032707644566893575 +INFO:root: +[Valid][25][UTT][ MSE = 0.1594 | LCC = 0.9634 | SRCC = 0.9634 ] +INFO:root:Average loss=0.03191980791091919 +INFO:root: +[Valid][26][UTT][ MSE = 0.1868 | LCC = 0.9621 | SRCC = 0.9627 ] +INFO:root:Average loss=0.031062034323811532 +INFO:root: +[Valid][27][UTT][ MSE = 0.1700 | LCC = 0.9633 | SRCC = 0.9631 ] +INFO:root:Average loss=0.029993731617927552 +INFO:root: +[Valid][28][UTT][ MSE = 0.1999 | LCC = 0.9627 | SRCC = 0.9625 ] +INFO:root:Average loss=0.028640477143228055 +INFO:root: +[Valid][29][UTT][ MSE = 0.1726 | LCC = 0.9635 | SRCC = 0.9633 ] +INFO:root:Average loss=0.029769282177090645 +INFO:root: +[Valid][30][UTT][ MSE = 0.1774 | LCC = 0.9628 | SRCC = 0.9627 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1804 | LCC = 0.9618 | SRCC = 0.9628 ] diff --git a/results_tencent/w2v2_large_asr_en/layer8/run1/config.gin b/results_tencent/w2v2_large_asr_en/layer8/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer8/run1/model.pt b/results_tencent/w2v2_large_asr_en/layer8/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6c6a64ebe88104a60e0028e26e3f340a21c767b9 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16b0dcf3e2a12f1d9af86a7a93798bc914365c8d531822a2c99190e47aecfddf +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer8/run1/model_best.pt b/results_tencent/w2v2_large_asr_en/layer8/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2226e1f4efde98cae502bd89217b5585651597ba --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:037a37841077a9c20c9e4a35be9966c2530343e00633ee791cd5f341845d836d +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer8/run1/train.log b/results_tencent/w2v2_large_asr_en/layer8/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a8e1ba12723010da186c6a223187d53cbfbc7879 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46598492515087125 +INFO:root: +[Valid][0][UTT][ MSE = 0.3815 | LCC = 0.9147 | SRCC = 0.9143 ] +INFO:root:Average loss=0.21148688286542894 +INFO:root: +[Valid][1][UTT][ MSE = 0.2756 | LCC = 0.9372 | SRCC = 0.9367 ] +INFO:root:Average loss=0.16088824570178986 +INFO:root: +[Valid][2][UTT][ MSE = 0.2322 | LCC = 0.9471 | SRCC = 0.9474 ] +INFO:root:Average loss=0.13886086243391038 +INFO:root: +[Valid][3][UTT][ MSE = 0.2412 | LCC = 0.9531 | SRCC = 0.9534 ] +INFO:root:Average loss=0.12511237689852714 +INFO:root: +[Valid][4][UTT][ MSE = 0.2064 | LCC = 0.9561 | SRCC = 0.9560 ] +INFO:root:Average loss=0.12106458270549775 +INFO:root: +[Valid][5][UTT][ MSE = 0.1750 | LCC = 0.9575 | SRCC = 0.9585 ] +INFO:root:Average loss=0.09979512333869935 +INFO:root: +[Valid][6][UTT][ MSE = 0.2128 | LCC = 0.9568 | SRCC = 0.9577 ] +INFO:root:Average loss=0.09217605531215668 +INFO:root: +[Valid][7][UTT][ MSE = 0.1945 | LCC = 0.9605 | SRCC = 0.9606 ] +INFO:root:Average loss=0.08813777405023575 +INFO:root: +[Valid][8][UTT][ MSE = 0.1797 | LCC = 0.9572 | SRCC = 0.9570 ] +INFO:root:Average loss=0.08088360965251923 +INFO:root: +[Valid][9][UTT][ MSE = 0.1971 | LCC = 0.9615 | SRCC = 0.9619 ] +INFO:root:Average loss=0.07447268551588059 +INFO:root: +[Valid][10][UTT][ MSE = 0.1836 | LCC = 0.9605 | SRCC = 0.9603 ] +INFO:root:Average loss=0.07052150762081147 +INFO:root: +[Valid][11][UTT][ MSE = 0.2057 | LCC = 0.9621 | SRCC = 0.9622 ] +INFO:root:Average loss=0.06799553960561752 +INFO:root: +[Valid][12][UTT][ MSE = 0.1641 | LCC = 0.9614 | SRCC = 0.9611 ] +INFO:root:Average loss=0.06337766307592392 +INFO:root: +[Valid][13][UTT][ MSE = 0.1787 | LCC = 0.9636 | SRCC = 0.9635 ] +INFO:root:Average loss=0.05597265699505806 +INFO:root: +[Valid][14][UTT][ MSE = 0.1769 | LCC = 0.9637 | SRCC = 0.9632 ] +INFO:root:Average loss=0.0540192497074604 +INFO:root: +[Valid][15][UTT][ MSE = 0.1675 | LCC = 0.9596 | SRCC = 0.9584 ] +INFO:root:Average loss=0.05314879684150219 +INFO:root: +[Valid][16][UTT][ MSE = 0.1904 | LCC = 0.9625 | SRCC = 0.9618 ] +INFO:root:Average loss=0.04889477540552616 +INFO:root: +[Valid][17][UTT][ MSE = 0.1813 | LCC = 0.9633 | SRCC = 0.9628 ] +INFO:root:Average loss=0.044220374077558515 +INFO:root: +[Valid][18][UTT][ MSE = 0.1647 | LCC = 0.9639 | SRCC = 0.9630 ] +INFO:root:Average loss=0.04362687616050243 +INFO:root: +[Valid][19][UTT][ MSE = 0.1988 | LCC = 0.9619 | SRCC = 0.9616 ] +INFO:root:Average loss=0.04197048784792423 +INFO:root: +[Valid][20][UTT][ MSE = 0.1840 | LCC = 0.9625 | SRCC = 0.9624 ] +INFO:root:Average loss=0.03993211455643177 +INFO:root: +[Valid][21][UTT][ MSE = 0.1853 | LCC = 0.9645 | SRCC = 0.9637 ] +INFO:root:Average loss=0.037832736670970916 +INFO:root: +[Valid][22][UTT][ MSE = 0.1962 | LCC = 0.9619 | SRCC = 0.9602 ] +INFO:root:Average loss=0.035413215920329096 +INFO:root: +[Valid][23][UTT][ MSE = 0.1860 | LCC = 0.9638 | SRCC = 0.9631 ] +INFO:root:Average loss=0.03317230319976806 +INFO:root: +[Valid][24][UTT][ MSE = 0.2011 | LCC = 0.9632 | SRCC = 0.9623 ] +INFO:root:Average loss=0.03293273451924324 +INFO:root: +[Valid][25][UTT][ MSE = 0.1834 | LCC = 0.9639 | SRCC = 0.9634 ] +INFO:root:Average loss=0.030953787118196487 +INFO:root: +[Valid][26][UTT][ MSE = 0.1952 | LCC = 0.9633 | SRCC = 0.9628 ] +INFO:root:Average loss=0.030788473680615425 +INFO:root: +[Valid][27][UTT][ MSE = 0.1784 | LCC = 0.9639 | SRCC = 0.9636 ] +INFO:root:Average loss=0.0273007809817791 +INFO:root: +[Valid][28][UTT][ MSE = 0.1880 | LCC = 0.9641 | SRCC = 0.9634 ] +INFO:root:Average loss=0.02735107219219208 +INFO:root: +[Valid][29][UTT][ MSE = 0.2075 | LCC = 0.9639 | SRCC = 0.9635 ] +INFO:root:Average loss=0.02812838540971279 +INFO:root: +[Valid][30][UTT][ MSE = 0.1772 | LCC = 0.9648 | SRCC = 0.9639 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1868 | LCC = 0.9621 | SRCC = 0.9631 ] diff --git a/results_tencent/w2v2_large_asr_en/layer8/run2/config.gin b/results_tencent/w2v2_large_asr_en/layer8/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer8/run2/model.pt b/results_tencent/w2v2_large_asr_en/layer8/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..19ded5a94239c16cc66c71f6e41385734623b654 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b1b1ab9a578b376295d9d6a4de401de578e45e14fb8f0e0d8797824571a48ecb +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer8/run2/model_best.pt b/results_tencent/w2v2_large_asr_en/layer8/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..85ed8d6522a3261208b03f26b076622d7782197c --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b5d7950f9013e2687c45141dfb754b9a45930b6bbeaf897373bfc09fc442d378 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer8/run2/train.log b/results_tencent/w2v2_large_asr_en/layer8/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3e31535a046cb28f449511b16d03d1b147dc1da2 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4564903973340988 +INFO:root: +[Valid][0][UTT][ MSE = 0.2957 | LCC = 0.9154 | SRCC = 0.9161 ] +INFO:root:Average loss=0.20269430434703828 +INFO:root: +[Valid][1][UTT][ MSE = 0.2356 | LCC = 0.9404 | SRCC = 0.9396 ] +INFO:root:Average loss=0.1622093763947487 +INFO:root: +[Valid][2][UTT][ MSE = 0.2012 | LCC = 0.9486 | SRCC = 0.9482 ] +INFO:root:Average loss=0.13702476197481156 +INFO:root: +[Valid][3][UTT][ MSE = 0.1711 | LCC = 0.9512 | SRCC = 0.9520 ] +INFO:root:Average loss=0.12742454355955124 +INFO:root: +[Valid][4][UTT][ MSE = 0.1774 | LCC = 0.9544 | SRCC = 0.9549 ] +INFO:root:Average loss=0.10993161696195602 +INFO:root: +[Valid][5][UTT][ MSE = 0.1944 | LCC = 0.9577 | SRCC = 0.9575 ] +INFO:root:Average loss=0.10622764232754707 +INFO:root: +[Valid][6][UTT][ MSE = 0.2222 | LCC = 0.9596 | SRCC = 0.9600 ] +INFO:root:Average loss=0.09332282420992852 +INFO:root: +[Valid][7][UTT][ MSE = 0.1641 | LCC = 0.9601 | SRCC = 0.9607 ] +INFO:root:Average loss=0.09346237000823021 +INFO:root: +[Valid][8][UTT][ MSE = 0.2107 | LCC = 0.9598 | SRCC = 0.9604 ] +INFO:root:Average loss=0.08568111175298691 +INFO:root: +[Valid][9][UTT][ MSE = 0.1968 | LCC = 0.9609 | SRCC = 0.9611 ] +INFO:root:Average loss=0.07956001809239388 +INFO:root: +[Valid][10][UTT][ MSE = 0.1693 | LCC = 0.9625 | SRCC = 0.9623 ] +INFO:root:Average loss=0.07304101893305778 +INFO:root: +[Valid][11][UTT][ MSE = 0.1974 | LCC = 0.9633 | SRCC = 0.9631 ] +INFO:root:Average loss=0.06463754883408547 +INFO:root: +[Valid][12][UTT][ MSE = 0.1853 | LCC = 0.9628 | SRCC = 0.9625 ] +INFO:root:Average loss=0.06422908455133439 +INFO:root: +[Valid][13][UTT][ MSE = 0.1671 | LCC = 0.9640 | SRCC = 0.9636 ] +INFO:root:Average loss=0.05972053860127926 +INFO:root: +[Valid][14][UTT][ MSE = 0.1597 | LCC = 0.9636 | SRCC = 0.9634 ] +INFO:root:Average loss=0.06017948964238167 +INFO:root: +[Valid][15][UTT][ MSE = 0.1512 | LCC = 0.9607 | SRCC = 0.9622 ] +INFO:root:Average loss=0.052074773967266084 +INFO:root: +[Valid][16][UTT][ MSE = 0.1541 | LCC = 0.9634 | SRCC = 0.9632 ] +INFO:root:Average loss=0.04889074918627739 +INFO:root: +[Valid][17][UTT][ MSE = 0.1588 | LCC = 0.9643 | SRCC = 0.9641 ] +INFO:root:Average loss=0.04618220663070679 +INFO:root: +[Valid][18][UTT][ MSE = 0.1913 | LCC = 0.9652 | SRCC = 0.9649 ] +INFO:root:Average loss=0.04449784877896309 +INFO:root: +[Valid][19][UTT][ MSE = 0.1684 | LCC = 0.9645 | SRCC = 0.9649 ] +INFO:root:Average loss=0.04609560303390026 +INFO:root: +[Valid][20][UTT][ MSE = 0.2017 | LCC = 0.9626 | SRCC = 0.9637 ] +INFO:root:Average loss=0.041393720284104346 +INFO:root: +[Valid][21][UTT][ MSE = 0.1655 | LCC = 0.9648 | SRCC = 0.9643 ] +INFO:root:Average loss=0.035559043690562245 +INFO:root: +[Valid][22][UTT][ MSE = 0.1695 | LCC = 0.9653 | SRCC = 0.9649 ] +INFO:root:Average loss=0.03748587034642696 +INFO:root: +[Valid][23][UTT][ MSE = 0.1626 | LCC = 0.9640 | SRCC = 0.9637 ] +INFO:root:Average loss=0.03442882694303989 +INFO:root: +[Valid][24][UTT][ MSE = 0.2109 | LCC = 0.9624 | SRCC = 0.9638 ] +INFO:root:Average loss=0.036189740136265754 +INFO:root: +[Valid][25][UTT][ MSE = 0.1660 | LCC = 0.9646 | SRCC = 0.9642 ] +INFO:root:Average loss=0.03268098366260529 +INFO:root: +[Valid][26][UTT][ MSE = 0.1623 | LCC = 0.9653 | SRCC = 0.9651 ] +INFO:root:Average loss=0.02961081923544407 +INFO:root: +[Valid][27][UTT][ MSE = 0.1724 | LCC = 0.9651 | SRCC = 0.9651 ] +INFO:root:Average loss=0.030132284939289093 +INFO:root: +[Valid][28][UTT][ MSE = 0.1706 | LCC = 0.9660 | SRCC = 0.9655 ] +INFO:root:Average loss=0.03091816933453083 +INFO:root: +[Valid][29][UTT][ MSE = 0.1739 | LCC = 0.9659 | SRCC = 0.9650 ] +INFO:root:Average loss=0.02691946256905794 +INFO:root: +[Valid][30][UTT][ MSE = 0.1878 | LCC = 0.9641 | SRCC = 0.9650 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1843 | LCC = 0.9622 | SRCC = 0.9629 ] diff --git a/results_tencent/w2v2_large_asr_en/layer8/run3/config.gin b/results_tencent/w2v2_large_asr_en/layer8/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer8/run3/model.pt b/results_tencent/w2v2_large_asr_en/layer8/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..02ccb41560cee80dc3289ebd022b6a6acc95021d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e4f1b144cb4fc9d49101fde740905b5904d35b2e54555dc6e71c5df38b85e6eb +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer8/run3/model_best.pt b/results_tencent/w2v2_large_asr_en/layer8/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..201e6e4b7f2d985f26452616ccd7f4d0b1259caf --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2ea79b53098652bd0859ce6df5faf38bfbc9a1fa47e5b576e3647a053d5c5533 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer8/run3/train.log b/results_tencent/w2v2_large_asr_en/layer8/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..247f62ed165265a5e7705e4bd13111fe6fadc0ed --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5933385223150254 +INFO:root: +[Valid][0][UTT][ MSE = 0.3669 | LCC = 0.9049 | SRCC = 0.9042 ] +INFO:root:Average loss=0.2081800788640976 +INFO:root: +[Valid][1][UTT][ MSE = 0.2702 | LCC = 0.9369 | SRCC = 0.9372 ] +INFO:root:Average loss=0.17481973278522492 +INFO:root: +[Valid][2][UTT][ MSE = 0.2846 | LCC = 0.9453 | SRCC = 0.9468 ] +INFO:root:Average loss=0.14443707782030105 +INFO:root: +[Valid][3][UTT][ MSE = 0.2364 | LCC = 0.9496 | SRCC = 0.9499 ] +INFO:root:Average loss=0.12407432425022125 +INFO:root: +[Valid][4][UTT][ MSE = 0.2163 | LCC = 0.9520 | SRCC = 0.9525 ] +INFO:root:Average loss=0.11037368577718734 +INFO:root: +[Valid][5][UTT][ MSE = 0.2019 | LCC = 0.9554 | SRCC = 0.9557 ] +INFO:root:Average loss=0.1037197342813015 +INFO:root: +[Valid][6][UTT][ MSE = 0.2082 | LCC = 0.9564 | SRCC = 0.9568 ] +INFO:root:Average loss=0.09253348553180694 +INFO:root: +[Valid][7][UTT][ MSE = 0.2081 | LCC = 0.9584 | SRCC = 0.9589 ] +INFO:root:Average loss=0.09111560261249542 +INFO:root: +[Valid][8][UTT][ MSE = 0.1883 | LCC = 0.9596 | SRCC = 0.9598 ] +INFO:root:Average loss=0.0839040909409523 +INFO:root: +[Valid][9][UTT][ MSE = 0.1688 | LCC = 0.9584 | SRCC = 0.9592 ] +INFO:root:Average loss=0.07781177660822869 +INFO:root: +[Valid][10][UTT][ MSE = 0.1841 | LCC = 0.9606 | SRCC = 0.9600 ] +INFO:root:Average loss=0.07276378986239433 +INFO:root: +[Valid][11][UTT][ MSE = 0.1853 | LCC = 0.9621 | SRCC = 0.9616 ] +INFO:root:Average loss=0.0659175574183464 +INFO:root: +[Valid][12][UTT][ MSE = 0.1944 | LCC = 0.9619 | SRCC = 0.9621 ] +INFO:root:Average loss=0.06246680951118469 +INFO:root: +[Valid][13][UTT][ MSE = 0.1706 | LCC = 0.9616 | SRCC = 0.9614 ] +INFO:root:Average loss=0.05945025384426117 +INFO:root: +[Valid][14][UTT][ MSE = 0.1938 | LCC = 0.9622 | SRCC = 0.9623 ] +INFO:root:Average loss=0.06112117832899094 +INFO:root: +[Valid][15][UTT][ MSE = 0.2041 | LCC = 0.9625 | SRCC = 0.9619 ] +INFO:root:Average loss=0.0593561005294323 +INFO:root: +[Valid][16][UTT][ MSE = 0.1782 | LCC = 0.9620 | SRCC = 0.9617 ] +INFO:root:Average loss=0.05120430760085583 +INFO:root: +[Valid][17][UTT][ MSE = 0.2113 | LCC = 0.9624 | SRCC = 0.9623 ] +INFO:root:Average loss=0.045796249613165856 +INFO:root: +[Valid][18][UTT][ MSE = 0.1867 | LCC = 0.9629 | SRCC = 0.9623 ] +INFO:root:Average loss=0.043513678789138796 +INFO:root: +[Valid][19][UTT][ MSE = 0.1742 | LCC = 0.9634 | SRCC = 0.9630 ] +INFO:root:Average loss=0.04203731814026833 +INFO:root: +[Valid][20][UTT][ MSE = 0.1758 | LCC = 0.9644 | SRCC = 0.9636 ] +INFO:root:Average loss=0.039740458473563195 +INFO:root: +[Valid][21][UTT][ MSE = 0.1766 | LCC = 0.9632 | SRCC = 0.9631 ] +INFO:root:Average loss=0.04094577912986279 +INFO:root: +[Valid][22][UTT][ MSE = 0.1885 | LCC = 0.9631 | SRCC = 0.9627 ] +INFO:root:Average loss=0.038807891279459 +INFO:root: +[Valid][23][UTT][ MSE = 0.1993 | LCC = 0.9635 | SRCC = 0.9632 ] +INFO:root:Average loss=0.03573849248886109 +INFO:root: +[Valid][24][UTT][ MSE = 0.1823 | LCC = 0.9640 | SRCC = 0.9634 ] +INFO:root:Average loss=0.0319470532387495 +INFO:root: +[Valid][25][UTT][ MSE = 0.1920 | LCC = 0.9641 | SRCC = 0.9638 ] +INFO:root:Average loss=0.03196916301548481 +INFO:root: +[Valid][26][UTT][ MSE = 0.1961 | LCC = 0.9639 | SRCC = 0.9637 ] +INFO:root:Average loss=0.034076146990060806 +INFO:root: +[Valid][27][UTT][ MSE = 0.1953 | LCC = 0.9634 | SRCC = 0.9630 ] +INFO:root:Average loss=0.031364118114113805 +INFO:root: +[Valid][28][UTT][ MSE = 0.1838 | LCC = 0.9639 | SRCC = 0.9634 ] +INFO:root:Average loss=0.028194803468883037 +INFO:root: +[Valid][29][UTT][ MSE = 0.1852 | LCC = 0.9635 | SRCC = 0.9633 ] +INFO:root:Average loss=0.02743151694536209 +INFO:root: +[Valid][30][UTT][ MSE = 0.1793 | LCC = 0.9636 | SRCC = 0.9632 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1854 | LCC = 0.9623 | SRCC = 0.9632 ] diff --git a/results_tencent/w2v2_large_asr_en/layer8/run4/config.gin b/results_tencent/w2v2_large_asr_en/layer8/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_en/layer8/run4/model.pt b/results_tencent/w2v2_large_asr_en/layer8/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f1025bca08e2881a9cbfb74d3b6f95559fb27f30 --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ef1fe7959f8f64cd9169fb04a8e89feb069737886393ee1283fd4a43b00bc396 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_en/layer8/run4/model_best.pt b/results_tencent/w2v2_large_asr_en/layer8/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..caa934445b8a75daed0f3973a159c691bb8f5a3e --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ccd4571bde97d81657bcea64e8f312315f982ffcdb8c90ded6a2ede4c416bfbc +size 3405062 diff --git a/results_tencent/w2v2_large_asr_en/layer8/run4/train.log b/results_tencent/w2v2_large_asr_en/layer8/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..935eb2addcf711987b3a1926b5e287f5c337d85e --- /dev/null +++ b/results_tencent/w2v2_large_asr_en/layer8/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.468673248052597 +INFO:root: +[Valid][0][UTT][ MSE = 0.3274 | LCC = 0.9208 | SRCC = 0.9201 ] +INFO:root:Average loss=0.20766886359453202 +INFO:root: +[Valid][1][UTT][ MSE = 0.3069 | LCC = 0.9371 | SRCC = 0.9361 ] +INFO:root:Average loss=0.16355902171134948 +INFO:root: +[Valid][2][UTT][ MSE = 0.2414 | LCC = 0.9461 | SRCC = 0.9463 ] +INFO:root:Average loss=0.13893747103214263 +INFO:root: +[Valid][3][UTT][ MSE = 0.2348 | LCC = 0.9508 | SRCC = 0.9512 ] +INFO:root:Average loss=0.11781968307495118 +INFO:root: +[Valid][4][UTT][ MSE = 0.2480 | LCC = 0.9474 | SRCC = 0.9483 ] +INFO:root:Average loss=0.11170800986886024 +INFO:root: +[Valid][5][UTT][ MSE = 0.2666 | LCC = 0.9526 | SRCC = 0.9544 ] +INFO:root:Average loss=0.10396323025226593 +INFO:root: +[Valid][6][UTT][ MSE = 0.2388 | LCC = 0.9562 | SRCC = 0.9576 ] +INFO:root:Average loss=0.09261626896262169 +INFO:root: +[Valid][7][UTT][ MSE = 0.2121 | LCC = 0.9589 | SRCC = 0.9588 ] +INFO:root:Average loss=0.08722768440842628 +INFO:root: +[Valid][8][UTT][ MSE = 0.1949 | LCC = 0.9582 | SRCC = 0.9584 ] +INFO:root:Average loss=0.08335587108135223 +INFO:root: +[Valid][9][UTT][ MSE = 0.2030 | LCC = 0.9587 | SRCC = 0.9588 ] +INFO:root:Average loss=0.07329895770549774 +INFO:root: +[Valid][10][UTT][ MSE = 0.2299 | LCC = 0.9574 | SRCC = 0.9586 ] +INFO:root:Average loss=0.0668146134018898 +INFO:root: +[Valid][11][UTT][ MSE = 0.1626 | LCC = 0.9599 | SRCC = 0.9605 ] +INFO:root:Average loss=0.061949410900473595 +INFO:root: +[Valid][12][UTT][ MSE = 0.1744 | LCC = 0.9594 | SRCC = 0.9587 ] +INFO:root:Average loss=0.0644526814520359 +INFO:root: +[Valid][13][UTT][ MSE = 0.1702 | LCC = 0.9595 | SRCC = 0.9595 ] +INFO:root:Average loss=0.05650112204253674 +INFO:root: +[Valid][14][UTT][ MSE = 0.1910 | LCC = 0.9605 | SRCC = 0.9598 ] +INFO:root:Average loss=0.05352893011271954 +INFO:root: +[Valid][15][UTT][ MSE = 0.1610 | LCC = 0.9610 | SRCC = 0.9606 ] +INFO:root:Average loss=0.04975971084833145 +INFO:root: +[Valid][16][UTT][ MSE = 0.1739 | LCC = 0.9606 | SRCC = 0.9607 ] +INFO:root:Average loss=0.04819870586693287 +INFO:root: +[Valid][17][UTT][ MSE = 0.1714 | LCC = 0.9609 | SRCC = 0.9610 ] +INFO:root:Average loss=0.0484014754742384 +INFO:root: +[Valid][18][UTT][ MSE = 0.1813 | LCC = 0.9619 | SRCC = 0.9617 ] +INFO:root:Average loss=0.03994495619833469 +INFO:root: +[Valid][19][UTT][ MSE = 0.1850 | LCC = 0.9612 | SRCC = 0.9610 ] +INFO:root:Average loss=0.03848662582039833 +INFO:root: +[Valid][20][UTT][ MSE = 0.1787 | LCC = 0.9614 | SRCC = 0.9613 ] +INFO:root:Average loss=0.037732948571443554 +INFO:root: +[Valid][21][UTT][ MSE = 0.1774 | LCC = 0.9601 | SRCC = 0.9603 ] +INFO:root:Average loss=0.037082235977053644 +INFO:root: +[Valid][22][UTT][ MSE = 0.2139 | LCC = 0.9611 | SRCC = 0.9611 ] +INFO:root:Average loss=0.034830958530306814 +INFO:root: +[Valid][23][UTT][ MSE = 0.1689 | LCC = 0.9615 | SRCC = 0.9618 ] +INFO:root:Average loss=0.03369467629492283 +INFO:root: +[Valid][24][UTT][ MSE = 0.1883 | LCC = 0.9618 | SRCC = 0.9617 ] +INFO:root:Average loss=0.031339191183447836 +INFO:root: +[Valid][25][UTT][ MSE = 0.1693 | LCC = 0.9624 | SRCC = 0.9621 ] +INFO:root:Average loss=0.03274203526973724 +INFO:root: +[Valid][26][UTT][ MSE = 0.1828 | LCC = 0.9621 | SRCC = 0.9617 ] +INFO:root:Average loss=0.02816258670389652 +INFO:root: +[Valid][27][UTT][ MSE = 0.1824 | LCC = 0.9621 | SRCC = 0.9615 ] +INFO:root:Average loss=0.02812367734313011 +INFO:root: +[Valid][28][UTT][ MSE = 0.2020 | LCC = 0.9613 | SRCC = 0.9613 ] +INFO:root:Average loss=0.027844575479626654 +INFO:root: +[Valid][29][UTT][ MSE = 0.1850 | LCC = 0.9623 | SRCC = 0.9617 ] +INFO:root:Average loss=0.02660474905371666 +INFO:root: +[Valid][30][UTT][ MSE = 0.1791 | LCC = 0.9617 | SRCC = 0.9615 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1755 | LCC = 0.9616 | SRCC = 0.9622 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run0/config.gin b/results_tencent/w2v2_large_asr_zh/layer0/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run0/model.pt b/results_tencent/w2v2_large_asr_zh/layer0/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..dc433bd4658427e4cdc658db97d3e545b74b5bae --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a26a56eeca37f9c6f6cd8497f2eb4f3459f5b5652df82d4edf0ff58815c2a8e3 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run0/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer0/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a21f18eadd1ff544d459b7204146a9db47798c9f --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:88503c392ea457a5a5d2796840e3f4370ce6635b69d2237853f6f2ab4b5e4786 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run0/train.log b/results_tencent/w2v2_large_asr_zh/layer0/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3db518a8bcbc0891982e46e8540f792bfe6e2ae3 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6681449954509735 +INFO:root: +[Valid][0][UTT][ MSE = 0.5345 | LCC = 0.8547 | SRCC = 0.8581 ] +INFO:root:Average loss=0.4356583127975464 +INFO:root: +[Valid][1][UTT][ MSE = 0.3879 | LCC = 0.8802 | SRCC = 0.8803 ] +INFO:root:Average loss=0.338454541683197 +INFO:root: +[Valid][2][UTT][ MSE = 0.3678 | LCC = 0.9004 | SRCC = 0.8988 ] +INFO:root:Average loss=0.30310092139244077 +INFO:root: +[Valid][3][UTT][ MSE = 0.3587 | LCC = 0.9049 | SRCC = 0.9022 ] +INFO:root:Average loss=0.2804183163642883 +INFO:root: +[Valid][4][UTT][ MSE = 0.3907 | LCC = 0.9051 | SRCC = 0.9073 ] +INFO:root:Average loss=0.2539649299383163 +INFO:root: +[Valid][5][UTT][ MSE = 0.3986 | LCC = 0.9081 | SRCC = 0.9100 ] +INFO:root:Average loss=0.22332588696479796 +INFO:root: +[Valid][6][UTT][ MSE = 0.3712 | LCC = 0.9232 | SRCC = 0.9248 ] +INFO:root:Average loss=0.23375817215442657 +INFO:root: +[Valid][7][UTT][ MSE = 0.3057 | LCC = 0.9200 | SRCC = 0.9199 ] +INFO:root:Average loss=0.2218710550069809 +INFO:root: +[Valid][8][UTT][ MSE = 0.2919 | LCC = 0.9264 | SRCC = 0.9262 ] +INFO:root:Average loss=0.19481675189733505 +INFO:root: +[Valid][9][UTT][ MSE = 0.2706 | LCC = 0.9235 | SRCC = 0.9261 ] +INFO:root:Average loss=0.18566503816843033 +INFO:root: +[Valid][10][UTT][ MSE = 0.2747 | LCC = 0.9293 | SRCC = 0.9296 ] +INFO:root:Average loss=0.1802762544155121 +INFO:root: +[Valid][11][UTT][ MSE = 0.3049 | LCC = 0.9294 | SRCC = 0.9302 ] +INFO:root:Average loss=0.17138692551851273 +INFO:root: +[Valid][12][UTT][ MSE = 0.3047 | LCC = 0.9312 | SRCC = 0.9313 ] +INFO:root:Average loss=0.16226623779535293 +INFO:root: +[Valid][13][UTT][ MSE = 0.2608 | LCC = 0.9302 | SRCC = 0.9311 ] +INFO:root:Average loss=0.15335773068666458 +INFO:root: +[Valid][14][UTT][ MSE = 0.2691 | LCC = 0.9302 | SRCC = 0.9303 ] +INFO:root:Average loss=0.15015802186727523 +INFO:root: +[Valid][15][UTT][ MSE = 0.3591 | LCC = 0.9324 | SRCC = 0.9344 ] +INFO:root:Average loss=0.14574045842885971 +INFO:root: +[Valid][16][UTT][ MSE = 0.2963 | LCC = 0.9329 | SRCC = 0.9341 ] +INFO:root:Average loss=0.1397637113928795 +INFO:root: +[Valid][17][UTT][ MSE = 0.2842 | LCC = 0.9360 | SRCC = 0.9363 ] +INFO:root:Average loss=0.14045220255851745 +INFO:root: +[Valid][18][UTT][ MSE = 0.3012 | LCC = 0.9343 | SRCC = 0.9352 ] +INFO:root:Average loss=0.13350885260105133 +INFO:root: +[Valid][19][UTT][ MSE = 0.2754 | LCC = 0.9351 | SRCC = 0.9358 ] +INFO:root:Average loss=0.12293113759160042 +INFO:root: +[Valid][20][UTT][ MSE = 0.2258 | LCC = 0.9357 | SRCC = 0.9370 ] +INFO:root:Average loss=0.1152601221203804 +INFO:root: +[Valid][21][UTT][ MSE = 0.2449 | LCC = 0.9379 | SRCC = 0.9379 ] +INFO:root:Average loss=0.11558894449472427 +INFO:root: +[Valid][22][UTT][ MSE = 0.2349 | LCC = 0.9330 | SRCC = 0.9331 ] +INFO:root:Average loss=0.11086820414662361 +INFO:root: +[Valid][23][UTT][ MSE = 0.2177 | LCC = 0.9385 | SRCC = 0.9388 ] +INFO:root:Average loss=0.1075833226442337 +INFO:root: +[Valid][24][UTT][ MSE = 0.2528 | LCC = 0.9383 | SRCC = 0.9391 ] +INFO:root:Average loss=0.10966210454702377 +INFO:root: +[Valid][25][UTT][ MSE = 0.2780 | LCC = 0.9370 | SRCC = 0.9374 ] +INFO:root:Average loss=0.10150853097438813 +INFO:root: +[Valid][26][UTT][ MSE = 0.2457 | LCC = 0.9401 | SRCC = 0.9405 ] +INFO:root:Average loss=0.10056563666462898 +INFO:root: +[Valid][27][UTT][ MSE = 0.2403 | LCC = 0.9401 | SRCC = 0.9407 ] +INFO:root:Average loss=0.09606211775541305 +INFO:root: +[Valid][28][UTT][ MSE = 0.2615 | LCC = 0.9394 | SRCC = 0.9391 ] +INFO:root:Average loss=0.09326884958148003 +INFO:root: +[Valid][29][UTT][ MSE = 0.2242 | LCC = 0.9394 | SRCC = 0.9397 ] +INFO:root:Average loss=0.09744221231341361 +INFO:root: +[Valid][30][UTT][ MSE = 0.2653 | LCC = 0.9412 | SRCC = 0.9411 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2830 | LCC = 0.9373 | SRCC = 0.9376 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run1/config.gin b/results_tencent/w2v2_large_asr_zh/layer0/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run1/model.pt b/results_tencent/w2v2_large_asr_zh/layer0/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0594c0f4c027b1ecb3bca9f4249ad71af853173e --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2d9ae0d9fa700fa8e324c35998c015e224dbd65cf9677e3991a88ccac5ca8e3a +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run1/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer0/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4c874e84790b32c0483b8800fca01f2c662b7a60 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e5f31c8835afe1e2e7c3218e7aa5d91528e488daa5d10a88c4bf7fa0eea438a +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run1/train.log b/results_tencent/w2v2_large_asr_zh/layer0/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..955989a0fd59d77a6873e2d537256c931c017f67 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6774428796768188 +INFO:root: +[Valid][0][UTT][ MSE = 0.4484 | LCC = 0.8644 | SRCC = 0.8628 ] +INFO:root:Average loss=0.36737520289421083 +INFO:root: +[Valid][1][UTT][ MSE = 0.3548 | LCC = 0.8996 | SRCC = 0.8991 ] +INFO:root:Average loss=0.34033866930007933 +INFO:root: +[Valid][2][UTT][ MSE = 0.3445 | LCC = 0.9024 | SRCC = 0.9046 ] +INFO:root:Average loss=0.3222711981534958 +INFO:root: +[Valid][3][UTT][ MSE = 0.3618 | LCC = 0.9096 | SRCC = 0.9095 ] +INFO:root:Average loss=0.2557679291963577 +INFO:root: +[Valid][4][UTT][ MSE = 0.3154 | LCC = 0.9161 | SRCC = 0.9188 ] +INFO:root:Average loss=0.25259732764959336 +INFO:root: +[Valid][5][UTT][ MSE = 0.3968 | LCC = 0.9152 | SRCC = 0.9157 ] +INFO:root:Average loss=0.22093553566932678 +INFO:root: +[Valid][6][UTT][ MSE = 0.2458 | LCC = 0.9216 | SRCC = 0.9240 ] +INFO:root:Average loss=0.20640818375349046 +INFO:root: +[Valid][7][UTT][ MSE = 0.2961 | LCC = 0.9236 | SRCC = 0.9243 ] +INFO:root:Average loss=0.20242829990386962 +INFO:root: +[Valid][8][UTT][ MSE = 0.2606 | LCC = 0.9258 | SRCC = 0.9275 ] +INFO:root:Average loss=0.18776507258415223 +INFO:root: +[Valid][9][UTT][ MSE = 0.2838 | LCC = 0.9260 | SRCC = 0.9272 ] +INFO:root:Average loss=0.176601620554924 +INFO:root: +[Valid][10][UTT][ MSE = 0.2740 | LCC = 0.9268 | SRCC = 0.9273 ] +INFO:root:Average loss=0.16918697345256806 +INFO:root: +[Valid][11][UTT][ MSE = 0.2645 | LCC = 0.9331 | SRCC = 0.9341 ] +INFO:root:Average loss=0.16208453044295312 +INFO:root: +[Valid][12][UTT][ MSE = 0.2373 | LCC = 0.9285 | SRCC = 0.9300 ] +INFO:root:Average loss=0.1541780369281769 +INFO:root: +[Valid][13][UTT][ MSE = 0.2822 | LCC = 0.9305 | SRCC = 0.9306 ] +INFO:root:Average loss=0.15003037375211717 +INFO:root: +[Valid][14][UTT][ MSE = 0.2662 | LCC = 0.9307 | SRCC = 0.9319 ] +INFO:root:Average loss=0.14604324996471404 +INFO:root: +[Valid][15][UTT][ MSE = 0.2921 | LCC = 0.9344 | SRCC = 0.9349 ] +INFO:root:Average loss=0.14078605648875236 +INFO:root: +[Valid][16][UTT][ MSE = 0.2641 | LCC = 0.9346 | SRCC = 0.9363 ] +INFO:root:Average loss=0.13616577970981597 +INFO:root: +[Valid][17][UTT][ MSE = 0.2628 | LCC = 0.9375 | SRCC = 0.9388 ] +INFO:root:Average loss=0.1256192890405655 +INFO:root: +[Valid][18][UTT][ MSE = 0.2800 | LCC = 0.9354 | SRCC = 0.9384 ] +INFO:root:Average loss=0.12565742042660713 +INFO:root: +[Valid][19][UTT][ MSE = 0.2394 | LCC = 0.9370 | SRCC = 0.9380 ] +INFO:root:Average loss=0.12321622985601426 +INFO:root: +[Valid][20][UTT][ MSE = 0.2493 | LCC = 0.9391 | SRCC = 0.9400 ] +INFO:root:Average loss=0.11871963250637055 +INFO:root: +[Valid][21][UTT][ MSE = 0.2693 | LCC = 0.9373 | SRCC = 0.9399 ] +INFO:root:Average loss=0.11325651663541794 +INFO:root: +[Valid][22][UTT][ MSE = 0.2413 | LCC = 0.9398 | SRCC = 0.9408 ] +INFO:root:Average loss=0.11122071394324302 +INFO:root: +[Valid][23][UTT][ MSE = 0.2289 | LCC = 0.9414 | SRCC = 0.9432 ] +INFO:root:Average loss=0.10656094828248024 +INFO:root: +[Valid][24][UTT][ MSE = 0.2477 | LCC = 0.9391 | SRCC = 0.9400 ] +INFO:root:Average loss=0.1046516716182232 +INFO:root: +[Valid][25][UTT][ MSE = 0.2490 | LCC = 0.9423 | SRCC = 0.9433 ] +INFO:root:Average loss=0.10479879027605057 +INFO:root: +[Valid][26][UTT][ MSE = 0.2381 | LCC = 0.9420 | SRCC = 0.9430 ] +INFO:root:Average loss=0.10390762928128243 +INFO:root: +[Valid][27][UTT][ MSE = 0.2074 | LCC = 0.9425 | SRCC = 0.9432 ] +INFO:root:Average loss=0.09662379071116448 +INFO:root: +[Valid][28][UTT][ MSE = 0.2263 | LCC = 0.9423 | SRCC = 0.9425 ] +INFO:root:Average loss=0.09885993927717208 +INFO:root: +[Valid][29][UTT][ MSE = 0.2251 | LCC = 0.9409 | SRCC = 0.9418 ] +INFO:root:Average loss=0.09621339827775956 +INFO:root: +[Valid][30][UTT][ MSE = 0.2166 | LCC = 0.9434 | SRCC = 0.9443 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2210 | LCC = 0.9424 | SRCC = 0.9430 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run2/config.gin b/results_tencent/w2v2_large_asr_zh/layer0/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run2/model.pt b/results_tencent/w2v2_large_asr_zh/layer0/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ba157fd8cb96379058ec411ae73a53c42a4da8ab --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:88a0ea401486cb2b9a7fc4f7c05d885c67ffd4fd465dd3cccd0d6989d598b398 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run2/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer0/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ad40b00dc74cdc2853e432b0957402f080779f65 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:65e16b19d9f06aefcfec7382ec2d9552e221a0286337db49812128d0854bbaa5 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run2/train.log b/results_tencent/w2v2_large_asr_zh/layer0/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..fbaf44e1f094abc9ed8d14adf486dee0507deb09 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.671284348487854 +INFO:root: +[Valid][0][UTT][ MSE = 0.5515 | LCC = 0.8601 | SRCC = 0.8556 ] +INFO:root:Average loss=0.37865333318710326 +INFO:root: +[Valid][1][UTT][ MSE = 0.3919 | LCC = 0.8910 | SRCC = 0.8933 ] +INFO:root:Average loss=0.32285331678390505 +INFO:root: +[Valid][2][UTT][ MSE = 0.3473 | LCC = 0.9082 | SRCC = 0.9104 ] +INFO:root:Average loss=0.27132480597496034 +INFO:root: +[Valid][3][UTT][ MSE = 0.3118 | LCC = 0.9124 | SRCC = 0.9139 ] +INFO:root:Average loss=0.24931260979175568 +INFO:root: +[Valid][4][UTT][ MSE = 0.3075 | LCC = 0.9198 | SRCC = 0.9209 ] +INFO:root:Average loss=0.23649326264858245 +INFO:root: +[Valid][5][UTT][ MSE = 0.2954 | LCC = 0.9231 | SRCC = 0.9244 ] +INFO:root:Average loss=0.2086027085185051 +INFO:root: +[Valid][6][UTT][ MSE = 0.2815 | LCC = 0.9251 | SRCC = 0.9264 ] +INFO:root:Average loss=0.20666488236188887 +INFO:root: +[Valid][7][UTT][ MSE = 0.3094 | LCC = 0.9241 | SRCC = 0.9264 ] +INFO:root:Average loss=0.19049295830726623 +INFO:root: +[Valid][8][UTT][ MSE = 0.2852 | LCC = 0.9188 | SRCC = 0.9208 ] +INFO:root:Average loss=0.19248049592971803 +INFO:root: +[Valid][9][UTT][ MSE = 0.3731 | LCC = 0.9269 | SRCC = 0.9306 ] +INFO:root:Average loss=0.1792094714641571 +INFO:root: +[Valid][10][UTT][ MSE = 0.2674 | LCC = 0.9314 | SRCC = 0.9331 ] +INFO:root:Average loss=0.1749591562151909 +INFO:root: +[Valid][11][UTT][ MSE = 0.2510 | LCC = 0.9320 | SRCC = 0.9336 ] +INFO:root:Average loss=0.16173267805576325 +INFO:root: +[Valid][12][UTT][ MSE = 0.3101 | LCC = 0.9315 | SRCC = 0.9338 ] +INFO:root:Average loss=0.1628376722931862 +INFO:root: +[Valid][13][UTT][ MSE = 0.2708 | LCC = 0.9322 | SRCC = 0.9337 ] +INFO:root:Average loss=0.15016700392961502 +INFO:root: +[Valid][14][UTT][ MSE = 0.2615 | LCC = 0.9348 | SRCC = 0.9364 ] +INFO:root:Average loss=0.14576880395412445 +INFO:root: +[Valid][15][UTT][ MSE = 0.2626 | LCC = 0.9361 | SRCC = 0.9381 ] +INFO:root:Average loss=0.14273862248659133 +INFO:root: +[Valid][16][UTT][ MSE = 0.2584 | LCC = 0.9356 | SRCC = 0.9370 ] +INFO:root:Average loss=0.1345158749818802 +INFO:root: +[Valid][17][UTT][ MSE = 0.2625 | LCC = 0.9359 | SRCC = 0.9374 ] +INFO:root:Average loss=0.12624723452329636 +INFO:root: +[Valid][18][UTT][ MSE = 0.3389 | LCC = 0.9327 | SRCC = 0.9360 ] +INFO:root:Average loss=0.12553563064336776 +INFO:root: +[Valid][19][UTT][ MSE = 0.2600 | LCC = 0.9369 | SRCC = 0.9386 ] +INFO:root:Average loss=0.12917184120416642 +INFO:root: +[Valid][20][UTT][ MSE = 0.2694 | LCC = 0.9398 | SRCC = 0.9411 ] +INFO:root:Average loss=0.11657251730561256 +INFO:root: +[Valid][21][UTT][ MSE = 0.2460 | LCC = 0.9414 | SRCC = 0.9424 ] +INFO:root:Average loss=0.11403440564870834 +INFO:root: +[Valid][22][UTT][ MSE = 0.2396 | LCC = 0.9391 | SRCC = 0.9403 ] +INFO:root:Average loss=0.1132515015900135 +INFO:root: +[Valid][23][UTT][ MSE = 0.2467 | LCC = 0.9408 | SRCC = 0.9411 ] +INFO:root:Average loss=0.10445835381746292 +INFO:root: +[Valid][24][UTT][ MSE = 0.2739 | LCC = 0.9388 | SRCC = 0.9399 ] +INFO:root:Average loss=0.10128957387804985 +INFO:root: +[Valid][25][UTT][ MSE = 0.2383 | LCC = 0.9419 | SRCC = 0.9426 ] +INFO:root:Average loss=0.09798848980665206 +INFO:root: +[Valid][26][UTT][ MSE = 0.2442 | LCC = 0.9419 | SRCC = 0.9427 ] +INFO:root:Average loss=0.09886642688512802 +INFO:root: +[Valid][27][UTT][ MSE = 0.2610 | LCC = 0.9413 | SRCC = 0.9427 ] +INFO:root:Average loss=0.09761891204118729 +INFO:root: +[Valid][28][UTT][ MSE = 0.2330 | LCC = 0.9419 | SRCC = 0.9429 ] +INFO:root:Average loss=0.09060594302415848 +INFO:root: +[Valid][29][UTT][ MSE = 0.2343 | LCC = 0.9422 | SRCC = 0.9432 ] +INFO:root:Average loss=0.08746518477797509 +INFO:root: +[Valid][30][UTT][ MSE = 0.2255 | LCC = 0.9435 | SRCC = 0.9443 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2395 | LCC = 0.9423 | SRCC = 0.9422 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run3/config.gin b/results_tencent/w2v2_large_asr_zh/layer0/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run3/model.pt b/results_tencent/w2v2_large_asr_zh/layer0/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..60958b0af1d9826c7f555c5495d360fb600d54ae --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bddbafea4a68d115b220bc5d80583397903e2fd77227f55596502f5a8c16d89b +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run3/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer0/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..5e3eeff43574327c2fff7db96f96803849555da6 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b878163961497941b7f49cfa1a7f9acf08676ea7bf49da85887d7f9677a0e79c +size 3404998 diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run3/train.log b/results_tencent/w2v2_large_asr_zh/layer0/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e7a0f4114c1b7a4462b7dd3b5bdbe976b20da342 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6378090085983277 +INFO:root: +[Valid][0][UTT][ MSE = 0.4466 | LCC = 0.8654 | SRCC = 0.8597 ] +INFO:root:Average loss=0.38814903116226196 +INFO:root: +[Valid][1][UTT][ MSE = 0.4618 | LCC = 0.8849 | SRCC = 0.8848 ] +INFO:root:Average loss=0.34171477949619294 +INFO:root: +[Valid][2][UTT][ MSE = 0.3472 | LCC = 0.9051 | SRCC = 0.9046 ] +INFO:root:Average loss=0.3100585802793503 +INFO:root: +[Valid][3][UTT][ MSE = 0.3809 | LCC = 0.8938 | SRCC = 0.8931 ] +INFO:root:Average loss=0.2890544899702072 +INFO:root: +[Valid][4][UTT][ MSE = 0.3729 | LCC = 0.9144 | SRCC = 0.9181 ] +INFO:root:Average loss=0.25150664389133454 +INFO:root: +[Valid][5][UTT][ MSE = 0.3514 | LCC = 0.9194 | SRCC = 0.9202 ] +INFO:root:Average loss=0.2318825935125351 +INFO:root: +[Valid][6][UTT][ MSE = 0.3086 | LCC = 0.9209 | SRCC = 0.9219 ] +INFO:root:Average loss=0.21250194734334946 +INFO:root: +[Valid][7][UTT][ MSE = 0.2951 | LCC = 0.9239 | SRCC = 0.9243 ] +INFO:root:Average loss=0.20158285307884216 +INFO:root: +[Valid][8][UTT][ MSE = 0.3114 | LCC = 0.9260 | SRCC = 0.9287 ] +INFO:root:Average loss=0.18736524367332458 +INFO:root: +[Valid][9][UTT][ MSE = 0.2841 | LCC = 0.9259 | SRCC = 0.9260 ] +INFO:root:Average loss=0.17559041440486908 +INFO:root: +[Valid][10][UTT][ MSE = 0.2723 | LCC = 0.9288 | SRCC = 0.9295 ] +INFO:root:Average loss=0.16201377695798874 +INFO:root: +[Valid][11][UTT][ MSE = 0.2789 | LCC = 0.9309 | SRCC = 0.9319 ] +INFO:root:Average loss=0.1669875425696373 +INFO:root: +[Valid][12][UTT][ MSE = 0.2633 | LCC = 0.9320 | SRCC = 0.9334 ] +INFO:root:Average loss=0.16092655509710313 +INFO:root: +[Valid][13][UTT][ MSE = 0.3908 | LCC = 0.9098 | SRCC = 0.9147 ] +INFO:root:Average loss=0.15435853493213653 +INFO:root: +[Valid][14][UTT][ MSE = 0.2327 | LCC = 0.9338 | SRCC = 0.9345 ] +INFO:root:Average loss=0.14266090387105942 +INFO:root: +[Valid][15][UTT][ MSE = 0.2385 | LCC = 0.9357 | SRCC = 0.9359 ] +INFO:root:Average loss=0.1364672564268112 +INFO:root: +[Valid][16][UTT][ MSE = 0.2289 | LCC = 0.9367 | SRCC = 0.9375 ] +INFO:root:Average loss=0.1438102298974991 +INFO:root: +[Valid][17][UTT][ MSE = 0.2592 | LCC = 0.9354 | SRCC = 0.9356 ] +INFO:root:Average loss=0.1323671189546585 +INFO:root: +[Valid][18][UTT][ MSE = 0.2486 | LCC = 0.9376 | SRCC = 0.9383 ] +INFO:root:Average loss=0.12522184282541274 +INFO:root: +[Valid][19][UTT][ MSE = 0.2808 | LCC = 0.9387 | SRCC = 0.9397 ] +INFO:root:Average loss=0.12643874680995942 +INFO:root: +[Valid][20][UTT][ MSE = 0.2697 | LCC = 0.9385 | SRCC = 0.9387 ] +INFO:root:Average loss=0.11385537025332451 +INFO:root: +[Valid][21][UTT][ MSE = 0.2456 | LCC = 0.9378 | SRCC = 0.9391 ] +INFO:root:Average loss=0.11057625186443329 +INFO:root: +[Valid][22][UTT][ MSE = 0.2501 | LCC = 0.9326 | SRCC = 0.9337 ] +INFO:root:Average loss=0.12147750186920166 +INFO:root: +[Valid][23][UTT][ MSE = 0.2418 | LCC = 0.9330 | SRCC = 0.9329 ] +INFO:root:Average loss=0.10804466310143471 +INFO:root: +[Valid][24][UTT][ MSE = 0.2495 | LCC = 0.9389 | SRCC = 0.9396 ] +INFO:root:Average loss=0.10328111261129379 +INFO:root: +[Valid][25][UTT][ MSE = 0.2604 | LCC = 0.9404 | SRCC = 0.9418 ] +INFO:root:Average loss=0.10042049905657768 +INFO:root: +[Valid][26][UTT][ MSE = 0.2601 | LCC = 0.9394 | SRCC = 0.9396 ] +INFO:root:Average loss=0.09856345680356025 +INFO:root: +[Valid][27][UTT][ MSE = 0.2273 | LCC = 0.9405 | SRCC = 0.9406 ] +INFO:root:Average loss=0.09198298481106758 +INFO:root: +[Valid][28][UTT][ MSE = 0.2651 | LCC = 0.9391 | SRCC = 0.9402 ] +INFO:root:Average loss=0.09729321014881134 +INFO:root: +[Valid][29][UTT][ MSE = 0.2383 | LCC = 0.9419 | SRCC = 0.9428 ] +INFO:root:Average loss=0.08812451922893524 +INFO:root: +[Valid][30][UTT][ MSE = 0.2410 | LCC = 0.9417 | SRCC = 0.9415 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2452 | LCC = 0.9423 | SRCC = 0.9426 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run4/config.gin b/results_tencent/w2v2_large_asr_zh/layer0/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run4/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer0/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..fc7cc1325327e086e5522a5905ff9fed2427c8f6 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b48a8c00fd8508ab653a2a0d6ffa38250f029c7270770e0097fa8dd0f9c09c83 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer0/run4/train.log b/results_tencent/w2v2_large_asr_zh/layer0/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..900e2d440ea8911f3c46a9e0dec7ecd3d7721a9b --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer0/run4/train.log @@ -0,0 +1,61 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.7586560635566711 +INFO:root: +[Valid][0][UTT][ MSE = 0.6712 | LCC = 0.8337 | SRCC = 0.8415 ] +INFO:root:Average loss=0.403533731341362 +INFO:root: +[Valid][1][UTT][ MSE = 0.3974 | LCC = 0.8872 | SRCC = 0.8881 ] +INFO:root:Average loss=0.33163184320926664 +INFO:root: +[Valid][2][UTT][ MSE = 0.3581 | LCC = 0.9077 | SRCC = 0.9096 ] +INFO:root:Average loss=0.31768014991283416 +INFO:root: +[Valid][3][UTT][ MSE = 0.3733 | LCC = 0.9111 | SRCC = 0.9139 ] +INFO:root:Average loss=0.28076802587509153 +INFO:root: +[Valid][4][UTT][ MSE = 0.3249 | LCC = 0.9167 | SRCC = 0.9186 ] +INFO:root:Average loss=0.23750319755077362 +INFO:root: +[Valid][5][UTT][ MSE = 0.3440 | LCC = 0.9188 | SRCC = 0.9200 ] +INFO:root:Average loss=0.23187036621570586 +INFO:root: +[Valid][6][UTT][ MSE = 0.2968 | LCC = 0.9213 | SRCC = 0.9215 ] +INFO:root:Average loss=0.2114708326458931 +INFO:root: +[Valid][7][UTT][ MSE = 0.3635 | LCC = 0.9164 | SRCC = 0.9205 ] +INFO:root:Average loss=0.20089914709329604 +INFO:root: +[Valid][8][UTT][ MSE = 0.3298 | LCC = 0.9296 | SRCC = 0.9311 ] +INFO:root:Average loss=0.18600933986902238 +INFO:root: +[Valid][9][UTT][ MSE = 0.2914 | LCC = 0.9257 | SRCC = 0.9267 ] +INFO:root:Average loss=0.18360183924436568 +INFO:root: +[Valid][10][UTT][ MSE = 0.2764 | LCC = 0.9323 | SRCC = 0.9345 ] +INFO:root:Average loss=0.17260625225305556 +INFO:root: +[Valid][11][UTT][ MSE = 0.3346 | LCC = 0.9304 | SRCC = 0.9350 ] +INFO:root:Average loss=0.1652604493498802 +INFO:root: +[Valid][12][UTT][ MSE = 0.2823 | LCC = 0.9320 | SRCC = 0.9334 ] +INFO:root:Average loss=0.15598573511838912 +INFO:root: +[Valid][13][UTT][ MSE = 0.2627 | LCC = 0.9352 | SRCC = 0.9359 ] +INFO:root:Average loss=0.14708843672275543 +INFO:root: +[Valid][14][UTT][ MSE = 0.2777 | LCC = 0.9366 | SRCC = 0.9378 ] +INFO:root:Average loss=0.14652638894319533 +INFO:root: +[Valid][15][UTT][ MSE = 0.2748 | LCC = 0.9331 | SRCC = 0.9332 ] +INFO:root:Average loss=0.14176087921857833 +INFO:root: +[Valid][16][UTT][ MSE = 0.2865 | LCC = 0.9376 | SRCC = 0.9388 ] +INFO:root:Average loss=0.13331753897666931 +INFO:root: +[Valid][17][UTT][ MSE = 0.2515 | LCC = 0.9386 | SRCC = 0.9392 ] +INFO:root:Average loss=0.1337440876364708 +INFO:root: +[Valid][18][UTT][ MSE = 0.3153 | LCC = 0.9341 | SRCC = 0.9376 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run0/config.gin b/results_tencent/w2v2_large_asr_zh/layer10/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run0/model.pt b/results_tencent/w2v2_large_asr_zh/layer10/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2c74b47c189084299aba5711268867c448da90a6 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:493e77f3dc890cee0e374f3132d24c3ff3b35d61e191fcdcbb634c5e0cfd78da +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run0/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer10/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7dd121b6ffcedd9dcb2b8f7101c36a9fb3162b87 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c60e107e6b1762c40c222baa2b6947b48c0d61cc6dbe45f5c3a0d802e23118af +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run0/train.log b/results_tencent/w2v2_large_asr_zh/layer10/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..917469f40709d5263f66d03be98ebf044030c0c8 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46963083696365354 +INFO:root: +[Valid][0][UTT][ MSE = 0.3233 | LCC = 0.9051 | SRCC = 0.9038 ] +INFO:root:Average loss=0.2260573353767395 +INFO:root: +[Valid][1][UTT][ MSE = 0.2594 | LCC = 0.9363 | SRCC = 0.9362 ] +INFO:root:Average loss=0.18393055558204652 +INFO:root: +[Valid][2][UTT][ MSE = 0.2232 | LCC = 0.9431 | SRCC = 0.9437 ] +INFO:root:Average loss=0.14720117390155793 +INFO:root: +[Valid][3][UTT][ MSE = 0.2399 | LCC = 0.9485 | SRCC = 0.9496 ] +INFO:root:Average loss=0.13193895131349564 +INFO:root: +[Valid][4][UTT][ MSE = 0.2282 | LCC = 0.9502 | SRCC = 0.9525 ] +INFO:root:Average loss=0.12401282066106796 +INFO:root: +[Valid][5][UTT][ MSE = 0.1832 | LCC = 0.9529 | SRCC = 0.9544 ] +INFO:root:Average loss=0.10593218463659286 +INFO:root: +[Valid][6][UTT][ MSE = 0.2114 | LCC = 0.9541 | SRCC = 0.9550 ] +INFO:root:Average loss=0.09713965299725533 +INFO:root: +[Valid][7][UTT][ MSE = 0.1976 | LCC = 0.9558 | SRCC = 0.9567 ] +INFO:root:Average loss=0.09321482619643211 +INFO:root: +[Valid][8][UTT][ MSE = 0.2212 | LCC = 0.9581 | SRCC = 0.9588 ] +INFO:root:Average loss=0.08568807560205459 +INFO:root: +[Valid][9][UTT][ MSE = 0.1738 | LCC = 0.9562 | SRCC = 0.9575 ] +INFO:root:Average loss=0.08033579754829406 +INFO:root: +[Valid][10][UTT][ MSE = 0.1858 | LCC = 0.9564 | SRCC = 0.9573 ] +INFO:root:Average loss=0.07666086515784264 +INFO:root: +[Valid][11][UTT][ MSE = 0.2137 | LCC = 0.9607 | SRCC = 0.9610 ] +INFO:root:Average loss=0.07691530016064643 +INFO:root: +[Valid][12][UTT][ MSE = 0.2126 | LCC = 0.9589 | SRCC = 0.9599 ] +INFO:root:Average loss=0.06965039739012718 +INFO:root: +[Valid][13][UTT][ MSE = 0.1838 | LCC = 0.9592 | SRCC = 0.9595 ] +INFO:root:Average loss=0.06444608667492867 +INFO:root: +[Valid][14][UTT][ MSE = 0.1818 | LCC = 0.9612 | SRCC = 0.9603 ] +INFO:root:Average loss=0.06199332654476166 +INFO:root: +[Valid][15][UTT][ MSE = 0.2232 | LCC = 0.9610 | SRCC = 0.9607 ] +INFO:root:Average loss=0.05717923586070538 +INFO:root: +[Valid][16][UTT][ MSE = 0.1961 | LCC = 0.9621 | SRCC = 0.9619 ] +INFO:root:Average loss=0.05387803688645363 +INFO:root: +[Valid][17][UTT][ MSE = 0.2382 | LCC = 0.9614 | SRCC = 0.9616 ] +INFO:root:Average loss=0.05292967812716961 +INFO:root: +[Valid][18][UTT][ MSE = 0.2074 | LCC = 0.9618 | SRCC = 0.9612 ] +INFO:root:Average loss=0.05227358937263489 +INFO:root: +[Valid][19][UTT][ MSE = 0.1820 | LCC = 0.9600 | SRCC = 0.9596 ] +INFO:root:Average loss=0.05033288994431496 +INFO:root: +[Valid][20][UTT][ MSE = 0.1954 | LCC = 0.9618 | SRCC = 0.9616 ] +INFO:root:Average loss=0.047310817658901216 +INFO:root: +[Valid][21][UTT][ MSE = 0.1939 | LCC = 0.9620 | SRCC = 0.9618 ] +INFO:root:Average loss=0.0411057538986206 +INFO:root: +[Valid][22][UTT][ MSE = 0.1881 | LCC = 0.9619 | SRCC = 0.9618 ] +INFO:root:Average loss=0.04237282167375088 +INFO:root: +[Valid][23][UTT][ MSE = 0.1971 | LCC = 0.9611 | SRCC = 0.9610 ] +INFO:root:Average loss=0.04239663626253605 +INFO:root: +[Valid][24][UTT][ MSE = 0.1997 | LCC = 0.9616 | SRCC = 0.9611 ] +INFO:root:Average loss=0.042016334623098374 +INFO:root: +[Valid][25][UTT][ MSE = 0.1982 | LCC = 0.9617 | SRCC = 0.9613 ] +INFO:root:Average loss=0.03702243514358997 +INFO:root: +[Valid][26][UTT][ MSE = 0.2094 | LCC = 0.9608 | SRCC = 0.9605 ] +INFO:root:Average loss=0.034179866060614586 +INFO:root: +[Valid][27][UTT][ MSE = 0.1822 | LCC = 0.9607 | SRCC = 0.9602 ] +INFO:root:Average loss=0.03422879511117935 +INFO:root: +[Valid][28][UTT][ MSE = 0.2022 | LCC = 0.9617 | SRCC = 0.9612 ] +INFO:root:Average loss=0.032586742773652075 +INFO:root: +[Valid][29][UTT][ MSE = 0.2196 | LCC = 0.9612 | SRCC = 0.9617 ] +INFO:root:Average loss=0.03542439687252045 +INFO:root: +[Valid][30][UTT][ MSE = 0.1974 | LCC = 0.9620 | SRCC = 0.9618 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2085 | LCC = 0.9587 | SRCC = 0.9595 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run1/config.gin b/results_tencent/w2v2_large_asr_zh/layer10/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run1/model.pt b/results_tencent/w2v2_large_asr_zh/layer10/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3b75bd7ec0066c5b2fa47b61e7cc4003cee0f2d1 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3623711b12390ae5f0875fbc52acf872e60dc5b590f73de82d090c83634316fe +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run1/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer10/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4188c4244aaa9b771b254f46dd44e8704c23d896 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f427e2ad9636f1d1ca57c701de2b2e7f1eca5d14e0c0e9fbbb70e6ff1b57aa0 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run1/train.log b/results_tencent/w2v2_large_asr_zh/layer10/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0414896ff3e56140f594c09990cd7e8cb6164351 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4566160742044449 +INFO:root: +[Valid][0][UTT][ MSE = 0.3092 | LCC = 0.9127 | SRCC = 0.9129 ] +INFO:root:Average loss=0.21247813487052916 +INFO:root: +[Valid][1][UTT][ MSE = 0.3450 | LCC = 0.9264 | SRCC = 0.9272 ] +INFO:root:Average loss=0.17184787321090697 +INFO:root: +[Valid][2][UTT][ MSE = 0.2450 | LCC = 0.9422 | SRCC = 0.9429 ] +INFO:root:Average loss=0.15245940804481506 +INFO:root: +[Valid][3][UTT][ MSE = 0.2330 | LCC = 0.9484 | SRCC = 0.9484 ] +INFO:root:Average loss=0.12639596354961397 +INFO:root: +[Valid][4][UTT][ MSE = 0.2231 | LCC = 0.9500 | SRCC = 0.9507 ] +INFO:root:Average loss=0.11629383999109268 +INFO:root: +[Valid][5][UTT][ MSE = 0.2476 | LCC = 0.9535 | SRCC = 0.9549 ] +INFO:root:Average loss=0.10751358166337013 +INFO:root: +[Valid][6][UTT][ MSE = 0.2189 | LCC = 0.9552 | SRCC = 0.9558 ] +INFO:root:Average loss=0.10305665695667267 +INFO:root: +[Valid][7][UTT][ MSE = 0.2282 | LCC = 0.9565 | SRCC = 0.9572 ] +INFO:root:Average loss=0.09011200073361397 +INFO:root: +[Valid][8][UTT][ MSE = 0.2317 | LCC = 0.9572 | SRCC = 0.9584 ] +INFO:root:Average loss=0.08512359115481377 +INFO:root: +[Valid][9][UTT][ MSE = 0.2182 | LCC = 0.9576 | SRCC = 0.9581 ] +INFO:root:Average loss=0.08059692573547363 +INFO:root: +[Valid][10][UTT][ MSE = 0.1892 | LCC = 0.9568 | SRCC = 0.9576 ] +INFO:root:Average loss=0.07597542726993561 +INFO:root: +[Valid][11][UTT][ MSE = 0.2027 | LCC = 0.9585 | SRCC = 0.9592 ] +INFO:root:Average loss=0.0692131852209568 +INFO:root: +[Valid][12][UTT][ MSE = 0.2157 | LCC = 0.9593 | SRCC = 0.9604 ] +INFO:root:Average loss=0.06775573804974556 +INFO:root: +[Valid][13][UTT][ MSE = 0.2126 | LCC = 0.9600 | SRCC = 0.9604 ] +INFO:root:Average loss=0.06276753172278404 +INFO:root: +[Valid][14][UTT][ MSE = 0.2291 | LCC = 0.9596 | SRCC = 0.9601 ] +INFO:root:Average loss=0.05941283115744591 +INFO:root: +[Valid][15][UTT][ MSE = 0.2057 | LCC = 0.9601 | SRCC = 0.9610 ] +INFO:root:Average loss=0.05699885445833206 +INFO:root: +[Valid][16][UTT][ MSE = 0.1715 | LCC = 0.9594 | SRCC = 0.9598 ] +INFO:root:Average loss=0.05580769999325275 +INFO:root: +[Valid][17][UTT][ MSE = 0.1917 | LCC = 0.9609 | SRCC = 0.9621 ] +INFO:root:Average loss=0.05437469276785851 +INFO:root: +[Valid][18][UTT][ MSE = 0.2111 | LCC = 0.9601 | SRCC = 0.9607 ] +INFO:root:Average loss=0.04939914040267467 +INFO:root: +[Valid][19][UTT][ MSE = 0.1959 | LCC = 0.9604 | SRCC = 0.9608 ] +INFO:root:Average loss=0.04594221593439579 +INFO:root: +[Valid][20][UTT][ MSE = 0.2164 | LCC = 0.9595 | SRCC = 0.9606 ] +INFO:root:Average loss=0.04391225336492061 +INFO:root: +[Valid][21][UTT][ MSE = 0.2238 | LCC = 0.9596 | SRCC = 0.9597 ] +INFO:root:Average loss=0.043600158765912056 +INFO:root: +[Valid][22][UTT][ MSE = 0.1781 | LCC = 0.9611 | SRCC = 0.9613 ] +INFO:root:Average loss=0.04000585363805294 +INFO:root: +[Valid][23][UTT][ MSE = 0.1858 | LCC = 0.9599 | SRCC = 0.9598 ] +INFO:root:Average loss=0.0393199023604393 +INFO:root: +[Valid][24][UTT][ MSE = 0.1887 | LCC = 0.9588 | SRCC = 0.9597 ] +INFO:root:Average loss=0.03890284825861454 +INFO:root: +[Valid][25][UTT][ MSE = 0.1983 | LCC = 0.9615 | SRCC = 0.9617 ] +INFO:root:Average loss=0.03701491817831993 +INFO:root: +[Valid][26][UTT][ MSE = 0.1994 | LCC = 0.9607 | SRCC = 0.9611 ] +INFO:root:Average loss=0.034705630153417585 +INFO:root: +[Valid][27][UTT][ MSE = 0.1769 | LCC = 0.9601 | SRCC = 0.9604 ] +INFO:root:Average loss=0.033234213680028914 +INFO:root: +[Valid][28][UTT][ MSE = 0.1968 | LCC = 0.9607 | SRCC = 0.9611 ] +INFO:root:Average loss=0.032939401656389235 +INFO:root: +[Valid][29][UTT][ MSE = 0.1887 | LCC = 0.9620 | SRCC = 0.9618 ] +INFO:root:Average loss=0.030511742874979972 +INFO:root: +[Valid][30][UTT][ MSE = 0.1983 | LCC = 0.9612 | SRCC = 0.9615 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1962 | LCC = 0.9595 | SRCC = 0.9604 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run2/config.gin b/results_tencent/w2v2_large_asr_zh/layer10/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run2/model.pt b/results_tencent/w2v2_large_asr_zh/layer10/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6bbd770cdb7f1162d892842eea02924e5395dab8 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:80ec2dbf16327650b9a84ef13f4f4f7e27f82cca4c320219d9db31eeb969d280 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run2/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer10/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6b9607f3f45129c87e50abb72a04121e95fba191 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e8bf7923e36df695206f984d8d9a31c81eb38417970e0040591592a9e711281f +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run2/train.log b/results_tencent/w2v2_large_asr_zh/layer10/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d79ca7e17712fc6b7ecd86ef5915d26571d15e48 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4500124242305756 +INFO:root: +[Valid][0][UTT][ MSE = 0.3576 | LCC = 0.9151 | SRCC = 0.9142 ] +INFO:root:Average loss=0.22403869378566743 +INFO:root: +[Valid][1][UTT][ MSE = 0.3275 | LCC = 0.9294 | SRCC = 0.9300 ] +INFO:root:Average loss=0.17526149982213973 +INFO:root: +[Valid][2][UTT][ MSE = 0.3071 | LCC = 0.9413 | SRCC = 0.9415 ] +INFO:root:Average loss=0.1520601834654808 +INFO:root: +[Valid][3][UTT][ MSE = 0.3071 | LCC = 0.9467 | SRCC = 0.9484 ] +INFO:root:Average loss=0.13330467891693115 +INFO:root: +[Valid][4][UTT][ MSE = 0.2308 | LCC = 0.9470 | SRCC = 0.9490 ] +INFO:root:Average loss=0.1231534985601902 +INFO:root: +[Valid][5][UTT][ MSE = 0.2462 | LCC = 0.9533 | SRCC = 0.9539 ] +INFO:root:Average loss=0.1162192610502243 +INFO:root: +[Valid][6][UTT][ MSE = 0.2076 | LCC = 0.9513 | SRCC = 0.9527 ] +INFO:root:Average loss=0.10693024858832359 +INFO:root: +[Valid][7][UTT][ MSE = 0.2220 | LCC = 0.9577 | SRCC = 0.9582 ] +INFO:root:Average loss=0.09435814279317856 +INFO:root: +[Valid][8][UTT][ MSE = 0.2369 | LCC = 0.9588 | SRCC = 0.9598 ] +INFO:root:Average loss=0.09260213986039162 +INFO:root: +[Valid][9][UTT][ MSE = 0.1968 | LCC = 0.9581 | SRCC = 0.9590 ] +INFO:root:Average loss=0.08818524342775345 +INFO:root: +[Valid][10][UTT][ MSE = 0.2290 | LCC = 0.9601 | SRCC = 0.9607 ] +INFO:root:Average loss=0.0839541592001915 +INFO:root: +[Valid][11][UTT][ MSE = 0.1954 | LCC = 0.9576 | SRCC = 0.9587 ] +INFO:root:Average loss=0.07759246948361397 +INFO:root: +[Valid][12][UTT][ MSE = 0.2346 | LCC = 0.9606 | SRCC = 0.9616 ] +INFO:root:Average loss=0.07260602459311485 +INFO:root: +[Valid][13][UTT][ MSE = 0.2055 | LCC = 0.9612 | SRCC = 0.9615 ] +INFO:root:Average loss=0.06854302272200584 +INFO:root: +[Valid][14][UTT][ MSE = 0.2027 | LCC = 0.9594 | SRCC = 0.9592 ] +INFO:root:Average loss=0.06397987435758114 +INFO:root: +[Valid][15][UTT][ MSE = 0.1940 | LCC = 0.9608 | SRCC = 0.9605 ] +INFO:root:Average loss=0.06261644667387009 +INFO:root: +[Valid][16][UTT][ MSE = 0.2288 | LCC = 0.9600 | SRCC = 0.9603 ] +INFO:root:Average loss=0.057207971274852755 +INFO:root: +[Valid][17][UTT][ MSE = 0.1958 | LCC = 0.9616 | SRCC = 0.9618 ] +INFO:root:Average loss=0.05785218204557896 +INFO:root: +[Valid][18][UTT][ MSE = 0.2527 | LCC = 0.9583 | SRCC = 0.9595 ] +INFO:root:Average loss=0.05415861986577511 +INFO:root: +[Valid][19][UTT][ MSE = 0.2154 | LCC = 0.9618 | SRCC = 0.9621 ] +INFO:root:Average loss=0.05243546226620674 +INFO:root: +[Valid][20][UTT][ MSE = 0.2167 | LCC = 0.9603 | SRCC = 0.9608 ] +INFO:root:Average loss=0.05071640777587891 +INFO:root: +[Valid][21][UTT][ MSE = 0.2054 | LCC = 0.9621 | SRCC = 0.9620 ] +INFO:root:Average loss=0.04829455555975437 +INFO:root: +[Valid][22][UTT][ MSE = 0.2070 | LCC = 0.9615 | SRCC = 0.9611 ] +INFO:root:Average loss=0.045248620018363 +INFO:root: +[Valid][23][UTT][ MSE = 0.2129 | LCC = 0.9626 | SRCC = 0.9622 ] +INFO:root:Average loss=0.04551428741216659 +INFO:root: +[Valid][24][UTT][ MSE = 0.2200 | LCC = 0.9615 | SRCC = 0.9617 ] +INFO:root:Average loss=0.04238621339201927 +INFO:root: +[Valid][25][UTT][ MSE = 0.2035 | LCC = 0.9623 | SRCC = 0.9619 ] +INFO:root:Average loss=0.041494322776794434 +INFO:root: +[Valid][26][UTT][ MSE = 0.2189 | LCC = 0.9614 | SRCC = 0.9612 ] +INFO:root:Average loss=0.04057692076265812 +INFO:root: +[Valid][27][UTT][ MSE = 0.2164 | LCC = 0.9619 | SRCC = 0.9620 ] +INFO:root:Average loss=0.03745567962527275 +INFO:root: +[Valid][28][UTT][ MSE = 0.2170 | LCC = 0.9619 | SRCC = 0.9618 ] +INFO:root:Average loss=0.037110941648483275 +INFO:root: +[Valid][29][UTT][ MSE = 0.2334 | LCC = 0.9608 | SRCC = 0.9608 ] +INFO:root:Average loss=0.03576038928329944 +INFO:root: +[Valid][30][UTT][ MSE = 0.2394 | LCC = 0.9610 | SRCC = 0.9614 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2260 | LCC = 0.9580 | SRCC = 0.9589 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run3/config.gin b/results_tencent/w2v2_large_asr_zh/layer10/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run3/model.pt b/results_tencent/w2v2_large_asr_zh/layer10/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4436f4a1829625c00c079a94acd686a75ec56c84 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:681ee6bcc61f622db7ded07d94e87c987e28f0071bbad7512bba9b0e5e245870 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run3/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer10/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..99238bfca1f6a4cf17c42b717551ee6d2cfa671e --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6537439c978829c237686372e9295e95dedf0313ab16fae63762065daedf31d0 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run3/train.log b/results_tencent/w2v2_large_asr_zh/layer10/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..855758beed9ddfa92c25fb670e7967304f3a2658 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4528689475059509 +INFO:root: +[Valid][0][UTT][ MSE = 0.3214 | LCC = 0.9159 | SRCC = 0.9134 ] +INFO:root:Average loss=0.21400415128469466 +INFO:root: +[Valid][1][UTT][ MSE = 0.2182 | LCC = 0.9348 | SRCC = 0.9351 ] +INFO:root:Average loss=0.1659979482293129 +INFO:root: +[Valid][2][UTT][ MSE = 0.2001 | LCC = 0.9463 | SRCC = 0.9478 ] +INFO:root:Average loss=0.1426703977584839 +INFO:root: +[Valid][3][UTT][ MSE = 0.1984 | LCC = 0.9519 | SRCC = 0.9529 ] +INFO:root:Average loss=0.1222051472067833 +INFO:root: +[Valid][4][UTT][ MSE = 0.1712 | LCC = 0.9534 | SRCC = 0.9552 ] +INFO:root:Average loss=0.11961574599146843 +INFO:root: +[Valid][5][UTT][ MSE = 0.1888 | LCC = 0.9556 | SRCC = 0.9561 ] +INFO:root:Average loss=0.10399472802877426 +INFO:root: +[Valid][6][UTT][ MSE = 0.2118 | LCC = 0.9543 | SRCC = 0.9558 ] +INFO:root:Average loss=0.10179201292991638 +INFO:root: +[Valid][7][UTT][ MSE = 0.1872 | LCC = 0.9542 | SRCC = 0.9554 ] +INFO:root:Average loss=0.09208247178792954 +INFO:root: +[Valid][8][UTT][ MSE = 0.2068 | LCC = 0.9597 | SRCC = 0.9598 ] +INFO:root:Average loss=0.08716560819745064 +INFO:root: +[Valid][9][UTT][ MSE = 0.2085 | LCC = 0.9606 | SRCC = 0.9606 ] +INFO:root:Average loss=0.08027985140681267 +INFO:root: +[Valid][10][UTT][ MSE = 0.2034 | LCC = 0.9601 | SRCC = 0.9604 ] +INFO:root:Average loss=0.07660102251172066 +INFO:root: +[Valid][11][UTT][ MSE = 0.1848 | LCC = 0.9596 | SRCC = 0.9592 ] +INFO:root:Average loss=0.07390498501062394 +INFO:root: +[Valid][12][UTT][ MSE = 0.1675 | LCC = 0.9606 | SRCC = 0.9603 ] +INFO:root:Average loss=0.06797313094139099 +INFO:root: +[Valid][13][UTT][ MSE = 0.1534 | LCC = 0.9606 | SRCC = 0.9612 ] +INFO:root:Average loss=0.06432241824269294 +INFO:root: +[Valid][14][UTT][ MSE = 0.1697 | LCC = 0.9611 | SRCC = 0.9610 ] +INFO:root:Average loss=0.06150408512353897 +INFO:root: +[Valid][15][UTT][ MSE = 0.1781 | LCC = 0.9614 | SRCC = 0.9615 ] +INFO:root:Average loss=0.05840297782421112 +INFO:root: +[Valid][16][UTT][ MSE = 0.1644 | LCC = 0.9609 | SRCC = 0.9607 ] +INFO:root:Average loss=0.054611950546503064 +INFO:root: +[Valid][17][UTT][ MSE = 0.1709 | LCC = 0.9598 | SRCC = 0.9600 ] +INFO:root:Average loss=0.05217350909113884 +INFO:root: +[Valid][18][UTT][ MSE = 0.1628 | LCC = 0.9627 | SRCC = 0.9628 ] +INFO:root:Average loss=0.04994829298555851 +INFO:root: +[Valid][19][UTT][ MSE = 0.2124 | LCC = 0.9616 | SRCC = 0.9622 ] +INFO:root:Average loss=0.04812727865576744 +INFO:root: +[Valid][20][UTT][ MSE = 0.1677 | LCC = 0.9613 | SRCC = 0.9617 ] +INFO:root:Average loss=0.049901160821318624 +INFO:root: +[Valid][21][UTT][ MSE = 0.1758 | LCC = 0.9609 | SRCC = 0.9616 ] +INFO:root:Average loss=0.0445510821044445 +INFO:root: +[Valid][22][UTT][ MSE = 0.1544 | LCC = 0.9593 | SRCC = 0.9595 ] +INFO:root:Average loss=0.0423289385586977 +INFO:root: +[Valid][23][UTT][ MSE = 0.1612 | LCC = 0.9625 | SRCC = 0.9621 ] +INFO:root:Average loss=0.0414222606420517 +INFO:root: +[Valid][24][UTT][ MSE = 0.1807 | LCC = 0.9634 | SRCC = 0.9633 ] +INFO:root:Average loss=0.039435860514640805 +INFO:root: +[Valid][25][UTT][ MSE = 0.1734 | LCC = 0.9624 | SRCC = 0.9617 ] +INFO:root:Average loss=0.0378688570857048 +INFO:root: +[Valid][26][UTT][ MSE = 0.1757 | LCC = 0.9616 | SRCC = 0.9617 ] +INFO:root:Average loss=0.04022929859161377 +INFO:root: +[Valid][27][UTT][ MSE = 0.1730 | LCC = 0.9631 | SRCC = 0.9629 ] +INFO:root:Average loss=0.0361565348058939 +INFO:root: +[Valid][28][UTT][ MSE = 0.1630 | LCC = 0.9618 | SRCC = 0.9614 ] +INFO:root:Average loss=0.03459809985756874 +INFO:root: +[Valid][29][UTT][ MSE = 0.1807 | LCC = 0.9617 | SRCC = 0.9615 ] +INFO:root:Average loss=0.0330853786021471 +INFO:root: +[Valid][30][UTT][ MSE = 0.1734 | LCC = 0.9616 | SRCC = 0.9614 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1932 | LCC = 0.9580 | SRCC = 0.9586 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run4/config.gin b/results_tencent/w2v2_large_asr_zh/layer10/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run4/model.pt b/results_tencent/w2v2_large_asr_zh/layer10/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a6a697a7d00c1e55baaa0fa418ff1896a31e7d60 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2c474c15b7b80b6ddc96d64a8c20649beba2dfce5d7b3d7a8e177ac377c9a39b +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run4/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer10/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..17ade9da3478b1d42fc332d2e1914808b0712eed --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e78d9034d5ffc84de6dafcd874dc5a9b77947e571ad23653737097f2e42c50c9 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer10/run4/train.log b/results_tencent/w2v2_large_asr_zh/layer10/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..8bc8ee0397eddc55c327b84a692b0d80201fab29 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer10/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46268958568573 +INFO:root: +[Valid][0][UTT][ MSE = 0.3193 | LCC = 0.9089 | SRCC = 0.9092 ] +INFO:root:Average loss=0.2049077730178833 +INFO:root: +[Valid][1][UTT][ MSE = 0.2365 | LCC = 0.9417 | SRCC = 0.9413 ] +INFO:root:Average loss=0.16188013136386872 +INFO:root: +[Valid][2][UTT][ MSE = 0.2473 | LCC = 0.9485 | SRCC = 0.9494 ] +INFO:root:Average loss=0.13414291900396347 +INFO:root: +[Valid][3][UTT][ MSE = 0.2949 | LCC = 0.9498 | SRCC = 0.9482 ] +INFO:root:Average loss=0.12371917772293091 +INFO:root: +[Valid][4][UTT][ MSE = 0.2160 | LCC = 0.9526 | SRCC = 0.9532 ] +INFO:root:Average loss=0.11495767903327941 +INFO:root: +[Valid][5][UTT][ MSE = 0.2107 | LCC = 0.9563 | SRCC = 0.9562 ] +INFO:root:Average loss=0.10611912021040916 +INFO:root: +[Valid][6][UTT][ MSE = 0.2025 | LCC = 0.9556 | SRCC = 0.9567 ] +INFO:root:Average loss=0.10116399136185646 +INFO:root: +[Valid][7][UTT][ MSE = 0.1992 | LCC = 0.9578 | SRCC = 0.9576 ] +INFO:root:Average loss=0.09151470878720283 +INFO:root: +[Valid][8][UTT][ MSE = 0.1821 | LCC = 0.9580 | SRCC = 0.9580 ] +INFO:root:Average loss=0.08363859713077545 +INFO:root: +[Valid][9][UTT][ MSE = 0.2139 | LCC = 0.9601 | SRCC = 0.9595 ] +INFO:root:Average loss=0.08361868757009507 +INFO:root: +[Valid][10][UTT][ MSE = 0.1953 | LCC = 0.9580 | SRCC = 0.9581 ] +INFO:root:Average loss=0.07599332171678543 +INFO:root: +[Valid][11][UTT][ MSE = 0.1643 | LCC = 0.9599 | SRCC = 0.9594 ] +INFO:root:Average loss=0.06989672940969467 +INFO:root: +[Valid][12][UTT][ MSE = 0.1887 | LCC = 0.9591 | SRCC = 0.9587 ] +INFO:root:Average loss=0.06687596854567528 +INFO:root: +[Valid][13][UTT][ MSE = 0.1818 | LCC = 0.9592 | SRCC = 0.9580 ] +INFO:root:Average loss=0.06191864213347435 +INFO:root: +[Valid][14][UTT][ MSE = 0.1680 | LCC = 0.9602 | SRCC = 0.9599 ] +INFO:root:Average loss=0.05988443008065224 +INFO:root: +[Valid][15][UTT][ MSE = 0.2053 | LCC = 0.9589 | SRCC = 0.9587 ] +INFO:root:Average loss=0.05588047304749489 +INFO:root: +[Valid][16][UTT][ MSE = 0.1780 | LCC = 0.9609 | SRCC = 0.9604 ] +INFO:root:Average loss=0.05352094940841198 +INFO:root: +[Valid][17][UTT][ MSE = 0.1731 | LCC = 0.9610 | SRCC = 0.9605 ] +INFO:root:Average loss=0.04896980437636375 +INFO:root: +[Valid][18][UTT][ MSE = 0.1717 | LCC = 0.9614 | SRCC = 0.9608 ] +INFO:root:Average loss=0.04669969440996647 +INFO:root: +[Valid][19][UTT][ MSE = 0.1542 | LCC = 0.9620 | SRCC = 0.9616 ] +INFO:root:Average loss=0.04605697312951088 +INFO:root: +[Valid][20][UTT][ MSE = 0.2057 | LCC = 0.9613 | SRCC = 0.9607 ] +INFO:root:Average loss=0.045427177801728245 +INFO:root: +[Valid][21][UTT][ MSE = 0.1699 | LCC = 0.9625 | SRCC = 0.9616 ] +INFO:root:Average loss=0.03942062783241272 +INFO:root: +[Valid][22][UTT][ MSE = 0.1736 | LCC = 0.9599 | SRCC = 0.9593 ] +INFO:root:Average loss=0.03929334336519241 +INFO:root: +[Valid][23][UTT][ MSE = 0.1795 | LCC = 0.9614 | SRCC = 0.9608 ] +INFO:root:Average loss=0.04097801089286804 +INFO:root: +[Valid][24][UTT][ MSE = 0.1470 | LCC = 0.9619 | SRCC = 0.9617 ] +INFO:root:Average loss=0.03988350565731526 +INFO:root: +[Valid][25][UTT][ MSE = 0.2109 | LCC = 0.9604 | SRCC = 0.9596 ] +INFO:root:Average loss=0.035795210361480714 +INFO:root: +[Valid][26][UTT][ MSE = 0.1984 | LCC = 0.9611 | SRCC = 0.9600 ] +INFO:root:Average loss=0.0344837718307972 +INFO:root: +[Valid][27][UTT][ MSE = 0.1836 | LCC = 0.9606 | SRCC = 0.9601 ] +INFO:root:Average loss=0.03251996760070324 +INFO:root: +[Valid][28][UTT][ MSE = 0.1667 | LCC = 0.9610 | SRCC = 0.9604 ] +INFO:root:Average loss=0.03048785537481308 +INFO:root: +[Valid][29][UTT][ MSE = 0.1934 | LCC = 0.9608 | SRCC = 0.9605 ] +INFO:root:Average loss=0.03173080126941204 +INFO:root: +[Valid][30][UTT][ MSE = 0.1886 | LCC = 0.9623 | SRCC = 0.9624 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1792 | LCC = 0.9588 | SRCC = 0.9591 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run0/config.gin b/results_tencent/w2v2_large_asr_zh/layer12/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run0/model.pt b/results_tencent/w2v2_large_asr_zh/layer12/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e5494ad331cfa1d4f8f91f6bb1175c7e3f5dcecc --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:15d47752a50391615731d04dc5a4d9d15f0ac99f4588b3de4f5879b96994b6a6 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run0/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer12/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..074fd72e2b31332fff06f8a4414479d5d639722a --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:04d4b065417355f11e3778e10718f1af29d060dc0da3b13264fa68f6bda77d5c +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run0/train.log b/results_tencent/w2v2_large_asr_zh/layer12/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ca1a666112bc703e21dfd2e70c147e2e0dc92912 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5340030844211578 +INFO:root: +[Valid][0][UTT][ MSE = 0.3442 | LCC = 0.8973 | SRCC = 0.8961 ] +INFO:root:Average loss=0.23346925961971282 +INFO:root: +[Valid][1][UTT][ MSE = 0.2635 | LCC = 0.9292 | SRCC = 0.9290 ] +INFO:root:Average loss=0.18988298684358595 +INFO:root: +[Valid][2][UTT][ MSE = 0.2302 | LCC = 0.9391 | SRCC = 0.9402 ] +INFO:root:Average loss=0.1525003604888916 +INFO:root: +[Valid][3][UTT][ MSE = 0.2505 | LCC = 0.9472 | SRCC = 0.9472 ] +INFO:root:Average loss=0.1393222314119339 +INFO:root: +[Valid][4][UTT][ MSE = 0.2511 | LCC = 0.9477 | SRCC = 0.9481 ] +INFO:root:Average loss=0.12607791757583617 +INFO:root: +[Valid][5][UTT][ MSE = 0.2284 | LCC = 0.9508 | SRCC = 0.9514 ] +INFO:root:Average loss=0.1194343977868557 +INFO:root: +[Valid][6][UTT][ MSE = 0.2198 | LCC = 0.9524 | SRCC = 0.9536 ] +INFO:root:Average loss=0.10980276477336884 +INFO:root: +[Valid][7][UTT][ MSE = 0.2289 | LCC = 0.9510 | SRCC = 0.9531 ] +INFO:root:Average loss=0.1027402735054493 +INFO:root: +[Valid][8][UTT][ MSE = 0.2148 | LCC = 0.9556 | SRCC = 0.9559 ] +INFO:root:Average loss=0.0900670405626297 +INFO:root: +[Valid][9][UTT][ MSE = 0.1928 | LCC = 0.9537 | SRCC = 0.9543 ] +INFO:root:Average loss=0.08789718118309975 +INFO:root: +[Valid][10][UTT][ MSE = 0.1776 | LCC = 0.9551 | SRCC = 0.9570 ] +INFO:root:Average loss=0.08197462573647499 +INFO:root: +[Valid][11][UTT][ MSE = 0.1889 | LCC = 0.9553 | SRCC = 0.9557 ] +INFO:root:Average loss=0.07829704350233078 +INFO:root: +[Valid][12][UTT][ MSE = 0.1885 | LCC = 0.9572 | SRCC = 0.9574 ] +INFO:root:Average loss=0.07141891250014305 +INFO:root: +[Valid][13][UTT][ MSE = 0.1688 | LCC = 0.9589 | SRCC = 0.9593 ] +INFO:root:Average loss=0.06626197478175164 +INFO:root: +[Valid][14][UTT][ MSE = 0.1740 | LCC = 0.9590 | SRCC = 0.9597 ] +INFO:root:Average loss=0.06468490770459175 +INFO:root: +[Valid][15][UTT][ MSE = 0.2214 | LCC = 0.9576 | SRCC = 0.9581 ] +INFO:root:Average loss=0.05968265596032143 +INFO:root: +[Valid][16][UTT][ MSE = 0.2184 | LCC = 0.9590 | SRCC = 0.9595 ] +INFO:root:Average loss=0.05577167831361294 +INFO:root: +[Valid][17][UTT][ MSE = 0.1773 | LCC = 0.9599 | SRCC = 0.9595 ] +INFO:root:Average loss=0.05448625442385673 +INFO:root: +[Valid][18][UTT][ MSE = 0.2236 | LCC = 0.9591 | SRCC = 0.9599 ] +INFO:root:Average loss=0.05276918022334576 +INFO:root: +[Valid][19][UTT][ MSE = 0.1970 | LCC = 0.9607 | SRCC = 0.9603 ] +INFO:root:Average loss=0.04842992940545082 +INFO:root: +[Valid][20][UTT][ MSE = 0.1760 | LCC = 0.9603 | SRCC = 0.9605 ] +INFO:root:Average loss=0.0478650354295969 +INFO:root: +[Valid][21][UTT][ MSE = 0.1855 | LCC = 0.9595 | SRCC = 0.9595 ] +INFO:root:Average loss=0.0452844977080822 +INFO:root: +[Valid][22][UTT][ MSE = 0.1932 | LCC = 0.9592 | SRCC = 0.9591 ] +INFO:root:Average loss=0.041892413079738615 +INFO:root: +[Valid][23][UTT][ MSE = 0.1900 | LCC = 0.9596 | SRCC = 0.9589 ] +INFO:root:Average loss=0.041055249229073526 +INFO:root: +[Valid][24][UTT][ MSE = 0.2244 | LCC = 0.9599 | SRCC = 0.9595 ] +INFO:root:Average loss=0.04075822429358959 +INFO:root: +[Valid][25][UTT][ MSE = 0.1838 | LCC = 0.9596 | SRCC = 0.9592 ] +INFO:root:Average loss=0.04065913672745228 +INFO:root: +[Valid][26][UTT][ MSE = 0.1768 | LCC = 0.9580 | SRCC = 0.9578 ] +INFO:root:Average loss=0.03709329788386822 +INFO:root: +[Valid][27][UTT][ MSE = 0.1898 | LCC = 0.9590 | SRCC = 0.9592 ] +INFO:root:Average loss=0.03505465844273567 +INFO:root: +[Valid][28][UTT][ MSE = 0.1951 | LCC = 0.9602 | SRCC = 0.9602 ] +INFO:root:Average loss=0.03496843619644642 +INFO:root: +[Valid][29][UTT][ MSE = 0.1966 | LCC = 0.9592 | SRCC = 0.9586 ] +INFO:root:Average loss=0.03336858317255974 +INFO:root: +[Valid][30][UTT][ MSE = 0.1797 | LCC = 0.9587 | SRCC = 0.9585 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2108 | LCC = 0.9558 | SRCC = 0.9565 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run1/config.gin b/results_tencent/w2v2_large_asr_zh/layer12/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run1/model.pt b/results_tencent/w2v2_large_asr_zh/layer12/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f6fc14d5a826133b8bdcda2818a14b826ffc3f75 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe2928eb7a7bf04cf847aa03fb3e608edd6bbbc0740aa173625d8e684301b678 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run1/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer12/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4da6b29a21467d3e93036d96ae085f3ce353b46b --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1255a813838ee55c30eeb4f3a750696de834ac6187d3e8537f6f8930f3526c27 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run1/train.log b/results_tencent/w2v2_large_asr_zh/layer12/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b27d5eea106e74405762e9fa69cf792648e7f895 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.453061047077179 +INFO:root: +[Valid][0][UTT][ MSE = 0.3243 | LCC = 0.9066 | SRCC = 0.9098 ] +INFO:root:Average loss=0.2346823782324791 +INFO:root: +[Valid][1][UTT][ MSE = 0.2612 | LCC = 0.9332 | SRCC = 0.9329 ] +INFO:root:Average loss=0.18186927074193954 +INFO:root: +[Valid][2][UTT][ MSE = 0.2676 | LCC = 0.9432 | SRCC = 0.9440 ] +INFO:root:Average loss=0.15326055359840393 +INFO:root: +[Valid][3][UTT][ MSE = 0.2568 | LCC = 0.9461 | SRCC = 0.9470 ] +INFO:root:Average loss=0.14244453620910644 +INFO:root: +[Valid][4][UTT][ MSE = 0.2234 | LCC = 0.9477 | SRCC = 0.9478 ] +INFO:root:Average loss=0.12487248736619949 +INFO:root: +[Valid][5][UTT][ MSE = 0.2389 | LCC = 0.9518 | SRCC = 0.9524 ] +INFO:root:Average loss=0.11470503228902816 +INFO:root: +[Valid][6][UTT][ MSE = 0.2242 | LCC = 0.9530 | SRCC = 0.9527 ] +INFO:root:Average loss=0.10412597611546516 +INFO:root: +[Valid][7][UTT][ MSE = 0.2048 | LCC = 0.9547 | SRCC = 0.9557 ] +INFO:root:Average loss=0.0953001204431057 +INFO:root: +[Valid][8][UTT][ MSE = 0.2188 | LCC = 0.9548 | SRCC = 0.9548 ] +INFO:root:Average loss=0.09428560972213745 +INFO:root: +[Valid][9][UTT][ MSE = 0.2264 | LCC = 0.9552 | SRCC = 0.9550 ] +INFO:root:Average loss=0.08449947938323021 +INFO:root: +[Valid][10][UTT][ MSE = 0.2120 | LCC = 0.9554 | SRCC = 0.9558 ] +INFO:root:Average loss=0.07998243656754493 +INFO:root: +[Valid][11][UTT][ MSE = 0.1947 | LCC = 0.9555 | SRCC = 0.9569 ] +INFO:root:Average loss=0.07267769977450371 +INFO:root: +[Valid][12][UTT][ MSE = 0.2151 | LCC = 0.9580 | SRCC = 0.9579 ] +INFO:root:Average loss=0.0696639542579651 +INFO:root: +[Valid][13][UTT][ MSE = 0.1918 | LCC = 0.9581 | SRCC = 0.9578 ] +INFO:root:Average loss=0.06603167244791984 +INFO:root: +[Valid][14][UTT][ MSE = 0.1861 | LCC = 0.9575 | SRCC = 0.9570 ] +INFO:root:Average loss=0.06244253200292587 +INFO:root: +[Valid][15][UTT][ MSE = 0.1824 | LCC = 0.9569 | SRCC = 0.9573 ] +INFO:root:Average loss=0.060098151743412015 +INFO:root: +[Valid][16][UTT][ MSE = 0.2240 | LCC = 0.9546 | SRCC = 0.9552 ] +INFO:root:Average loss=0.05768802681565285 +INFO:root: +[Valid][17][UTT][ MSE = 0.2227 | LCC = 0.9566 | SRCC = 0.9566 ] +INFO:root:Average loss=0.05668943458795547 +INFO:root: +[Valid][18][UTT][ MSE = 0.2012 | LCC = 0.9551 | SRCC = 0.9563 ] +INFO:root:Average loss=0.05093151849508286 +INFO:root: +[Valid][19][UTT][ MSE = 0.2116 | LCC = 0.9559 | SRCC = 0.9557 ] +INFO:root:Average loss=0.04831788124144077 +INFO:root: +[Valid][20][UTT][ MSE = 0.2018 | LCC = 0.9590 | SRCC = 0.9596 ] +INFO:root:Average loss=0.04888919344544411 +INFO:root: +[Valid][21][UTT][ MSE = 0.1775 | LCC = 0.9582 | SRCC = 0.9574 ] +INFO:root:Average loss=0.04543403254449368 +INFO:root: +[Valid][22][UTT][ MSE = 0.1812 | LCC = 0.9570 | SRCC = 0.9564 ] +INFO:root:Average loss=0.04297624234855175 +INFO:root: +[Valid][23][UTT][ MSE = 0.1997 | LCC = 0.9566 | SRCC = 0.9565 ] +INFO:root:Average loss=0.04216683270037174 +INFO:root: +[Valid][24][UTT][ MSE = 0.1933 | LCC = 0.9569 | SRCC = 0.9564 ] +INFO:root:Average loss=0.03769187326729297 +INFO:root: +[Valid][25][UTT][ MSE = 0.1899 | LCC = 0.9575 | SRCC = 0.9576 ] +INFO:root:Average loss=0.042115257769823077 +INFO:root: +[Valid][26][UTT][ MSE = 0.1786 | LCC = 0.9572 | SRCC = 0.9572 ] +INFO:root:Average loss=0.037926492154598236 +INFO:root: +[Valid][27][UTT][ MSE = 0.1949 | LCC = 0.9588 | SRCC = 0.9586 ] +INFO:root:Average loss=0.03671346783638001 +INFO:root: +[Valid][28][UTT][ MSE = 0.1981 | LCC = 0.9574 | SRCC = 0.9572 ] +INFO:root:Average loss=0.034113550528883936 +INFO:root: +[Valid][29][UTT][ MSE = 0.1914 | LCC = 0.9579 | SRCC = 0.9578 ] +INFO:root:Average loss=0.03320007961988449 +INFO:root: +[Valid][30][UTT][ MSE = 0.1861 | LCC = 0.9584 | SRCC = 0.9574 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2107 | LCC = 0.9556 | SRCC = 0.9562 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run2/config.gin b/results_tencent/w2v2_large_asr_zh/layer12/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run2/model.pt b/results_tencent/w2v2_large_asr_zh/layer12/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d796daa0def5cb7b89cc45c26744c938857d0266 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ad55645015e44bc4764a00f6388c524dee3ecd23c7acec6a549eaf0d65989329 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run2/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer12/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7437d8243bba82cfff65d1ecf3e6d950dd81cac0 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fcbfe4d4104ad835bf0d7fb7b1c7e0fc34dd5ab51988e32bc4738a21749956c2 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run2/train.log b/results_tencent/w2v2_large_asr_zh/layer12/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..dadfcf9035a6c2816481c9ae20eac9af02bc0c80 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4666772435903549 +INFO:root: +[Valid][0][UTT][ MSE = 0.3358 | LCC = 0.9098 | SRCC = 0.9085 ] +INFO:root:Average loss=0.23036661267280578 +INFO:root: +[Valid][1][UTT][ MSE = 0.2857 | LCC = 0.9336 | SRCC = 0.9324 ] +INFO:root:Average loss=0.177925564289093 +INFO:root: +[Valid][2][UTT][ MSE = 0.2935 | LCC = 0.9448 | SRCC = 0.9443 ] +INFO:root:Average loss=0.14518453723192215 +INFO:root: +[Valid][3][UTT][ MSE = 0.2562 | LCC = 0.9468 | SRCC = 0.9478 ] +INFO:root:Average loss=0.13104577505588533 +INFO:root: +[Valid][4][UTT][ MSE = 0.2348 | LCC = 0.9533 | SRCC = 0.9534 ] +INFO:root:Average loss=0.11858146888017654 +INFO:root: +[Valid][5][UTT][ MSE = 0.2151 | LCC = 0.9548 | SRCC = 0.9545 ] +INFO:root:Average loss=0.10653630217909812 +INFO:root: +[Valid][6][UTT][ MSE = 0.2087 | LCC = 0.9509 | SRCC = 0.9526 ] +INFO:root:Average loss=0.1023320375084877 +INFO:root: +[Valid][7][UTT][ MSE = 0.2489 | LCC = 0.9576 | SRCC = 0.9575 ] +INFO:root:Average loss=0.09229568755626678 +INFO:root: +[Valid][8][UTT][ MSE = 0.2068 | LCC = 0.9576 | SRCC = 0.9578 ] +INFO:root:Average loss=0.08490486544370651 +INFO:root: +[Valid][9][UTT][ MSE = 0.2236 | LCC = 0.9581 | SRCC = 0.9582 ] +INFO:root:Average loss=0.08093266597390175 +INFO:root: +[Valid][10][UTT][ MSE = 0.2166 | LCC = 0.9602 | SRCC = 0.9598 ] +INFO:root:Average loss=0.07915576037764549 +INFO:root: +[Valid][11][UTT][ MSE = 0.2278 | LCC = 0.9584 | SRCC = 0.9588 ] +INFO:root:Average loss=0.07177600678801536 +INFO:root: +[Valid][12][UTT][ MSE = 0.2136 | LCC = 0.9561 | SRCC = 0.9566 ] +INFO:root:Average loss=0.07089734610915184 +INFO:root: +[Valid][13][UTT][ MSE = 0.2150 | LCC = 0.9550 | SRCC = 0.9549 ] +INFO:root:Average loss=0.0637262798845768 +INFO:root: +[Valid][14][UTT][ MSE = 0.1985 | LCC = 0.9605 | SRCC = 0.9604 ] +INFO:root:Average loss=0.05847865101695061 +INFO:root: +[Valid][15][UTT][ MSE = 0.2216 | LCC = 0.9590 | SRCC = 0.9595 ] +INFO:root:Average loss=0.05627876496315003 +INFO:root: +[Valid][16][UTT][ MSE = 0.2200 | LCC = 0.9612 | SRCC = 0.9610 ] +INFO:root:Average loss=0.055609015583992005 +INFO:root: +[Valid][17][UTT][ MSE = 0.2054 | LCC = 0.9600 | SRCC = 0.9595 ] +INFO:root:Average loss=0.05069049814343452 +INFO:root: +[Valid][18][UTT][ MSE = 0.1869 | LCC = 0.9600 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04875809708237648 +INFO:root: +[Valid][19][UTT][ MSE = 0.1823 | LCC = 0.9601 | SRCC = 0.9596 ] +INFO:root:Average loss=0.04743185582756996 +INFO:root: +[Valid][20][UTT][ MSE = 0.1894 | LCC = 0.9604 | SRCC = 0.9607 ] +INFO:root:Average loss=0.045364531919360164 +INFO:root: +[Valid][21][UTT][ MSE = 0.1772 | LCC = 0.9610 | SRCC = 0.9604 ] +INFO:root:Average loss=0.04412487341463566 +INFO:root: +[Valid][22][UTT][ MSE = 0.1761 | LCC = 0.9600 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04275245264172554 +INFO:root: +[Valid][23][UTT][ MSE = 0.1992 | LCC = 0.9609 | SRCC = 0.9609 ] +INFO:root:Average loss=0.04344027255475521 +INFO:root: +[Valid][24][UTT][ MSE = 0.1919 | LCC = 0.9611 | SRCC = 0.9604 ] +INFO:root:Average loss=0.03904172259569168 +INFO:root: +[Valid][25][UTT][ MSE = 0.1965 | LCC = 0.9618 | SRCC = 0.9612 ] +INFO:root:Average loss=0.036685889169573785 +INFO:root: +[Valid][26][UTT][ MSE = 0.2072 | LCC = 0.9592 | SRCC = 0.9589 ] +INFO:root:Average loss=0.036253662660717965 +INFO:root: +[Valid][27][UTT][ MSE = 0.2060 | LCC = 0.9599 | SRCC = 0.9590 ] +INFO:root:Average loss=0.03371203579008579 +INFO:root: +[Valid][28][UTT][ MSE = 0.2114 | LCC = 0.9607 | SRCC = 0.9600 ] +INFO:root:Average loss=0.03234697186946869 +INFO:root: +[Valid][29][UTT][ MSE = 0.2123 | LCC = 0.9608 | SRCC = 0.9601 ] +INFO:root:Average loss=0.033169435575604436 +INFO:root: +[Valid][30][UTT][ MSE = 0.1849 | LCC = 0.9610 | SRCC = 0.9603 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2055 | LCC = 0.9556 | SRCC = 0.9561 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run3/config.gin b/results_tencent/w2v2_large_asr_zh/layer12/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run3/model.pt b/results_tencent/w2v2_large_asr_zh/layer12/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b3cec9b2c9e59b3c33cfaf8eda46b282965eba32 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b2210cfe30bbed7daa099c5e72bb9e6df7a9cc10ddd2c5249f711c097ea8771a +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run3/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer12/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..041c25022927654a0fc89fe30cbb73da62ffec6d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b4299dbdbdba14e69152ee92b4680d918320a6a3263a8ea2868ebd4d0a1482f +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run3/train.log b/results_tencent/w2v2_large_asr_zh/layer12/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f7465e0b38a21b272e8226393ec406d8c1ae121b --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5380023988485336 +INFO:root: +[Valid][0][UTT][ MSE = 0.3819 | LCC = 0.8960 | SRCC = 0.8972 ] +INFO:root:Average loss=0.2415346650481224 +INFO:root: +[Valid][1][UTT][ MSE = 0.2712 | LCC = 0.9314 | SRCC = 0.9305 ] +INFO:root:Average loss=0.18295622432231903 +INFO:root: +[Valid][2][UTT][ MSE = 0.2478 | LCC = 0.9435 | SRCC = 0.9428 ] +INFO:root:Average loss=0.15236105424165725 +INFO:root: +[Valid][3][UTT][ MSE = 0.2217 | LCC = 0.9469 | SRCC = 0.9480 ] +INFO:root:Average loss=0.13011339366436003 +INFO:root: +[Valid][4][UTT][ MSE = 0.2060 | LCC = 0.9510 | SRCC = 0.9521 ] +INFO:root:Average loss=0.1191167438030243 +INFO:root: +[Valid][5][UTT][ MSE = 0.1925 | LCC = 0.9525 | SRCC = 0.9533 ] +INFO:root:Average loss=0.11332428008317948 +INFO:root: +[Valid][6][UTT][ MSE = 0.2035 | LCC = 0.9546 | SRCC = 0.9546 ] +INFO:root:Average loss=0.1036290803551674 +INFO:root: +[Valid][7][UTT][ MSE = 0.2289 | LCC = 0.9558 | SRCC = 0.9558 ] +INFO:root:Average loss=0.095241081148386 +INFO:root: +[Valid][8][UTT][ MSE = 0.2076 | LCC = 0.9577 | SRCC = 0.9583 ] +INFO:root:Average loss=0.0910743063390255 +INFO:root: +[Valid][9][UTT][ MSE = 0.1962 | LCC = 0.9576 | SRCC = 0.9577 ] +INFO:root:Average loss=0.08247675797343254 +INFO:root: +[Valid][10][UTT][ MSE = 0.1863 | LCC = 0.9576 | SRCC = 0.9574 ] +INFO:root:Average loss=0.07641833779215813 +INFO:root: +[Valid][11][UTT][ MSE = 0.2128 | LCC = 0.9562 | SRCC = 0.9558 ] +INFO:root:Average loss=0.07773380219936371 +INFO:root: +[Valid][12][UTT][ MSE = 0.1693 | LCC = 0.9569 | SRCC = 0.9570 ] +INFO:root:Average loss=0.07288208115100861 +INFO:root: +[Valid][13][UTT][ MSE = 0.2009 | LCC = 0.9571 | SRCC = 0.9575 ] +INFO:root:Average loss=0.06682690498232842 +INFO:root: +[Valid][14][UTT][ MSE = 0.1687 | LCC = 0.9583 | SRCC = 0.9576 ] +INFO:root:Average loss=0.06567499443888664 +INFO:root: +[Valid][15][UTT][ MSE = 0.1770 | LCC = 0.9578 | SRCC = 0.9579 ] +INFO:root:Average loss=0.06025143709778786 +INFO:root: +[Valid][16][UTT][ MSE = 0.1957 | LCC = 0.9566 | SRCC = 0.9564 ] +INFO:root:Average loss=0.05586822670698166 +INFO:root: +[Valid][17][UTT][ MSE = 0.1714 | LCC = 0.9572 | SRCC = 0.9568 ] +INFO:root:Average loss=0.05716824838519096 +INFO:root: +[Valid][18][UTT][ MSE = 0.1889 | LCC = 0.9586 | SRCC = 0.9592 ] +INFO:root:Average loss=0.053685881227254865 +INFO:root: +[Valid][19][UTT][ MSE = 0.1708 | LCC = 0.9563 | SRCC = 0.9565 ] +INFO:root:Average loss=0.05073609074950218 +INFO:root: +[Valid][20][UTT][ MSE = 0.1828 | LCC = 0.9586 | SRCC = 0.9580 ] +INFO:root:Average loss=0.05064345797896385 +INFO:root: +[Valid][21][UTT][ MSE = 0.1896 | LCC = 0.9595 | SRCC = 0.9588 ] +INFO:root:Average loss=0.04644735848903656 +INFO:root: +[Valid][22][UTT][ MSE = 0.1795 | LCC = 0.9587 | SRCC = 0.9578 ] +INFO:root:Average loss=0.043514652416110036 +INFO:root: +[Valid][23][UTT][ MSE = 0.1717 | LCC = 0.9586 | SRCC = 0.9577 ] +INFO:root:Average loss=0.04099229171872139 +INFO:root: +[Valid][24][UTT][ MSE = 0.1925 | LCC = 0.9600 | SRCC = 0.9595 ] +INFO:root:Average loss=0.04129734568297863 +INFO:root: +[Valid][25][UTT][ MSE = 0.1865 | LCC = 0.9582 | SRCC = 0.9580 ] +INFO:root:Average loss=0.03944625408947468 +INFO:root: +[Valid][26][UTT][ MSE = 0.1711 | LCC = 0.9580 | SRCC = 0.9578 ] +INFO:root:Average loss=0.037224850833415986 +INFO:root: +[Valid][27][UTT][ MSE = 0.1732 | LCC = 0.9597 | SRCC = 0.9591 ] +INFO:root:Average loss=0.03630506594479084 +INFO:root: +[Valid][28][UTT][ MSE = 0.1962 | LCC = 0.9596 | SRCC = 0.9593 ] +INFO:root:Average loss=0.03485654509067535 +INFO:root: +[Valid][29][UTT][ MSE = 0.1721 | LCC = 0.9594 | SRCC = 0.9592 ] +INFO:root:Average loss=0.033309288695454596 +INFO:root: +[Valid][30][UTT][ MSE = 0.1698 | LCC = 0.9589 | SRCC = 0.9582 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2072 | LCC = 0.9555 | SRCC = 0.9563 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run4/config.gin b/results_tencent/w2v2_large_asr_zh/layer12/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run4/model.pt b/results_tencent/w2v2_large_asr_zh/layer12/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..690621f3555f1fc3225879d11ae80316537447b2 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe2309a51df09e6ac11a18624001d0eeff2f8164abf418250ee42059e958f2a7 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run4/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer12/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..70bda852715fdfa854d0eb4253403ac7d32141ee --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bdafb0b71695846002e4f4fc34f271801b0b80bd0b3e880990f5d394fa19b26d +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer12/run4/train.log b/results_tencent/w2v2_large_asr_zh/layer12/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a88cd5b679063cdf2f01ea86d36bf4b7f907ff55 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer12/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5002886242866517 +INFO:root: +[Valid][0][UTT][ MSE = 0.3571 | LCC = 0.9096 | SRCC = 0.9069 ] +INFO:root:Average loss=0.2326894006729126 +INFO:root: +[Valid][1][UTT][ MSE = 0.2899 | LCC = 0.9317 | SRCC = 0.9317 ] +INFO:root:Average loss=0.19663790810108184 +INFO:root: +[Valid][2][UTT][ MSE = 0.2878 | LCC = 0.9385 | SRCC = 0.9389 ] +INFO:root:Average loss=0.16016513162851334 +INFO:root: +[Valid][3][UTT][ MSE = 0.2312 | LCC = 0.9440 | SRCC = 0.9451 ] +INFO:root:Average loss=0.13494513338804245 +INFO:root: +[Valid][4][UTT][ MSE = 0.2393 | LCC = 0.9490 | SRCC = 0.9490 ] +INFO:root:Average loss=0.12010506105422973 +INFO:root: +[Valid][5][UTT][ MSE = 0.2214 | LCC = 0.9493 | SRCC = 0.9507 ] +INFO:root:Average loss=0.11611504745483399 +INFO:root: +[Valid][6][UTT][ MSE = 0.2275 | LCC = 0.9529 | SRCC = 0.9535 ] +INFO:root:Average loss=0.10615294775366783 +INFO:root: +[Valid][7][UTT][ MSE = 0.2621 | LCC = 0.9540 | SRCC = 0.9542 ] +INFO:root:Average loss=0.10185268160700799 +INFO:root: +[Valid][8][UTT][ MSE = 0.2073 | LCC = 0.9543 | SRCC = 0.9540 ] +INFO:root:Average loss=0.09323562970757485 +INFO:root: +[Valid][9][UTT][ MSE = 0.2002 | LCC = 0.9548 | SRCC = 0.9548 ] +INFO:root:Average loss=0.08525131967663765 +INFO:root: +[Valid][10][UTT][ MSE = 0.2043 | LCC = 0.9571 | SRCC = 0.9566 ] +INFO:root:Average loss=0.07840751469135285 +INFO:root: +[Valid][11][UTT][ MSE = 0.2185 | LCC = 0.9556 | SRCC = 0.9556 ] +INFO:root:Average loss=0.07456205379962921 +INFO:root: +[Valid][12][UTT][ MSE = 0.2100 | LCC = 0.9542 | SRCC = 0.9539 ] +INFO:root:Average loss=0.07150216519832611 +INFO:root: +[Valid][13][UTT][ MSE = 0.2021 | LCC = 0.9559 | SRCC = 0.9562 ] +INFO:root:Average loss=0.06960974222421647 +INFO:root: +[Valid][14][UTT][ MSE = 0.2050 | LCC = 0.9562 | SRCC = 0.9564 ] +INFO:root:Average loss=0.06254170425236225 +INFO:root: +[Valid][15][UTT][ MSE = 0.2014 | LCC = 0.9562 | SRCC = 0.9561 ] +INFO:root:Average loss=0.05929239848256111 +INFO:root: +[Valid][16][UTT][ MSE = 0.1892 | LCC = 0.9582 | SRCC = 0.9577 ] +INFO:root:Average loss=0.054157464131712914 +INFO:root: +[Valid][17][UTT][ MSE = 0.2234 | LCC = 0.9578 | SRCC = 0.9572 ] +INFO:root:Average loss=0.053138326972723006 +INFO:root: +[Valid][18][UTT][ MSE = 0.1952 | LCC = 0.9567 | SRCC = 0.9561 ] +INFO:root:Average loss=0.04966131646931171 +INFO:root: +[Valid][19][UTT][ MSE = 0.1715 | LCC = 0.9574 | SRCC = 0.9572 ] +INFO:root:Average loss=0.049701078861951825 +INFO:root: +[Valid][20][UTT][ MSE = 0.2111 | LCC = 0.9591 | SRCC = 0.9581 ] +INFO:root:Average loss=0.04591572254896164 +INFO:root: +[Valid][21][UTT][ MSE = 0.1976 | LCC = 0.9593 | SRCC = 0.9585 ] +INFO:root:Average loss=0.044481241151690484 +INFO:root: +[Valid][22][UTT][ MSE = 0.2030 | LCC = 0.9584 | SRCC = 0.9579 ] +INFO:root:Average loss=0.04376052284240722 +INFO:root: +[Valid][23][UTT][ MSE = 0.1826 | LCC = 0.9568 | SRCC = 0.9572 ] +INFO:root:Average loss=0.04125824378430843 +INFO:root: +[Valid][24][UTT][ MSE = 0.1677 | LCC = 0.9578 | SRCC = 0.9583 ] +INFO:root:Average loss=0.03914964163303375 +INFO:root: +[Valid][25][UTT][ MSE = 0.1720 | LCC = 0.9593 | SRCC = 0.9584 ] +INFO:root:Average loss=0.03763870859146118 +INFO:root: +[Valid][26][UTT][ MSE = 0.1896 | LCC = 0.9592 | SRCC = 0.9586 ] +INFO:root:Average loss=0.036958078637719154 +INFO:root: +[Valid][27][UTT][ MSE = 0.2188 | LCC = 0.9580 | SRCC = 0.9575 ] +INFO:root:Average loss=0.03420894558727741 +INFO:root: +[Valid][28][UTT][ MSE = 0.1947 | LCC = 0.9590 | SRCC = 0.9584 ] +INFO:root:Average loss=0.03543461112678051 +INFO:root: +[Valid][29][UTT][ MSE = 0.2008 | LCC = 0.9602 | SRCC = 0.9595 ] +INFO:root:Average loss=0.032351033344864846 +INFO:root: +[Valid][30][UTT][ MSE = 0.1881 | LCC = 0.9580 | SRCC = 0.9582 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2154 | LCC = 0.9550 | SRCC = 0.9555 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run0/config.gin b/results_tencent/w2v2_large_asr_zh/layer14/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run0/model.pt b/results_tencent/w2v2_large_asr_zh/layer14/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bdde2883415338273a7607134cc4d797b29f6fd3 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:888708f0f01efcb656298875244717a6ce44241c2950b847b7f8943f5392082c +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run0/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer14/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..93e3505d5eeed82b32123b0fb8ba48ae81c7a92a --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6a78e5edeb7e190a07f7afeae4653606a058742b6982b634de2b4402e138afd +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run0/train.log b/results_tencent/w2v2_large_asr_zh/layer14/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5da942f4eb0a6771da5253275ec5e4a90d5ad4a8 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5109517790079117 +INFO:root: +[Valid][0][UTT][ MSE = 0.3435 | LCC = 0.9044 | SRCC = 0.9048 ] +INFO:root:Average loss=0.22975022971630096 +INFO:root: +[Valid][1][UTT][ MSE = 0.3447 | LCC = 0.9283 | SRCC = 0.9285 ] +INFO:root:Average loss=0.18773834902048112 +INFO:root: +[Valid][2][UTT][ MSE = 0.2784 | LCC = 0.9403 | SRCC = 0.9398 ] +INFO:root:Average loss=0.16000844901800157 +INFO:root: +[Valid][3][UTT][ MSE = 0.2593 | LCC = 0.9444 | SRCC = 0.9449 ] +INFO:root:Average loss=0.14849042201042176 +INFO:root: +[Valid][4][UTT][ MSE = 0.2700 | LCC = 0.9485 | SRCC = 0.9496 ] +INFO:root:Average loss=0.1311000314950943 +INFO:root: +[Valid][5][UTT][ MSE = 0.2148 | LCC = 0.9503 | SRCC = 0.9513 ] +INFO:root:Average loss=0.1282151637673378 +INFO:root: +[Valid][6][UTT][ MSE = 0.2475 | LCC = 0.9465 | SRCC = 0.9504 ] +INFO:root:Average loss=0.11570179206132888 +INFO:root: +[Valid][7][UTT][ MSE = 0.2373 | LCC = 0.9507 | SRCC = 0.9522 ] +INFO:root:Average loss=0.10137863618135452 +INFO:root: +[Valid][8][UTT][ MSE = 0.2438 | LCC = 0.9516 | SRCC = 0.9525 ] +INFO:root:Average loss=0.0982653997540474 +INFO:root: +[Valid][9][UTT][ MSE = 0.2244 | LCC = 0.9553 | SRCC = 0.9555 ] +INFO:root:Average loss=0.08935850164294243 +INFO:root: +[Valid][10][UTT][ MSE = 0.2708 | LCC = 0.9555 | SRCC = 0.9565 ] +INFO:root:Average loss=0.08741755786538125 +INFO:root: +[Valid][11][UTT][ MSE = 0.2083 | LCC = 0.9545 | SRCC = 0.9551 ] +INFO:root:Average loss=0.08191058614850044 +INFO:root: +[Valid][12][UTT][ MSE = 0.2442 | LCC = 0.9529 | SRCC = 0.9533 ] +INFO:root:Average loss=0.07760071060061455 +INFO:root: +[Valid][13][UTT][ MSE = 0.2384 | LCC = 0.9509 | SRCC = 0.9516 ] +INFO:root:Average loss=0.06933818557858468 +INFO:root: +[Valid][14][UTT][ MSE = 0.2120 | LCC = 0.9536 | SRCC = 0.9538 ] +INFO:root:Average loss=0.06979879131913184 +INFO:root: +[Valid][15][UTT][ MSE = 0.2696 | LCC = 0.9555 | SRCC = 0.9565 ] +INFO:root:Average loss=0.062396496504545214 +INFO:root: +[Valid][16][UTT][ MSE = 0.2311 | LCC = 0.9569 | SRCC = 0.9565 ] +INFO:root:Average loss=0.062099581837654114 +INFO:root: +[Valid][17][UTT][ MSE = 0.2358 | LCC = 0.9551 | SRCC = 0.9548 ] +INFO:root:Average loss=0.060333644449710844 +INFO:root: +[Valid][18][UTT][ MSE = 0.2239 | LCC = 0.9568 | SRCC = 0.9569 ] +INFO:root:Average loss=0.05604608097672462 +INFO:root: +[Valid][19][UTT][ MSE = 0.2220 | LCC = 0.9539 | SRCC = 0.9535 ] +INFO:root:Average loss=0.052055047884583476 +INFO:root: +[Valid][20][UTT][ MSE = 0.2233 | LCC = 0.9560 | SRCC = 0.9556 ] +INFO:root:Average loss=0.05003929430246353 +INFO:root: +[Valid][21][UTT][ MSE = 0.2047 | LCC = 0.9569 | SRCC = 0.9560 ] +INFO:root:Average loss=0.04939912715554237 +INFO:root: +[Valid][22][UTT][ MSE = 0.2324 | LCC = 0.9549 | SRCC = 0.9538 ] +INFO:root:Average loss=0.049068619191646574 +INFO:root: +[Valid][23][UTT][ MSE = 0.2362 | LCC = 0.9567 | SRCC = 0.9565 ] +INFO:root:Average loss=0.04447971346974373 +INFO:root: +[Valid][24][UTT][ MSE = 0.2316 | LCC = 0.9550 | SRCC = 0.9554 ] +INFO:root:Average loss=0.045427131652832034 +INFO:root: +[Valid][25][UTT][ MSE = 0.2224 | LCC = 0.9529 | SRCC = 0.9522 ] +INFO:root:Average loss=0.041364862814545635 +INFO:root: +[Valid][26][UTT][ MSE = 0.2265 | LCC = 0.9567 | SRCC = 0.9554 ] +INFO:root:Average loss=0.041981010615825654 +INFO:root: +[Valid][27][UTT][ MSE = 0.2445 | LCC = 0.9543 | SRCC = 0.9546 ] +INFO:root:Average loss=0.03976295858621597 +INFO:root: +[Valid][28][UTT][ MSE = 0.2104 | LCC = 0.9566 | SRCC = 0.9558 ] +INFO:root:Average loss=0.03768674306571484 +INFO:root: +[Valid][29][UTT][ MSE = 0.2400 | LCC = 0.9542 | SRCC = 0.9541 ] +INFO:root:Average loss=0.035563146829605104 +INFO:root: +[Valid][30][UTT][ MSE = 0.1990 | LCC = 0.9552 | SRCC = 0.9548 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2171 | LCC = 0.9529 | SRCC = 0.9530 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run1/config.gin b/results_tencent/w2v2_large_asr_zh/layer14/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run1/model.pt b/results_tencent/w2v2_large_asr_zh/layer14/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..90f084c60494630b75f3092ab8f9c02dd08c0ecc --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:691bd7595512aadd62926cc079bf65e9dde8b12d789b24f6a267a1819b4432c7 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run1/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer14/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d5af3dad989b08bd7b03fda7532e2996a215a4e1 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:de37bfb33530b1ce7c623b8eddf6dba73bef4d2178afd11bb25b3d2e94f830b5 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run1/train.log b/results_tencent/w2v2_large_asr_zh/layer14/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..8aad4c993893116cf85cd61d0bd5b5c19b4adf75 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4180321968793869 +INFO:root: +[Valid][0][UTT][ MSE = 0.3941 | LCC = 0.9080 | SRCC = 0.9059 ] +INFO:root:Average loss=0.23571402305364608 +INFO:root: +[Valid][1][UTT][ MSE = 0.3034 | LCC = 0.9267 | SRCC = 0.9249 ] +INFO:root:Average loss=0.19024450862407685 +INFO:root: +[Valid][2][UTT][ MSE = 0.2827 | LCC = 0.9411 | SRCC = 0.9416 ] +INFO:root:Average loss=0.15712190669775009 +INFO:root: +[Valid][3][UTT][ MSE = 0.2661 | LCC = 0.9436 | SRCC = 0.9456 ] +INFO:root:Average loss=0.1441310303211212 +INFO:root: +[Valid][4][UTT][ MSE = 0.2453 | LCC = 0.9485 | SRCC = 0.9495 ] +INFO:root:Average loss=0.13045112192630767 +INFO:root: +[Valid][5][UTT][ MSE = 0.2137 | LCC = 0.9490 | SRCC = 0.9490 ] +INFO:root:Average loss=0.1214940379858017 +INFO:root: +[Valid][6][UTT][ MSE = 0.2194 | LCC = 0.9510 | SRCC = 0.9514 ] +INFO:root:Average loss=0.11440473812818527 +INFO:root: +[Valid][7][UTT][ MSE = 0.2962 | LCC = 0.9486 | SRCC = 0.9503 ] +INFO:root:Average loss=0.10572516211867332 +INFO:root: +[Valid][8][UTT][ MSE = 0.1984 | LCC = 0.9521 | SRCC = 0.9521 ] +INFO:root:Average loss=0.0973666490316391 +INFO:root: +[Valid][9][UTT][ MSE = 0.1903 | LCC = 0.9528 | SRCC = 0.9525 ] +INFO:root:Average loss=0.09301068216562271 +INFO:root: +[Valid][10][UTT][ MSE = 0.2450 | LCC = 0.9534 | SRCC = 0.9542 ] +INFO:root:Average loss=0.08934582236409187 +INFO:root: +[Valid][11][UTT][ MSE = 0.2082 | LCC = 0.9542 | SRCC = 0.9537 ] +INFO:root:Average loss=0.08109873166680336 +INFO:root: +[Valid][12][UTT][ MSE = 0.2237 | LCC = 0.9527 | SRCC = 0.9539 ] +INFO:root:Average loss=0.07865429690480233 +INFO:root: +[Valid][13][UTT][ MSE = 0.2154 | LCC = 0.9542 | SRCC = 0.9546 ] +INFO:root:Average loss=0.07278497123718261 +INFO:root: +[Valid][14][UTT][ MSE = 0.2225 | LCC = 0.9557 | SRCC = 0.9555 ] +INFO:root:Average loss=0.0700671385526657 +INFO:root: +[Valid][15][UTT][ MSE = 0.2116 | LCC = 0.9553 | SRCC = 0.9553 ] +INFO:root:Average loss=0.06690411594510079 +INFO:root: +[Valid][16][UTT][ MSE = 0.2072 | LCC = 0.9562 | SRCC = 0.9563 ] +INFO:root:Average loss=0.06181754437088966 +INFO:root: +[Valid][17][UTT][ MSE = 0.2242 | LCC = 0.9542 | SRCC = 0.9554 ] +INFO:root:Average loss=0.06084394192695618 +INFO:root: +[Valid][18][UTT][ MSE = 0.2192 | LCC = 0.9550 | SRCC = 0.9559 ] +INFO:root:Average loss=0.05616339416801929 +INFO:root: +[Valid][19][UTT][ MSE = 0.1786 | LCC = 0.9565 | SRCC = 0.9563 ] +INFO:root:Average loss=0.054617374062538145 +INFO:root: +[Valid][20][UTT][ MSE = 0.2039 | LCC = 0.9564 | SRCC = 0.9564 ] +INFO:root:Average loss=0.055642233669757846 +INFO:root: +[Valid][21][UTT][ MSE = 0.2012 | LCC = 0.9574 | SRCC = 0.9574 ] +INFO:root:Average loss=0.05086420366168022 +INFO:root: +[Valid][22][UTT][ MSE = 0.2141 | LCC = 0.9561 | SRCC = 0.9555 ] +INFO:root:Average loss=0.04933376707136631 +INFO:root: +[Valid][23][UTT][ MSE = 0.2009 | LCC = 0.9570 | SRCC = 0.9565 ] +INFO:root:Average loss=0.04818790504336357 +INFO:root: +[Valid][24][UTT][ MSE = 0.2130 | LCC = 0.9565 | SRCC = 0.9572 ] +INFO:root:Average loss=0.04599669805169106 +INFO:root: +[Valid][25][UTT][ MSE = 0.2080 | LCC = 0.9571 | SRCC = 0.9572 ] +INFO:root:Average loss=0.04423443399369717 +INFO:root: +[Valid][26][UTT][ MSE = 0.2200 | LCC = 0.9577 | SRCC = 0.9576 ] +INFO:root:Average loss=0.042514045476913453 +INFO:root: +[Valid][27][UTT][ MSE = 0.1984 | LCC = 0.9579 | SRCC = 0.9581 ] +INFO:root:Average loss=0.04050559562444687 +INFO:root: +[Valid][28][UTT][ MSE = 0.2047 | LCC = 0.9573 | SRCC = 0.9574 ] +INFO:root:Average loss=0.03914749452471733 +INFO:root: +[Valid][29][UTT][ MSE = 0.2085 | LCC = 0.9573 | SRCC = 0.9566 ] +INFO:root:Average loss=0.04008727651834488 +INFO:root: +[Valid][30][UTT][ MSE = 0.2103 | LCC = 0.9559 | SRCC = 0.9558 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2092 | LCC = 0.9536 | SRCC = 0.9543 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run2/config.gin b/results_tencent/w2v2_large_asr_zh/layer14/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run2/model.pt b/results_tencent/w2v2_large_asr_zh/layer14/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2d8236a1ca6f094817e39454413a181dbcc8136a --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f772765df4f35e99751231d9236c54fb40846e1b97ae87f9202eaa7eb76545ce +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run2/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer14/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ee2db3ad6acb6ef92bd8fc1a9a5c955166676d5f --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:52d3822e60789cec95b207f541df286e040eb7d0f91ff5abd4cc91c5310b430b +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run2/train.log b/results_tencent/w2v2_large_asr_zh/layer14/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b9ebc8924d81c15bfd014acf17d3c2f5cb263bd2 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5025919516086579 +INFO:root: +[Valid][0][UTT][ MSE = 0.3663 | LCC = 0.9083 | SRCC = 0.9054 ] +INFO:root:Average loss=0.22855446976423263 +INFO:root: +[Valid][1][UTT][ MSE = 0.3421 | LCC = 0.9297 | SRCC = 0.9309 ] +INFO:root:Average loss=0.1896907719373703 +INFO:root: +[Valid][2][UTT][ MSE = 0.3340 | LCC = 0.9345 | SRCC = 0.9362 ] +INFO:root:Average loss=0.16921808105707167 +INFO:root: +[Valid][3][UTT][ MSE = 0.2586 | LCC = 0.9436 | SRCC = 0.9447 ] +INFO:root:Average loss=0.1473254761695862 +INFO:root: +[Valid][4][UTT][ MSE = 0.3062 | LCC = 0.9461 | SRCC = 0.9481 ] +INFO:root:Average loss=0.13110121822357176 +INFO:root: +[Valid][5][UTT][ MSE = 0.2587 | LCC = 0.9465 | SRCC = 0.9484 ] +INFO:root:Average loss=0.11516794019937515 +INFO:root: +[Valid][6][UTT][ MSE = 0.2954 | LCC = 0.9470 | SRCC = 0.9506 ] +INFO:root:Average loss=0.11334882068634033 +INFO:root: +[Valid][7][UTT][ MSE = 0.2179 | LCC = 0.9532 | SRCC = 0.9534 ] +INFO:root:Average loss=0.09957668495178222 +INFO:root: +[Valid][8][UTT][ MSE = 0.2359 | LCC = 0.9523 | SRCC = 0.9528 ] +INFO:root:Average loss=0.09524314248561859 +INFO:root: +[Valid][9][UTT][ MSE = 0.2258 | LCC = 0.9506 | SRCC = 0.9515 ] +INFO:root:Average loss=0.08930075114965438 +INFO:root: +[Valid][10][UTT][ MSE = 0.1946 | LCC = 0.9554 | SRCC = 0.9552 ] +INFO:root:Average loss=0.08460972034931183 +INFO:root: +[Valid][11][UTT][ MSE = 0.2241 | LCC = 0.9542 | SRCC = 0.9536 ] +INFO:root:Average loss=0.07716690281033516 +INFO:root: +[Valid][12][UTT][ MSE = 0.2042 | LCC = 0.9547 | SRCC = 0.9539 ] +INFO:root:Average loss=0.07205391511321067 +INFO:root: +[Valid][13][UTT][ MSE = 0.2430 | LCC = 0.9540 | SRCC = 0.9544 ] +INFO:root:Average loss=0.07408451586961746 +INFO:root: +[Valid][14][UTT][ MSE = 0.2281 | LCC = 0.9542 | SRCC = 0.9547 ] +INFO:root:Average loss=0.06719810771942139 +INFO:root: +[Valid][15][UTT][ MSE = 0.1963 | LCC = 0.9551 | SRCC = 0.9547 ] +INFO:root:Average loss=0.06686576125025749 +INFO:root: +[Valid][16][UTT][ MSE = 0.2640 | LCC = 0.9534 | SRCC = 0.9555 ] +INFO:root:Average loss=0.05867238262295723 +INFO:root: +[Valid][17][UTT][ MSE = 0.2093 | LCC = 0.9549 | SRCC = 0.9548 ] +INFO:root:Average loss=0.05815768441557884 +INFO:root: +[Valid][18][UTT][ MSE = 0.2280 | LCC = 0.9555 | SRCC = 0.9558 ] +INFO:root:Average loss=0.05472321280837059 +INFO:root: +[Valid][19][UTT][ MSE = 0.2204 | LCC = 0.9546 | SRCC = 0.9548 ] +INFO:root:Average loss=0.05435580968856812 +INFO:root: +[Valid][20][UTT][ MSE = 0.2378 | LCC = 0.9546 | SRCC = 0.9554 ] +INFO:root:Average loss=0.049431065022945406 +INFO:root: +[Valid][21][UTT][ MSE = 0.2280 | LCC = 0.9564 | SRCC = 0.9559 ] +INFO:root:Average loss=0.04782814660668373 +INFO:root: +[Valid][22][UTT][ MSE = 0.2002 | LCC = 0.9565 | SRCC = 0.9559 ] +INFO:root:Average loss=0.04789072445034981 +INFO:root: +[Valid][23][UTT][ MSE = 0.2228 | LCC = 0.9554 | SRCC = 0.9554 ] +INFO:root:Average loss=0.0445939195305109 +INFO:root: +[Valid][24][UTT][ MSE = 0.2638 | LCC = 0.9556 | SRCC = 0.9551 ] +INFO:root:Average loss=0.04262061858177185 +INFO:root: +[Valid][25][UTT][ MSE = 0.2298 | LCC = 0.9558 | SRCC = 0.9557 ] +INFO:root:Average loss=0.04133531369268894 +INFO:root: +[Valid][26][UTT][ MSE = 0.2151 | LCC = 0.9566 | SRCC = 0.9563 ] +INFO:root:Average loss=0.040036014661192894 +INFO:root: +[Valid][27][UTT][ MSE = 0.2295 | LCC = 0.9547 | SRCC = 0.9542 ] +INFO:root:Average loss=0.03915920922160149 +INFO:root: +[Valid][28][UTT][ MSE = 0.2104 | LCC = 0.9558 | SRCC = 0.9551 ] +INFO:root:Average loss=0.037663949131965635 +INFO:root: +[Valid][29][UTT][ MSE = 0.2345 | LCC = 0.9552 | SRCC = 0.9548 ] +INFO:root:Average loss=0.035415126070380214 +INFO:root: +[Valid][30][UTT][ MSE = 0.2277 | LCC = 0.9557 | SRCC = 0.9557 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2295 | LCC = 0.9528 | SRCC = 0.9538 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run3/config.gin b/results_tencent/w2v2_large_asr_zh/layer14/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run3/model.pt b/results_tencent/w2v2_large_asr_zh/layer14/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7eea76d3377d5106ca7d1e4ef136d34c886393f7 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6bc708335fae03d480cb2f76eb123745aaa29396b82777e31af28493ad8bd59 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run3/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer14/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7a2e9ce7e750d07be81efe1c6e4a6bbc8f06cdc0 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f262140c40d64a9de41f9461688cb3e262a9f6643b2a01e2d0931ebcf7e5fc9b +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run3/train.log b/results_tencent/w2v2_large_asr_zh/layer14/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..36a5fd31b6e1d277b82b69eed443d1a552590802 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.49836868810653684 +INFO:root: +[Valid][0][UTT][ MSE = 0.3632 | LCC = 0.9022 | SRCC = 0.9019 ] +INFO:root:Average loss=0.2421994045972824 +INFO:root: +[Valid][1][UTT][ MSE = 0.3329 | LCC = 0.9274 | SRCC = 0.9269 ] +INFO:root:Average loss=0.1950742652416229 +INFO:root: +[Valid][2][UTT][ MSE = 0.3239 | LCC = 0.9337 | SRCC = 0.9343 ] +INFO:root:Average loss=0.17389718687534333 +INFO:root: +[Valid][3][UTT][ MSE = 0.2378 | LCC = 0.9416 | SRCC = 0.9409 ] +INFO:root:Average loss=0.1527961004972458 +INFO:root: +[Valid][4][UTT][ MSE = 0.2533 | LCC = 0.9428 | SRCC = 0.9434 ] +INFO:root:Average loss=0.13105271810293198 +INFO:root: +[Valid][5][UTT][ MSE = 0.2028 | LCC = 0.9476 | SRCC = 0.9477 ] +INFO:root:Average loss=0.12602961313724517 +INFO:root: +[Valid][6][UTT][ MSE = 0.2178 | LCC = 0.9467 | SRCC = 0.9468 ] +INFO:root:Average loss=0.11472795128822327 +INFO:root: +[Valid][7][UTT][ MSE = 0.2008 | LCC = 0.9506 | SRCC = 0.9506 ] +INFO:root:Average loss=0.10557521864771843 +INFO:root: +[Valid][8][UTT][ MSE = 0.1996 | LCC = 0.9522 | SRCC = 0.9524 ] +INFO:root:Average loss=0.09859285223484039 +INFO:root: +[Valid][9][UTT][ MSE = 0.2244 | LCC = 0.9517 | SRCC = 0.9526 ] +INFO:root:Average loss=0.09703372317552567 +INFO:root: +[Valid][10][UTT][ MSE = 0.2547 | LCC = 0.9505 | SRCC = 0.9515 ] +INFO:root:Average loss=0.08827392423152923 +INFO:root: +[Valid][11][UTT][ MSE = 0.2209 | LCC = 0.9527 | SRCC = 0.9534 ] +INFO:root:Average loss=0.08300836768746377 +INFO:root: +[Valid][12][UTT][ MSE = 0.1744 | LCC = 0.9530 | SRCC = 0.9535 ] +INFO:root:Average loss=0.07814411553740501 +INFO:root: +[Valid][13][UTT][ MSE = 0.2003 | LCC = 0.9541 | SRCC = 0.9543 ] +INFO:root:Average loss=0.07942025178670883 +INFO:root: +[Valid][14][UTT][ MSE = 0.1939 | LCC = 0.9555 | SRCC = 0.9554 ] +INFO:root:Average loss=0.0693521134853363 +INFO:root: +[Valid][15][UTT][ MSE = 0.2151 | LCC = 0.9532 | SRCC = 0.9534 ] +INFO:root:Average loss=0.06736412751674652 +INFO:root: +[Valid][16][UTT][ MSE = 0.1916 | LCC = 0.9547 | SRCC = 0.9555 ] +INFO:root:Average loss=0.06389132609963417 +INFO:root: +[Valid][17][UTT][ MSE = 0.2145 | LCC = 0.9556 | SRCC = 0.9557 ] +INFO:root:Average loss=0.06368098014593124 +INFO:root: +[Valid][18][UTT][ MSE = 0.2112 | LCC = 0.9550 | SRCC = 0.9554 ] +INFO:root:Average loss=0.05833317330479622 +INFO:root: +[Valid][19][UTT][ MSE = 0.1812 | LCC = 0.9555 | SRCC = 0.9550 ] +INFO:root:Average loss=0.05615196403861046 +INFO:root: +[Valid][20][UTT][ MSE = 0.1933 | LCC = 0.9566 | SRCC = 0.9565 ] +INFO:root:Average loss=0.0538218843638897 +INFO:root: +[Valid][21][UTT][ MSE = 0.2136 | LCC = 0.9552 | SRCC = 0.9552 ] +INFO:root:Average loss=0.051029512137174604 +INFO:root: +[Valid][22][UTT][ MSE = 0.2045 | LCC = 0.9558 | SRCC = 0.9555 ] +INFO:root:Average loss=0.04895744940638542 +INFO:root: +[Valid][23][UTT][ MSE = 0.1990 | LCC = 0.9569 | SRCC = 0.9564 ] +INFO:root:Average loss=0.04652087301015854 +INFO:root: +[Valid][24][UTT][ MSE = 0.1792 | LCC = 0.9542 | SRCC = 0.9543 ] +INFO:root:Average loss=0.04536440840363502 +INFO:root: +[Valid][25][UTT][ MSE = 0.2055 | LCC = 0.9567 | SRCC = 0.9565 ] +INFO:root:Average loss=0.04336391785740852 +INFO:root: +[Valid][26][UTT][ MSE = 0.1961 | LCC = 0.9572 | SRCC = 0.9569 ] +INFO:root:Average loss=0.04249380278587341 +INFO:root: +[Valid][27][UTT][ MSE = 0.1918 | LCC = 0.9566 | SRCC = 0.9562 ] +INFO:root:Average loss=0.04165319356322288 +INFO:root: +[Valid][28][UTT][ MSE = 0.2207 | LCC = 0.9550 | SRCC = 0.9556 ] +INFO:root:Average loss=0.042219074353575706 +INFO:root: +[Valid][29][UTT][ MSE = 0.1810 | LCC = 0.9548 | SRCC = 0.9546 ] +INFO:root:Average loss=0.03716967561841011 +INFO:root: +[Valid][30][UTT][ MSE = 0.1951 | LCC = 0.9558 | SRCC = 0.9555 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2100 | LCC = 0.9532 | SRCC = 0.9530 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run4/config.gin b/results_tencent/w2v2_large_asr_zh/layer14/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run4/model.pt b/results_tencent/w2v2_large_asr_zh/layer14/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..18e7253322282b49b9fccb215e934a3d90009b2b --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6040aedb8fdcf13e7ddb100fc7298860fc07f90ef99d6124f582b6d23644bea5 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run4/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer14/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..113a07619249e1e651377f2fe26b297eae2dbb30 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f866b6c380668448d3b91144fd6749ccb42c9345a04f92e5ef8973b80b3b3a66 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer14/run4/train.log b/results_tencent/w2v2_large_asr_zh/layer14/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f672199f98c7f69d9b40ed36be64133914069399 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer14/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5369585739374161 +INFO:root: +[Valid][0][UTT][ MSE = 0.4284 | LCC = 0.9036 | SRCC = 0.9001 ] +INFO:root:Average loss=0.23928037637472152 +INFO:root: +[Valid][1][UTT][ MSE = 0.2880 | LCC = 0.9241 | SRCC = 0.9270 ] +INFO:root:Average loss=0.1880605424642563 +INFO:root: +[Valid][2][UTT][ MSE = 0.2544 | LCC = 0.9347 | SRCC = 0.9361 ] +INFO:root:Average loss=0.15847524058818818 +INFO:root: +[Valid][3][UTT][ MSE = 0.2921 | LCC = 0.9458 | SRCC = 0.9455 ] +INFO:root:Average loss=0.14372475838661194 +INFO:root: +[Valid][4][UTT][ MSE = 0.2185 | LCC = 0.9426 | SRCC = 0.9451 ] +INFO:root:Average loss=0.1266924131512642 +INFO:root: +[Valid][5][UTT][ MSE = 0.2546 | LCC = 0.9498 | SRCC = 0.9504 ] +INFO:root:Average loss=0.12032174742221832 +INFO:root: +[Valid][6][UTT][ MSE = 0.2169 | LCC = 0.9524 | SRCC = 0.9526 ] +INFO:root:Average loss=0.10984767737984658 +INFO:root: +[Valid][7][UTT][ MSE = 0.2015 | LCC = 0.9466 | SRCC = 0.9493 ] +INFO:root:Average loss=0.10389167276024819 +INFO:root: +[Valid][8][UTT][ MSE = 0.2239 | LCC = 0.9539 | SRCC = 0.9545 ] +INFO:root:Average loss=0.09432206773757934 +INFO:root: +[Valid][9][UTT][ MSE = 0.2078 | LCC = 0.9542 | SRCC = 0.9543 ] +INFO:root:Average loss=0.09264223203063011 +INFO:root: +[Valid][10][UTT][ MSE = 0.2141 | LCC = 0.9527 | SRCC = 0.9535 ] +INFO:root:Average loss=0.08900809180736542 +INFO:root: +[Valid][11][UTT][ MSE = 0.2177 | LCC = 0.9546 | SRCC = 0.9545 ] +INFO:root:Average loss=0.08227638754248619 +INFO:root: +[Valid][12][UTT][ MSE = 0.2036 | LCC = 0.9533 | SRCC = 0.9546 ] +INFO:root:Average loss=0.0751623746752739 +INFO:root: +[Valid][13][UTT][ MSE = 0.2368 | LCC = 0.9524 | SRCC = 0.9530 ] +INFO:root:Average loss=0.07194585919380188 +INFO:root: +[Valid][14][UTT][ MSE = 0.2184 | LCC = 0.9568 | SRCC = 0.9576 ] +INFO:root:Average loss=0.06650600278377533 +INFO:root: +[Valid][15][UTT][ MSE = 0.2431 | LCC = 0.9556 | SRCC = 0.9565 ] +INFO:root:Average loss=0.06255708822607994 +INFO:root: +[Valid][16][UTT][ MSE = 0.2038 | LCC = 0.9555 | SRCC = 0.9558 ] +INFO:root:Average loss=0.06287981671094894 +INFO:root: +[Valid][17][UTT][ MSE = 0.2272 | LCC = 0.9551 | SRCC = 0.9546 ] +INFO:root:Average loss=0.057202013179659845 +INFO:root: +[Valid][18][UTT][ MSE = 0.2150 | LCC = 0.9565 | SRCC = 0.9561 ] +INFO:root:Average loss=0.05358145308494568 +INFO:root: +[Valid][19][UTT][ MSE = 0.1996 | LCC = 0.9537 | SRCC = 0.9538 ] +INFO:root:Average loss=0.05429888555407524 +INFO:root: +[Valid][20][UTT][ MSE = 0.2071 | LCC = 0.9570 | SRCC = 0.9567 ] +INFO:root:Average loss=0.050962016105651854 +INFO:root: +[Valid][21][UTT][ MSE = 0.2243 | LCC = 0.9560 | SRCC = 0.9559 ] +INFO:root:Average loss=0.051420026317238805 +INFO:root: +[Valid][22][UTT][ MSE = 0.2061 | LCC = 0.9542 | SRCC = 0.9547 ] +INFO:root:Average loss=0.04529250229895115 +INFO:root: +[Valid][23][UTT][ MSE = 0.2230 | LCC = 0.9558 | SRCC = 0.9552 ] +INFO:root:Average loss=0.044302967712283134 +INFO:root: +[Valid][24][UTT][ MSE = 0.2187 | LCC = 0.9539 | SRCC = 0.9544 ] +INFO:root:Average loss=0.0421888590157032 +INFO:root: +[Valid][25][UTT][ MSE = 0.2238 | LCC = 0.9562 | SRCC = 0.9565 ] +INFO:root:Average loss=0.04413069443404675 +INFO:root: +[Valid][26][UTT][ MSE = 0.2424 | LCC = 0.9549 | SRCC = 0.9550 ] +INFO:root:Average loss=0.040094147577881815 +INFO:root: +[Valid][27][UTT][ MSE = 0.1947 | LCC = 0.9568 | SRCC = 0.9567 ] +INFO:root:Average loss=0.03770466734468937 +INFO:root: +[Valid][28][UTT][ MSE = 0.2077 | LCC = 0.9566 | SRCC = 0.9561 ] +INFO:root:Average loss=0.0407445517629385 +INFO:root: +[Valid][29][UTT][ MSE = 0.2217 | LCC = 0.9569 | SRCC = 0.9569 ] +INFO:root:Average loss=0.03588453122973442 +INFO:root: +[Valid][30][UTT][ MSE = 0.2191 | LCC = 0.9559 | SRCC = 0.9554 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2244 | LCC = 0.9515 | SRCC = 0.9523 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run0/config.gin b/results_tencent/w2v2_large_asr_zh/layer16/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run0/model.pt b/results_tencent/w2v2_large_asr_zh/layer16/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3f8e756c34634d959cfdd764e438b5ca3b4b44ee --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aa38c67fc3e3a247a4f81f2d8c518372643dd8a560e0c692cf7d82f693a3f83a +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run0/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer16/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a34afcc077c3a747fa5f8461279f76c5606f7201 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e88f00c1a8edb081f2f08fe0d9cb315821ac780cfcd26aa09a586352d129a0b +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run0/train.log b/results_tencent/w2v2_large_asr_zh/layer16/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d05c8cd35b9216ce598508ebafd2229c0ae50cbd --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4643911528587341 +INFO:root: +[Valid][0][UTT][ MSE = 0.4167 | LCC = 0.8914 | SRCC = 0.8903 ] +INFO:root:Average loss=0.24255873614549636 +INFO:root: +[Valid][1][UTT][ MSE = 0.2802 | LCC = 0.9206 | SRCC = 0.9198 ] +INFO:root:Average loss=0.205590192258358 +INFO:root: +[Valid][2][UTT][ MSE = 0.3330 | LCC = 0.9282 | SRCC = 0.9287 ] +INFO:root:Average loss=0.17399183440208435 +INFO:root: +[Valid][3][UTT][ MSE = 0.2443 | LCC = 0.9405 | SRCC = 0.9393 ] +INFO:root:Average loss=0.15439943408966064 +INFO:root: +[Valid][4][UTT][ MSE = 0.2368 | LCC = 0.9422 | SRCC = 0.9412 ] +INFO:root:Average loss=0.1415457330942154 +INFO:root: +[Valid][5][UTT][ MSE = 0.2259 | LCC = 0.9449 | SRCC = 0.9456 ] +INFO:root:Average loss=0.12699961286783218 +INFO:root: +[Valid][6][UTT][ MSE = 0.2324 | LCC = 0.9477 | SRCC = 0.9485 ] +INFO:root:Average loss=0.1214002260863781 +INFO:root: +[Valid][7][UTT][ MSE = 0.2238 | LCC = 0.9484 | SRCC = 0.9487 ] +INFO:root:Average loss=0.10949325466156005 +INFO:root: +[Valid][8][UTT][ MSE = 0.2183 | LCC = 0.9476 | SRCC = 0.9484 ] +INFO:root:Average loss=0.10390700098872185 +INFO:root: +[Valid][9][UTT][ MSE = 0.2116 | LCC = 0.9509 | SRCC = 0.9510 ] +INFO:root:Average loss=0.09434378644824028 +INFO:root: +[Valid][10][UTT][ MSE = 0.2161 | LCC = 0.9514 | SRCC = 0.9519 ] +INFO:root:Average loss=0.0896070617735386 +INFO:root: +[Valid][11][UTT][ MSE = 0.2252 | LCC = 0.9504 | SRCC = 0.9509 ] +INFO:root:Average loss=0.09001068097352982 +INFO:root: +[Valid][12][UTT][ MSE = 0.2162 | LCC = 0.9527 | SRCC = 0.9524 ] +INFO:root:Average loss=0.08461929789185524 +INFO:root: +[Valid][13][UTT][ MSE = 0.2533 | LCC = 0.9530 | SRCC = 0.9535 ] +INFO:root:Average loss=0.0783948303759098 +INFO:root: +[Valid][14][UTT][ MSE = 0.2273 | LCC = 0.9531 | SRCC = 0.9525 ] +INFO:root:Average loss=0.07570121887326241 +INFO:root: +[Valid][15][UTT][ MSE = 0.2238 | LCC = 0.9543 | SRCC = 0.9539 ] +INFO:root:Average loss=0.0725297309756279 +INFO:root: +[Valid][16][UTT][ MSE = 0.2575 | LCC = 0.9544 | SRCC = 0.9545 ] +INFO:root:Average loss=0.0680395959019661 +INFO:root: +[Valid][17][UTT][ MSE = 0.2136 | LCC = 0.9545 | SRCC = 0.9545 ] +INFO:root:Average loss=0.06451768431067467 +INFO:root: +[Valid][18][UTT][ MSE = 0.2122 | LCC = 0.9538 | SRCC = 0.9533 ] +INFO:root:Average loss=0.062465428203344345 +INFO:root: +[Valid][19][UTT][ MSE = 0.2447 | LCC = 0.9543 | SRCC = 0.9539 ] +INFO:root:Average loss=0.062321854859590534 +INFO:root: +[Valid][20][UTT][ MSE = 0.2176 | LCC = 0.9541 | SRCC = 0.9542 ] +INFO:root:Average loss=0.05838440242409706 +INFO:root: +[Valid][21][UTT][ MSE = 0.2078 | LCC = 0.9546 | SRCC = 0.9544 ] +INFO:root:Average loss=0.057285005748271944 +INFO:root: +[Valid][22][UTT][ MSE = 0.2208 | LCC = 0.9551 | SRCC = 0.9545 ] +INFO:root:Average loss=0.05505483016371727 +INFO:root: +[Valid][23][UTT][ MSE = 0.1914 | LCC = 0.9535 | SRCC = 0.9530 ] +INFO:root:Average loss=0.04852144350111485 +INFO:root: +[Valid][24][UTT][ MSE = 0.2145 | LCC = 0.9562 | SRCC = 0.9559 ] +INFO:root:Average loss=0.04707871559262276 +INFO:root: +[Valid][25][UTT][ MSE = 0.2032 | LCC = 0.9551 | SRCC = 0.9547 ] +INFO:root:Average loss=0.0460852260440588 +INFO:root: +[Valid][26][UTT][ MSE = 0.2069 | LCC = 0.9548 | SRCC = 0.9543 ] +INFO:root:Average loss=0.04803316493332386 +INFO:root: +[Valid][27][UTT][ MSE = 0.2143 | LCC = 0.9549 | SRCC = 0.9545 ] +INFO:root:Average loss=0.043982309967279434 +INFO:root: +[Valid][28][UTT][ MSE = 0.2200 | LCC = 0.9542 | SRCC = 0.9535 ] +INFO:root:Average loss=0.04159953168034553 +INFO:root: +[Valid][29][UTT][ MSE = 0.2182 | LCC = 0.9558 | SRCC = 0.9557 ] +INFO:root:Average loss=0.041540339455008506 +INFO:root: +[Valid][30][UTT][ MSE = 0.2238 | LCC = 0.9534 | SRCC = 0.9535 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2340 | LCC = 0.9494 | SRCC = 0.9496 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run1/config.gin b/results_tencent/w2v2_large_asr_zh/layer16/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run1/model.pt b/results_tencent/w2v2_large_asr_zh/layer16/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4271dc70a96a8eabc266288573402416209979e4 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:61ca7ee8b608089ae21d637bf688600186e653634d94535b4ad7aa447de8923e +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run1/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer16/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..12062fdf703c4ac7e1ba5b17a85ae1242f723fb0 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bc4139cdf57822a5e80a487e695595c6ce1464531feabcbe8f2eaab8ee54edbe +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run1/train.log b/results_tencent/w2v2_large_asr_zh/layer16/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..44e4449b8dbcfcb257cedf36f1514d1830f7ba17 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.496307311296463 +INFO:root: +[Valid][0][UTT][ MSE = 0.3954 | LCC = 0.8903 | SRCC = 0.8910 ] +INFO:root:Average loss=0.26902598202228545 +INFO:root: +[Valid][1][UTT][ MSE = 0.3555 | LCC = 0.9112 | SRCC = 0.9111 ] +INFO:root:Average loss=0.21239678865671158 +INFO:root: +[Valid][2][UTT][ MSE = 0.3248 | LCC = 0.9303 | SRCC = 0.9295 ] +INFO:root:Average loss=0.18025650411844255 +INFO:root: +[Valid][3][UTT][ MSE = 0.2458 | LCC = 0.9388 | SRCC = 0.9376 ] +INFO:root:Average loss=0.15821766662597656 +INFO:root: +[Valid][4][UTT][ MSE = 0.2613 | LCC = 0.9432 | SRCC = 0.9432 ] +INFO:root:Average loss=0.13662532484531403 +INFO:root: +[Valid][5][UTT][ MSE = 0.2061 | LCC = 0.9445 | SRCC = 0.9434 ] +INFO:root:Average loss=0.12389708530902863 +INFO:root: +[Valid][6][UTT][ MSE = 0.2540 | LCC = 0.9454 | SRCC = 0.9444 ] +INFO:root:Average loss=0.12241653925180435 +INFO:root: +[Valid][7][UTT][ MSE = 0.2156 | LCC = 0.9465 | SRCC = 0.9460 ] +INFO:root:Average loss=0.10763822549581528 +INFO:root: +[Valid][8][UTT][ MSE = 0.2398 | LCC = 0.9493 | SRCC = 0.9501 ] +INFO:root:Average loss=0.10378679606318474 +INFO:root: +[Valid][9][UTT][ MSE = 0.2229 | LCC = 0.9490 | SRCC = 0.9491 ] +INFO:root:Average loss=0.09394497308135033 +INFO:root: +[Valid][10][UTT][ MSE = 0.1960 | LCC = 0.9501 | SRCC = 0.9497 ] +INFO:root:Average loss=0.08924173551797866 +INFO:root: +[Valid][11][UTT][ MSE = 0.2117 | LCC = 0.9513 | SRCC = 0.9510 ] +INFO:root:Average loss=0.0816667338013649 +INFO:root: +[Valid][12][UTT][ MSE = 0.2048 | LCC = 0.9532 | SRCC = 0.9520 ] +INFO:root:Average loss=0.07666373708844185 +INFO:root: +[Valid][13][UTT][ MSE = 0.2029 | LCC = 0.9535 | SRCC = 0.9531 ] +INFO:root:Average loss=0.07257224455475807 +INFO:root: +[Valid][14][UTT][ MSE = 0.1999 | LCC = 0.9520 | SRCC = 0.9506 ] +INFO:root:Average loss=0.07041085797548294 +INFO:root: +[Valid][15][UTT][ MSE = 0.1979 | LCC = 0.9478 | SRCC = 0.9470 ] +INFO:root:Average loss=0.06845167207717895 +INFO:root: +[Valid][16][UTT][ MSE = 0.2393 | LCC = 0.9538 | SRCC = 0.9530 ] +INFO:root:Average loss=0.06685524532198905 +INFO:root: +[Valid][17][UTT][ MSE = 0.2002 | LCC = 0.9530 | SRCC = 0.9529 ] +INFO:root:Average loss=0.057929718464612964 +INFO:root: +[Valid][18][UTT][ MSE = 0.2096 | LCC = 0.9521 | SRCC = 0.9521 ] +INFO:root:Average loss=0.06010167986154556 +INFO:root: +[Valid][19][UTT][ MSE = 0.2214 | LCC = 0.9530 | SRCC = 0.9522 ] +INFO:root:Average loss=0.055213855028152466 +INFO:root: +[Valid][20][UTT][ MSE = 0.2207 | LCC = 0.9490 | SRCC = 0.9499 ] +INFO:root:Average loss=0.05258133317530155 +INFO:root: +[Valid][21][UTT][ MSE = 0.2248 | LCC = 0.9560 | SRCC = 0.9556 ] +INFO:root:Average loss=0.05197100326418877 +INFO:root: +[Valid][22][UTT][ MSE = 0.1981 | LCC = 0.9527 | SRCC = 0.9518 ] +INFO:root:Average loss=0.050907883182168004 +INFO:root: +[Valid][23][UTT][ MSE = 0.1862 | LCC = 0.9513 | SRCC = 0.9505 ] +INFO:root:Average loss=0.04538945837318897 +INFO:root: +[Valid][24][UTT][ MSE = 0.1959 | LCC = 0.9529 | SRCC = 0.9522 ] +INFO:root:Average loss=0.04940945838391781 +INFO:root: +[Valid][25][UTT][ MSE = 0.1906 | LCC = 0.9548 | SRCC = 0.9551 ] +INFO:root:Average loss=0.04392101763188839 +INFO:root: +[Valid][26][UTT][ MSE = 0.2016 | LCC = 0.9550 | SRCC = 0.9546 ] +INFO:root:Average loss=0.04134537677466869 +INFO:root: +[Valid][27][UTT][ MSE = 0.2075 | LCC = 0.9523 | SRCC = 0.9517 ] +INFO:root:Average loss=0.04270209275186062 +INFO:root: +[Valid][28][UTT][ MSE = 0.2381 | LCC = 0.9519 | SRCC = 0.9512 ] +INFO:root:Average loss=0.04017895114421845 +INFO:root: +[Valid][29][UTT][ MSE = 0.1784 | LCC = 0.9533 | SRCC = 0.9530 ] +INFO:root:Average loss=0.039793145805597306 +INFO:root: +[Valid][30][UTT][ MSE = 0.2042 | LCC = 0.9543 | SRCC = 0.9536 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2396 | LCC = 0.9495 | SRCC = 0.9498 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run2/config.gin b/results_tencent/w2v2_large_asr_zh/layer16/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run2/model.pt b/results_tencent/w2v2_large_asr_zh/layer16/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1c40c22c4328b179077e0b0bddda1b7147c005a0 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4016855a60a8adb0bb01a33bc733b20a0da169cfecb7606b6adc9c2b3d219745 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run2/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer16/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..26ae0ce029d19cd1004b1713ffddb00349e6a3da --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b99e1abee0056285d50e87b490dc688109533c795590104b98cc971ddafbce4e +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run2/train.log b/results_tencent/w2v2_large_asr_zh/layer16/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e0a7b5443604615f0fa696723c0417623ac2f98c --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44665410089492796 +INFO:root: +[Valid][0][UTT][ MSE = 0.3778 | LCC = 0.9034 | SRCC = 0.8994 ] +INFO:root:Average loss=0.2429692680835724 +INFO:root: +[Valid][1][UTT][ MSE = 0.2983 | LCC = 0.9241 | SRCC = 0.9227 ] +INFO:root:Average loss=0.19900283896923066 +INFO:root: +[Valid][2][UTT][ MSE = 0.2864 | LCC = 0.9357 | SRCC = 0.9357 ] +INFO:root:Average loss=0.16559576809406282 +INFO:root: +[Valid][3][UTT][ MSE = 0.2403 | LCC = 0.9407 | SRCC = 0.9405 ] +INFO:root:Average loss=0.154331957757473 +INFO:root: +[Valid][4][UTT][ MSE = 0.2734 | LCC = 0.9414 | SRCC = 0.9433 ] +INFO:root:Average loss=0.14457446002960206 +INFO:root: +[Valid][5][UTT][ MSE = 0.2617 | LCC = 0.9460 | SRCC = 0.9456 ] +INFO:root:Average loss=0.12752207791805267 +INFO:root: +[Valid][6][UTT][ MSE = 0.2621 | LCC = 0.9480 | SRCC = 0.9492 ] +INFO:root:Average loss=0.12279097765684127 +INFO:root: +[Valid][7][UTT][ MSE = 0.2608 | LCC = 0.9493 | SRCC = 0.9500 ] +INFO:root:Average loss=0.10709672948718071 +INFO:root: +[Valid][8][UTT][ MSE = 0.2167 | LCC = 0.9482 | SRCC = 0.9490 ] +INFO:root:Average loss=0.10229060578346252 +INFO:root: +[Valid][9][UTT][ MSE = 0.2616 | LCC = 0.9461 | SRCC = 0.9492 ] +INFO:root:Average loss=0.09778247526288032 +INFO:root: +[Valid][10][UTT][ MSE = 0.2407 | LCC = 0.9512 | SRCC = 0.9517 ] +INFO:root:Average loss=0.08823997044563293 +INFO:root: +[Valid][11][UTT][ MSE = 0.2260 | LCC = 0.9514 | SRCC = 0.9515 ] +INFO:root:Average loss=0.0822425527870655 +INFO:root: +[Valid][12][UTT][ MSE = 0.2205 | LCC = 0.9528 | SRCC = 0.9524 ] +INFO:root:Average loss=0.07942181852459908 +INFO:root: +[Valid][13][UTT][ MSE = 0.2067 | LCC = 0.9536 | SRCC = 0.9536 ] +INFO:root:Average loss=0.0742854299545288 +INFO:root: +[Valid][14][UTT][ MSE = 0.2278 | LCC = 0.9520 | SRCC = 0.9518 ] +INFO:root:Average loss=0.07209867060184479 +INFO:root: +[Valid][15][UTT][ MSE = 0.2484 | LCC = 0.9507 | SRCC = 0.9518 ] +INFO:root:Average loss=0.06626401215791702 +INFO:root: +[Valid][16][UTT][ MSE = 0.2184 | LCC = 0.9507 | SRCC = 0.9508 ] +INFO:root:Average loss=0.06842042315006257 +INFO:root: +[Valid][17][UTT][ MSE = 0.2200 | LCC = 0.9508 | SRCC = 0.9506 ] +INFO:root:Average loss=0.06292507418990136 +INFO:root: +[Valid][18][UTT][ MSE = 0.2176 | LCC = 0.9532 | SRCC = 0.9530 ] +INFO:root:Average loss=0.05821116864681244 +INFO:root: +[Valid][19][UTT][ MSE = 0.2413 | LCC = 0.9528 | SRCC = 0.9530 ] +INFO:root:Average loss=0.056909106940031054 +INFO:root: +[Valid][20][UTT][ MSE = 0.2060 | LCC = 0.9519 | SRCC = 0.9511 ] +INFO:root:Average loss=0.05555037879943848 +INFO:root: +[Valid][21][UTT][ MSE = 0.2348 | LCC = 0.9525 | SRCC = 0.9528 ] +INFO:root:Average loss=0.05132908689975738 +INFO:root: +[Valid][22][UTT][ MSE = 0.2447 | LCC = 0.9523 | SRCC = 0.9527 ] +INFO:root:Average loss=0.05099305027723312 +INFO:root: +[Valid][23][UTT][ MSE = 0.2012 | LCC = 0.9480 | SRCC = 0.9473 ] +INFO:root:Average loss=0.04962517005205155 +INFO:root: +[Valid][24][UTT][ MSE = 0.2260 | LCC = 0.9531 | SRCC = 0.9532 ] +INFO:root:Average loss=0.04518632647395134 +INFO:root: +[Valid][25][UTT][ MSE = 0.2126 | LCC = 0.9540 | SRCC = 0.9538 ] +INFO:root:Average loss=0.046123330801725385 +INFO:root: +[Valid][26][UTT][ MSE = 0.2140 | LCC = 0.9517 | SRCC = 0.9520 ] +INFO:root:Average loss=0.04342205914855003 +INFO:root: +[Valid][27][UTT][ MSE = 0.2248 | LCC = 0.9519 | SRCC = 0.9523 ] +INFO:root:Average loss=0.04063433700799942 +INFO:root: +[Valid][28][UTT][ MSE = 0.2374 | LCC = 0.9512 | SRCC = 0.9528 ] +INFO:root:Average loss=0.038934530943632126 +INFO:root: +[Valid][29][UTT][ MSE = 0.2121 | LCC = 0.9537 | SRCC = 0.9536 ] +INFO:root:Average loss=0.037940112262964246 +INFO:root: +[Valid][30][UTT][ MSE = 0.2059 | LCC = 0.9540 | SRCC = 0.9536 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2196 | LCC = 0.9496 | SRCC = 0.9497 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run3/config.gin b/results_tencent/w2v2_large_asr_zh/layer16/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run3/model.pt b/results_tencent/w2v2_large_asr_zh/layer16/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..176bd55ab3e098e8de2cd6771e4a4768873e1dad --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8e7222fd414ddb1072d57a0f36813d674a83d803d55778d6b661919e31b44000 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run3/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer16/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d325940dc0b7c7b45d162ffa6c3798b29b86083b --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16d52d12372de79d922225a021ca204ee6f3e41fb507706bd1644d94b0a13fdc +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run3/train.log b/results_tencent/w2v2_large_asr_zh/layer16/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..30a8ff81d99421f6925da88d6a7af828ce13a7f0 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5677301704883575 +INFO:root: +[Valid][0][UTT][ MSE = 0.3672 | LCC = 0.8839 | SRCC = 0.8814 ] +INFO:root:Average loss=0.2568521981239319 +INFO:root: +[Valid][1][UTT][ MSE = 0.3278 | LCC = 0.9196 | SRCC = 0.9176 ] +INFO:root:Average loss=0.20210639822483062 +INFO:root: +[Valid][2][UTT][ MSE = 0.2878 | LCC = 0.9267 | SRCC = 0.9260 ] +INFO:root:Average loss=0.18300412374734878 +INFO:root: +[Valid][3][UTT][ MSE = 0.3865 | LCC = 0.9242 | SRCC = 0.9237 ] +INFO:root:Average loss=0.16625583297014238 +INFO:root: +[Valid][4][UTT][ MSE = 0.2810 | LCC = 0.9385 | SRCC = 0.9382 ] +INFO:root:Average loss=0.14419450849294663 +INFO:root: +[Valid][5][UTT][ MSE = 0.2561 | LCC = 0.9427 | SRCC = 0.9420 ] +INFO:root:Average loss=0.12871039682626725 +INFO:root: +[Valid][6][UTT][ MSE = 0.3110 | LCC = 0.9448 | SRCC = 0.9449 ] +INFO:root:Average loss=0.12211888030171394 +INFO:root: +[Valid][7][UTT][ MSE = 0.3308 | LCC = 0.9447 | SRCC = 0.9467 ] +INFO:root:Average loss=0.12009826958179474 +INFO:root: +[Valid][8][UTT][ MSE = 0.2353 | LCC = 0.9475 | SRCC = 0.9466 ] +INFO:root:Average loss=0.10746979433298111 +INFO:root: +[Valid][9][UTT][ MSE = 0.2736 | LCC = 0.9483 | SRCC = 0.9479 ] +INFO:root:Average loss=0.09874257430434227 +INFO:root: +[Valid][10][UTT][ MSE = 0.2493 | LCC = 0.9469 | SRCC = 0.9470 ] +INFO:root:Average loss=0.09301790207624436 +INFO:root: +[Valid][11][UTT][ MSE = 0.2436 | LCC = 0.9451 | SRCC = 0.9454 ] +INFO:root:Average loss=0.08931910610198975 +INFO:root: +[Valid][12][UTT][ MSE = 0.2197 | LCC = 0.9511 | SRCC = 0.9509 ] +INFO:root:Average loss=0.08270152708888054 +INFO:root: +[Valid][13][UTT][ MSE = 0.2206 | LCC = 0.9500 | SRCC = 0.9494 ] +INFO:root:Average loss=0.07879424181580544 +INFO:root: +[Valid][14][UTT][ MSE = 0.2682 | LCC = 0.9397 | SRCC = 0.9397 ] +INFO:root:Average loss=0.07381060814857483 +INFO:root: +[Valid][15][UTT][ MSE = 0.2221 | LCC = 0.9511 | SRCC = 0.9504 ] +INFO:root:Average loss=0.06961893534660339 +INFO:root: +[Valid][16][UTT][ MSE = 0.1995 | LCC = 0.9518 | SRCC = 0.9512 ] +INFO:root:Average loss=0.06937167629599571 +INFO:root: +[Valid][17][UTT][ MSE = 0.2514 | LCC = 0.9490 | SRCC = 0.9485 ] +INFO:root:Average loss=0.06840432485938072 +INFO:root: +[Valid][18][UTT][ MSE = 0.2482 | LCC = 0.9488 | SRCC = 0.9480 ] +INFO:root:Average loss=0.06154615643620491 +INFO:root: +[Valid][19][UTT][ MSE = 0.2211 | LCC = 0.9516 | SRCC = 0.9512 ] +INFO:root:Average loss=0.05856920006871223 +INFO:root: +[Valid][20][UTT][ MSE = 0.2278 | LCC = 0.9525 | SRCC = 0.9517 ] +INFO:root:Average loss=0.05641262269020081 +INFO:root: +[Valid][21][UTT][ MSE = 0.2428 | LCC = 0.9508 | SRCC = 0.9519 ] +INFO:root:Average loss=0.05498359233140945 +INFO:root: +[Valid][22][UTT][ MSE = 0.2267 | LCC = 0.9507 | SRCC = 0.9508 ] +INFO:root:Average loss=0.05305710491538048 +INFO:root: +[Valid][23][UTT][ MSE = 0.2375 | LCC = 0.9523 | SRCC = 0.9523 ] +INFO:root:Average loss=0.05109260793030262 +INFO:root: +[Valid][24][UTT][ MSE = 0.2185 | LCC = 0.9529 | SRCC = 0.9523 ] +INFO:root:Average loss=0.04976301722228527 +INFO:root: +[Valid][25][UTT][ MSE = 0.2411 | LCC = 0.9487 | SRCC = 0.9479 ] +INFO:root:Average loss=0.04774068789184093 +INFO:root: +[Valid][26][UTT][ MSE = 0.2208 | LCC = 0.9522 | SRCC = 0.9518 ] +INFO:root:Average loss=0.04393444062769413 +INFO:root: +[Valid][27][UTT][ MSE = 0.2412 | LCC = 0.9498 | SRCC = 0.9506 ] +INFO:root:Average loss=0.04445560309290886 +INFO:root: +[Valid][28][UTT][ MSE = 0.2213 | LCC = 0.9476 | SRCC = 0.9469 ] +INFO:root:Average loss=0.042677387088537216 +INFO:root: +[Valid][29][UTT][ MSE = 0.2278 | LCC = 0.9475 | SRCC = 0.9478 ] +INFO:root:Average loss=0.04050610598921776 +INFO:root: +[Valid][30][UTT][ MSE = 0.2077 | LCC = 0.9519 | SRCC = 0.9513 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2375 | LCC = 0.9473 | SRCC = 0.9473 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run4/config.gin b/results_tencent/w2v2_large_asr_zh/layer16/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run4/model.pt b/results_tencent/w2v2_large_asr_zh/layer16/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..43d21aaf31f98f16d09f6936c696e1030288df7a --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e6607e7cea48e808103ef1c90b9811a63c00fa58e634af5ec764872184d270e +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run4/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer16/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a4aefd622b0f72ca52cdef590df43d5a5a0de725 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:33bf04eb456db746798e6bd25a30071c1cd58e0550b1bd534021d8fc0f622a80 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer16/run4/train.log b/results_tencent/w2v2_large_asr_zh/layer16/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..aa231d62344088cfcb19c85554c8210b9432e288 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer16/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.47799645590782164 +INFO:root: +[Valid][0][UTT][ MSE = 0.3806 | LCC = 0.8945 | SRCC = 0.8915 ] +INFO:root:Average loss=0.23539694738388062 +INFO:root: +[Valid][1][UTT][ MSE = 0.2688 | LCC = 0.9219 | SRCC = 0.9217 ] +INFO:root:Average loss=0.19653142219781874 +INFO:root: +[Valid][2][UTT][ MSE = 0.2778 | LCC = 0.9334 | SRCC = 0.9317 ] +INFO:root:Average loss=0.1702085683941841 +INFO:root: +[Valid][3][UTT][ MSE = 0.2409 | LCC = 0.9381 | SRCC = 0.9380 ] +INFO:root:Average loss=0.15103831571340562 +INFO:root: +[Valid][4][UTT][ MSE = 0.2010 | LCC = 0.9389 | SRCC = 0.9394 ] +INFO:root:Average loss=0.134738005399704 +INFO:root: +[Valid][5][UTT][ MSE = 0.2286 | LCC = 0.9432 | SRCC = 0.9423 ] +INFO:root:Average loss=0.1270338397026062 +INFO:root: +[Valid][6][UTT][ MSE = 0.2880 | LCC = 0.9462 | SRCC = 0.9463 ] +INFO:root:Average loss=0.1183849699497223 +INFO:root: +[Valid][7][UTT][ MSE = 0.2640 | LCC = 0.9467 | SRCC = 0.9480 ] +INFO:root:Average loss=0.10700648531317711 +INFO:root: +[Valid][8][UTT][ MSE = 0.2153 | LCC = 0.9513 | SRCC = 0.9505 ] +INFO:root:Average loss=0.0983788639307022 +INFO:root: +[Valid][9][UTT][ MSE = 0.1958 | LCC = 0.9507 | SRCC = 0.9506 ] +INFO:root:Average loss=0.09374068900942803 +INFO:root: +[Valid][10][UTT][ MSE = 0.1928 | LCC = 0.9504 | SRCC = 0.9507 ] +INFO:root:Average loss=0.08617385813593864 +INFO:root: +[Valid][11][UTT][ MSE = 0.2085 | LCC = 0.9523 | SRCC = 0.9519 ] +INFO:root:Average loss=0.08045380645990372 +INFO:root: +[Valid][12][UTT][ MSE = 0.2002 | LCC = 0.9541 | SRCC = 0.9531 ] +INFO:root:Average loss=0.0783419548869133 +INFO:root: +[Valid][13][UTT][ MSE = 0.1931 | LCC = 0.9523 | SRCC = 0.9518 ] +INFO:root:Average loss=0.07622395491600037 +INFO:root: +[Valid][14][UTT][ MSE = 0.2088 | LCC = 0.9447 | SRCC = 0.9460 ] +INFO:root:Average loss=0.0773960277736187 +INFO:root: +[Valid][15][UTT][ MSE = 0.2010 | LCC = 0.9532 | SRCC = 0.9527 ] +INFO:root:Average loss=0.07044993591308593 +INFO:root: +[Valid][16][UTT][ MSE = 0.2137 | LCC = 0.9544 | SRCC = 0.9533 ] +INFO:root:Average loss=0.06493438041210174 +INFO:root: +[Valid][17][UTT][ MSE = 0.2083 | LCC = 0.9517 | SRCC = 0.9513 ] +INFO:root:Average loss=0.06413402020931244 +INFO:root: +[Valid][18][UTT][ MSE = 0.2155 | LCC = 0.9510 | SRCC = 0.9504 ] +INFO:root:Average loss=0.060676126658916474 +INFO:root: +[Valid][19][UTT][ MSE = 0.2465 | LCC = 0.9512 | SRCC = 0.9508 ] +INFO:root:Average loss=0.058847860485315324 +INFO:root: +[Valid][20][UTT][ MSE = 0.1989 | LCC = 0.9528 | SRCC = 0.9522 ] +INFO:root:Average loss=0.0544984866976738 +INFO:root: +[Valid][21][UTT][ MSE = 0.1914 | LCC = 0.9533 | SRCC = 0.9525 ] +INFO:root:Average loss=0.052231346994638446 +INFO:root: +[Valid][22][UTT][ MSE = 0.1906 | LCC = 0.9526 | SRCC = 0.9516 ] +INFO:root:Average loss=0.05174759244918823 +INFO:root: +[Valid][23][UTT][ MSE = 0.2115 | LCC = 0.9484 | SRCC = 0.9487 ] +INFO:root:Average loss=0.04693572810292244 +INFO:root: +[Valid][24][UTT][ MSE = 0.1825 | LCC = 0.9518 | SRCC = 0.9510 ] +INFO:root:Average loss=0.04749105504155159 +INFO:root: +[Valid][25][UTT][ MSE = 0.2333 | LCC = 0.9530 | SRCC = 0.9527 ] +INFO:root:Average loss=0.04545638009905815 +INFO:root: +[Valid][26][UTT][ MSE = 0.1962 | LCC = 0.9534 | SRCC = 0.9522 ] +INFO:root:Average loss=0.04219263072311878 +INFO:root: +[Valid][27][UTT][ MSE = 0.1884 | LCC = 0.9505 | SRCC = 0.9495 ] +INFO:root:Average loss=0.0410209891051054 +INFO:root: +[Valid][28][UTT][ MSE = 0.2092 | LCC = 0.9532 | SRCC = 0.9529 ] +INFO:root:Average loss=0.041345334649086 +INFO:root: +[Valid][29][UTT][ MSE = 0.2086 | LCC = 0.9523 | SRCC = 0.9518 ] +INFO:root:Average loss=0.04179089519381523 +INFO:root: +[Valid][30][UTT][ MSE = 0.2123 | LCC = 0.9498 | SRCC = 0.9488 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2275 | LCC = 0.9507 | SRCC = 0.9508 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run0/config.gin b/results_tencent/w2v2_large_asr_zh/layer18/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run0/model.pt b/results_tencent/w2v2_large_asr_zh/layer18/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..40b190e6dc7bea16526fdcd9ecac8e100d8c90b9 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:099f94d9ce2f9f3ddc14ad1dd19633e3ab8d75f2d38a0986665d2730a7b79788 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run0/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer18/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b0ef27f1c9ac0ef7d6f6ea1de6638c0edbe60373 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:582f1bdfdeb4294840adb571a91129e5755e89fafab3ce0e0454f0f81f99fab3 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run0/train.log b/results_tencent/w2v2_large_asr_zh/layer18/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b8bdc6791841e856697f2e79a79545bc82ba1e42 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6028406143188476 +INFO:root: +[Valid][0][UTT][ MSE = 0.4290 | LCC = 0.8673 | SRCC = 0.8721 ] +INFO:root:Average loss=0.2742081034183502 +INFO:root: +[Valid][1][UTT][ MSE = 0.3622 | LCC = 0.9115 | SRCC = 0.9089 ] +INFO:root:Average loss=0.23044080924987792 +INFO:root: +[Valid][2][UTT][ MSE = 0.2668 | LCC = 0.9184 | SRCC = 0.9178 ] +INFO:root:Average loss=0.19718024414777754 +INFO:root: +[Valid][3][UTT][ MSE = 0.2794 | LCC = 0.9307 | SRCC = 0.9285 ] +INFO:root:Average loss=0.1725174720287323 +INFO:root: +[Valid][4][UTT][ MSE = 0.2366 | LCC = 0.9339 | SRCC = 0.9330 ] +INFO:root:Average loss=0.1625106731057167 +INFO:root: +[Valid][5][UTT][ MSE = 0.2230 | LCC = 0.9363 | SRCC = 0.9361 ] +INFO:root:Average loss=0.1445128858089447 +INFO:root: +[Valid][6][UTT][ MSE = 0.2457 | LCC = 0.9403 | SRCC = 0.9380 ] +INFO:root:Average loss=0.12900306755304336 +INFO:root: +[Valid][7][UTT][ MSE = 0.2235 | LCC = 0.9426 | SRCC = 0.9421 ] +INFO:root:Average loss=0.12007031163573265 +INFO:root: +[Valid][8][UTT][ MSE = 0.2211 | LCC = 0.9447 | SRCC = 0.9437 ] +INFO:root:Average loss=0.1151491026878357 +INFO:root: +[Valid][9][UTT][ MSE = 0.2341 | LCC = 0.9467 | SRCC = 0.9449 ] +INFO:root:Average loss=0.10576128721237182 +INFO:root: +[Valid][10][UTT][ MSE = 0.2934 | LCC = 0.9470 | SRCC = 0.9467 ] +INFO:root:Average loss=0.09576701349020005 +INFO:root: +[Valid][11][UTT][ MSE = 0.2303 | LCC = 0.9386 | SRCC = 0.9374 ] +INFO:root:Average loss=0.09459289294481278 +INFO:root: +[Valid][12][UTT][ MSE = 0.2097 | LCC = 0.9467 | SRCC = 0.9462 ] +INFO:root:Average loss=0.0847588192820549 +INFO:root: +[Valid][13][UTT][ MSE = 0.2180 | LCC = 0.9485 | SRCC = 0.9473 ] +INFO:root:Average loss=0.0782886206805706 +INFO:root: +[Valid][14][UTT][ MSE = 0.2148 | LCC = 0.9483 | SRCC = 0.9462 ] +INFO:root:Average loss=0.07737697333097458 +INFO:root: +[Valid][15][UTT][ MSE = 0.2127 | LCC = 0.9475 | SRCC = 0.9456 ] +INFO:root:Average loss=0.07257261651754379 +INFO:root: +[Valid][16][UTT][ MSE = 0.2165 | LCC = 0.9467 | SRCC = 0.9457 ] +INFO:root:Average loss=0.07044503718614578 +INFO:root: +[Valid][17][UTT][ MSE = 0.2138 | LCC = 0.9502 | SRCC = 0.9491 ] +INFO:root:Average loss=0.07203435964882374 +INFO:root: +[Valid][18][UTT][ MSE = 0.2023 | LCC = 0.9476 | SRCC = 0.9474 ] +INFO:root:Average loss=0.0626768234372139 +INFO:root: +[Valid][19][UTT][ MSE = 0.2052 | LCC = 0.9472 | SRCC = 0.9458 ] +INFO:root:Average loss=0.05935683995485306 +INFO:root: +[Valid][20][UTT][ MSE = 0.2105 | LCC = 0.9475 | SRCC = 0.9450 ] +INFO:root:Average loss=0.058704062163829805 +INFO:root: +[Valid][21][UTT][ MSE = 0.2146 | LCC = 0.9500 | SRCC = 0.9474 ] +INFO:root:Average loss=0.054803917914628984 +INFO:root: +[Valid][22][UTT][ MSE = 0.2102 | LCC = 0.9495 | SRCC = 0.9476 ] +INFO:root:Average loss=0.05555885271728039 +INFO:root: +[Valid][23][UTT][ MSE = 0.2071 | LCC = 0.9478 | SRCC = 0.9472 ] +INFO:root:Average loss=0.05092754481732845 +INFO:root: +[Valid][24][UTT][ MSE = 0.2453 | LCC = 0.9494 | SRCC = 0.9482 ] +INFO:root:Average loss=0.0481285240650177 +INFO:root: +[Valid][25][UTT][ MSE = 0.2203 | LCC = 0.9455 | SRCC = 0.9443 ] +INFO:root:Average loss=0.04890142716467381 +INFO:root: +[Valid][26][UTT][ MSE = 0.2333 | LCC = 0.9473 | SRCC = 0.9453 ] +INFO:root:Average loss=0.04818002325296402 +INFO:root: +[Valid][27][UTT][ MSE = 0.2104 | LCC = 0.9490 | SRCC = 0.9478 ] +INFO:root:Average loss=0.04252480047941208 +INFO:root: +[Valid][28][UTT][ MSE = 0.2435 | LCC = 0.9487 | SRCC = 0.9475 ] +INFO:root:Average loss=0.04265177878737449 +INFO:root: +[Valid][29][UTT][ MSE = 0.2400 | LCC = 0.9491 | SRCC = 0.9480 ] +INFO:root:Average loss=0.04345430694520473 +INFO:root: +[Valid][30][UTT][ MSE = 0.2187 | LCC = 0.9494 | SRCC = 0.9475 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2295 | LCC = 0.9451 | SRCC = 0.9456 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run1/config.gin b/results_tencent/w2v2_large_asr_zh/layer18/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run1/model.pt b/results_tencent/w2v2_large_asr_zh/layer18/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..84845cf97569caeed0cb62ffff15a33da8797287 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e66c4b21c69e26dc2ecea19f97de7c8b3b073af74fa63eb51a70624373ba3016 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run1/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer18/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..1dcf2b946a31e713623da0347c5aa42681796cab --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bf456b32efaa1e5c6a751067e6e47f9fc5348d7c11d6403dbf39d9f1cd421bc9 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run1/train.log b/results_tencent/w2v2_large_asr_zh/layer18/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..fbc72f36e4d6c5aa4e50b7a4d5c64b2fba71f0da --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5079883902072907 +INFO:root: +[Valid][0][UTT][ MSE = 0.3582 | LCC = 0.8934 | SRCC = 0.8928 ] +INFO:root:Average loss=0.2547494387626648 +INFO:root: +[Valid][1][UTT][ MSE = 0.3340 | LCC = 0.9178 | SRCC = 0.9155 ] +INFO:root:Average loss=0.2179391016960144 +INFO:root: +[Valid][2][UTT][ MSE = 0.3366 | LCC = 0.9291 | SRCC = 0.9280 ] +INFO:root:Average loss=0.18774591988325118 +INFO:root: +[Valid][3][UTT][ MSE = 0.2931 | LCC = 0.9356 | SRCC = 0.9335 ] +INFO:root:Average loss=0.16108200794458388 +INFO:root: +[Valid][4][UTT][ MSE = 0.2264 | LCC = 0.9340 | SRCC = 0.9328 ] +INFO:root:Average loss=0.14645003062486647 +INFO:root: +[Valid][5][UTT][ MSE = 0.2843 | LCC = 0.9414 | SRCC = 0.9406 ] +INFO:root:Average loss=0.13271687960624695 +INFO:root: +[Valid][6][UTT][ MSE = 0.2081 | LCC = 0.9422 | SRCC = 0.9412 ] +INFO:root:Average loss=0.12683075326681137 +INFO:root: +[Valid][7][UTT][ MSE = 0.2429 | LCC = 0.9425 | SRCC = 0.9415 ] +INFO:root:Average loss=0.1153348223567009 +INFO:root: +[Valid][8][UTT][ MSE = 0.2522 | LCC = 0.9418 | SRCC = 0.9403 ] +INFO:root:Average loss=0.10512111195921899 +INFO:root: +[Valid][9][UTT][ MSE = 0.2396 | LCC = 0.9468 | SRCC = 0.9459 ] +INFO:root:Average loss=0.10454344463348389 +INFO:root: +[Valid][10][UTT][ MSE = 0.2401 | LCC = 0.9483 | SRCC = 0.9457 ] +INFO:root:Average loss=0.09354752370715141 +INFO:root: +[Valid][11][UTT][ MSE = 0.1995 | LCC = 0.9470 | SRCC = 0.9455 ] +INFO:root:Average loss=0.08890461197495461 +INFO:root: +[Valid][12][UTT][ MSE = 0.2339 | LCC = 0.9489 | SRCC = 0.9471 ] +INFO:root:Average loss=0.07999263271689415 +INFO:root: +[Valid][13][UTT][ MSE = 0.2238 | LCC = 0.9501 | SRCC = 0.9484 ] +INFO:root:Average loss=0.07513271045684815 +INFO:root: +[Valid][14][UTT][ MSE = 0.2530 | LCC = 0.9489 | SRCC = 0.9475 ] +INFO:root:Average loss=0.07512917572259903 +INFO:root: +[Valid][15][UTT][ MSE = 0.2455 | LCC = 0.9505 | SRCC = 0.9485 ] +INFO:root:Average loss=0.07101388052105903 +INFO:root: +[Valid][16][UTT][ MSE = 0.2163 | LCC = 0.9507 | SRCC = 0.9485 ] +INFO:root:Average loss=0.0658650263249874 +INFO:root: +[Valid][17][UTT][ MSE = 0.2269 | LCC = 0.9499 | SRCC = 0.9476 ] +INFO:root:Average loss=0.06290690323710442 +INFO:root: +[Valid][18][UTT][ MSE = 0.2152 | LCC = 0.9508 | SRCC = 0.9487 ] +INFO:root:Average loss=0.062492994502186774 +INFO:root: +[Valid][19][UTT][ MSE = 0.2240 | LCC = 0.9505 | SRCC = 0.9488 ] +INFO:root:Average loss=0.05733041015267372 +INFO:root: +[Valid][20][UTT][ MSE = 0.2244 | LCC = 0.9485 | SRCC = 0.9472 ] +INFO:root:Average loss=0.05534169483184814 +INFO:root: +[Valid][21][UTT][ MSE = 0.2154 | LCC = 0.9491 | SRCC = 0.9465 ] +INFO:root:Average loss=0.05590243837237358 +INFO:root: +[Valid][22][UTT][ MSE = 0.2138 | LCC = 0.9487 | SRCC = 0.9465 ] +INFO:root:Average loss=0.051037631198763846 +INFO:root: +[Valid][23][UTT][ MSE = 0.2123 | LCC = 0.9502 | SRCC = 0.9485 ] +INFO:root:Average loss=0.05276788458228111 +INFO:root: +[Valid][24][UTT][ MSE = 0.2460 | LCC = 0.9513 | SRCC = 0.9499 ] +INFO:root:Average loss=0.052267211988568305 +INFO:root: +[Valid][25][UTT][ MSE = 0.2012 | LCC = 0.9492 | SRCC = 0.9474 ] +INFO:root:Average loss=0.044920409321784975 +INFO:root: +[Valid][26][UTT][ MSE = 0.1972 | LCC = 0.9507 | SRCC = 0.9481 ] +INFO:root:Average loss=0.045828700765967366 +INFO:root: +[Valid][27][UTT][ MSE = 0.2125 | LCC = 0.9497 | SRCC = 0.9479 ] +INFO:root:Average loss=0.04458714313805103 +INFO:root: +[Valid][28][UTT][ MSE = 0.1837 | LCC = 0.9515 | SRCC = 0.9497 ] +INFO:root:Average loss=0.0424871474057436 +INFO:root: +[Valid][29][UTT][ MSE = 0.2245 | LCC = 0.9490 | SRCC = 0.9475 ] +INFO:root:Average loss=0.04059001825749874 +INFO:root: +[Valid][30][UTT][ MSE = 0.2093 | LCC = 0.9523 | SRCC = 0.9498 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2250 | LCC = 0.9463 | SRCC = 0.9459 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run2/config.gin b/results_tencent/w2v2_large_asr_zh/layer18/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run2/model.pt b/results_tencent/w2v2_large_asr_zh/layer18/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..549bb325c348ccf0b247300ed0c74206757efbf3 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:33b3363681dd047f012270faf60dffec7a66c29425825e385471b45888751d2e +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run2/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer18/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f9755cf5da5506f660bcf91411970cd389e53a7e --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2bf9c85b04a8a85bc2e3b18329c840a539cd50c4692d96bbe1244627fa0dc5f3 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run2/train.log b/results_tencent/w2v2_large_asr_zh/layer18/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..bbc39bbe28d7fb4ad037c95b107d2f3c6e3054c1 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.48365336763858796 +INFO:root: +[Valid][0][UTT][ MSE = 0.3946 | LCC = 0.8880 | SRCC = 0.8853 ] +INFO:root:Average loss=0.2839484529495239 +INFO:root: +[Valid][1][UTT][ MSE = 0.2922 | LCC = 0.9124 | SRCC = 0.9103 ] +INFO:root:Average loss=0.2094278106689453 +INFO:root: +[Valid][2][UTT][ MSE = 0.2477 | LCC = 0.9227 | SRCC = 0.9227 ] +INFO:root:Average loss=0.18582757812738418 +INFO:root: +[Valid][3][UTT][ MSE = 0.2604 | LCC = 0.9339 | SRCC = 0.9334 ] +INFO:root:Average loss=0.1608265368938446 +INFO:root: +[Valid][4][UTT][ MSE = 0.3095 | LCC = 0.9371 | SRCC = 0.9370 ] +INFO:root:Average loss=0.14632067441940308 +INFO:root: +[Valid][5][UTT][ MSE = 0.2588 | LCC = 0.9402 | SRCC = 0.9388 ] +INFO:root:Average loss=0.1315524709224701 +INFO:root: +[Valid][6][UTT][ MSE = 0.2435 | LCC = 0.9411 | SRCC = 0.9393 ] +INFO:root:Average loss=0.12240381163358688 +INFO:root: +[Valid][7][UTT][ MSE = 0.2173 | LCC = 0.9447 | SRCC = 0.9434 ] +INFO:root:Average loss=0.11295355838537216 +INFO:root: +[Valid][8][UTT][ MSE = 0.2702 | LCC = 0.9409 | SRCC = 0.9414 ] +INFO:root:Average loss=0.10361893936991691 +INFO:root: +[Valid][9][UTT][ MSE = 0.1993 | LCC = 0.9453 | SRCC = 0.9439 ] +INFO:root:Average loss=0.0977837483882904 +INFO:root: +[Valid][10][UTT][ MSE = 0.2449 | LCC = 0.9456 | SRCC = 0.9441 ] +INFO:root:Average loss=0.08961265516281128 +INFO:root: +[Valid][11][UTT][ MSE = 0.2103 | LCC = 0.9468 | SRCC = 0.9456 ] +INFO:root:Average loss=0.08649627766013146 +INFO:root: +[Valid][12][UTT][ MSE = 0.2363 | LCC = 0.9480 | SRCC = 0.9464 ] +INFO:root:Average loss=0.08224886524677276 +INFO:root: +[Valid][13][UTT][ MSE = 0.2729 | LCC = 0.9453 | SRCC = 0.9444 ] +INFO:root:Average loss=0.07498408663272858 +INFO:root: +[Valid][14][UTT][ MSE = 0.2076 | LCC = 0.9446 | SRCC = 0.9445 ] +INFO:root:Average loss=0.07492872622609138 +INFO:root: +[Valid][15][UTT][ MSE = 0.2124 | LCC = 0.9481 | SRCC = 0.9470 ] +INFO:root:Average loss=0.06787966588139534 +INFO:root: +[Valid][16][UTT][ MSE = 0.2117 | LCC = 0.9491 | SRCC = 0.9477 ] +INFO:root:Average loss=0.06908968463540077 +INFO:root: +[Valid][17][UTT][ MSE = 0.2089 | LCC = 0.9473 | SRCC = 0.9462 ] +INFO:root:Average loss=0.06518303427100182 +INFO:root: +[Valid][18][UTT][ MSE = 0.2137 | LCC = 0.9446 | SRCC = 0.9456 ] +INFO:root:Average loss=0.062063238590955735 +INFO:root: +[Valid][19][UTT][ MSE = 0.2233 | LCC = 0.9482 | SRCC = 0.9468 ] +INFO:root:Average loss=0.05643564674258232 +INFO:root: +[Valid][20][UTT][ MSE = 0.2088 | LCC = 0.9485 | SRCC = 0.9467 ] +INFO:root:Average loss=0.05700470146536827 +INFO:root: +[Valid][21][UTT][ MSE = 0.2414 | LCC = 0.9493 | SRCC = 0.9477 ] +INFO:root:Average loss=0.053007304444909094 +INFO:root: +[Valid][22][UTT][ MSE = 0.2346 | LCC = 0.9476 | SRCC = 0.9463 ] +INFO:root:Average loss=0.050046776816248896 +INFO:root: +[Valid][23][UTT][ MSE = 0.2217 | LCC = 0.9474 | SRCC = 0.9458 ] +INFO:root:Average loss=0.04660888907313347 +INFO:root: +[Valid][24][UTT][ MSE = 0.2034 | LCC = 0.9488 | SRCC = 0.9473 ] +INFO:root:Average loss=0.04833644752204418 +INFO:root: +[Valid][25][UTT][ MSE = 0.2193 | LCC = 0.9489 | SRCC = 0.9472 ] +INFO:root:Average loss=0.04548919688165188 +INFO:root: +[Valid][26][UTT][ MSE = 0.2216 | LCC = 0.9493 | SRCC = 0.9478 ] +INFO:root:Average loss=0.046455352619290353 +INFO:root: +[Valid][27][UTT][ MSE = 0.2160 | LCC = 0.9476 | SRCC = 0.9460 ] +INFO:root:Average loss=0.04428763833642006 +INFO:root: +[Valid][28][UTT][ MSE = 0.2224 | LCC = 0.9498 | SRCC = 0.9481 ] +INFO:root:Average loss=0.04053117655217647 +INFO:root: +[Valid][29][UTT][ MSE = 0.2185 | LCC = 0.9472 | SRCC = 0.9458 ] +INFO:root:Average loss=0.040738252878189087 +INFO:root: +[Valid][30][UTT][ MSE = 0.2300 | LCC = 0.9479 | SRCC = 0.9463 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2304 | LCC = 0.9463 | SRCC = 0.9466 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run3/config.gin b/results_tencent/w2v2_large_asr_zh/layer18/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run3/model.pt b/results_tencent/w2v2_large_asr_zh/layer18/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..29378fa9208ae31de52e123068c39406a0961281 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:de8aeb901ab52a7afc3754831ba008c3d437a382c78a5c145676e8642db71506 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run3/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer18/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..8be8fcce39aafb777b7b0c69b9f694d8f6be26ca --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eb842b09dabd7f1b4f2c2498b4e2b80ec7cab5561235365fd8a1e1f132bfec15 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run3/train.log b/results_tencent/w2v2_large_asr_zh/layer18/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f7291b9db578f209a02b31c8f81ebd3b34f7cdd4 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.45311253702640536 +INFO:root: +[Valid][0][UTT][ MSE = 0.4298 | LCC = 0.8946 | SRCC = 0.8896 ] +INFO:root:Average loss=0.25667422318458555 +INFO:root: +[Valid][1][UTT][ MSE = 0.3224 | LCC = 0.9108 | SRCC = 0.9078 ] +INFO:root:Average loss=0.2246887503862381 +INFO:root: +[Valid][2][UTT][ MSE = 0.3062 | LCC = 0.9245 | SRCC = 0.9226 ] +INFO:root:Average loss=0.18944965493679047 +INFO:root: +[Valid][3][UTT][ MSE = 0.3041 | LCC = 0.9312 | SRCC = 0.9296 ] +INFO:root:Average loss=0.17184562838077544 +INFO:root: +[Valid][4][UTT][ MSE = 0.2983 | LCC = 0.9350 | SRCC = 0.9323 ] +INFO:root:Average loss=0.15403742706775667 +INFO:root: +[Valid][5][UTT][ MSE = 0.2885 | LCC = 0.9343 | SRCC = 0.9320 ] +INFO:root:Average loss=0.13592820304632186 +INFO:root: +[Valid][6][UTT][ MSE = 0.2447 | LCC = 0.9418 | SRCC = 0.9405 ] +INFO:root:Average loss=0.12840838733315468 +INFO:root: +[Valid][7][UTT][ MSE = 0.2318 | LCC = 0.9429 | SRCC = 0.9410 ] +INFO:root:Average loss=0.11946760645508767 +INFO:root: +[Valid][8][UTT][ MSE = 0.2084 | LCC = 0.9447 | SRCC = 0.9421 ] +INFO:root:Average loss=0.1047293963432312 +INFO:root: +[Valid][9][UTT][ MSE = 0.2768 | LCC = 0.9452 | SRCC = 0.9430 ] +INFO:root:Average loss=0.0988310289978981 +INFO:root: +[Valid][10][UTT][ MSE = 0.2431 | LCC = 0.9430 | SRCC = 0.9414 ] +INFO:root:Average loss=0.09651858633756638 +INFO:root: +[Valid][11][UTT][ MSE = 0.2727 | LCC = 0.9456 | SRCC = 0.9439 ] +INFO:root:Average loss=0.0863243019580841 +INFO:root: +[Valid][12][UTT][ MSE = 0.2275 | LCC = 0.9442 | SRCC = 0.9415 ] +INFO:root:Average loss=0.08461924082040786 +INFO:root: +[Valid][13][UTT][ MSE = 0.2370 | LCC = 0.9431 | SRCC = 0.9416 ] +INFO:root:Average loss=0.07940385055541992 +INFO:root: +[Valid][14][UTT][ MSE = 0.2553 | LCC = 0.9453 | SRCC = 0.9439 ] +INFO:root:Average loss=0.07821654748916626 +INFO:root: +[Valid][15][UTT][ MSE = 0.2429 | LCC = 0.9465 | SRCC = 0.9448 ] +INFO:root:Average loss=0.07274247562885285 +INFO:root: +[Valid][16][UTT][ MSE = 0.2817 | LCC = 0.9396 | SRCC = 0.9383 ] +INFO:root:Average loss=0.06969283720850944 +INFO:root: +[Valid][17][UTT][ MSE = 0.2719 | LCC = 0.9442 | SRCC = 0.9417 ] +INFO:root:Average loss=0.06972084540128708 +INFO:root: +[Valid][18][UTT][ MSE = 0.2376 | LCC = 0.9481 | SRCC = 0.9458 ] +INFO:root:Average loss=0.061033073514699936 +INFO:root: +[Valid][19][UTT][ MSE = 0.2291 | LCC = 0.9424 | SRCC = 0.9402 ] +INFO:root:Average loss=0.06125300088524818 +INFO:root: +[Valid][20][UTT][ MSE = 0.2774 | LCC = 0.9459 | SRCC = 0.9440 ] +INFO:root:Average loss=0.059976189106702806 +INFO:root: +[Valid][21][UTT][ MSE = 0.2293 | LCC = 0.9480 | SRCC = 0.9457 ] +INFO:root:Average loss=0.05980843377113342 +INFO:root: +[Valid][22][UTT][ MSE = 0.2445 | LCC = 0.9461 | SRCC = 0.9435 ] +INFO:root:Average loss=0.05141303327679634 +INFO:root: +[Valid][23][UTT][ MSE = 0.2255 | LCC = 0.9460 | SRCC = 0.9438 ] +INFO:root:Average loss=0.05052255429327488 +INFO:root: +[Valid][24][UTT][ MSE = 0.2174 | LCC = 0.9456 | SRCC = 0.9433 ] +INFO:root:Average loss=0.05284749336540699 +INFO:root: +[Valid][25][UTT][ MSE = 0.2259 | LCC = 0.9474 | SRCC = 0.9447 ] +INFO:root:Average loss=0.05210843586921692 +INFO:root: +[Valid][26][UTT][ MSE = 0.2691 | LCC = 0.9456 | SRCC = 0.9434 ] +INFO:root:Average loss=0.047080849021673206 +INFO:root: +[Valid][27][UTT][ MSE = 0.2817 | LCC = 0.9418 | SRCC = 0.9401 ] +INFO:root:Average loss=0.04739635135233402 +INFO:root: +[Valid][28][UTT][ MSE = 0.2522 | LCC = 0.9460 | SRCC = 0.9436 ] +INFO:root:Average loss=0.0447641125023365 +INFO:root: +[Valid][29][UTT][ MSE = 0.2567 | LCC = 0.9448 | SRCC = 0.9430 ] +INFO:root:Average loss=0.04173234732449055 +INFO:root: +[Valid][30][UTT][ MSE = 0.2706 | LCC = 0.9460 | SRCC = 0.9435 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2463 | LCC = 0.9460 | SRCC = 0.9465 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run4/config.gin b/results_tencent/w2v2_large_asr_zh/layer18/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run4/model.pt b/results_tencent/w2v2_large_asr_zh/layer18/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0daedfe63f779e14acf63083abd5f7720d868182 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ee35a025ad2c24bfdaf5dcf12b04a6ac5b13df85468870e99a336b4e7ed8c7f0 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run4/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer18/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..999083d26ba264dff7a0b31d9a94dbd59595c892 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e3fdb0af6296550f848558ea20c0e3b1ffac250e9d73d154adb5518422c2270 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer18/run4/train.log b/results_tencent/w2v2_large_asr_zh/layer18/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e722d4d7df3b0a71e22bba7e53b8aece08f05f77 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer18/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4721053327322006 +INFO:root: +[Valid][0][UTT][ MSE = 0.3719 | LCC = 0.8890 | SRCC = 0.8846 ] +INFO:root:Average loss=0.26043229126930234 +INFO:root: +[Valid][1][UTT][ MSE = 0.3176 | LCC = 0.9134 | SRCC = 0.9120 ] +INFO:root:Average loss=0.21430920195579528 +INFO:root: +[Valid][2][UTT][ MSE = 0.3045 | LCC = 0.9239 | SRCC = 0.9229 ] +INFO:root:Average loss=0.18607171058654784 +INFO:root: +[Valid][3][UTT][ MSE = 0.2493 | LCC = 0.9336 | SRCC = 0.9323 ] +INFO:root:Average loss=0.16393293452262878 +INFO:root: +[Valid][4][UTT][ MSE = 0.3240 | LCC = 0.9289 | SRCC = 0.9303 ] +INFO:root:Average loss=0.14848281610012054 +INFO:root: +[Valid][5][UTT][ MSE = 0.2451 | LCC = 0.9415 | SRCC = 0.9412 ] +INFO:root:Average loss=0.1334966564178467 +INFO:root: +[Valid][6][UTT][ MSE = 0.2139 | LCC = 0.9408 | SRCC = 0.9407 ] +INFO:root:Average loss=0.1278004448413849 +INFO:root: +[Valid][7][UTT][ MSE = 0.2213 | LCC = 0.9449 | SRCC = 0.9430 ] +INFO:root:Average loss=0.11455334213376045 +INFO:root: +[Valid][8][UTT][ MSE = 0.2294 | LCC = 0.9425 | SRCC = 0.9421 ] +INFO:root:Average loss=0.10402699410915375 +INFO:root: +[Valid][9][UTT][ MSE = 0.2295 | LCC = 0.9441 | SRCC = 0.9426 ] +INFO:root:Average loss=0.10058455842733383 +INFO:root: +[Valid][10][UTT][ MSE = 0.2277 | LCC = 0.9441 | SRCC = 0.9428 ] +INFO:root:Average loss=0.09467375987768173 +INFO:root: +[Valid][11][UTT][ MSE = 0.2160 | LCC = 0.9417 | SRCC = 0.9425 ] +INFO:root:Average loss=0.0874363964498043 +INFO:root: +[Valid][12][UTT][ MSE = 0.2139 | LCC = 0.9467 | SRCC = 0.9447 ] +INFO:root:Average loss=0.08607588896155358 +INFO:root: +[Valid][13][UTT][ MSE = 0.2163 | LCC = 0.9455 | SRCC = 0.9434 ] +INFO:root:Average loss=0.0798625095486641 +INFO:root: +[Valid][14][UTT][ MSE = 0.2031 | LCC = 0.9464 | SRCC = 0.9454 ] +INFO:root:Average loss=0.07292502719163894 +INFO:root: +[Valid][15][UTT][ MSE = 0.2317 | LCC = 0.9406 | SRCC = 0.9404 ] +INFO:root:Average loss=0.07547279039025306 +INFO:root: +[Valid][16][UTT][ MSE = 0.2557 | LCC = 0.9430 | SRCC = 0.9420 ] +INFO:root:Average loss=0.06774990153312684 +INFO:root: +[Valid][17][UTT][ MSE = 0.2350 | LCC = 0.9442 | SRCC = 0.9429 ] +INFO:root:Average loss=0.0638851688504219 +INFO:root: +[Valid][18][UTT][ MSE = 0.2195 | LCC = 0.9460 | SRCC = 0.9434 ] +INFO:root:Average loss=0.06282660788297653 +INFO:root: +[Valid][19][UTT][ MSE = 0.2118 | LCC = 0.9453 | SRCC = 0.9432 ] +INFO:root:Average loss=0.06026726233959198 +INFO:root: +[Valid][20][UTT][ MSE = 0.2031 | LCC = 0.9446 | SRCC = 0.9429 ] +INFO:root:Average loss=0.058043807670474054 +INFO:root: +[Valid][21][UTT][ MSE = 0.2171 | LCC = 0.9402 | SRCC = 0.9415 ] +INFO:root:Average loss=0.05321419809758663 +INFO:root: +[Valid][22][UTT][ MSE = 0.2426 | LCC = 0.9423 | SRCC = 0.9409 ] +INFO:root:Average loss=0.052031919449567796 +INFO:root: +[Valid][23][UTT][ MSE = 0.1989 | LCC = 0.9464 | SRCC = 0.9449 ] +INFO:root:Average loss=0.05105883267521858 +INFO:root: +[Valid][24][UTT][ MSE = 0.1934 | LCC = 0.9439 | SRCC = 0.9429 ] +INFO:root:Average loss=0.04801124709844589 +INFO:root: +[Valid][25][UTT][ MSE = 0.2418 | LCC = 0.9432 | SRCC = 0.9425 ] +INFO:root:Average loss=0.04725202031433582 +INFO:root: +[Valid][26][UTT][ MSE = 0.2244 | LCC = 0.9437 | SRCC = 0.9426 ] +INFO:root:Average loss=0.044648383021354675 +INFO:root: +[Valid][27][UTT][ MSE = 0.2140 | LCC = 0.9450 | SRCC = 0.9442 ] +INFO:root:Average loss=0.04285421498119831 +INFO:root: +[Valid][28][UTT][ MSE = 0.2409 | LCC = 0.9441 | SRCC = 0.9424 ] +INFO:root:Average loss=0.04028838866949081 +INFO:root: +[Valid][29][UTT][ MSE = 0.2382 | LCC = 0.9433 | SRCC = 0.9423 ] +INFO:root:Average loss=0.038553927019238474 +INFO:root: +[Valid][30][UTT][ MSE = 0.2203 | LCC = 0.9439 | SRCC = 0.9421 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2259 | LCC = 0.9437 | SRCC = 0.9419 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run0/config.gin b/results_tencent/w2v2_large_asr_zh/layer2/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run0/model.pt b/results_tencent/w2v2_large_asr_zh/layer2/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c8e9ebe1a7d87377a8e91da6ee155aca1887dbed --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eddfceebef53b08482b5d6cbd40b651011d85eb1266435e903d917f099603ed4 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run0/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer2/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f8b57a75d226e816821f1146b8cd902ad060d601 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:23c5b57984aae98c0de5e12c1456be6639565b96f99c13ee2fcce2bcf604468d +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run0/train.log b/results_tencent/w2v2_large_asr_zh/layer2/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..fb811e690f59a57ad059127828e9db85a01b523a --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4681169190406799 +INFO:root: +[Valid][0][UTT][ MSE = 0.3557 | LCC = 0.9105 | SRCC = 0.9094 ] +INFO:root:Average loss=0.23805804538726807 +INFO:root: +[Valid][1][UTT][ MSE = 0.3848 | LCC = 0.9218 | SRCC = 0.9235 ] +INFO:root:Average loss=0.22052563071250916 +INFO:root: +[Valid][2][UTT][ MSE = 0.2575 | LCC = 0.9375 | SRCC = 0.9382 ] +INFO:root:Average loss=0.17463290977478027 +INFO:root: +[Valid][3][UTT][ MSE = 0.2695 | LCC = 0.9411 | SRCC = 0.9407 ] +INFO:root:Average loss=0.15224865090847015 +INFO:root: +[Valid][4][UTT][ MSE = 0.2828 | LCC = 0.9455 | SRCC = 0.9462 ] +INFO:root:Average loss=0.1417994801402092 +INFO:root: +[Valid][5][UTT][ MSE = 0.3213 | LCC = 0.9424 | SRCC = 0.9444 ] +INFO:root:Average loss=0.13556998115777968 +INFO:root: +[Valid][6][UTT][ MSE = 0.2202 | LCC = 0.9468 | SRCC = 0.9481 ] +INFO:root:Average loss=0.12207067501544952 +INFO:root: +[Valid][7][UTT][ MSE = 0.2665 | LCC = 0.9518 | SRCC = 0.9527 ] +INFO:root:Average loss=0.11857432392239571 +INFO:root: +[Valid][8][UTT][ MSE = 0.2475 | LCC = 0.9514 | SRCC = 0.9527 ] +INFO:root:Average loss=0.11461901500821113 +INFO:root: +[Valid][9][UTT][ MSE = 0.2052 | LCC = 0.9522 | SRCC = 0.9532 ] +INFO:root:Average loss=0.10420870098471642 +INFO:root: +[Valid][10][UTT][ MSE = 0.2527 | LCC = 0.9530 | SRCC = 0.9542 ] +INFO:root:Average loss=0.09913648521900177 +INFO:root: +[Valid][11][UTT][ MSE = 0.2339 | LCC = 0.9542 | SRCC = 0.9551 ] +INFO:root:Average loss=0.09189422819018364 +INFO:root: +[Valid][12][UTT][ MSE = 0.2274 | LCC = 0.9548 | SRCC = 0.9564 ] +INFO:root:Average loss=0.08738162660598756 +INFO:root: +[Valid][13][UTT][ MSE = 0.2348 | LCC = 0.9581 | SRCC = 0.9589 ] +INFO:root:Average loss=0.08335158145427704 +INFO:root: +[Valid][14][UTT][ MSE = 0.1993 | LCC = 0.9582 | SRCC = 0.9591 ] +INFO:root:Average loss=0.08070916193723679 +INFO:root: +[Valid][15][UTT][ MSE = 0.2172 | LCC = 0.9575 | SRCC = 0.9578 ] +INFO:root:Average loss=0.07628752067685127 +INFO:root: +[Valid][16][UTT][ MSE = 0.2001 | LCC = 0.9583 | SRCC = 0.9590 ] +INFO:root:Average loss=0.07225843435525894 +INFO:root: +[Valid][17][UTT][ MSE = 0.1791 | LCC = 0.9572 | SRCC = 0.9589 ] +INFO:root:Average loss=0.07026224863529205 +INFO:root: +[Valid][18][UTT][ MSE = 0.2528 | LCC = 0.9590 | SRCC = 0.9603 ] +INFO:root:Average loss=0.07210698732733727 +INFO:root: +[Valid][19][UTT][ MSE = 0.2546 | LCC = 0.9574 | SRCC = 0.9588 ] +INFO:root:Average loss=0.06389093515276908 +INFO:root: +[Valid][20][UTT][ MSE = 0.2050 | LCC = 0.9599 | SRCC = 0.9604 ] +INFO:root:Average loss=0.061298935383558274 +INFO:root: +[Valid][21][UTT][ MSE = 0.2265 | LCC = 0.9595 | SRCC = 0.9598 ] +INFO:root:Average loss=0.05922232410311699 +INFO:root: +[Valid][22][UTT][ MSE = 0.2208 | LCC = 0.9602 | SRCC = 0.9604 ] +INFO:root:Average loss=0.05818450555205345 +INFO:root: +[Valid][23][UTT][ MSE = 0.2662 | LCC = 0.9565 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05638441574573517 +INFO:root: +[Valid][24][UTT][ MSE = 0.1950 | LCC = 0.9602 | SRCC = 0.9605 ] +INFO:root:Average loss=0.05551509363949299 +INFO:root: +[Valid][25][UTT][ MSE = 0.2111 | LCC = 0.9600 | SRCC = 0.9598 ] +INFO:root:Average loss=0.0524526614099741 +INFO:root: +[Valid][26][UTT][ MSE = 0.2186 | LCC = 0.9597 | SRCC = 0.9600 ] +INFO:root:Average loss=0.04848689216375351 +INFO:root: +[Valid][27][UTT][ MSE = 0.1939 | LCC = 0.9599 | SRCC = 0.9605 ] +INFO:root:Average loss=0.05037837316095829 +INFO:root: +[Valid][28][UTT][ MSE = 0.2002 | LCC = 0.9598 | SRCC = 0.9601 ] +INFO:root:Average loss=0.04517105294764042 +INFO:root: +[Valid][29][UTT][ MSE = 0.1761 | LCC = 0.9573 | SRCC = 0.9590 ] +INFO:root:Average loss=0.0464861676543951 +INFO:root: +[Valid][30][UTT][ MSE = 0.2049 | LCC = 0.9596 | SRCC = 0.9601 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2362 | LCC = 0.9568 | SRCC = 0.9575 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run1/config.gin b/results_tencent/w2v2_large_asr_zh/layer2/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run1/model.pt b/results_tencent/w2v2_large_asr_zh/layer2/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4054adf29ac9a3d7d56db53328f59420d45b457c --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c56efe418b691a41bda06f919f3282508a2976357a40997f978cebeacfdc88cc +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run1/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer2/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..edcd6ccb08d56384e73d463f2e3bf71cc2ed4d5a --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:66d9e2f231b9fc6b7970f15d39790f3de64541ceed33811e1f7dd19dc4baf226 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run1/train.log b/results_tencent/w2v2_large_asr_zh/layer2/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..cb0ab8635a0316db6119052f7f694455bae4684f --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5091792130470276 +INFO:root: +[Valid][0][UTT][ MSE = 0.3902 | LCC = 0.9133 | SRCC = 0.9118 ] +INFO:root:Average loss=0.23494216585159303 +INFO:root: +[Valid][1][UTT][ MSE = 0.4144 | LCC = 0.9182 | SRCC = 0.9210 ] +INFO:root:Average loss=0.20495502066612242 +INFO:root: +[Valid][2][UTT][ MSE = 0.2911 | LCC = 0.9382 | SRCC = 0.9376 ] +INFO:root:Average loss=0.16757819330692292 +INFO:root: +[Valid][3][UTT][ MSE = 0.3516 | LCC = 0.9366 | SRCC = 0.9377 ] +INFO:root:Average loss=0.14981945419311524 +INFO:root: +[Valid][4][UTT][ MSE = 0.2318 | LCC = 0.9436 | SRCC = 0.9442 ] +INFO:root:Average loss=0.1379434367120266 +INFO:root: +[Valid][5][UTT][ MSE = 0.2163 | LCC = 0.9499 | SRCC = 0.9509 ] +INFO:root:Average loss=0.13116892391443252 +INFO:root: +[Valid][6][UTT][ MSE = 0.2486 | LCC = 0.9501 | SRCC = 0.9524 ] +INFO:root:Average loss=0.12351484861969948 +INFO:root: +[Valid][7][UTT][ MSE = 0.2714 | LCC = 0.9522 | SRCC = 0.9524 ] +INFO:root:Average loss=0.10831589794158936 +INFO:root: +[Valid][8][UTT][ MSE = 0.2168 | LCC = 0.9465 | SRCC = 0.9483 ] +INFO:root:Average loss=0.107062265843153 +INFO:root: +[Valid][9][UTT][ MSE = 0.2472 | LCC = 0.9483 | SRCC = 0.9507 ] +INFO:root:Average loss=0.09900643080472946 +INFO:root: +[Valid][10][UTT][ MSE = 0.2407 | LCC = 0.9560 | SRCC = 0.9569 ] +INFO:root:Average loss=0.09448891505599022 +INFO:root: +[Valid][11][UTT][ MSE = 0.2333 | LCC = 0.9565 | SRCC = 0.9574 ] +INFO:root:Average loss=0.08869941127300263 +INFO:root: +[Valid][12][UTT][ MSE = 0.1825 | LCC = 0.9581 | SRCC = 0.9588 ] +INFO:root:Average loss=0.08268752437829971 +INFO:root: +[Valid][13][UTT][ MSE = 0.2266 | LCC = 0.9547 | SRCC = 0.9558 ] +INFO:root:Average loss=0.0839217828810215 +INFO:root: +[Valid][14][UTT][ MSE = 0.1873 | LCC = 0.9593 | SRCC = 0.9592 ] +INFO:root:Average loss=0.073330198854208 +INFO:root: +[Valid][15][UTT][ MSE = 0.2313 | LCC = 0.9589 | SRCC = 0.9588 ] +INFO:root:Average loss=0.07289407214522362 +INFO:root: +[Valid][16][UTT][ MSE = 0.2434 | LCC = 0.9574 | SRCC = 0.9572 ] +INFO:root:Average loss=0.07005614048242569 +INFO:root: +[Valid][17][UTT][ MSE = 0.2154 | LCC = 0.9608 | SRCC = 0.9612 ] +INFO:root:Average loss=0.0692681941986084 +INFO:root: +[Valid][18][UTT][ MSE = 0.2316 | LCC = 0.9582 | SRCC = 0.9588 ] +INFO:root:Average loss=0.06360407924652099 +INFO:root: +[Valid][19][UTT][ MSE = 0.1799 | LCC = 0.9612 | SRCC = 0.9612 ] +INFO:root:Average loss=0.05945955023169518 +INFO:root: +[Valid][20][UTT][ MSE = 0.1847 | LCC = 0.9613 | SRCC = 0.9613 ] +INFO:root:Average loss=0.06058333596587181 +INFO:root: +[Valid][21][UTT][ MSE = 0.1864 | LCC = 0.9612 | SRCC = 0.9616 ] +INFO:root:Average loss=0.05906153392791748 +INFO:root: +[Valid][22][UTT][ MSE = 0.1978 | LCC = 0.9591 | SRCC = 0.9592 ] +INFO:root:Average loss=0.0547967726290226 +INFO:root: +[Valid][23][UTT][ MSE = 0.1684 | LCC = 0.9606 | SRCC = 0.9606 ] +INFO:root:Average loss=0.05627043282985687 +INFO:root: +[Valid][24][UTT][ MSE = 0.1949 | LCC = 0.9601 | SRCC = 0.9597 ] +INFO:root:Average loss=0.05244363410770893 +INFO:root: +[Valid][25][UTT][ MSE = 0.1925 | LCC = 0.9604 | SRCC = 0.9603 ] +INFO:root:Average loss=0.05219132420420647 +INFO:root: +[Valid][26][UTT][ MSE = 0.2072 | LCC = 0.9610 | SRCC = 0.9612 ] +INFO:root:Average loss=0.04961608174443245 +INFO:root: +[Valid][27][UTT][ MSE = 0.2219 | LCC = 0.9572 | SRCC = 0.9596 ] +INFO:root:Average loss=0.04662946012616158 +INFO:root: +[Valid][28][UTT][ MSE = 0.1732 | LCC = 0.9614 | SRCC = 0.9610 ] +INFO:root:Average loss=0.043468851298093795 +INFO:root: +[Valid][29][UTT][ MSE = 0.1779 | LCC = 0.9584 | SRCC = 0.9578 ] +INFO:root:Average loss=0.046316970199346545 +INFO:root: +[Valid][30][UTT][ MSE = 0.1762 | LCC = 0.9615 | SRCC = 0.9611 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1831 | LCC = 0.9569 | SRCC = 0.9582 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run2/config.gin b/results_tencent/w2v2_large_asr_zh/layer2/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run2/model.pt b/results_tencent/w2v2_large_asr_zh/layer2/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5bf981e9f4e1482ad43a2f6c60ec098bb187441f --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6aa17243fa7821f3e874652fba8729a753fe7efe1b629cda00b48cfa91f35f02 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run2/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer2/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..1d6e1c8b2ea6a22a3aa28bef15a3b5e048e5d3e8 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0d99f3df8d21870c3beb494b91e82d0f7e98a9de6bfa893c4c038a2740f6130a +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run2/train.log b/results_tencent/w2v2_large_asr_zh/layer2/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..71c215ca382bc59141f667196506a39fecbd5586 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5316642671823502 +INFO:root: +[Valid][0][UTT][ MSE = 0.3750 | LCC = 0.9111 | SRCC = 0.9100 ] +INFO:root:Average loss=0.23696059274673462 +INFO:root: +[Valid][1][UTT][ MSE = 0.3533 | LCC = 0.9270 | SRCC = 0.9294 ] +INFO:root:Average loss=0.20994119316339493 +INFO:root: +[Valid][2][UTT][ MSE = 0.2977 | LCC = 0.9403 | SRCC = 0.9407 ] +INFO:root:Average loss=0.1712069076895714 +INFO:root: +[Valid][3][UTT][ MSE = 0.2626 | LCC = 0.9439 | SRCC = 0.9443 ] +INFO:root:Average loss=0.15409370106458664 +INFO:root: +[Valid][4][UTT][ MSE = 0.2834 | LCC = 0.9464 | SRCC = 0.9479 ] +INFO:root:Average loss=0.14908762896060943 +INFO:root: +[Valid][5][UTT][ MSE = 0.2764 | LCC = 0.9373 | SRCC = 0.9404 ] +INFO:root:Average loss=0.12867475429177283 +INFO:root: +[Valid][6][UTT][ MSE = 0.2099 | LCC = 0.9487 | SRCC = 0.9518 ] +INFO:root:Average loss=0.12042511069774628 +INFO:root: +[Valid][7][UTT][ MSE = 0.2376 | LCC = 0.9550 | SRCC = 0.9558 ] +INFO:root:Average loss=0.10880176132917405 +INFO:root: +[Valid][8][UTT][ MSE = 0.2152 | LCC = 0.9550 | SRCC = 0.9557 ] +INFO:root:Average loss=0.10282927626371384 +INFO:root: +[Valid][9][UTT][ MSE = 0.2155 | LCC = 0.9557 | SRCC = 0.9564 ] +INFO:root:Average loss=0.10123639327287674 +INFO:root: +[Valid][10][UTT][ MSE = 0.2060 | LCC = 0.9564 | SRCC = 0.9574 ] +INFO:root:Average loss=0.09272337546944619 +INFO:root: +[Valid][11][UTT][ MSE = 0.1913 | LCC = 0.9553 | SRCC = 0.9564 ] +INFO:root:Average loss=0.09036252215504646 +INFO:root: +[Valid][12][UTT][ MSE = 0.2324 | LCC = 0.9575 | SRCC = 0.9579 ] +INFO:root:Average loss=0.08488731247186661 +INFO:root: +[Valid][13][UTT][ MSE = 0.2262 | LCC = 0.9578 | SRCC = 0.9580 ] +INFO:root:Average loss=0.08216285100579261 +INFO:root: +[Valid][14][UTT][ MSE = 0.2203 | LCC = 0.9581 | SRCC = 0.9583 ] +INFO:root:Average loss=0.08217434194684028 +INFO:root: +[Valid][15][UTT][ MSE = 0.2342 | LCC = 0.9555 | SRCC = 0.9564 ] +INFO:root:Average loss=0.07231089282035828 +INFO:root: +[Valid][16][UTT][ MSE = 0.2291 | LCC = 0.9587 | SRCC = 0.9587 ] +INFO:root:Average loss=0.06860787743330002 +INFO:root: +[Valid][17][UTT][ MSE = 0.2226 | LCC = 0.9581 | SRCC = 0.9582 ] +INFO:root:Average loss=0.06847847720980645 +INFO:root: +[Valid][18][UTT][ MSE = 0.2011 | LCC = 0.9585 | SRCC = 0.9591 ] +INFO:root:Average loss=0.07042139810323715 +INFO:root: +[Valid][19][UTT][ MSE = 0.2035 | LCC = 0.9588 | SRCC = 0.9589 ] +INFO:root:Average loss=0.06599068033695221 +INFO:root: +[Valid][20][UTT][ MSE = 0.2010 | LCC = 0.9590 | SRCC = 0.9596 ] +INFO:root:Average loss=0.0638447357416153 +INFO:root: +[Valid][21][UTT][ MSE = 0.1932 | LCC = 0.9587 | SRCC = 0.9587 ] +INFO:root:Average loss=0.05942314603924751 +INFO:root: +[Valid][22][UTT][ MSE = 0.1987 | LCC = 0.9599 | SRCC = 0.9603 ] +INFO:root:Average loss=0.05564791642129421 +INFO:root: +[Valid][23][UTT][ MSE = 0.1854 | LCC = 0.9604 | SRCC = 0.9604 ] +INFO:root:Average loss=0.05388548256456852 +INFO:root: +[Valid][24][UTT][ MSE = 0.2002 | LCC = 0.9607 | SRCC = 0.9609 ] +INFO:root:Average loss=0.05155213335156441 +INFO:root: +[Valid][25][UTT][ MSE = 0.2103 | LCC = 0.9595 | SRCC = 0.9595 ] +INFO:root:Average loss=0.050477638781070706 +INFO:root: +[Valid][26][UTT][ MSE = 0.2186 | LCC = 0.9599 | SRCC = 0.9602 ] +INFO:root:Average loss=0.04723711064457893 +INFO:root: +[Valid][27][UTT][ MSE = 0.2048 | LCC = 0.9615 | SRCC = 0.9610 ] +INFO:root:Average loss=0.04427620892226696 +INFO:root: +[Valid][28][UTT][ MSE = 0.2039 | LCC = 0.9616 | SRCC = 0.9616 ] +INFO:root:Average loss=0.044257597997784615 +INFO:root: +[Valid][29][UTT][ MSE = 0.1917 | LCC = 0.9588 | SRCC = 0.9596 ] +INFO:root:Average loss=0.04306997548043728 +INFO:root: +[Valid][30][UTT][ MSE = 0.1868 | LCC = 0.9618 | SRCC = 0.9615 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1945 | LCC = 0.9579 | SRCC = 0.9590 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run3/config.gin b/results_tencent/w2v2_large_asr_zh/layer2/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run3/model.pt b/results_tencent/w2v2_large_asr_zh/layer2/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c345e0227e753997ed21a8c2341679f293a92fda --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aae26fcf465fe3a4ff2d2b5fc0a310d8b4889b017e3736fda7fb17d2460f5e34 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run3/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer2/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..414454c6f55eedb79e93a78a29fa2799d9aaa21a --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b5078cd10593506effc07d6ef40a8da78fbc76b0a775ebb18712b80f1b2021c +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run3/train.log b/results_tencent/w2v2_large_asr_zh/layer2/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f42f656da905022ab2fbc7cac63044762550f3cd --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.48801482081413267 +INFO:root: +[Valid][0][UTT][ MSE = 0.3252 | LCC = 0.9067 | SRCC = 0.9045 ] +INFO:root:Average loss=0.24319308078289031 +INFO:root: +[Valid][1][UTT][ MSE = 0.2786 | LCC = 0.9275 | SRCC = 0.9265 ] +INFO:root:Average loss=0.20536955034732818 +INFO:root: +[Valid][2][UTT][ MSE = 0.2836 | LCC = 0.9376 | SRCC = 0.9379 ] +INFO:root:Average loss=0.18084545361995696 +INFO:root: +[Valid][3][UTT][ MSE = 0.2406 | LCC = 0.9392 | SRCC = 0.9391 ] +INFO:root:Average loss=0.1704311438202858 +INFO:root: +[Valid][4][UTT][ MSE = 0.2556 | LCC = 0.9373 | SRCC = 0.9378 ] +INFO:root:Average loss=0.15240766280889512 +INFO:root: +[Valid][5][UTT][ MSE = 0.2571 | LCC = 0.9468 | SRCC = 0.9472 ] +INFO:root:Average loss=0.13598234552145005 +INFO:root: +[Valid][6][UTT][ MSE = 0.2452 | LCC = 0.9455 | SRCC = 0.9473 ] +INFO:root:Average loss=0.12696826487779617 +INFO:root: +[Valid][7][UTT][ MSE = 0.1979 | LCC = 0.9465 | SRCC = 0.9485 ] +INFO:root:Average loss=0.11963838285207748 +INFO:root: +[Valid][8][UTT][ MSE = 0.2314 | LCC = 0.9494 | SRCC = 0.9517 ] +INFO:root:Average loss=0.11196453070640563 +INFO:root: +[Valid][9][UTT][ MSE = 0.2042 | LCC = 0.9527 | SRCC = 0.9539 ] +INFO:root:Average loss=0.10611038848757744 +INFO:root: +[Valid][10][UTT][ MSE = 0.1902 | LCC = 0.9524 | SRCC = 0.9531 ] +INFO:root:Average loss=0.10271013617515563 +INFO:root: +[Valid][11][UTT][ MSE = 0.1966 | LCC = 0.9541 | SRCC = 0.9552 ] +INFO:root:Average loss=0.09459375041723252 +INFO:root: +[Valid][12][UTT][ MSE = 0.1888 | LCC = 0.9536 | SRCC = 0.9552 ] +INFO:root:Average loss=0.09303631070256234 +INFO:root: +[Valid][13][UTT][ MSE = 0.2098 | LCC = 0.9557 | SRCC = 0.9561 ] +INFO:root:Average loss=0.08569142320752143 +INFO:root: +[Valid][14][UTT][ MSE = 0.1972 | LCC = 0.9566 | SRCC = 0.9570 ] +INFO:root:Average loss=0.08877525654435157 +INFO:root: +[Valid][15][UTT][ MSE = 0.1937 | LCC = 0.9539 | SRCC = 0.9549 ] +INFO:root:Average loss=0.0868169517815113 +INFO:root: +[Valid][16][UTT][ MSE = 0.1821 | LCC = 0.9551 | SRCC = 0.9561 ] +INFO:root:Average loss=0.08143976211547851 +INFO:root: +[Valid][17][UTT][ MSE = 0.1844 | LCC = 0.9561 | SRCC = 0.9571 ] +INFO:root:Average loss=0.07181838700175286 +INFO:root: +[Valid][18][UTT][ MSE = 0.2028 | LCC = 0.9567 | SRCC = 0.9570 ] +INFO:root:Average loss=0.06792970246076584 +INFO:root: +[Valid][19][UTT][ MSE = 0.2005 | LCC = 0.9565 | SRCC = 0.9569 ] +INFO:root:Average loss=0.06752907329797744 +INFO:root: +[Valid][20][UTT][ MSE = 0.1830 | LCC = 0.9562 | SRCC = 0.9566 ] +INFO:root:Average loss=0.06491609182953835 +INFO:root: +[Valid][21][UTT][ MSE = 0.1965 | LCC = 0.9569 | SRCC = 0.9569 ] +INFO:root:Average loss=0.06364136278629302 +INFO:root: +[Valid][22][UTT][ MSE = 0.1932 | LCC = 0.9543 | SRCC = 0.9557 ] +INFO:root:Average loss=0.06015642800927162 +INFO:root: +[Valid][23][UTT][ MSE = 0.1936 | LCC = 0.9566 | SRCC = 0.9583 ] +INFO:root:Average loss=0.05538818867504597 +INFO:root: +[Valid][24][UTT][ MSE = 0.2101 | LCC = 0.9573 | SRCC = 0.9575 ] +INFO:root:Average loss=0.056319781452417374 +INFO:root: +[Valid][25][UTT][ MSE = 0.2046 | LCC = 0.9571 | SRCC = 0.9573 ] +INFO:root:Average loss=0.05343953104317188 +INFO:root: +[Valid][26][UTT][ MSE = 0.2103 | LCC = 0.9599 | SRCC = 0.9605 ] +INFO:root:Average loss=0.052108921736478804 +INFO:root: +[Valid][27][UTT][ MSE = 0.2136 | LCC = 0.9563 | SRCC = 0.9565 ] +INFO:root:Average loss=0.049813504457473755 +INFO:root: +[Valid][28][UTT][ MSE = 0.1960 | LCC = 0.9589 | SRCC = 0.9591 ] +INFO:root:Average loss=0.051273649543523786 +INFO:root: +[Valid][29][UTT][ MSE = 0.2019 | LCC = 0.9581 | SRCC = 0.9581 ] +INFO:root:Average loss=0.04778134101629257 +INFO:root: +[Valid][30][UTT][ MSE = 0.1847 | LCC = 0.9573 | SRCC = 0.9580 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2154 | LCC = 0.9580 | SRCC = 0.9585 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run4/config.gin b/results_tencent/w2v2_large_asr_zh/layer2/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run4/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer2/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..109609c58fd63cd52a178421fde6fc4367689d45 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8e9b8e27294723d365716c8e24a67aed0109e08d28299fb2b8d33fafa3baff48 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer2/run4/train.log b/results_tencent/w2v2_large_asr_zh/layer2/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2c84b096cec0a44c20c08fcf2cf977024e8c8f25 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer2/run4/train.log @@ -0,0 +1,62 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.45539915788173674 +INFO:root: +[Valid][0][UTT][ MSE = 0.3346 | LCC = 0.9122 | SRCC = 0.9114 ] +INFO:root:Average loss=0.2477571499347687 +INFO:root: +[Valid][1][UTT][ MSE = 0.3407 | LCC = 0.9183 | SRCC = 0.9199 ] +INFO:root:Average loss=0.21767711496353148 +INFO:root: +[Valid][2][UTT][ MSE = 0.2719 | LCC = 0.9287 | SRCC = 0.9282 ] +INFO:root:Average loss=0.18880634987354278 +INFO:root: +[Valid][3][UTT][ MSE = 0.3418 | LCC = 0.9347 | SRCC = 0.9345 ] +INFO:root:Average loss=0.16688509142398833 +INFO:root: +[Valid][4][UTT][ MSE = 0.2201 | LCC = 0.9427 | SRCC = 0.9438 ] +INFO:root:Average loss=0.1473585286140442 +INFO:root: +[Valid][5][UTT][ MSE = 0.2635 | LCC = 0.9448 | SRCC = 0.9458 ] +INFO:root:Average loss=0.13636645847558976 +INFO:root: +[Valid][6][UTT][ MSE = 0.2598 | LCC = 0.9465 | SRCC = 0.9482 ] +INFO:root:Average loss=0.13347371464967728 +INFO:root: +[Valid][7][UTT][ MSE = 0.2680 | LCC = 0.9506 | SRCC = 0.9508 ] +INFO:root:Average loss=0.11265562748908997 +INFO:root: +[Valid][8][UTT][ MSE = 0.2459 | LCC = 0.9516 | SRCC = 0.9527 ] +INFO:root:Average loss=0.11776879334449768 +INFO:root: +[Valid][9][UTT][ MSE = 0.2428 | LCC = 0.9520 | SRCC = 0.9531 ] +INFO:root:Average loss=0.10471377128362655 +INFO:root: +[Valid][10][UTT][ MSE = 0.2406 | LCC = 0.9522 | SRCC = 0.9533 ] +INFO:root:Average loss=0.10232694831490517 +INFO:root: +[Valid][11][UTT][ MSE = 0.2410 | LCC = 0.9544 | SRCC = 0.9545 ] +INFO:root:Average loss=0.09234482216835022 +INFO:root: +[Valid][12][UTT][ MSE = 0.2222 | LCC = 0.9565 | SRCC = 0.9568 ] +INFO:root:Average loss=0.08687102243304252 +INFO:root: +[Valid][13][UTT][ MSE = 0.2419 | LCC = 0.9555 | SRCC = 0.9561 ] +INFO:root:Average loss=0.08811984416842461 +INFO:root: +[Valid][14][UTT][ MSE = 0.2047 | LCC = 0.9561 | SRCC = 0.9568 ] +INFO:root:Average loss=0.07727862852811813 +INFO:root: +[Valid][15][UTT][ MSE = 0.2365 | LCC = 0.9563 | SRCC = 0.9567 ] +INFO:root:Average loss=0.0744583768248558 +INFO:root: +[Valid][16][UTT][ MSE = 0.1967 | LCC = 0.9559 | SRCC = 0.9570 ] +INFO:root:Average loss=0.07146639177203179 +INFO:root: +[Valid][17][UTT][ MSE = 0.2056 | LCC = 0.9572 | SRCC = 0.9578 ] +INFO:root:Average loss=0.07759963351488114 +INFO:root: +[Valid][18][UTT][ MSE = 0.2013 | LCC = 0.9585 | SRCC = 0.9587 ] +INFO:root:Average loss=0.06717864948511124 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run0/config.gin b/results_tencent/w2v2_large_asr_zh/layer20/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run0/model.pt b/results_tencent/w2v2_large_asr_zh/layer20/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ff2d4ab0d318e7ac71bcb9571e69ed30af082d91 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6a619d4ab9f2900ca344859d44d3883a2aebdf2d52835a924e6601ed4c4eae84 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run0/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer20/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ccb489421b6ad39af45174b064a6f29a259546fd --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1daae71675a02588a2214417d3450342803ca51ada694b509f9beb71d04a0e64 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run0/train.log b/results_tencent/w2v2_large_asr_zh/layer20/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..4074f5f4f08be87b1f1528ea446a5932ab5414fa --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44365488076210025 +INFO:root: +[Valid][0][UTT][ MSE = 0.4298 | LCC = 0.8891 | SRCC = 0.8846 ] +INFO:root:Average loss=0.30146607959270477 +INFO:root: +[Valid][1][UTT][ MSE = 0.4615 | LCC = 0.8981 | SRCC = 0.8959 ] +INFO:root:Average loss=0.2568985047340393 +INFO:root: +[Valid][2][UTT][ MSE = 0.3460 | LCC = 0.9117 | SRCC = 0.9099 ] +INFO:root:Average loss=0.22798057997226714 +INFO:root: +[Valid][3][UTT][ MSE = 0.3436 | LCC = 0.9141 | SRCC = 0.9128 ] +INFO:root:Average loss=0.20664231836795807 +INFO:root: +[Valid][4][UTT][ MSE = 0.3003 | LCC = 0.9268 | SRCC = 0.9257 ] +INFO:root:Average loss=0.19034575629234313 +INFO:root: +[Valid][5][UTT][ MSE = 0.3673 | LCC = 0.9266 | SRCC = 0.9274 ] +INFO:root:Average loss=0.17908014959096907 +INFO:root: +[Valid][6][UTT][ MSE = 0.3054 | LCC = 0.9298 | SRCC = 0.9279 ] +INFO:root:Average loss=0.15793970853090286 +INFO:root: +[Valid][7][UTT][ MSE = 0.3217 | LCC = 0.9343 | SRCC = 0.9338 ] +INFO:root:Average loss=0.14985852640867234 +INFO:root: +[Valid][8][UTT][ MSE = 0.2601 | LCC = 0.9341 | SRCC = 0.9325 ] +INFO:root:Average loss=0.14008123821020127 +INFO:root: +[Valid][9][UTT][ MSE = 0.2945 | LCC = 0.9334 | SRCC = 0.9340 ] +INFO:root:Average loss=0.12902878123521805 +INFO:root: +[Valid][10][UTT][ MSE = 0.2788 | LCC = 0.9385 | SRCC = 0.9371 ] +INFO:root:Average loss=0.1300810508131981 +INFO:root: +[Valid][11][UTT][ MSE = 0.2841 | LCC = 0.9396 | SRCC = 0.9379 ] +INFO:root:Average loss=0.11922032761573792 +INFO:root: +[Valid][12][UTT][ MSE = 0.2489 | LCC = 0.9393 | SRCC = 0.9376 ] +INFO:root:Average loss=0.10920166540145874 +INFO:root: +[Valid][13][UTT][ MSE = 0.2666 | LCC = 0.9393 | SRCC = 0.9380 ] +INFO:root:Average loss=0.10210755050182342 +INFO:root: +[Valid][14][UTT][ MSE = 0.2560 | LCC = 0.9403 | SRCC = 0.9404 ] +INFO:root:Average loss=0.10394879934191704 +INFO:root: +[Valid][15][UTT][ MSE = 0.2736 | LCC = 0.9408 | SRCC = 0.9411 ] +INFO:root:Average loss=0.09526275074481964 +INFO:root: +[Valid][16][UTT][ MSE = 0.2402 | LCC = 0.9405 | SRCC = 0.9385 ] +INFO:root:Average loss=0.09178234401345253 +INFO:root: +[Valid][17][UTT][ MSE = 0.2513 | LCC = 0.9409 | SRCC = 0.9398 ] +INFO:root:Average loss=0.09068240812420846 +INFO:root: +[Valid][18][UTT][ MSE = 0.2838 | LCC = 0.9375 | SRCC = 0.9363 ] +INFO:root:Average loss=0.08487303793430329 +INFO:root: +[Valid][19][UTT][ MSE = 0.2556 | LCC = 0.9433 | SRCC = 0.9421 ] +INFO:root:Average loss=0.07767895305156708 +INFO:root: +[Valid][20][UTT][ MSE = 0.2532 | LCC = 0.9419 | SRCC = 0.9408 ] +INFO:root:Average loss=0.0809382204413414 +INFO:root: +[Valid][21][UTT][ MSE = 0.2772 | LCC = 0.9420 | SRCC = 0.9404 ] +INFO:root:Average loss=0.07247763586044312 +INFO:root: +[Valid][22][UTT][ MSE = 0.2683 | LCC = 0.9353 | SRCC = 0.9340 ] +INFO:root:Average loss=0.06634619149565697 +INFO:root: +[Valid][23][UTT][ MSE = 0.2622 | LCC = 0.9408 | SRCC = 0.9399 ] +INFO:root:Average loss=0.06466134411096573 +INFO:root: +[Valid][24][UTT][ MSE = 0.2522 | LCC = 0.9386 | SRCC = 0.9380 ] +INFO:root:Average loss=0.06460033002495766 +INFO:root: +[Valid][25][UTT][ MSE = 0.2688 | LCC = 0.9428 | SRCC = 0.9415 ] +INFO:root:Average loss=0.05962935376167297 +INFO:root: +[Valid][26][UTT][ MSE = 0.2510 | LCC = 0.9421 | SRCC = 0.9409 ] +INFO:root:Average loss=0.05881214872002602 +INFO:root: +[Valid][27][UTT][ MSE = 0.2812 | LCC = 0.9424 | SRCC = 0.9420 ] +INFO:root:Average loss=0.05704365959763527 +INFO:root: +[Valid][28][UTT][ MSE = 0.2626 | LCC = 0.9411 | SRCC = 0.9413 ] +INFO:root:Average loss=0.05820665538311005 +INFO:root: +[Valid][29][UTT][ MSE = 0.2617 | LCC = 0.9427 | SRCC = 0.9415 ] +INFO:root:Average loss=0.052883592277765276 +INFO:root: +[Valid][30][UTT][ MSE = 0.2770 | LCC = 0.9432 | SRCC = 0.9424 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2776 | LCC = 0.9358 | SRCC = 0.9355 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run1/config.gin b/results_tencent/w2v2_large_asr_zh/layer20/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run1/model.pt b/results_tencent/w2v2_large_asr_zh/layer20/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1c358ce3a09c4ffd3b0e7fb101a6adf60c6abf6b --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:02e6296587d40e3ccb595a00fdad12094480024da7bdf259d4ad1a6ad5a61ef3 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run1/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer20/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f5f8b8b361a2f82fd1acfd2f1ebd617181062a8a --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ea4508174a6e69c66868edc4279c6a7234844a7cc6c5f94f7d61a5aaa827dc8b +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run1/train.log b/results_tencent/w2v2_large_asr_zh/layer20/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d1f8bc48dba4399b113636ec58a795ab9b24f731 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.43554911482334135 +INFO:root: +[Valid][0][UTT][ MSE = 0.4236 | LCC = 0.8865 | SRCC = 0.8847 ] +INFO:root:Average loss=0.29209227120876313 +INFO:root: +[Valid][1][UTT][ MSE = 0.4411 | LCC = 0.9010 | SRCC = 0.8989 ] +INFO:root:Average loss=0.26327430617809294 +INFO:root: +[Valid][2][UTT][ MSE = 0.3477 | LCC = 0.9119 | SRCC = 0.9085 ] +INFO:root:Average loss=0.23602463126182557 +INFO:root: +[Valid][3][UTT][ MSE = 0.3557 | LCC = 0.9187 | SRCC = 0.9166 ] +INFO:root:Average loss=0.21326426947116853 +INFO:root: +[Valid][4][UTT][ MSE = 0.2972 | LCC = 0.9191 | SRCC = 0.9189 ] +INFO:root:Average loss=0.1920730465054512 +INFO:root: +[Valid][5][UTT][ MSE = 0.2960 | LCC = 0.9255 | SRCC = 0.9237 ] +INFO:root:Average loss=0.17126756650209427 +INFO:root: +[Valid][6][UTT][ MSE = 0.3522 | LCC = 0.9226 | SRCC = 0.9217 ] +INFO:root:Average loss=0.15833556765317916 +INFO:root: +[Valid][7][UTT][ MSE = 0.2586 | LCC = 0.9325 | SRCC = 0.9295 ] +INFO:root:Average loss=0.1463786120414734 +INFO:root: +[Valid][8][UTT][ MSE = 0.3259 | LCC = 0.9333 | SRCC = 0.9304 ] +INFO:root:Average loss=0.1391373751759529 +INFO:root: +[Valid][9][UTT][ MSE = 0.3456 | LCC = 0.9256 | SRCC = 0.9268 ] +INFO:root:Average loss=0.12718455404043197 +INFO:root: +[Valid][10][UTT][ MSE = 0.2538 | LCC = 0.9342 | SRCC = 0.9324 ] +INFO:root:Average loss=0.11863614851236344 +INFO:root: +[Valid][11][UTT][ MSE = 0.3400 | LCC = 0.9347 | SRCC = 0.9340 ] +INFO:root:Average loss=0.11306249713897705 +INFO:root: +[Valid][12][UTT][ MSE = 0.2714 | LCC = 0.9328 | SRCC = 0.9324 ] +INFO:root:Average loss=0.10567360728979111 +INFO:root: +[Valid][13][UTT][ MSE = 0.2612 | LCC = 0.9398 | SRCC = 0.9382 ] +INFO:root:Average loss=0.09804627841711044 +INFO:root: +[Valid][14][UTT][ MSE = 0.2517 | LCC = 0.9388 | SRCC = 0.9361 ] +INFO:root:Average loss=0.09448482716083527 +INFO:root: +[Valid][15][UTT][ MSE = 0.2652 | LCC = 0.9346 | SRCC = 0.9318 ] +INFO:root:Average loss=0.09005463635921478 +INFO:root: +[Valid][16][UTT][ MSE = 0.3244 | LCC = 0.9365 | SRCC = 0.9359 ] +INFO:root:Average loss=0.09099337765574456 +INFO:root: +[Valid][17][UTT][ MSE = 0.2828 | LCC = 0.9306 | SRCC = 0.9297 ] +INFO:root:Average loss=0.08395336389541626 +INFO:root: +[Valid][18][UTT][ MSE = 0.2624 | LCC = 0.9380 | SRCC = 0.9364 ] +INFO:root:Average loss=0.07868605968356132 +INFO:root: +[Valid][19][UTT][ MSE = 0.2733 | LCC = 0.9346 | SRCC = 0.9332 ] +INFO:root:Average loss=0.07638608705997467 +INFO:root: +[Valid][20][UTT][ MSE = 0.2657 | LCC = 0.9387 | SRCC = 0.9364 ] +INFO:root:Average loss=0.07094721123576164 +INFO:root: +[Valid][21][UTT][ MSE = 0.2772 | LCC = 0.9406 | SRCC = 0.9377 ] +INFO:root:Average loss=0.06845826789736748 +INFO:root: +[Valid][22][UTT][ MSE = 0.2614 | LCC = 0.9415 | SRCC = 0.9380 ] +INFO:root:Average loss=0.0700337139070034 +INFO:root: +[Valid][23][UTT][ MSE = 0.2661 | LCC = 0.9386 | SRCC = 0.9379 ] +INFO:root:Average loss=0.06658325725793839 +INFO:root: +[Valid][24][UTT][ MSE = 0.2691 | LCC = 0.9404 | SRCC = 0.9380 ] +INFO:root:Average loss=0.06417040196061134 +INFO:root: +[Valid][25][UTT][ MSE = 0.2733 | LCC = 0.9390 | SRCC = 0.9374 ] +INFO:root:Average loss=0.056378580898046496 +INFO:root: +[Valid][26][UTT][ MSE = 0.3032 | LCC = 0.9383 | SRCC = 0.9358 ] +INFO:root:Average loss=0.056257043957710265 +INFO:root: +[Valid][27][UTT][ MSE = 0.2742 | LCC = 0.9388 | SRCC = 0.9363 ] +INFO:root:Average loss=0.05365578165650368 +INFO:root: +[Valid][28][UTT][ MSE = 0.2903 | LCC = 0.9406 | SRCC = 0.9384 ] +INFO:root:Average loss=0.05219737106561661 +INFO:root: +[Valid][29][UTT][ MSE = 0.3031 | LCC = 0.9364 | SRCC = 0.9342 ] +INFO:root:Average loss=0.049792263478040694 +INFO:root: +[Valid][30][UTT][ MSE = 0.2988 | LCC = 0.9404 | SRCC = 0.9391 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2821 | LCC = 0.9356 | SRCC = 0.9331 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run2/config.gin b/results_tencent/w2v2_large_asr_zh/layer20/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run2/model.pt b/results_tencent/w2v2_large_asr_zh/layer20/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..660d320da8ea05b6d39da4bf83baa7160d574a3e --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d16081a1c3fa2c531ec33b9d5be1583d8fbef9f597ffb9aae89c27e8daa5ebcb +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run2/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer20/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a8afe28d2641707926840e30c693b644261f3392 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0b92a42ebffea485cb43094a6f9fb9a5f8f9914621a136f1f804d28550c68f57 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run2/train.log b/results_tencent/w2v2_large_asr_zh/layer20/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..88afdc8bcbd680a811020e6466bc4ac4914a7713 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44408620238304136 +INFO:root: +[Valid][0][UTT][ MSE = 0.3761 | LCC = 0.8885 | SRCC = 0.8842 ] +INFO:root:Average loss=0.281201375246048 +INFO:root: +[Valid][1][UTT][ MSE = 0.3555 | LCC = 0.9004 | SRCC = 0.9010 ] +INFO:root:Average loss=0.2441364631652832 +INFO:root: +[Valid][2][UTT][ MSE = 0.3531 | LCC = 0.9076 | SRCC = 0.9041 ] +INFO:root:Average loss=0.23078054577112198 +INFO:root: +[Valid][3][UTT][ MSE = 0.3274 | LCC = 0.9205 | SRCC = 0.9169 ] +INFO:root:Average loss=0.2091682575941086 +INFO:root: +[Valid][4][UTT][ MSE = 0.3075 | LCC = 0.9196 | SRCC = 0.9180 ] +INFO:root:Average loss=0.18648098731040955 +INFO:root: +[Valid][5][UTT][ MSE = 0.2993 | LCC = 0.9270 | SRCC = 0.9264 ] +INFO:root:Average loss=0.17202079397439957 +INFO:root: +[Valid][6][UTT][ MSE = 0.2609 | LCC = 0.9286 | SRCC = 0.9260 ] +INFO:root:Average loss=0.15701868283748627 +INFO:root: +[Valid][7][UTT][ MSE = 0.2889 | LCC = 0.9315 | SRCC = 0.9297 ] +INFO:root:Average loss=0.1459222701191902 +INFO:root: +[Valid][8][UTT][ MSE = 0.2709 | LCC = 0.9331 | SRCC = 0.9310 ] +INFO:root:Average loss=0.13687228095531465 +INFO:root: +[Valid][9][UTT][ MSE = 0.2592 | LCC = 0.9304 | SRCC = 0.9289 ] +INFO:root:Average loss=0.13003526321053505 +INFO:root: +[Valid][10][UTT][ MSE = 0.2484 | LCC = 0.9354 | SRCC = 0.9336 ] +INFO:root:Average loss=0.119804019510746 +INFO:root: +[Valid][11][UTT][ MSE = 0.2795 | LCC = 0.9358 | SRCC = 0.9343 ] +INFO:root:Average loss=0.11506181120872498 +INFO:root: +[Valid][12][UTT][ MSE = 0.2109 | LCC = 0.9347 | SRCC = 0.9343 ] +INFO:root:Average loss=0.10376830863952637 +INFO:root: +[Valid][13][UTT][ MSE = 0.2545 | LCC = 0.9352 | SRCC = 0.9339 ] +INFO:root:Average loss=0.09997296813130378 +INFO:root: +[Valid][14][UTT][ MSE = 0.2233 | LCC = 0.9385 | SRCC = 0.9369 ] +INFO:root:Average loss=0.0918651013970375 +INFO:root: +[Valid][15][UTT][ MSE = 0.2336 | LCC = 0.9402 | SRCC = 0.9386 ] +INFO:root:Average loss=0.09385097107291221 +INFO:root: +[Valid][16][UTT][ MSE = 0.2510 | LCC = 0.9365 | SRCC = 0.9368 ] +INFO:root:Average loss=0.08681431943178176 +INFO:root: +[Valid][17][UTT][ MSE = 0.2621 | LCC = 0.9347 | SRCC = 0.9334 ] +INFO:root:Average loss=0.07836205244064331 +INFO:root: +[Valid][18][UTT][ MSE = 0.2255 | LCC = 0.9412 | SRCC = 0.9398 ] +INFO:root:Average loss=0.0748845112323761 +INFO:root: +[Valid][19][UTT][ MSE = 0.2367 | LCC = 0.9375 | SRCC = 0.9359 ] +INFO:root:Average loss=0.07535287314653397 +INFO:root: +[Valid][20][UTT][ MSE = 0.2318 | LCC = 0.9427 | SRCC = 0.9410 ] +INFO:root:Average loss=0.07183374950289727 +INFO:root: +[Valid][21][UTT][ MSE = 0.2184 | LCC = 0.9369 | SRCC = 0.9375 ] +INFO:root:Average loss=0.07265736064314843 +INFO:root: +[Valid][22][UTT][ MSE = 0.2325 | LCC = 0.9329 | SRCC = 0.9336 ] +INFO:root:Average loss=0.06379371964931488 +INFO:root: +[Valid][23][UTT][ MSE = 0.2181 | LCC = 0.9403 | SRCC = 0.9386 ] +INFO:root:Average loss=0.06324146249890328 +INFO:root: +[Valid][24][UTT][ MSE = 0.2371 | LCC = 0.9415 | SRCC = 0.9409 ] +INFO:root:Average loss=0.062080407947301866 +INFO:root: +[Valid][25][UTT][ MSE = 0.2693 | LCC = 0.9372 | SRCC = 0.9374 ] +INFO:root:Average loss=0.05755383232235908 +INFO:root: +[Valid][26][UTT][ MSE = 0.2287 | LCC = 0.9409 | SRCC = 0.9391 ] +INFO:root:Average loss=0.05325020705163479 +INFO:root: +[Valid][27][UTT][ MSE = 0.2290 | LCC = 0.9398 | SRCC = 0.9377 ] +INFO:root:Average loss=0.052282418116927146 +INFO:root: +[Valid][28][UTT][ MSE = 0.2181 | LCC = 0.9408 | SRCC = 0.9396 ] +INFO:root:Average loss=0.05332356825470924 +INFO:root: +[Valid][29][UTT][ MSE = 0.2262 | LCC = 0.9421 | SRCC = 0.9404 ] +INFO:root:Average loss=0.05330299389362335 +INFO:root: +[Valid][30][UTT][ MSE = 0.2091 | LCC = 0.9386 | SRCC = 0.9385 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2532 | LCC = 0.9343 | SRCC = 0.9330 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run3/config.gin b/results_tencent/w2v2_large_asr_zh/layer20/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run3/model.pt b/results_tencent/w2v2_large_asr_zh/layer20/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..73c20972026002fff28172ab332f176ba93bbea4 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1df27a06ab40a4304e1dacb5e9cfcbc29f23f178ee0af022815e2939e63b0860 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run3/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer20/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9cec489765e9f87bb5943ba80547645b2ef3a6c2 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:79b4f7e28f84962bb4dc8b038804b6d9629ab4b5481338bb1b8a59511a224296 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run3/train.log b/results_tencent/w2v2_large_asr_zh/layer20/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..82f8138763bb8bf38c23c07b2c22de78929c3658 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.48498155295848844 +INFO:root: +[Valid][0][UTT][ MSE = 0.5751 | LCC = 0.8872 | SRCC = 0.8820 ] +INFO:root:Average loss=0.2994481554031372 +INFO:root: +[Valid][1][UTT][ MSE = 0.4170 | LCC = 0.9020 | SRCC = 0.8999 ] +INFO:root:Average loss=0.24630956757068634 +INFO:root: +[Valid][2][UTT][ MSE = 0.3548 | LCC = 0.9122 | SRCC = 0.9095 ] +INFO:root:Average loss=0.21355943769216537 +INFO:root: +[Valid][3][UTT][ MSE = 0.3318 | LCC = 0.9223 | SRCC = 0.9194 ] +INFO:root:Average loss=0.1935590623021126 +INFO:root: +[Valid][4][UTT][ MSE = 0.3721 | LCC = 0.9257 | SRCC = 0.9238 ] +INFO:root:Average loss=0.1787436689734459 +INFO:root: +[Valid][5][UTT][ MSE = 0.3449 | LCC = 0.9311 | SRCC = 0.9300 ] +INFO:root:Average loss=0.16294836264848708 +INFO:root: +[Valid][6][UTT][ MSE = 0.3016 | LCC = 0.9341 | SRCC = 0.9316 ] +INFO:root:Average loss=0.15097577041387558 +INFO:root: +[Valid][7][UTT][ MSE = 0.2989 | LCC = 0.9357 | SRCC = 0.9354 ] +INFO:root:Average loss=0.14117570316791533 +INFO:root: +[Valid][8][UTT][ MSE = 0.2735 | LCC = 0.9371 | SRCC = 0.9352 ] +INFO:root:Average loss=0.13413269644975662 +INFO:root: +[Valid][9][UTT][ MSE = 0.2936 | LCC = 0.9362 | SRCC = 0.9348 ] +INFO:root:Average loss=0.12666071540117263 +INFO:root: +[Valid][10][UTT][ MSE = 0.2595 | LCC = 0.9387 | SRCC = 0.9364 ] +INFO:root:Average loss=0.11497615468502044 +INFO:root: +[Valid][11][UTT][ MSE = 0.2639 | LCC = 0.9378 | SRCC = 0.9365 ] +INFO:root:Average loss=0.10976570063829422 +INFO:root: +[Valid][12][UTT][ MSE = 0.3230 | LCC = 0.9390 | SRCC = 0.9387 ] +INFO:root:Average loss=0.10728880244493484 +INFO:root: +[Valid][13][UTT][ MSE = 0.2545 | LCC = 0.9409 | SRCC = 0.9390 ] +INFO:root:Average loss=0.09508484300971032 +INFO:root: +[Valid][14][UTT][ MSE = 0.2731 | LCC = 0.9410 | SRCC = 0.9398 ] +INFO:root:Average loss=0.09015534418821335 +INFO:root: +[Valid][15][UTT][ MSE = 0.2785 | LCC = 0.9429 | SRCC = 0.9414 ] +INFO:root:Average loss=0.0924232206940651 +INFO:root: +[Valid][16][UTT][ MSE = 0.2660 | LCC = 0.9412 | SRCC = 0.9399 ] +INFO:root:Average loss=0.08291665691137313 +INFO:root: +[Valid][17][UTT][ MSE = 0.2843 | LCC = 0.9414 | SRCC = 0.9402 ] +INFO:root:Average loss=0.08046780958771706 +INFO:root: +[Valid][18][UTT][ MSE = 0.2542 | LCC = 0.9427 | SRCC = 0.9403 ] +INFO:root:Average loss=0.0757774432003498 +INFO:root: +[Valid][19][UTT][ MSE = 0.2741 | LCC = 0.9424 | SRCC = 0.9412 ] +INFO:root:Average loss=0.07680464044213295 +INFO:root: +[Valid][20][UTT][ MSE = 0.2850 | LCC = 0.9436 | SRCC = 0.9427 ] +INFO:root:Average loss=0.07246995574235916 +INFO:root: +[Valid][21][UTT][ MSE = 0.2796 | LCC = 0.9419 | SRCC = 0.9399 ] +INFO:root:Average loss=0.06624887490272521 +INFO:root: +[Valid][22][UTT][ MSE = 0.2661 | LCC = 0.9431 | SRCC = 0.9403 ] +INFO:root:Average loss=0.06610964220762253 +INFO:root: +[Valid][23][UTT][ MSE = 0.2757 | LCC = 0.9417 | SRCC = 0.9402 ] +INFO:root:Average loss=0.06339628019928932 +INFO:root: +[Valid][24][UTT][ MSE = 0.2607 | LCC = 0.9442 | SRCC = 0.9418 ] +INFO:root:Average loss=0.06099192681908607 +INFO:root: +[Valid][25][UTT][ MSE = 0.2381 | LCC = 0.9418 | SRCC = 0.9392 ] +INFO:root:Average loss=0.0600903491973877 +INFO:root: +[Valid][26][UTT][ MSE = 0.2182 | LCC = 0.9401 | SRCC = 0.9375 ] +INFO:root:Average loss=0.05595304603874683 +INFO:root: +[Valid][27][UTT][ MSE = 0.2385 | LCC = 0.9402 | SRCC = 0.9366 ] +INFO:root:Average loss=0.05561095394194126 +INFO:root: +[Valid][28][UTT][ MSE = 0.2770 | LCC = 0.9424 | SRCC = 0.9407 ] +INFO:root:Average loss=0.05093849024176598 +INFO:root: +[Valid][29][UTT][ MSE = 0.2730 | LCC = 0.9409 | SRCC = 0.9392 ] +INFO:root:Average loss=0.052509452298283577 +INFO:root: +[Valid][30][UTT][ MSE = 0.2408 | LCC = 0.9431 | SRCC = 0.9405 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2821 | LCC = 0.9376 | SRCC = 0.9367 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run4/config.gin b/results_tencent/w2v2_large_asr_zh/layer20/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run4/model.pt b/results_tencent/w2v2_large_asr_zh/layer20/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..cccde8f6e6f22e46c59b46c0c76d6c61720d047b --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f5ff3f15486b0541609791edf63368c8e7370452538fecdf75b870c0f5c7bd58 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run4/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer20/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f7808c324d6dc8a6193f50bf5b01a6d34e041ad7 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0225729ec8aa089fa26f3e3adcee34be09fbf26218a636d71dcb0860a484d546 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer20/run4/train.log b/results_tencent/w2v2_large_asr_zh/layer20/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e1aecab7f4d8f21b83275b079ae27c7fd0fddd6f --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer20/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.50238991522789 +INFO:root: +[Valid][0][UTT][ MSE = 0.3905 | LCC = 0.8818 | SRCC = 0.8834 ] +INFO:root:Average loss=0.306810026884079 +INFO:root: +[Valid][1][UTT][ MSE = 0.4319 | LCC = 0.9000 | SRCC = 0.8969 ] +INFO:root:Average loss=0.27040499019622805 +INFO:root: +[Valid][2][UTT][ MSE = 0.3047 | LCC = 0.9108 | SRCC = 0.9093 ] +INFO:root:Average loss=0.2273172711133957 +INFO:root: +[Valid][3][UTT][ MSE = 0.3302 | LCC = 0.9198 | SRCC = 0.9172 ] +INFO:root:Average loss=0.2124250031709671 +INFO:root: +[Valid][4][UTT][ MSE = 0.2442 | LCC = 0.9218 | SRCC = 0.9198 ] +INFO:root:Average loss=0.1898493950366974 +INFO:root: +[Valid][5][UTT][ MSE = 0.3125 | LCC = 0.9265 | SRCC = 0.9252 ] +INFO:root:Average loss=0.16962408477067947 +INFO:root: +[Valid][6][UTT][ MSE = 0.2982 | LCC = 0.9335 | SRCC = 0.9322 ] +INFO:root:Average loss=0.1514303880929947 +INFO:root: +[Valid][7][UTT][ MSE = 0.2338 | LCC = 0.9374 | SRCC = 0.9341 ] +INFO:root:Average loss=0.1479457870721817 +INFO:root: +[Valid][8][UTT][ MSE = 0.3014 | LCC = 0.9381 | SRCC = 0.9373 ] +INFO:root:Average loss=0.13088696891069412 +INFO:root: +[Valid][9][UTT][ MSE = 0.2404 | LCC = 0.9408 | SRCC = 0.9386 ] +INFO:root:Average loss=0.11979520988464355 +INFO:root: +[Valid][10][UTT][ MSE = 0.2451 | LCC = 0.9419 | SRCC = 0.9396 ] +INFO:root:Average loss=0.11442206758260727 +INFO:root: +[Valid][11][UTT][ MSE = 0.2676 | LCC = 0.9413 | SRCC = 0.9398 ] +INFO:root:Average loss=0.10491722571849824 +INFO:root: +[Valid][12][UTT][ MSE = 0.2860 | LCC = 0.9425 | SRCC = 0.9403 ] +INFO:root:Average loss=0.10744428700208664 +INFO:root: +[Valid][13][UTT][ MSE = 0.2280 | LCC = 0.9437 | SRCC = 0.9409 ] +INFO:root:Average loss=0.09429724538326263 +INFO:root: +[Valid][14][UTT][ MSE = 0.2742 | LCC = 0.9399 | SRCC = 0.9388 ] +INFO:root:Average loss=0.09057559052109718 +INFO:root: +[Valid][15][UTT][ MSE = 0.2369 | LCC = 0.9428 | SRCC = 0.9396 ] +INFO:root:Average loss=0.08604418671131134 +INFO:root: +[Valid][16][UTT][ MSE = 0.2398 | LCC = 0.9442 | SRCC = 0.9408 ] +INFO:root:Average loss=0.08364660683274269 +INFO:root: +[Valid][17][UTT][ MSE = 0.2365 | LCC = 0.9447 | SRCC = 0.9427 ] +INFO:root:Average loss=0.0753620329797268 +INFO:root: +[Valid][18][UTT][ MSE = 0.2334 | LCC = 0.9447 | SRCC = 0.9419 ] +INFO:root:Average loss=0.07307341668009758 +INFO:root: +[Valid][19][UTT][ MSE = 0.2515 | LCC = 0.9431 | SRCC = 0.9413 ] +INFO:root:Average loss=0.06944124695658684 +INFO:root: +[Valid][20][UTT][ MSE = 0.2540 | LCC = 0.9451 | SRCC = 0.9430 ] +INFO:root:Average loss=0.06689742895960808 +INFO:root: +[Valid][21][UTT][ MSE = 0.2361 | LCC = 0.9445 | SRCC = 0.9428 ] +INFO:root:Average loss=0.06258686423301697 +INFO:root: +[Valid][22][UTT][ MSE = 0.2344 | LCC = 0.9453 | SRCC = 0.9439 ] +INFO:root:Average loss=0.062168100506067275 +INFO:root: +[Valid][23][UTT][ MSE = 0.2854 | LCC = 0.9403 | SRCC = 0.9411 ] +INFO:root:Average loss=0.06190564513206482 +INFO:root: +[Valid][24][UTT][ MSE = 0.2166 | LCC = 0.9437 | SRCC = 0.9410 ] +INFO:root:Average loss=0.05899948254227638 +INFO:root: +[Valid][25][UTT][ MSE = 0.2255 | LCC = 0.9441 | SRCC = 0.9420 ] +INFO:root:Average loss=0.05523637697100639 +INFO:root: +[Valid][26][UTT][ MSE = 0.2572 | LCC = 0.9437 | SRCC = 0.9419 ] +INFO:root:Average loss=0.053171177342534064 +INFO:root: +[Valid][27][UTT][ MSE = 0.2461 | LCC = 0.9428 | SRCC = 0.9412 ] +INFO:root:Average loss=0.05265368017554283 +INFO:root: +[Valid][28][UTT][ MSE = 0.2493 | LCC = 0.9448 | SRCC = 0.9429 ] +INFO:root:Average loss=0.04849704186618328 +INFO:root: +[Valid][29][UTT][ MSE = 0.2518 | LCC = 0.9433 | SRCC = 0.9422 ] +INFO:root:Average loss=0.051949161276221276 +INFO:root: +[Valid][30][UTT][ MSE = 0.2351 | LCC = 0.9440 | SRCC = 0.9428 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2465 | LCC = 0.9397 | SRCC = 0.9390 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run0/config.gin b/results_tencent/w2v2_large_asr_zh/layer22/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run0/model.pt b/results_tencent/w2v2_large_asr_zh/layer22/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8484cf5450decc44acd21fe32e6f5908b8694b5b --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f682a80331611a73dee7817bf6f4e84f896b087c4132da1187688b4937449ca8 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run0/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer22/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4e8921ea2a12b90b706a4329850c97d7dd873c1a --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8d422696f9a152f36645e5d407ed08b2580593e56d647d65f059489822f6414a +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run0/train.log b/results_tencent/w2v2_large_asr_zh/layer22/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9877898c3747204f2f87bd1bb9a55e74b7b100d6 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5161304049491883 +INFO:root: +[Valid][0][UTT][ MSE = 0.4423 | LCC = 0.8735 | SRCC = 0.8652 ] +INFO:root:Average loss=0.319636528134346 +INFO:root: +[Valid][1][UTT][ MSE = 0.4181 | LCC = 0.8946 | SRCC = 0.8894 ] +INFO:root:Average loss=0.2730831460952759 +INFO:root: +[Valid][2][UTT][ MSE = 0.3803 | LCC = 0.9063 | SRCC = 0.9029 ] +INFO:root:Average loss=0.24268186795711516 +INFO:root: +[Valid][3][UTT][ MSE = 0.3171 | LCC = 0.9155 | SRCC = 0.9124 ] +INFO:root:Average loss=0.2162714533805847 +INFO:root: +[Valid][4][UTT][ MSE = 0.3008 | LCC = 0.9214 | SRCC = 0.9192 ] +INFO:root:Average loss=0.19069519233703613 +INFO:root: +[Valid][5][UTT][ MSE = 0.3005 | LCC = 0.9261 | SRCC = 0.9239 ] +INFO:root:Average loss=0.1710389699935913 +INFO:root: +[Valid][6][UTT][ MSE = 0.2960 | LCC = 0.9269 | SRCC = 0.9265 ] +INFO:root:Average loss=0.16081360018253327 +INFO:root: +[Valid][7][UTT][ MSE = 0.3126 | LCC = 0.9254 | SRCC = 0.9233 ] +INFO:root:Average loss=0.15177860045433045 +INFO:root: +[Valid][8][UTT][ MSE = 0.2710 | LCC = 0.9303 | SRCC = 0.9279 ] +INFO:root:Average loss=0.1399022770524025 +INFO:root: +[Valid][9][UTT][ MSE = 0.2498 | LCC = 0.9306 | SRCC = 0.9289 ] +INFO:root:Average loss=0.13215042746067046 +INFO:root: +[Valid][10][UTT][ MSE = 0.2761 | LCC = 0.9320 | SRCC = 0.9306 ] +INFO:root:Average loss=0.11831232172250748 +INFO:root: +[Valid][11][UTT][ MSE = 0.2575 | LCC = 0.9358 | SRCC = 0.9341 ] +INFO:root:Average loss=0.10925416547060013 +INFO:root: +[Valid][12][UTT][ MSE = 0.2529 | LCC = 0.9345 | SRCC = 0.9323 ] +INFO:root:Average loss=0.10083406347036361 +INFO:root: +[Valid][13][UTT][ MSE = 0.2811 | LCC = 0.9343 | SRCC = 0.9327 ] +INFO:root:Average loss=0.09790593031048775 +INFO:root: +[Valid][14][UTT][ MSE = 0.2831 | LCC = 0.9366 | SRCC = 0.9348 ] +INFO:root:Average loss=0.09140561011433601 +INFO:root: +[Valid][15][UTT][ MSE = 0.2813 | LCC = 0.9347 | SRCC = 0.9319 ] +INFO:root:Average loss=0.08826278138160705 +INFO:root: +[Valid][16][UTT][ MSE = 0.2491 | LCC = 0.9352 | SRCC = 0.9331 ] +INFO:root:Average loss=0.08481054413318634 +INFO:root: +[Valid][17][UTT][ MSE = 0.2349 | LCC = 0.9321 | SRCC = 0.9323 ] +INFO:root:Average loss=0.07854460147023201 +INFO:root: +[Valid][18][UTT][ MSE = 0.2668 | LCC = 0.9325 | SRCC = 0.9305 ] +INFO:root:Average loss=0.07415925261378288 +INFO:root: +[Valid][19][UTT][ MSE = 0.2793 | LCC = 0.9326 | SRCC = 0.9309 ] +INFO:root:Average loss=0.06880004739761353 +INFO:root: +[Valid][20][UTT][ MSE = 0.2563 | LCC = 0.9351 | SRCC = 0.9335 ] +INFO:root:Average loss=0.06585723006725311 +INFO:root: +[Valid][21][UTT][ MSE = 0.2506 | LCC = 0.9365 | SRCC = 0.9344 ] +INFO:root:Average loss=0.06334813979268074 +INFO:root: +[Valid][22][UTT][ MSE = 0.2753 | LCC = 0.9356 | SRCC = 0.9342 ] +INFO:root:Average loss=0.062103536427021024 +INFO:root: +[Valid][23][UTT][ MSE = 0.2650 | LCC = 0.9361 | SRCC = 0.9334 ] +INFO:root:Average loss=0.0594316743016243 +INFO:root: +[Valid][24][UTT][ MSE = 0.2774 | LCC = 0.9293 | SRCC = 0.9276 ] +INFO:root:Average loss=0.054557727694511414 +INFO:root: +[Valid][25][UTT][ MSE = 0.2690 | LCC = 0.9321 | SRCC = 0.9314 ] +INFO:root:Average loss=0.054882973462343214 +INFO:root: +[Valid][26][UTT][ MSE = 0.2709 | LCC = 0.9354 | SRCC = 0.9350 ] +INFO:root:Average loss=0.051579668030142786 +INFO:root: +[Valid][27][UTT][ MSE = 0.2688 | LCC = 0.9367 | SRCC = 0.9339 ] +INFO:root:Average loss=0.05256671354174614 +INFO:root: +[Valid][28][UTT][ MSE = 0.2573 | LCC = 0.9360 | SRCC = 0.9334 ] +INFO:root:Average loss=0.04999315929412842 +INFO:root: +[Valid][29][UTT][ MSE = 0.2795 | LCC = 0.9353 | SRCC = 0.9329 ] +INFO:root:Average loss=0.04656087619066238 +INFO:root: +[Valid][30][UTT][ MSE = 0.2527 | LCC = 0.9373 | SRCC = 0.9347 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2697 | LCC = 0.9315 | SRCC = 0.9288 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run1/config.gin b/results_tencent/w2v2_large_asr_zh/layer22/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run1/model.pt b/results_tencent/w2v2_large_asr_zh/layer22/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ea183ed8e569473f4a4d4d0ad8740093717dceb3 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:74bc1a1f87031a586bf48fd345bf45f976d5f478cf6155385a37db0f9931c935 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run1/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer22/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..06f5997c18466988317d137d46fb231e49258630 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:668d5b2c963e069a1a10697711b2d95b48f0e317e42360605dd3110c11199ac9 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run1/train.log b/results_tencent/w2v2_large_asr_zh/layer22/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c309e41fb9a9e53a0f6f9223fdce53fdd931ded1 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.47858500277996063 +INFO:root: +[Valid][0][UTT][ MSE = 0.4354 | LCC = 0.8820 | SRCC = 0.8725 ] +INFO:root:Average loss=0.30118340003490446 +INFO:root: +[Valid][1][UTT][ MSE = 0.3688 | LCC = 0.8999 | SRCC = 0.8945 ] +INFO:root:Average loss=0.2616541152000427 +INFO:root: +[Valid][2][UTT][ MSE = 0.3926 | LCC = 0.9044 | SRCC = 0.9007 ] +INFO:root:Average loss=0.22926308298110962 +INFO:root: +[Valid][3][UTT][ MSE = 0.3352 | LCC = 0.9120 | SRCC = 0.9089 ] +INFO:root:Average loss=0.20540566593408585 +INFO:root: +[Valid][4][UTT][ MSE = 0.3016 | LCC = 0.9183 | SRCC = 0.9154 ] +INFO:root:Average loss=0.18986390578746795 +INFO:root: +[Valid][5][UTT][ MSE = 0.3412 | LCC = 0.9196 | SRCC = 0.9175 ] +INFO:root:Average loss=0.16862846076488494 +INFO:root: +[Valid][6][UTT][ MSE = 0.3209 | LCC = 0.9274 | SRCC = 0.9245 ] +INFO:root:Average loss=0.15790762084722518 +INFO:root: +[Valid][7][UTT][ MSE = 0.2591 | LCC = 0.9313 | SRCC = 0.9284 ] +INFO:root:Average loss=0.14800862723588942 +INFO:root: +[Valid][8][UTT][ MSE = 0.2533 | LCC = 0.9222 | SRCC = 0.9193 ] +INFO:root:Average loss=0.13517638963460923 +INFO:root: +[Valid][9][UTT][ MSE = 0.2810 | LCC = 0.9313 | SRCC = 0.9278 ] +INFO:root:Average loss=0.12488906919956207 +INFO:root: +[Valid][10][UTT][ MSE = 0.2739 | LCC = 0.9323 | SRCC = 0.9299 ] +INFO:root:Average loss=0.12026351642608643 +INFO:root: +[Valid][11][UTT][ MSE = 0.3034 | LCC = 0.9334 | SRCC = 0.9304 ] +INFO:root:Average loss=0.11020589148998261 +INFO:root: +[Valid][12][UTT][ MSE = 0.2701 | LCC = 0.9265 | SRCC = 0.9257 ] +INFO:root:Average loss=0.10330897510051727 +INFO:root: +[Valid][13][UTT][ MSE = 0.2512 | LCC = 0.9361 | SRCC = 0.9337 ] +INFO:root:Average loss=0.1009384842813015 +INFO:root: +[Valid][14][UTT][ MSE = 0.3243 | LCC = 0.9317 | SRCC = 0.9298 ] +INFO:root:Average loss=0.09490015670657158 +INFO:root: +[Valid][15][UTT][ MSE = 0.2524 | LCC = 0.9344 | SRCC = 0.9306 ] +INFO:root:Average loss=0.08226436248421669 +INFO:root: +[Valid][16][UTT][ MSE = 0.2166 | LCC = 0.9336 | SRCC = 0.9327 ] +INFO:root:Average loss=0.08183574131131172 +INFO:root: +[Valid][17][UTT][ MSE = 0.2598 | LCC = 0.9372 | SRCC = 0.9344 ] +INFO:root:Average loss=0.07673193201422691 +INFO:root: +[Valid][18][UTT][ MSE = 0.2585 | LCC = 0.9370 | SRCC = 0.9344 ] +INFO:root:Average loss=0.07411426311731338 +INFO:root: +[Valid][19][UTT][ MSE = 0.2252 | LCC = 0.9364 | SRCC = 0.9347 ] +INFO:root:Average loss=0.06920907178521156 +INFO:root: +[Valid][20][UTT][ MSE = 0.2612 | LCC = 0.9348 | SRCC = 0.9313 ] +INFO:root:Average loss=0.06895376527309417 +INFO:root: +[Valid][21][UTT][ MSE = 0.2519 | LCC = 0.9367 | SRCC = 0.9334 ] +INFO:root:Average loss=0.06548010662198067 +INFO:root: +[Valid][22][UTT][ MSE = 0.2410 | LCC = 0.9351 | SRCC = 0.9311 ] +INFO:root:Average loss=0.062162138521671295 +INFO:root: +[Valid][23][UTT][ MSE = 0.3075 | LCC = 0.9339 | SRCC = 0.9334 ] +INFO:root:Average loss=0.058095529928803444 +INFO:root: +[Valid][24][UTT][ MSE = 0.2361 | LCC = 0.9346 | SRCC = 0.9319 ] +INFO:root:Average loss=0.05699200057983399 +INFO:root: +[Valid][25][UTT][ MSE = 0.2359 | LCC = 0.9369 | SRCC = 0.9341 ] +INFO:root:Average loss=0.05272714561223984 +INFO:root: +[Valid][26][UTT][ MSE = 0.2319 | LCC = 0.9370 | SRCC = 0.9355 ] +INFO:root:Average loss=0.05566129893064499 +INFO:root: +[Valid][27][UTT][ MSE = 0.2387 | LCC = 0.9353 | SRCC = 0.9341 ] +INFO:root:Average loss=0.04965858379006386 +INFO:root: +[Valid][28][UTT][ MSE = 0.2278 | LCC = 0.9381 | SRCC = 0.9362 ] +INFO:root:Average loss=0.051240454882383345 +INFO:root: +[Valid][29][UTT][ MSE = 0.2558 | LCC = 0.9361 | SRCC = 0.9344 ] +INFO:root:Average loss=0.048939605012536046 +INFO:root: +[Valid][30][UTT][ MSE = 0.2300 | LCC = 0.9394 | SRCC = 0.9376 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2502 | LCC = 0.9311 | SRCC = 0.9297 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run2/config.gin b/results_tencent/w2v2_large_asr_zh/layer22/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run2/model.pt b/results_tencent/w2v2_large_asr_zh/layer22/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b783981a60a4643ed4d7e0d2df62800c9ce9db60 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b827a243efff69d120e296eee7a555b6c77d27f6fe8027ff84f8e5adde68323 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run2/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer22/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d42b9af504691d6b10ce72984e2fb9e5e1aa51ee --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:54a7ed112d79775b5eb67cf09e43e6dd261f51541c780179087fba7b3b4edea7 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run2/train.log b/results_tencent/w2v2_large_asr_zh/layer22/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..aa0ba0cbb79120e4c8b1d70c820eb2e007e55acc --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4744691332578659 +INFO:root: +[Valid][0][UTT][ MSE = 0.4652 | LCC = 0.8753 | SRCC = 0.8717 ] +INFO:root:Average loss=0.31531331765651704 +INFO:root: +[Valid][1][UTT][ MSE = 0.4023 | LCC = 0.8856 | SRCC = 0.8827 ] +INFO:root:Average loss=0.26966627502441404 +INFO:root: +[Valid][2][UTT][ MSE = 0.4358 | LCC = 0.9050 | SRCC = 0.9016 ] +INFO:root:Average loss=0.24080849587917327 +INFO:root: +[Valid][3][UTT][ MSE = 0.3524 | LCC = 0.9128 | SRCC = 0.9085 ] +INFO:root:Average loss=0.2190892105102539 +INFO:root: +[Valid][4][UTT][ MSE = 0.3620 | LCC = 0.9192 | SRCC = 0.9186 ] +INFO:root:Average loss=0.1934830181002617 +INFO:root: +[Valid][5][UTT][ MSE = 0.2889 | LCC = 0.9252 | SRCC = 0.9218 ] +INFO:root:Average loss=0.1784970177412033 +INFO:root: +[Valid][6][UTT][ MSE = 0.3001 | LCC = 0.9273 | SRCC = 0.9242 ] +INFO:root:Average loss=0.15917595565319062 +INFO:root: +[Valid][7][UTT][ MSE = 0.2728 | LCC = 0.9288 | SRCC = 0.9256 ] +INFO:root:Average loss=0.15963305723667146 +INFO:root: +[Valid][8][UTT][ MSE = 0.2693 | LCC = 0.9310 | SRCC = 0.9275 ] +INFO:root:Average loss=0.14285312342643738 +INFO:root: +[Valid][9][UTT][ MSE = 0.3136 | LCC = 0.9339 | SRCC = 0.9313 ] +INFO:root:Average loss=0.12460046309232711 +INFO:root: +[Valid][10][UTT][ MSE = 0.2674 | LCC = 0.9339 | SRCC = 0.9314 ] +INFO:root:Average loss=0.11825232648849487 +INFO:root: +[Valid][11][UTT][ MSE = 0.2588 | LCC = 0.9360 | SRCC = 0.9331 ] +INFO:root:Average loss=0.11295532810688019 +INFO:root: +[Valid][12][UTT][ MSE = 0.3137 | LCC = 0.9356 | SRCC = 0.9333 ] +INFO:root:Average loss=0.10817132666707038 +INFO:root: +[Valid][13][UTT][ MSE = 0.2589 | LCC = 0.9361 | SRCC = 0.9343 ] +INFO:root:Average loss=0.09686256504058838 +INFO:root: +[Valid][14][UTT][ MSE = 0.2629 | LCC = 0.9362 | SRCC = 0.9334 ] +INFO:root:Average loss=0.09385626301169396 +INFO:root: +[Valid][15][UTT][ MSE = 0.3070 | LCC = 0.9356 | SRCC = 0.9335 ] +INFO:root:Average loss=0.08684332877397537 +INFO:root: +[Valid][16][UTT][ MSE = 0.2218 | LCC = 0.9319 | SRCC = 0.9283 ] +INFO:root:Average loss=0.08694724425673485 +INFO:root: +[Valid][17][UTT][ MSE = 0.2415 | LCC = 0.9352 | SRCC = 0.9327 ] +INFO:root:Average loss=0.08149479576945305 +INFO:root: +[Valid][18][UTT][ MSE = 0.2495 | LCC = 0.9362 | SRCC = 0.9343 ] +INFO:root:Average loss=0.07624710807204246 +INFO:root: +[Valid][19][UTT][ MSE = 0.2626 | LCC = 0.9350 | SRCC = 0.9321 ] +INFO:root:Average loss=0.07431483775377273 +INFO:root: +[Valid][20][UTT][ MSE = 0.2463 | LCC = 0.9360 | SRCC = 0.9330 ] +INFO:root:Average loss=0.07197210884094238 +INFO:root: +[Valid][21][UTT][ MSE = 0.2806 | LCC = 0.9358 | SRCC = 0.9328 ] +INFO:root:Average loss=0.0686326133608818 +INFO:root: +[Valid][22][UTT][ MSE = 0.2336 | LCC = 0.9367 | SRCC = 0.9337 ] +INFO:root:Average loss=0.06125649008154869 +INFO:root: +[Valid][23][UTT][ MSE = 0.2692 | LCC = 0.9346 | SRCC = 0.9319 ] +INFO:root:Average loss=0.05916280049085617 +INFO:root: +[Valid][24][UTT][ MSE = 0.2682 | LCC = 0.9339 | SRCC = 0.9320 ] +INFO:root:Average loss=0.05613142907619476 +INFO:root: +[Valid][25][UTT][ MSE = 0.2353 | LCC = 0.9374 | SRCC = 0.9342 ] +INFO:root:Average loss=0.055397563457489016 +INFO:root: +[Valid][26][UTT][ MSE = 0.2642 | LCC = 0.9362 | SRCC = 0.9332 ] +INFO:root:Average loss=0.054119858980178834 +INFO:root: +[Valid][27][UTT][ MSE = 0.2597 | LCC = 0.9347 | SRCC = 0.9315 ] +INFO:root:Average loss=0.05529325014352798 +INFO:root: +[Valid][28][UTT][ MSE = 0.2543 | LCC = 0.9354 | SRCC = 0.9331 ] +INFO:root:Average loss=0.049486020654439924 +INFO:root: +[Valid][29][UTT][ MSE = 0.2647 | LCC = 0.9350 | SRCC = 0.9326 ] +INFO:root:Average loss=0.047445611089468 +INFO:root: +[Valid][30][UTT][ MSE = 0.2647 | LCC = 0.9335 | SRCC = 0.9309 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2554 | LCC = 0.9307 | SRCC = 0.9282 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run3/config.gin b/results_tencent/w2v2_large_asr_zh/layer22/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run3/model.pt b/results_tencent/w2v2_large_asr_zh/layer22/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..abf2d99e228c622f8e613a001933511deb296ae8 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0c34a9e8eaf22e1231d9df15d1c969f7526848c12c90f524f357fd65bdec9b70 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run3/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer22/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..96c6dac6af443e223e58d7b15677c09c58a0fed7 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:42be13535614cac4da6d6b7dab44e39f0c8b64797e5359cf445bb8d2375dbe33 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run3/train.log b/results_tencent/w2v2_large_asr_zh/layer22/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..119a279a09f5b54785b64486de063e46e5f643a5 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5161679733991623 +INFO:root: +[Valid][0][UTT][ MSE = 0.4116 | LCC = 0.8675 | SRCC = 0.8542 ] +INFO:root:Average loss=0.31319540095329285 +INFO:root: +[Valid][1][UTT][ MSE = 0.3502 | LCC = 0.8936 | SRCC = 0.8891 ] +INFO:root:Average loss=0.27086159646511077 +INFO:root: +[Valid][2][UTT][ MSE = 0.3422 | LCC = 0.9082 | SRCC = 0.9035 ] +INFO:root:Average loss=0.23495869034528732 +INFO:root: +[Valid][3][UTT][ MSE = 0.3286 | LCC = 0.9142 | SRCC = 0.9120 ] +INFO:root:Average loss=0.20486092454195023 +INFO:root: +[Valid][4][UTT][ MSE = 0.3125 | LCC = 0.9141 | SRCC = 0.9119 ] +INFO:root:Average loss=0.1850064714550972 +INFO:root: +[Valid][5][UTT][ MSE = 0.3199 | LCC = 0.9221 | SRCC = 0.9205 ] +INFO:root:Average loss=0.16989127975702287 +INFO:root: +[Valid][6][UTT][ MSE = 0.2532 | LCC = 0.9279 | SRCC = 0.9252 ] +INFO:root:Average loss=0.15697341299057008 +INFO:root: +[Valid][7][UTT][ MSE = 0.2542 | LCC = 0.9267 | SRCC = 0.9242 ] +INFO:root:Average loss=0.1418006989955902 +INFO:root: +[Valid][8][UTT][ MSE = 0.2496 | LCC = 0.9288 | SRCC = 0.9282 ] +INFO:root:Average loss=0.1311200531721115 +INFO:root: +[Valid][9][UTT][ MSE = 0.2446 | LCC = 0.9321 | SRCC = 0.9292 ] +INFO:root:Average loss=0.12789783537387847 +INFO:root: +[Valid][10][UTT][ MSE = 0.2361 | LCC = 0.9333 | SRCC = 0.9335 ] +INFO:root:Average loss=0.12231555593013764 +INFO:root: +[Valid][11][UTT][ MSE = 0.2212 | LCC = 0.9351 | SRCC = 0.9311 ] +INFO:root:Average loss=0.10714401829242706 +INFO:root: +[Valid][12][UTT][ MSE = 0.2497 | LCC = 0.9361 | SRCC = 0.9332 ] +INFO:root:Average loss=0.09776117467880249 +INFO:root: +[Valid][13][UTT][ MSE = 0.2402 | LCC = 0.9361 | SRCC = 0.9359 ] +INFO:root:Average loss=0.09914157500863076 +INFO:root: +[Valid][14][UTT][ MSE = 0.2876 | LCC = 0.9318 | SRCC = 0.9305 ] +INFO:root:Average loss=0.09228146708011627 +INFO:root: +[Valid][15][UTT][ MSE = 0.2588 | LCC = 0.9371 | SRCC = 0.9341 ] +INFO:root:Average loss=0.08454653698205948 +INFO:root: +[Valid][16][UTT][ MSE = 0.2700 | LCC = 0.9317 | SRCC = 0.9300 ] +INFO:root:Average loss=0.08199073261022567 +INFO:root: +[Valid][17][UTT][ MSE = 0.2835 | LCC = 0.9374 | SRCC = 0.9350 ] +INFO:root:Average loss=0.07907140386104584 +INFO:root: +[Valid][18][UTT][ MSE = 0.2755 | LCC = 0.9345 | SRCC = 0.9340 ] +INFO:root:Average loss=0.07510945722460746 +INFO:root: +[Valid][19][UTT][ MSE = 0.3022 | LCC = 0.9344 | SRCC = 0.9304 ] +INFO:root:Average loss=0.07328527158498764 +INFO:root: +[Valid][20][UTT][ MSE = 0.2405 | LCC = 0.9349 | SRCC = 0.9343 ] +INFO:root:Average loss=0.06657692775130272 +INFO:root: +[Valid][21][UTT][ MSE = 0.2702 | LCC = 0.9352 | SRCC = 0.9333 ] +INFO:root:Average loss=0.0646938986480236 +INFO:root: +[Valid][22][UTT][ MSE = 0.2579 | LCC = 0.9365 | SRCC = 0.9347 ] +INFO:root:Average loss=0.06215720161795616 +INFO:root: +[Valid][23][UTT][ MSE = 0.2432 | LCC = 0.9362 | SRCC = 0.9332 ] +INFO:root:Average loss=0.06174397096037865 +INFO:root: +[Valid][24][UTT][ MSE = 0.2900 | LCC = 0.9360 | SRCC = 0.9341 ] +INFO:root:Average loss=0.05452759262919426 +INFO:root: +[Valid][25][UTT][ MSE = 0.2637 | LCC = 0.9330 | SRCC = 0.9321 ] +INFO:root:Average loss=0.05302037248015404 +INFO:root: +[Valid][26][UTT][ MSE = 0.2618 | LCC = 0.9367 | SRCC = 0.9346 ] +INFO:root:Average loss=0.05109695869684219 +INFO:root: +[Valid][27][UTT][ MSE = 0.2707 | LCC = 0.9331 | SRCC = 0.9318 ] +INFO:root:Average loss=0.048795543044805524 +INFO:root: +[Valid][28][UTT][ MSE = 0.2763 | LCC = 0.9332 | SRCC = 0.9325 ] +INFO:root:Average loss=0.047056369543075564 +INFO:root: +[Valid][29][UTT][ MSE = 0.2748 | LCC = 0.9363 | SRCC = 0.9340 ] +INFO:root:Average loss=0.04603902712464333 +INFO:root: +[Valid][30][UTT][ MSE = 0.2612 | LCC = 0.9357 | SRCC = 0.9333 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2944 | LCC = 0.9301 | SRCC = 0.9287 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run4/config.gin b/results_tencent/w2v2_large_asr_zh/layer22/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run4/model.pt b/results_tencent/w2v2_large_asr_zh/layer22/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7c565f48af7fecbb59c6a8379a4d85edde2057b8 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f8b3aa91d6a0c3cecc98500956f35267d2d85ec53819317ac8b73c1bd6025549 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run4/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer22/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a0e0c9517e72c784c8c0991355a261e56b60d7b8 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3e0441dcf0e5c851f0f920b7d60ff6a1b08df1c4b0862032cfdd8ffff7393ea8 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer22/run4/train.log b/results_tencent/w2v2_large_asr_zh/layer22/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0938a8f6906c7ab9bd0ca8a63b4f9751a986628e --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer22/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.51064084482193 +INFO:root: +[Valid][0][UTT][ MSE = 0.4279 | LCC = 0.8741 | SRCC = 0.8665 ] +INFO:root:Average loss=0.3319871335029602 +INFO:root: +[Valid][1][UTT][ MSE = 0.4623 | LCC = 0.8858 | SRCC = 0.8818 ] +INFO:root:Average loss=0.26963654643297197 +INFO:root: +[Valid][2][UTT][ MSE = 0.3516 | LCC = 0.9056 | SRCC = 0.9025 ] +INFO:root:Average loss=0.2535505484342575 +INFO:root: +[Valid][3][UTT][ MSE = 0.3412 | LCC = 0.9128 | SRCC = 0.9086 ] +INFO:root:Average loss=0.21596389639377594 +INFO:root: +[Valid][4][UTT][ MSE = 0.5167 | LCC = 0.9084 | SRCC = 0.9049 ] +INFO:root:Average loss=0.22063978600502016 +INFO:root: +[Valid][5][UTT][ MSE = 0.3536 | LCC = 0.9216 | SRCC = 0.9181 ] +INFO:root:Average loss=0.1786582607626915 +INFO:root: +[Valid][6][UTT][ MSE = 0.2531 | LCC = 0.9218 | SRCC = 0.9199 ] +INFO:root:Average loss=0.1602436902523041 +INFO:root: +[Valid][7][UTT][ MSE = 0.2988 | LCC = 0.9278 | SRCC = 0.9250 ] +INFO:root:Average loss=0.1570935863852501 +INFO:root: +[Valid][8][UTT][ MSE = 0.3002 | LCC = 0.9266 | SRCC = 0.9240 ] +INFO:root:Average loss=0.13882913464307786 +INFO:root: +[Valid][9][UTT][ MSE = 0.3379 | LCC = 0.9219 | SRCC = 0.9222 ] +INFO:root:Average loss=0.12796907770633698 +INFO:root: +[Valid][10][UTT][ MSE = 0.2759 | LCC = 0.9309 | SRCC = 0.9291 ] +INFO:root:Average loss=0.12577569258213042 +INFO:root: +[Valid][11][UTT][ MSE = 0.2863 | LCC = 0.9342 | SRCC = 0.9319 ] +INFO:root:Average loss=0.1090001215338707 +INFO:root: +[Valid][12][UTT][ MSE = 0.2479 | LCC = 0.9323 | SRCC = 0.9306 ] +INFO:root:Average loss=0.10550730714201927 +INFO:root: +[Valid][13][UTT][ MSE = 0.2393 | LCC = 0.9331 | SRCC = 0.9314 ] +INFO:root:Average loss=0.09749835106730462 +INFO:root: +[Valid][14][UTT][ MSE = 0.2531 | LCC = 0.9314 | SRCC = 0.9278 ] +INFO:root:Average loss=0.09040098363161087 +INFO:root: +[Valid][15][UTT][ MSE = 0.2439 | LCC = 0.9339 | SRCC = 0.9319 ] +INFO:root:Average loss=0.08410962989926338 +INFO:root: +[Valid][16][UTT][ MSE = 0.2588 | LCC = 0.9346 | SRCC = 0.9314 ] +INFO:root:Average loss=0.08351880076527596 +INFO:root: +[Valid][17][UTT][ MSE = 0.2560 | LCC = 0.9348 | SRCC = 0.9322 ] +INFO:root:Average loss=0.07820993947982788 +INFO:root: +[Valid][18][UTT][ MSE = 0.2716 | LCC = 0.9327 | SRCC = 0.9325 ] +INFO:root:Average loss=0.07711214110255242 +INFO:root: +[Valid][19][UTT][ MSE = 0.2506 | LCC = 0.9359 | SRCC = 0.9331 ] +INFO:root:Average loss=0.06930626383423805 +INFO:root: +[Valid][20][UTT][ MSE = 0.2503 | LCC = 0.9365 | SRCC = 0.9328 ] +INFO:root:Average loss=0.06483596585690975 +INFO:root: +[Valid][21][UTT][ MSE = 0.2959 | LCC = 0.9332 | SRCC = 0.9292 ] +INFO:root:Average loss=0.06269636437296867 +INFO:root: +[Valid][22][UTT][ MSE = 0.2643 | LCC = 0.9264 | SRCC = 0.9237 ] +INFO:root:Average loss=0.0601298486739397 +INFO:root: +[Valid][23][UTT][ MSE = 0.2515 | LCC = 0.9346 | SRCC = 0.9312 ] +INFO:root:Average loss=0.05747359216213226 +INFO:root: +[Valid][24][UTT][ MSE = 0.2394 | LCC = 0.9360 | SRCC = 0.9328 ] +INFO:root:Average loss=0.05603893445432186 +INFO:root: +[Valid][25][UTT][ MSE = 0.2365 | LCC = 0.9329 | SRCC = 0.9305 ] +INFO:root:Average loss=0.05384180098772049 +INFO:root: +[Valid][26][UTT][ MSE = 0.2604 | LCC = 0.9355 | SRCC = 0.9319 ] +INFO:root:Average loss=0.05008410167694092 +INFO:root: +[Valid][27][UTT][ MSE = 0.2704 | LCC = 0.9325 | SRCC = 0.9294 ] +INFO:root:Average loss=0.05272351640462875 +INFO:root: +[Valid][28][UTT][ MSE = 0.2721 | LCC = 0.9333 | SRCC = 0.9291 ] +INFO:root:Average loss=0.04569090819358826 +INFO:root: +[Valid][29][UTT][ MSE = 0.2632 | LCC = 0.9326 | SRCC = 0.9304 ] +INFO:root:Average loss=0.046122729629278186 +INFO:root: +[Valid][30][UTT][ MSE = 0.2417 | LCC = 0.9331 | SRCC = 0.9292 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2680 | LCC = 0.9299 | SRCC = 0.9276 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run0/config.gin b/results_tencent/w2v2_large_asr_zh/layer4/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run0/model.pt b/results_tencent/w2v2_large_asr_zh/layer4/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..991e2194440adea3485ac94c4f8e1371ca7473c1 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7806ea86654013b531b434dadbc8a4b71793527c1e5221346f4d85ce516b8965 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run0/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer4/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7e493ec6290c620e2cbad09c1f54c09869c4fa70 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d88364ceeaa8611c89afa2c8911741abe11681334cfd917c055d1fdbd3dd1ed7 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run0/train.log b/results_tencent/w2v2_large_asr_zh/layer4/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..bb0d400176acc32c1b143916afe40766c369fbdd --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4671582734584808 +INFO:root: +[Valid][0][UTT][ MSE = 0.2892 | LCC = 0.9218 | SRCC = 0.9231 ] +INFO:root:Average loss=0.21727392601966858 +INFO:root: +[Valid][1][UTT][ MSE = 0.2593 | LCC = 0.9379 | SRCC = 0.9400 ] +INFO:root:Average loss=0.1798443005681038 +INFO:root: +[Valid][2][UTT][ MSE = 0.2695 | LCC = 0.9447 | SRCC = 0.9445 ] +INFO:root:Average loss=0.1508446066379547 +INFO:root: +[Valid][3][UTT][ MSE = 0.2156 | LCC = 0.9492 | SRCC = 0.9521 ] +INFO:root:Average loss=0.1344508199095726 +INFO:root: +[Valid][4][UTT][ MSE = 0.2456 | LCC = 0.9541 | SRCC = 0.9551 ] +INFO:root:Average loss=0.12147651362419129 +INFO:root: +[Valid][5][UTT][ MSE = 0.1812 | LCC = 0.9556 | SRCC = 0.9580 ] +INFO:root:Average loss=0.11504247885942459 +INFO:root: +[Valid][6][UTT][ MSE = 0.2253 | LCC = 0.9593 | SRCC = 0.9603 ] +INFO:root:Average loss=0.10327682510018349 +INFO:root: +[Valid][7][UTT][ MSE = 0.2102 | LCC = 0.9592 | SRCC = 0.9611 ] +INFO:root:Average loss=0.09231534826755523 +INFO:root: +[Valid][8][UTT][ MSE = 0.1592 | LCC = 0.9597 | SRCC = 0.9615 ] +INFO:root:Average loss=0.08757446876168251 +INFO:root: +[Valid][9][UTT][ MSE = 0.1572 | LCC = 0.9581 | SRCC = 0.9623 ] +INFO:root:Average loss=0.08259823921322823 +INFO:root: +[Valid][10][UTT][ MSE = 0.1872 | LCC = 0.9630 | SRCC = 0.9638 ] +INFO:root:Average loss=0.07898971948027611 +INFO:root: +[Valid][11][UTT][ MSE = 0.1718 | LCC = 0.9644 | SRCC = 0.9647 ] +INFO:root:Average loss=0.07423610916733742 +INFO:root: +[Valid][12][UTT][ MSE = 0.2118 | LCC = 0.9636 | SRCC = 0.9643 ] +INFO:root:Average loss=0.07109002423286438 +INFO:root: +[Valid][13][UTT][ MSE = 0.1890 | LCC = 0.9629 | SRCC = 0.9640 ] +INFO:root:Average loss=0.06924077841639519 +INFO:root: +[Valid][14][UTT][ MSE = 0.1823 | LCC = 0.9658 | SRCC = 0.9664 ] +INFO:root:Average loss=0.06678023666143418 +INFO:root: +[Valid][15][UTT][ MSE = 0.1800 | LCC = 0.9644 | SRCC = 0.9648 ] +INFO:root:Average loss=0.05991326171159744 +INFO:root: +[Valid][16][UTT][ MSE = 0.1846 | LCC = 0.9663 | SRCC = 0.9664 ] +INFO:root:Average loss=0.058552637189626694 +INFO:root: +[Valid][17][UTT][ MSE = 0.1976 | LCC = 0.9649 | SRCC = 0.9649 ] +INFO:root:Average loss=0.057108530372381214 +INFO:root: +[Valid][18][UTT][ MSE = 0.1677 | LCC = 0.9658 | SRCC = 0.9663 ] +INFO:root:Average loss=0.05402103389799595 +INFO:root: +[Valid][19][UTT][ MSE = 0.1766 | LCC = 0.9640 | SRCC = 0.9644 ] +INFO:root:Average loss=0.05267316478490829 +INFO:root: +[Valid][20][UTT][ MSE = 0.1808 | LCC = 0.9661 | SRCC = 0.9657 ] +INFO:root:Average loss=0.050527893885970114 +INFO:root: +[Valid][21][UTT][ MSE = 0.1867 | LCC = 0.9659 | SRCC = 0.9658 ] +INFO:root:Average loss=0.04730535684525967 +INFO:root: +[Valid][22][UTT][ MSE = 0.1649 | LCC = 0.9654 | SRCC = 0.9659 ] +INFO:root:Average loss=0.0459296455681324 +INFO:root: +[Valid][23][UTT][ MSE = 0.1909 | LCC = 0.9661 | SRCC = 0.9667 ] +INFO:root:Average loss=0.04417700645327568 +INFO:root: +[Valid][24][UTT][ MSE = 0.1455 | LCC = 0.9645 | SRCC = 0.9651 ] +INFO:root:Average loss=0.042902085840702056 +INFO:root: +[Valid][25][UTT][ MSE = 0.1863 | LCC = 0.9676 | SRCC = 0.9671 ] +INFO:root:Average loss=0.040491386488080025 +INFO:root: +[Valid][26][UTT][ MSE = 0.1749 | LCC = 0.9642 | SRCC = 0.9642 ] +INFO:root:Average loss=0.039115276739001274 +INFO:root: +[Valid][27][UTT][ MSE = 0.1885 | LCC = 0.9660 | SRCC = 0.9660 ] +INFO:root:Average loss=0.039569177746772766 +INFO:root: +[Valid][28][UTT][ MSE = 0.1954 | LCC = 0.9652 | SRCC = 0.9648 ] +INFO:root:Average loss=0.03594486711919308 +INFO:root: +[Valid][29][UTT][ MSE = 0.1832 | LCC = 0.9661 | SRCC = 0.9654 ] +INFO:root:Average loss=0.034316850632429126 +INFO:root: +[Valid][30][UTT][ MSE = 0.1761 | LCC = 0.9665 | SRCC = 0.9659 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1998 | LCC = 0.9617 | SRCC = 0.9629 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run1/config.gin b/results_tencent/w2v2_large_asr_zh/layer4/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run1/model.pt b/results_tencent/w2v2_large_asr_zh/layer4/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8667f43d329d6b3cfb8132f7e93408967ac4ff63 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:886566fbcff8d16fd89da26f76740757df5ec46655333c2a26e19cdf001d1522 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run1/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer4/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ec818a60242f4f6c948e6bd53871a53d2b00809e --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dd44c4492fe0d0acf165922e29670a168da60bb1a06de203c86758328096c724 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run1/train.log b/results_tencent/w2v2_large_asr_zh/layer4/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9dfedda25839d3e737f61d952e02481f776ab7a2 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.41175501644611356 +INFO:root: +[Valid][0][UTT][ MSE = 0.2667 | LCC = 0.9247 | SRCC = 0.9258 ] +INFO:root:Average loss=0.20096661353111267 +INFO:root: +[Valid][1][UTT][ MSE = 0.2474 | LCC = 0.9393 | SRCC = 0.9406 ] +INFO:root:Average loss=0.16188838398456573 +INFO:root: +[Valid][2][UTT][ MSE = 0.2422 | LCC = 0.9512 | SRCC = 0.9526 ] +INFO:root:Average loss=0.13054424166679382 +INFO:root: +[Valid][3][UTT][ MSE = 0.2094 | LCC = 0.9557 | SRCC = 0.9570 ] +INFO:root:Average loss=0.11951795023679733 +INFO:root: +[Valid][4][UTT][ MSE = 0.2200 | LCC = 0.9567 | SRCC = 0.9579 ] +INFO:root:Average loss=0.11248419618606567 +INFO:root: +[Valid][5][UTT][ MSE = 0.1734 | LCC = 0.9589 | SRCC = 0.9595 ] +INFO:root:Average loss=0.10530249536037445 +INFO:root: +[Valid][6][UTT][ MSE = 0.1975 | LCC = 0.9605 | SRCC = 0.9610 ] +INFO:root:Average loss=0.09281405571103096 +INFO:root: +[Valid][7][UTT][ MSE = 0.2308 | LCC = 0.9613 | SRCC = 0.9637 ] +INFO:root:Average loss=0.09150765004754066 +INFO:root: +[Valid][8][UTT][ MSE = 0.2033 | LCC = 0.9622 | SRCC = 0.9631 ] +INFO:root:Average loss=0.0894284320473671 +INFO:root: +[Valid][9][UTT][ MSE = 0.2253 | LCC = 0.9572 | SRCC = 0.9620 ] +INFO:root:Average loss=0.08401640465855599 +INFO:root: +[Valid][10][UTT][ MSE = 0.1691 | LCC = 0.9633 | SRCC = 0.9643 ] +INFO:root:Average loss=0.0806079881489277 +INFO:root: +[Valid][11][UTT][ MSE = 0.2237 | LCC = 0.9601 | SRCC = 0.9631 ] +INFO:root:Average loss=0.07551345634460449 +INFO:root: +[Valid][12][UTT][ MSE = 0.1843 | LCC = 0.9642 | SRCC = 0.9649 ] +INFO:root:Average loss=0.07027350610494613 +INFO:root: +[Valid][13][UTT][ MSE = 0.2000 | LCC = 0.9638 | SRCC = 0.9648 ] +INFO:root:Average loss=0.06874603250622749 +INFO:root: +[Valid][14][UTT][ MSE = 0.1723 | LCC = 0.9627 | SRCC = 0.9642 ] +INFO:root:Average loss=0.06515091714262962 +INFO:root: +[Valid][15][UTT][ MSE = 0.1997 | LCC = 0.9630 | SRCC = 0.9638 ] +INFO:root:Average loss=0.05837149491906166 +INFO:root: +[Valid][16][UTT][ MSE = 0.1813 | LCC = 0.9636 | SRCC = 0.9642 ] +INFO:root:Average loss=0.05706763634085655 +INFO:root: +[Valid][17][UTT][ MSE = 0.1869 | LCC = 0.9655 | SRCC = 0.9661 ] +INFO:root:Average loss=0.05439202037453651 +INFO:root: +[Valid][18][UTT][ MSE = 0.2117 | LCC = 0.9602 | SRCC = 0.9634 ] +INFO:root:Average loss=0.05490518471598625 +INFO:root: +[Valid][19][UTT][ MSE = 0.2036 | LCC = 0.9650 | SRCC = 0.9654 ] +INFO:root:Average loss=0.056752555042505266 +INFO:root: +[Valid][20][UTT][ MSE = 0.1970 | LCC = 0.9640 | SRCC = 0.9650 ] +INFO:root:Average loss=0.050609809175133705 +INFO:root: +[Valid][21][UTT][ MSE = 0.1961 | LCC = 0.9653 | SRCC = 0.9659 ] +INFO:root:Average loss=0.045911633655428885 +INFO:root: +[Valid][22][UTT][ MSE = 0.1749 | LCC = 0.9628 | SRCC = 0.9631 ] +INFO:root:Average loss=0.04332199047505855 +INFO:root: +[Valid][23][UTT][ MSE = 0.1737 | LCC = 0.9659 | SRCC = 0.9666 ] +INFO:root:Average loss=0.04232223811745644 +INFO:root: +[Valid][24][UTT][ MSE = 0.1766 | LCC = 0.9648 | SRCC = 0.9656 ] +INFO:root:Average loss=0.042647529527544976 +INFO:root: +[Valid][25][UTT][ MSE = 0.1926 | LCC = 0.9646 | SRCC = 0.9648 ] +INFO:root:Average loss=0.03865544086694717 +INFO:root: +[Valid][26][UTT][ MSE = 0.1738 | LCC = 0.9655 | SRCC = 0.9656 ] +INFO:root:Average loss=0.038987865298986436 +INFO:root: +[Valid][27][UTT][ MSE = 0.1633 | LCC = 0.9653 | SRCC = 0.9656 ] +INFO:root:Average loss=0.03620562508702278 +INFO:root: +[Valid][28][UTT][ MSE = 0.1807 | LCC = 0.9655 | SRCC = 0.9656 ] +INFO:root:Average loss=0.03576530151069164 +INFO:root: +[Valid][29][UTT][ MSE = 0.1752 | LCC = 0.9632 | SRCC = 0.9632 ] +INFO:root:Average loss=0.03365074297785759 +INFO:root: +[Valid][30][UTT][ MSE = 0.1651 | LCC = 0.9648 | SRCC = 0.9655 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1828 | LCC = 0.9618 | SRCC = 0.9631 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run2/config.gin b/results_tencent/w2v2_large_asr_zh/layer4/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run2/model.pt b/results_tencent/w2v2_large_asr_zh/layer4/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..555322b693bfe1f6652c37a40d71a982dc7bec52 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b52ba3209f87135dbb2c2da47fc7050af8d95de2e319521897d9695ced981d67 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run2/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer4/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..463eddb62fca221532a2d28c3f8dde5d2bb0c828 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:084f83184f97dd22cf62d5dc56e62e6742f5900ad8bbfd1b75abeacb557feaad +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run2/train.log b/results_tencent/w2v2_large_asr_zh/layer4/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..91dc9432ab53c1562b1d149fef657d0ed7eb4014 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.45331639778614047 +INFO:root: +[Valid][0][UTT][ MSE = 0.3164 | LCC = 0.9198 | SRCC = 0.9202 ] +INFO:root:Average loss=0.22142437344789506 +INFO:root: +[Valid][1][UTT][ MSE = 0.2258 | LCC = 0.9384 | SRCC = 0.9393 ] +INFO:root:Average loss=0.17671009957790373 +INFO:root: +[Valid][2][UTT][ MSE = 0.2607 | LCC = 0.9450 | SRCC = 0.9464 ] +INFO:root:Average loss=0.1501505309343338 +INFO:root: +[Valid][3][UTT][ MSE = 0.2534 | LCC = 0.9517 | SRCC = 0.9525 ] +INFO:root:Average loss=0.12883328020572662 +INFO:root: +[Valid][4][UTT][ MSE = 0.2076 | LCC = 0.9563 | SRCC = 0.9569 ] +INFO:root:Average loss=0.12623460161685943 +INFO:root: +[Valid][5][UTT][ MSE = 0.2343 | LCC = 0.9555 | SRCC = 0.9584 ] +INFO:root:Average loss=0.11267994132637978 +INFO:root: +[Valid][6][UTT][ MSE = 0.2732 | LCC = 0.9599 | SRCC = 0.9610 ] +INFO:root:Average loss=0.10503767144680023 +INFO:root: +[Valid][7][UTT][ MSE = 0.1896 | LCC = 0.9584 | SRCC = 0.9600 ] +INFO:root:Average loss=0.10119824361801147 +INFO:root: +[Valid][8][UTT][ MSE = 0.1944 | LCC = 0.9607 | SRCC = 0.9623 ] +INFO:root:Average loss=0.08719756415486336 +INFO:root: +[Valid][9][UTT][ MSE = 0.1668 | LCC = 0.9615 | SRCC = 0.9628 ] +INFO:root:Average loss=0.08426101487874985 +INFO:root: +[Valid][10][UTT][ MSE = 0.1898 | LCC = 0.9637 | SRCC = 0.9646 ] +INFO:root:Average loss=0.07655566722154618 +INFO:root: +[Valid][11][UTT][ MSE = 0.2042 | LCC = 0.9630 | SRCC = 0.9630 ] +INFO:root:Average loss=0.07678084596991538 +INFO:root: +[Valid][12][UTT][ MSE = 0.1745 | LCC = 0.9645 | SRCC = 0.9645 ] +INFO:root:Average loss=0.06936757075786591 +INFO:root: +[Valid][13][UTT][ MSE = 0.1745 | LCC = 0.9629 | SRCC = 0.9644 ] +INFO:root:Average loss=0.06844581326842308 +INFO:root: +[Valid][14][UTT][ MSE = 0.2065 | LCC = 0.9645 | SRCC = 0.9651 ] +INFO:root:Average loss=0.06283505660295487 +INFO:root: +[Valid][15][UTT][ MSE = 0.1856 | LCC = 0.9644 | SRCC = 0.9651 ] +INFO:root:Average loss=0.060035881251096726 +INFO:root: +[Valid][16][UTT][ MSE = 0.1689 | LCC = 0.9643 | SRCC = 0.9647 ] +INFO:root:Average loss=0.058048069149255754 +INFO:root: +[Valid][17][UTT][ MSE = 0.1880 | LCC = 0.9639 | SRCC = 0.9647 ] +INFO:root:Average loss=0.054372497811913494 +INFO:root: +[Valid][18][UTT][ MSE = 0.1750 | LCC = 0.9630 | SRCC = 0.9636 ] +INFO:root:Average loss=0.05275444401800632 +INFO:root: +[Valid][19][UTT][ MSE = 0.1942 | LCC = 0.9649 | SRCC = 0.9649 ] +INFO:root:Average loss=0.050702056795358656 +INFO:root: +[Valid][20][UTT][ MSE = 0.1715 | LCC = 0.9656 | SRCC = 0.9661 ] +INFO:root:Average loss=0.047710788369178775 +INFO:root: +[Valid][21][UTT][ MSE = 0.1899 | LCC = 0.9650 | SRCC = 0.9648 ] +INFO:root:Average loss=0.04772174316644669 +INFO:root: +[Valid][22][UTT][ MSE = 0.1981 | LCC = 0.9643 | SRCC = 0.9649 ] +INFO:root:Average loss=0.04379465852677822 +INFO:root: +[Valid][23][UTT][ MSE = 0.2092 | LCC = 0.9650 | SRCC = 0.9651 ] +INFO:root:Average loss=0.04392327529191971 +INFO:root: +[Valid][24][UTT][ MSE = 0.1738 | LCC = 0.9655 | SRCC = 0.9652 ] +INFO:root:Average loss=0.03984903173148632 +INFO:root: +[Valid][25][UTT][ MSE = 0.1662 | LCC = 0.9648 | SRCC = 0.9649 ] +INFO:root:Average loss=0.041132183745503426 +INFO:root: +[Valid][26][UTT][ MSE = 0.2022 | LCC = 0.9640 | SRCC = 0.9646 ] +INFO:root:Average loss=0.039336875230073926 +INFO:root: +[Valid][27][UTT][ MSE = 0.1859 | LCC = 0.9656 | SRCC = 0.9658 ] +INFO:root:Average loss=0.03785022056102753 +INFO:root: +[Valid][28][UTT][ MSE = 0.1769 | LCC = 0.9652 | SRCC = 0.9650 ] +INFO:root:Average loss=0.03453517657518387 +INFO:root: +[Valid][29][UTT][ MSE = 0.1919 | LCC = 0.9654 | SRCC = 0.9656 ] +INFO:root:Average loss=0.03484609822928905 +INFO:root: +[Valid][30][UTT][ MSE = 0.1800 | LCC = 0.9641 | SRCC = 0.9646 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1953 | LCC = 0.9628 | SRCC = 0.9637 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run3/config.gin b/results_tencent/w2v2_large_asr_zh/layer4/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run3/model.pt b/results_tencent/w2v2_large_asr_zh/layer4/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5ccb6ca2148d7fe4f2141c6c4420816071712148 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4fc70dcf0885d52b904a887995161777b4cbad587fe124b53b1c4877f5f9d612 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run3/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer4/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9b103e1dc11f7055e7cf19495a1b60e9e7907159 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:21dbf8fbdb3f83b46d17d00465c4db75139373873b21c998db383b75a1b00bfa +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run3/train.log b/results_tencent/w2v2_large_asr_zh/layer4/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c1cbf2186633759d2d8d24cbbcbc5b3bffb7e948 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42968086087703705 +INFO:root: +[Valid][0][UTT][ MSE = 0.2876 | LCC = 0.9242 | SRCC = 0.9256 ] +INFO:root:Average loss=0.19687780517339706 +INFO:root: +[Valid][1][UTT][ MSE = 0.2445 | LCC = 0.9385 | SRCC = 0.9397 ] +INFO:root:Average loss=0.1644321881532669 +INFO:root: +[Valid][2][UTT][ MSE = 0.2169 | LCC = 0.9447 | SRCC = 0.9463 ] +INFO:root:Average loss=0.14549614429473878 +INFO:root: +[Valid][3][UTT][ MSE = 0.2077 | LCC = 0.9501 | SRCC = 0.9515 ] +INFO:root:Average loss=0.12840441274642944 +INFO:root: +[Valid][4][UTT][ MSE = 0.2286 | LCC = 0.9556 | SRCC = 0.9570 ] +INFO:root:Average loss=0.1168073292374611 +INFO:root: +[Valid][5][UTT][ MSE = 0.1928 | LCC = 0.9572 | SRCC = 0.9585 ] +INFO:root:Average loss=0.11150916436314583 +INFO:root: +[Valid][6][UTT][ MSE = 0.2059 | LCC = 0.9599 | SRCC = 0.9609 ] +INFO:root:Average loss=0.09926950109004974 +INFO:root: +[Valid][7][UTT][ MSE = 0.1981 | LCC = 0.9602 | SRCC = 0.9606 ] +INFO:root:Average loss=0.09323716256022453 +INFO:root: +[Valid][8][UTT][ MSE = 0.1784 | LCC = 0.9614 | SRCC = 0.9620 ] +INFO:root:Average loss=0.08609352880716324 +INFO:root: +[Valid][9][UTT][ MSE = 0.1854 | LCC = 0.9629 | SRCC = 0.9635 ] +INFO:root:Average loss=0.08623604106903077 +INFO:root: +[Valid][10][UTT][ MSE = 0.1965 | LCC = 0.9611 | SRCC = 0.9615 ] +INFO:root:Average loss=0.07979818049073219 +INFO:root: +[Valid][11][UTT][ MSE = 0.2006 | LCC = 0.9620 | SRCC = 0.9623 ] +INFO:root:Average loss=0.07435156485438346 +INFO:root: +[Valid][12][UTT][ MSE = 0.1707 | LCC = 0.9626 | SRCC = 0.9629 ] +INFO:root:Average loss=0.06719306457042694 +INFO:root: +[Valid][13][UTT][ MSE = 0.2195 | LCC = 0.9621 | SRCC = 0.9639 ] +INFO:root:Average loss=0.066817069709301 +INFO:root: +[Valid][14][UTT][ MSE = 0.1757 | LCC = 0.9634 | SRCC = 0.9637 ] +INFO:root:Average loss=0.06196819368004799 +INFO:root: +[Valid][15][UTT][ MSE = 0.1693 | LCC = 0.9643 | SRCC = 0.9644 ] +INFO:root:Average loss=0.06005227303504944 +INFO:root: +[Valid][16][UTT][ MSE = 0.1803 | LCC = 0.9640 | SRCC = 0.9639 ] +INFO:root:Average loss=0.056905188605189325 +INFO:root: +[Valid][17][UTT][ MSE = 0.1790 | LCC = 0.9635 | SRCC = 0.9643 ] +INFO:root:Average loss=0.053996617883443836 +INFO:root: +[Valid][18][UTT][ MSE = 0.1864 | LCC = 0.9612 | SRCC = 0.9622 ] +INFO:root:Average loss=0.052521275326609614 +INFO:root: +[Valid][19][UTT][ MSE = 0.1699 | LCC = 0.9625 | SRCC = 0.9629 ] +INFO:root:Average loss=0.04899167729914188 +INFO:root: +[Valid][20][UTT][ MSE = 0.1797 | LCC = 0.9636 | SRCC = 0.9639 ] +INFO:root:Average loss=0.04684244357049465 +INFO:root: +[Valid][21][UTT][ MSE = 0.1692 | LCC = 0.9640 | SRCC = 0.9641 ] +INFO:root:Average loss=0.04832861177623272 +INFO:root: +[Valid][22][UTT][ MSE = 0.1722 | LCC = 0.9638 | SRCC = 0.9639 ] +INFO:root:Average loss=0.04364089412987232 +INFO:root: +[Valid][23][UTT][ MSE = 0.1649 | LCC = 0.9646 | SRCC = 0.9647 ] +INFO:root:Average loss=0.04261976550519466 +INFO:root: +[Valid][24][UTT][ MSE = 0.1819 | LCC = 0.9617 | SRCC = 0.9628 ] +INFO:root:Average loss=0.04105141884088516 +INFO:root: +[Valid][25][UTT][ MSE = 0.2002 | LCC = 0.9620 | SRCC = 0.9621 ] +INFO:root:Average loss=0.03997696042060852 +INFO:root: +[Valid][26][UTT][ MSE = 0.1816 | LCC = 0.9635 | SRCC = 0.9637 ] +INFO:root:Average loss=0.038016442060470584 +INFO:root: +[Valid][27][UTT][ MSE = 0.1728 | LCC = 0.9635 | SRCC = 0.9639 ] +INFO:root:Average loss=0.03673151823878288 +INFO:root: +[Valid][28][UTT][ MSE = 0.1689 | LCC = 0.9638 | SRCC = 0.9637 ] +INFO:root:Average loss=0.03523783655464649 +INFO:root: +[Valid][29][UTT][ MSE = 0.1588 | LCC = 0.9645 | SRCC = 0.9643 ] +INFO:root:Average loss=0.03470280539989472 +INFO:root: +[Valid][30][UTT][ MSE = 0.1608 | LCC = 0.9632 | SRCC = 0.9637 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1740 | LCC = 0.9602 | SRCC = 0.9609 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run4/config.gin b/results_tencent/w2v2_large_asr_zh/layer4/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run4/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer4/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f55125717795753feccabcefbbad11084ce066e1 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:311c21a767abd0bb69fc13a6b9d83cb1d13f8142f1e50c626b5c9e819552f4af +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer4/run4/train.log b/results_tencent/w2v2_large_asr_zh/layer4/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..02a8def7e1f0018eea833bb277e3f1a4d8fbd9b8 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer4/run4/train.log @@ -0,0 +1,95 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3822605036497116 +INFO:root: +[Valid][0][UTT][ MSE = 0.3011 | LCC = 0.9274 | SRCC = 0.9296 ] +INFO:root:Average loss=0.20790721261501313 +INFO:root: +[Valid][1][UTT][ MSE = 0.2757 | LCC = 0.9388 | SRCC = 0.9407 ] +INFO:root:Average loss=0.17119460988044738 +INFO:root: +[Valid][2][UTT][ MSE = 0.2368 | LCC = 0.9479 | SRCC = 0.9493 ] +INFO:root:Average loss=0.14581843560934066 +INFO:root: +[Valid][3][UTT][ MSE = 0.1746 | LCC = 0.9518 | SRCC = 0.9540 ] +INFO:root:Average loss=0.1304865774512291 +INFO:root: +[Valid][4][UTT][ MSE = 0.2391 | LCC = 0.9519 | SRCC = 0.9554 ] +INFO:root:Average loss=0.11987171798944474 +INFO:root: +[Valid][5][UTT][ MSE = 0.2241 | LCC = 0.9559 | SRCC = 0.9577 ] +INFO:root:Average loss=0.10539954066276551 +INFO:root: +[Valid][6][UTT][ MSE = 0.1948 | LCC = 0.9567 | SRCC = 0.9592 ] +INFO:root:Average loss=0.1013393726348877 +INFO:root: +[Valid][7][UTT][ MSE = 0.2082 | LCC = 0.9540 | SRCC = 0.9571 ] +INFO:root:Average loss=0.09760866659879684 +INFO:root: +[Valid][8][UTT][ MSE = 0.2098 | LCC = 0.9613 | SRCC = 0.9623 ] +INFO:root:Average loss=0.09071674072742462 +INFO:root: +[Valid][9][UTT][ MSE = 0.1893 | LCC = 0.9599 | SRCC = 0.9624 ] +INFO:root:Average loss=0.0864864268898964 +INFO:root: +[Valid][10][UTT][ MSE = 0.2464 | LCC = 0.9592 | SRCC = 0.9616 ] +INFO:root:Average loss=0.07980864155292511 +INFO:root: +[Valid][11][UTT][ MSE = 0.2142 | LCC = 0.9605 | SRCC = 0.9602 ] +INFO:root:Average loss=0.0760100708603859 +INFO:root: +[Valid][12][UTT][ MSE = 0.1775 | LCC = 0.9615 | SRCC = 0.9625 ] +INFO:root:Average loss=0.07033617961406707 +INFO:root: +[Valid][13][UTT][ MSE = 0.1990 | LCC = 0.9629 | SRCC = 0.9631 ] +INFO:root:Average loss=0.07077038703858853 +INFO:root: +[Valid][14][UTT][ MSE = 0.2001 | LCC = 0.9637 | SRCC = 0.9644 ] +INFO:root:Average loss=0.06461135172843933 +INFO:root: +[Valid][15][UTT][ MSE = 0.2145 | LCC = 0.9625 | SRCC = 0.9633 ] +INFO:root:Average loss=0.06050260657072067 +INFO:root: +[Valid][16][UTT][ MSE = 0.1550 | LCC = 0.9619 | SRCC = 0.9625 ] +INFO:root:Average loss=0.05966469129920006 +INFO:root: +[Valid][17][UTT][ MSE = 0.2047 | LCC = 0.9639 | SRCC = 0.9636 ] +INFO:root:Average loss=0.05391359049081802 +INFO:root: +[Valid][18][UTT][ MSE = 0.2039 | LCC = 0.9619 | SRCC = 0.9627 ] +INFO:root:Average loss=0.053831557810306546 +INFO:root: +[Valid][19][UTT][ MSE = 0.2065 | LCC = 0.9631 | SRCC = 0.9637 ] +INFO:root:Average loss=0.053799903705716134 +INFO:root: +[Valid][20][UTT][ MSE = 0.2142 | LCC = 0.9585 | SRCC = 0.9582 ] +INFO:root:Average loss=0.0494664055109024 +INFO:root: +[Valid][21][UTT][ MSE = 0.1735 | LCC = 0.9636 | SRCC = 0.9638 ] +INFO:root:Average loss=0.04716902570426464 +INFO:root: +[Valid][22][UTT][ MSE = 0.1986 | LCC = 0.9592 | SRCC = 0.9599 ] +INFO:root:Average loss=0.04430631305277347 +INFO:root: +[Valid][23][UTT][ MSE = 0.1870 | LCC = 0.9627 | SRCC = 0.9627 ] +INFO:root:Average loss=0.045120879754424095 +INFO:root: +[Valid][24][UTT][ MSE = 0.1878 | LCC = 0.9638 | SRCC = 0.9636 ] +INFO:root:Average loss=0.04382745546102524 +INFO:root: +[Valid][25][UTT][ MSE = 0.1880 | LCC = 0.9629 | SRCC = 0.9626 ] +INFO:root:Average loss=0.04046052248775959 +INFO:root: +[Valid][26][UTT][ MSE = 0.1963 | LCC = 0.9640 | SRCC = 0.9638 ] +INFO:root:Average loss=0.03966558988392353 +INFO:root: +[Valid][27][UTT][ MSE = 0.1882 | LCC = 0.9612 | SRCC = 0.9612 ] +INFO:root:Average loss=0.040354699596762654 +INFO:root: +[Valid][28][UTT][ MSE = 0.1850 | LCC = 0.9616 | SRCC = 0.9625 ] +INFO:root:Average loss=0.037696974575519564 +INFO:root: +[Valid][29][UTT][ MSE = 0.1772 | LCC = 0.9644 | SRCC = 0.9644 ] +INFO:root:Average loss=0.03671452908217907 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run0/config.gin b/results_tencent/w2v2_large_asr_zh/layer6/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run0/model.pt b/results_tencent/w2v2_large_asr_zh/layer6/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ec4b7dcd73a06a4a1515d2ebbf075671a5f64785 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:33a69f0201e152484ea2b26d3687824ebe3a5e3b9a253ecdd459ba3205330670 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run0/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer6/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b4ad695971f95b62fe91f18f66442133ce34a9e0 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:63d09ed3d495defacf812c9c2539676c5f7e5b7654dfb8ed4f1e37387cb2baa4 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run0/train.log b/results_tencent/w2v2_large_asr_zh/layer6/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..99416c00de47764e656ccc228f9e558abb8f2326 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46997705388069155 +INFO:root: +[Valid][0][UTT][ MSE = 0.3551 | LCC = 0.9176 | SRCC = 0.9185 ] +INFO:root:Average loss=0.21464821928739547 +INFO:root: +[Valid][1][UTT][ MSE = 0.3207 | LCC = 0.9340 | SRCC = 0.9357 ] +INFO:root:Average loss=0.15784029561281204 +INFO:root: +[Valid][2][UTT][ MSE = 0.2873 | LCC = 0.9495 | SRCC = 0.9513 ] +INFO:root:Average loss=0.13801561164855958 +INFO:root: +[Valid][3][UTT][ MSE = 0.2813 | LCC = 0.9527 | SRCC = 0.9550 ] +INFO:root:Average loss=0.12076298102736473 +INFO:root: +[Valid][4][UTT][ MSE = 0.2210 | LCC = 0.9550 | SRCC = 0.9565 ] +INFO:root:Average loss=0.10681923002004623 +INFO:root: +[Valid][5][UTT][ MSE = 0.2419 | LCC = 0.9563 | SRCC = 0.9583 ] +INFO:root:Average loss=0.10068125694990158 +INFO:root: +[Valid][6][UTT][ MSE = 0.2168 | LCC = 0.9595 | SRCC = 0.9614 ] +INFO:root:Average loss=0.0940373702943325 +INFO:root: +[Valid][7][UTT][ MSE = 0.2371 | LCC = 0.9610 | SRCC = 0.9629 ] +INFO:root:Average loss=0.08791943129897117 +INFO:root: +[Valid][8][UTT][ MSE = 0.2027 | LCC = 0.9634 | SRCC = 0.9642 ] +INFO:root:Average loss=0.07876252141594887 +INFO:root: +[Valid][9][UTT][ MSE = 0.2252 | LCC = 0.9621 | SRCC = 0.9629 ] +INFO:root:Average loss=0.07837297582626343 +INFO:root: +[Valid][10][UTT][ MSE = 0.2122 | LCC = 0.9623 | SRCC = 0.9639 ] +INFO:root:Average loss=0.07170398172736168 +INFO:root: +[Valid][11][UTT][ MSE = 0.2085 | LCC = 0.9633 | SRCC = 0.9641 ] +INFO:root:Average loss=0.06730459782481193 +INFO:root: +[Valid][12][UTT][ MSE = 0.2193 | LCC = 0.9651 | SRCC = 0.9656 ] +INFO:root:Average loss=0.06543908002972602 +INFO:root: +[Valid][13][UTT][ MSE = 0.2145 | LCC = 0.9627 | SRCC = 0.9641 ] +INFO:root:Average loss=0.06234961900115013 +INFO:root: +[Valid][14][UTT][ MSE = 0.1937 | LCC = 0.9653 | SRCC = 0.9655 ] +INFO:root:Average loss=0.05954968143999577 +INFO:root: +[Valid][15][UTT][ MSE = 0.2068 | LCC = 0.9642 | SRCC = 0.9648 ] +INFO:root:Average loss=0.05466793131828308 +INFO:root: +[Valid][16][UTT][ MSE = 0.1976 | LCC = 0.9638 | SRCC = 0.9650 ] +INFO:root:Average loss=0.05234649342298508 +INFO:root: +[Valid][17][UTT][ MSE = 0.1945 | LCC = 0.9655 | SRCC = 0.9661 ] +INFO:root:Average loss=0.05146839146316051 +INFO:root: +[Valid][18][UTT][ MSE = 0.2023 | LCC = 0.9662 | SRCC = 0.9668 ] +INFO:root:Average loss=0.05078490921854973 +INFO:root: +[Valid][19][UTT][ MSE = 0.1969 | LCC = 0.9653 | SRCC = 0.9659 ] +INFO:root:Average loss=0.04616724872589111 +INFO:root: +[Valid][20][UTT][ MSE = 0.2020 | LCC = 0.9645 | SRCC = 0.9650 ] +INFO:root:Average loss=0.0513097290545702 +INFO:root: +[Valid][21][UTT][ MSE = 0.2426 | LCC = 0.9654 | SRCC = 0.9671 ] +INFO:root:Average loss=0.04169975613057613 +INFO:root: +[Valid][22][UTT][ MSE = 0.2223 | LCC = 0.9655 | SRCC = 0.9668 ] +INFO:root:Average loss=0.04147836172580719 +INFO:root: +[Valid][23][UTT][ MSE = 0.1865 | LCC = 0.9667 | SRCC = 0.9669 ] +INFO:root:Average loss=0.04404013474285603 +INFO:root: +[Valid][24][UTT][ MSE = 0.1993 | LCC = 0.9671 | SRCC = 0.9669 ] +INFO:root:Average loss=0.03730929562449455 +INFO:root: +[Valid][25][UTT][ MSE = 0.2100 | LCC = 0.9673 | SRCC = 0.9676 ] +INFO:root:Average loss=0.03443968121707439 +INFO:root: +[Valid][26][UTT][ MSE = 0.2160 | LCC = 0.9655 | SRCC = 0.9657 ] +INFO:root:Average loss=0.03513623471558094 +INFO:root: +[Valid][27][UTT][ MSE = 0.1978 | LCC = 0.9654 | SRCC = 0.9647 ] +INFO:root:Average loss=0.035157064840197566 +INFO:root: +[Valid][28][UTT][ MSE = 0.2075 | LCC = 0.9657 | SRCC = 0.9666 ] +INFO:root:Average loss=0.03318715414404869 +INFO:root: +[Valid][29][UTT][ MSE = 0.1864 | LCC = 0.9670 | SRCC = 0.9665 ] +INFO:root:Average loss=0.033613938048481944 +INFO:root: +[Valid][30][UTT][ MSE = 0.1988 | LCC = 0.9672 | SRCC = 0.9673 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2190 | LCC = 0.9639 | SRCC = 0.9652 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run1/config.gin b/results_tencent/w2v2_large_asr_zh/layer6/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run1/model.pt b/results_tencent/w2v2_large_asr_zh/layer6/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..049b5e26775b3389dc3083030212e28428ca232f --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1be05615dd3b394dc05d644e522f6d5713a088239ea162e78b96a57462b991fa +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run1/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer6/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6c2e3916176a8bc7dba44f7985c08635b1ab9550 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eef8723692cdf113cfa69895a11172168f0309037009c0e8b706bf6cb10a6428 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run1/train.log b/results_tencent/w2v2_large_asr_zh/layer6/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3b9261ab4d40bc4ff36aad76f449f6e9b74b3d50 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44165295565128326 +INFO:root: +[Valid][0][UTT][ MSE = 0.2664 | LCC = 0.9231 | SRCC = 0.9235 ] +INFO:root:Average loss=0.20915446031093599 +INFO:root: +[Valid][1][UTT][ MSE = 0.2614 | LCC = 0.9414 | SRCC = 0.9419 ] +INFO:root:Average loss=0.17101287865638734 +INFO:root: +[Valid][2][UTT][ MSE = 0.2201 | LCC = 0.9488 | SRCC = 0.9500 ] +INFO:root:Average loss=0.14249089032411574 +INFO:root: +[Valid][3][UTT][ MSE = 0.1884 | LCC = 0.9510 | SRCC = 0.9525 ] +INFO:root:Average loss=0.12839415925741196 +INFO:root: +[Valid][4][UTT][ MSE = 0.2443 | LCC = 0.9561 | SRCC = 0.9567 ] +INFO:root:Average loss=0.11417144790291786 +INFO:root: +[Valid][5][UTT][ MSE = 0.1876 | LCC = 0.9583 | SRCC = 0.9593 ] +INFO:root:Average loss=0.10798488163948058 +INFO:root: +[Valid][6][UTT][ MSE = 0.2147 | LCC = 0.9589 | SRCC = 0.9596 ] +INFO:root:Average loss=0.09928062132000923 +INFO:root: +[Valid][7][UTT][ MSE = 0.1985 | LCC = 0.9590 | SRCC = 0.9602 ] +INFO:root:Average loss=0.08991031774878502 +INFO:root: +[Valid][8][UTT][ MSE = 0.1775 | LCC = 0.9621 | SRCC = 0.9625 ] +INFO:root:Average loss=0.08369644093513488 +INFO:root: +[Valid][9][UTT][ MSE = 0.2273 | LCC = 0.9623 | SRCC = 0.9629 ] +INFO:root:Average loss=0.08167951294779778 +INFO:root: +[Valid][10][UTT][ MSE = 0.2215 | LCC = 0.9559 | SRCC = 0.9592 ] +INFO:root:Average loss=0.07732427963614463 +INFO:root: +[Valid][11][UTT][ MSE = 0.2071 | LCC = 0.9622 | SRCC = 0.9627 ] +INFO:root:Average loss=0.07417945212125779 +INFO:root: +[Valid][12][UTT][ MSE = 0.2035 | LCC = 0.9625 | SRCC = 0.9637 ] +INFO:root:Average loss=0.07297048431634903 +INFO:root: +[Valid][13][UTT][ MSE = 0.1891 | LCC = 0.9614 | SRCC = 0.9630 ] +INFO:root:Average loss=0.062484087437391284 +INFO:root: +[Valid][14][UTT][ MSE = 0.1955 | LCC = 0.9635 | SRCC = 0.9653 ] +INFO:root:Average loss=0.058055303871631624 +INFO:root: +[Valid][15][UTT][ MSE = 0.1580 | LCC = 0.9653 | SRCC = 0.9652 ] +INFO:root:Average loss=0.05747188597917557 +INFO:root: +[Valid][16][UTT][ MSE = 0.1846 | LCC = 0.9640 | SRCC = 0.9641 ] +INFO:root:Average loss=0.05379266914725304 +INFO:root: +[Valid][17][UTT][ MSE = 0.1934 | LCC = 0.9630 | SRCC = 0.9634 ] +INFO:root:Average loss=0.05494304272532463 +INFO:root: +[Valid][18][UTT][ MSE = 0.1820 | LCC = 0.9617 | SRCC = 0.9636 ] +INFO:root:Average loss=0.05152850815653801 +INFO:root: +[Valid][19][UTT][ MSE = 0.1995 | LCC = 0.9630 | SRCC = 0.9638 ] +INFO:root:Average loss=0.04816105529665947 +INFO:root: +[Valid][20][UTT][ MSE = 0.1874 | LCC = 0.9614 | SRCC = 0.9618 ] +INFO:root:Average loss=0.0460609909594059 +INFO:root: +[Valid][21][UTT][ MSE = 0.1873 | LCC = 0.9640 | SRCC = 0.9643 ] +INFO:root:Average loss=0.042664244368672374 +INFO:root: +[Valid][22][UTT][ MSE = 0.1662 | LCC = 0.9637 | SRCC = 0.9639 ] +INFO:root:Average loss=0.04130357326567173 +INFO:root: +[Valid][23][UTT][ MSE = 0.1712 | LCC = 0.9637 | SRCC = 0.9641 ] +INFO:root:Average loss=0.03922898165881634 +INFO:root: +[Valid][24][UTT][ MSE = 0.1795 | LCC = 0.9631 | SRCC = 0.9638 ] +INFO:root:Average loss=0.039682214230298996 +INFO:root: +[Valid][25][UTT][ MSE = 0.1684 | LCC = 0.9654 | SRCC = 0.9653 ] +INFO:root:Average loss=0.03726142320036888 +INFO:root: +[Valid][26][UTT][ MSE = 0.1650 | LCC = 0.9628 | SRCC = 0.9629 ] +INFO:root:Average loss=0.03831013199687004 +INFO:root: +[Valid][27][UTT][ MSE = 0.1643 | LCC = 0.9647 | SRCC = 0.9642 ] +INFO:root:Average loss=0.03429539588093758 +INFO:root: +[Valid][28][UTT][ MSE = 0.1777 | LCC = 0.9640 | SRCC = 0.9641 ] +INFO:root:Average loss=0.03420083796977997 +INFO:root: +[Valid][29][UTT][ MSE = 0.1851 | LCC = 0.9658 | SRCC = 0.9658 ] +INFO:root:Average loss=0.0327231660336256 +INFO:root: +[Valid][30][UTT][ MSE = 0.1834 | LCC = 0.9626 | SRCC = 0.9634 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1921 | LCC = 0.9627 | SRCC = 0.9636 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run2/config.gin b/results_tencent/w2v2_large_asr_zh/layer6/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run2/model.pt b/results_tencent/w2v2_large_asr_zh/layer6/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..541541fd04f2e99a1ddbfc68835025c3ea90b31c --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:47e35ecf12611af945486351a4da9f4faacfbfb0f3f5bf57e8ebaa0fe0994b66 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run2/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer6/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0eede758382437e7b2f162fde9cf1664277838f8 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f32b689873f8dda2a8b30c6b9b8f6cb40d469bdc9a1811cb533afd1011edbafd +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run2/train.log b/results_tencent/w2v2_large_asr_zh/layer6/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1d9b2dd9e30f5dfc7c0f63613f339925619a2c91 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5509365005493164 +INFO:root: +[Valid][0][UTT][ MSE = 0.3087 | LCC = 0.9224 | SRCC = 0.9219 ] +INFO:root:Average loss=0.19527860110998155 +INFO:root: +[Valid][1][UTT][ MSE = 0.2675 | LCC = 0.9450 | SRCC = 0.9458 ] +INFO:root:Average loss=0.15866012126207352 +INFO:root: +[Valid][2][UTT][ MSE = 0.2707 | LCC = 0.9481 | SRCC = 0.9487 ] +INFO:root:Average loss=0.12714210709929466 +INFO:root: +[Valid][3][UTT][ MSE = 0.2659 | LCC = 0.9501 | SRCC = 0.9534 ] +INFO:root:Average loss=0.11531522366404533 +INFO:root: +[Valid][4][UTT][ MSE = 0.2688 | LCC = 0.9578 | SRCC = 0.9585 ] +INFO:root:Average loss=0.10388939094543458 +INFO:root: +[Valid][5][UTT][ MSE = 0.2275 | LCC = 0.9599 | SRCC = 0.9611 ] +INFO:root:Average loss=0.10098067906498909 +INFO:root: +[Valid][6][UTT][ MSE = 0.2118 | LCC = 0.9613 | SRCC = 0.9622 ] +INFO:root:Average loss=0.09394397148489952 +INFO:root: +[Valid][7][UTT][ MSE = 0.2387 | LCC = 0.9612 | SRCC = 0.9629 ] +INFO:root:Average loss=0.08734305337071419 +INFO:root: +[Valid][8][UTT][ MSE = 0.2071 | LCC = 0.9616 | SRCC = 0.9626 ] +INFO:root:Average loss=0.08229627785086632 +INFO:root: +[Valid][9][UTT][ MSE = 0.2667 | LCC = 0.9616 | SRCC = 0.9633 ] +INFO:root:Average loss=0.08087533423304558 +INFO:root: +[Valid][10][UTT][ MSE = 0.2168 | LCC = 0.9639 | SRCC = 0.9647 ] +INFO:root:Average loss=0.06989785787463189 +INFO:root: +[Valid][11][UTT][ MSE = 0.1911 | LCC = 0.9642 | SRCC = 0.9643 ] +INFO:root:Average loss=0.0678867725431919 +INFO:root: +[Valid][12][UTT][ MSE = 0.2026 | LCC = 0.9633 | SRCC = 0.9638 ] +INFO:root:Average loss=0.06257673099637032 +INFO:root: +[Valid][13][UTT][ MSE = 0.1921 | LCC = 0.9640 | SRCC = 0.9644 ] +INFO:root:Average loss=0.06523363426327705 +INFO:root: +[Valid][14][UTT][ MSE = 0.2132 | LCC = 0.9645 | SRCC = 0.9653 ] +INFO:root:Average loss=0.058340576604008675 +INFO:root: +[Valid][15][UTT][ MSE = 0.2072 | LCC = 0.9650 | SRCC = 0.9653 ] +INFO:root:Average loss=0.05695232152938843 +INFO:root: +[Valid][16][UTT][ MSE = 0.2129 | LCC = 0.9638 | SRCC = 0.9645 ] +INFO:root:Average loss=0.05149370673298836 +INFO:root: +[Valid][17][UTT][ MSE = 0.2301 | LCC = 0.9626 | SRCC = 0.9649 ] +INFO:root:Average loss=0.050906504705548285 +INFO:root: +[Valid][18][UTT][ MSE = 0.2198 | LCC = 0.9630 | SRCC = 0.9652 ] +INFO:root:Average loss=0.04863957804441452 +INFO:root: +[Valid][19][UTT][ MSE = 0.2038 | LCC = 0.9639 | SRCC = 0.9641 ] +INFO:root:Average loss=0.04693237833678723 +INFO:root: +[Valid][20][UTT][ MSE = 0.2102 | LCC = 0.9660 | SRCC = 0.9662 ] +INFO:root:Average loss=0.04566435036063194 +INFO:root: +[Valid][21][UTT][ MSE = 0.1970 | LCC = 0.9649 | SRCC = 0.9653 ] +INFO:root:Average loss=0.041751346781849864 +INFO:root: +[Valid][22][UTT][ MSE = 0.1764 | LCC = 0.9646 | SRCC = 0.9643 ] +INFO:root:Average loss=0.041696571767330166 +INFO:root: +[Valid][23][UTT][ MSE = 0.1979 | LCC = 0.9641 | SRCC = 0.9639 ] +INFO:root:Average loss=0.03983059670031071 +INFO:root: +[Valid][24][UTT][ MSE = 0.2093 | LCC = 0.9650 | SRCC = 0.9654 ] +INFO:root:Average loss=0.03792686876654625 +INFO:root: +[Valid][25][UTT][ MSE = 0.2083 | LCC = 0.9648 | SRCC = 0.9653 ] +INFO:root:Average loss=0.03692148046195507 +INFO:root: +[Valid][26][UTT][ MSE = 0.2042 | LCC = 0.9657 | SRCC = 0.9661 ] +INFO:root:Average loss=0.03470493367314339 +INFO:root: +[Valid][27][UTT][ MSE = 0.1866 | LCC = 0.9647 | SRCC = 0.9642 ] +INFO:root:Average loss=0.033991225361824036 +INFO:root: +[Valid][28][UTT][ MSE = 0.2050 | LCC = 0.9644 | SRCC = 0.9649 ] +INFO:root:Average loss=0.03103117652237415 +INFO:root: +[Valid][29][UTT][ MSE = 0.1927 | LCC = 0.9650 | SRCC = 0.9650 ] +INFO:root:Average loss=0.03155846448242664 +INFO:root: +[Valid][30][UTT][ MSE = 0.1899 | LCC = 0.9654 | SRCC = 0.9653 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2264 | LCC = 0.9600 | SRCC = 0.9618 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run3/config.gin b/results_tencent/w2v2_large_asr_zh/layer6/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run3/model.pt b/results_tencent/w2v2_large_asr_zh/layer6/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..928b0d3183d562587a4ce9a0064ec99a5148b3ec --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6ec5beaeb1e7a88498f103485b3145c4868cd29a5925a45eb4d5227afbdc3ee1 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run3/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer6/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c585bb1a4bb63e55971bbc8f26777976b476134c --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:07ed0ce53936fdfdc50319e04c6c5ad0adcff468e037066f8d505bb2d01ad7d2 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run3/train.log b/results_tencent/w2v2_large_asr_zh/layer6/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6bbfe06824868bf37fa9a41c4b835502fc7bcfa5 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4465181201696396 +INFO:root: +[Valid][0][UTT][ MSE = 0.2650 | LCC = 0.9262 | SRCC = 0.9275 ] +INFO:root:Average loss=0.19305557376146318 +INFO:root: +[Valid][1][UTT][ MSE = 0.2863 | LCC = 0.9378 | SRCC = 0.9399 ] +INFO:root:Average loss=0.16050119709968566 +INFO:root: +[Valid][2][UTT][ MSE = 0.2415 | LCC = 0.9496 | SRCC = 0.9505 ] +INFO:root:Average loss=0.13735432690382005 +INFO:root: +[Valid][3][UTT][ MSE = 0.2209 | LCC = 0.9530 | SRCC = 0.9547 ] +INFO:root:Average loss=0.12282881468534469 +INFO:root: +[Valid][4][UTT][ MSE = 0.2076 | LCC = 0.9567 | SRCC = 0.9583 ] +INFO:root:Average loss=0.10944483596086502 +INFO:root: +[Valid][5][UTT][ MSE = 0.2567 | LCC = 0.9557 | SRCC = 0.9594 ] +INFO:root:Average loss=0.10215889298915863 +INFO:root: +[Valid][6][UTT][ MSE = 0.2237 | LCC = 0.9589 | SRCC = 0.9599 ] +INFO:root:Average loss=0.09727239200472831 +INFO:root: +[Valid][7][UTT][ MSE = 0.2119 | LCC = 0.9617 | SRCC = 0.9622 ] +INFO:root:Average loss=0.09002619487047195 +INFO:root: +[Valid][8][UTT][ MSE = 0.2019 | LCC = 0.9614 | SRCC = 0.9622 ] +INFO:root:Average loss=0.08662121194601059 +INFO:root: +[Valid][9][UTT][ MSE = 0.1917 | LCC = 0.9623 | SRCC = 0.9628 ] +INFO:root:Average loss=0.07991273403167724 +INFO:root: +[Valid][10][UTT][ MSE = 0.1809 | LCC = 0.9626 | SRCC = 0.9627 ] +INFO:root:Average loss=0.07446411919593811 +INFO:root: +[Valid][11][UTT][ MSE = 0.2099 | LCC = 0.9641 | SRCC = 0.9647 ] +INFO:root:Average loss=0.07181980448961257 +INFO:root: +[Valid][12][UTT][ MSE = 0.2106 | LCC = 0.9645 | SRCC = 0.9650 ] +INFO:root:Average loss=0.06730194482207298 +INFO:root: +[Valid][13][UTT][ MSE = 0.1971 | LCC = 0.9633 | SRCC = 0.9635 ] +INFO:root:Average loss=0.0647856606245041 +INFO:root: +[Valid][14][UTT][ MSE = 0.2130 | LCC = 0.9641 | SRCC = 0.9651 ] +INFO:root:Average loss=0.06017742556333542 +INFO:root: +[Valid][15][UTT][ MSE = 0.1769 | LCC = 0.9637 | SRCC = 0.9643 ] +INFO:root:Average loss=0.05692466643452644 +INFO:root: +[Valid][16][UTT][ MSE = 0.1975 | LCC = 0.9617 | SRCC = 0.9630 ] +INFO:root:Average loss=0.05344939991831779 +INFO:root: +[Valid][17][UTT][ MSE = 0.1678 | LCC = 0.9647 | SRCC = 0.9650 ] +INFO:root:Average loss=0.05208696317672729 +INFO:root: +[Valid][18][UTT][ MSE = 0.1913 | LCC = 0.9643 | SRCC = 0.9647 ] +INFO:root:Average loss=0.049777655258774756 +INFO:root: +[Valid][19][UTT][ MSE = 0.1748 | LCC = 0.9652 | SRCC = 0.9648 ] +INFO:root:Average loss=0.0460881065428257 +INFO:root: +[Valid][20][UTT][ MSE = 0.2254 | LCC = 0.9620 | SRCC = 0.9637 ] +INFO:root:Average loss=0.0515649426728487 +INFO:root: +[Valid][21][UTT][ MSE = 0.2083 | LCC = 0.9653 | SRCC = 0.9656 ] +INFO:root:Average loss=0.044025276735424994 +INFO:root: +[Valid][22][UTT][ MSE = 0.1709 | LCC = 0.9652 | SRCC = 0.9653 ] +INFO:root:Average loss=0.0411250873953104 +INFO:root: +[Valid][23][UTT][ MSE = 0.2009 | LCC = 0.9651 | SRCC = 0.9651 ] +INFO:root:Average loss=0.04177295686304569 +INFO:root: +[Valid][24][UTT][ MSE = 0.1822 | LCC = 0.9663 | SRCC = 0.9656 ] +INFO:root:Average loss=0.039300519838929175 +INFO:root: +[Valid][25][UTT][ MSE = 0.2064 | LCC = 0.9653 | SRCC = 0.9651 ] +INFO:root:Average loss=0.03706240986287594 +INFO:root: +[Valid][26][UTT][ MSE = 0.2035 | LCC = 0.9646 | SRCC = 0.9651 ] +INFO:root:Average loss=0.038822579964995385 +INFO:root: +[Valid][27][UTT][ MSE = 0.1922 | LCC = 0.9653 | SRCC = 0.9653 ] +INFO:root:Average loss=0.035678254917263987 +INFO:root: +[Valid][28][UTT][ MSE = 0.1824 | LCC = 0.9654 | SRCC = 0.9647 ] +INFO:root:Average loss=0.03234036664664745 +INFO:root: +[Valid][29][UTT][ MSE = 0.1779 | LCC = 0.9652 | SRCC = 0.9650 ] +INFO:root:Average loss=0.031237986877560617 +INFO:root: +[Valid][30][UTT][ MSE = 0.2122 | LCC = 0.9650 | SRCC = 0.9648 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1984 | LCC = 0.9600 | SRCC = 0.9611 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run4/config.gin b/results_tencent/w2v2_large_asr_zh/layer6/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run4/model.pt b/results_tencent/w2v2_large_asr_zh/layer6/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5f5e0c9e8b1a411f21b07ccf3767988225fc7e3a --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:64d0fc2d64006e7486ffad2279a9a5a76877c2b25a18cd004ae638904b4cfb69 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run4/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer6/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c5367d7dac36f6ea6929fd3d4d4ea8b0ede3209c --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:682daf865247d89b75401c84f524a0ce5932ff68d15c375fcdf2d63f6ceb5308 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer6/run4/train.log b/results_tencent/w2v2_large_asr_zh/layer6/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..49ac5a6904e08c681ffeea7a3383936fba99a5aa --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer6/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5010172157287598 +INFO:root: +[Valid][0][UTT][ MSE = 0.3282 | LCC = 0.9208 | SRCC = 0.9210 ] +INFO:root:Average loss=0.19930376774072647 +INFO:root: +[Valid][1][UTT][ MSE = 0.2451 | LCC = 0.9416 | SRCC = 0.9439 ] +INFO:root:Average loss=0.15182938826084136 +INFO:root: +[Valid][2][UTT][ MSE = 0.2707 | LCC = 0.9512 | SRCC = 0.9536 ] +INFO:root:Average loss=0.12967975044250488 +INFO:root: +[Valid][3][UTT][ MSE = 0.2379 | LCC = 0.9541 | SRCC = 0.9570 ] +INFO:root:Average loss=0.11866761142015457 +INFO:root: +[Valid][4][UTT][ MSE = 0.1785 | LCC = 0.9577 | SRCC = 0.9596 ] +INFO:root:Average loss=0.1083133562207222 +INFO:root: +[Valid][5][UTT][ MSE = 0.2181 | LCC = 0.9594 | SRCC = 0.9612 ] +INFO:root:Average loss=0.09764782670140267 +INFO:root: +[Valid][6][UTT][ MSE = 0.2238 | LCC = 0.9620 | SRCC = 0.9633 ] +INFO:root:Average loss=0.09451647138595581 +INFO:root: +[Valid][7][UTT][ MSE = 0.1876 | LCC = 0.9621 | SRCC = 0.9626 ] +INFO:root:Average loss=0.08792094340920448 +INFO:root: +[Valid][8][UTT][ MSE = 0.1842 | LCC = 0.9624 | SRCC = 0.9630 ] +INFO:root:Average loss=0.08312950742244721 +INFO:root: +[Valid][9][UTT][ MSE = 0.2056 | LCC = 0.9559 | SRCC = 0.9572 ] +INFO:root:Average loss=0.07717286732792854 +INFO:root: +[Valid][10][UTT][ MSE = 0.1823 | LCC = 0.9604 | SRCC = 0.9618 ] +INFO:root:Average loss=0.07464797487854957 +INFO:root: +[Valid][11][UTT][ MSE = 0.1882 | LCC = 0.9639 | SRCC = 0.9639 ] +INFO:root:Average loss=0.06913211280107498 +INFO:root: +[Valid][12][UTT][ MSE = 0.1949 | LCC = 0.9617 | SRCC = 0.9630 ] +INFO:root:Average loss=0.0671195158958435 +INFO:root: +[Valid][13][UTT][ MSE = 0.1878 | LCC = 0.9647 | SRCC = 0.9649 ] +INFO:root:Average loss=0.059940279245376586 +INFO:root: +[Valid][14][UTT][ MSE = 0.1741 | LCC = 0.9644 | SRCC = 0.9648 ] +INFO:root:Average loss=0.05683079069852829 +INFO:root: +[Valid][15][UTT][ MSE = 0.1804 | LCC = 0.9637 | SRCC = 0.9644 ] +INFO:root:Average loss=0.05561216548085213 +INFO:root: +[Valid][16][UTT][ MSE = 0.1843 | LCC = 0.9647 | SRCC = 0.9646 ] +INFO:root:Average loss=0.05653582146763802 +INFO:root: +[Valid][17][UTT][ MSE = 0.1842 | LCC = 0.9656 | SRCC = 0.9656 ] +INFO:root:Average loss=0.05048408168554306 +INFO:root: +[Valid][18][UTT][ MSE = 0.2021 | LCC = 0.9645 | SRCC = 0.9648 ] +INFO:root:Average loss=0.048808831870555874 +INFO:root: +[Valid][19][UTT][ MSE = 0.1708 | LCC = 0.9646 | SRCC = 0.9648 ] +INFO:root:Average loss=0.04870910240709782 +INFO:root: +[Valid][20][UTT][ MSE = 0.1782 | LCC = 0.9649 | SRCC = 0.9649 ] +INFO:root:Average loss=0.047378852978348734 +INFO:root: +[Valid][21][UTT][ MSE = 0.1927 | LCC = 0.9633 | SRCC = 0.9635 ] +INFO:root:Average loss=0.04077916073799133 +INFO:root: +[Valid][22][UTT][ MSE = 0.1817 | LCC = 0.9652 | SRCC = 0.9652 ] +INFO:root:Average loss=0.038992570400238034 +INFO:root: +[Valid][23][UTT][ MSE = 0.1794 | LCC = 0.9655 | SRCC = 0.9652 ] +INFO:root:Average loss=0.0416384946256876 +INFO:root: +[Valid][24][UTT][ MSE = 0.1922 | LCC = 0.9652 | SRCC = 0.9656 ] +INFO:root:Average loss=0.039458199858665464 +INFO:root: +[Valid][25][UTT][ MSE = 0.1905 | LCC = 0.9658 | SRCC = 0.9655 ] +INFO:root:Average loss=0.03778443178534508 +INFO:root: +[Valid][26][UTT][ MSE = 0.1776 | LCC = 0.9658 | SRCC = 0.9656 ] +INFO:root:Average loss=0.03413491803407669 +INFO:root: +[Valid][27][UTT][ MSE = 0.1791 | LCC = 0.9656 | SRCC = 0.9652 ] +INFO:root:Average loss=0.03258153565227986 +INFO:root: +[Valid][28][UTT][ MSE = 0.1768 | LCC = 0.9643 | SRCC = 0.9642 ] +INFO:root:Average loss=0.03214940628409386 +INFO:root: +[Valid][29][UTT][ MSE = 0.1793 | LCC = 0.9650 | SRCC = 0.9643 ] +INFO:root:Average loss=0.03113426822423935 +INFO:root: +[Valid][30][UTT][ MSE = 0.1899 | LCC = 0.9654 | SRCC = 0.9654 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2036 | LCC = 0.9598 | SRCC = 0.9606 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run0/config.gin b/results_tencent/w2v2_large_asr_zh/layer8/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run0/model.pt b/results_tencent/w2v2_large_asr_zh/layer8/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9fbccdb536f30abf3a5c07c4c7c16f9f2e0f6a2f --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c1f01e98a44f24993878d6bf2d7334f4b45399d661d5578c0e1ce180b073016b +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run0/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer8/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c896dd9c4025439147d8c795eaa43129fe3b2df6 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:97ef18cdb66b74ee69a1f5af1c828f4e554bbf23947238470087cf7f47596240 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run0/train.log b/results_tencent/w2v2_large_asr_zh/layer8/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c904bc16a17fa3b139f4494125d50e2943ed52fb --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46138367831707 +INFO:root: +[Valid][0][UTT][ MSE = 0.3424 | LCC = 0.9177 | SRCC = 0.9168 ] +INFO:root:Average loss=0.20701831638813017 +INFO:root: +[Valid][1][UTT][ MSE = 0.2591 | LCC = 0.9415 | SRCC = 0.9416 ] +INFO:root:Average loss=0.15760618495941162 +INFO:root: +[Valid][2][UTT][ MSE = 0.2538 | LCC = 0.9493 | SRCC = 0.9494 ] +INFO:root:Average loss=0.1390376667380333 +INFO:root: +[Valid][3][UTT][ MSE = 0.2282 | LCC = 0.9553 | SRCC = 0.9555 ] +INFO:root:Average loss=0.12109712082147599 +INFO:root: +[Valid][4][UTT][ MSE = 0.2610 | LCC = 0.9543 | SRCC = 0.9553 ] +INFO:root:Average loss=0.11492239779233933 +INFO:root: +[Valid][5][UTT][ MSE = 0.2162 | LCC = 0.9579 | SRCC = 0.9591 ] +INFO:root:Average loss=0.10268074586987495 +INFO:root: +[Valid][6][UTT][ MSE = 0.2119 | LCC = 0.9592 | SRCC = 0.9597 ] +INFO:root:Average loss=0.09742216116189957 +INFO:root: +[Valid][7][UTT][ MSE = 0.2318 | LCC = 0.9603 | SRCC = 0.9606 ] +INFO:root:Average loss=0.09157710289955139 +INFO:root: +[Valid][8][UTT][ MSE = 0.2252 | LCC = 0.9620 | SRCC = 0.9629 ] +INFO:root:Average loss=0.08422402414679528 +INFO:root: +[Valid][9][UTT][ MSE = 0.1924 | LCC = 0.9636 | SRCC = 0.9638 ] +INFO:root:Average loss=0.08265314891934394 +INFO:root: +[Valid][10][UTT][ MSE = 0.2088 | LCC = 0.9638 | SRCC = 0.9632 ] +INFO:root:Average loss=0.0775372671186924 +INFO:root: +[Valid][11][UTT][ MSE = 0.2337 | LCC = 0.9614 | SRCC = 0.9634 ] +INFO:root:Average loss=0.07205580985546112 +INFO:root: +[Valid][12][UTT][ MSE = 0.1971 | LCC = 0.9645 | SRCC = 0.9644 ] +INFO:root:Average loss=0.06830484089255333 +INFO:root: +[Valid][13][UTT][ MSE = 0.1888 | LCC = 0.9645 | SRCC = 0.9646 ] +INFO:root:Average loss=0.06326693525910378 +INFO:root: +[Valid][14][UTT][ MSE = 0.1615 | LCC = 0.9620 | SRCC = 0.9627 ] +INFO:root:Average loss=0.06046172082424164 +INFO:root: +[Valid][15][UTT][ MSE = 0.2000 | LCC = 0.9630 | SRCC = 0.9631 ] +INFO:root:Average loss=0.057687247395515444 +INFO:root: +[Valid][16][UTT][ MSE = 0.1996 | LCC = 0.9640 | SRCC = 0.9647 ] +INFO:root:Average loss=0.05693952426314354 +INFO:root: +[Valid][17][UTT][ MSE = 0.1952 | LCC = 0.9621 | SRCC = 0.9627 ] +INFO:root:Average loss=0.05214262722432613 +INFO:root: +[Valid][18][UTT][ MSE = 0.1863 | LCC = 0.9643 | SRCC = 0.9645 ] +INFO:root:Average loss=0.0539250063598156 +INFO:root: +[Valid][19][UTT][ MSE = 0.1957 | LCC = 0.9646 | SRCC = 0.9646 ] +INFO:root:Average loss=0.04797383730113507 +INFO:root: +[Valid][20][UTT][ MSE = 0.1964 | LCC = 0.9655 | SRCC = 0.9652 ] +INFO:root:Average loss=0.0456393369436264 +INFO:root: +[Valid][21][UTT][ MSE = 0.1731 | LCC = 0.9643 | SRCC = 0.9642 ] +INFO:root:Average loss=0.04533799628913403 +INFO:root: +[Valid][22][UTT][ MSE = 0.1994 | LCC = 0.9648 | SRCC = 0.9654 ] +INFO:root:Average loss=0.04178673611581325 +INFO:root: +[Valid][23][UTT][ MSE = 0.1662 | LCC = 0.9653 | SRCC = 0.9653 ] +INFO:root:Average loss=0.0420809670239687 +INFO:root: +[Valid][24][UTT][ MSE = 0.1890 | LCC = 0.9649 | SRCC = 0.9650 ] +INFO:root:Average loss=0.040371906034648415 +INFO:root: +[Valid][25][UTT][ MSE = 0.1769 | LCC = 0.9618 | SRCC = 0.9636 ] +INFO:root:Average loss=0.04116623093187809 +INFO:root: +[Valid][26][UTT][ MSE = 0.2110 | LCC = 0.9640 | SRCC = 0.9647 ] +INFO:root:Average loss=0.03861736960709095 +INFO:root: +[Valid][27][UTT][ MSE = 0.1891 | LCC = 0.9648 | SRCC = 0.9649 ] +INFO:root:Average loss=0.0356969318985939 +INFO:root: +[Valid][28][UTT][ MSE = 0.1901 | LCC = 0.9651 | SRCC = 0.9650 ] +INFO:root:Average loss=0.034693168625235554 +INFO:root: +[Valid][29][UTT][ MSE = 0.1945 | LCC = 0.9646 | SRCC = 0.9649 ] +INFO:root:Average loss=0.03376957568526268 +INFO:root: +[Valid][30][UTT][ MSE = 0.1731 | LCC = 0.9652 | SRCC = 0.9654 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2147 | LCC = 0.9606 | SRCC = 0.9617 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run1/config.gin b/results_tencent/w2v2_large_asr_zh/layer8/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run1/model.pt b/results_tencent/w2v2_large_asr_zh/layer8/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0e0331ebbec7ddb5ba684980f495471a1a21da89 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4910176ee26da8aae7159a98ad5148afdb9c8dc3e75213e15c7e5682ae989ac3 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run1/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer8/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..7a824ac69725b503838c6ba6aa262e3b6aa44347 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:76804e051b9236a366fd5095b6cb930658e30120c2aa02daf500d6e515f3337e +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run1/train.log b/results_tencent/w2v2_large_asr_zh/layer8/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..64eea42d3a7c62c2beb5425e7c69b0d0fa4700da --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4496051656007767 +INFO:root: +[Valid][0][UTT][ MSE = 0.3389 | LCC = 0.9174 | SRCC = 0.9172 ] +INFO:root:Average loss=0.20654586750268936 +INFO:root: +[Valid][1][UTT][ MSE = 0.2696 | LCC = 0.9401 | SRCC = 0.9402 ] +INFO:root:Average loss=0.1582761858701706 +INFO:root: +[Valid][2][UTT][ MSE = 0.2469 | LCC = 0.9500 | SRCC = 0.9502 ] +INFO:root:Average loss=0.1374310314655304 +INFO:root: +[Valid][3][UTT][ MSE = 0.2632 | LCC = 0.9524 | SRCC = 0.9527 ] +INFO:root:Average loss=0.12517544519901275 +INFO:root: +[Valid][4][UTT][ MSE = 0.2027 | LCC = 0.9551 | SRCC = 0.9550 ] +INFO:root:Average loss=0.11111151161789894 +INFO:root: +[Valid][5][UTT][ MSE = 0.2091 | LCC = 0.9558 | SRCC = 0.9561 ] +INFO:root:Average loss=0.10250683417916298 +INFO:root: +[Valid][6][UTT][ MSE = 0.1949 | LCC = 0.9579 | SRCC = 0.9585 ] +INFO:root:Average loss=0.09843589195609093 +INFO:root: +[Valid][7][UTT][ MSE = 0.2314 | LCC = 0.9596 | SRCC = 0.9601 ] +INFO:root:Average loss=0.08641524282097816 +INFO:root: +[Valid][8][UTT][ MSE = 0.2262 | LCC = 0.9586 | SRCC = 0.9600 ] +INFO:root:Average loss=0.08083641356229782 +INFO:root: +[Valid][9][UTT][ MSE = 0.1985 | LCC = 0.9608 | SRCC = 0.9612 ] +INFO:root:Average loss=0.07655464324355125 +INFO:root: +[Valid][10][UTT][ MSE = 0.2125 | LCC = 0.9606 | SRCC = 0.9608 ] +INFO:root:Average loss=0.07297244080901145 +INFO:root: +[Valid][11][UTT][ MSE = 0.1957 | LCC = 0.9623 | SRCC = 0.9624 ] +INFO:root:Average loss=0.07180207020044327 +INFO:root: +[Valid][12][UTT][ MSE = 0.1919 | LCC = 0.9621 | SRCC = 0.9616 ] +INFO:root:Average loss=0.06916336116194725 +INFO:root: +[Valid][13][UTT][ MSE = 0.1794 | LCC = 0.9614 | SRCC = 0.9612 ] +INFO:root:Average loss=0.06463627254962921 +INFO:root: +[Valid][14][UTT][ MSE = 0.1853 | LCC = 0.9634 | SRCC = 0.9633 ] +INFO:root:Average loss=0.06019887875020504 +INFO:root: +[Valid][15][UTT][ MSE = 0.1892 | LCC = 0.9640 | SRCC = 0.9636 ] +INFO:root:Average loss=0.0579149899482727 +INFO:root: +[Valid][16][UTT][ MSE = 0.1892 | LCC = 0.9618 | SRCC = 0.9617 ] +INFO:root:Average loss=0.05416258892416954 +INFO:root: +[Valid][17][UTT][ MSE = 0.1868 | LCC = 0.9620 | SRCC = 0.9619 ] +INFO:root:Average loss=0.05079864464700222 +INFO:root: +[Valid][18][UTT][ MSE = 0.1973 | LCC = 0.9611 | SRCC = 0.9611 ] +INFO:root:Average loss=0.05151952062547207 +INFO:root: +[Valid][19][UTT][ MSE = 0.1753 | LCC = 0.9628 | SRCC = 0.9634 ] +INFO:root:Average loss=0.0475256222486496 +INFO:root: +[Valid][20][UTT][ MSE = 0.1784 | LCC = 0.9631 | SRCC = 0.9628 ] +INFO:root:Average loss=0.04477994441986084 +INFO:root: +[Valid][21][UTT][ MSE = 0.1903 | LCC = 0.9630 | SRCC = 0.9625 ] +INFO:root:Average loss=0.04524632741510868 +INFO:root: +[Valid][22][UTT][ MSE = 0.1636 | LCC = 0.9643 | SRCC = 0.9637 ] +INFO:root:Average loss=0.04235901454091072 +INFO:root: +[Valid][23][UTT][ MSE = 0.1884 | LCC = 0.9628 | SRCC = 0.9626 ] +INFO:root:Average loss=0.04161695189774037 +INFO:root: +[Valid][24][UTT][ MSE = 0.1827 | LCC = 0.9642 | SRCC = 0.9635 ] +INFO:root:Average loss=0.03800996070355177 +INFO:root: +[Valid][25][UTT][ MSE = 0.1590 | LCC = 0.9628 | SRCC = 0.9627 ] +INFO:root:Average loss=0.0362367824614048 +INFO:root: +[Valid][26][UTT][ MSE = 0.1729 | LCC = 0.9627 | SRCC = 0.9632 ] +INFO:root:Average loss=0.03945870842039585 +INFO:root: +[Valid][27][UTT][ MSE = 0.1714 | LCC = 0.9636 | SRCC = 0.9633 ] +INFO:root:Average loss=0.03593271751701832 +INFO:root: +[Valid][28][UTT][ MSE = 0.1678 | LCC = 0.9644 | SRCC = 0.9636 ] +INFO:root:Average loss=0.03160584644973278 +INFO:root: +[Valid][29][UTT][ MSE = 0.1867 | LCC = 0.9642 | SRCC = 0.9638 ] +INFO:root:Average loss=0.03191978771984577 +INFO:root: +[Valid][30][UTT][ MSE = 0.2113 | LCC = 0.9648 | SRCC = 0.9646 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2210 | LCC = 0.9603 | SRCC = 0.9618 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run2/config.gin b/results_tencent/w2v2_large_asr_zh/layer8/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run2/model.pt b/results_tencent/w2v2_large_asr_zh/layer8/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..74c401a1b3bdc74ab5996b0af01d5b6255f2c5c3 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bfba309716dcf1f707aec85a2c5b735d33177d81af8a1b28c83867eac78d2197 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run2/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer8/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..635bef32e4179432139423cbf03b02856be2ef97 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:92a4416bf52c2b066e48e738c2440eae702b8e28fe5bed9b86b5e98366203507 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run2/train.log b/results_tencent/w2v2_large_asr_zh/layer8/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e727a74d9ee209d2480f5ec9274441a868fa7b49 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44774922943115236 +INFO:root: +[Valid][0][UTT][ MSE = 0.3424 | LCC = 0.9196 | SRCC = 0.9196 ] +INFO:root:Average loss=0.1948332362174988 +INFO:root: +[Valid][1][UTT][ MSE = 0.2509 | LCC = 0.9399 | SRCC = 0.9391 ] +INFO:root:Average loss=0.16167582112550735 +INFO:root: +[Valid][2][UTT][ MSE = 0.2198 | LCC = 0.9483 | SRCC = 0.9499 ] +INFO:root:Average loss=0.14660974299907684 +INFO:root: +[Valid][3][UTT][ MSE = 0.2599 | LCC = 0.9521 | SRCC = 0.9531 ] +INFO:root:Average loss=0.1279889994263649 +INFO:root: +[Valid][4][UTT][ MSE = 0.2536 | LCC = 0.9550 | SRCC = 0.9552 ] +INFO:root:Average loss=0.11570348888635636 +INFO:root: +[Valid][5][UTT][ MSE = 0.2104 | LCC = 0.9562 | SRCC = 0.9564 ] +INFO:root:Average loss=0.10766507813334465 +INFO:root: +[Valid][6][UTT][ MSE = 0.1640 | LCC = 0.9573 | SRCC = 0.9580 ] +INFO:root:Average loss=0.09603279766440391 +INFO:root: +[Valid][7][UTT][ MSE = 0.2038 | LCC = 0.9580 | SRCC = 0.9590 ] +INFO:root:Average loss=0.09185554948449134 +INFO:root: +[Valid][8][UTT][ MSE = 0.1945 | LCC = 0.9586 | SRCC = 0.9582 ] +INFO:root:Average loss=0.08796309566497802 +INFO:root: +[Valid][9][UTT][ MSE = 0.1998 | LCC = 0.9579 | SRCC = 0.9580 ] +INFO:root:Average loss=0.07796754083037376 +INFO:root: +[Valid][10][UTT][ MSE = 0.1830 | LCC = 0.9590 | SRCC = 0.9587 ] +INFO:root:Average loss=0.07273478761315345 +INFO:root: +[Valid][11][UTT][ MSE = 0.1679 | LCC = 0.9613 | SRCC = 0.9605 ] +INFO:root:Average loss=0.07172715041041375 +INFO:root: +[Valid][12][UTT][ MSE = 0.1706 | LCC = 0.9612 | SRCC = 0.9602 ] +INFO:root:Average loss=0.06561289596557617 +INFO:root: +[Valid][13][UTT][ MSE = 0.1806 | LCC = 0.9624 | SRCC = 0.9618 ] +INFO:root:Average loss=0.06323986351490021 +INFO:root: +[Valid][14][UTT][ MSE = 0.1894 | LCC = 0.9593 | SRCC = 0.9596 ] +INFO:root:Average loss=0.061477187231183054 +INFO:root: +[Valid][15][UTT][ MSE = 0.1651 | LCC = 0.9623 | SRCC = 0.9617 ] +INFO:root:Average loss=0.05642421653866768 +INFO:root: +[Valid][16][UTT][ MSE = 0.1773 | LCC = 0.9630 | SRCC = 0.9620 ] +INFO:root:Average loss=0.05536876100301742 +INFO:root: +[Valid][17][UTT][ MSE = 0.1811 | LCC = 0.9628 | SRCC = 0.9620 ] +INFO:root:Average loss=0.052510504633188244 +INFO:root: +[Valid][18][UTT][ MSE = 0.1620 | LCC = 0.9620 | SRCC = 0.9613 ] +INFO:root:Average loss=0.04750273796916008 +INFO:root: +[Valid][19][UTT][ MSE = 0.1715 | LCC = 0.9627 | SRCC = 0.9619 ] +INFO:root:Average loss=0.04641787739098072 +INFO:root: +[Valid][20][UTT][ MSE = 0.2100 | LCC = 0.9588 | SRCC = 0.9592 ] +INFO:root:Average loss=0.04446340523660183 +INFO:root: +[Valid][21][UTT][ MSE = 0.2026 | LCC = 0.9605 | SRCC = 0.9606 ] +INFO:root:Average loss=0.0469063820540905 +INFO:root: +[Valid][22][UTT][ MSE = 0.1559 | LCC = 0.9621 | SRCC = 0.9610 ] +INFO:root:Average loss=0.04370194587111473 +INFO:root: +[Valid][23][UTT][ MSE = 0.2133 | LCC = 0.9607 | SRCC = 0.9594 ] +INFO:root:Average loss=0.040708862274885176 +INFO:root: +[Valid][24][UTT][ MSE = 0.2067 | LCC = 0.9615 | SRCC = 0.9606 ] +INFO:root:Average loss=0.03908954948186874 +INFO:root: +[Valid][25][UTT][ MSE = 0.1773 | LCC = 0.9609 | SRCC = 0.9608 ] +INFO:root:Average loss=0.03676729027926922 +INFO:root: +[Valid][26][UTT][ MSE = 0.1908 | LCC = 0.9616 | SRCC = 0.9605 ] +INFO:root:Average loss=0.03760829408466816 +INFO:root: +[Valid][27][UTT][ MSE = 0.1916 | LCC = 0.9607 | SRCC = 0.9599 ] +INFO:root:Average loss=0.03483388356864452 +INFO:root: +[Valid][28][UTT][ MSE = 0.1710 | LCC = 0.9619 | SRCC = 0.9612 ] +INFO:root:Average loss=0.033547310635447504 +INFO:root: +[Valid][29][UTT][ MSE = 0.1794 | LCC = 0.9616 | SRCC = 0.9606 ] +INFO:root:Average loss=0.03157063837349415 +INFO:root: +[Valid][30][UTT][ MSE = 0.1834 | LCC = 0.9616 | SRCC = 0.9602 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1882 | LCC = 0.9580 | SRCC = 0.9587 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run3/config.gin b/results_tencent/w2v2_large_asr_zh/layer8/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run3/model.pt b/results_tencent/w2v2_large_asr_zh/layer8/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c111358c0a6fe882c374f4fd76fa8a4eefe820fb --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b2433b29ef3763243b75f99a13898e11f7573e2b197883015512d6a27e01c8d5 +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run3/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer8/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..24815afd4c4581905edecfa9fbc93970ae364a1d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e141797a8a32ab171701da8fe84a02eecfad0e374ca5531d50e1a1807a29b44b +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run3/train.log b/results_tencent/w2v2_large_asr_zh/layer8/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1bf8e99d6537262982e0d764a755a93ae91837d1 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4836697701215744 +INFO:root: +[Valid][0][UTT][ MSE = 0.3321 | LCC = 0.9167 | SRCC = 0.9177 ] +INFO:root:Average loss=0.20295148223638534 +INFO:root: +[Valid][1][UTT][ MSE = 0.2160 | LCC = 0.9421 | SRCC = 0.9420 ] +INFO:root:Average loss=0.15257879561185836 +INFO:root: +[Valid][2][UTT][ MSE = 0.1689 | LCC = 0.9496 | SRCC = 0.9503 ] +INFO:root:Average loss=0.13194075447320938 +INFO:root: +[Valid][3][UTT][ MSE = 0.2507 | LCC = 0.9536 | SRCC = 0.9555 ] +INFO:root:Average loss=0.12195882266759872 +INFO:root: +[Valid][4][UTT][ MSE = 0.2251 | LCC = 0.9569 | SRCC = 0.9580 ] +INFO:root:Average loss=0.10806738954782485 +INFO:root: +[Valid][5][UTT][ MSE = 0.2310 | LCC = 0.9586 | SRCC = 0.9584 ] +INFO:root:Average loss=0.10387509047985077 +INFO:root: +[Valid][6][UTT][ MSE = 0.2805 | LCC = 0.9549 | SRCC = 0.9555 ] +INFO:root:Average loss=0.09632400399446488 +INFO:root: +[Valid][7][UTT][ MSE = 0.1931 | LCC = 0.9590 | SRCC = 0.9597 ] +INFO:root:Average loss=0.0854670133292675 +INFO:root: +[Valid][8][UTT][ MSE = 0.2162 | LCC = 0.9595 | SRCC = 0.9599 ] +INFO:root:Average loss=0.08202901071310044 +INFO:root: +[Valid][9][UTT][ MSE = 0.2162 | LCC = 0.9625 | SRCC = 0.9623 ] +INFO:root:Average loss=0.07803321161866188 +INFO:root: +[Valid][10][UTT][ MSE = 0.2075 | LCC = 0.9610 | SRCC = 0.9618 ] +INFO:root:Average loss=0.07308437421917915 +INFO:root: +[Valid][11][UTT][ MSE = 0.1960 | LCC = 0.9614 | SRCC = 0.9609 ] +INFO:root:Average loss=0.06760757827758788 +INFO:root: +[Valid][12][UTT][ MSE = 0.2075 | LCC = 0.9603 | SRCC = 0.9615 ] +INFO:root:Average loss=0.06691402433812618 +INFO:root: +[Valid][13][UTT][ MSE = 0.1871 | LCC = 0.9608 | SRCC = 0.9613 ] +INFO:root:Average loss=0.06258558478951454 +INFO:root: +[Valid][14][UTT][ MSE = 0.1923 | LCC = 0.9624 | SRCC = 0.9623 ] +INFO:root:Average loss=0.057625084191560746 +INFO:root: +[Valid][15][UTT][ MSE = 0.1791 | LCC = 0.9628 | SRCC = 0.9633 ] +INFO:root:Average loss=0.0580637826025486 +INFO:root: +[Valid][16][UTT][ MSE = 0.1782 | LCC = 0.9638 | SRCC = 0.9635 ] +INFO:root:Average loss=0.05546923714876175 +INFO:root: +[Valid][17][UTT][ MSE = 0.2003 | LCC = 0.9627 | SRCC = 0.9627 ] +INFO:root:Average loss=0.05402531063556671 +INFO:root: +[Valid][18][UTT][ MSE = 0.1975 | LCC = 0.9622 | SRCC = 0.9618 ] +INFO:root:Average loss=0.04999537990987301 +INFO:root: +[Valid][19][UTT][ MSE = 0.1772 | LCC = 0.9635 | SRCC = 0.9629 ] +INFO:root:Average loss=0.04388232888281345 +INFO:root: +[Valid][20][UTT][ MSE = 0.1919 | LCC = 0.9633 | SRCC = 0.9632 ] +INFO:root:Average loss=0.044390634700655934 +INFO:root: +[Valid][21][UTT][ MSE = 0.1791 | LCC = 0.9642 | SRCC = 0.9636 ] +INFO:root:Average loss=0.041242802619934085 +INFO:root: +[Valid][22][UTT][ MSE = 0.1822 | LCC = 0.9636 | SRCC = 0.9642 ] +INFO:root:Average loss=0.04066494563221931 +INFO:root: +[Valid][23][UTT][ MSE = 0.2032 | LCC = 0.9633 | SRCC = 0.9633 ] +INFO:root:Average loss=0.03926120997965336 +INFO:root: +[Valid][24][UTT][ MSE = 0.1777 | LCC = 0.9635 | SRCC = 0.9630 ] +INFO:root:Average loss=0.037379612520337106 +INFO:root: +[Valid][25][UTT][ MSE = 0.1762 | LCC = 0.9630 | SRCC = 0.9630 ] +INFO:root:Average loss=0.03639578554034233 +INFO:root: +[Valid][26][UTT][ MSE = 0.1711 | LCC = 0.9646 | SRCC = 0.9640 ] +INFO:root:Average loss=0.036559226378798486 +INFO:root: +[Valid][27][UTT][ MSE = 0.1823 | LCC = 0.9632 | SRCC = 0.9639 ] +INFO:root:Average loss=0.033565830841660496 +INFO:root: +[Valid][28][UTT][ MSE = 0.1975 | LCC = 0.9635 | SRCC = 0.9636 ] +INFO:root:Average loss=0.03490517760813236 +INFO:root: +[Valid][29][UTT][ MSE = 0.1728 | LCC = 0.9632 | SRCC = 0.9629 ] +INFO:root:Average loss=0.03267583341896534 +INFO:root: +[Valid][30][UTT][ MSE = 0.1827 | LCC = 0.9647 | SRCC = 0.9641 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1857 | LCC = 0.9629 | SRCC = 0.9631 ] diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run4/config.gin b/results_tencent/w2v2_large_asr_zh/layer8/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run4/model.pt b/results_tencent/w2v2_large_asr_zh/layer8/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c652001251678795f4f1b1c313d9157f6e08fdc0 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39c61b411166be1e7b929276f345e165a33e947761072f968b395f5cdb39bc5d +size 3403339 diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run4/model_best.pt b/results_tencent/w2v2_large_asr_zh/layer8/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f248d37580953999991cbd710266c06edb06ce16 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a99f02fa042f9720251a71d2867ddc1f8e57ac91a2d616995966bf59c3b0b99 +size 3405062 diff --git a/results_tencent/w2v2_large_asr_zh/layer8/run4/train.log b/results_tencent/w2v2_large_asr_zh/layer8/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..44e6bb2a471b45b97992bbf5a0fd9e3e9fccf7c2 --- /dev/null +++ b/results_tencent/w2v2_large_asr_zh/layer8/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.49412175261974334 +INFO:root: +[Valid][0][UTT][ MSE = 0.3270 | LCC = 0.9112 | SRCC = 0.9136 ] +INFO:root:Average loss=0.20865376114845277 +INFO:root: +[Valid][1][UTT][ MSE = 0.2963 | LCC = 0.9357 | SRCC = 0.9350 ] +INFO:root:Average loss=0.16605838137865067 +INFO:root: +[Valid][2][UTT][ MSE = 0.2624 | LCC = 0.9493 | SRCC = 0.9495 ] +INFO:root:Average loss=0.1375709785223007 +INFO:root: +[Valid][3][UTT][ MSE = 0.1959 | LCC = 0.9512 | SRCC = 0.9521 ] +INFO:root:Average loss=0.12683605194091796 +INFO:root: +[Valid][4][UTT][ MSE = 0.2143 | LCC = 0.9544 | SRCC = 0.9555 ] +INFO:root:Average loss=0.11514004904031754 +INFO:root: +[Valid][5][UTT][ MSE = 0.1837 | LCC = 0.9556 | SRCC = 0.9566 ] +INFO:root:Average loss=0.10114439269900322 +INFO:root: +[Valid][6][UTT][ MSE = 0.2189 | LCC = 0.9601 | SRCC = 0.9600 ] +INFO:root:Average loss=0.09557655394077301 +INFO:root: +[Valid][7][UTT][ MSE = 0.2070 | LCC = 0.9610 | SRCC = 0.9608 ] +INFO:root:Average loss=0.09130334115028381 +INFO:root: +[Valid][8][UTT][ MSE = 0.2149 | LCC = 0.9610 | SRCC = 0.9609 ] +INFO:root:Average loss=0.08409697943925858 +INFO:root: +[Valid][9][UTT][ MSE = 0.1872 | LCC = 0.9622 | SRCC = 0.9633 ] +INFO:root:Average loss=0.08637213379144669 +INFO:root: +[Valid][10][UTT][ MSE = 0.2347 | LCC = 0.9611 | SRCC = 0.9621 ] +INFO:root:Average loss=0.0769672112762928 +INFO:root: +[Valid][11][UTT][ MSE = 0.2162 | LCC = 0.9611 | SRCC = 0.9630 ] +INFO:root:Average loss=0.07498842957615852 +INFO:root: +[Valid][12][UTT][ MSE = 0.2027 | LCC = 0.9625 | SRCC = 0.9630 ] +INFO:root:Average loss=0.07305460196733475 +INFO:root: +[Valid][13][UTT][ MSE = 0.1814 | LCC = 0.9631 | SRCC = 0.9628 ] +INFO:root:Average loss=0.07278050461411475 +INFO:root: +[Valid][14][UTT][ MSE = 0.2115 | LCC = 0.9637 | SRCC = 0.9639 ] +INFO:root:Average loss=0.06420149096846581 +INFO:root: +[Valid][15][UTT][ MSE = 0.1700 | LCC = 0.9653 | SRCC = 0.9649 ] +INFO:root:Average loss=0.06254185220599175 +INFO:root: +[Valid][16][UTT][ MSE = 0.1734 | LCC = 0.9645 | SRCC = 0.9642 ] +INFO:root:Average loss=0.055992768928408626 +INFO:root: +[Valid][17][UTT][ MSE = 0.1657 | LCC = 0.9643 | SRCC = 0.9643 ] +INFO:root:Average loss=0.05459943506121635 +INFO:root: +[Valid][18][UTT][ MSE = 0.1992 | LCC = 0.9644 | SRCC = 0.9639 ] +INFO:root:Average loss=0.051207719922065735 +INFO:root: +[Valid][19][UTT][ MSE = 0.1966 | LCC = 0.9647 | SRCC = 0.9639 ] +INFO:root:Average loss=0.048543695613741875 +INFO:root: +[Valid][20][UTT][ MSE = 0.1996 | LCC = 0.9645 | SRCC = 0.9637 ] +INFO:root:Average loss=0.0470619645267725 +INFO:root: +[Valid][21][UTT][ MSE = 0.1641 | LCC = 0.9648 | SRCC = 0.9641 ] +INFO:root:Average loss=0.04560992541909218 +INFO:root: +[Valid][22][UTT][ MSE = 0.1979 | LCC = 0.9650 | SRCC = 0.9647 ] +INFO:root:Average loss=0.04447022522985935 +INFO:root: +[Valid][23][UTT][ MSE = 0.2033 | LCC = 0.9647 | SRCC = 0.9641 ] +INFO:root:Average loss=0.04189255174994469 +INFO:root: +[Valid][24][UTT][ MSE = 0.1836 | LCC = 0.9641 | SRCC = 0.9637 ] +INFO:root:Average loss=0.04142818939685822 +INFO:root: +[Valid][25][UTT][ MSE = 0.2108 | LCC = 0.9639 | SRCC = 0.9635 ] +INFO:root:Average loss=0.04071401704847813 +INFO:root: +[Valid][26][UTT][ MSE = 0.1697 | LCC = 0.9642 | SRCC = 0.9632 ] +INFO:root:Average loss=0.03768039238452912 +INFO:root: +[Valid][27][UTT][ MSE = 0.1882 | LCC = 0.9623 | SRCC = 0.9618 ] +INFO:root:Average loss=0.037324674323201176 +INFO:root: +[Valid][28][UTT][ MSE = 0.1924 | LCC = 0.9647 | SRCC = 0.9643 ] +INFO:root:Average loss=0.03742805194854736 +INFO:root: +[Valid][29][UTT][ MSE = 0.1978 | LCC = 0.9633 | SRCC = 0.9631 ] +INFO:root:Average loss=0.032943107977509496 +INFO:root: +[Valid][30][UTT][ MSE = 0.1874 | LCC = 0.9629 | SRCC = 0.9625 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1843 | LCC = 0.9593 | SRCC = 0.9599 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run0/config.gin b/results_tencent/w2v2_large_xlsr53/layer0/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run0/model.pt b/results_tencent/w2v2_large_xlsr53/layer0/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..fb0d77463b9b7efe9c0f9ed138c65c7eccd197fe --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f97a365978a0f65869abbbefe867da9ef28fb2678e9a30076c663a47dfbbc168 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run0/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer0/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..8abccaf609ef8b03c4da6eee48bbb1952247fa7b --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:141fb650eabeb934e20f76ff62e95b2b568c77a1f2bc28957f57fcf2f681fb38 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run0/train.log b/results_tencent/w2v2_large_xlsr53/layer0/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..39712b006cf1425b91e53f2fbbe415ba3fb4416d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6684663178920746 +INFO:root: +[Valid][0][UTT][ MSE = 0.4687 | LCC = 0.8689 | SRCC = 0.8685 ] +INFO:root:Average loss=0.4354150129556656 +INFO:root: +[Valid][1][UTT][ MSE = 0.4271 | LCC = 0.8899 | SRCC = 0.8888 ] +INFO:root:Average loss=0.35578377890586854 +INFO:root: +[Valid][2][UTT][ MSE = 0.3866 | LCC = 0.8834 | SRCC = 0.8865 ] +INFO:root:Average loss=0.30127158868312837 +INFO:root: +[Valid][3][UTT][ MSE = 0.3489 | LCC = 0.9074 | SRCC = 0.9070 ] +INFO:root:Average loss=0.29030870163440703 +INFO:root: +[Valid][4][UTT][ MSE = 0.4047 | LCC = 0.9137 | SRCC = 0.9166 ] +INFO:root:Average loss=0.24374581927061081 +INFO:root: +[Valid][5][UTT][ MSE = 0.3189 | LCC = 0.9195 | SRCC = 0.9198 ] +INFO:root:Average loss=0.22644798529148102 +INFO:root: +[Valid][6][UTT][ MSE = 0.2885 | LCC = 0.9188 | SRCC = 0.9192 ] +INFO:root:Average loss=0.2253140875697136 +INFO:root: +[Valid][7][UTT][ MSE = 0.4105 | LCC = 0.9053 | SRCC = 0.9064 ] +INFO:root:Average loss=0.22001235473155975 +INFO:root: +[Valid][8][UTT][ MSE = 0.2631 | LCC = 0.9253 | SRCC = 0.9263 ] +INFO:root:Average loss=0.21736555325984955 +INFO:root: +[Valid][9][UTT][ MSE = 0.3144 | LCC = 0.9259 | SRCC = 0.9276 ] +INFO:root:Average loss=0.19208678823709488 +INFO:root: +[Valid][10][UTT][ MSE = 0.2399 | LCC = 0.9274 | SRCC = 0.9294 ] +INFO:root:Average loss=0.18190429538488387 +INFO:root: +[Valid][11][UTT][ MSE = 0.2613 | LCC = 0.9292 | SRCC = 0.9306 ] +INFO:root:Average loss=0.1689336720108986 +INFO:root: +[Valid][12][UTT][ MSE = 0.2752 | LCC = 0.9301 | SRCC = 0.9310 ] +INFO:root:Average loss=0.16152018314599992 +INFO:root: +[Valid][13][UTT][ MSE = 0.2535 | LCC = 0.9329 | SRCC = 0.9333 ] +INFO:root:Average loss=0.16312175130844117 +INFO:root: +[Valid][14][UTT][ MSE = 0.2726 | LCC = 0.9328 | SRCC = 0.9339 ] +INFO:root:Average loss=0.15329030668735505 +INFO:root: +[Valid][15][UTT][ MSE = 0.2377 | LCC = 0.9358 | SRCC = 0.9364 ] +INFO:root:Average loss=0.15228110897541047 +INFO:root: +[Valid][16][UTT][ MSE = 0.2501 | LCC = 0.9218 | SRCC = 0.9224 ] +INFO:root:Average loss=0.1467117264866829 +INFO:root: +[Valid][17][UTT][ MSE = 0.2652 | LCC = 0.9306 | SRCC = 0.9324 ] +INFO:root:Average loss=0.143916130900383 +INFO:root: +[Valid][18][UTT][ MSE = 0.2755 | LCC = 0.9319 | SRCC = 0.9323 ] +INFO:root:Average loss=0.1377168068289757 +INFO:root: +[Valid][19][UTT][ MSE = 0.2767 | LCC = 0.9358 | SRCC = 0.9377 ] +INFO:root:Average loss=0.131955485701561 +INFO:root: +[Valid][20][UTT][ MSE = 0.2735 | LCC = 0.9340 | SRCC = 0.9349 ] +INFO:root:Average loss=0.1268673693239689 +INFO:root: +[Valid][21][UTT][ MSE = 0.2367 | LCC = 0.9357 | SRCC = 0.9362 ] +INFO:root:Average loss=0.1299217260479927 +INFO:root: +[Valid][22][UTT][ MSE = 0.2606 | LCC = 0.9382 | SRCC = 0.9388 ] +INFO:root:Average loss=0.12682672441005707 +INFO:root: +[Valid][23][UTT][ MSE = 0.2499 | LCC = 0.9383 | SRCC = 0.9390 ] +INFO:root:Average loss=0.11710813307762145 +INFO:root: +[Valid][24][UTT][ MSE = 0.2401 | LCC = 0.9382 | SRCC = 0.9389 ] +INFO:root:Average loss=0.11410166466236114 +INFO:root: +[Valid][25][UTT][ MSE = 0.2240 | LCC = 0.9395 | SRCC = 0.9403 ] +INFO:root:Average loss=0.1124019986987114 +INFO:root: +[Valid][26][UTT][ MSE = 0.2574 | LCC = 0.9407 | SRCC = 0.9406 ] +INFO:root:Average loss=0.10972828075289726 +INFO:root: +[Valid][27][UTT][ MSE = 0.2664 | LCC = 0.9383 | SRCC = 0.9398 ] +INFO:root:Average loss=0.10562221109867095 +INFO:root: +[Valid][28][UTT][ MSE = 0.2343 | LCC = 0.9409 | SRCC = 0.9412 ] +INFO:root:Average loss=0.10213524806499481 +INFO:root: +[Valid][29][UTT][ MSE = 0.2650 | LCC = 0.9363 | SRCC = 0.9373 ] +INFO:root:Average loss=0.10028490921854973 +INFO:root: +[Valid][30][UTT][ MSE = 0.2544 | LCC = 0.9402 | SRCC = 0.9408 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2477 | LCC = 0.9397 | SRCC = 0.9400 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run1/config.gin b/results_tencent/w2v2_large_xlsr53/layer0/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run1/model.pt b/results_tencent/w2v2_large_xlsr53/layer0/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..aa3627f5377684160592ff2d08790f4ce0546645 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8ef19c6b9bc767c21434c527956b6724488f5e1a82a9a4c8db305440df41524a +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run1/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer0/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..edd00af37a7f84f154f9199fbe52687716abaad3 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5fee9ab4c644351d7210da2328e11e00c72c27bf74da1fca77e4e89257e63b12 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run1/train.log b/results_tencent/w2v2_large_xlsr53/layer0/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f747c0d3d6d551cd599ed2af9a6a404acad72c53 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5837222230434418 +INFO:root: +[Valid][0][UTT][ MSE = 0.3990 | LCC = 0.8827 | SRCC = 0.8808 ] +INFO:root:Average loss=0.37926537561416623 +INFO:root: +[Valid][1][UTT][ MSE = 0.5191 | LCC = 0.8680 | SRCC = 0.8759 ] +INFO:root:Average loss=0.3750261504650116 +INFO:root: +[Valid][2][UTT][ MSE = 0.4609 | LCC = 0.8969 | SRCC = 0.9024 ] +INFO:root:Average loss=0.34612031865119935 +INFO:root: +[Valid][3][UTT][ MSE = 0.3146 | LCC = 0.9122 | SRCC = 0.9125 ] +INFO:root:Average loss=0.2709216994047165 +INFO:root: +[Valid][4][UTT][ MSE = 0.2990 | LCC = 0.9171 | SRCC = 0.9165 ] +INFO:root:Average loss=0.24233896654844284 +INFO:root: +[Valid][5][UTT][ MSE = 0.3041 | LCC = 0.9202 | SRCC = 0.9208 ] +INFO:root:Average loss=0.23145117390155792 +INFO:root: +[Valid][6][UTT][ MSE = 0.2945 | LCC = 0.9198 | SRCC = 0.9216 ] +INFO:root:Average loss=0.22357566475868226 +INFO:root: +[Valid][7][UTT][ MSE = 0.2697 | LCC = 0.9254 | SRCC = 0.9266 ] +INFO:root:Average loss=0.2168325988650322 +INFO:root: +[Valid][8][UTT][ MSE = 0.2626 | LCC = 0.9231 | SRCC = 0.9254 ] +INFO:root:Average loss=0.1895653783082962 +INFO:root: +[Valid][9][UTT][ MSE = 0.2390 | LCC = 0.9285 | SRCC = 0.9295 ] +INFO:root:Average loss=0.198195929646492 +INFO:root: +[Valid][10][UTT][ MSE = 0.2816 | LCC = 0.9272 | SRCC = 0.9293 ] +INFO:root:Average loss=0.18136970341205597 +INFO:root: +[Valid][11][UTT][ MSE = 0.2518 | LCC = 0.9330 | SRCC = 0.9337 ] +INFO:root:Average loss=0.17546321886777877 +INFO:root: +[Valid][12][UTT][ MSE = 0.2469 | LCC = 0.9272 | SRCC = 0.9285 ] +INFO:root:Average loss=0.16934628415107728 +INFO:root: +[Valid][13][UTT][ MSE = 0.2389 | LCC = 0.9349 | SRCC = 0.9363 ] +INFO:root:Average loss=0.1620507578253746 +INFO:root: +[Valid][14][UTT][ MSE = 0.2264 | LCC = 0.9348 | SRCC = 0.9368 ] +INFO:root:Average loss=0.15236909711360933 +INFO:root: +[Valid][15][UTT][ MSE = 0.2348 | LCC = 0.9335 | SRCC = 0.9336 ] +INFO:root:Average loss=0.1540885056257248 +INFO:root: +[Valid][16][UTT][ MSE = 0.2566 | LCC = 0.9364 | SRCC = 0.9385 ] +INFO:root:Average loss=0.14943861323595048 +INFO:root: +[Valid][17][UTT][ MSE = 0.2207 | LCC = 0.9358 | SRCC = 0.9370 ] +INFO:root:Average loss=0.13905636501312255 +INFO:root: +[Valid][18][UTT][ MSE = 0.2376 | LCC = 0.9369 | SRCC = 0.9381 ] +INFO:root:Average loss=0.1339317115545273 +INFO:root: +[Valid][19][UTT][ MSE = 0.2155 | LCC = 0.9378 | SRCC = 0.9399 ] +INFO:root:Average loss=0.13255828499794006 +INFO:root: +[Valid][20][UTT][ MSE = 0.2127 | LCC = 0.9368 | SRCC = 0.9393 ] +INFO:root:Average loss=0.12660698729753495 +INFO:root: +[Valid][21][UTT][ MSE = 0.2330 | LCC = 0.9416 | SRCC = 0.9422 ] +INFO:root:Average loss=0.12383249640464783 +INFO:root: +[Valid][22][UTT][ MSE = 0.2178 | LCC = 0.9399 | SRCC = 0.9405 ] +INFO:root:Average loss=0.11614035254716873 +INFO:root: +[Valid][23][UTT][ MSE = 0.2651 | LCC = 0.9372 | SRCC = 0.9375 ] +INFO:root:Average loss=0.11248802757263184 +INFO:root: +[Valid][24][UTT][ MSE = 0.2220 | LCC = 0.9395 | SRCC = 0.9406 ] +INFO:root:Average loss=0.1144346199631691 +INFO:root: +[Valid][25][UTT][ MSE = 0.2044 | LCC = 0.9414 | SRCC = 0.9423 ] +INFO:root:Average loss=0.10589844101667405 +INFO:root: +[Valid][26][UTT][ MSE = 0.2159 | LCC = 0.9441 | SRCC = 0.9444 ] +INFO:root:Average loss=0.10525080543756485 +INFO:root: +[Valid][27][UTT][ MSE = 0.2110 | LCC = 0.9426 | SRCC = 0.9435 ] +INFO:root:Average loss=0.10194379246234894 +INFO:root: +[Valid][28][UTT][ MSE = 0.2163 | LCC = 0.9421 | SRCC = 0.9427 ] +INFO:root:Average loss=0.1059630334675312 +INFO:root: +[Valid][29][UTT][ MSE = 0.1988 | LCC = 0.9394 | SRCC = 0.9427 ] +INFO:root:Average loss=0.10036472716927529 +INFO:root: +[Valid][30][UTT][ MSE = 0.2193 | LCC = 0.9400 | SRCC = 0.9404 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2274 | LCC = 0.9434 | SRCC = 0.9446 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run2/config.gin b/results_tencent/w2v2_large_xlsr53/layer0/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run2/model.pt b/results_tencent/w2v2_large_xlsr53/layer0/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3f614a0588e519d792ab7dcc91f1044cedc1c9cb --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:47a3591f8b4e02ad7962804116aad08aef4960856bd0c65e590cf769abe2a8bb +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run2/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer0/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..5a8fa23076b56d8ff6eab2d0a7e4378e2139f1a8 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8ec41e85e2da69341c723347fb18ee999831dd2ce5107410e4c9df0bfcbd21b9 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run2/train.log b/results_tencent/w2v2_large_xlsr53/layer0/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0e28f96c96cbd9d3514c837c791b28e7ea315650 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.750185741186142 +INFO:root: +[Valid][0][UTT][ MSE = 0.6281 | LCC = 0.7969 | SRCC = 0.8163 ] +INFO:root:Average loss=0.4492122069597244 +INFO:root: +[Valid][1][UTT][ MSE = 0.4190 | LCC = 0.8819 | SRCC = 0.8829 ] +INFO:root:Average loss=0.32458782112598417 +INFO:root: +[Valid][2][UTT][ MSE = 0.3128 | LCC = 0.8958 | SRCC = 0.9004 ] +INFO:root:Average loss=0.2905689994096756 +INFO:root: +[Valid][3][UTT][ MSE = 0.3339 | LCC = 0.9105 | SRCC = 0.9129 ] +INFO:root:Average loss=0.28524586391448975 +INFO:root: +[Valid][4][UTT][ MSE = 0.3165 | LCC = 0.9043 | SRCC = 0.9045 ] +INFO:root:Average loss=0.24182996785640717 +INFO:root: +[Valid][5][UTT][ MSE = 0.3110 | LCC = 0.9179 | SRCC = 0.9186 ] +INFO:root:Average loss=0.2296307705640793 +INFO:root: +[Valid][6][UTT][ MSE = 0.2766 | LCC = 0.9197 | SRCC = 0.9223 ] +INFO:root:Average loss=0.21911851423978806 +INFO:root: +[Valid][7][UTT][ MSE = 0.2569 | LCC = 0.9237 | SRCC = 0.9255 ] +INFO:root:Average loss=0.2030495409965515 +INFO:root: +[Valid][8][UTT][ MSE = 0.3311 | LCC = 0.9217 | SRCC = 0.9236 ] +INFO:root:Average loss=0.20859082287549974 +INFO:root: +[Valid][9][UTT][ MSE = 0.2652 | LCC = 0.9281 | SRCC = 0.9287 ] +INFO:root:Average loss=0.18486490589380264 +INFO:root: +[Valid][10][UTT][ MSE = 0.2672 | LCC = 0.9304 | SRCC = 0.9304 ] +INFO:root:Average loss=0.17921914285421373 +INFO:root: +[Valid][11][UTT][ MSE = 0.2556 | LCC = 0.9293 | SRCC = 0.9291 ] +INFO:root:Average loss=0.17137171697616577 +INFO:root: +[Valid][12][UTT][ MSE = 0.2348 | LCC = 0.9212 | SRCC = 0.9236 ] +INFO:root:Average loss=0.1651897394657135 +INFO:root: +[Valid][13][UTT][ MSE = 0.3299 | LCC = 0.9305 | SRCC = 0.9332 ] +INFO:root:Average loss=0.15801941329240798 +INFO:root: +[Valid][14][UTT][ MSE = 0.2255 | LCC = 0.9349 | SRCC = 0.9354 ] +INFO:root:Average loss=0.15537130773067476 +INFO:root: +[Valid][15][UTT][ MSE = 0.2337 | LCC = 0.9375 | SRCC = 0.9371 ] +INFO:root:Average loss=0.14964659357070922 +INFO:root: +[Valid][16][UTT][ MSE = 0.2526 | LCC = 0.9330 | SRCC = 0.9330 ] +INFO:root:Average loss=0.13908950865268707 +INFO:root: +[Valid][17][UTT][ MSE = 0.2787 | LCC = 0.9361 | SRCC = 0.9367 ] +INFO:root:Average loss=0.14417859655618667 +INFO:root: +[Valid][18][UTT][ MSE = 0.2460 | LCC = 0.9355 | SRCC = 0.9376 ] +INFO:root:Average loss=0.13567350792884827 +INFO:root: +[Valid][19][UTT][ MSE = 0.3075 | LCC = 0.9342 | SRCC = 0.9349 ] +INFO:root:Average loss=0.14547885394096374 +INFO:root: +[Valid][20][UTT][ MSE = 0.2600 | LCC = 0.9412 | SRCC = 0.9410 ] +INFO:root:Average loss=0.12422155395150185 +INFO:root: +[Valid][21][UTT][ MSE = 0.2252 | LCC = 0.9412 | SRCC = 0.9405 ] +INFO:root:Average loss=0.12368121713399888 +INFO:root: +[Valid][22][UTT][ MSE = 0.2623 | LCC = 0.9391 | SRCC = 0.9390 ] +INFO:root:Average loss=0.11992577153444291 +INFO:root: +[Valid][23][UTT][ MSE = 0.2407 | LCC = 0.9383 | SRCC = 0.9385 ] +INFO:root:Average loss=0.11639479985833168 +INFO:root: +[Valid][24][UTT][ MSE = 0.2498 | LCC = 0.9373 | SRCC = 0.9381 ] +INFO:root:Average loss=0.11037976777553558 +INFO:root: +[Valid][25][UTT][ MSE = 0.2258 | LCC = 0.9345 | SRCC = 0.9373 ] +INFO:root:Average loss=0.11191148030757904 +INFO:root: +[Valid][26][UTT][ MSE = 0.2235 | LCC = 0.9383 | SRCC = 0.9390 ] +INFO:root:Average loss=0.10570981842279434 +INFO:root: +[Valid][27][UTT][ MSE = 0.2223 | LCC = 0.9381 | SRCC = 0.9392 ] +INFO:root:Average loss=0.10534641256928444 +INFO:root: +[Valid][28][UTT][ MSE = 0.2173 | LCC = 0.9411 | SRCC = 0.9409 ] +INFO:root:Average loss=0.1030001725256443 +INFO:root: +[Valid][29][UTT][ MSE = 0.2196 | LCC = 0.9418 | SRCC = 0.9421 ] +INFO:root:Average loss=0.09647428539395332 +INFO:root: +[Valid][30][UTT][ MSE = 0.2317 | LCC = 0.9424 | SRCC = 0.9414 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2418 | LCC = 0.9421 | SRCC = 0.9425 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run3/config.gin b/results_tencent/w2v2_large_xlsr53/layer0/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run3/model.pt b/results_tencent/w2v2_large_xlsr53/layer0/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..11361bc7a8eb4c64fc42730691ff4180b20f1035 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df95a1c6e9ba9d522528252e6e0e20f8cd570216472f172ff1b899a531559a67 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run3/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer0/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..25827715b3d40c83fe45f491abc875282c0a7099 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1d2fe3ef91595a6d43653795660734ebd65a263b4ae5d0220b58ad3eb8e80dc8 +size 3404998 diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run3/train.log b/results_tencent/w2v2_large_xlsr53/layer0/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..bcf6d8a449bfae4a74d20059c799dd7d63ca27bb --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.65171271276474 +INFO:root: +[Valid][0][UTT][ MSE = 0.6311 | LCC = 0.8548 | SRCC = 0.8568 ] +INFO:root:Average loss=0.4105572341680527 +INFO:root: +[Valid][1][UTT][ MSE = 0.4241 | LCC = 0.8907 | SRCC = 0.8916 ] +INFO:root:Average loss=0.3335506502389908 +INFO:root: +[Valid][2][UTT][ MSE = 0.3691 | LCC = 0.9035 | SRCC = 0.9043 ] +INFO:root:Average loss=0.3025468566417694 +INFO:root: +[Valid][3][UTT][ MSE = 0.3230 | LCC = 0.9055 | SRCC = 0.9056 ] +INFO:root:Average loss=0.2882475498914719 +INFO:root: +[Valid][4][UTT][ MSE = 0.3621 | LCC = 0.9150 | SRCC = 0.9153 ] +INFO:root:Average loss=0.2636498420238495 +INFO:root: +[Valid][5][UTT][ MSE = 0.4386 | LCC = 0.9155 | SRCC = 0.9160 ] +INFO:root:Average loss=0.24349230802059174 +INFO:root: +[Valid][6][UTT][ MSE = 0.4062 | LCC = 0.9115 | SRCC = 0.9109 ] +INFO:root:Average loss=0.24722016316652298 +INFO:root: +[Valid][7][UTT][ MSE = 0.3471 | LCC = 0.9193 | SRCC = 0.9208 ] +INFO:root:Average loss=0.21488330298662187 +INFO:root: +[Valid][8][UTT][ MSE = 0.3055 | LCC = 0.9222 | SRCC = 0.9224 ] +INFO:root:Average loss=0.21903651022911072 +INFO:root: +[Valid][9][UTT][ MSE = 0.2805 | LCC = 0.9213 | SRCC = 0.9229 ] +INFO:root:Average loss=0.18750878965854645 +INFO:root: +[Valid][10][UTT][ MSE = 0.3200 | LCC = 0.9293 | SRCC = 0.9307 ] +INFO:root:Average loss=0.20242166632413863 +INFO:root: +[Valid][11][UTT][ MSE = 0.3408 | LCC = 0.9272 | SRCC = 0.9268 ] +INFO:root:Average loss=0.18443584787845613 +INFO:root: +[Valid][12][UTT][ MSE = 0.2378 | LCC = 0.9313 | SRCC = 0.9317 ] +INFO:root:Average loss=0.1865711720585823 +INFO:root: +[Valid][13][UTT][ MSE = 0.2871 | LCC = 0.9288 | SRCC = 0.9304 ] +INFO:root:Average loss=0.16657097190618514 +INFO:root: +[Valid][14][UTT][ MSE = 0.2766 | LCC = 0.9276 | SRCC = 0.9269 ] +INFO:root:Average loss=0.1593109455704689 +INFO:root: +[Valid][15][UTT][ MSE = 0.3304 | LCC = 0.9268 | SRCC = 0.9258 ] +INFO:root:Average loss=0.15423468750715255 +INFO:root: +[Valid][16][UTT][ MSE = 0.2329 | LCC = 0.9339 | SRCC = 0.9345 ] +INFO:root:Average loss=0.14898351126909257 +INFO:root: +[Valid][17][UTT][ MSE = 0.3446 | LCC = 0.9310 | SRCC = 0.9313 ] +INFO:root:Average loss=0.1464483929872513 +INFO:root: +[Valid][18][UTT][ MSE = 0.2322 | LCC = 0.9352 | SRCC = 0.9358 ] +INFO:root:Average loss=0.13828378164768218 +INFO:root: +[Valid][19][UTT][ MSE = 0.2555 | LCC = 0.9332 | SRCC = 0.9326 ] +INFO:root:Average loss=0.13082974749803544 +INFO:root: +[Valid][20][UTT][ MSE = 0.3206 | LCC = 0.9303 | SRCC = 0.9297 ] +INFO:root:Average loss=0.13151658272743225 +INFO:root: +[Valid][21][UTT][ MSE = 0.3055 | LCC = 0.9329 | SRCC = 0.9329 ] +INFO:root:Average loss=0.12544846487045289 +INFO:root: +[Valid][22][UTT][ MSE = 0.2624 | LCC = 0.9381 | SRCC = 0.9376 ] +INFO:root:Average loss=0.11823065984249115 +INFO:root: +[Valid][23][UTT][ MSE = 0.2661 | LCC = 0.9394 | SRCC = 0.9387 ] +INFO:root:Average loss=0.1216283804178238 +INFO:root: +[Valid][24][UTT][ MSE = 0.2948 | LCC = 0.9373 | SRCC = 0.9371 ] +INFO:root:Average loss=0.11601371490955353 +INFO:root: +[Valid][25][UTT][ MSE = 0.2598 | LCC = 0.9387 | SRCC = 0.9383 ] +INFO:root:Average loss=0.11022782409191131 +INFO:root: +[Valid][26][UTT][ MSE = 0.2628 | LCC = 0.9385 | SRCC = 0.9382 ] +INFO:root:Average loss=0.11287579074501991 +INFO:root: +[Valid][27][UTT][ MSE = 0.3204 | LCC = 0.9292 | SRCC = 0.9289 ] +INFO:root:Average loss=0.10381621474027633 +INFO:root: +[Valid][28][UTT][ MSE = 0.2661 | LCC = 0.9400 | SRCC = 0.9397 ] +INFO:root:Average loss=0.10066493353247642 +INFO:root: +[Valid][29][UTT][ MSE = 0.2614 | LCC = 0.9366 | SRCC = 0.9358 ] +INFO:root:Average loss=0.09712157186865807 +INFO:root: +[Valid][30][UTT][ MSE = 0.2888 | LCC = 0.9387 | SRCC = 0.9379 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2672 | LCC = 0.9419 | SRCC = 0.9426 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run4/config.gin b/results_tencent/w2v2_large_xlsr53/layer0/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run4/model.pt b/results_tencent/w2v2_large_xlsr53/layer0/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3977a00948926a59268191bd05b0f015483eb195 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b2ef180e5bbaafdb1c22b761593f4122a293043c31c6e15eb79127e61e63c561 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run4/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer0/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..658ff9a085f9be20dbd58d4dc8b6ec4a54f09e08 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2cbb5c0a847ea5babe9faf7348bd30563e9f793d9ae0ab78425d3d9084147145 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer0/run4/train.log b/results_tencent/w2v2_large_xlsr53/layer0/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e0eaeb03dbfd152106079dc135d1794d99add11b --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer0/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6321425845623017 +INFO:root: +[Valid][0][UTT][ MSE = 0.4385 | LCC = 0.8573 | SRCC = 0.8578 ] +INFO:root:Average loss=0.3828275818824768 +INFO:root: +[Valid][1][UTT][ MSE = 0.3644 | LCC = 0.8965 | SRCC = 0.8970 ] +INFO:root:Average loss=0.3474248902797699 +INFO:root: +[Valid][2][UTT][ MSE = 0.3837 | LCC = 0.9046 | SRCC = 0.9071 ] +INFO:root:Average loss=0.30191039919853213 +INFO:root: +[Valid][3][UTT][ MSE = 0.3051 | LCC = 0.9137 | SRCC = 0.9157 ] +INFO:root:Average loss=0.2846943528652191 +INFO:root: +[Valid][4][UTT][ MSE = 0.3446 | LCC = 0.9168 | SRCC = 0.9175 ] +INFO:root:Average loss=0.24959957081079484 +INFO:root: +[Valid][5][UTT][ MSE = 0.3330 | LCC = 0.9147 | SRCC = 0.9145 ] +INFO:root:Average loss=0.23334274631738663 +INFO:root: +[Valid][6][UTT][ MSE = 0.4057 | LCC = 0.9089 | SRCC = 0.9146 ] +INFO:root:Average loss=0.23392452991008758 +INFO:root: +[Valid][7][UTT][ MSE = 0.4346 | LCC = 0.9060 | SRCC = 0.9058 ] +INFO:root:Average loss=0.20908035123348237 +INFO:root: +[Valid][8][UTT][ MSE = 0.3137 | LCC = 0.9200 | SRCC = 0.9200 ] +INFO:root:Average loss=0.20712517166137695 +INFO:root: +[Valid][9][UTT][ MSE = 0.3535 | LCC = 0.9220 | SRCC = 0.9220 ] +INFO:root:Average loss=0.18636799710989 +INFO:root: +[Valid][10][UTT][ MSE = 0.3409 | LCC = 0.9110 | SRCC = 0.9106 ] +INFO:root:Average loss=0.1817897207736969 +INFO:root: +[Valid][11][UTT][ MSE = 0.3421 | LCC = 0.9273 | SRCC = 0.9277 ] +INFO:root:Average loss=0.17374174445867538 +INFO:root: +[Valid][12][UTT][ MSE = 0.2554 | LCC = 0.9308 | SRCC = 0.9312 ] +INFO:root:Average loss=0.1732175037264824 +INFO:root: +[Valid][13][UTT][ MSE = 0.2713 | LCC = 0.9333 | SRCC = 0.9334 ] +INFO:root:Average loss=0.16199878281354904 +INFO:root: +[Valid][14][UTT][ MSE = 0.3053 | LCC = 0.9205 | SRCC = 0.9239 ] +INFO:root:Average loss=0.1650044282078743 +INFO:root: +[Valid][15][UTT][ MSE = 0.2122 | LCC = 0.9356 | SRCC = 0.9364 ] +INFO:root:Average loss=0.1523437914252281 +INFO:root: +[Valid][16][UTT][ MSE = 0.2825 | LCC = 0.9353 | SRCC = 0.9354 ] +INFO:root:Average loss=0.1537971231341362 +INFO:root: +[Valid][17][UTT][ MSE = 0.2467 | LCC = 0.9370 | SRCC = 0.9374 ] +INFO:root:Average loss=0.14875829672813415 +INFO:root: +[Valid][18][UTT][ MSE = 0.2226 | LCC = 0.9382 | SRCC = 0.9386 ] +INFO:root:Average loss=0.1429619482755661 +INFO:root: +[Valid][19][UTT][ MSE = 0.2521 | LCC = 0.9343 | SRCC = 0.9341 ] +INFO:root:Average loss=0.13168285804986954 +INFO:root: +[Valid][20][UTT][ MSE = 0.2270 | LCC = 0.9405 | SRCC = 0.9418 ] +INFO:root:Average loss=0.13141158604621886 +INFO:root: +[Valid][21][UTT][ MSE = 0.2506 | LCC = 0.9384 | SRCC = 0.9386 ] +INFO:root:Average loss=0.12673782378435136 +INFO:root: +[Valid][22][UTT][ MSE = 0.2492 | LCC = 0.9367 | SRCC = 0.9366 ] +INFO:root:Average loss=0.12586901545524598 +INFO:root: +[Valid][23][UTT][ MSE = 0.2438 | LCC = 0.9397 | SRCC = 0.9395 ] +INFO:root:Average loss=0.12591976511478425 +INFO:root: +[Valid][24][UTT][ MSE = 0.2063 | LCC = 0.9443 | SRCC = 0.9445 ] +INFO:root:Average loss=0.10909104654192925 +INFO:root: +[Valid][25][UTT][ MSE = 0.2193 | LCC = 0.9431 | SRCC = 0.9430 ] +INFO:root:Average loss=0.11723135602474212 +INFO:root: +[Valid][26][UTT][ MSE = 0.2075 | LCC = 0.9424 | SRCC = 0.9426 ] +INFO:root:Average loss=0.1111500728726387 +INFO:root: +[Valid][27][UTT][ MSE = 0.2149 | LCC = 0.9441 | SRCC = 0.9457 ] +INFO:root:Average loss=0.10385756361484527 +INFO:root: +[Valid][28][UTT][ MSE = 0.2094 | LCC = 0.9418 | SRCC = 0.9418 ] +INFO:root:Average loss=0.10500214487314224 +INFO:root: +[Valid][29][UTT][ MSE = 0.2202 | LCC = 0.9450 | SRCC = 0.9443 ] +INFO:root:Average loss=0.1033059732913971 +INFO:root: +[Valid][30][UTT][ MSE = 0.2482 | LCC = 0.9421 | SRCC = 0.9418 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2271 | LCC = 0.9449 | SRCC = 0.9455 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run0/config.gin b/results_tencent/w2v2_large_xlsr53/layer10/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run0/model.pt b/results_tencent/w2v2_large_xlsr53/layer10/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6cc974a4efe8d390e63cf1ed0ddee50a664544ae --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9eea6bf9d4245a9f1431aca476ee0040c3a53b0edab08e455bef3735b2182a80 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run0/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer10/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..459f1560a03a69d75f931e4f3d5bea8cc5008549 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:45d17e8b09a9ee7c9e952b7a308880984622c12e0d5d4228784822b1338d30a0 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run0/train.log b/results_tencent/w2v2_large_xlsr53/layer10/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..dd84caf05121500a766fe8911cee161df552be0d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.49194937896728513 +INFO:root: +[Valid][0][UTT][ MSE = 0.3086 | LCC = 0.9208 | SRCC = 0.9228 ] +INFO:root:Average loss=0.18682294541597366 +INFO:root: +[Valid][1][UTT][ MSE = 0.2684 | LCC = 0.9410 | SRCC = 0.9416 ] +INFO:root:Average loss=0.15890808475017548 +INFO:root: +[Valid][2][UTT][ MSE = 0.2767 | LCC = 0.9477 | SRCC = 0.9480 ] +INFO:root:Average loss=0.13523971116542816 +INFO:root: +[Valid][3][UTT][ MSE = 0.2006 | LCC = 0.9512 | SRCC = 0.9536 ] +INFO:root:Average loss=0.12526522880792618 +INFO:root: +[Valid][4][UTT][ MSE = 0.2170 | LCC = 0.9526 | SRCC = 0.9553 ] +INFO:root:Average loss=0.11398429852724075 +INFO:root: +[Valid][5][UTT][ MSE = 0.1833 | LCC = 0.9565 | SRCC = 0.9591 ] +INFO:root:Average loss=0.11247365731000901 +INFO:root: +[Valid][6][UTT][ MSE = 0.2147 | LCC = 0.9595 | SRCC = 0.9598 ] +INFO:root:Average loss=0.10564139196276665 +INFO:root: +[Valid][7][UTT][ MSE = 0.1951 | LCC = 0.9599 | SRCC = 0.9607 ] +INFO:root:Average loss=0.09610296747088433 +INFO:root: +[Valid][8][UTT][ MSE = 0.1964 | LCC = 0.9595 | SRCC = 0.9599 ] +INFO:root:Average loss=0.09158537337183953 +INFO:root: +[Valid][9][UTT][ MSE = 0.1918 | LCC = 0.9601 | SRCC = 0.9600 ] +INFO:root:Average loss=0.0803801615536213 +INFO:root: +[Valid][10][UTT][ MSE = 0.1893 | LCC = 0.9592 | SRCC = 0.9602 ] +INFO:root:Average loss=0.07953756794333458 +INFO:root: +[Valid][11][UTT][ MSE = 0.1609 | LCC = 0.9594 | SRCC = 0.9609 ] +INFO:root:Average loss=0.0789550712108612 +INFO:root: +[Valid][12][UTT][ MSE = 0.1760 | LCC = 0.9611 | SRCC = 0.9616 ] +INFO:root:Average loss=0.07677059468626976 +INFO:root: +[Valid][13][UTT][ MSE = 0.2074 | LCC = 0.9606 | SRCC = 0.9617 ] +INFO:root:Average loss=0.07221611040830613 +INFO:root: +[Valid][14][UTT][ MSE = 0.2055 | LCC = 0.9615 | SRCC = 0.9624 ] +INFO:root:Average loss=0.06603033423423767 +INFO:root: +[Valid][15][UTT][ MSE = 0.1742 | LCC = 0.9601 | SRCC = 0.9610 ] +INFO:root:Average loss=0.06475516818463803 +INFO:root: +[Valid][16][UTT][ MSE = 0.1642 | LCC = 0.9609 | SRCC = 0.9618 ] +INFO:root:Average loss=0.06467474883794784 +INFO:root: +[Valid][17][UTT][ MSE = 0.1753 | LCC = 0.9629 | SRCC = 0.9631 ] +INFO:root:Average loss=0.05885009694099426 +INFO:root: +[Valid][18][UTT][ MSE = 0.1729 | LCC = 0.9642 | SRCC = 0.9645 ] +INFO:root:Average loss=0.05495568184554577 +INFO:root: +[Valid][19][UTT][ MSE = 0.1959 | LCC = 0.9573 | SRCC = 0.9612 ] +INFO:root:Average loss=0.05154720142483711 +INFO:root: +[Valid][20][UTT][ MSE = 0.1738 | LCC = 0.9601 | SRCC = 0.9610 ] +INFO:root:Average loss=0.05158018715679646 +INFO:root: +[Valid][21][UTT][ MSE = 0.1748 | LCC = 0.9639 | SRCC = 0.9636 ] +INFO:root:Average loss=0.05310294950008392 +INFO:root: +[Valid][22][UTT][ MSE = 0.2013 | LCC = 0.9635 | SRCC = 0.9640 ] +INFO:root:Average loss=0.04943451949954033 +INFO:root: +[Valid][23][UTT][ MSE = 0.1722 | LCC = 0.9653 | SRCC = 0.9651 ] +INFO:root:Average loss=0.051012931242585186 +INFO:root: +[Valid][24][UTT][ MSE = 0.1624 | LCC = 0.9643 | SRCC = 0.9641 ] +INFO:root:Average loss=0.046172823801636696 +INFO:root: +[Valid][25][UTT][ MSE = 0.1853 | LCC = 0.9626 | SRCC = 0.9628 ] +INFO:root:Average loss=0.04445196205377579 +INFO:root: +[Valid][26][UTT][ MSE = 0.1650 | LCC = 0.9639 | SRCC = 0.9636 ] +INFO:root:Average loss=0.0422253877222538 +INFO:root: +[Valid][27][UTT][ MSE = 0.1724 | LCC = 0.9646 | SRCC = 0.9645 ] +INFO:root:Average loss=0.03897864189743996 +INFO:root: +[Valid][28][UTT][ MSE = 0.1967 | LCC = 0.9640 | SRCC = 0.9638 ] +INFO:root:Average loss=0.037580706998705866 +INFO:root: +[Valid][29][UTT][ MSE = 0.1797 | LCC = 0.9626 | SRCC = 0.9622 ] +INFO:root:Average loss=0.03727917438745499 +INFO:root: +[Valid][30][UTT][ MSE = 0.1870 | LCC = 0.9620 | SRCC = 0.9618 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1818 | LCC = 0.9623 | SRCC = 0.9632 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run1/config.gin b/results_tencent/w2v2_large_xlsr53/layer10/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run1/model.pt b/results_tencent/w2v2_large_xlsr53/layer10/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..366c6e7f0299a7aaf71cb7db7ed988fb82213f21 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ed9c5d5986a7db1353abf34e4cda781e539732afd62bbe410018df3e30ea6b7 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run1/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer10/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..1c1b559227068876e837b0de1ef47f6f137c60bb --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e9c8e2e43816bee712f4c40861208ef46116462f09d8cf8e1297f97df6ed98c1 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run1/train.log b/results_tencent/w2v2_large_xlsr53/layer10/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..74859aedfad41d91c0ce2aa21d5d58d980f0f64d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.589994236588478 +INFO:root: +[Valid][0][UTT][ MSE = 0.3572 | LCC = 0.9165 | SRCC = 0.9190 ] +INFO:root:Average loss=0.2069593849182129 +INFO:root: +[Valid][1][UTT][ MSE = 0.2935 | LCC = 0.9391 | SRCC = 0.9404 ] +INFO:root:Average loss=0.1600075197815895 +INFO:root: +[Valid][2][UTT][ MSE = 0.2974 | LCC = 0.9482 | SRCC = 0.9497 ] +INFO:root:Average loss=0.14127581369876863 +INFO:root: +[Valid][3][UTT][ MSE = 0.2378 | LCC = 0.9515 | SRCC = 0.9531 ] +INFO:root:Average loss=0.127774620115757 +INFO:root: +[Valid][4][UTT][ MSE = 0.2332 | LCC = 0.9545 | SRCC = 0.9564 ] +INFO:root:Average loss=0.11429979968070984 +INFO:root: +[Valid][5][UTT][ MSE = 0.2469 | LCC = 0.9541 | SRCC = 0.9563 ] +INFO:root:Average loss=0.10992519119381905 +INFO:root: +[Valid][6][UTT][ MSE = 0.2135 | LCC = 0.9589 | SRCC = 0.9600 ] +INFO:root:Average loss=0.09963910520076752 +INFO:root: +[Valid][7][UTT][ MSE = 0.1993 | LCC = 0.9595 | SRCC = 0.9608 ] +INFO:root:Average loss=0.09676647412776947 +INFO:root: +[Valid][8][UTT][ MSE = 0.2235 | LCC = 0.9603 | SRCC = 0.9619 ] +INFO:root:Average loss=0.09040240812301635 +INFO:root: +[Valid][9][UTT][ MSE = 0.1931 | LCC = 0.9582 | SRCC = 0.9603 ] +INFO:root:Average loss=0.0871525117456913 +INFO:root: +[Valid][10][UTT][ MSE = 0.1850 | LCC = 0.9615 | SRCC = 0.9625 ] +INFO:root:Average loss=0.08238345539569855 +INFO:root: +[Valid][11][UTT][ MSE = 0.2168 | LCC = 0.9620 | SRCC = 0.9628 ] +INFO:root:Average loss=0.0793175051510334 +INFO:root: +[Valid][12][UTT][ MSE = 0.2122 | LCC = 0.9625 | SRCC = 0.9636 ] +INFO:root:Average loss=0.07117924946546554 +INFO:root: +[Valid][13][UTT][ MSE = 0.2356 | LCC = 0.9615 | SRCC = 0.9632 ] +INFO:root:Average loss=0.07087549650669098 +INFO:root: +[Valid][14][UTT][ MSE = 0.1999 | LCC = 0.9641 | SRCC = 0.9644 ] +INFO:root:Average loss=0.0660397481918335 +INFO:root: +[Valid][15][UTT][ MSE = 0.2202 | LCC = 0.9609 | SRCC = 0.9618 ] +INFO:root:Average loss=0.06390922117233276 +INFO:root: +[Valid][16][UTT][ MSE = 0.2491 | LCC = 0.9620 | SRCC = 0.9637 ] +INFO:root:Average loss=0.06319706566631794 +INFO:root: +[Valid][17][UTT][ MSE = 0.1820 | LCC = 0.9622 | SRCC = 0.9629 ] +INFO:root:Average loss=0.05783869808912277 +INFO:root: +[Valid][18][UTT][ MSE = 0.1833 | LCC = 0.9626 | SRCC = 0.9630 ] +INFO:root:Average loss=0.05669530290365219 +INFO:root: +[Valid][19][UTT][ MSE = 0.1772 | LCC = 0.9635 | SRCC = 0.9638 ] +INFO:root:Average loss=0.05297011962532997 +INFO:root: +[Valid][20][UTT][ MSE = 0.2076 | LCC = 0.9631 | SRCC = 0.9638 ] +INFO:root:Average loss=0.053233875960111615 +INFO:root: +[Valid][21][UTT][ MSE = 0.1877 | LCC = 0.9639 | SRCC = 0.9641 ] +INFO:root:Average loss=0.05061652390658856 +INFO:root: +[Valid][22][UTT][ MSE = 0.1824 | LCC = 0.9638 | SRCC = 0.9637 ] +INFO:root:Average loss=0.04622306197881699 +INFO:root: +[Valid][23][UTT][ MSE = 0.2094 | LCC = 0.9637 | SRCC = 0.9640 ] +INFO:root:Average loss=0.04332860912382603 +INFO:root: +[Valid][24][UTT][ MSE = 0.1823 | LCC = 0.9647 | SRCC = 0.9646 ] +INFO:root:Average loss=0.04311965523660183 +INFO:root: +[Valid][25][UTT][ MSE = 0.1955 | LCC = 0.9641 | SRCC = 0.9643 ] +INFO:root:Average loss=0.04104310449957847 +INFO:root: +[Valid][26][UTT][ MSE = 0.2362 | LCC = 0.9624 | SRCC = 0.9644 ] +INFO:root:Average loss=0.04382402950525284 +INFO:root: +[Valid][27][UTT][ MSE = 0.1713 | LCC = 0.9637 | SRCC = 0.9636 ] +INFO:root:Average loss=0.038354770347476 +INFO:root: +[Valid][28][UTT][ MSE = 0.2128 | LCC = 0.9639 | SRCC = 0.9646 ] +INFO:root:Average loss=0.04016078971326351 +INFO:root: +[Valid][29][UTT][ MSE = 0.2064 | LCC = 0.9634 | SRCC = 0.9638 ] +INFO:root:Average loss=0.035424732506275175 +INFO:root: +[Valid][30][UTT][ MSE = 0.2000 | LCC = 0.9641 | SRCC = 0.9642 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1961 | LCC = 0.9627 | SRCC = 0.9638 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run2/config.gin b/results_tencent/w2v2_large_xlsr53/layer10/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run2/model.pt b/results_tencent/w2v2_large_xlsr53/layer10/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8687fc1eba5e5f94fcb524b2fb934e23cedb6565 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:664acaeba6a179462f54408089dc15277600c3f78c9a51aec876a1dc6cf1cb5c +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run2/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer10/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..daf6df985c462e921cb76cb03e201891e76fdc92 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16292cd0f674917c95c9f94d1ef15b356f8dbef9d3485e6cb5929e289290d5c3 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run2/train.log b/results_tencent/w2v2_large_xlsr53/layer10/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d1f79cb14a4a68ffa125f8616d7bd2195e9bd84f --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5004845705032349 +INFO:root: +[Valid][0][UTT][ MSE = 0.2853 | LCC = 0.9208 | SRCC = 0.9215 ] +INFO:root:Average loss=0.19378316354751587 +INFO:root: +[Valid][1][UTT][ MSE = 0.2353 | LCC = 0.9413 | SRCC = 0.9426 ] +INFO:root:Average loss=0.16035386085510253 +INFO:root: +[Valid][2][UTT][ MSE = 0.2352 | LCC = 0.9486 | SRCC = 0.9503 ] +INFO:root:Average loss=0.13330268520116806 +INFO:root: +[Valid][3][UTT][ MSE = 0.2145 | LCC = 0.9531 | SRCC = 0.9544 ] +INFO:root:Average loss=0.12864966839551925 +INFO:root: +[Valid][4][UTT][ MSE = 0.1774 | LCC = 0.9548 | SRCC = 0.9574 ] +INFO:root:Average loss=0.11516807574033737 +INFO:root: +[Valid][5][UTT][ MSE = 0.1976 | LCC = 0.9548 | SRCC = 0.9567 ] +INFO:root:Average loss=0.10760714119672775 +INFO:root: +[Valid][6][UTT][ MSE = 0.1723 | LCC = 0.9578 | SRCC = 0.9605 ] +INFO:root:Average loss=0.10300295257568359 +INFO:root: +[Valid][7][UTT][ MSE = 0.1565 | LCC = 0.9600 | SRCC = 0.9616 ] +INFO:root:Average loss=0.09252279818058014 +INFO:root: +[Valid][8][UTT][ MSE = 0.1813 | LCC = 0.9617 | SRCC = 0.9628 ] +INFO:root:Average loss=0.09080757039785385 +INFO:root: +[Valid][9][UTT][ MSE = 0.1960 | LCC = 0.9626 | SRCC = 0.9636 ] +INFO:root:Average loss=0.0824408213198185 +INFO:root: +[Valid][10][UTT][ MSE = 0.1729 | LCC = 0.9616 | SRCC = 0.9624 ] +INFO:root:Average loss=0.08096962240338325 +INFO:root: +[Valid][11][UTT][ MSE = 0.1972 | LCC = 0.9634 | SRCC = 0.9642 ] +INFO:root:Average loss=0.07308240950107575 +INFO:root: +[Valid][12][UTT][ MSE = 0.1925 | LCC = 0.9635 | SRCC = 0.9640 ] +INFO:root:Average loss=0.07283768558502197 +INFO:root: +[Valid][13][UTT][ MSE = 0.1932 | LCC = 0.9628 | SRCC = 0.9639 ] +INFO:root:Average loss=0.07020581403374672 +INFO:root: +[Valid][14][UTT][ MSE = 0.1684 | LCC = 0.9643 | SRCC = 0.9644 ] +INFO:root:Average loss=0.06856219604611397 +INFO:root: +[Valid][15][UTT][ MSE = 0.1734 | LCC = 0.9642 | SRCC = 0.9646 ] +INFO:root:Average loss=0.06561361601948738 +INFO:root: +[Valid][16][UTT][ MSE = 0.1595 | LCC = 0.9644 | SRCC = 0.9657 ] +INFO:root:Average loss=0.06259106150269508 +INFO:root: +[Valid][17][UTT][ MSE = 0.1492 | LCC = 0.9645 | SRCC = 0.9652 ] +INFO:root:Average loss=0.05940805557370186 +INFO:root: +[Valid][18][UTT][ MSE = 0.1726 | LCC = 0.9643 | SRCC = 0.9641 ] +INFO:root:Average loss=0.05994418051838875 +INFO:root: +[Valid][19][UTT][ MSE = 0.1750 | LCC = 0.9650 | SRCC = 0.9651 ] +INFO:root:Average loss=0.05108153739571571 +INFO:root: +[Valid][20][UTT][ MSE = 0.1802 | LCC = 0.9653 | SRCC = 0.9652 ] +INFO:root:Average loss=0.05092921777069569 +INFO:root: +[Valid][21][UTT][ MSE = 0.1605 | LCC = 0.9663 | SRCC = 0.9662 ] +INFO:root:Average loss=0.04904427345097065 +INFO:root: +[Valid][22][UTT][ MSE = 0.1825 | LCC = 0.9633 | SRCC = 0.9637 ] +INFO:root:Average loss=0.04470578716695309 +INFO:root: +[Valid][23][UTT][ MSE = 0.1642 | LCC = 0.9648 | SRCC = 0.9649 ] +INFO:root:Average loss=0.04672740918397904 +INFO:root: +[Valid][24][UTT][ MSE = 0.1734 | LCC = 0.9653 | SRCC = 0.9652 ] +INFO:root:Average loss=0.04207502967119217 +INFO:root: +[Valid][25][UTT][ MSE = 0.1855 | LCC = 0.9642 | SRCC = 0.9648 ] +INFO:root:Average loss=0.043298830285668374 +INFO:root: +[Valid][26][UTT][ MSE = 0.1762 | LCC = 0.9650 | SRCC = 0.9649 ] +INFO:root:Average loss=0.03873851075768471 +INFO:root: +[Valid][27][UTT][ MSE = 0.1704 | LCC = 0.9645 | SRCC = 0.9643 ] +INFO:root:Average loss=0.0380872597694397 +INFO:root: +[Valid][28][UTT][ MSE = 0.1635 | LCC = 0.9645 | SRCC = 0.9639 ] +INFO:root:Average loss=0.03930583472549915 +INFO:root: +[Valid][29][UTT][ MSE = 0.1823 | LCC = 0.9650 | SRCC = 0.9654 ] +INFO:root:Average loss=0.038609858095645905 +INFO:root: +[Valid][30][UTT][ MSE = 0.1560 | LCC = 0.9653 | SRCC = 0.9650 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1735 | LCC = 0.9620 | SRCC = 0.9624 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run3/config.gin b/results_tencent/w2v2_large_xlsr53/layer10/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run3/model.pt b/results_tencent/w2v2_large_xlsr53/layer10/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3fbdeff660db4af05b784974b2fbcf3379ea40bc --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:faab126bbe81de4b245f65f8090eab2c91856f48545efd1a39ae08cbea80847b +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run3/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer10/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e37bceae5edba88414dcad9d3bf5bb57c3d11414 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d1beb4df16ac952e22daa943715246b65b86b8e06bf579a44e2cf93d40a91da2 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run3/train.log b/results_tencent/w2v2_large_xlsr53/layer10/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9818166fee1753c2281fb1fb9835e26fb1f58f39 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4619197800159454 +INFO:root: +[Valid][0][UTT][ MSE = 0.3093 | LCC = 0.9203 | SRCC = 0.9212 ] +INFO:root:Average loss=0.19219168764352798 +INFO:root: +[Valid][1][UTT][ MSE = 0.2858 | LCC = 0.9389 | SRCC = 0.9408 ] +INFO:root:Average loss=0.15832208460569383 +INFO:root: +[Valid][2][UTT][ MSE = 0.2539 | LCC = 0.9480 | SRCC = 0.9503 ] +INFO:root:Average loss=0.1412235242128372 +INFO:root: +[Valid][3][UTT][ MSE = 0.2359 | LCC = 0.9482 | SRCC = 0.9493 ] +INFO:root:Average loss=0.12984488946199418 +INFO:root: +[Valid][4][UTT][ MSE = 0.2044 | LCC = 0.9556 | SRCC = 0.9570 ] +INFO:root:Average loss=0.11266167736053467 +INFO:root: +[Valid][5][UTT][ MSE = 0.1987 | LCC = 0.9588 | SRCC = 0.9596 ] +INFO:root:Average loss=0.10444804883003235 +INFO:root: +[Valid][6][UTT][ MSE = 0.2289 | LCC = 0.9588 | SRCC = 0.9594 ] +INFO:root:Average loss=0.10450960999727249 +INFO:root: +[Valid][7][UTT][ MSE = 0.1975 | LCC = 0.9578 | SRCC = 0.9595 ] +INFO:root:Average loss=0.0961440417766571 +INFO:root: +[Valid][8][UTT][ MSE = 0.2274 | LCC = 0.9618 | SRCC = 0.9621 ] +INFO:root:Average loss=0.08683554244041443 +INFO:root: +[Valid][9][UTT][ MSE = 0.2108 | LCC = 0.9620 | SRCC = 0.9624 ] +INFO:root:Average loss=0.08357620358467102 +INFO:root: +[Valid][10][UTT][ MSE = 0.2075 | LCC = 0.9625 | SRCC = 0.9633 ] +INFO:root:Average loss=0.08212664771080017 +INFO:root: +[Valid][11][UTT][ MSE = 0.1915 | LCC = 0.9617 | SRCC = 0.9621 ] +INFO:root:Average loss=0.07695048427581787 +INFO:root: +[Valid][12][UTT][ MSE = 0.1958 | LCC = 0.9633 | SRCC = 0.9631 ] +INFO:root:Average loss=0.07468674343824387 +INFO:root: +[Valid][13][UTT][ MSE = 0.2176 | LCC = 0.9626 | SRCC = 0.9639 ] +INFO:root:Average loss=0.07040891137719155 +INFO:root: +[Valid][14][UTT][ MSE = 0.1963 | LCC = 0.9625 | SRCC = 0.9629 ] +INFO:root:Average loss=0.06582528941333293 +INFO:root: +[Valid][15][UTT][ MSE = 0.1868 | LCC = 0.9635 | SRCC = 0.9640 ] +INFO:root:Average loss=0.06571033602952957 +INFO:root: +[Valid][16][UTT][ MSE = 0.1780 | LCC = 0.9634 | SRCC = 0.9639 ] +INFO:root:Average loss=0.06355382731556893 +INFO:root: +[Valid][17][UTT][ MSE = 0.1673 | LCC = 0.9640 | SRCC = 0.9641 ] +INFO:root:Average loss=0.058947583615779876 +INFO:root: +[Valid][18][UTT][ MSE = 0.1560 | LCC = 0.9633 | SRCC = 0.9639 ] +INFO:root:Average loss=0.061134441196918486 +INFO:root: +[Valid][19][UTT][ MSE = 0.1896 | LCC = 0.9640 | SRCC = 0.9642 ] +INFO:root:Average loss=0.054744407385587696 +INFO:root: +[Valid][20][UTT][ MSE = 0.1990 | LCC = 0.9643 | SRCC = 0.9642 ] +INFO:root:Average loss=0.05135272155702114 +INFO:root: +[Valid][21][UTT][ MSE = 0.2059 | LCC = 0.9596 | SRCC = 0.9617 ] +INFO:root:Average loss=0.05218017259240151 +INFO:root: +[Valid][22][UTT][ MSE = 0.1688 | LCC = 0.9615 | SRCC = 0.9629 ] +INFO:root:Average loss=0.04792819321155548 +INFO:root: +[Valid][23][UTT][ MSE = 0.1753 | LCC = 0.9627 | SRCC = 0.9633 ] +INFO:root:Average loss=0.046923725098371505 +INFO:root: +[Valid][24][UTT][ MSE = 0.1783 | LCC = 0.9638 | SRCC = 0.9646 ] +INFO:root:Average loss=0.04468590821325779 +INFO:root: +[Valid][25][UTT][ MSE = 0.1686 | LCC = 0.9640 | SRCC = 0.9641 ] +INFO:root:Average loss=0.04217187985777855 +INFO:root: +[Valid][26][UTT][ MSE = 0.1947 | LCC = 0.9645 | SRCC = 0.9645 ] +INFO:root:Average loss=0.04172761204838753 +INFO:root: +[Valid][27][UTT][ MSE = 0.1707 | LCC = 0.9637 | SRCC = 0.9636 ] +INFO:root:Average loss=0.042383626386523245 +INFO:root: +[Valid][28][UTT][ MSE = 0.1797 | LCC = 0.9646 | SRCC = 0.9641 ] +INFO:root:Average loss=0.0402790439426899 +INFO:root: +[Valid][29][UTT][ MSE = 0.1847 | LCC = 0.9613 | SRCC = 0.9624 ] +INFO:root:Average loss=0.03741925297677517 +INFO:root: +[Valid][30][UTT][ MSE = 0.1726 | LCC = 0.9637 | SRCC = 0.9638 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1840 | LCC = 0.9632 | SRCC = 0.9641 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run4/config.gin b/results_tencent/w2v2_large_xlsr53/layer10/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run4/model.pt b/results_tencent/w2v2_large_xlsr53/layer10/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a908bb52b333c84964571bb2b90b20ce4d4fc68c --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ab2f37d8f4bbbee3ed880073f2b9566a4eb4dc476bc51453816c984c2fada254 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run4/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer10/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a3998e0c8fa083c554cc902db6df59a2a9e909e3 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24c04ad5f54c9ee5dbe7ac0a6ed76e1ab768ab6141e7a080820abffa433b48a9 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer10/run4/train.log b/results_tencent/w2v2_large_xlsr53/layer10/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ed6ed1508d46c055f47094ada0160e21f73e9e50 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer10/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.465950489282608 +INFO:root: +[Valid][0][UTT][ MSE = 0.3854 | LCC = 0.9214 | SRCC = 0.9221 ] +INFO:root:Average loss=0.19537348246574401 +INFO:root: +[Valid][1][UTT][ MSE = 0.2557 | LCC = 0.9408 | SRCC = 0.9416 ] +INFO:root:Average loss=0.15982520121335983 +INFO:root: +[Valid][2][UTT][ MSE = 0.2041 | LCC = 0.9476 | SRCC = 0.9496 ] +INFO:root:Average loss=0.1521505823135376 +INFO:root: +[Valid][3][UTT][ MSE = 0.2216 | LCC = 0.9499 | SRCC = 0.9540 ] +INFO:root:Average loss=0.12343611907958985 +INFO:root: +[Valid][4][UTT][ MSE = 0.1970 | LCC = 0.9536 | SRCC = 0.9565 ] +INFO:root:Average loss=0.1139386162161827 +INFO:root: +[Valid][5][UTT][ MSE = 0.1781 | LCC = 0.9566 | SRCC = 0.9595 ] +INFO:root:Average loss=0.10592125007510185 +INFO:root: +[Valid][6][UTT][ MSE = 0.1894 | LCC = 0.9567 | SRCC = 0.9586 ] +INFO:root:Average loss=0.10577095153927803 +INFO:root: +[Valid][7][UTT][ MSE = 0.2088 | LCC = 0.9597 | SRCC = 0.9619 ] +INFO:root:Average loss=0.10063423156738281 +INFO:root: +[Valid][8][UTT][ MSE = 0.1922 | LCC = 0.9563 | SRCC = 0.9597 ] +INFO:root:Average loss=0.09149724236130714 +INFO:root: +[Valid][9][UTT][ MSE = 0.1796 | LCC = 0.9618 | SRCC = 0.9639 ] +INFO:root:Average loss=0.08805279847979545 +INFO:root: +[Valid][10][UTT][ MSE = 0.2115 | LCC = 0.9623 | SRCC = 0.9636 ] +INFO:root:Average loss=0.08146461459994316 +INFO:root: +[Valid][11][UTT][ MSE = 0.1803 | LCC = 0.9628 | SRCC = 0.9642 ] +INFO:root:Average loss=0.07503565132617951 +INFO:root: +[Valid][12][UTT][ MSE = 0.1819 | LCC = 0.9630 | SRCC = 0.9651 ] +INFO:root:Average loss=0.07461747214198113 +INFO:root: +[Valid][13][UTT][ MSE = 0.1797 | LCC = 0.9634 | SRCC = 0.9646 ] +INFO:root:Average loss=0.07024546605348587 +INFO:root: +[Valid][14][UTT][ MSE = 0.1968 | LCC = 0.9647 | SRCC = 0.9652 ] +INFO:root:Average loss=0.06656740188598632 +INFO:root: +[Valid][15][UTT][ MSE = 0.1946 | LCC = 0.9626 | SRCC = 0.9640 ] +INFO:root:Average loss=0.064614864975214 +INFO:root: +[Valid][16][UTT][ MSE = 0.1767 | LCC = 0.9635 | SRCC = 0.9640 ] +INFO:root:Average loss=0.06020263250172138 +INFO:root: +[Valid][17][UTT][ MSE = 0.1900 | LCC = 0.9622 | SRCC = 0.9635 ] +INFO:root:Average loss=0.05693288603425026 +INFO:root: +[Valid][18][UTT][ MSE = 0.1701 | LCC = 0.9612 | SRCC = 0.9635 ] +INFO:root:Average loss=0.05698802652955055 +INFO:root: +[Valid][19][UTT][ MSE = 0.1945 | LCC = 0.9645 | SRCC = 0.9654 ] +INFO:root:Average loss=0.05189521300792694 +INFO:root: +[Valid][20][UTT][ MSE = 0.1775 | LCC = 0.9653 | SRCC = 0.9657 ] +INFO:root:Average loss=0.049194841355085374 +INFO:root: +[Valid][21][UTT][ MSE = 0.1640 | LCC = 0.9644 | SRCC = 0.9651 ] +INFO:root:Average loss=0.0515929659307003 +INFO:root: +[Valid][22][UTT][ MSE = 0.1927 | LCC = 0.9653 | SRCC = 0.9659 ] +INFO:root:Average loss=0.046298128947615626 +INFO:root: +[Valid][23][UTT][ MSE = 0.2236 | LCC = 0.9627 | SRCC = 0.9635 ] +INFO:root:Average loss=0.05064973932504654 +INFO:root: +[Valid][24][UTT][ MSE = 0.2162 | LCC = 0.9647 | SRCC = 0.9659 ] +INFO:root:Average loss=0.04338656595349312 +INFO:root: +[Valid][25][UTT][ MSE = 0.1678 | LCC = 0.9644 | SRCC = 0.9656 ] +INFO:root:Average loss=0.04170184341073036 +INFO:root: +[Valid][26][UTT][ MSE = 0.1768 | LCC = 0.9646 | SRCC = 0.9651 ] +INFO:root:Average loss=0.04196528844535351 +INFO:root: +[Valid][27][UTT][ MSE = 0.1860 | LCC = 0.9623 | SRCC = 0.9638 ] +INFO:root:Average loss=0.04091385701298714 +INFO:root: +[Valid][28][UTT][ MSE = 0.1856 | LCC = 0.9647 | SRCC = 0.9646 ] +INFO:root:Average loss=0.03732149773836136 +INFO:root: +[Valid][29][UTT][ MSE = 0.1862 | LCC = 0.9658 | SRCC = 0.9660 ] +INFO:root:Average loss=0.03750532639026642 +INFO:root: +[Valid][30][UTT][ MSE = 0.1692 | LCC = 0.9645 | SRCC = 0.9649 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1973 | LCC = 0.9624 | SRCC = 0.9639 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run0/config.gin b/results_tencent/w2v2_large_xlsr53/layer12/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run0/model.pt b/results_tencent/w2v2_large_xlsr53/layer12/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a796a4cf35513be3f652f2e187d6378e540b6839 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e0a494e0fa71112929b385bbea510f341be6d5db70f0c4b39fcf30f82e75d20 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run0/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer12/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..829c2cb6c88a3a872ad5ef4a1b3622dbb7b28931 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8928061b66b4a353b45ae0b6d4e14b4594908f9e75f6f246bae3131e5d57b776 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run0/train.log b/results_tencent/w2v2_large_xlsr53/layer12/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a5772a83918edeef6627e24d42e0a33e93456864 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5567583624124527 +INFO:root: +[Valid][0][UTT][ MSE = 0.3414 | LCC = 0.9164 | SRCC = 0.9200 ] +INFO:root:Average loss=0.20373304319381713 +INFO:root: +[Valid][1][UTT][ MSE = 0.2983 | LCC = 0.9392 | SRCC = 0.9426 ] +INFO:root:Average loss=0.1548291957974434 +INFO:root: +[Valid][2][UTT][ MSE = 0.2250 | LCC = 0.9490 | SRCC = 0.9504 ] +INFO:root:Average loss=0.13630943393707276 +INFO:root: +[Valid][3][UTT][ MSE = 0.1978 | LCC = 0.9511 | SRCC = 0.9531 ] +INFO:root:Average loss=0.12468462973833085 +INFO:root: +[Valid][4][UTT][ MSE = 0.1976 | LCC = 0.9558 | SRCC = 0.9569 ] +INFO:root:Average loss=0.12072055789828301 +INFO:root: +[Valid][5][UTT][ MSE = 0.2169 | LCC = 0.9566 | SRCC = 0.9579 ] +INFO:root:Average loss=0.10517207235097885 +INFO:root: +[Valid][6][UTT][ MSE = 0.2368 | LCC = 0.9570 | SRCC = 0.9584 ] +INFO:root:Average loss=0.10046509537100792 +INFO:root: +[Valid][7][UTT][ MSE = 0.2159 | LCC = 0.9584 | SRCC = 0.9602 ] +INFO:root:Average loss=0.0953511278629303 +INFO:root: +[Valid][8][UTT][ MSE = 0.2173 | LCC = 0.9602 | SRCC = 0.9612 ] +INFO:root:Average loss=0.09196847182512283 +INFO:root: +[Valid][9][UTT][ MSE = 0.1966 | LCC = 0.9618 | SRCC = 0.9621 ] +INFO:root:Average loss=0.09047107237577438 +INFO:root: +[Valid][10][UTT][ MSE = 0.2196 | LCC = 0.9604 | SRCC = 0.9609 ] +INFO:root:Average loss=0.07860218036174774 +INFO:root: +[Valid][11][UTT][ MSE = 0.1910 | LCC = 0.9621 | SRCC = 0.9620 ] +INFO:root:Average loss=0.07787407892942429 +INFO:root: +[Valid][12][UTT][ MSE = 0.2060 | LCC = 0.9621 | SRCC = 0.9624 ] +INFO:root:Average loss=0.07617492499947548 +INFO:root: +[Valid][13][UTT][ MSE = 0.1597 | LCC = 0.9618 | SRCC = 0.9620 ] +INFO:root:Average loss=0.07052921777963639 +INFO:root: +[Valid][14][UTT][ MSE = 0.1901 | LCC = 0.9616 | SRCC = 0.9616 ] +INFO:root:Average loss=0.0674685334265232 +INFO:root: +[Valid][15][UTT][ MSE = 0.1960 | LCC = 0.9616 | SRCC = 0.9621 ] +INFO:root:Average loss=0.06386832761764526 +INFO:root: +[Valid][16][UTT][ MSE = 0.1789 | LCC = 0.9633 | SRCC = 0.9631 ] +INFO:root:Average loss=0.058927383348345755 +INFO:root: +[Valid][17][UTT][ MSE = 0.1830 | LCC = 0.9629 | SRCC = 0.9628 ] +INFO:root:Average loss=0.05772418797016144 +INFO:root: +[Valid][18][UTT][ MSE = 0.1789 | LCC = 0.9630 | SRCC = 0.9631 ] +INFO:root:Average loss=0.05818222707509994 +INFO:root: +[Valid][19][UTT][ MSE = 0.1704 | LCC = 0.9615 | SRCC = 0.9613 ] +INFO:root:Average loss=0.05442635345458984 +INFO:root: +[Valid][20][UTT][ MSE = 0.1865 | LCC = 0.9627 | SRCC = 0.9629 ] +INFO:root:Average loss=0.048616977304220196 +INFO:root: +[Valid][21][UTT][ MSE = 0.1915 | LCC = 0.9637 | SRCC = 0.9632 ] +INFO:root:Average loss=0.04816257543861866 +INFO:root: +[Valid][22][UTT][ MSE = 0.1994 | LCC = 0.9624 | SRCC = 0.9624 ] +INFO:root:Average loss=0.047697880774736405 +INFO:root: +[Valid][23][UTT][ MSE = 0.2206 | LCC = 0.9607 | SRCC = 0.9613 ] +INFO:root:Average loss=0.04338037821650505 +INFO:root: +[Valid][24][UTT][ MSE = 0.1722 | LCC = 0.9599 | SRCC = 0.9600 ] +INFO:root:Average loss=0.044218697160482406 +INFO:root: +[Valid][25][UTT][ MSE = 0.1944 | LCC = 0.9622 | SRCC = 0.9621 ] +INFO:root:Average loss=0.04255992244184017 +INFO:root: +[Valid][26][UTT][ MSE = 0.1753 | LCC = 0.9630 | SRCC = 0.9626 ] +INFO:root:Average loss=0.04405056242644787 +INFO:root: +[Valid][27][UTT][ MSE = 0.1997 | LCC = 0.9604 | SRCC = 0.9610 ] +INFO:root:Average loss=0.03844554027915001 +INFO:root: +[Valid][28][UTT][ MSE = 0.1980 | LCC = 0.9605 | SRCC = 0.9605 ] +INFO:root:Average loss=0.04030302755534649 +INFO:root: +[Valid][29][UTT][ MSE = 0.1645 | LCC = 0.9630 | SRCC = 0.9627 ] +INFO:root:Average loss=0.03575048178434372 +INFO:root: +[Valid][30][UTT][ MSE = 0.1858 | LCC = 0.9622 | SRCC = 0.9619 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1957 | LCC = 0.9631 | SRCC = 0.9644 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run1/config.gin b/results_tencent/w2v2_large_xlsr53/layer12/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run1/model.pt b/results_tencent/w2v2_large_xlsr53/layer12/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..22610c6a1aa562d573f22999afafe4c64298c278 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:526c80ce78d8ec025beff771143a2f6593fb7eeacc59398056792bd19f8412ec +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run1/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer12/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..21647003d8761a51d9c474aecfa74e6c7d9d9572 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b434438b88e2702df1a38b05be0f52fb402344e63611bd6e060bff4c7e3d0d3 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run1/train.log b/results_tencent/w2v2_large_xlsr53/layer12/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f93bdbfb55adb8f4106ed6657184772e1dec9456 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4395606951713562 +INFO:root: +[Valid][0][UTT][ MSE = 0.2899 | LCC = 0.9238 | SRCC = 0.9254 ] +INFO:root:Average loss=0.2021771702170372 +INFO:root: +[Valid][1][UTT][ MSE = 0.2597 | LCC = 0.9406 | SRCC = 0.9415 ] +INFO:root:Average loss=0.1639694768190384 +INFO:root: +[Valid][2][UTT][ MSE = 0.2468 | LCC = 0.9491 | SRCC = 0.9513 ] +INFO:root:Average loss=0.14393633776903153 +INFO:root: +[Valid][3][UTT][ MSE = 0.2397 | LCC = 0.9522 | SRCC = 0.9541 ] +INFO:root:Average loss=0.1240381845831871 +INFO:root: +[Valid][4][UTT][ MSE = 0.2554 | LCC = 0.9539 | SRCC = 0.9565 ] +INFO:root:Average loss=0.12195248371362687 +INFO:root: +[Valid][5][UTT][ MSE = 0.2070 | LCC = 0.9576 | SRCC = 0.9585 ] +INFO:root:Average loss=0.11117646059393883 +INFO:root: +[Valid][6][UTT][ MSE = 0.2490 | LCC = 0.9551 | SRCC = 0.9583 ] +INFO:root:Average loss=0.10706942820549012 +INFO:root: +[Valid][7][UTT][ MSE = 0.2059 | LCC = 0.9589 | SRCC = 0.9608 ] +INFO:root:Average loss=0.09304742768406868 +INFO:root: +[Valid][8][UTT][ MSE = 0.1743 | LCC = 0.9599 | SRCC = 0.9616 ] +INFO:root:Average loss=0.09199388960003853 +INFO:root: +[Valid][9][UTT][ MSE = 0.1869 | LCC = 0.9607 | SRCC = 0.9618 ] +INFO:root:Average loss=0.08495884078741074 +INFO:root: +[Valid][10][UTT][ MSE = 0.1710 | LCC = 0.9624 | SRCC = 0.9635 ] +INFO:root:Average loss=0.08070180767774582 +INFO:root: +[Valid][11][UTT][ MSE = 0.2068 | LCC = 0.9631 | SRCC = 0.9635 ] +INFO:root:Average loss=0.07966940566897392 +INFO:root: +[Valid][12][UTT][ MSE = 0.1945 | LCC = 0.9626 | SRCC = 0.9635 ] +INFO:root:Average loss=0.08117986279726029 +INFO:root: +[Valid][13][UTT][ MSE = 0.1964 | LCC = 0.9588 | SRCC = 0.9621 ] +INFO:root:Average loss=0.07082140362262726 +INFO:root: +[Valid][14][UTT][ MSE = 0.1819 | LCC = 0.9609 | SRCC = 0.9630 ] +INFO:root:Average loss=0.06665449929237366 +INFO:root: +[Valid][15][UTT][ MSE = 0.1833 | LCC = 0.9637 | SRCC = 0.9645 ] +INFO:root:Average loss=0.06787472060322762 +INFO:root: +[Valid][16][UTT][ MSE = 0.1811 | LCC = 0.9637 | SRCC = 0.9636 ] +INFO:root:Average loss=0.06693326656520367 +INFO:root: +[Valid][17][UTT][ MSE = 0.1939 | LCC = 0.9618 | SRCC = 0.9628 ] +INFO:root:Average loss=0.06311931253969669 +INFO:root: +[Valid][18][UTT][ MSE = 0.1948 | LCC = 0.9629 | SRCC = 0.9642 ] +INFO:root:Average loss=0.05628118495643139 +INFO:root: +[Valid][19][UTT][ MSE = 0.1674 | LCC = 0.9627 | SRCC = 0.9635 ] +INFO:root:Average loss=0.054987579494714736 +INFO:root: +[Valid][20][UTT][ MSE = 0.1673 | LCC = 0.9634 | SRCC = 0.9634 ] +INFO:root:Average loss=0.05215098501741886 +INFO:root: +[Valid][21][UTT][ MSE = 0.1867 | LCC = 0.9628 | SRCC = 0.9633 ] +INFO:root:Average loss=0.051539886236190796 +INFO:root: +[Valid][22][UTT][ MSE = 0.2010 | LCC = 0.9640 | SRCC = 0.9640 ] +INFO:root:Average loss=0.048934295758605 +INFO:root: +[Valid][23][UTT][ MSE = 0.1785 | LCC = 0.9630 | SRCC = 0.9641 ] +INFO:root:Average loss=0.04694494904577732 +INFO:root: +[Valid][24][UTT][ MSE = 0.1951 | LCC = 0.9637 | SRCC = 0.9646 ] +INFO:root:Average loss=0.04339496164023876 +INFO:root: +[Valid][25][UTT][ MSE = 0.1648 | LCC = 0.9641 | SRCC = 0.9644 ] +INFO:root:Average loss=0.04294469563663006 +INFO:root: +[Valid][26][UTT][ MSE = 0.1899 | LCC = 0.9618 | SRCC = 0.9629 ] +INFO:root:Average loss=0.04084075547754765 +INFO:root: +[Valid][27][UTT][ MSE = 0.2050 | LCC = 0.9636 | SRCC = 0.9636 ] +INFO:root:Average loss=0.03963747139275074 +INFO:root: +[Valid][28][UTT][ MSE = 0.1652 | LCC = 0.9641 | SRCC = 0.9640 ] +INFO:root:Average loss=0.03832261748611927 +INFO:root: +[Valid][29][UTT][ MSE = 0.2019 | LCC = 0.9630 | SRCC = 0.9634 ] +INFO:root:Average loss=0.03636516426503658 +INFO:root: +[Valid][30][UTT][ MSE = 0.1792 | LCC = 0.9637 | SRCC = 0.9634 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1736 | LCC = 0.9622 | SRCC = 0.9630 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run2/config.gin b/results_tencent/w2v2_large_xlsr53/layer12/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run2/model.pt b/results_tencent/w2v2_large_xlsr53/layer12/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..183981d01e73655b1cb1032ca24cfc1b07c795ed --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6323b1066115ef8446cbea20b9a2ee46bcd5b806dc16b54f66c204af97b55523 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run2/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer12/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2342a7fb7af4e6c549cf59b17da230b8506cc1db --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f41abea6c2337deaaf97693b2a1b64ede8853977f2eaecfc8243f1dd9d163592 +size 3404998 diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run2/train.log b/results_tencent/w2v2_large_xlsr53/layer12/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6c06bfbe8ced878e189219cb30ab543356e6fecf --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4489307544827461 +INFO:root: +[Valid][0][UTT][ MSE = 0.3014 | LCC = 0.9222 | SRCC = 0.9280 ] +INFO:root:Average loss=0.18823837625980377 +INFO:root: +[Valid][1][UTT][ MSE = 0.2694 | LCC = 0.9336 | SRCC = 0.9395 ] +INFO:root:Average loss=0.15721866303682328 +INFO:root: +[Valid][2][UTT][ MSE = 0.2268 | LCC = 0.9444 | SRCC = 0.9491 ] +INFO:root:Average loss=0.1419783883690834 +INFO:root: +[Valid][3][UTT][ MSE = 0.2159 | LCC = 0.9510 | SRCC = 0.9538 ] +INFO:root:Average loss=0.1279665949344635 +INFO:root: +[Valid][4][UTT][ MSE = 0.1910 | LCC = 0.9534 | SRCC = 0.9562 ] +INFO:root:Average loss=0.12001747462153435 +INFO:root: +[Valid][5][UTT][ MSE = 0.2556 | LCC = 0.9504 | SRCC = 0.9522 ] +INFO:root:Average loss=0.11637951990962028 +INFO:root: +[Valid][6][UTT][ MSE = 0.2090 | LCC = 0.9579 | SRCC = 0.9605 ] +INFO:root:Average loss=0.10813490188121795 +INFO:root: +[Valid][7][UTT][ MSE = 0.1900 | LCC = 0.9580 | SRCC = 0.9597 ] +INFO:root:Average loss=0.09641235807538033 +INFO:root: +[Valid][8][UTT][ MSE = 0.1961 | LCC = 0.9574 | SRCC = 0.9590 ] +INFO:root:Average loss=0.09330880147218704 +INFO:root: +[Valid][9][UTT][ MSE = 0.1848 | LCC = 0.9568 | SRCC = 0.9594 ] +INFO:root:Average loss=0.09063287034630775 +INFO:root: +[Valid][10][UTT][ MSE = 0.1685 | LCC = 0.9612 | SRCC = 0.9624 ] +INFO:root:Average loss=0.0828113350570202 +INFO:root: +[Valid][11][UTT][ MSE = 0.2086 | LCC = 0.9620 | SRCC = 0.9621 ] +INFO:root:Average loss=0.08067986425757408 +INFO:root: +[Valid][12][UTT][ MSE = 0.1975 | LCC = 0.9608 | SRCC = 0.9613 ] +INFO:root:Average loss=0.07544185760617256 +INFO:root: +[Valid][13][UTT][ MSE = 0.1873 | LCC = 0.9631 | SRCC = 0.9633 ] +INFO:root:Average loss=0.0715216031074524 +INFO:root: +[Valid][14][UTT][ MSE = 0.1887 | LCC = 0.9637 | SRCC = 0.9639 ] +INFO:root:Average loss=0.07155561542510987 +INFO:root: +[Valid][15][UTT][ MSE = 0.1681 | LCC = 0.9609 | SRCC = 0.9629 ] +INFO:root:Average loss=0.06824248865246772 +INFO:root: +[Valid][16][UTT][ MSE = 0.1670 | LCC = 0.9630 | SRCC = 0.9639 ] +INFO:root:Average loss=0.060062739670276644 +INFO:root: +[Valid][17][UTT][ MSE = 0.1751 | LCC = 0.9634 | SRCC = 0.9634 ] +INFO:root:Average loss=0.0599928814470768 +INFO:root: +[Valid][18][UTT][ MSE = 0.2095 | LCC = 0.9628 | SRCC = 0.9631 ] +INFO:root:Average loss=0.0550942288339138 +INFO:root: +[Valid][19][UTT][ MSE = 0.1634 | LCC = 0.9639 | SRCC = 0.9639 ] +INFO:root:Average loss=0.05720131346583367 +INFO:root: +[Valid][20][UTT][ MSE = 0.1599 | LCC = 0.9570 | SRCC = 0.9616 ] +INFO:root:Average loss=0.05501071906089783 +INFO:root: +[Valid][21][UTT][ MSE = 0.1928 | LCC = 0.9636 | SRCC = 0.9634 ] +INFO:root:Average loss=0.05214022044837475 +INFO:root: +[Valid][22][UTT][ MSE = 0.1839 | LCC = 0.9638 | SRCC = 0.9636 ] +INFO:root:Average loss=0.04848587277531624 +INFO:root: +[Valid][23][UTT][ MSE = 0.1980 | LCC = 0.9629 | SRCC = 0.9631 ] +INFO:root:Average loss=0.04591811585426331 +INFO:root: +[Valid][24][UTT][ MSE = 0.1865 | LCC = 0.9610 | SRCC = 0.9625 ] +INFO:root:Average loss=0.04732720021903515 +INFO:root: +[Valid][25][UTT][ MSE = 0.1792 | LCC = 0.9650 | SRCC = 0.9649 ] +INFO:root:Average loss=0.04680764399468899 +INFO:root: +[Valid][26][UTT][ MSE = 0.1611 | LCC = 0.9599 | SRCC = 0.9607 ] +INFO:root:Average loss=0.043694478139281276 +INFO:root: +[Valid][27][UTT][ MSE = 0.1565 | LCC = 0.9633 | SRCC = 0.9633 ] +INFO:root:Average loss=0.0398775297999382 +INFO:root: +[Valid][28][UTT][ MSE = 0.1584 | LCC = 0.9626 | SRCC = 0.9636 ] +INFO:root:Average loss=0.03746545796096325 +INFO:root: +[Valid][29][UTT][ MSE = 0.1671 | LCC = 0.9634 | SRCC = 0.9637 ] +INFO:root:Average loss=0.03941305935382843 +INFO:root: +[Valid][30][UTT][ MSE = 0.1726 | LCC = 0.9651 | SRCC = 0.9650 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1794 | LCC = 0.9651 | SRCC = 0.9660 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run3/config.gin b/results_tencent/w2v2_large_xlsr53/layer12/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run3/model.pt b/results_tencent/w2v2_large_xlsr53/layer12/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e5583e32c1ea030b051883052acc43c7e4bda098 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:69e1da0f4b4a289af8d73ee106626f645e48754e2de00fbe5fd767ab56ae2980 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run3/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer12/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..00d0efa9417412ea93f669689bc06dae35a4bac2 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:33ba9e884562b971f56be41a69a10b9298d3ec6ac59d6202fedbfdbf328e6f11 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run3/train.log b/results_tencent/w2v2_large_xlsr53/layer12/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..df143ef215c655eefa3b0a7a0c1269ba642b5bf4 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5163194895982742 +INFO:root: +[Valid][0][UTT][ MSE = 0.3566 | LCC = 0.9229 | SRCC = 0.9241 ] +INFO:root:Average loss=0.20165350365638732 +INFO:root: +[Valid][1][UTT][ MSE = 0.3154 | LCC = 0.9395 | SRCC = 0.9409 ] +INFO:root:Average loss=0.15779387694597244 +INFO:root: +[Valid][2][UTT][ MSE = 0.2568 | LCC = 0.9469 | SRCC = 0.9482 ] +INFO:root:Average loss=0.15209601438045503 +INFO:root: +[Valid][3][UTT][ MSE = 0.2676 | LCC = 0.9506 | SRCC = 0.9510 ] +INFO:root:Average loss=0.12459499698877334 +INFO:root: +[Valid][4][UTT][ MSE = 0.2890 | LCC = 0.9551 | SRCC = 0.9577 ] +INFO:root:Average loss=0.11883345621824265 +INFO:root: +[Valid][5][UTT][ MSE = 0.2107 | LCC = 0.9563 | SRCC = 0.9577 ] +INFO:root:Average loss=0.11028675228357315 +INFO:root: +[Valid][6][UTT][ MSE = 0.2264 | LCC = 0.9594 | SRCC = 0.9601 ] +INFO:root:Average loss=0.10264493155479432 +INFO:root: +[Valid][7][UTT][ MSE = 0.1907 | LCC = 0.9583 | SRCC = 0.9601 ] +INFO:root:Average loss=0.09602992951869965 +INFO:root: +[Valid][8][UTT][ MSE = 0.1829 | LCC = 0.9603 | SRCC = 0.9618 ] +INFO:root:Average loss=0.09105345648527145 +INFO:root: +[Valid][9][UTT][ MSE = 0.2583 | LCC = 0.9594 | SRCC = 0.9613 ] +INFO:root:Average loss=0.0860619255900383 +INFO:root: +[Valid][10][UTT][ MSE = 0.1958 | LCC = 0.9607 | SRCC = 0.9619 ] +INFO:root:Average loss=0.08212137949466705 +INFO:root: +[Valid][11][UTT][ MSE = 0.2329 | LCC = 0.9613 | SRCC = 0.9619 ] +INFO:root:Average loss=0.07695173737406731 +INFO:root: +[Valid][12][UTT][ MSE = 0.2239 | LCC = 0.9625 | SRCC = 0.9629 ] +INFO:root:Average loss=0.07574874268472195 +INFO:root: +[Valid][13][UTT][ MSE = 0.2246 | LCC = 0.9616 | SRCC = 0.9617 ] +INFO:root:Average loss=0.07692946231365204 +INFO:root: +[Valid][14][UTT][ MSE = 0.1913 | LCC = 0.9634 | SRCC = 0.9635 ] +INFO:root:Average loss=0.06718684265017509 +INFO:root: +[Valid][15][UTT][ MSE = 0.1943 | LCC = 0.9620 | SRCC = 0.9616 ] +INFO:root:Average loss=0.06965228191018105 +INFO:root: +[Valid][16][UTT][ MSE = 0.1986 | LCC = 0.9635 | SRCC = 0.9641 ] +INFO:root:Average loss=0.06135375554859638 +INFO:root: +[Valid][17][UTT][ MSE = 0.1940 | LCC = 0.9605 | SRCC = 0.9605 ] +INFO:root:Average loss=0.06132237443327904 +INFO:root: +[Valid][18][UTT][ MSE = 0.2027 | LCC = 0.9610 | SRCC = 0.9614 ] +INFO:root:Average loss=0.05549145331978798 +INFO:root: +[Valid][19][UTT][ MSE = 0.2148 | LCC = 0.9615 | SRCC = 0.9621 ] +INFO:root:Average loss=0.05570977082848549 +INFO:root: +[Valid][20][UTT][ MSE = 0.1956 | LCC = 0.9606 | SRCC = 0.9621 ] +INFO:root:Average loss=0.05518102541565895 +INFO:root: +[Valid][21][UTT][ MSE = 0.1769 | LCC = 0.9626 | SRCC = 0.9628 ] +INFO:root:Average loss=0.05450099161267281 +INFO:root: +[Valid][22][UTT][ MSE = 0.1936 | LCC = 0.9632 | SRCC = 0.9634 ] +INFO:root:Average loss=0.0488316802084446 +INFO:root: +[Valid][23][UTT][ MSE = 0.2113 | LCC = 0.9626 | SRCC = 0.9626 ] +INFO:root:Average loss=0.047034714207053185 +INFO:root: +[Valid][24][UTT][ MSE = 0.1828 | LCC = 0.9638 | SRCC = 0.9632 ] +INFO:root:Average loss=0.04249962149560452 +INFO:root: +[Valid][25][UTT][ MSE = 0.1746 | LCC = 0.9615 | SRCC = 0.9626 ] +INFO:root:Average loss=0.044381737157702446 +INFO:root: +[Valid][26][UTT][ MSE = 0.1793 | LCC = 0.9619 | SRCC = 0.9625 ] +INFO:root:Average loss=0.040839234933257106 +INFO:root: +[Valid][27][UTT][ MSE = 0.1930 | LCC = 0.9633 | SRCC = 0.9631 ] +INFO:root:Average loss=0.03946636483073235 +INFO:root: +[Valid][28][UTT][ MSE = 0.1876 | LCC = 0.9616 | SRCC = 0.9614 ] +INFO:root:Average loss=0.03832454067468643 +INFO:root: +[Valid][29][UTT][ MSE = 0.2145 | LCC = 0.9622 | SRCC = 0.9623 ] +INFO:root:Average loss=0.03666528575122356 +INFO:root: +[Valid][30][UTT][ MSE = 0.1868 | LCC = 0.9626 | SRCC = 0.9624 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1926 | LCC = 0.9623 | SRCC = 0.9631 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run4/config.gin b/results_tencent/w2v2_large_xlsr53/layer12/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run4/model.pt b/results_tencent/w2v2_large_xlsr53/layer12/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6c406d9a4fdf35cef40f1b1e5e3ba75ec02f6ee4 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7df1cb94e7659b5c277ed9d8ea9816f1af8949b1f12e21787e4af83d8881a408 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run4/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer12/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b08a7ef4ecaf80eaec524e478b171485e4ca0d98 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b6318d9d138eb97a7dae89aa26315e3c53e85783f0edb8746d56566324b1cfd8 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer12/run4/train.log b/results_tencent/w2v2_large_xlsr53/layer12/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..af17f24ebaf53f9a1c2e4c86bce84864656404d5 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer12/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.521131572008133 +INFO:root: +[Valid][0][UTT][ MSE = 0.3366 | LCC = 0.9200 | SRCC = 0.9215 ] +INFO:root:Average loss=0.20839888477325438 +INFO:root: +[Valid][1][UTT][ MSE = 0.2655 | LCC = 0.9393 | SRCC = 0.9413 ] +INFO:root:Average loss=0.159680739402771 +INFO:root: +[Valid][2][UTT][ MSE = 0.2333 | LCC = 0.9498 | SRCC = 0.9507 ] +INFO:root:Average loss=0.1376642231941223 +INFO:root: +[Valid][3][UTT][ MSE = 0.2047 | LCC = 0.9530 | SRCC = 0.9543 ] +INFO:root:Average loss=0.1286624995470047 +INFO:root: +[Valid][4][UTT][ MSE = 0.2603 | LCC = 0.9520 | SRCC = 0.9546 ] +INFO:root:Average loss=0.12319485089182854 +INFO:root: +[Valid][5][UTT][ MSE = 0.2020 | LCC = 0.9579 | SRCC = 0.9591 ] +INFO:root:Average loss=0.11017193618416786 +INFO:root: +[Valid][6][UTT][ MSE = 0.2419 | LCC = 0.9578 | SRCC = 0.9599 ] +INFO:root:Average loss=0.10586695086956024 +INFO:root: +[Valid][7][UTT][ MSE = 0.2280 | LCC = 0.9587 | SRCC = 0.9600 ] +INFO:root:Average loss=0.09641392922401429 +INFO:root: +[Valid][8][UTT][ MSE = 0.1797 | LCC = 0.9606 | SRCC = 0.9617 ] +INFO:root:Average loss=0.09203581431508065 +INFO:root: +[Valid][9][UTT][ MSE = 0.2226 | LCC = 0.9619 | SRCC = 0.9627 ] +INFO:root:Average loss=0.08256971630454063 +INFO:root: +[Valid][10][UTT][ MSE = 0.2250 | LCC = 0.9610 | SRCC = 0.9617 ] +INFO:root:Average loss=0.08494431799650193 +INFO:root: +[Valid][11][UTT][ MSE = 0.2390 | LCC = 0.9621 | SRCC = 0.9628 ] +INFO:root:Average loss=0.07706684476137161 +INFO:root: +[Valid][12][UTT][ MSE = 0.2201 | LCC = 0.9617 | SRCC = 0.9629 ] +INFO:root:Average loss=0.07648142418265343 +INFO:root: +[Valid][13][UTT][ MSE = 0.2238 | LCC = 0.9615 | SRCC = 0.9628 ] +INFO:root:Average loss=0.07107394576072693 +INFO:root: +[Valid][14][UTT][ MSE = 0.1766 | LCC = 0.9627 | SRCC = 0.9624 ] +INFO:root:Average loss=0.0671486259996891 +INFO:root: +[Valid][15][UTT][ MSE = 0.1658 | LCC = 0.9630 | SRCC = 0.9634 ] +INFO:root:Average loss=0.06372616580128669 +INFO:root: +[Valid][16][UTT][ MSE = 0.1834 | LCC = 0.9641 | SRCC = 0.9642 ] +INFO:root:Average loss=0.06230780509114266 +INFO:root: +[Valid][17][UTT][ MSE = 0.1944 | LCC = 0.9627 | SRCC = 0.9630 ] +INFO:root:Average loss=0.05620904602110386 +INFO:root: +[Valid][18][UTT][ MSE = 0.1696 | LCC = 0.9641 | SRCC = 0.9639 ] +INFO:root:Average loss=0.056338201105594635 +INFO:root: +[Valid][19][UTT][ MSE = 0.1931 | LCC = 0.9644 | SRCC = 0.9645 ] +INFO:root:Average loss=0.053728263795375826 +INFO:root: +[Valid][20][UTT][ MSE = 0.2045 | LCC = 0.9606 | SRCC = 0.9620 ] +INFO:root:Average loss=0.052487892299890516 +INFO:root: +[Valid][21][UTT][ MSE = 0.1823 | LCC = 0.9643 | SRCC = 0.9645 ] +INFO:root:Average loss=0.048717605367302894 +INFO:root: +[Valid][22][UTT][ MSE = 0.2055 | LCC = 0.9611 | SRCC = 0.9624 ] +INFO:root:Average loss=0.048349027901887894 +INFO:root: +[Valid][23][UTT][ MSE = 0.1803 | LCC = 0.9649 | SRCC = 0.9649 ] +INFO:root:Average loss=0.047826089873909954 +INFO:root: +[Valid][24][UTT][ MSE = 0.1739 | LCC = 0.9641 | SRCC = 0.9640 ] +INFO:root:Average loss=0.04767277082800865 +INFO:root: +[Valid][25][UTT][ MSE = 0.1895 | LCC = 0.9647 | SRCC = 0.9646 ] +INFO:root:Average loss=0.043207256719470025 +INFO:root: +[Valid][26][UTT][ MSE = 0.1860 | LCC = 0.9639 | SRCC = 0.9640 ] +INFO:root:Average loss=0.04120388886332512 +INFO:root: +[Valid][27][UTT][ MSE = 0.1831 | LCC = 0.9644 | SRCC = 0.9642 ] +INFO:root:Average loss=0.04163418781757355 +INFO:root: +[Valid][28][UTT][ MSE = 0.1812 | LCC = 0.9641 | SRCC = 0.9636 ] +INFO:root:Average loss=0.03700066202878952 +INFO:root: +[Valid][29][UTT][ MSE = 0.1845 | LCC = 0.9648 | SRCC = 0.9647 ] +INFO:root:Average loss=0.03616278003156185 +INFO:root: +[Valid][30][UTT][ MSE = 0.1816 | LCC = 0.9638 | SRCC = 0.9638 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1887 | LCC = 0.9631 | SRCC = 0.9642 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run0/config.gin b/results_tencent/w2v2_large_xlsr53/layer14/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run0/model.pt b/results_tencent/w2v2_large_xlsr53/layer14/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6749125471f12daf287ca2b086d4e57c63d64e64 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fdfbbfb60a26635e8f020ef296c422a7d1f61f718f529d6ec74a04495ad9c9ef +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run0/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer14/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..515a1e54178ea72da568d947afaa396ff180ce7b --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f68693c8feaaae307572247f20c3885119665ed229baaf86c0dd85523b39091c +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run0/train.log b/results_tencent/w2v2_large_xlsr53/layer14/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9134eb0dbaf75a32fa035ec1ea43dd0fa015a916 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.43006341099739076 +INFO:root: +[Valid][0][UTT][ MSE = 0.3371 | LCC = 0.9245 | SRCC = 0.9247 ] +INFO:root:Average loss=0.1959560704231262 +INFO:root: +[Valid][1][UTT][ MSE = 0.2386 | LCC = 0.9414 | SRCC = 0.9435 ] +INFO:root:Average loss=0.1608742135167122 +INFO:root: +[Valid][2][UTT][ MSE = 0.2414 | LCC = 0.9477 | SRCC = 0.9495 ] +INFO:root:Average loss=0.13917632645368577 +INFO:root: +[Valid][3][UTT][ MSE = 0.2239 | LCC = 0.9519 | SRCC = 0.9536 ] +INFO:root:Average loss=0.12387573248147965 +INFO:root: +[Valid][4][UTT][ MSE = 0.2532 | LCC = 0.9554 | SRCC = 0.9580 ] +INFO:root:Average loss=0.12128651797771454 +INFO:root: +[Valid][5][UTT][ MSE = 0.1845 | LCC = 0.9587 | SRCC = 0.9603 ] +INFO:root:Average loss=0.11308632755279541 +INFO:root: +[Valid][6][UTT][ MSE = 0.2055 | LCC = 0.9575 | SRCC = 0.9592 ] +INFO:root:Average loss=0.09946422854065895 +INFO:root: +[Valid][7][UTT][ MSE = 0.1984 | LCC = 0.9607 | SRCC = 0.9618 ] +INFO:root:Average loss=0.09858786675333976 +INFO:root: +[Valid][8][UTT][ MSE = 0.1812 | LCC = 0.9598 | SRCC = 0.9617 ] +INFO:root:Average loss=0.09248992142081261 +INFO:root: +[Valid][9][UTT][ MSE = 0.2117 | LCC = 0.9614 | SRCC = 0.9625 ] +INFO:root:Average loss=0.09676281091570854 +INFO:root: +[Valid][10][UTT][ MSE = 0.1895 | LCC = 0.9571 | SRCC = 0.9587 ] +INFO:root:Average loss=0.08610827389359474 +INFO:root: +[Valid][11][UTT][ MSE = 0.2145 | LCC = 0.9623 | SRCC = 0.9638 ] +INFO:root:Average loss=0.08022017043828965 +INFO:root: +[Valid][12][UTT][ MSE = 0.2010 | LCC = 0.9641 | SRCC = 0.9645 ] +INFO:root:Average loss=0.07753248780965805 +INFO:root: +[Valid][13][UTT][ MSE = 0.1883 | LCC = 0.9625 | SRCC = 0.9637 ] +INFO:root:Average loss=0.07361738282442093 +INFO:root: +[Valid][14][UTT][ MSE = 0.2284 | LCC = 0.9633 | SRCC = 0.9639 ] +INFO:root:Average loss=0.06815737429261208 +INFO:root: +[Valid][15][UTT][ MSE = 0.1761 | LCC = 0.9628 | SRCC = 0.9640 ] +INFO:root:Average loss=0.06861218604445457 +INFO:root: +[Valid][16][UTT][ MSE = 0.1840 | LCC = 0.9631 | SRCC = 0.9634 ] +INFO:root:Average loss=0.06241007998585701 +INFO:root: +[Valid][17][UTT][ MSE = 0.1964 | LCC = 0.9634 | SRCC = 0.9634 ] +INFO:root:Average loss=0.0620883908867836 +INFO:root: +[Valid][18][UTT][ MSE = 0.1972 | LCC = 0.9631 | SRCC = 0.9631 ] +INFO:root:Average loss=0.05615733551979065 +INFO:root: +[Valid][19][UTT][ MSE = 0.1959 | LCC = 0.9639 | SRCC = 0.9642 ] +INFO:root:Average loss=0.05558582085371017 +INFO:root: +[Valid][20][UTT][ MSE = 0.1983 | LCC = 0.9644 | SRCC = 0.9647 ] +INFO:root:Average loss=0.05306707048416138 +INFO:root: +[Valid][21][UTT][ MSE = 0.2067 | LCC = 0.9609 | SRCC = 0.9629 ] +INFO:root:Average loss=0.05317243053019047 +INFO:root: +[Valid][22][UTT][ MSE = 0.1876 | LCC = 0.9609 | SRCC = 0.9612 ] +INFO:root:Average loss=0.04992939782142639 +INFO:root: +[Valid][23][UTT][ MSE = 0.1692 | LCC = 0.9640 | SRCC = 0.9640 ] +INFO:root:Average loss=0.05136477077007294 +INFO:root: +[Valid][24][UTT][ MSE = 0.1804 | LCC = 0.9637 | SRCC = 0.9635 ] +INFO:root:Average loss=0.04569811172783375 +INFO:root: +[Valid][25][UTT][ MSE = 0.1906 | LCC = 0.9639 | SRCC = 0.9633 ] +INFO:root:Average loss=0.046844424366950986 +INFO:root: +[Valid][26][UTT][ MSE = 0.1619 | LCC = 0.9637 | SRCC = 0.9645 ] +INFO:root:Average loss=0.04330252918601036 +INFO:root: +[Valid][27][UTT][ MSE = 0.1929 | LCC = 0.9640 | SRCC = 0.9646 ] +INFO:root:Average loss=0.040420472398400305 +INFO:root: +[Valid][28][UTT][ MSE = 0.1780 | LCC = 0.9643 | SRCC = 0.9639 ] +INFO:root:Average loss=0.03984687416255474 +INFO:root: +[Valid][29][UTT][ MSE = 0.1789 | LCC = 0.9642 | SRCC = 0.9639 ] +INFO:root:Average loss=0.03884649394452572 +INFO:root: +[Valid][30][UTT][ MSE = 0.2022 | LCC = 0.9626 | SRCC = 0.9629 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2038 | LCC = 0.9637 | SRCC = 0.9646 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run1/config.gin b/results_tencent/w2v2_large_xlsr53/layer14/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run1/model.pt b/results_tencent/w2v2_large_xlsr53/layer14/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c4f585d7c7d05508db4074933f1d0e77076075e0 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c09dec428515f3f99f4fc5b3e4c96ceac3070b0aa3238a520d06ffeb3a8eab2d +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run1/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer14/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..53d70bc29d2d21fe7e6cc251e59355d4b854e1c5 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d8616d7943bbd3b24b29f80575956a59a92f1b6731a0388b3a1d4638b87eeba5 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run1/train.log b/results_tencent/w2v2_large_xlsr53/layer14/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1245ac922f9be3e451294e2062d5e03636a2d8f4 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.415326157271862 +INFO:root: +[Valid][0][UTT][ MSE = 0.2931 | LCC = 0.9293 | SRCC = 0.9302 ] +INFO:root:Average loss=0.19925858825445175 +INFO:root: +[Valid][1][UTT][ MSE = 0.2605 | LCC = 0.9407 | SRCC = 0.9427 ] +INFO:root:Average loss=0.16692903953790664 +INFO:root: +[Valid][2][UTT][ MSE = 0.2176 | LCC = 0.9493 | SRCC = 0.9503 ] +INFO:root:Average loss=0.1362878983616829 +INFO:root: +[Valid][3][UTT][ MSE = 0.2256 | LCC = 0.9542 | SRCC = 0.9562 ] +INFO:root:Average loss=0.12515751731395722 +INFO:root: +[Valid][4][UTT][ MSE = 0.1901 | LCC = 0.9558 | SRCC = 0.9576 ] +INFO:root:Average loss=0.11087127518653869 +INFO:root: +[Valid][5][UTT][ MSE = 0.2330 | LCC = 0.9584 | SRCC = 0.9591 ] +INFO:root:Average loss=0.11119184812903404 +INFO:root: +[Valid][6][UTT][ MSE = 0.2274 | LCC = 0.9604 | SRCC = 0.9611 ] +INFO:root:Average loss=0.09700139161944389 +INFO:root: +[Valid][7][UTT][ MSE = 0.2095 | LCC = 0.9617 | SRCC = 0.9628 ] +INFO:root:Average loss=0.09368971315026284 +INFO:root: +[Valid][8][UTT][ MSE = 0.1875 | LCC = 0.9622 | SRCC = 0.9630 ] +INFO:root:Average loss=0.08742821833491325 +INFO:root: +[Valid][9][UTT][ MSE = 0.2264 | LCC = 0.9617 | SRCC = 0.9633 ] +INFO:root:Average loss=0.08482777234911919 +INFO:root: +[Valid][10][UTT][ MSE = 0.2088 | LCC = 0.9622 | SRCC = 0.9630 ] +INFO:root:Average loss=0.07962606278061866 +INFO:root: +[Valid][11][UTT][ MSE = 0.1683 | LCC = 0.9633 | SRCC = 0.9637 ] +INFO:root:Average loss=0.07624728271365165 +INFO:root: +[Valid][12][UTT][ MSE = 0.1526 | LCC = 0.9613 | SRCC = 0.9620 ] +INFO:root:Average loss=0.0721889854669571 +INFO:root: +[Valid][13][UTT][ MSE = 0.2013 | LCC = 0.9636 | SRCC = 0.9636 ] +INFO:root:Average loss=0.07040946629643441 +INFO:root: +[Valid][14][UTT][ MSE = 0.1948 | LCC = 0.9648 | SRCC = 0.9648 ] +INFO:root:Average loss=0.06533477461338043 +INFO:root: +[Valid][15][UTT][ MSE = 0.1827 | LCC = 0.9637 | SRCC = 0.9635 ] +INFO:root:Average loss=0.0656157087981701 +INFO:root: +[Valid][16][UTT][ MSE = 0.1988 | LCC = 0.9643 | SRCC = 0.9640 ] +INFO:root:Average loss=0.06161342599987984 +INFO:root: +[Valid][17][UTT][ MSE = 0.2074 | LCC = 0.9623 | SRCC = 0.9635 ] +INFO:root:Average loss=0.060269093722105024 +INFO:root: +[Valid][18][UTT][ MSE = 0.1787 | LCC = 0.9642 | SRCC = 0.9643 ] +INFO:root:Average loss=0.055758605152368544 +INFO:root: +[Valid][19][UTT][ MSE = 0.1773 | LCC = 0.9649 | SRCC = 0.9648 ] +INFO:root:Average loss=0.05531493593752384 +INFO:root: +[Valid][20][UTT][ MSE = 0.1984 | LCC = 0.9632 | SRCC = 0.9639 ] +INFO:root:Average loss=0.052917904749512676 +INFO:root: +[Valid][21][UTT][ MSE = 0.2070 | LCC = 0.9642 | SRCC = 0.9646 ] +INFO:root:Average loss=0.05037883326411247 +INFO:root: +[Valid][22][UTT][ MSE = 0.1995 | LCC = 0.9644 | SRCC = 0.9641 ] +INFO:root:Average loss=0.04836542618274689 +INFO:root: +[Valid][23][UTT][ MSE = 0.2042 | LCC = 0.9634 | SRCC = 0.9638 ] +INFO:root:Average loss=0.04691313527524471 +INFO:root: +[Valid][24][UTT][ MSE = 0.1899 | LCC = 0.9639 | SRCC = 0.9634 ] +INFO:root:Average loss=0.046109191730618476 +INFO:root: +[Valid][25][UTT][ MSE = 0.1855 | LCC = 0.9646 | SRCC = 0.9643 ] +INFO:root:Average loss=0.04602241881191731 +INFO:root: +[Valid][26][UTT][ MSE = 0.1794 | LCC = 0.9644 | SRCC = 0.9637 ] +INFO:root:Average loss=0.0421087788939476 +INFO:root: +[Valid][27][UTT][ MSE = 0.1809 | LCC = 0.9632 | SRCC = 0.9636 ] +INFO:root:Average loss=0.041331661760807034 +INFO:root: +[Valid][28][UTT][ MSE = 0.1955 | LCC = 0.9642 | SRCC = 0.9638 ] +INFO:root:Average loss=0.04260330605506897 +INFO:root: +[Valid][29][UTT][ MSE = 0.1890 | LCC = 0.9651 | SRCC = 0.9647 ] +INFO:root:Average loss=0.03849607263505459 +INFO:root: +[Valid][30][UTT][ MSE = 0.1822 | LCC = 0.9645 | SRCC = 0.9640 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2004 | LCC = 0.9612 | SRCC = 0.9624 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run2/config.gin b/results_tencent/w2v2_large_xlsr53/layer14/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run2/model.pt b/results_tencent/w2v2_large_xlsr53/layer14/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..db2f9dc58377db2de3002b9bd87ecd69716b1a95 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0d588392cc8cfe6badaaa0d9255b46bdf8a7fb5fee9812cd6f1a2b353b8ab150 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run2/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer14/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d3bda98f40889974d922110731ddd034889d9b21 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:def158b0b64d6ea4f2edb103dc27f93bfb1ab4e4e11ebbdb8d141122c7f41b2f +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run2/train.log b/results_tencent/w2v2_large_xlsr53/layer14/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5939b8f739ab4550c542e79c8e8886ad9db9ae5f --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46885611033439634 +INFO:root: +[Valid][0][UTT][ MSE = 0.3052 | LCC = 0.9158 | SRCC = 0.9148 ] +INFO:root:Average loss=0.23354758286476135 +INFO:root: +[Valid][1][UTT][ MSE = 0.2713 | LCC = 0.9343 | SRCC = 0.9351 ] +INFO:root:Average loss=0.17347564214468003 +INFO:root: +[Valid][2][UTT][ MSE = 0.2496 | LCC = 0.9442 | SRCC = 0.9455 ] +INFO:root:Average loss=0.14594601047039032 +INFO:root: +[Valid][3][UTT][ MSE = 0.2616 | LCC = 0.9507 | SRCC = 0.9519 ] +INFO:root:Average loss=0.13397228503227235 +INFO:root: +[Valid][4][UTT][ MSE = 0.2203 | LCC = 0.9544 | SRCC = 0.9561 ] +INFO:root:Average loss=0.1198594667315483 +INFO:root: +[Valid][5][UTT][ MSE = 0.2317 | LCC = 0.9565 | SRCC = 0.9582 ] +INFO:root:Average loss=0.11267294389009476 +INFO:root: +[Valid][6][UTT][ MSE = 0.1730 | LCC = 0.9592 | SRCC = 0.9607 ] +INFO:root:Average loss=0.10555411458015442 +INFO:root: +[Valid][7][UTT][ MSE = 0.1796 | LCC = 0.9612 | SRCC = 0.9625 ] +INFO:root:Average loss=0.10109587872028351 +INFO:root: +[Valid][8][UTT][ MSE = 0.1884 | LCC = 0.9617 | SRCC = 0.9627 ] +INFO:root:Average loss=0.09797879147529602 +INFO:root: +[Valid][9][UTT][ MSE = 0.1824 | LCC = 0.9621 | SRCC = 0.9632 ] +INFO:root:Average loss=0.09043451261520385 +INFO:root: +[Valid][10][UTT][ MSE = 0.1772 | LCC = 0.9638 | SRCC = 0.9643 ] +INFO:root:Average loss=0.08575934559106826 +INFO:root: +[Valid][11][UTT][ MSE = 0.1708 | LCC = 0.9630 | SRCC = 0.9637 ] +INFO:root:Average loss=0.08174925565719604 +INFO:root: +[Valid][12][UTT][ MSE = 0.1988 | LCC = 0.9641 | SRCC = 0.9643 ] +INFO:root:Average loss=0.0768416319489479 +INFO:root: +[Valid][13][UTT][ MSE = 0.1781 | LCC = 0.9634 | SRCC = 0.9641 ] +INFO:root:Average loss=0.073387117177248 +INFO:root: +[Valid][14][UTT][ MSE = 0.1627 | LCC = 0.9633 | SRCC = 0.9645 ] +INFO:root:Average loss=0.07090455016493798 +INFO:root: +[Valid][15][UTT][ MSE = 0.2149 | LCC = 0.9652 | SRCC = 0.9650 ] +INFO:root:Average loss=0.06796280667185783 +INFO:root: +[Valid][16][UTT][ MSE = 0.1872 | LCC = 0.9636 | SRCC = 0.9643 ] +INFO:root:Average loss=0.06395047229528428 +INFO:root: +[Valid][17][UTT][ MSE = 0.1651 | LCC = 0.9653 | SRCC = 0.9649 ] +INFO:root:Average loss=0.06480462989211082 +INFO:root: +[Valid][18][UTT][ MSE = 0.1451 | LCC = 0.9636 | SRCC = 0.9635 ] +INFO:root:Average loss=0.05971020904183388 +INFO:root: +[Valid][19][UTT][ MSE = 0.1757 | LCC = 0.9655 | SRCC = 0.9657 ] +INFO:root:Average loss=0.05537576755881309 +INFO:root: +[Valid][20][UTT][ MSE = 0.1822 | LCC = 0.9648 | SRCC = 0.9651 ] +INFO:root:Average loss=0.0538552660048008 +INFO:root: +[Valid][21][UTT][ MSE = 0.1780 | LCC = 0.9664 | SRCC = 0.9661 ] +INFO:root:Average loss=0.052482214942574504 +INFO:root: +[Valid][22][UTT][ MSE = 0.1725 | LCC = 0.9651 | SRCC = 0.9645 ] +INFO:root:Average loss=0.05185001397132873 +INFO:root: +[Valid][23][UTT][ MSE = 0.1763 | LCC = 0.9663 | SRCC = 0.9660 ] +INFO:root:Average loss=0.05220246440172195 +INFO:root: +[Valid][24][UTT][ MSE = 0.1873 | LCC = 0.9658 | SRCC = 0.9661 ] +INFO:root:Average loss=0.0474115351587534 +INFO:root: +[Valid][25][UTT][ MSE = 0.1838 | LCC = 0.9654 | SRCC = 0.9656 ] +INFO:root:Average loss=0.04507037170231342 +INFO:root: +[Valid][26][UTT][ MSE = 0.1654 | LCC = 0.9656 | SRCC = 0.9655 ] +INFO:root:Average loss=0.042423386111855504 +INFO:root: +[Valid][27][UTT][ MSE = 0.1754 | LCC = 0.9643 | SRCC = 0.9644 ] +INFO:root:Average loss=0.04512192437052727 +INFO:root: +[Valid][28][UTT][ MSE = 0.1707 | LCC = 0.9645 | SRCC = 0.9639 ] +INFO:root:Average loss=0.04205025750398636 +INFO:root: +[Valid][29][UTT][ MSE = 0.1788 | LCC = 0.9647 | SRCC = 0.9646 ] +INFO:root:Average loss=0.041835583552718165 +INFO:root: +[Valid][30][UTT][ MSE = 0.1770 | LCC = 0.9641 | SRCC = 0.9639 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1902 | LCC = 0.9624 | SRCC = 0.9636 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run3/config.gin b/results_tencent/w2v2_large_xlsr53/layer14/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run3/model.pt b/results_tencent/w2v2_large_xlsr53/layer14/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..5bbe6c726e374bd172a8a41ce7fa2e20fe916853 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ec92f294d497ba65b040ac80ef476f12d0bc1f787c8d4be360651233bc9485b1 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run3/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer14/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c68d514910a9ad03802d8a5624f1b5c29e2e3298 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6a4062b39c80d55458983de9d3ed91270195fde53cbd5f2fa1f3ccc2fdb066a4 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run3/train.log b/results_tencent/w2v2_large_xlsr53/layer14/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b9e06655092c8618a932c6b6f9d88b13d49c3a65 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4483659917116165 +INFO:root: +[Valid][0][UTT][ MSE = 0.2972 | LCC = 0.9253 | SRCC = 0.9261 ] +INFO:root:Average loss=0.20874171990156173 +INFO:root: +[Valid][1][UTT][ MSE = 0.2542 | LCC = 0.9431 | SRCC = 0.9437 ] +INFO:root:Average loss=0.15612542307376862 +INFO:root: +[Valid][2][UTT][ MSE = 0.2083 | LCC = 0.9508 | SRCC = 0.9540 ] +INFO:root:Average loss=0.1377876558303833 +INFO:root: +[Valid][3][UTT][ MSE = 0.2449 | LCC = 0.9542 | SRCC = 0.9549 ] +INFO:root:Average loss=0.13032397615909577 +INFO:root: +[Valid][4][UTT][ MSE = 0.2138 | LCC = 0.9566 | SRCC = 0.9582 ] +INFO:root:Average loss=0.12557058346271516 +INFO:root: +[Valid][5][UTT][ MSE = 0.2314 | LCC = 0.9599 | SRCC = 0.9607 ] +INFO:root:Average loss=0.10811115768551827 +INFO:root: +[Valid][6][UTT][ MSE = 0.2082 | LCC = 0.9608 | SRCC = 0.9616 ] +INFO:root:Average loss=0.09863081783056259 +INFO:root: +[Valid][7][UTT][ MSE = 0.1988 | LCC = 0.9607 | SRCC = 0.9618 ] +INFO:root:Average loss=0.0983964530825615 +INFO:root: +[Valid][8][UTT][ MSE = 0.2254 | LCC = 0.9619 | SRCC = 0.9635 ] +INFO:root:Average loss=0.08917754945158958 +INFO:root: +[Valid][9][UTT][ MSE = 0.2095 | LCC = 0.9620 | SRCC = 0.9630 ] +INFO:root:Average loss=0.086341039955616 +INFO:root: +[Valid][10][UTT][ MSE = 0.1905 | LCC = 0.9637 | SRCC = 0.9642 ] +INFO:root:Average loss=0.08650208076834678 +INFO:root: +[Valid][11][UTT][ MSE = 0.1578 | LCC = 0.9635 | SRCC = 0.9643 ] +INFO:root:Average loss=0.0771184087395668 +INFO:root: +[Valid][12][UTT][ MSE = 0.1743 | LCC = 0.9639 | SRCC = 0.9641 ] +INFO:root:Average loss=0.07496028679609298 +INFO:root: +[Valid][13][UTT][ MSE = 0.1898 | LCC = 0.9635 | SRCC = 0.9642 ] +INFO:root:Average loss=0.0737803162932396 +INFO:root: +[Valid][14][UTT][ MSE = 0.1691 | LCC = 0.9644 | SRCC = 0.9644 ] +INFO:root:Average loss=0.06864249107241631 +INFO:root: +[Valid][15][UTT][ MSE = 0.2172 | LCC = 0.9612 | SRCC = 0.9636 ] +INFO:root:Average loss=0.07043195566534996 +INFO:root: +[Valid][16][UTT][ MSE = 0.2301 | LCC = 0.9622 | SRCC = 0.9643 ] +INFO:root:Average loss=0.061761035293340684 +INFO:root: +[Valid][17][UTT][ MSE = 0.2035 | LCC = 0.9638 | SRCC = 0.9646 ] +INFO:root:Average loss=0.0592706435918808 +INFO:root: +[Valid][18][UTT][ MSE = 0.1808 | LCC = 0.9644 | SRCC = 0.9650 ] +INFO:root:Average loss=0.057525914043188096 +INFO:root: +[Valid][19][UTT][ MSE = 0.2003 | LCC = 0.9638 | SRCC = 0.9638 ] +INFO:root:Average loss=0.05814719775319099 +INFO:root: +[Valid][20][UTT][ MSE = 0.2027 | LCC = 0.9642 | SRCC = 0.9654 ] +INFO:root:Average loss=0.05304221297800541 +INFO:root: +[Valid][21][UTT][ MSE = 0.1741 | LCC = 0.9624 | SRCC = 0.9641 ] +INFO:root:Average loss=0.052149691045284274 +INFO:root: +[Valid][22][UTT][ MSE = 0.1683 | LCC = 0.9655 | SRCC = 0.9654 ] +INFO:root:Average loss=0.049863875299692155 +INFO:root: +[Valid][23][UTT][ MSE = 0.1512 | LCC = 0.9653 | SRCC = 0.9650 ] +INFO:root:Average loss=0.04693413060903549 +INFO:root: +[Valid][24][UTT][ MSE = 0.1979 | LCC = 0.9658 | SRCC = 0.9657 ] +INFO:root:Average loss=0.047396782115101815 +INFO:root: +[Valid][25][UTT][ MSE = 0.1779 | LCC = 0.9640 | SRCC = 0.9644 ] +INFO:root:Average loss=0.04479567849636078 +INFO:root: +[Valid][26][UTT][ MSE = 0.1833 | LCC = 0.9647 | SRCC = 0.9643 ] +INFO:root:Average loss=0.04388544228672981 +INFO:root: +[Valid][27][UTT][ MSE = 0.1779 | LCC = 0.9658 | SRCC = 0.9656 ] +INFO:root:Average loss=0.043171820878982546 +INFO:root: +[Valid][28][UTT][ MSE = 0.1816 | LCC = 0.9658 | SRCC = 0.9658 ] +INFO:root:Average loss=0.041556056648492816 +INFO:root: +[Valid][29][UTT][ MSE = 0.1819 | LCC = 0.9651 | SRCC = 0.9654 ] +INFO:root:Average loss=0.04072307884693146 +INFO:root: +[Valid][30][UTT][ MSE = 0.1587 | LCC = 0.9642 | SRCC = 0.9657 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1872 | LCC = 0.9636 | SRCC = 0.9642 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run4/config.gin b/results_tencent/w2v2_large_xlsr53/layer14/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run4/model.pt b/results_tencent/w2v2_large_xlsr53/layer14/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f6b38c098aa51f5a261d0ef9807ff024685d0226 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9f8330e82f6e064ab41f5de28124da9461dadb11795c41a152465ba3d4f2ef80 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run4/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer14/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..eb141f8b8b818a7dc5cb1f59654761a3498fb0d8 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f23917dfba7e8f4b4b06dedd803876a4dc73a98db31b3c72f1440c1b9ce43989 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer14/run4/train.log b/results_tencent/w2v2_large_xlsr53/layer14/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..78a78259c479753294544020d11d9efc32ca535a --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer14/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4695589963197708 +INFO:root: +[Valid][0][UTT][ MSE = 0.3156 | LCC = 0.9278 | SRCC = 0.9294 ] +INFO:root:Average loss=0.19061370813846587 +INFO:root: +[Valid][1][UTT][ MSE = 0.2224 | LCC = 0.9435 | SRCC = 0.9444 ] +INFO:root:Average loss=0.15410245150327684 +INFO:root: +[Valid][2][UTT][ MSE = 0.2590 | LCC = 0.9505 | SRCC = 0.9531 ] +INFO:root:Average loss=0.14158218532800673 +INFO:root: +[Valid][3][UTT][ MSE = 0.2189 | LCC = 0.9523 | SRCC = 0.9548 ] +INFO:root:Average loss=0.12598831355571746 +INFO:root: +[Valid][4][UTT][ MSE = 0.1754 | LCC = 0.9561 | SRCC = 0.9571 ] +INFO:root:Average loss=0.11341007459163666 +INFO:root: +[Valid][5][UTT][ MSE = 0.2347 | LCC = 0.9573 | SRCC = 0.9593 ] +INFO:root:Average loss=0.1047786939740181 +INFO:root: +[Valid][6][UTT][ MSE = 0.1968 | LCC = 0.9607 | SRCC = 0.9617 ] +INFO:root:Average loss=0.10048063841462135 +INFO:root: +[Valid][7][UTT][ MSE = 0.2199 | LCC = 0.9617 | SRCC = 0.9629 ] +INFO:root:Average loss=0.10343266585469246 +INFO:root: +[Valid][8][UTT][ MSE = 0.1786 | LCC = 0.9619 | SRCC = 0.9620 ] +INFO:root:Average loss=0.0885175532400608 +INFO:root: +[Valid][9][UTT][ MSE = 0.2057 | LCC = 0.9611 | SRCC = 0.9621 ] +INFO:root:Average loss=0.08433436533808708 +INFO:root: +[Valid][10][UTT][ MSE = 0.1651 | LCC = 0.9628 | SRCC = 0.9635 ] +INFO:root:Average loss=0.08138748893141746 +INFO:root: +[Valid][11][UTT][ MSE = 0.2165 | LCC = 0.9631 | SRCC = 0.9647 ] +INFO:root:Average loss=0.07843477934598923 +INFO:root: +[Valid][12][UTT][ MSE = 0.1928 | LCC = 0.9646 | SRCC = 0.9647 ] +INFO:root:Average loss=0.07241060262918472 +INFO:root: +[Valid][13][UTT][ MSE = 0.1709 | LCC = 0.9629 | SRCC = 0.9642 ] +INFO:root:Average loss=0.06950125387310982 +INFO:root: +[Valid][14][UTT][ MSE = 0.1763 | LCC = 0.9644 | SRCC = 0.9646 ] +INFO:root:Average loss=0.06787063285708428 +INFO:root: +[Valid][15][UTT][ MSE = 0.2017 | LCC = 0.9652 | SRCC = 0.9660 ] +INFO:root:Average loss=0.06711900749802589 +INFO:root: +[Valid][16][UTT][ MSE = 0.1981 | LCC = 0.9645 | SRCC = 0.9655 ] +INFO:root:Average loss=0.06259139367938042 +INFO:root: +[Valid][17][UTT][ MSE = 0.2109 | LCC = 0.9660 | SRCC = 0.9661 ] +INFO:root:Average loss=0.06044326141476631 +INFO:root: +[Valid][18][UTT][ MSE = 0.1767 | LCC = 0.9653 | SRCC = 0.9657 ] +INFO:root:Average loss=0.058259370237588884 +INFO:root: +[Valid][19][UTT][ MSE = 0.1649 | LCC = 0.9653 | SRCC = 0.9651 ] +INFO:root:Average loss=0.05561916673183441 +INFO:root: +[Valid][20][UTT][ MSE = 0.1875 | LCC = 0.9649 | SRCC = 0.9651 ] +INFO:root:Average loss=0.05710024492442608 +INFO:root: +[Valid][21][UTT][ MSE = 0.1950 | LCC = 0.9646 | SRCC = 0.9651 ] +INFO:root:Average loss=0.05199528260529041 +INFO:root: +[Valid][22][UTT][ MSE = 0.1537 | LCC = 0.9660 | SRCC = 0.9656 ] +INFO:root:Average loss=0.05187561985850334 +INFO:root: +[Valid][23][UTT][ MSE = 0.1830 | LCC = 0.9657 | SRCC = 0.9653 ] +INFO:root:Average loss=0.04866536739468574 +INFO:root: +[Valid][24][UTT][ MSE = 0.1630 | LCC = 0.9662 | SRCC = 0.9661 ] +INFO:root:Average loss=0.04703013996779919 +INFO:root: +[Valid][25][UTT][ MSE = 0.1553 | LCC = 0.9666 | SRCC = 0.9664 ] +INFO:root:Average loss=0.04489597032964229 +INFO:root: +[Valid][26][UTT][ MSE = 0.1730 | LCC = 0.9665 | SRCC = 0.9663 ] +INFO:root:Average loss=0.044461611568927764 +INFO:root: +[Valid][27][UTT][ MSE = 0.1848 | LCC = 0.9655 | SRCC = 0.9664 ] +INFO:root:Average loss=0.04242924191057682 +INFO:root: +[Valid][28][UTT][ MSE = 0.1585 | LCC = 0.9663 | SRCC = 0.9661 ] +INFO:root:Average loss=0.04056888072192669 +INFO:root: +[Valid][29][UTT][ MSE = 0.1663 | LCC = 0.9672 | SRCC = 0.9667 ] +INFO:root:Average loss=0.039839939057827 +INFO:root: +[Valid][30][UTT][ MSE = 0.1915 | LCC = 0.9655 | SRCC = 0.9659 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1726 | LCC = 0.9644 | SRCC = 0.9653 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run0/config.gin b/results_tencent/w2v2_large_xlsr53/layer16/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run0/model.pt b/results_tencent/w2v2_large_xlsr53/layer16/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..421bf55778da64e64d9e69b47e36f9187947aa52 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2cb141ef8e29e2d230f130b2a087aa0f9874121a30fb962b525e69cd097c507d +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run0/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer16/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..102186f9dff70a3d071d18a7b522d6e12070a14f --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c9d582b58a5c25aa2e2ae948961691dcc949b6f09a0f87c600f494312aa697a5 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run0/train.log b/results_tencent/w2v2_large_xlsr53/layer16/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..efb4a43f506728aee59894bac8775eb0e50fc6f9 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5365224579572677 +INFO:root: +[Valid][0][UTT][ MSE = 0.3723 | LCC = 0.9086 | SRCC = 0.9188 ] +INFO:root:Average loss=0.2110677530169487 +INFO:root: +[Valid][1][UTT][ MSE = 0.2695 | LCC = 0.9423 | SRCC = 0.9424 ] +INFO:root:Average loss=0.1601612930893898 +INFO:root: +[Valid][2][UTT][ MSE = 0.2646 | LCC = 0.9453 | SRCC = 0.9480 ] +INFO:root:Average loss=0.13576393288373947 +INFO:root: +[Valid][3][UTT][ MSE = 0.2801 | LCC = 0.9512 | SRCC = 0.9534 ] +INFO:root:Average loss=0.1204810222685337 +INFO:root: +[Valid][4][UTT][ MSE = 0.2024 | LCC = 0.9559 | SRCC = 0.9573 ] +INFO:root:Average loss=0.11143402206897736 +INFO:root: +[Valid][5][UTT][ MSE = 0.2437 | LCC = 0.9581 | SRCC = 0.9592 ] +INFO:root:Average loss=0.10945748373866081 +INFO:root: +[Valid][6][UTT][ MSE = 0.2176 | LCC = 0.9540 | SRCC = 0.9573 ] +INFO:root:Average loss=0.1028214792907238 +INFO:root: +[Valid][7][UTT][ MSE = 0.1991 | LCC = 0.9607 | SRCC = 0.9617 ] +INFO:root:Average loss=0.0940398035645485 +INFO:root: +[Valid][8][UTT][ MSE = 0.2093 | LCC = 0.9568 | SRCC = 0.9572 ] +INFO:root:Average loss=0.08877258017659187 +INFO:root: +[Valid][9][UTT][ MSE = 0.2530 | LCC = 0.9593 | SRCC = 0.9617 ] +INFO:root:Average loss=0.08587017610669136 +INFO:root: +[Valid][10][UTT][ MSE = 0.2398 | LCC = 0.9599 | SRCC = 0.9617 ] +INFO:root:Average loss=0.08061012399196625 +INFO:root: +[Valid][11][UTT][ MSE = 0.2552 | LCC = 0.9603 | SRCC = 0.9620 ] +INFO:root:Average loss=0.07564945343136788 +INFO:root: +[Valid][12][UTT][ MSE = 0.2246 | LCC = 0.9615 | SRCC = 0.9629 ] +INFO:root:Average loss=0.07226022350788117 +INFO:root: +[Valid][13][UTT][ MSE = 0.2380 | LCC = 0.9619 | SRCC = 0.9617 ] +INFO:root:Average loss=0.07173343873023987 +INFO:root: +[Valid][14][UTT][ MSE = 0.1719 | LCC = 0.9619 | SRCC = 0.9622 ] +INFO:root:Average loss=0.06518207663297654 +INFO:root: +[Valid][15][UTT][ MSE = 0.1733 | LCC = 0.9636 | SRCC = 0.9638 ] +INFO:root:Average loss=0.06437010020017624 +INFO:root: +[Valid][16][UTT][ MSE = 0.1864 | LCC = 0.9629 | SRCC = 0.9631 ] +INFO:root:Average loss=0.06097273793816566 +INFO:root: +[Valid][17][UTT][ MSE = 0.1881 | LCC = 0.9630 | SRCC = 0.9629 ] +INFO:root:Average loss=0.06076639668643474 +INFO:root: +[Valid][18][UTT][ MSE = 0.2011 | LCC = 0.9634 | SRCC = 0.9637 ] +INFO:root:Average loss=0.05579030965268612 +INFO:root: +[Valid][19][UTT][ MSE = 0.1676 | LCC = 0.9633 | SRCC = 0.9628 ] +INFO:root:Average loss=0.05405898566544056 +INFO:root: +[Valid][20][UTT][ MSE = 0.1970 | LCC = 0.9637 | SRCC = 0.9634 ] +INFO:root:Average loss=0.051065088659524915 +INFO:root: +[Valid][21][UTT][ MSE = 0.1895 | LCC = 0.9631 | SRCC = 0.9631 ] +INFO:root:Average loss=0.05158224330842495 +INFO:root: +[Valid][22][UTT][ MSE = 0.1907 | LCC = 0.9620 | SRCC = 0.9636 ] +INFO:root:Average loss=0.047852037101984025 +INFO:root: +[Valid][23][UTT][ MSE = 0.2099 | LCC = 0.9638 | SRCC = 0.9636 ] +INFO:root:Average loss=0.047303221255540845 +INFO:root: +[Valid][24][UTT][ MSE = 0.1839 | LCC = 0.9643 | SRCC = 0.9637 ] +INFO:root:Average loss=0.043887838542461394 +INFO:root: +[Valid][25][UTT][ MSE = 0.1760 | LCC = 0.9634 | SRCC = 0.9625 ] +INFO:root:Average loss=0.042267488598823545 +INFO:root: +[Valid][26][UTT][ MSE = 0.1850 | LCC = 0.9635 | SRCC = 0.9636 ] +INFO:root:Average loss=0.04026079931855202 +INFO:root: +[Valid][27][UTT][ MSE = 0.2132 | LCC = 0.9618 | SRCC = 0.9629 ] +INFO:root:Average loss=0.04230159409344196 +INFO:root: +[Valid][28][UTT][ MSE = 0.1876 | LCC = 0.9594 | SRCC = 0.9603 ] +INFO:root:Average loss=0.04027805107831955 +INFO:root: +[Valid][29][UTT][ MSE = 0.2351 | LCC = 0.9613 | SRCC = 0.9632 ] +INFO:root:Average loss=0.037914634570479396 +INFO:root: +[Valid][30][UTT][ MSE = 0.1833 | LCC = 0.9633 | SRCC = 0.9633 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1886 | LCC = 0.9636 | SRCC = 0.9647 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run1/config.gin b/results_tencent/w2v2_large_xlsr53/layer16/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run1/model.pt b/results_tencent/w2v2_large_xlsr53/layer16/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..70ea652fdd8eaa368636ab6b54e6c9a418685026 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f84d3d78cbaa374d319eb932c267eab5026d1ab5d09255009b7ba308ca53cf04 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run1/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer16/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0dba69eb53b3f99466cd03fbf166fcbfa66d8a30 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:78fd150e89390dcad26916791f229cb90828c2ba8c0acb45e3dabb31f8dc411b +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run1/train.log b/results_tencent/w2v2_large_xlsr53/layer16/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6d279b3101930c29bec4777f63db40173c5db8d4 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6183173854351044 +INFO:root: +[Valid][0][UTT][ MSE = 0.3288 | LCC = 0.9211 | SRCC = 0.9202 ] +INFO:root:Average loss=0.1893882941007614 +INFO:root: +[Valid][1][UTT][ MSE = 0.2668 | LCC = 0.9442 | SRCC = 0.9454 ] +INFO:root:Average loss=0.14511840242147445 +INFO:root: +[Valid][2][UTT][ MSE = 0.2452 | LCC = 0.9512 | SRCC = 0.9534 ] +INFO:root:Average loss=0.13064431029558182 +INFO:root: +[Valid][3][UTT][ MSE = 0.2050 | LCC = 0.9530 | SRCC = 0.9555 ] +INFO:root:Average loss=0.11861791014671326 +INFO:root: +[Valid][4][UTT][ MSE = 0.1808 | LCC = 0.9584 | SRCC = 0.9597 ] +INFO:root:Average loss=0.11042832696437836 +INFO:root: +[Valid][5][UTT][ MSE = 0.2143 | LCC = 0.9597 | SRCC = 0.9607 ] +INFO:root:Average loss=0.1056757342517376 +INFO:root: +[Valid][6][UTT][ MSE = 0.2152 | LCC = 0.9586 | SRCC = 0.9604 ] +INFO:root:Average loss=0.0954638059437275 +INFO:root: +[Valid][7][UTT][ MSE = 0.2394 | LCC = 0.9591 | SRCC = 0.9608 ] +INFO:root:Average loss=0.09664845624566078 +INFO:root: +[Valid][8][UTT][ MSE = 0.2563 | LCC = 0.9604 | SRCC = 0.9622 ] +INFO:root:Average loss=0.08754548168182373 +INFO:root: +[Valid][9][UTT][ MSE = 0.3049 | LCC = 0.9546 | SRCC = 0.9596 ] +INFO:root:Average loss=0.08589071109890938 +INFO:root: +[Valid][10][UTT][ MSE = 0.2129 | LCC = 0.9621 | SRCC = 0.9630 ] +INFO:root:Average loss=0.07721202975511551 +INFO:root: +[Valid][11][UTT][ MSE = 0.2049 | LCC = 0.9624 | SRCC = 0.9633 ] +INFO:root:Average loss=0.07486471658945083 +INFO:root: +[Valid][12][UTT][ MSE = 0.2065 | LCC = 0.9629 | SRCC = 0.9635 ] +INFO:root:Average loss=0.07516126170754432 +INFO:root: +[Valid][13][UTT][ MSE = 0.2221 | LCC = 0.9602 | SRCC = 0.9613 ] +INFO:root:Average loss=0.07551193663477898 +INFO:root: +[Valid][14][UTT][ MSE = 0.1758 | LCC = 0.9633 | SRCC = 0.9640 ] +INFO:root:Average loss=0.0650109140276909 +INFO:root: +[Valid][15][UTT][ MSE = 0.1741 | LCC = 0.9610 | SRCC = 0.9604 ] +INFO:root:Average loss=0.0625930609703064 +INFO:root: +[Valid][16][UTT][ MSE = 0.2136 | LCC = 0.9622 | SRCC = 0.9630 ] +INFO:root:Average loss=0.06434573218226433 +INFO:root: +[Valid][17][UTT][ MSE = 0.1819 | LCC = 0.9629 | SRCC = 0.9625 ] +INFO:root:Average loss=0.05706654697656632 +INFO:root: +[Valid][18][UTT][ MSE = 0.2008 | LCC = 0.9647 | SRCC = 0.9645 ] +INFO:root:Average loss=0.05604425531625748 +INFO:root: +[Valid][19][UTT][ MSE = 0.1847 | LCC = 0.9639 | SRCC = 0.9642 ] +INFO:root:Average loss=0.05131010131537914 +INFO:root: +[Valid][20][UTT][ MSE = 0.1821 | LCC = 0.9624 | SRCC = 0.9627 ] +INFO:root:Average loss=0.049133721515536306 +INFO:root: +[Valid][21][UTT][ MSE = 0.1840 | LCC = 0.9630 | SRCC = 0.9628 ] +INFO:root:Average loss=0.04867157761752605 +INFO:root: +[Valid][22][UTT][ MSE = 0.2128 | LCC = 0.9628 | SRCC = 0.9634 ] +INFO:root:Average loss=0.04585587035119534 +INFO:root: +[Valid][23][UTT][ MSE = 0.1767 | LCC = 0.9645 | SRCC = 0.9647 ] +INFO:root:Average loss=0.04468769682943821 +INFO:root: +[Valid][24][UTT][ MSE = 0.2006 | LCC = 0.9622 | SRCC = 0.9619 ] +INFO:root:Average loss=0.044113707408308986 +INFO:root: +[Valid][25][UTT][ MSE = 0.1910 | LCC = 0.9644 | SRCC = 0.9638 ] +INFO:root:Average loss=0.039614924147725106 +INFO:root: +[Valid][26][UTT][ MSE = 0.1969 | LCC = 0.9649 | SRCC = 0.9648 ] +INFO:root:Average loss=0.03898917280137539 +INFO:root: +[Valid][27][UTT][ MSE = 0.2074 | LCC = 0.9631 | SRCC = 0.9632 ] +INFO:root:Average loss=0.03882655744254589 +INFO:root: +[Valid][28][UTT][ MSE = 0.1864 | LCC = 0.9645 | SRCC = 0.9646 ] +INFO:root:Average loss=0.03885403011739254 +INFO:root: +[Valid][29][UTT][ MSE = 0.1836 | LCC = 0.9645 | SRCC = 0.9647 ] +INFO:root:Average loss=0.036026495680212975 +INFO:root: +[Valid][30][UTT][ MSE = 0.2098 | LCC = 0.9631 | SRCC = 0.9637 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2083 | LCC = 0.9636 | SRCC = 0.9642 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run2/config.gin b/results_tencent/w2v2_large_xlsr53/layer16/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run2/model.pt b/results_tencent/w2v2_large_xlsr53/layer16/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..32fc85252eacc71eec91022a0b717dfe8b7d6097 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d752f8b21a62303cea20bda6e5f2f5ba0fbe72f50308fd228185b203cc3b3930 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run2/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer16/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a92d3044100ac336b5225ff5ddd490dff36d024c --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:58c72f1c0b8678c46c19198bb8d83e6fb5610d9ea8ed550deb8a0eb88fe759fb +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run2/train.log b/results_tencent/w2v2_large_xlsr53/layer16/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9eb7a90fb276985e8688f724370899b355ea56ab --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5397572714090347 +INFO:root: +[Valid][0][UTT][ MSE = 0.3238 | LCC = 0.9218 | SRCC = 0.9223 ] +INFO:root:Average loss=0.18620825093984603 +INFO:root: +[Valid][1][UTT][ MSE = 0.2943 | LCC = 0.9434 | SRCC = 0.9450 ] +INFO:root:Average loss=0.15570630145072936 +INFO:root: +[Valid][2][UTT][ MSE = 0.2300 | LCC = 0.9480 | SRCC = 0.9502 ] +INFO:root:Average loss=0.1420606295466423 +INFO:root: +[Valid][3][UTT][ MSE = 0.2140 | LCC = 0.9523 | SRCC = 0.9545 ] +INFO:root:Average loss=0.12198758959770202 +INFO:root: +[Valid][4][UTT][ MSE = 0.1999 | LCC = 0.9566 | SRCC = 0.9585 ] +INFO:root:Average loss=0.12092717123031617 +INFO:root: +[Valid][5][UTT][ MSE = 0.2079 | LCC = 0.9568 | SRCC = 0.9579 ] +INFO:root:Average loss=0.10801135697960854 +INFO:root: +[Valid][6][UTT][ MSE = 0.1929 | LCC = 0.9605 | SRCC = 0.9612 ] +INFO:root:Average loss=0.10067010554671288 +INFO:root: +[Valid][7][UTT][ MSE = 0.2070 | LCC = 0.9613 | SRCC = 0.9620 ] +INFO:root:Average loss=0.0954327901005745 +INFO:root: +[Valid][8][UTT][ MSE = 0.2134 | LCC = 0.9615 | SRCC = 0.9625 ] +INFO:root:Average loss=0.08958659368753433 +INFO:root: +[Valid][9][UTT][ MSE = 0.2030 | LCC = 0.9625 | SRCC = 0.9633 ] +INFO:root:Average loss=0.08362657609581947 +INFO:root: +[Valid][10][UTT][ MSE = 0.2151 | LCC = 0.9634 | SRCC = 0.9644 ] +INFO:root:Average loss=0.0822958414554596 +INFO:root: +[Valid][11][UTT][ MSE = 0.2115 | LCC = 0.9621 | SRCC = 0.9627 ] +INFO:root:Average loss=0.07738843768835067 +INFO:root: +[Valid][12][UTT][ MSE = 0.1829 | LCC = 0.9634 | SRCC = 0.9645 ] +INFO:root:Average loss=0.07749402779340744 +INFO:root: +[Valid][13][UTT][ MSE = 0.2076 | LCC = 0.9632 | SRCC = 0.9637 ] +INFO:root:Average loss=0.07073966133594513 +INFO:root: +[Valid][14][UTT][ MSE = 0.1986 | LCC = 0.9639 | SRCC = 0.9650 ] +INFO:root:Average loss=0.06738584861159325 +INFO:root: +[Valid][15][UTT][ MSE = 0.1737 | LCC = 0.9632 | SRCC = 0.9635 ] +INFO:root:Average loss=0.06386353322863579 +INFO:root: +[Valid][16][UTT][ MSE = 0.1963 | LCC = 0.9640 | SRCC = 0.9649 ] +INFO:root:Average loss=0.06159870655834675 +INFO:root: +[Valid][17][UTT][ MSE = 0.1986 | LCC = 0.9643 | SRCC = 0.9650 ] +INFO:root:Average loss=0.0600799420773983 +INFO:root: +[Valid][18][UTT][ MSE = 0.1717 | LCC = 0.9624 | SRCC = 0.9635 ] +INFO:root:Average loss=0.05727346232533455 +INFO:root: +[Valid][19][UTT][ MSE = 0.1735 | LCC = 0.9648 | SRCC = 0.9652 ] +INFO:root:Average loss=0.05508535692095756 +INFO:root: +[Valid][20][UTT][ MSE = 0.1675 | LCC = 0.9646 | SRCC = 0.9654 ] +INFO:root:Average loss=0.05084998880326748 +INFO:root: +[Valid][21][UTT][ MSE = 0.2010 | LCC = 0.9648 | SRCC = 0.9655 ] +INFO:root:Average loss=0.05016200940310955 +INFO:root: +[Valid][22][UTT][ MSE = 0.1819 | LCC = 0.9644 | SRCC = 0.9646 ] +INFO:root:Average loss=0.04670997634530068 +INFO:root: +[Valid][23][UTT][ MSE = 0.1791 | LCC = 0.9641 | SRCC = 0.9643 ] +INFO:root:Average loss=0.0481313890516758 +INFO:root: +[Valid][24][UTT][ MSE = 0.2124 | LCC = 0.9648 | SRCC = 0.9660 ] +INFO:root:Average loss=0.04283826990425587 +INFO:root: +[Valid][25][UTT][ MSE = 0.1622 | LCC = 0.9655 | SRCC = 0.9660 ] +INFO:root:Average loss=0.04327587200701237 +INFO:root: +[Valid][26][UTT][ MSE = 0.1799 | LCC = 0.9652 | SRCC = 0.9650 ] +INFO:root:Average loss=0.04203475396335125 +INFO:root: +[Valid][27][UTT][ MSE = 0.1777 | LCC = 0.9641 | SRCC = 0.9652 ] +INFO:root:Average loss=0.04009726122021675 +INFO:root: +[Valid][28][UTT][ MSE = 0.1802 | LCC = 0.9640 | SRCC = 0.9646 ] +INFO:root:Average loss=0.03925517615675926 +INFO:root: +[Valid][29][UTT][ MSE = 0.1655 | LCC = 0.9646 | SRCC = 0.9654 ] +INFO:root:Average loss=0.039064822137355806 +INFO:root: +[Valid][30][UTT][ MSE = 0.1708 | LCC = 0.9646 | SRCC = 0.9647 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1714 | LCC = 0.9628 | SRCC = 0.9637 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run3/config.gin b/results_tencent/w2v2_large_xlsr53/layer16/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run3/model.pt b/results_tencent/w2v2_large_xlsr53/layer16/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..957cc50869e886d1da8fc44b66920a2792618e73 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a472f571e145aa6954032e1041ea5cdada4b5551783a52068b41888aaa01a814 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run3/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer16/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ac91fb60b42ce5ecf7f1d0a3828f366694c653b5 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:48131967728c01ea4ecb0625c88a41891394974429d0d76b2fe34dca400c2a97 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run3/train.log b/results_tencent/w2v2_large_xlsr53/layer16/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c3558d8e30df5132e9486913b66092f441864802 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3867712896466255 +INFO:root: +[Valid][0][UTT][ MSE = 0.2952 | LCC = 0.9344 | SRCC = 0.9359 ] +INFO:root:Average loss=0.17220624601840973 +INFO:root: +[Valid][1][UTT][ MSE = 0.2383 | LCC = 0.9469 | SRCC = 0.9486 ] +INFO:root:Average loss=0.14255747246742248 +INFO:root: +[Valid][2][UTT][ MSE = 0.2767 | LCC = 0.9509 | SRCC = 0.9524 ] +INFO:root:Average loss=0.1278865799307823 +INFO:root: +[Valid][3][UTT][ MSE = 0.2396 | LCC = 0.9547 | SRCC = 0.9558 ] +INFO:root:Average loss=0.11417831346392632 +INFO:root: +[Valid][4][UTT][ MSE = 0.2266 | LCC = 0.9592 | SRCC = 0.9607 ] +INFO:root:Average loss=0.11100525206327438 +INFO:root: +[Valid][5][UTT][ MSE = 0.2814 | LCC = 0.9573 | SRCC = 0.9589 ] +INFO:root:Average loss=0.0985394096672535 +INFO:root: +[Valid][6][UTT][ MSE = 0.1956 | LCC = 0.9614 | SRCC = 0.9617 ] +INFO:root:Average loss=0.09173552915453911 +INFO:root: +[Valid][7][UTT][ MSE = 0.2224 | LCC = 0.9615 | SRCC = 0.9624 ] +INFO:root:Average loss=0.08915138491988182 +INFO:root: +[Valid][8][UTT][ MSE = 0.2074 | LCC = 0.9620 | SRCC = 0.9627 ] +INFO:root:Average loss=0.08321616068482399 +INFO:root: +[Valid][9][UTT][ MSE = 0.2125 | LCC = 0.9617 | SRCC = 0.9630 ] +INFO:root:Average loss=0.08062839359045029 +INFO:root: +[Valid][10][UTT][ MSE = 0.2604 | LCC = 0.9623 | SRCC = 0.9634 ] +INFO:root:Average loss=0.0746037275493145 +INFO:root: +[Valid][11][UTT][ MSE = 0.1775 | LCC = 0.9621 | SRCC = 0.9631 ] +INFO:root:Average loss=0.07233942970633507 +INFO:root: +[Valid][12][UTT][ MSE = 0.2329 | LCC = 0.9640 | SRCC = 0.9649 ] +INFO:root:Average loss=0.0687949711382389 +INFO:root: +[Valid][13][UTT][ MSE = 0.2036 | LCC = 0.9631 | SRCC = 0.9639 ] +INFO:root:Average loss=0.06983149147033692 +INFO:root: +[Valid][14][UTT][ MSE = 0.2234 | LCC = 0.9627 | SRCC = 0.9638 ] +INFO:root:Average loss=0.062275347352027895 +INFO:root: +[Valid][15][UTT][ MSE = 0.2307 | LCC = 0.9616 | SRCC = 0.9641 ] +INFO:root:Average loss=0.06717544862627983 +INFO:root: +[Valid][16][UTT][ MSE = 0.2127 | LCC = 0.9647 | SRCC = 0.9653 ] +INFO:root:Average loss=0.05822282910346985 +INFO:root: +[Valid][17][UTT][ MSE = 0.1933 | LCC = 0.9647 | SRCC = 0.9651 ] +INFO:root:Average loss=0.05589316287636757 +INFO:root: +[Valid][18][UTT][ MSE = 0.1971 | LCC = 0.9647 | SRCC = 0.9654 ] +INFO:root:Average loss=0.05462197203934192 +INFO:root: +[Valid][19][UTT][ MSE = 0.2047 | LCC = 0.9646 | SRCC = 0.9654 ] +INFO:root:Average loss=0.05305630677938461 +INFO:root: +[Valid][20][UTT][ MSE = 0.1760 | LCC = 0.9647 | SRCC = 0.9647 ] +INFO:root:Average loss=0.051809060096740724 +INFO:root: +[Valid][21][UTT][ MSE = 0.1958 | LCC = 0.9643 | SRCC = 0.9645 ] +INFO:root:Average loss=0.04792577981948853 +INFO:root: +[Valid][22][UTT][ MSE = 0.1961 | LCC = 0.9649 | SRCC = 0.9647 ] +INFO:root:Average loss=0.04641508440673351 +INFO:root: +[Valid][23][UTT][ MSE = 0.1823 | LCC = 0.9651 | SRCC = 0.9650 ] +INFO:root:Average loss=0.042392783507704736 +INFO:root: +[Valid][24][UTT][ MSE = 0.1844 | LCC = 0.9658 | SRCC = 0.9658 ] +INFO:root:Average loss=0.04413654506206512 +INFO:root: +[Valid][25][UTT][ MSE = 0.1965 | LCC = 0.9645 | SRCC = 0.9654 ] +INFO:root:Average loss=0.04122439403831959 +INFO:root: +[Valid][26][UTT][ MSE = 0.2024 | LCC = 0.9643 | SRCC = 0.9651 ] +INFO:root:Average loss=0.040303936496376994 +INFO:root: +[Valid][27][UTT][ MSE = 0.1893 | LCC = 0.9649 | SRCC = 0.9645 ] +INFO:root:Average loss=0.03909369133412838 +INFO:root: +[Valid][28][UTT][ MSE = 0.1739 | LCC = 0.9640 | SRCC = 0.9641 ] +INFO:root:Average loss=0.035445221692323683 +INFO:root: +[Valid][29][UTT][ MSE = 0.2020 | LCC = 0.9638 | SRCC = 0.9644 ] +INFO:root:Average loss=0.038194004774093626 +INFO:root: +[Valid][30][UTT][ MSE = 0.1880 | LCC = 0.9640 | SRCC = 0.9642 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1903 | LCC = 0.9646 | SRCC = 0.9654 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run4/config.gin b/results_tencent/w2v2_large_xlsr53/layer16/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run4/model.pt b/results_tencent/w2v2_large_xlsr53/layer16/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f3f0960a588569c899b5775b0ddfa52f5ffc2123 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0badcbf181d0128aa40c72434b38e5f5836d4dd4334a21ed3694d79a147dc46d +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run4/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer16/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..94f7c2ff8db6f6e63455beb6960a1eed6fe02299 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:35d602f39593bf0d619e511b775fbd0d8a4cfa7751a30969cff32f9025315e20 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer16/run4/train.log b/results_tencent/w2v2_large_xlsr53/layer16/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..23a58dc6902c14d125610e0851ff742c0b72e5f2 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer16/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42471618187427523 +INFO:root: +[Valid][0][UTT][ MSE = 0.2783 | LCC = 0.9227 | SRCC = 0.9248 ] +INFO:root:Average loss=0.19946184504032136 +INFO:root: +[Valid][1][UTT][ MSE = 0.2417 | LCC = 0.9426 | SRCC = 0.9448 ] +INFO:root:Average loss=0.16566746759414672 +INFO:root: +[Valid][2][UTT][ MSE = 0.2267 | LCC = 0.9501 | SRCC = 0.9519 ] +INFO:root:Average loss=0.13649350398778914 +INFO:root: +[Valid][3][UTT][ MSE = 0.1953 | LCC = 0.9535 | SRCC = 0.9558 ] +INFO:root:Average loss=0.12105500382184982 +INFO:root: +[Valid][4][UTT][ MSE = 0.2199 | LCC = 0.9545 | SRCC = 0.9577 ] +INFO:root:Average loss=0.11417307758331299 +INFO:root: +[Valid][5][UTT][ MSE = 0.1957 | LCC = 0.9578 | SRCC = 0.9585 ] +INFO:root:Average loss=0.10884602305293083 +INFO:root: +[Valid][6][UTT][ MSE = 0.1982 | LCC = 0.9572 | SRCC = 0.9588 ] +INFO:root:Average loss=0.09843042230606079 +INFO:root: +[Valid][7][UTT][ MSE = 0.2183 | LCC = 0.9591 | SRCC = 0.9608 ] +INFO:root:Average loss=0.08833120930194854 +INFO:root: +[Valid][8][UTT][ MSE = 0.1877 | LCC = 0.9605 | SRCC = 0.9622 ] +INFO:root:Average loss=0.08820579817891121 +INFO:root: +[Valid][9][UTT][ MSE = 0.1815 | LCC = 0.9617 | SRCC = 0.9626 ] +INFO:root:Average loss=0.08029557153582573 +INFO:root: +[Valid][10][UTT][ MSE = 0.1705 | LCC = 0.9615 | SRCC = 0.9623 ] +INFO:root:Average loss=0.08015782138705253 +INFO:root: +[Valid][11][UTT][ MSE = 0.1834 | LCC = 0.9612 | SRCC = 0.9629 ] +INFO:root:Average loss=0.07582131958007812 +INFO:root: +[Valid][12][UTT][ MSE = 0.2280 | LCC = 0.9600 | SRCC = 0.9625 ] +INFO:root:Average loss=0.07151784321665763 +INFO:root: +[Valid][13][UTT][ MSE = 0.1763 | LCC = 0.9611 | SRCC = 0.9630 ] +INFO:root:Average loss=0.06686930578947067 +INFO:root: +[Valid][14][UTT][ MSE = 0.1883 | LCC = 0.9622 | SRCC = 0.9632 ] +INFO:root:Average loss=0.06422859543561936 +INFO:root: +[Valid][15][UTT][ MSE = 0.1606 | LCC = 0.9645 | SRCC = 0.9651 ] +INFO:root:Average loss=0.06030838352441788 +INFO:root: +[Valid][16][UTT][ MSE = 0.1902 | LCC = 0.9639 | SRCC = 0.9644 ] +INFO:root:Average loss=0.057179634153842925 +INFO:root: +[Valid][17][UTT][ MSE = 0.1687 | LCC = 0.9650 | SRCC = 0.9652 ] +INFO:root:Average loss=0.057527063429355624 +INFO:root: +[Valid][18][UTT][ MSE = 0.1697 | LCC = 0.9640 | SRCC = 0.9644 ] +INFO:root:Average loss=0.05328825770318508 +INFO:root: +[Valid][19][UTT][ MSE = 0.1879 | LCC = 0.9622 | SRCC = 0.9630 ] +INFO:root:Average loss=0.052249226272106174 +INFO:root: +[Valid][20][UTT][ MSE = 0.1659 | LCC = 0.9636 | SRCC = 0.9649 ] +INFO:root:Average loss=0.04978218297660351 +INFO:root: +[Valid][21][UTT][ MSE = 0.1778 | LCC = 0.9634 | SRCC = 0.9637 ] +INFO:root:Average loss=0.04660342912375927 +INFO:root: +[Valid][22][UTT][ MSE = 0.1804 | LCC = 0.9639 | SRCC = 0.9642 ] +INFO:root:Average loss=0.04470174925029278 +INFO:root: +[Valid][23][UTT][ MSE = 0.1795 | LCC = 0.9638 | SRCC = 0.9645 ] +INFO:root:Average loss=0.04327949111163616 +INFO:root: +[Valid][24][UTT][ MSE = 0.1940 | LCC = 0.9641 | SRCC = 0.9642 ] +INFO:root:Average loss=0.0418251443952322 +INFO:root: +[Valid][25][UTT][ MSE = 0.1782 | LCC = 0.9639 | SRCC = 0.9641 ] +INFO:root:Average loss=0.03978823095560074 +INFO:root: +[Valid][26][UTT][ MSE = 0.1820 | LCC = 0.9618 | SRCC = 0.9620 ] +INFO:root:Average loss=0.0383830453902483 +INFO:root: +[Valid][27][UTT][ MSE = 0.1911 | LCC = 0.9626 | SRCC = 0.9632 ] +INFO:root:Average loss=0.037997677430510524 +INFO:root: +[Valid][28][UTT][ MSE = 0.1725 | LCC = 0.9631 | SRCC = 0.9632 ] +INFO:root:Average loss=0.03766191820800304 +INFO:root: +[Valid][29][UTT][ MSE = 0.1690 | LCC = 0.9629 | SRCC = 0.9629 ] +INFO:root:Average loss=0.035681977346539495 +INFO:root: +[Valid][30][UTT][ MSE = 0.1700 | LCC = 0.9640 | SRCC = 0.9644 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1737 | LCC = 0.9634 | SRCC = 0.9640 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run0/config.gin b/results_tencent/w2v2_large_xlsr53/layer18/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run0/model.pt b/results_tencent/w2v2_large_xlsr53/layer18/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bf6286ef15c29aeae6b6be10e938cf56338887fc --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d1233313b3f5a6aa47dea0cc1a45fb45beb05478b6b736cdb88a26b7234213d4 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run0/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer18/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..47f9e9f614806d3166669bf79a3e600726b1bf5d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0bf01cfe020dfb72c07053cf7ba1b07fa45893b2d2d1301ee15978f51c66a7c4 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run0/train.log b/results_tencent/w2v2_large_xlsr53/layer18/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c5ccf9c7dd81ab520fb36c57ec2e0405b11f0913 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42341373109817504 +INFO:root: +[Valid][0][UTT][ MSE = 0.3899 | LCC = 0.9185 | SRCC = 0.9246 ] +INFO:root:Average loss=0.17956506073474884 +INFO:root: +[Valid][1][UTT][ MSE = 0.2589 | LCC = 0.9488 | SRCC = 0.9495 ] +INFO:root:Average loss=0.14962590074539184 +INFO:root: +[Valid][2][UTT][ MSE = 0.2137 | LCC = 0.9525 | SRCC = 0.9549 ] +INFO:root:Average loss=0.13649285623431207 +INFO:root: +[Valid][3][UTT][ MSE = 0.2144 | LCC = 0.9562 | SRCC = 0.9563 ] +INFO:root:Average loss=0.12012034165859223 +INFO:root: +[Valid][4][UTT][ MSE = 0.2350 | LCC = 0.9590 | SRCC = 0.9596 ] +INFO:root:Average loss=0.11072437453269958 +INFO:root: +[Valid][5][UTT][ MSE = 0.2531 | LCC = 0.9579 | SRCC = 0.9601 ] +INFO:root:Average loss=0.10029066696763039 +INFO:root: +[Valid][6][UTT][ MSE = 0.2576 | LCC = 0.9603 | SRCC = 0.9616 ] +INFO:root:Average loss=0.09942198863625526 +INFO:root: +[Valid][7][UTT][ MSE = 0.2152 | LCC = 0.9614 | SRCC = 0.9622 ] +INFO:root:Average loss=0.09302461525797844 +INFO:root: +[Valid][8][UTT][ MSE = 0.2121 | LCC = 0.9591 | SRCC = 0.9603 ] +INFO:root:Average loss=0.08847957274317741 +INFO:root: +[Valid][9][UTT][ MSE = 0.2052 | LCC = 0.9619 | SRCC = 0.9624 ] +INFO:root:Average loss=0.08289432799816132 +INFO:root: +[Valid][10][UTT][ MSE = 0.1854 | LCC = 0.9635 | SRCC = 0.9637 ] +INFO:root:Average loss=0.07872265613079071 +INFO:root: +[Valid][11][UTT][ MSE = 0.1891 | LCC = 0.9628 | SRCC = 0.9625 ] +INFO:root:Average loss=0.07266114568710327 +INFO:root: +[Valid][12][UTT][ MSE = 0.2229 | LCC = 0.9621 | SRCC = 0.9633 ] +INFO:root:Average loss=0.07666573294997216 +INFO:root: +[Valid][13][UTT][ MSE = 0.2038 | LCC = 0.9623 | SRCC = 0.9630 ] +INFO:root:Average loss=0.0727660091817379 +INFO:root: +[Valid][14][UTT][ MSE = 0.2166 | LCC = 0.9629 | SRCC = 0.9635 ] +INFO:root:Average loss=0.06529303961992264 +INFO:root: +[Valid][15][UTT][ MSE = 0.1925 | LCC = 0.9644 | SRCC = 0.9643 ] +INFO:root:Average loss=0.06479722318053245 +INFO:root: +[Valid][16][UTT][ MSE = 0.1959 | LCC = 0.9631 | SRCC = 0.9632 ] +INFO:root:Average loss=0.05788569304347038 +INFO:root: +[Valid][17][UTT][ MSE = 0.1833 | LCC = 0.9650 | SRCC = 0.9646 ] +INFO:root:Average loss=0.06002878201007843 +INFO:root: +[Valid][18][UTT][ MSE = 0.2150 | LCC = 0.9640 | SRCC = 0.9641 ] +INFO:root:Average loss=0.05767410808801651 +INFO:root: +[Valid][19][UTT][ MSE = 0.1855 | LCC = 0.9645 | SRCC = 0.9642 ] +INFO:root:Average loss=0.05588714936375618 +INFO:root: +[Valid][20][UTT][ MSE = 0.1742 | LCC = 0.9632 | SRCC = 0.9629 ] +INFO:root:Average loss=0.051100949302315714 +INFO:root: +[Valid][21][UTT][ MSE = 0.2234 | LCC = 0.9637 | SRCC = 0.9644 ] +INFO:root:Average loss=0.05157324762642384 +INFO:root: +[Valid][22][UTT][ MSE = 0.1997 | LCC = 0.9651 | SRCC = 0.9652 ] +INFO:root:Average loss=0.04960222363471985 +INFO:root: +[Valid][23][UTT][ MSE = 0.1726 | LCC = 0.9653 | SRCC = 0.9647 ] +INFO:root:Average loss=0.04508243387937546 +INFO:root: +[Valid][24][UTT][ MSE = 0.1949 | LCC = 0.9652 | SRCC = 0.9647 ] +INFO:root:Average loss=0.04443148861825466 +INFO:root: +[Valid][25][UTT][ MSE = 0.1839 | LCC = 0.9649 | SRCC = 0.9645 ] +INFO:root:Average loss=0.04241157893836498 +INFO:root: +[Valid][26][UTT][ MSE = 0.2195 | LCC = 0.9650 | SRCC = 0.9646 ] +INFO:root:Average loss=0.040662298053503035 +INFO:root: +[Valid][27][UTT][ MSE = 0.1761 | LCC = 0.9648 | SRCC = 0.9644 ] +INFO:root:Average loss=0.03891611412167549 +INFO:root: +[Valid][28][UTT][ MSE = 0.2122 | LCC = 0.9645 | SRCC = 0.9640 ] +INFO:root:Average loss=0.0379370764195919 +INFO:root: +[Valid][29][UTT][ MSE = 0.1997 | LCC = 0.9646 | SRCC = 0.9641 ] +INFO:root:Average loss=0.037878991633653644 +INFO:root: +[Valid][30][UTT][ MSE = 0.2011 | LCC = 0.9645 | SRCC = 0.9648 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1723 | LCC = 0.9648 | SRCC = 0.9657 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run1/config.gin b/results_tencent/w2v2_large_xlsr53/layer18/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run1/model.pt b/results_tencent/w2v2_large_xlsr53/layer18/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9a3d71d5dfd54c97648c92130385fcb372ac2547 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ac52d7452348834b9d7bc8b6490e4699af43b9f33fc80dd209e15b41db34182c +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run1/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer18/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..5f47d10ce56e2f8c52d267fd1635cbd7f135ea1d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f943ae7d3313032ade92685352741922f8832d3ec870cdf5208dcde47cd2ec4 +size 3404998 diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run1/train.log b/results_tencent/w2v2_large_xlsr53/layer18/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..16dbdf4ca98d6cc32bb5ef4fa2a26274e2d851da --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.39853864228725433 +INFO:root: +[Valid][0][UTT][ MSE = 0.3073 | LCC = 0.9349 | SRCC = 0.9353 ] +INFO:root:Average loss=0.17729806166887283 +INFO:root: +[Valid][1][UTT][ MSE = 0.2562 | LCC = 0.9494 | SRCC = 0.9499 ] +INFO:root:Average loss=0.14858112353086472 +INFO:root: +[Valid][2][UTT][ MSE = 0.2722 | LCC = 0.9529 | SRCC = 0.9531 ] +INFO:root:Average loss=0.13632602047920228 +INFO:root: +[Valid][3][UTT][ MSE = 0.2131 | LCC = 0.9555 | SRCC = 0.9565 ] +INFO:root:Average loss=0.12097495585680008 +INFO:root: +[Valid][4][UTT][ MSE = 0.1984 | LCC = 0.9586 | SRCC = 0.9596 ] +INFO:root:Average loss=0.1081818773150444 +INFO:root: +[Valid][5][UTT][ MSE = 0.2513 | LCC = 0.9607 | SRCC = 0.9614 ] +INFO:root:Average loss=0.10127050822973252 +INFO:root: +[Valid][6][UTT][ MSE = 0.1920 | LCC = 0.9618 | SRCC = 0.9620 ] +INFO:root:Average loss=0.09361083370447158 +INFO:root: +[Valid][7][UTT][ MSE = 0.1833 | LCC = 0.9618 | SRCC = 0.9630 ] +INFO:root:Average loss=0.08802493944764138 +INFO:root: +[Valid][8][UTT][ MSE = 0.2068 | LCC = 0.9637 | SRCC = 0.9641 ] +INFO:root:Average loss=0.08503871887922287 +INFO:root: +[Valid][9][UTT][ MSE = 0.2000 | LCC = 0.9609 | SRCC = 0.9623 ] +INFO:root:Average loss=0.07932502844929695 +INFO:root: +[Valid][10][UTT][ MSE = 0.2289 | LCC = 0.9642 | SRCC = 0.9645 ] +INFO:root:Average loss=0.07711989444494248 +INFO:root: +[Valid][11][UTT][ MSE = 0.1979 | LCC = 0.9657 | SRCC = 0.9658 ] +INFO:root:Average loss=0.07370769396424294 +INFO:root: +[Valid][12][UTT][ MSE = 0.1829 | LCC = 0.9650 | SRCC = 0.9651 ] +INFO:root:Average loss=0.07328623697161675 +INFO:root: +[Valid][13][UTT][ MSE = 0.2055 | LCC = 0.9651 | SRCC = 0.9650 ] +INFO:root:Average loss=0.06468822398781776 +INFO:root: +[Valid][14][UTT][ MSE = 0.1865 | LCC = 0.9619 | SRCC = 0.9620 ] +INFO:root:Average loss=0.06154419207572937 +INFO:root: +[Valid][15][UTT][ MSE = 0.1850 | LCC = 0.9651 | SRCC = 0.9652 ] +INFO:root:Average loss=0.06318631002306938 +INFO:root: +[Valid][16][UTT][ MSE = 0.1849 | LCC = 0.9642 | SRCC = 0.9640 ] +INFO:root:Average loss=0.060917133927345275 +INFO:root: +[Valid][17][UTT][ MSE = 0.1903 | LCC = 0.9641 | SRCC = 0.9638 ] +INFO:root:Average loss=0.056624400749802586 +INFO:root: +[Valid][18][UTT][ MSE = 0.1743 | LCC = 0.9658 | SRCC = 0.9654 ] +INFO:root:Average loss=0.05614767688512802 +INFO:root: +[Valid][19][UTT][ MSE = 0.2079 | LCC = 0.9653 | SRCC = 0.9648 ] +INFO:root:Average loss=0.05205449151992798 +INFO:root: +[Valid][20][UTT][ MSE = 0.1824 | LCC = 0.9656 | SRCC = 0.9650 ] +INFO:root:Average loss=0.04913279983401299 +INFO:root: +[Valid][21][UTT][ MSE = 0.1870 | LCC = 0.9671 | SRCC = 0.9668 ] +INFO:root:Average loss=0.04768527881801128 +INFO:root: +[Valid][22][UTT][ MSE = 0.2297 | LCC = 0.9639 | SRCC = 0.9661 ] +INFO:root:Average loss=0.04900757835805416 +INFO:root: +[Valid][23][UTT][ MSE = 0.1819 | LCC = 0.9649 | SRCC = 0.9651 ] +INFO:root:Average loss=0.047629676014184955 +INFO:root: +[Valid][24][UTT][ MSE = 0.1793 | LCC = 0.9659 | SRCC = 0.9656 ] +INFO:root:Average loss=0.042987432166934014 +INFO:root: +[Valid][25][UTT][ MSE = 0.2202 | LCC = 0.9647 | SRCC = 0.9650 ] +INFO:root:Average loss=0.042977860406041145 +INFO:root: +[Valid][26][UTT][ MSE = 0.2196 | LCC = 0.9646 | SRCC = 0.9647 ] +INFO:root:Average loss=0.04017806440591812 +INFO:root: +[Valid][27][UTT][ MSE = 0.1908 | LCC = 0.9653 | SRCC = 0.9647 ] +INFO:root:Average loss=0.03777063684165478 +INFO:root: +[Valid][28][UTT][ MSE = 0.1982 | LCC = 0.9658 | SRCC = 0.9658 ] +INFO:root:Average loss=0.039022227227687833 +INFO:root: +[Valid][29][UTT][ MSE = 0.1855 | LCC = 0.9657 | SRCC = 0.9657 ] +INFO:root:Average loss=0.03590583598613739 +INFO:root: +[Valid][30][UTT][ MSE = 0.1752 | LCC = 0.9657 | SRCC = 0.9655 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1836 | LCC = 0.9675 | SRCC = 0.9685 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run2/config.gin b/results_tencent/w2v2_large_xlsr53/layer18/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run2/model.pt b/results_tencent/w2v2_large_xlsr53/layer18/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ccc08997c905265cfb3409e2432591f41de4e313 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8eebf0e5bf169af14dcf0778f139f6acea217781d28afa5cdbe85745c73b8b6b +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run2/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer18/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b13abbe05df303bca9fadebd43134a429e8c2c22 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fb8e4fe4307f95c5b11aaa81d613cecbf729cc7ad287d3d7315493cb6f87c860 +size 3404998 diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run2/train.log b/results_tencent/w2v2_large_xlsr53/layer18/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..52b1399cf38b32eaffd86acc3abb0a15bf054dd5 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4399495187997818 +INFO:root: +[Valid][0][UTT][ MSE = 0.3141 | LCC = 0.9258 | SRCC = 0.9277 ] +INFO:root:Average loss=0.18862748157978057 +INFO:root: +[Valid][1][UTT][ MSE = 0.2348 | LCC = 0.9468 | SRCC = 0.9468 ] +INFO:root:Average loss=0.1498182812333107 +INFO:root: +[Valid][2][UTT][ MSE = 0.2083 | LCC = 0.9508 | SRCC = 0.9514 ] +INFO:root:Average loss=0.13310633844137193 +INFO:root: +[Valid][3][UTT][ MSE = 0.2144 | LCC = 0.9571 | SRCC = 0.9578 ] +INFO:root:Average loss=0.1211050914824009 +INFO:root: +[Valid][4][UTT][ MSE = 0.2187 | LCC = 0.9575 | SRCC = 0.9589 ] +INFO:root:Average loss=0.11038151749968529 +INFO:root: +[Valid][5][UTT][ MSE = 0.2555 | LCC = 0.9563 | SRCC = 0.9586 ] +INFO:root:Average loss=0.09947734275460243 +INFO:root: +[Valid][6][UTT][ MSE = 0.2215 | LCC = 0.9597 | SRCC = 0.9599 ] +INFO:root:Average loss=0.102354670971632 +INFO:root: +[Valid][7][UTT][ MSE = 0.2034 | LCC = 0.9621 | SRCC = 0.9624 ] +INFO:root:Average loss=0.08646605578064918 +INFO:root: +[Valid][8][UTT][ MSE = 0.2098 | LCC = 0.9635 | SRCC = 0.9644 ] +INFO:root:Average loss=0.08803321471810341 +INFO:root: +[Valid][9][UTT][ MSE = 0.1788 | LCC = 0.9625 | SRCC = 0.9633 ] +INFO:root:Average loss=0.08029684007167816 +INFO:root: +[Valid][10][UTT][ MSE = 0.1739 | LCC = 0.9638 | SRCC = 0.9639 ] +INFO:root:Average loss=0.07673537158966065 +INFO:root: +[Valid][11][UTT][ MSE = 0.1735 | LCC = 0.9636 | SRCC = 0.9644 ] +INFO:root:Average loss=0.07262680581212044 +INFO:root: +[Valid][12][UTT][ MSE = 0.1758 | LCC = 0.9642 | SRCC = 0.9641 ] +INFO:root:Average loss=0.07151822119951248 +INFO:root: +[Valid][13][UTT][ MSE = 0.1972 | LCC = 0.9641 | SRCC = 0.9647 ] +INFO:root:Average loss=0.06634227791428567 +INFO:root: +[Valid][14][UTT][ MSE = 0.1726 | LCC = 0.9651 | SRCC = 0.9651 ] +INFO:root:Average loss=0.06411724156141281 +INFO:root: +[Valid][15][UTT][ MSE = 0.1781 | LCC = 0.9653 | SRCC = 0.9656 ] +INFO:root:Average loss=0.065232392847538 +INFO:root: +[Valid][16][UTT][ MSE = 0.1697 | LCC = 0.9645 | SRCC = 0.9647 ] +INFO:root:Average loss=0.060212334215641025 +INFO:root: +[Valid][17][UTT][ MSE = 0.1890 | LCC = 0.9650 | SRCC = 0.9648 ] +INFO:root:Average loss=0.05661498248577118 +INFO:root: +[Valid][18][UTT][ MSE = 0.1889 | LCC = 0.9649 | SRCC = 0.9651 ] +INFO:root:Average loss=0.057761523962020875 +INFO:root: +[Valid][19][UTT][ MSE = 0.2032 | LCC = 0.9648 | SRCC = 0.9648 ] +INFO:root:Average loss=0.05305258908867836 +INFO:root: +[Valid][20][UTT][ MSE = 0.1826 | LCC = 0.9646 | SRCC = 0.9648 ] +INFO:root:Average loss=0.050264503121376036 +INFO:root: +[Valid][21][UTT][ MSE = 0.1851 | LCC = 0.9646 | SRCC = 0.9650 ] +INFO:root:Average loss=0.05033259184658528 +INFO:root: +[Valid][22][UTT][ MSE = 0.1635 | LCC = 0.9635 | SRCC = 0.9634 ] +INFO:root:Average loss=0.04892983755469322 +INFO:root: +[Valid][23][UTT][ MSE = 0.1719 | LCC = 0.9639 | SRCC = 0.9645 ] +INFO:root:Average loss=0.0474307521134615 +INFO:root: +[Valid][24][UTT][ MSE = 0.1882 | LCC = 0.9647 | SRCC = 0.9647 ] +INFO:root:Average loss=0.04296626025438309 +INFO:root: +[Valid][25][UTT][ MSE = 0.1611 | LCC = 0.9651 | SRCC = 0.9649 ] +INFO:root:Average loss=0.04239057499170303 +INFO:root: +[Valid][26][UTT][ MSE = 0.1632 | LCC = 0.9635 | SRCC = 0.9638 ] +INFO:root:Average loss=0.0421104221791029 +INFO:root: +[Valid][27][UTT][ MSE = 0.1773 | LCC = 0.9612 | SRCC = 0.9629 ] +INFO:root:Average loss=0.04306957690417767 +INFO:root: +[Valid][28][UTT][ MSE = 0.1643 | LCC = 0.9626 | SRCC = 0.9627 ] +INFO:root:Average loss=0.041108330562710765 +INFO:root: +[Valid][29][UTT][ MSE = 0.1881 | LCC = 0.9636 | SRCC = 0.9630 ] +INFO:root:Average loss=0.03793383139371872 +INFO:root: +[Valid][30][UTT][ MSE = 0.1790 | LCC = 0.9646 | SRCC = 0.9643 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1803 | LCC = 0.9653 | SRCC = 0.9660 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run3/config.gin b/results_tencent/w2v2_large_xlsr53/layer18/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run3/model.pt b/results_tencent/w2v2_large_xlsr53/layer18/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..eef75772d1b7f38de136418ceb3ce869979e728e --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0269c125731e4a06eeb4b2be38b2be31c30c8f0574747091e078508ac71ae4c +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run3/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer18/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..214b1362e0d71c3cde1e96b860033feffb73e009 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d88a8b96f8ba5f618cd1dfd55c5a546c3e35cc5f0a09911388a4d67e0ffe4daa +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run3/train.log b/results_tencent/w2v2_large_xlsr53/layer18/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..06236eb06a5d853eed08dc832eb7c7893381490d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4204043707847595 +INFO:root: +[Valid][0][UTT][ MSE = 0.2748 | LCC = 0.9301 | SRCC = 0.9312 ] +INFO:root:Average loss=0.19298544615507127 +INFO:root: +[Valid][1][UTT][ MSE = 0.2442 | LCC = 0.9456 | SRCC = 0.9468 ] +INFO:root:Average loss=0.15440139311552048 +INFO:root: +[Valid][2][UTT][ MSE = 0.2167 | LCC = 0.9501 | SRCC = 0.9525 ] +INFO:root:Average loss=0.13240109819173812 +INFO:root: +[Valid][3][UTT][ MSE = 0.2206 | LCC = 0.9528 | SRCC = 0.9540 ] +INFO:root:Average loss=0.1216648034453392 +INFO:root: +[Valid][4][UTT][ MSE = 0.1886 | LCC = 0.9573 | SRCC = 0.9582 ] +INFO:root:Average loss=0.10811400824785232 +INFO:root: +[Valid][5][UTT][ MSE = 0.2117 | LCC = 0.9586 | SRCC = 0.9598 ] +INFO:root:Average loss=0.10155244809389115 +INFO:root: +[Valid][6][UTT][ MSE = 0.2273 | LCC = 0.9608 | SRCC = 0.9617 ] +INFO:root:Average loss=0.09624984952807426 +INFO:root: +[Valid][7][UTT][ MSE = 0.2201 | LCC = 0.9613 | SRCC = 0.9619 ] +INFO:root:Average loss=0.0907769635617733 +INFO:root: +[Valid][8][UTT][ MSE = 0.2376 | LCC = 0.9589 | SRCC = 0.9607 ] +INFO:root:Average loss=0.0904927014708519 +INFO:root: +[Valid][9][UTT][ MSE = 0.1831 | LCC = 0.9613 | SRCC = 0.9623 ] +INFO:root:Average loss=0.08095414623618126 +INFO:root: +[Valid][10][UTT][ MSE = 0.2640 | LCC = 0.9606 | SRCC = 0.9624 ] +INFO:root:Average loss=0.08546674504876137 +INFO:root: +[Valid][11][UTT][ MSE = 0.2042 | LCC = 0.9639 | SRCC = 0.9640 ] +INFO:root:Average loss=0.07496187788248063 +INFO:root: +[Valid][12][UTT][ MSE = 0.2190 | LCC = 0.9635 | SRCC = 0.9647 ] +INFO:root:Average loss=0.07010074478387833 +INFO:root: +[Valid][13][UTT][ MSE = 0.2308 | LCC = 0.9633 | SRCC = 0.9639 ] +INFO:root:Average loss=0.06842223933339119 +INFO:root: +[Valid][14][UTT][ MSE = 0.1745 | LCC = 0.9623 | SRCC = 0.9633 ] +INFO:root:Average loss=0.06550950625538826 +INFO:root: +[Valid][15][UTT][ MSE = 0.2229 | LCC = 0.9640 | SRCC = 0.9645 ] +INFO:root:Average loss=0.06519931954145432 +INFO:root: +[Valid][16][UTT][ MSE = 0.1980 | LCC = 0.9639 | SRCC = 0.9639 ] +INFO:root:Average loss=0.059506081730127336 +INFO:root: +[Valid][17][UTT][ MSE = 0.1915 | LCC = 0.9646 | SRCC = 0.9647 ] +INFO:root:Average loss=0.06170048554241657 +INFO:root: +[Valid][18][UTT][ MSE = 0.2241 | LCC = 0.9641 | SRCC = 0.9652 ] +INFO:root:Average loss=0.05632220643758774 +INFO:root: +[Valid][19][UTT][ MSE = 0.1740 | LCC = 0.9645 | SRCC = 0.9648 ] +INFO:root:Average loss=0.05420557284355164 +INFO:root: +[Valid][20][UTT][ MSE = 0.2093 | LCC = 0.9654 | SRCC = 0.9654 ] +INFO:root:Average loss=0.052693799555301665 +INFO:root: +[Valid][21][UTT][ MSE = 0.2016 | LCC = 0.9645 | SRCC = 0.9652 ] +INFO:root:Average loss=0.05089446373283863 +INFO:root: +[Valid][22][UTT][ MSE = 0.1973 | LCC = 0.9646 | SRCC = 0.9652 ] +INFO:root:Average loss=0.047380027562379835 +INFO:root: +[Valid][23][UTT][ MSE = 0.1669 | LCC = 0.9644 | SRCC = 0.9646 ] +INFO:root:Average loss=0.04539400291442871 +INFO:root: +[Valid][24][UTT][ MSE = 0.1872 | LCC = 0.9641 | SRCC = 0.9641 ] +INFO:root:Average loss=0.0452081459313631 +INFO:root: +[Valid][25][UTT][ MSE = 0.1801 | LCC = 0.9644 | SRCC = 0.9643 ] +INFO:root:Average loss=0.0431467694491148 +INFO:root: +[Valid][26][UTT][ MSE = 0.2155 | LCC = 0.9650 | SRCC = 0.9653 ] +INFO:root:Average loss=0.045350357592105865 +INFO:root: +[Valid][27][UTT][ MSE = 0.2051 | LCC = 0.9632 | SRCC = 0.9636 ] +INFO:root:Average loss=0.03970570962131023 +INFO:root: +[Valid][28][UTT][ MSE = 0.1646 | LCC = 0.9640 | SRCC = 0.9641 ] +INFO:root:Average loss=0.04123353563249111 +INFO:root: +[Valid][29][UTT][ MSE = 0.1918 | LCC = 0.9652 | SRCC = 0.9654 ] +INFO:root:Average loss=0.038597080647945405 +INFO:root: +[Valid][30][UTT][ MSE = 0.2004 | LCC = 0.9634 | SRCC = 0.9637 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2111 | LCC = 0.9650 | SRCC = 0.9662 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run4/config.gin b/results_tencent/w2v2_large_xlsr53/layer18/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run4/model.pt b/results_tencent/w2v2_large_xlsr53/layer18/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f9f2ead3b69df83d5e4b8a01494ee0e98c7bb8e7 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a8a5960bbf89f7aadd28f22ec594ffee6be9b227f6749e372459bdbe3f8dec5d +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run4/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer18/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2834b583567c16ef18ccf81b333b45df14802cdc --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:abbfecd743acf4fc848433182d87d0817a0bbad9318fa55e9d10d807db101694 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer18/run4/train.log b/results_tencent/w2v2_large_xlsr53/layer18/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..fe073ffbd52f05df2985a6f4629dbe6540dc325e --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer18/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4016834222078323 +INFO:root: +[Valid][0][UTT][ MSE = 0.3069 | LCC = 0.9362 | SRCC = 0.9391 ] +INFO:root:Average loss=0.17967270010709763 +INFO:root: +[Valid][1][UTT][ MSE = 0.2415 | LCC = 0.9449 | SRCC = 0.9459 ] +INFO:root:Average loss=0.13969410234689714 +INFO:root: +[Valid][2][UTT][ MSE = 0.2409 | LCC = 0.9515 | SRCC = 0.9530 ] +INFO:root:Average loss=0.12816126883029938 +INFO:root: +[Valid][3][UTT][ MSE = 0.2210 | LCC = 0.9561 | SRCC = 0.9571 ] +INFO:root:Average loss=0.11233428257703781 +INFO:root: +[Valid][4][UTT][ MSE = 0.2373 | LCC = 0.9574 | SRCC = 0.9589 ] +INFO:root:Average loss=0.10814861559867858 +INFO:root: +[Valid][5][UTT][ MSE = 0.2150 | LCC = 0.9604 | SRCC = 0.9615 ] +INFO:root:Average loss=0.10180462548136711 +INFO:root: +[Valid][6][UTT][ MSE = 0.2121 | LCC = 0.9601 | SRCC = 0.9614 ] +INFO:root:Average loss=0.0946351068019867 +INFO:root: +[Valid][7][UTT][ MSE = 0.1872 | LCC = 0.9607 | SRCC = 0.9621 ] +INFO:root:Average loss=0.08786112824082375 +INFO:root: +[Valid][8][UTT][ MSE = 0.2642 | LCC = 0.9623 | SRCC = 0.9625 ] +INFO:root:Average loss=0.08622520527243614 +INFO:root: +[Valid][9][UTT][ MSE = 0.1978 | LCC = 0.9622 | SRCC = 0.9631 ] +INFO:root:Average loss=0.08370713439583778 +INFO:root: +[Valid][10][UTT][ MSE = 0.2128 | LCC = 0.9600 | SRCC = 0.9619 ] +INFO:root:Average loss=0.07935188236832619 +INFO:root: +[Valid][11][UTT][ MSE = 0.1806 | LCC = 0.9628 | SRCC = 0.9637 ] +INFO:root:Average loss=0.07007203906774521 +INFO:root: +[Valid][12][UTT][ MSE = 0.1645 | LCC = 0.9642 | SRCC = 0.9646 ] +INFO:root:Average loss=0.06918533524870872 +INFO:root: +[Valid][13][UTT][ MSE = 0.2268 | LCC = 0.9602 | SRCC = 0.9618 ] +INFO:root:Average loss=0.0675092308819294 +INFO:root: +[Valid][14][UTT][ MSE = 0.1762 | LCC = 0.9635 | SRCC = 0.9636 ] +INFO:root:Average loss=0.06305601412057876 +INFO:root: +[Valid][15][UTT][ MSE = 0.1832 | LCC = 0.9659 | SRCC = 0.9658 ] +INFO:root:Average loss=0.061647017657756806 +INFO:root: +[Valid][16][UTT][ MSE = 0.1932 | LCC = 0.9647 | SRCC = 0.9650 ] +INFO:root:Average loss=0.061268807232379915 +INFO:root: +[Valid][17][UTT][ MSE = 0.1882 | LCC = 0.9647 | SRCC = 0.9649 ] +INFO:root:Average loss=0.05591421929001808 +INFO:root: +[Valid][18][UTT][ MSE = 0.1901 | LCC = 0.9654 | SRCC = 0.9651 ] +INFO:root:Average loss=0.05281551019847393 +INFO:root: +[Valid][19][UTT][ MSE = 0.1791 | LCC = 0.9653 | SRCC = 0.9655 ] +INFO:root:Average loss=0.053677794635295865 +INFO:root: +[Valid][20][UTT][ MSE = 0.1794 | LCC = 0.9648 | SRCC = 0.9647 ] +INFO:root:Average loss=0.048982585951685906 +INFO:root: +[Valid][21][UTT][ MSE = 0.1930 | LCC = 0.9644 | SRCC = 0.9650 ] +INFO:root:Average loss=0.048170791998505595 +INFO:root: +[Valid][22][UTT][ MSE = 0.1914 | LCC = 0.9645 | SRCC = 0.9652 ] +INFO:root:Average loss=0.04672179460525513 +INFO:root: +[Valid][23][UTT][ MSE = 0.1900 | LCC = 0.9626 | SRCC = 0.9631 ] +INFO:root:Average loss=0.047025115326046946 +INFO:root: +[Valid][24][UTT][ MSE = 0.1787 | LCC = 0.9654 | SRCC = 0.9654 ] +INFO:root:Average loss=0.042933509409427646 +INFO:root: +[Valid][25][UTT][ MSE = 0.1923 | LCC = 0.9630 | SRCC = 0.9629 ] +INFO:root:Average loss=0.043557348974049094 +INFO:root: +[Valid][26][UTT][ MSE = 0.1845 | LCC = 0.9650 | SRCC = 0.9652 ] +INFO:root:Average loss=0.04106943705677986 +INFO:root: +[Valid][27][UTT][ MSE = 0.2055 | LCC = 0.9651 | SRCC = 0.9648 ] +INFO:root:Average loss=0.04132601591944694 +INFO:root: +[Valid][28][UTT][ MSE = 0.1670 | LCC = 0.9635 | SRCC = 0.9635 ] +INFO:root:Average loss=0.04250642742216587 +INFO:root: +[Valid][29][UTT][ MSE = 0.1730 | LCC = 0.9650 | SRCC = 0.9647 ] +INFO:root:Average loss=0.03747012421488762 +INFO:root: +[Valid][30][UTT][ MSE = 0.1995 | LCC = 0.9639 | SRCC = 0.9641 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1851 | LCC = 0.9656 | SRCC = 0.9666 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run0/config.gin b/results_tencent/w2v2_large_xlsr53/layer2/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run0/model.pt b/results_tencent/w2v2_large_xlsr53/layer2/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b678a0bbec849e7ca9ea4a236f0cfcd93f01824c --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a2723de73694e201e32528dada0f3b41263cf39ff1da2372575bf46d7151c41c +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run0/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer2/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4678cffda2e6dfb0ded499108136000e77f2e65b --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:614bd6f27a60c1a8206ad6bf9e960bb552259c5a02eca5324400a98619501d99 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run0/train.log b/results_tencent/w2v2_large_xlsr53/layer2/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..09e3cd15835c52e9982ce35996450becf265f9fc --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44609852623939517 +INFO:root: +[Valid][0][UTT][ MSE = 0.3260 | LCC = 0.9080 | SRCC = 0.9085 ] +INFO:root:Average loss=0.23833154356479644 +INFO:root: +[Valid][1][UTT][ MSE = 0.2567 | LCC = 0.9270 | SRCC = 0.9287 ] +INFO:root:Average loss=0.21026425963640213 +INFO:root: +[Valid][2][UTT][ MSE = 0.2424 | LCC = 0.9374 | SRCC = 0.9379 ] +INFO:root:Average loss=0.17998067587614058 +INFO:root: +[Valid][3][UTT][ MSE = 0.2618 | LCC = 0.9443 | SRCC = 0.9454 ] +INFO:root:Average loss=0.15720118457078933 +INFO:root: +[Valid][4][UTT][ MSE = 0.2209 | LCC = 0.9456 | SRCC = 0.9462 ] +INFO:root:Average loss=0.1487709168791771 +INFO:root: +[Valid][5][UTT][ MSE = 0.2116 | LCC = 0.9491 | SRCC = 0.9504 ] +INFO:root:Average loss=0.13343152916431428 +INFO:root: +[Valid][6][UTT][ MSE = 0.2214 | LCC = 0.9510 | SRCC = 0.9525 ] +INFO:root:Average loss=0.13299033570289612 +INFO:root: +[Valid][7][UTT][ MSE = 0.2397 | LCC = 0.9525 | SRCC = 0.9528 ] +INFO:root:Average loss=0.12205217537283898 +INFO:root: +[Valid][8][UTT][ MSE = 0.2230 | LCC = 0.9510 | SRCC = 0.9543 ] +INFO:root:Average loss=0.11192790335416794 +INFO:root: +[Valid][9][UTT][ MSE = 0.1994 | LCC = 0.9533 | SRCC = 0.9539 ] +INFO:root:Average loss=0.10973361909389495 +INFO:root: +[Valid][10][UTT][ MSE = 0.2271 | LCC = 0.9501 | SRCC = 0.9519 ] +INFO:root:Average loss=0.10106519895792007 +INFO:root: +[Valid][11][UTT][ MSE = 0.2096 | LCC = 0.9584 | SRCC = 0.9592 ] +INFO:root:Average loss=0.09630563578009606 +INFO:root: +[Valid][12][UTT][ MSE = 0.1953 | LCC = 0.9568 | SRCC = 0.9572 ] +INFO:root:Average loss=0.09022744387388229 +INFO:root: +[Valid][13][UTT][ MSE = 0.2217 | LCC = 0.9576 | SRCC = 0.9577 ] +INFO:root:Average loss=0.08978195306658744 +INFO:root: +[Valid][14][UTT][ MSE = 0.2234 | LCC = 0.9596 | SRCC = 0.9598 ] +INFO:root:Average loss=0.08498137480020523 +INFO:root: +[Valid][15][UTT][ MSE = 0.2139 | LCC = 0.9609 | SRCC = 0.9616 ] +INFO:root:Average loss=0.0798245558142662 +INFO:root: +[Valid][16][UTT][ MSE = 0.1901 | LCC = 0.9607 | SRCC = 0.9610 ] +INFO:root:Average loss=0.07848974671959877 +INFO:root: +[Valid][17][UTT][ MSE = 0.2149 | LCC = 0.9617 | SRCC = 0.9615 ] +INFO:root:Average loss=0.07384794455766677 +INFO:root: +[Valid][18][UTT][ MSE = 0.2097 | LCC = 0.9617 | SRCC = 0.9613 ] +INFO:root:Average loss=0.07390478670597077 +INFO:root: +[Valid][19][UTT][ MSE = 0.1889 | LCC = 0.9610 | SRCC = 0.9610 ] +INFO:root:Average loss=0.06417443534731865 +INFO:root: +[Valid][20][UTT][ MSE = 0.1616 | LCC = 0.9613 | SRCC = 0.9623 ] +INFO:root:Average loss=0.07049100410938262 +INFO:root: +[Valid][21][UTT][ MSE = 0.2057 | LCC = 0.9610 | SRCC = 0.9601 ] +INFO:root:Average loss=0.06375178435444832 +INFO:root: +[Valid][22][UTT][ MSE = 0.1852 | LCC = 0.9628 | SRCC = 0.9630 ] +INFO:root:Average loss=0.06109462293982506 +INFO:root: +[Valid][23][UTT][ MSE = 0.2016 | LCC = 0.9632 | SRCC = 0.9628 ] +INFO:root:Average loss=0.05667851032316685 +INFO:root: +[Valid][24][UTT][ MSE = 0.1789 | LCC = 0.9630 | SRCC = 0.9622 ] +INFO:root:Average loss=0.05805547061562538 +INFO:root: +[Valid][25][UTT][ MSE = 0.2311 | LCC = 0.9613 | SRCC = 0.9628 ] +INFO:root:Average loss=0.053344332575798036 +INFO:root: +[Valid][26][UTT][ MSE = 0.1963 | LCC = 0.9603 | SRCC = 0.9612 ] +INFO:root:Average loss=0.05805450719594955 +INFO:root: +[Valid][27][UTT][ MSE = 0.1856 | LCC = 0.9624 | SRCC = 0.9623 ] +INFO:root:Average loss=0.05289876037836075 +INFO:root: +[Valid][28][UTT][ MSE = 0.1768 | LCC = 0.9641 | SRCC = 0.9636 ] +INFO:root:Average loss=0.05044573411345482 +INFO:root: +[Valid][29][UTT][ MSE = 0.1800 | LCC = 0.9608 | SRCC = 0.9613 ] +INFO:root:Average loss=0.04804440629482269 +INFO:root: +[Valid][30][UTT][ MSE = 0.1775 | LCC = 0.9634 | SRCC = 0.9629 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1885 | LCC = 0.9599 | SRCC = 0.9602 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run1/config.gin b/results_tencent/w2v2_large_xlsr53/layer2/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run1/model.pt b/results_tencent/w2v2_large_xlsr53/layer2/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c57835582d6e88e406599ca9ed611308f314ffc6 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e17e34e185b7f2f964b9ede3a4b727b448ad450c8f281d13fb2657a660064322 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run1/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer2/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..aa28dbef70ad28b9d9ab82e0f4e3a3f1ac838dc7 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6299c7535f37ea573cacb73e2366225bf33f46860b8ba07f926f4563ec2c5d41 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run1/train.log b/results_tencent/w2v2_large_xlsr53/layer2/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..711a9831ee2136351bc3391126508d9bd0f83d62 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5135888406038285 +INFO:root: +[Valid][0][UTT][ MSE = 0.3430 | LCC = 0.9075 | SRCC = 0.9077 ] +INFO:root:Average loss=0.25969696807861326 +INFO:root: +[Valid][1][UTT][ MSE = 0.3719 | LCC = 0.9239 | SRCC = 0.9247 ] +INFO:root:Average loss=0.22443855732679366 +INFO:root: +[Valid][2][UTT][ MSE = 0.2917 | LCC = 0.9329 | SRCC = 0.9340 ] +INFO:root:Average loss=0.17771517914533616 +INFO:root: +[Valid][3][UTT][ MSE = 0.3256 | LCC = 0.9405 | SRCC = 0.9426 ] +INFO:root:Average loss=0.15716864007711412 +INFO:root: +[Valid][4][UTT][ MSE = 0.3282 | LCC = 0.9456 | SRCC = 0.9490 ] +INFO:root:Average loss=0.14285346585512163 +INFO:root: +[Valid][5][UTT][ MSE = 0.2583 | LCC = 0.9449 | SRCC = 0.9453 ] +INFO:root:Average loss=0.14296855002641679 +INFO:root: +[Valid][6][UTT][ MSE = 0.2342 | LCC = 0.9508 | SRCC = 0.9512 ] +INFO:root:Average loss=0.12921862483024596 +INFO:root: +[Valid][7][UTT][ MSE = 0.2602 | LCC = 0.9477 | SRCC = 0.9489 ] +INFO:root:Average loss=0.12902436661720276 +INFO:root: +[Valid][8][UTT][ MSE = 0.2120 | LCC = 0.9503 | SRCC = 0.9520 ] +INFO:root:Average loss=0.11463973596692086 +INFO:root: +[Valid][9][UTT][ MSE = 0.2414 | LCC = 0.9562 | SRCC = 0.9572 ] +INFO:root:Average loss=0.1038921508193016 +INFO:root: +[Valid][10][UTT][ MSE = 0.1838 | LCC = 0.9556 | SRCC = 0.9570 ] +INFO:root:Average loss=0.10651188996434212 +INFO:root: +[Valid][11][UTT][ MSE = 0.2537 | LCC = 0.9543 | SRCC = 0.9553 ] +INFO:root:Average loss=0.09722890165448189 +INFO:root: +[Valid][12][UTT][ MSE = 0.2076 | LCC = 0.9583 | SRCC = 0.9587 ] +INFO:root:Average loss=0.09511298221349716 +INFO:root: +[Valid][13][UTT][ MSE = 0.2296 | LCC = 0.9595 | SRCC = 0.9593 ] +INFO:root:Average loss=0.0887458536028862 +INFO:root: +[Valid][14][UTT][ MSE = 0.1983 | LCC = 0.9599 | SRCC = 0.9606 ] +INFO:root:Average loss=0.08680028867721558 +INFO:root: +[Valid][15][UTT][ MSE = 0.1897 | LCC = 0.9584 | SRCC = 0.9585 ] +INFO:root:Average loss=0.0812461057305336 +INFO:root: +[Valid][16][UTT][ MSE = 0.2123 | LCC = 0.9616 | SRCC = 0.9625 ] +INFO:root:Average loss=0.07829947060346604 +INFO:root: +[Valid][17][UTT][ MSE = 0.1922 | LCC = 0.9617 | SRCC = 0.9616 ] +INFO:root:Average loss=0.07313104006648063 +INFO:root: +[Valid][18][UTT][ MSE = 0.2094 | LCC = 0.9621 | SRCC = 0.9619 ] +INFO:root:Average loss=0.06782735189795494 +INFO:root: +[Valid][19][UTT][ MSE = 0.2014 | LCC = 0.9615 | SRCC = 0.9618 ] +INFO:root:Average loss=0.06777151447534562 +INFO:root: +[Valid][20][UTT][ MSE = 0.1957 | LCC = 0.9596 | SRCC = 0.9596 ] +INFO:root:Average loss=0.06590259677171707 +INFO:root: +[Valid][21][UTT][ MSE = 0.2383 | LCC = 0.9618 | SRCC = 0.9619 ] +INFO:root:Average loss=0.06097280925512314 +INFO:root: +[Valid][22][UTT][ MSE = 0.1979 | LCC = 0.9619 | SRCC = 0.9623 ] +INFO:root:Average loss=0.061339205533266064 +INFO:root: +[Valid][23][UTT][ MSE = 0.2037 | LCC = 0.9619 | SRCC = 0.9618 ] +INFO:root:Average loss=0.0558354367017746 +INFO:root: +[Valid][24][UTT][ MSE = 0.1908 | LCC = 0.9639 | SRCC = 0.9641 ] +INFO:root:Average loss=0.05762379065155983 +INFO:root: +[Valid][25][UTT][ MSE = 0.2249 | LCC = 0.9628 | SRCC = 0.9637 ] +INFO:root:Average loss=0.052513997167348864 +INFO:root: +[Valid][26][UTT][ MSE = 0.2177 | LCC = 0.9615 | SRCC = 0.9627 ] +INFO:root:Average loss=0.05441676266491413 +INFO:root: +[Valid][27][UTT][ MSE = 0.1873 | LCC = 0.9632 | SRCC = 0.9634 ] +INFO:root:Average loss=0.05195492020249367 +INFO:root: +[Valid][28][UTT][ MSE = 0.2116 | LCC = 0.9641 | SRCC = 0.9643 ] +INFO:root:Average loss=0.051719442114233974 +INFO:root: +[Valid][29][UTT][ MSE = 0.1824 | LCC = 0.9631 | SRCC = 0.9631 ] +INFO:root:Average loss=0.050187780246138576 +INFO:root: +[Valid][30][UTT][ MSE = 0.2086 | LCC = 0.9639 | SRCC = 0.9638 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2262 | LCC = 0.9619 | SRCC = 0.9633 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run2/config.gin b/results_tencent/w2v2_large_xlsr53/layer2/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run2/model.pt b/results_tencent/w2v2_large_xlsr53/layer2/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6ea72f80e79121230d2bf76698b85b4c44eec2c3 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:925469188bbf9e0714c3ff7d9fe82e9ace579619392a2ebdb153896549da1441 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run2/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer2/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..11963effde0801b0451154da7570d51c6cbf1115 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d0297ad3a26b58f2b7f6eccce2e95cbae18b6fc66df5486c58578285fc6e9c48 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run2/train.log b/results_tencent/w2v2_large_xlsr53/layer2/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1ed6e832afe638ce15899c2c426747657b12014d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4593293360471725 +INFO:root: +[Valid][0][UTT][ MSE = 0.3232 | LCC = 0.9144 | SRCC = 0.9134 ] +INFO:root:Average loss=0.22953666627407074 +INFO:root: +[Valid][1][UTT][ MSE = 0.3330 | LCC = 0.9269 | SRCC = 0.9266 ] +INFO:root:Average loss=0.2074994914531708 +INFO:root: +[Valid][2][UTT][ MSE = 0.2789 | LCC = 0.9377 | SRCC = 0.9381 ] +INFO:root:Average loss=0.17953446865081787 +INFO:root: +[Valid][3][UTT][ MSE = 0.3791 | LCC = 0.9416 | SRCC = 0.9432 ] +INFO:root:Average loss=0.16041931068897247 +INFO:root: +[Valid][4][UTT][ MSE = 0.2974 | LCC = 0.9448 | SRCC = 0.9475 ] +INFO:root:Average loss=0.15406611746549606 +INFO:root: +[Valid][5][UTT][ MSE = 0.2462 | LCC = 0.9480 | SRCC = 0.9482 ] +INFO:root:Average loss=0.14066838431358336 +INFO:root: +[Valid][6][UTT][ MSE = 0.2770 | LCC = 0.9512 | SRCC = 0.9528 ] +INFO:root:Average loss=0.12794737291336059 +INFO:root: +[Valid][7][UTT][ MSE = 0.2423 | LCC = 0.9519 | SRCC = 0.9537 ] +INFO:root:Average loss=0.11947393375635147 +INFO:root: +[Valid][8][UTT][ MSE = 0.2223 | LCC = 0.9541 | SRCC = 0.9555 ] +INFO:root:Average loss=0.11527088862657547 +INFO:root: +[Valid][9][UTT][ MSE = 0.2159 | LCC = 0.9538 | SRCC = 0.9556 ] +INFO:root:Average loss=0.10850246328115463 +INFO:root: +[Valid][10][UTT][ MSE = 0.2328 | LCC = 0.9560 | SRCC = 0.9568 ] +INFO:root:Average loss=0.10342880773544312 +INFO:root: +[Valid][11][UTT][ MSE = 0.2128 | LCC = 0.9578 | SRCC = 0.9589 ] +INFO:root:Average loss=0.0989182787835598 +INFO:root: +[Valid][12][UTT][ MSE = 0.2173 | LCC = 0.9562 | SRCC = 0.9573 ] +INFO:root:Average loss=0.0888883726298809 +INFO:root: +[Valid][13][UTT][ MSE = 0.2535 | LCC = 0.9553 | SRCC = 0.9559 ] +INFO:root:Average loss=0.0904968886077404 +INFO:root: +[Valid][14][UTT][ MSE = 0.1907 | LCC = 0.9599 | SRCC = 0.9602 ] +INFO:root:Average loss=0.08288402035832405 +INFO:root: +[Valid][15][UTT][ MSE = 0.1906 | LCC = 0.9601 | SRCC = 0.9606 ] +INFO:root:Average loss=0.0797882657945156 +INFO:root: +[Valid][16][UTT][ MSE = 0.1929 | LCC = 0.9593 | SRCC = 0.9597 ] +INFO:root:Average loss=0.07705901610851287 +INFO:root: +[Valid][17][UTT][ MSE = 0.1794 | LCC = 0.9604 | SRCC = 0.9609 ] +INFO:root:Average loss=0.07277630457282067 +INFO:root: +[Valid][18][UTT][ MSE = 0.2081 | LCC = 0.9611 | SRCC = 0.9608 ] +INFO:root:Average loss=0.0695393815934658 +INFO:root: +[Valid][19][UTT][ MSE = 0.1760 | LCC = 0.9605 | SRCC = 0.9607 ] +INFO:root:Average loss=0.06585526669025421 +INFO:root: +[Valid][20][UTT][ MSE = 0.2074 | LCC = 0.9622 | SRCC = 0.9624 ] +INFO:root:Average loss=0.0671180671453476 +INFO:root: +[Valid][21][UTT][ MSE = 0.1793 | LCC = 0.9631 | SRCC = 0.9629 ] +INFO:root:Average loss=0.06553419545292855 +INFO:root: +[Valid][22][UTT][ MSE = 0.2267 | LCC = 0.9609 | SRCC = 0.9618 ] +INFO:root:Average loss=0.06389310121536255 +INFO:root: +[Valid][23][UTT][ MSE = 0.2631 | LCC = 0.9580 | SRCC = 0.9610 ] +INFO:root:Average loss=0.061622502207756044 +INFO:root: +[Valid][24][UTT][ MSE = 0.1877 | LCC = 0.9618 | SRCC = 0.9621 ] +INFO:root:Average loss=0.05600602307915688 +INFO:root: +[Valid][25][UTT][ MSE = 0.1998 | LCC = 0.9630 | SRCC = 0.9632 ] +INFO:root:Average loss=0.055193663716316226 +INFO:root: +[Valid][26][UTT][ MSE = 0.2286 | LCC = 0.9625 | SRCC = 0.9637 ] +INFO:root:Average loss=0.056247327893972396 +INFO:root: +[Valid][27][UTT][ MSE = 0.2002 | LCC = 0.9627 | SRCC = 0.9633 ] +INFO:root:Average loss=0.052352641493082046 +INFO:root: +[Valid][28][UTT][ MSE = 0.1857 | LCC = 0.9632 | SRCC = 0.9640 ] +INFO:root:Average loss=0.05514593896269798 +INFO:root: +[Valid][29][UTT][ MSE = 0.1634 | LCC = 0.9620 | SRCC = 0.9627 ] +INFO:root:Average loss=0.04923662711679935 +INFO:root: +[Valid][30][UTT][ MSE = 0.2149 | LCC = 0.9609 | SRCC = 0.9625 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2028 | LCC = 0.9598 | SRCC = 0.9611 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run3/config.gin b/results_tencent/w2v2_large_xlsr53/layer2/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run3/model.pt b/results_tencent/w2v2_large_xlsr53/layer2/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c4eea7a10f50d6719bfe5fb6d1dbe53820f4a20d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0227bbf62748575544809dec5988f794d79e115f3d0825e47c168b90b3f3e42e +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run3/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer2/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4ae16d2d0f3e088e9a74d3e786c3b98f2dad9e36 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:97243df1b674bf6084fec0c5fb70b4efe2195085bae4df6189aba308c90e8ad7 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run3/train.log b/results_tencent/w2v2_large_xlsr53/layer2/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ff24f9859ed9f7d741b0b6513348f8045e8d66ae --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4562833994626999 +INFO:root: +[Valid][0][UTT][ MSE = 0.3765 | LCC = 0.9160 | SRCC = 0.9162 ] +INFO:root:Average loss=0.23653838413953782 +INFO:root: +[Valid][1][UTT][ MSE = 0.3087 | LCC = 0.9320 | SRCC = 0.9321 ] +INFO:root:Average loss=0.21741507542133331 +INFO:root: +[Valid][2][UTT][ MSE = 0.3581 | LCC = 0.9221 | SRCC = 0.9201 ] +INFO:root:Average loss=0.19501768243312836 +INFO:root: +[Valid][3][UTT][ MSE = 0.2655 | LCC = 0.9425 | SRCC = 0.9429 ] +INFO:root:Average loss=0.1622797245979309 +INFO:root: +[Valid][4][UTT][ MSE = 0.2510 | LCC = 0.9473 | SRCC = 0.9480 ] +INFO:root:Average loss=0.1507149466276169 +INFO:root: +[Valid][5][UTT][ MSE = 0.2427 | LCC = 0.9506 | SRCC = 0.9512 ] +INFO:root:Average loss=0.1376643219590187 +INFO:root: +[Valid][6][UTT][ MSE = 0.2209 | LCC = 0.9521 | SRCC = 0.9531 ] +INFO:root:Average loss=0.12826774108409883 +INFO:root: +[Valid][7][UTT][ MSE = 0.2360 | LCC = 0.9533 | SRCC = 0.9540 ] +INFO:root:Average loss=0.1204687474668026 +INFO:root: +[Valid][8][UTT][ MSE = 0.2200 | LCC = 0.9549 | SRCC = 0.9556 ] +INFO:root:Average loss=0.11655159151554108 +INFO:root: +[Valid][9][UTT][ MSE = 0.2399 | LCC = 0.9538 | SRCC = 0.9554 ] +INFO:root:Average loss=0.10875740474462509 +INFO:root: +[Valid][10][UTT][ MSE = 0.2154 | LCC = 0.9545 | SRCC = 0.9546 ] +INFO:root:Average loss=0.10597641411423683 +INFO:root: +[Valid][11][UTT][ MSE = 0.2244 | LCC = 0.9579 | SRCC = 0.9583 ] +INFO:root:Average loss=0.09732343754172325 +INFO:root: +[Valid][12][UTT][ MSE = 0.2014 | LCC = 0.9582 | SRCC = 0.9590 ] +INFO:root:Average loss=0.09095149698853493 +INFO:root: +[Valid][13][UTT][ MSE = 0.2262 | LCC = 0.9580 | SRCC = 0.9593 ] +INFO:root:Average loss=0.08686013081669808 +INFO:root: +[Valid][14][UTT][ MSE = 0.2062 | LCC = 0.9597 | SRCC = 0.9599 ] +INFO:root:Average loss=0.08680619505047799 +INFO:root: +[Valid][15][UTT][ MSE = 0.1918 | LCC = 0.9594 | SRCC = 0.9597 ] +INFO:root:Average loss=0.08003055828809738 +INFO:root: +[Valid][16][UTT][ MSE = 0.2458 | LCC = 0.9600 | SRCC = 0.9604 ] +INFO:root:Average loss=0.07810095965862274 +INFO:root: +[Valid][17][UTT][ MSE = 0.2226 | LCC = 0.9610 | SRCC = 0.9609 ] +INFO:root:Average loss=0.07092028328776359 +INFO:root: +[Valid][18][UTT][ MSE = 0.1829 | LCC = 0.9610 | SRCC = 0.9608 ] +INFO:root:Average loss=0.06838592681288719 +INFO:root: +[Valid][19][UTT][ MSE = 0.2020 | LCC = 0.9588 | SRCC = 0.9585 ] +INFO:root:Average loss=0.06741627958416939 +INFO:root: +[Valid][20][UTT][ MSE = 0.2171 | LCC = 0.9610 | SRCC = 0.9605 ] +INFO:root:Average loss=0.06785262542963028 +INFO:root: +[Valid][21][UTT][ MSE = 0.1911 | LCC = 0.9609 | SRCC = 0.9607 ] +INFO:root:Average loss=0.0635423118174076 +INFO:root: +[Valid][22][UTT][ MSE = 0.2001 | LCC = 0.9638 | SRCC = 0.9638 ] +INFO:root:Average loss=0.0595630339384079 +INFO:root: +[Valid][23][UTT][ MSE = 0.1874 | LCC = 0.9607 | SRCC = 0.9613 ] +INFO:root:Average loss=0.06005968725681305 +INFO:root: +[Valid][24][UTT][ MSE = 0.1693 | LCC = 0.9623 | SRCC = 0.9633 ] +INFO:root:Average loss=0.05575656974315643 +INFO:root: +[Valid][25][UTT][ MSE = 0.1781 | LCC = 0.9631 | SRCC = 0.9631 ] +INFO:root:Average loss=0.05308436116576195 +INFO:root: +[Valid][26][UTT][ MSE = 0.1945 | LCC = 0.9631 | SRCC = 0.9633 ] +INFO:root:Average loss=0.052127346009016036 +INFO:root: +[Valid][27][UTT][ MSE = 0.1776 | LCC = 0.9627 | SRCC = 0.9628 ] +INFO:root:Average loss=0.05397003518044949 +INFO:root: +[Valid][28][UTT][ MSE = 0.1874 | LCC = 0.9630 | SRCC = 0.9624 ] +INFO:root:Average loss=0.05174126599729061 +INFO:root: +[Valid][29][UTT][ MSE = 0.1887 | LCC = 0.9640 | SRCC = 0.9639 ] +INFO:root:Average loss=0.04706080523133278 +INFO:root: +[Valid][30][UTT][ MSE = 0.1780 | LCC = 0.9629 | SRCC = 0.9625 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1997 | LCC = 0.9608 | SRCC = 0.9620 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run4/config.gin b/results_tencent/w2v2_large_xlsr53/layer2/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run4/model.pt b/results_tencent/w2v2_large_xlsr53/layer2/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e7ae742ec26bc3cadd038ee212d8afcc2ece9925 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f8261e71d2d66dbcb49f536a09fddc21a49a8fdd27596d670aa0c9d40045959d +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run4/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer2/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6d8ad6dc6b805bbf038ac0b5ea53f018ec2cf779 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bc646cf3da48f530fb987e89aba8dcf759d878ebcf37f1011cf46cca8e50e12b +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer2/run4/train.log b/results_tencent/w2v2_large_xlsr53/layer2/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..206f11e179aeb9e80e91125ddb1b9e336ac9a2de --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer2/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5030979062318802 +INFO:root: +[Valid][0][UTT][ MSE = 0.3832 | LCC = 0.9108 | SRCC = 0.9082 ] +INFO:root:Average loss=0.2396397397518158 +INFO:root: +[Valid][1][UTT][ MSE = 0.2802 | LCC = 0.9280 | SRCC = 0.9284 ] +INFO:root:Average loss=0.2094955629706383 +INFO:root: +[Valid][2][UTT][ MSE = 0.3050 | LCC = 0.9385 | SRCC = 0.9386 ] +INFO:root:Average loss=0.18132342898845671 +INFO:root: +[Valid][3][UTT][ MSE = 0.3029 | LCC = 0.9404 | SRCC = 0.9415 ] +INFO:root:Average loss=0.1636745172739029 +INFO:root: +[Valid][4][UTT][ MSE = 0.2512 | LCC = 0.9445 | SRCC = 0.9468 ] +INFO:root:Average loss=0.15702207744121552 +INFO:root: +[Valid][5][UTT][ MSE = 0.2611 | LCC = 0.9476 | SRCC = 0.9488 ] +INFO:root:Average loss=0.13773655050992967 +INFO:root: +[Valid][6][UTT][ MSE = 0.2244 | LCC = 0.9491 | SRCC = 0.9497 ] +INFO:root:Average loss=0.1319179099202156 +INFO:root: +[Valid][7][UTT][ MSE = 0.1966 | LCC = 0.9511 | SRCC = 0.9531 ] +INFO:root:Average loss=0.12401845958828926 +INFO:root: +[Valid][8][UTT][ MSE = 0.2292 | LCC = 0.9541 | SRCC = 0.9539 ] +INFO:root:Average loss=0.11277573436498642 +INFO:root: +[Valid][9][UTT][ MSE = 0.2123 | LCC = 0.9551 | SRCC = 0.9556 ] +INFO:root:Average loss=0.1059305281341076 +INFO:root: +[Valid][10][UTT][ MSE = 0.2141 | LCC = 0.9553 | SRCC = 0.9553 ] +INFO:root:Average loss=0.1053193842768669 +INFO:root: +[Valid][11][UTT][ MSE = 0.2214 | LCC = 0.9520 | SRCC = 0.9530 ] +INFO:root:Average loss=0.10542487251758575 +INFO:root: +[Valid][12][UTT][ MSE = 0.1949 | LCC = 0.9540 | SRCC = 0.9561 ] +INFO:root:Average loss=0.09630312025547028 +INFO:root: +[Valid][13][UTT][ MSE = 0.2269 | LCC = 0.9576 | SRCC = 0.9591 ] +INFO:root:Average loss=0.08623543101549148 +INFO:root: +[Valid][14][UTT][ MSE = 0.2167 | LCC = 0.9594 | SRCC = 0.9588 ] +INFO:root:Average loss=0.08368380719423293 +INFO:root: +[Valid][15][UTT][ MSE = 0.1808 | LCC = 0.9589 | SRCC = 0.9591 ] +INFO:root:Average loss=0.08123496571183204 +INFO:root: +[Valid][16][UTT][ MSE = 0.1999 | LCC = 0.9600 | SRCC = 0.9609 ] +INFO:root:Average loss=0.07620237791538238 +INFO:root: +[Valid][17][UTT][ MSE = 0.1926 | LCC = 0.9591 | SRCC = 0.9597 ] +INFO:root:Average loss=0.07514574551582337 +INFO:root: +[Valid][18][UTT][ MSE = 0.2023 | LCC = 0.9613 | SRCC = 0.9614 ] +INFO:root:Average loss=0.07156422448158264 +INFO:root: +[Valid][19][UTT][ MSE = 0.2422 | LCC = 0.9601 | SRCC = 0.9595 ] +INFO:root:Average loss=0.06987842938303948 +INFO:root: +[Valid][20][UTT][ MSE = 0.1942 | LCC = 0.9615 | SRCC = 0.9623 ] +INFO:root:Average loss=0.06659885033965111 +INFO:root: +[Valid][21][UTT][ MSE = 0.1884 | LCC = 0.9623 | SRCC = 0.9625 ] +INFO:root:Average loss=0.0583419938236475 +INFO:root: +[Valid][22][UTT][ MSE = 0.1952 | LCC = 0.9618 | SRCC = 0.9612 ] +INFO:root:Average loss=0.06257145553827285 +INFO:root: +[Valid][23][UTT][ MSE = 0.2033 | LCC = 0.9579 | SRCC = 0.9608 ] +INFO:root:Average loss=0.06693692022562027 +INFO:root: +[Valid][24][UTT][ MSE = 0.1996 | LCC = 0.9624 | SRCC = 0.9617 ] +INFO:root:Average loss=0.05777599981427193 +INFO:root: +[Valid][25][UTT][ MSE = 0.2231 | LCC = 0.9620 | SRCC = 0.9612 ] +INFO:root:Average loss=0.0551169321835041 +INFO:root: +[Valid][26][UTT][ MSE = 0.1878 | LCC = 0.9628 | SRCC = 0.9624 ] +INFO:root:Average loss=0.054375190436840054 +INFO:root: +[Valid][27][UTT][ MSE = 0.2167 | LCC = 0.9622 | SRCC = 0.9622 ] +INFO:root:Average loss=0.05381190201640129 +INFO:root: +[Valid][28][UTT][ MSE = 0.1870 | LCC = 0.9634 | SRCC = 0.9635 ] +INFO:root:Average loss=0.05058666160702705 +INFO:root: +[Valid][29][UTT][ MSE = 0.1658 | LCC = 0.9627 | SRCC = 0.9622 ] +INFO:root:Average loss=0.04815906725823879 +INFO:root: +[Valid][30][UTT][ MSE = 0.2048 | LCC = 0.9638 | SRCC = 0.9636 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2110 | LCC = 0.9610 | SRCC = 0.9618 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run0/config.gin b/results_tencent/w2v2_large_xlsr53/layer20/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run0/model.pt b/results_tencent/w2v2_large_xlsr53/layer20/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..af88e08cd6b00e7d26dc6e9289b7884fe5998624 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:621d943ccd39e1ec753b82cfdd6e7007cecd3bd705dfa36a720e1e8d24ab9276 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run0/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer20/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e357f589e15653e71bd0aa624cd22872b41116f4 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:179e65c6c4acd4d2d406338d73aacff44b1bfef575f4494615e6c2c72d09a63b +size 3404998 diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run0/train.log b/results_tencent/w2v2_large_xlsr53/layer20/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5f753bcd050e117f9da2ed9be60ddd3f65db6508 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4629988824129105 +INFO:root: +[Valid][0][UTT][ MSE = 0.3114 | LCC = 0.9246 | SRCC = 0.9251 ] +INFO:root:Average loss=0.22366431391239167 +INFO:root: +[Valid][1][UTT][ MSE = 0.3548 | LCC = 0.9355 | SRCC = 0.9360 ] +INFO:root:Average loss=0.19278989297151566 +INFO:root: +[Valid][2][UTT][ MSE = 0.2481 | LCC = 0.9439 | SRCC = 0.9447 ] +INFO:root:Average loss=0.15878735518455506 +INFO:root: +[Valid][3][UTT][ MSE = 0.2127 | LCC = 0.9489 | SRCC = 0.9500 ] +INFO:root:Average loss=0.14561033445596694 +INFO:root: +[Valid][4][UTT][ MSE = 0.2575 | LCC = 0.9464 | SRCC = 0.9461 ] +INFO:root:Average loss=0.13305880004167556 +INFO:root: +[Valid][5][UTT][ MSE = 0.2155 | LCC = 0.9535 | SRCC = 0.9549 ] +INFO:root:Average loss=0.12477695709466934 +INFO:root: +[Valid][6][UTT][ MSE = 0.2228 | LCC = 0.9503 | SRCC = 0.9502 ] +INFO:root:Average loss=0.12468982422351838 +INFO:root: +[Valid][7][UTT][ MSE = 0.2121 | LCC = 0.9544 | SRCC = 0.9570 ] +INFO:root:Average loss=0.11153936570882797 +INFO:root: +[Valid][8][UTT][ MSE = 0.2146 | LCC = 0.9575 | SRCC = 0.9587 ] +INFO:root:Average loss=0.10729372012615204 +INFO:root: +[Valid][9][UTT][ MSE = 0.2479 | LCC = 0.9544 | SRCC = 0.9540 ] +INFO:root:Average loss=0.10579327291250229 +INFO:root: +[Valid][10][UTT][ MSE = 0.2627 | LCC = 0.9588 | SRCC = 0.9592 ] +INFO:root:Average loss=0.09813617685437202 +INFO:root: +[Valid][11][UTT][ MSE = 0.2157 | LCC = 0.9579 | SRCC = 0.9585 ] +INFO:root:Average loss=0.090685691177845 +INFO:root: +[Valid][12][UTT][ MSE = 0.2126 | LCC = 0.9557 | SRCC = 0.9576 ] +INFO:root:Average loss=0.09299306949973106 +INFO:root: +[Valid][13][UTT][ MSE = 0.2386 | LCC = 0.9611 | SRCC = 0.9617 ] +INFO:root:Average loss=0.08810516437888145 +INFO:root: +[Valid][14][UTT][ MSE = 0.1981 | LCC = 0.9588 | SRCC = 0.9598 ] +INFO:root:Average loss=0.08572147470712661 +INFO:root: +[Valid][15][UTT][ MSE = 0.2241 | LCC = 0.9617 | SRCC = 0.9618 ] +INFO:root:Average loss=0.07694192951917649 +INFO:root: +[Valid][16][UTT][ MSE = 0.2188 | LCC = 0.9594 | SRCC = 0.9595 ] +INFO:root:Average loss=0.07698564451932907 +INFO:root: +[Valid][17][UTT][ MSE = 0.1936 | LCC = 0.9574 | SRCC = 0.9586 ] +INFO:root:Average loss=0.07248524785041809 +INFO:root: +[Valid][18][UTT][ MSE = 0.2162 | LCC = 0.9621 | SRCC = 0.9618 ] +INFO:root:Average loss=0.06788701045513153 +INFO:root: +[Valid][19][UTT][ MSE = 0.1977 | LCC = 0.9615 | SRCC = 0.9616 ] +INFO:root:Average loss=0.06689307793974876 +INFO:root: +[Valid][20][UTT][ MSE = 0.2061 | LCC = 0.9556 | SRCC = 0.9589 ] +INFO:root:Average loss=0.06886657682061195 +INFO:root: +[Valid][21][UTT][ MSE = 0.1908 | LCC = 0.9540 | SRCC = 0.9567 ] +INFO:root:Average loss=0.06416540971398353 +INFO:root: +[Valid][22][UTT][ MSE = 0.1858 | LCC = 0.9598 | SRCC = 0.9610 ] +INFO:root:Average loss=0.061709255635738376 +INFO:root: +[Valid][23][UTT][ MSE = 0.2174 | LCC = 0.9612 | SRCC = 0.9608 ] +INFO:root:Average loss=0.06306987607479095 +INFO:root: +[Valid][24][UTT][ MSE = 0.1899 | LCC = 0.9632 | SRCC = 0.9627 ] +INFO:root:Average loss=0.05682410390675068 +INFO:root: +[Valid][25][UTT][ MSE = 0.2119 | LCC = 0.9612 | SRCC = 0.9610 ] +INFO:root:Average loss=0.05969626981019974 +INFO:root: +[Valid][26][UTT][ MSE = 0.2175 | LCC = 0.9617 | SRCC = 0.9611 ] +INFO:root:Average loss=0.05452480435371399 +INFO:root: +[Valid][27][UTT][ MSE = 0.2107 | LCC = 0.9605 | SRCC = 0.9600 ] +INFO:root:Average loss=0.05442394286394119 +INFO:root: +[Valid][28][UTT][ MSE = 0.2022 | LCC = 0.9624 | SRCC = 0.9623 ] +INFO:root:Average loss=0.05186582286655903 +INFO:root: +[Valid][29][UTT][ MSE = 0.2186 | LCC = 0.9617 | SRCC = 0.9612 ] +INFO:root:Average loss=0.04794477890431881 +INFO:root: +[Valid][30][UTT][ MSE = 0.1908 | LCC = 0.9578 | SRCC = 0.9581 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1907 | LCC = 0.9634 | SRCC = 0.9647 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run1/config.gin b/results_tencent/w2v2_large_xlsr53/layer20/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run1/model.pt b/results_tencent/w2v2_large_xlsr53/layer20/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b530e7c507c835bf8ab0974f3adb4225e9b4de30 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c93129992d216c08cf9142b35290d8ff07d37dba8e9dd68add4a3d0d947fc83 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run1/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer20/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d47e39345900df5e2eee80f1fcaa5122c89d5860 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f28ec17a9eb788068b4f85b097fefe779324f30d89b4a62bb3b9ebe91ac2275a +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run1/train.log b/results_tencent/w2v2_large_xlsr53/layer20/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..80cafbeef211f377102ab8a2156b91a73010c964 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5047453473806381 +INFO:root: +[Valid][0][UTT][ MSE = 0.3716 | LCC = 0.9164 | SRCC = 0.9183 ] +INFO:root:Average loss=0.2357911098599434 +INFO:root: +[Valid][1][UTT][ MSE = 0.2930 | LCC = 0.9364 | SRCC = 0.9358 ] +INFO:root:Average loss=0.19463507199287414 +INFO:root: +[Valid][2][UTT][ MSE = 0.3219 | LCC = 0.9411 | SRCC = 0.9437 ] +INFO:root:Average loss=0.16942326152324677 +INFO:root: +[Valid][3][UTT][ MSE = 0.2422 | LCC = 0.9478 | SRCC = 0.9494 ] +INFO:root:Average loss=0.1432472774386406 +INFO:root: +[Valid][4][UTT][ MSE = 0.2520 | LCC = 0.9501 | SRCC = 0.9504 ] +INFO:root:Average loss=0.1378837634921074 +INFO:root: +[Valid][5][UTT][ MSE = 0.2763 | LCC = 0.9529 | SRCC = 0.9538 ] +INFO:root:Average loss=0.13014211532473563 +INFO:root: +[Valid][6][UTT][ MSE = 0.2128 | LCC = 0.9568 | SRCC = 0.9575 ] +INFO:root:Average loss=0.1190508416891098 +INFO:root: +[Valid][7][UTT][ MSE = 0.2656 | LCC = 0.9524 | SRCC = 0.9563 ] +INFO:root:Average loss=0.11025733277201652 +INFO:root: +[Valid][8][UTT][ MSE = 0.2008 | LCC = 0.9565 | SRCC = 0.9569 ] +INFO:root:Average loss=0.10288125112652778 +INFO:root: +[Valid][9][UTT][ MSE = 0.2501 | LCC = 0.9565 | SRCC = 0.9589 ] +INFO:root:Average loss=0.09826641339063645 +INFO:root: +[Valid][10][UTT][ MSE = 0.2206 | LCC = 0.9596 | SRCC = 0.9599 ] +INFO:root:Average loss=0.09441669952869415 +INFO:root: +[Valid][11][UTT][ MSE = 0.2063 | LCC = 0.9598 | SRCC = 0.9603 ] +INFO:root:Average loss=0.09596970373392105 +INFO:root: +[Valid][12][UTT][ MSE = 0.2208 | LCC = 0.9606 | SRCC = 0.9611 ] +INFO:root:Average loss=0.09524157053232193 +INFO:root: +[Valid][13][UTT][ MSE = 0.2033 | LCC = 0.9600 | SRCC = 0.9602 ] +INFO:root:Average loss=0.08601947990059852 +INFO:root: +[Valid][14][UTT][ MSE = 0.2085 | LCC = 0.9610 | SRCC = 0.9618 ] +INFO:root:Average loss=0.08231701856851578 +INFO:root: +[Valid][15][UTT][ MSE = 0.2437 | LCC = 0.9610 | SRCC = 0.9614 ] +INFO:root:Average loss=0.07642404335737228 +INFO:root: +[Valid][16][UTT][ MSE = 0.2360 | LCC = 0.9604 | SRCC = 0.9599 ] +INFO:root:Average loss=0.07551264095306397 +INFO:root: +[Valid][17][UTT][ MSE = 0.2173 | LCC = 0.9606 | SRCC = 0.9610 ] +INFO:root:Average loss=0.07608897471427918 +INFO:root: +[Valid][18][UTT][ MSE = 0.2202 | LCC = 0.9613 | SRCC = 0.9613 ] +INFO:root:Average loss=0.06993224635720253 +INFO:root: +[Valid][19][UTT][ MSE = 0.2172 | LCC = 0.9624 | SRCC = 0.9623 ] +INFO:root:Average loss=0.07147807642817497 +INFO:root: +[Valid][20][UTT][ MSE = 0.1811 | LCC = 0.9622 | SRCC = 0.9622 ] +INFO:root:Average loss=0.0643931780308485 +INFO:root: +[Valid][21][UTT][ MSE = 0.1796 | LCC = 0.9602 | SRCC = 0.9606 ] +INFO:root:Average loss=0.06566146942973138 +INFO:root: +[Valid][22][UTT][ MSE = 0.1933 | LCC = 0.9624 | SRCC = 0.9627 ] +INFO:root:Average loss=0.06351164630055428 +INFO:root: +[Valid][23][UTT][ MSE = 0.2116 | LCC = 0.9629 | SRCC = 0.9630 ] +INFO:root:Average loss=0.059989070266485214 +INFO:root: +[Valid][24][UTT][ MSE = 0.1940 | LCC = 0.9588 | SRCC = 0.9601 ] +INFO:root:Average loss=0.06304998752474784 +INFO:root: +[Valid][25][UTT][ MSE = 0.1971 | LCC = 0.9611 | SRCC = 0.9625 ] +INFO:root:Average loss=0.05865181577205658 +INFO:root: +[Valid][26][UTT][ MSE = 0.2393 | LCC = 0.9611 | SRCC = 0.9628 ] +INFO:root:Average loss=0.05488476550579071 +INFO:root: +[Valid][27][UTT][ MSE = 0.1832 | LCC = 0.9630 | SRCC = 0.9628 ] +INFO:root:Average loss=0.053384332954883575 +INFO:root: +[Valid][28][UTT][ MSE = 0.2317 | LCC = 0.9623 | SRCC = 0.9622 ] +INFO:root:Average loss=0.05190641908347607 +INFO:root: +[Valid][29][UTT][ MSE = 0.2413 | LCC = 0.9607 | SRCC = 0.9624 ] +INFO:root:Average loss=0.051104773938655854 +INFO:root: +[Valid][30][UTT][ MSE = 0.1640 | LCC = 0.9616 | SRCC = 0.9627 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1865 | LCC = 0.9623 | SRCC = 0.9637 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run2/config.gin b/results_tencent/w2v2_large_xlsr53/layer20/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run2/model.pt b/results_tencent/w2v2_large_xlsr53/layer20/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4f63b3737bc606a41075ee594401e611db962b50 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:70b67097e1607c2c4e1688746a5122bdd0dc08b996cdfb319c434d46b5676701 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run2/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer20/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..885904d1a930a3b2c0b6bb403b3c26907f061d53 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e828430b7449334e3d16c0e3fa8e1b3ee158765e636cc37230e7ba45ebcb9399 +size 3404998 diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run2/train.log b/results_tencent/w2v2_large_xlsr53/layer20/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..bdbaeefdec03e17a93b188cd729db3d0ca2f948a --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4592245291471481 +INFO:root: +[Valid][0][UTT][ MSE = 0.3561 | LCC = 0.9202 | SRCC = 0.9194 ] +INFO:root:Average loss=0.22856857031583785 +INFO:root: +[Valid][1][UTT][ MSE = 0.2843 | LCC = 0.9351 | SRCC = 0.9348 ] +INFO:root:Average loss=0.20029717165231706 +INFO:root: +[Valid][2][UTT][ MSE = 0.2695 | LCC = 0.9437 | SRCC = 0.9445 ] +INFO:root:Average loss=0.16392665046453475 +INFO:root: +[Valid][3][UTT][ MSE = 0.2832 | LCC = 0.9466 | SRCC = 0.9490 ] +INFO:root:Average loss=0.15001720958948137 +INFO:root: +[Valid][4][UTT][ MSE = 0.2582 | LCC = 0.9513 | SRCC = 0.9523 ] +INFO:root:Average loss=0.1459993635416031 +INFO:root: +[Valid][5][UTT][ MSE = 0.2393 | LCC = 0.9546 | SRCC = 0.9542 ] +INFO:root:Average loss=0.13080942156910896 +INFO:root: +[Valid][6][UTT][ MSE = 0.2768 | LCC = 0.9527 | SRCC = 0.9549 ] +INFO:root:Average loss=0.1268283115029335 +INFO:root: +[Valid][7][UTT][ MSE = 0.2351 | LCC = 0.9572 | SRCC = 0.9580 ] +INFO:root:Average loss=0.11774727308750152 +INFO:root: +[Valid][8][UTT][ MSE = 0.2136 | LCC = 0.9573 | SRCC = 0.9587 ] +INFO:root:Average loss=0.10802944266796112 +INFO:root: +[Valid][9][UTT][ MSE = 0.2762 | LCC = 0.9584 | SRCC = 0.9589 ] +INFO:root:Average loss=0.10997072920203209 +INFO:root: +[Valid][10][UTT][ MSE = 0.1827 | LCC = 0.9592 | SRCC = 0.9606 ] +INFO:root:Average loss=0.10900064092874527 +INFO:root: +[Valid][11][UTT][ MSE = 0.2810 | LCC = 0.9563 | SRCC = 0.9601 ] +INFO:root:Average loss=0.0978016403913498 +INFO:root: +[Valid][12][UTT][ MSE = 0.2270 | LCC = 0.9611 | SRCC = 0.9610 ] +INFO:root:Average loss=0.0922300574183464 +INFO:root: +[Valid][13][UTT][ MSE = 0.2232 | LCC = 0.9562 | SRCC = 0.9578 ] +INFO:root:Average loss=0.09493628245592117 +INFO:root: +[Valid][14][UTT][ MSE = 0.2597 | LCC = 0.9613 | SRCC = 0.9613 ] +INFO:root:Average loss=0.08921448343992233 +INFO:root: +[Valid][15][UTT][ MSE = 0.2772 | LCC = 0.9611 | SRCC = 0.9612 ] +INFO:root:Average loss=0.08235558739304542 +INFO:root: +[Valid][16][UTT][ MSE = 0.2024 | LCC = 0.9608 | SRCC = 0.9607 ] +INFO:root:Average loss=0.07879251325130462 +INFO:root: +[Valid][17][UTT][ MSE = 0.2005 | LCC = 0.9618 | SRCC = 0.9623 ] +INFO:root:Average loss=0.07574979481101037 +INFO:root: +[Valid][18][UTT][ MSE = 0.2099 | LCC = 0.9618 | SRCC = 0.9625 ] +INFO:root:Average loss=0.07522772544622422 +INFO:root: +[Valid][19][UTT][ MSE = 0.2077 | LCC = 0.9613 | SRCC = 0.9621 ] +INFO:root:Average loss=0.0720227519273758 +INFO:root: +[Valid][20][UTT][ MSE = 0.1718 | LCC = 0.9630 | SRCC = 0.9631 ] +INFO:root:Average loss=0.0707611525952816 +INFO:root: +[Valid][21][UTT][ MSE = 0.2186 | LCC = 0.9619 | SRCC = 0.9613 ] +INFO:root:Average loss=0.06334253358840942 +INFO:root: +[Valid][22][UTT][ MSE = 0.2099 | LCC = 0.9623 | SRCC = 0.9620 ] +INFO:root:Average loss=0.0647425666898489 +INFO:root: +[Valid][23][UTT][ MSE = 0.2105 | LCC = 0.9609 | SRCC = 0.9606 ] +INFO:root:Average loss=0.06317376679182053 +INFO:root: +[Valid][24][UTT][ MSE = 0.1923 | LCC = 0.9628 | SRCC = 0.9630 ] +INFO:root:Average loss=0.06325192946195603 +INFO:root: +[Valid][25][UTT][ MSE = 0.2062 | LCC = 0.9622 | SRCC = 0.9618 ] +INFO:root:Average loss=0.06664031228423119 +INFO:root: +[Valid][26][UTT][ MSE = 0.2124 | LCC = 0.9643 | SRCC = 0.9637 ] +INFO:root:Average loss=0.056216356590390204 +INFO:root: +[Valid][27][UTT][ MSE = 0.1908 | LCC = 0.9609 | SRCC = 0.9611 ] +INFO:root:Average loss=0.05502391457557678 +INFO:root: +[Valid][28][UTT][ MSE = 0.1750 | LCC = 0.9631 | SRCC = 0.9629 ] +INFO:root:Average loss=0.0563949693441391 +INFO:root: +[Valid][29][UTT][ MSE = 0.2247 | LCC = 0.9645 | SRCC = 0.9637 ] +INFO:root:Average loss=0.051352586016058924 +INFO:root: +[Valid][30][UTT][ MSE = 0.2128 | LCC = 0.9622 | SRCC = 0.9610 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2230 | LCC = 0.9650 | SRCC = 0.9665 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run3/config.gin b/results_tencent/w2v2_large_xlsr53/layer20/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run3/model.pt b/results_tencent/w2v2_large_xlsr53/layer20/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c31f7463fefb06be160d71386da39eddd2e2cf32 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:26247bd58ed312aa0ac75d4f014b9244c3d24e5034969436b019c30df286bc4a +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run3/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer20/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d2f83ba6f0573e545b466f1de4f91397358e457c --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4c835eb02a6deb424d9a428b9b3c86a5997182c3656c5296e0821b1ef0e0b747 +size 3404998 diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run3/train.log b/results_tencent/w2v2_large_xlsr53/layer20/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..11a954ee6f1d6ee099f17f56469814357b7b893f --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4887297610044479 +INFO:root: +[Valid][0][UTT][ MSE = 0.3432 | LCC = 0.9180 | SRCC = 0.9194 ] +INFO:root:Average loss=0.22236785912513732 +INFO:root: +[Valid][1][UTT][ MSE = 0.2943 | LCC = 0.9363 | SRCC = 0.9380 ] +INFO:root:Average loss=0.20272053211927413 +INFO:root: +[Valid][2][UTT][ MSE = 0.2514 | LCC = 0.9415 | SRCC = 0.9433 ] +INFO:root:Average loss=0.20457129311561584 +INFO:root: +[Valid][3][UTT][ MSE = 0.2633 | LCC = 0.9469 | SRCC = 0.9465 ] +INFO:root:Average loss=0.1530159558057785 +INFO:root: +[Valid][4][UTT][ MSE = 0.2620 | LCC = 0.9499 | SRCC = 0.9508 ] +INFO:root:Average loss=0.1378568348288536 +INFO:root: +[Valid][5][UTT][ MSE = 0.2345 | LCC = 0.9498 | SRCC = 0.9515 ] +INFO:root:Average loss=0.13470689183473586 +INFO:root: +[Valid][6][UTT][ MSE = 0.2117 | LCC = 0.9534 | SRCC = 0.9551 ] +INFO:root:Average loss=0.11932743054628372 +INFO:root: +[Valid][7][UTT][ MSE = 0.2492 | LCC = 0.9552 | SRCC = 0.9566 ] +INFO:root:Average loss=0.12073795026540757 +INFO:root: +[Valid][8][UTT][ MSE = 0.2471 | LCC = 0.9581 | SRCC = 0.9581 ] +INFO:root:Average loss=0.11250528278946877 +INFO:root: +[Valid][9][UTT][ MSE = 0.2214 | LCC = 0.9544 | SRCC = 0.9550 ] +INFO:root:Average loss=0.10689688348770142 +INFO:root: +[Valid][10][UTT][ MSE = 0.2421 | LCC = 0.9604 | SRCC = 0.9604 ] +INFO:root:Average loss=0.10254013946652413 +INFO:root: +[Valid][11][UTT][ MSE = 0.1915 | LCC = 0.9580 | SRCC = 0.9587 ] +INFO:root:Average loss=0.09334642338752747 +INFO:root: +[Valid][12][UTT][ MSE = 0.2635 | LCC = 0.9570 | SRCC = 0.9570 ] +INFO:root:Average loss=0.09107991543412208 +INFO:root: +[Valid][13][UTT][ MSE = 0.2095 | LCC = 0.9589 | SRCC = 0.9590 ] +INFO:root:Average loss=0.08886212658882141 +INFO:root: +[Valid][14][UTT][ MSE = 0.2418 | LCC = 0.9565 | SRCC = 0.9563 ] +INFO:root:Average loss=0.09155965119600296 +INFO:root: +[Valid][15][UTT][ MSE = 0.2231 | LCC = 0.9587 | SRCC = 0.9598 ] +INFO:root:Average loss=0.08190883490443229 +INFO:root: +[Valid][16][UTT][ MSE = 0.2182 | LCC = 0.9596 | SRCC = 0.9606 ] +INFO:root:Average loss=0.07684870931506158 +INFO:root: +[Valid][17][UTT][ MSE = 0.2278 | LCC = 0.9597 | SRCC = 0.9608 ] +INFO:root:Average loss=0.07955517387390136 +INFO:root: +[Valid][18][UTT][ MSE = 0.2001 | LCC = 0.9606 | SRCC = 0.9607 ] +INFO:root:Average loss=0.07094107431173324 +INFO:root: +[Valid][19][UTT][ MSE = 0.2103 | LCC = 0.9622 | SRCC = 0.9624 ] +INFO:root:Average loss=0.07292112249135971 +INFO:root: +[Valid][20][UTT][ MSE = 0.1818 | LCC = 0.9617 | SRCC = 0.9616 ] +INFO:root:Average loss=0.07269875276088715 +INFO:root: +[Valid][21][UTT][ MSE = 0.2030 | LCC = 0.9590 | SRCC = 0.9590 ] +INFO:root:Average loss=0.06788848900794983 +INFO:root: +[Valid][22][UTT][ MSE = 0.2188 | LCC = 0.9604 | SRCC = 0.9619 ] +INFO:root:Average loss=0.06553642919659615 +INFO:root: +[Valid][23][UTT][ MSE = 0.1834 | LCC = 0.9619 | SRCC = 0.9617 ] +INFO:root:Average loss=0.0652249065041542 +INFO:root: +[Valid][24][UTT][ MSE = 0.2192 | LCC = 0.9610 | SRCC = 0.9620 ] +INFO:root:Average loss=0.06108393374085426 +INFO:root: +[Valid][25][UTT][ MSE = 0.2049 | LCC = 0.9606 | SRCC = 0.9613 ] +INFO:root:Average loss=0.06202440446615219 +INFO:root: +[Valid][26][UTT][ MSE = 0.2601 | LCC = 0.9514 | SRCC = 0.9517 ] +INFO:root:Average loss=0.0573408827483654 +INFO:root: +[Valid][27][UTT][ MSE = 0.2103 | LCC = 0.9628 | SRCC = 0.9626 ] +INFO:root:Average loss=0.055590807914733885 +INFO:root: +[Valid][28][UTT][ MSE = 0.2169 | LCC = 0.9613 | SRCC = 0.9613 ] +INFO:root:Average loss=0.05295767097175121 +INFO:root: +[Valid][29][UTT][ MSE = 0.2154 | LCC = 0.9618 | SRCC = 0.9618 ] +INFO:root:Average loss=0.05214070227742195 +INFO:root: +[Valid][30][UTT][ MSE = 0.2018 | LCC = 0.9617 | SRCC = 0.9617 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2082 | LCC = 0.9639 | SRCC = 0.9652 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run4/config.gin b/results_tencent/w2v2_large_xlsr53/layer20/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run4/model.pt b/results_tencent/w2v2_large_xlsr53/layer20/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1c822e20ef0e4615c6bc6af89f83879a4282f8e1 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a563b576e176571651eb114a94bd1866ce2d5b1927a9640959e4ca82367ac9c6 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run4/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer20/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e05819880c8905e000e1c28a51077e9fb724122e --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:64519c042a8f307fdb4189b358720ef9ddbb462d672662637981cacfd88ef1dc +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer20/run4/train.log b/results_tencent/w2v2_large_xlsr53/layer20/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a54a652d4ab62b9bcf50c226410ae61d041eed2c --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer20/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4438913307189941 +INFO:root: +[Valid][0][UTT][ MSE = 0.3031 | LCC = 0.9164 | SRCC = 0.9203 ] +INFO:root:Average loss=0.22037116223573686 +INFO:root: +[Valid][1][UTT][ MSE = 0.2380 | LCC = 0.9366 | SRCC = 0.9376 ] +INFO:root:Average loss=0.20937941604852675 +INFO:root: +[Valid][2][UTT][ MSE = 0.2327 | LCC = 0.9420 | SRCC = 0.9420 ] +INFO:root:Average loss=0.1756558896303177 +INFO:root: +[Valid][3][UTT][ MSE = 0.2516 | LCC = 0.9472 | SRCC = 0.9483 ] +INFO:root:Average loss=0.16959647619724275 +INFO:root: +[Valid][4][UTT][ MSE = 0.2567 | LCC = 0.9510 | SRCC = 0.9515 ] +INFO:root:Average loss=0.1494042741060257 +INFO:root: +[Valid][5][UTT][ MSE = 0.2476 | LCC = 0.9522 | SRCC = 0.9527 ] +INFO:root:Average loss=0.13452978736162186 +INFO:root: +[Valid][6][UTT][ MSE = 0.2149 | LCC = 0.9536 | SRCC = 0.9565 ] +INFO:root:Average loss=0.12278772836923599 +INFO:root: +[Valid][7][UTT][ MSE = 0.2638 | LCC = 0.9462 | SRCC = 0.9474 ] +INFO:root:Average loss=0.11878615748882294 +INFO:root: +[Valid][8][UTT][ MSE = 0.2596 | LCC = 0.9577 | SRCC = 0.9579 ] +INFO:root:Average loss=0.1165471718609333 +INFO:root: +[Valid][9][UTT][ MSE = 0.2007 | LCC = 0.9579 | SRCC = 0.9576 ] +INFO:root:Average loss=0.10512295883893967 +INFO:root: +[Valid][10][UTT][ MSE = 0.1997 | LCC = 0.9560 | SRCC = 0.9575 ] +INFO:root:Average loss=0.10762722018361091 +INFO:root: +[Valid][11][UTT][ MSE = 0.2147 | LCC = 0.9579 | SRCC = 0.9583 ] +INFO:root:Average loss=0.10033104828000068 +INFO:root: +[Valid][12][UTT][ MSE = 0.2267 | LCC = 0.9600 | SRCC = 0.9597 ] +INFO:root:Average loss=0.09317085373401642 +INFO:root: +[Valid][13][UTT][ MSE = 0.2352 | LCC = 0.9599 | SRCC = 0.9601 ] +INFO:root:Average loss=0.08515870162844658 +INFO:root: +[Valid][14][UTT][ MSE = 0.1938 | LCC = 0.9606 | SRCC = 0.9611 ] +INFO:root:Average loss=0.08635106354951859 +INFO:root: +[Valid][15][UTT][ MSE = 0.2025 | LCC = 0.9621 | SRCC = 0.9618 ] +INFO:root:Average loss=0.08063925221562386 +INFO:root: +[Valid][16][UTT][ MSE = 0.2134 | LCC = 0.9614 | SRCC = 0.9614 ] +INFO:root:Average loss=0.07857916888594628 +INFO:root: +[Valid][17][UTT][ MSE = 0.2023 | LCC = 0.9621 | SRCC = 0.9625 ] +INFO:root:Average loss=0.07518075999617577 +INFO:root: +[Valid][18][UTT][ MSE = 0.2204 | LCC = 0.9609 | SRCC = 0.9616 ] +INFO:root:Average loss=0.076436513453722 +INFO:root: +[Valid][19][UTT][ MSE = 0.1891 | LCC = 0.9612 | SRCC = 0.9618 ] +INFO:root:Average loss=0.0737953987121582 +INFO:root: +[Valid][20][UTT][ MSE = 0.1970 | LCC = 0.9616 | SRCC = 0.9633 ] +INFO:root:Average loss=0.0724681699872017 +INFO:root: +[Valid][21][UTT][ MSE = 0.1975 | LCC = 0.9608 | SRCC = 0.9606 ] +INFO:root:Average loss=0.07043879291415214 +INFO:root: +[Valid][22][UTT][ MSE = 0.1753 | LCC = 0.9628 | SRCC = 0.9629 ] +INFO:root:Average loss=0.06537891700863838 +INFO:root: +[Valid][23][UTT][ MSE = 0.2101 | LCC = 0.9634 | SRCC = 0.9636 ] +INFO:root:Average loss=0.06230460485816002 +INFO:root: +[Valid][24][UTT][ MSE = 0.1895 | LCC = 0.9626 | SRCC = 0.9628 ] +INFO:root:Average loss=0.06159735557436943 +INFO:root: +[Valid][25][UTT][ MSE = 0.1727 | LCC = 0.9633 | SRCC = 0.9633 ] +INFO:root:Average loss=0.05870636051893234 +INFO:root: +[Valid][26][UTT][ MSE = 0.1890 | LCC = 0.9628 | SRCC = 0.9637 ] +INFO:root:Average loss=0.05732084459066391 +INFO:root: +[Valid][27][UTT][ MSE = 0.1904 | LCC = 0.9631 | SRCC = 0.9631 ] +INFO:root:Average loss=0.05634220799803734 +INFO:root: +[Valid][28][UTT][ MSE = 0.1991 | LCC = 0.9625 | SRCC = 0.9627 ] +INFO:root:Average loss=0.05664846575260162 +INFO:root: +[Valid][29][UTT][ MSE = 0.1808 | LCC = 0.9629 | SRCC = 0.9632 ] +INFO:root:Average loss=0.0514111861884594 +INFO:root: +[Valid][30][UTT][ MSE = 0.2101 | LCC = 0.9631 | SRCC = 0.9633 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2110 | LCC = 0.9621 | SRCC = 0.9633 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run0/config.gin b/results_tencent/w2v2_large_xlsr53/layer22/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run0/model.pt b/results_tencent/w2v2_large_xlsr53/layer22/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d85be0a3559426c1ed7655569c8f4338e8229d50 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e106a421664ce3408b454f906d6ebae0972db1efa094087494f5488d43526de +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run0/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer22/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..afc63ad9715e3df8fa6a4f0cd761936a8ee184a6 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43369b5ab692a3e0b2c6921ae9980c561399577f5052d8a59d0ebe88cd4f1377 +size 3404998 diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run0/train.log b/results_tencent/w2v2_large_xlsr53/layer22/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..53025a828a973ddbbcedd52b835bf5d7d9e70491 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=1.3684512748718263 +INFO:root: +[Valid][0][UTT][ MSE = 1.3885 | LCC = 0.1672 | SRCC = 0.1790 ] +INFO:root:Average loss=1.3185780019760132 +INFO:root: +[Valid][1][UTT][ MSE = 1.2947 | LCC = 0.3349 | SRCC = 0.3752 ] +INFO:root:Average loss=1.2049365630149842 +INFO:root: +[Valid][2][UTT][ MSE = 1.2159 | LCC = 0.4026 | SRCC = 0.4624 ] +INFO:root:Average loss=1.090262647151947 +INFO:root: +[Valid][3][UTT][ MSE = 1.0047 | LCC = 0.5488 | SRCC = 0.5473 ] +INFO:root:Average loss=1.0320312538146972 +INFO:root: +[Valid][4][UTT][ MSE = 1.0459 | LCC = 0.5368 | SRCC = 0.5486 ] +INFO:root:Average loss=1.0240651378631591 +INFO:root: +[Valid][5][UTT][ MSE = 0.9897 | LCC = 0.5648 | SRCC = 0.5645 ] +INFO:root:Average loss=0.9661209816932679 +INFO:root: +[Valid][6][UTT][ MSE = 0.9609 | LCC = 0.5763 | SRCC = 0.5742 ] +INFO:root:Average loss=0.9734902033805847 +INFO:root: +[Valid][7][UTT][ MSE = 1.0325 | LCC = 0.5661 | SRCC = 0.5542 ] +INFO:root:Average loss=0.9104180889129638 +INFO:root: +[Valid][8][UTT][ MSE = 0.9325 | LCC = 0.5935 | SRCC = 0.5856 ] +INFO:root:Average loss=0.8978527464866638 +INFO:root: +[Valid][9][UTT][ MSE = 0.9123 | LCC = 0.6016 | SRCC = 0.5991 ] +INFO:root:Average loss=0.9005119414329529 +INFO:root: +[Valid][10][UTT][ MSE = 0.9748 | LCC = 0.5881 | SRCC = 0.5970 ] +INFO:root:Average loss=0.9230398035049439 +INFO:root: +[Valid][11][UTT][ MSE = 1.1257 | LCC = 0.5123 | SRCC = 0.5596 ] +INFO:root:Average loss=0.9067802004814148 +INFO:root: +[Valid][12][UTT][ MSE = 0.9734 | LCC = 0.5983 | SRCC = 0.5923 ] +INFO:root:Average loss=0.900419065952301 +INFO:root: +[Valid][13][UTT][ MSE = 0.9126 | LCC = 0.5959 | SRCC = 0.5994 ] +INFO:root:Average loss=0.8760300726890564 +INFO:root: +[Valid][14][UTT][ MSE = 1.1030 | LCC = 0.5393 | SRCC = 0.5718 ] +INFO:root:Average loss=0.8574440999031067 +INFO:root: +[Valid][15][UTT][ MSE = 0.9652 | LCC = 0.5960 | SRCC = 0.5896 ] +INFO:root:Average loss=0.8538707523345948 +INFO:root: +[Valid][16][UTT][ MSE = 0.8911 | LCC = 0.6155 | SRCC = 0.6149 ] +INFO:root:Average loss=0.8536452164649964 +INFO:root: +[Valid][17][UTT][ MSE = 0.9831 | LCC = 0.5643 | SRCC = 0.5883 ] +INFO:root:Average loss=0.8376441011428833 +INFO:root: +[Valid][18][UTT][ MSE = 0.9124 | LCC = 0.6135 | SRCC = 0.6142 ] +INFO:root:Average loss=0.840782039642334 +INFO:root: +[Valid][19][UTT][ MSE = 0.8765 | LCC = 0.6216 | SRCC = 0.6208 ] +INFO:root:Average loss=0.8377396245002746 +INFO:root: +[Valid][20][UTT][ MSE = 0.8734 | LCC = 0.6213 | SRCC = 0.6208 ] +INFO:root:Average loss=0.8281043164730072 +INFO:root: +[Valid][21][UTT][ MSE = 0.9440 | LCC = 0.6166 | SRCC = 0.6171 ] +INFO:root:Average loss=0.8215529828071594 +INFO:root: +[Valid][22][UTT][ MSE = 0.8736 | LCC = 0.6268 | SRCC = 0.6244 ] +INFO:root:Average loss=0.8266616544723511 +INFO:root: +[Valid][23][UTT][ MSE = 0.9041 | LCC = 0.6057 | SRCC = 0.6162 ] +INFO:root:Average loss=0.8205767998695374 +INFO:root: +[Valid][24][UTT][ MSE = 0.8622 | LCC = 0.6238 | SRCC = 0.6265 ] +INFO:root:Average loss=0.8137980906963348 +INFO:root: +[Valid][25][UTT][ MSE = 0.9343 | LCC = 0.6052 | SRCC = 0.6167 ] +INFO:root:Average loss=0.789811425447464 +INFO:root: +[Valid][26][UTT][ MSE = 0.8705 | LCC = 0.6293 | SRCC = 0.6319 ] +INFO:root:Average loss=0.7868204269409179 +INFO:root: +[Valid][27][UTT][ MSE = 0.8523 | LCC = 0.6340 | SRCC = 0.6341 ] +INFO:root:Average loss=0.8345686521530151 +INFO:root: +[Valid][28][UTT][ MSE = 0.8702 | LCC = 0.6369 | SRCC = 0.6375 ] +INFO:root:Average loss=0.770284089565277 +INFO:root: +[Valid][29][UTT][ MSE = 0.8641 | LCC = 0.6336 | SRCC = 0.6305 ] +INFO:root:Average loss=0.7804533228874206 +INFO:root: +[Valid][30][UTT][ MSE = 0.9878 | LCC = 0.5914 | SRCC = 0.5768 ] +INFO:root: +[Test][31][UTT][ MSE = 0.8585 | LCC = 0.6507 | SRCC = 0.6519 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run1/config.gin b/results_tencent/w2v2_large_xlsr53/layer22/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run1/model.pt b/results_tencent/w2v2_large_xlsr53/layer22/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..86935e7335c565e1295b6e283902e29e44141b67 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6c025dca2673d212a9ca2ebb7e8ea86743f4cff200b162c4c70d62f1d9c48eb4 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run1/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer22/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..47d237ec80a241131087c262964da541b26aef4d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5732e1bf70e9efdcde78b0d813483ccee7926cef2fd4b52160c769de003e9921 +size 3404998 diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run1/train.log b/results_tencent/w2v2_large_xlsr53/layer22/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b3fee8162265d1be67c77f14069dd3501b42f076 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=1.3697916808128356 +INFO:root: +[Valid][0][UTT][ MSE = 1.3900 | LCC = 0.1141 | SRCC = 0.1504 ] +INFO:root:Average loss=1.3497833404541015 +INFO:root: +[Valid][1][UTT][ MSE = 1.3658 | LCC = 0.2280 | SRCC = 0.2600 ] +INFO:root:Average loss=1.2831717114448546 +INFO:root: +[Valid][2][UTT][ MSE = 1.2712 | LCC = 0.3537 | SRCC = 0.4343 ] +INFO:root:Average loss=1.1827870154380797 +INFO:root: +[Valid][3][UTT][ MSE = 1.1449 | LCC = 0.4635 | SRCC = 0.4978 ] +INFO:root:Average loss=1.0618492345809936 +INFO:root: +[Valid][4][UTT][ MSE = 1.0259 | LCC = 0.5362 | SRCC = 0.5499 ] +INFO:root:Average loss=1.0068892731666566 +INFO:root: +[Valid][5][UTT][ MSE = 0.9767 | LCC = 0.5603 | SRCC = 0.5642 ] +INFO:root:Average loss=0.9612804822921753 +INFO:root: +[Valid][6][UTT][ MSE = 1.0078 | LCC = 0.5615 | SRCC = 0.5711 ] +INFO:root:Average loss=0.954794273853302 +INFO:root: +[Valid][7][UTT][ MSE = 1.1302 | LCC = 0.5182 | SRCC = 0.5544 ] +INFO:root:Average loss=0.9218985965251922 +INFO:root: +[Valid][8][UTT][ MSE = 1.0428 | LCC = 0.5451 | SRCC = 0.5695 ] +INFO:root:Average loss=0.9151750493049622 +INFO:root: +[Valid][9][UTT][ MSE = 1.0378 | LCC = 0.5566 | SRCC = 0.5748 ] +INFO:root:Average loss=0.9142542119026185 +INFO:root: +[Valid][10][UTT][ MSE = 1.1681 | LCC = 0.5072 | SRCC = 0.5552 ] +INFO:root:Average loss=0.9022729239463806 +INFO:root: +[Valid][11][UTT][ MSE = 1.0196 | LCC = 0.5663 | SRCC = 0.5843 ] +INFO:root:Average loss=0.9036704626083374 +INFO:root: +[Valid][12][UTT][ MSE = 1.0094 | LCC = 0.5626 | SRCC = 0.5848 ] +INFO:root:Average loss=0.8951854686737061 +INFO:root: +[Valid][13][UTT][ MSE = 1.2866 | LCC = 0.4620 | SRCC = 0.5410 ] +INFO:root:Average loss=0.8857113265991211 +INFO:root: +[Valid][14][UTT][ MSE = 1.2660 | LCC = 0.5052 | SRCC = 0.5573 ] +INFO:root:Average loss=0.900576762676239 +INFO:root: +[Valid][15][UTT][ MSE = 1.0779 | LCC = 0.5480 | SRCC = 0.5825 ] +INFO:root:Average loss=0.855360668182373 +INFO:root: +[Valid][16][UTT][ MSE = 0.9595 | LCC = 0.5948 | SRCC = 0.6065 ] +INFO:root:Average loss=0.8354804968833923 +INFO:root: +[Valid][17][UTT][ MSE = 1.1267 | LCC = 0.5485 | SRCC = 0.5850 ] +INFO:root:Average loss=0.8386798362731933 +INFO:root: +[Valid][18][UTT][ MSE = 0.8916 | LCC = 0.6111 | SRCC = 0.6130 ] +INFO:root:Average loss=0.8687886810302734 +INFO:root: +[Valid][19][UTT][ MSE = 0.9202 | LCC = 0.5984 | SRCC = 0.5904 ] +INFO:root:Average loss=0.8639036264419556 +INFO:root: +[Valid][20][UTT][ MSE = 0.9212 | LCC = 0.6086 | SRCC = 0.6132 ] +INFO:root:Average loss=0.8198338603973389 +INFO:root: +[Valid][21][UTT][ MSE = 0.9333 | LCC = 0.6052 | SRCC = 0.6148 ] +INFO:root:Average loss=0.833513179063797 +INFO:root: +[Valid][22][UTT][ MSE = 0.9582 | LCC = 0.6156 | SRCC = 0.6169 ] +INFO:root:Average loss=0.822482721567154 +INFO:root: +[Valid][23][UTT][ MSE = 0.8792 | LCC = 0.6197 | SRCC = 0.6211 ] +INFO:root:Average loss=0.8124448807239533 +INFO:root: +[Valid][24][UTT][ MSE = 0.9455 | LCC = 0.6078 | SRCC = 0.6206 ] +INFO:root:Average loss=0.833196713924408 +INFO:root: +[Valid][25][UTT][ MSE = 0.9011 | LCC = 0.6106 | SRCC = 0.6183 ] +INFO:root:Average loss=0.7988637018203736 +INFO:root: +[Valid][26][UTT][ MSE = 0.8858 | LCC = 0.6232 | SRCC = 0.6270 ] +INFO:root:Average loss=0.8376022176742554 +INFO:root: +[Valid][27][UTT][ MSE = 0.9049 | LCC = 0.6142 | SRCC = 0.6220 ] +INFO:root:Average loss=0.8085892419815064 +INFO:root: +[Valid][28][UTT][ MSE = 0.9282 | LCC = 0.6205 | SRCC = 0.6285 ] +INFO:root:Average loss=0.7840879125595093 +INFO:root: +[Valid][29][UTT][ MSE = 0.8664 | LCC = 0.6294 | SRCC = 0.6285 ] +INFO:root:Average loss=0.7818871817588806 +INFO:root: +[Valid][30][UTT][ MSE = 0.9238 | LCC = 0.6125 | SRCC = 0.6228 ] +INFO:root: +[Test][31][UTT][ MSE = 0.8643 | LCC = 0.6395 | SRCC = 0.6350 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run2/config.gin b/results_tencent/w2v2_large_xlsr53/layer22/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run2/model.pt b/results_tencent/w2v2_large_xlsr53/layer22/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..601c2605daa454286f08500929a3e21873d84a51 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e45396b2306f383e41c29afbbb7d8c9fefbfb6dd3bf482c10c77171f142e4db0 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run2/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer22/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f59b45b06d3213dc7e85a20927fc1f60ec3b0192 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:de96ac2256c12bea651704f235884302f3475b7a864bb7c976fee53df18d4e49 +size 3404998 diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run2/train.log b/results_tencent/w2v2_large_xlsr53/layer22/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e4243964570e6365fdff6baa79e32645c3c50b04 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=1.3495703072547913 +INFO:root: +[Valid][0][UTT][ MSE = 1.3218 | LCC = 0.3019 | SRCC = 0.3599 ] +INFO:root:Average loss=1.2545515756607055 +INFO:root: +[Valid][1][UTT][ MSE = 1.2101 | LCC = 0.3894 | SRCC = 0.4461 ] +INFO:root:Average loss=1.15214506816864 +INFO:root: +[Valid][2][UTT][ MSE = 1.0907 | LCC = 0.4933 | SRCC = 0.4980 ] +INFO:root:Average loss=1.107693200111389 +INFO:root: +[Valid][3][UTT][ MSE = 1.0432 | LCC = 0.5090 | SRCC = 0.5181 ] +INFO:root:Average loss=1.0077357177734374 +INFO:root: +[Valid][4][UTT][ MSE = 1.0721 | LCC = 0.5300 | SRCC = 0.5392 ] +INFO:root:Average loss=0.9847073545455932 +INFO:root: +[Valid][5][UTT][ MSE = 1.0560 | LCC = 0.5779 | SRCC = 0.5742 ] +INFO:root:Average loss=1.0210583400726319 +INFO:root: +[Valid][6][UTT][ MSE = 0.9725 | LCC = 0.5681 | SRCC = 0.5702 ] +INFO:root:Average loss=0.9871400418281555 +INFO:root: +[Valid][7][UTT][ MSE = 0.9701 | LCC = 0.5765 | SRCC = 0.5694 ] +INFO:root:Average loss=0.9574124264717102 +INFO:root: +[Valid][8][UTT][ MSE = 1.1179 | LCC = 0.5101 | SRCC = 0.5503 ] +INFO:root:Average loss=1.0059001529216767 +INFO:root: +[Valid][9][UTT][ MSE = 0.9923 | LCC = 0.5809 | SRCC = 0.5733 ] +INFO:root:Average loss=0.9298468236923217 +INFO:root: +[Valid][10][UTT][ MSE = 1.0001 | LCC = 0.5797 | SRCC = 0.5839 ] +INFO:root:Average loss=0.9162679376602173 +INFO:root: +[Valid][11][UTT][ MSE = 0.9727 | LCC = 0.5777 | SRCC = 0.5697 ] +INFO:root:Average loss=0.8648003997802735 +INFO:root: +[Valid][12][UTT][ MSE = 0.9134 | LCC = 0.6055 | SRCC = 0.5997 ] +INFO:root:Average loss=0.9620122857093811 +INFO:root: +[Valid][13][UTT][ MSE = 0.9522 | LCC = 0.5913 | SRCC = 0.5984 ] +INFO:root:Average loss=0.8928136167526245 +INFO:root: +[Valid][14][UTT][ MSE = 0.9032 | LCC = 0.5998 | SRCC = 0.5990 ] +INFO:root:Average loss=0.8825822958946228 +INFO:root: +[Valid][15][UTT][ MSE = 0.9475 | LCC = 0.5941 | SRCC = 0.5996 ] +INFO:root:Average loss=0.888066017150879 +INFO:root: +[Valid][16][UTT][ MSE = 1.0374 | LCC = 0.5741 | SRCC = 0.5777 ] +INFO:root:Average loss=0.874454731464386 +INFO:root: +[Valid][17][UTT][ MSE = 1.1985 | LCC = 0.5216 | SRCC = 0.5677 ] +INFO:root:Average loss=0.8628517601490021 +INFO:root: +[Valid][18][UTT][ MSE = 0.9129 | LCC = 0.6126 | SRCC = 0.6088 ] +INFO:root:Average loss=0.8409099450111389 +INFO:root: +[Valid][19][UTT][ MSE = 0.9060 | LCC = 0.6024 | SRCC = 0.6081 ] +INFO:root:Average loss=0.8411545405387878 +INFO:root: +[Valid][20][UTT][ MSE = 1.0800 | LCC = 0.5612 | SRCC = 0.5913 ] +INFO:root:Average loss=0.8665351037979125 +INFO:root: +[Valid][21][UTT][ MSE = 0.9124 | LCC = 0.6189 | SRCC = 0.6146 ] +INFO:root:Average loss=0.8310923213958741 +INFO:root: +[Valid][22][UTT][ MSE = 0.9211 | LCC = 0.6231 | SRCC = 0.6223 ] +INFO:root:Average loss=0.8151833081245422 +INFO:root: +[Valid][23][UTT][ MSE = 0.8684 | LCC = 0.6339 | SRCC = 0.6332 ] +INFO:root:Average loss=0.8241881957054138 +INFO:root: +[Valid][24][UTT][ MSE = 0.9038 | LCC = 0.6081 | SRCC = 0.6152 ] +INFO:root:Average loss=0.861776032447815 +INFO:root: +[Valid][25][UTT][ MSE = 0.9525 | LCC = 0.6089 | SRCC = 0.6167 ] +INFO:root:Average loss=0.8041440052986145 +INFO:root: +[Valid][26][UTT][ MSE = 0.8952 | LCC = 0.6234 | SRCC = 0.6239 ] +INFO:root:Average loss=0.794885272026062 +INFO:root: +[Valid][27][UTT][ MSE = 0.8376 | LCC = 0.6415 | SRCC = 0.6411 ] +INFO:root:Average loss=0.7781342220306396 +INFO:root: +[Valid][28][UTT][ MSE = 0.8619 | LCC = 0.6305 | SRCC = 0.6301 ] +INFO:root:Average loss=0.8157622120380401 +INFO:root: +[Valid][29][UTT][ MSE = 0.9124 | LCC = 0.6339 | SRCC = 0.6345 ] +INFO:root:Average loss=0.8012469696998596 +INFO:root: +[Valid][30][UTT][ MSE = 0.8473 | LCC = 0.6382 | SRCC = 0.6361 ] +INFO:root: +[Test][31][UTT][ MSE = 0.8268 | LCC = 0.6527 | SRCC = 0.6510 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run3/config.gin b/results_tencent/w2v2_large_xlsr53/layer22/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run3/model.pt b/results_tencent/w2v2_large_xlsr53/layer22/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..101e1f2dba5694f4e4b9c31183184de4d9da7bf5 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:21b30b2a81cc97b8af95e4d0135876569308abc0c57a3e53e04336de35d447e7 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run3/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer22/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..30e4972570ba80332bcab0bee0b4c080a36fd99c --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a15be03da4b5cdc8154f95fe510b7f7759b695f2a6013004e429980c661b5e0d +size 3404998 diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run3/train.log b/results_tencent/w2v2_large_xlsr53/layer22/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..15e13721078825b279a58c048eb222f1ccfbce50 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=1.369110429763794 +INFO:root: +[Valid][0][UTT][ MSE = 1.3837 | LCC = 0.1717 | SRCC = 0.2095 ] +INFO:root:Average loss=1.351394769668579 +INFO:root: +[Valid][1][UTT][ MSE = 1.3630 | LCC = 0.3155 | SRCC = 0.4007 ] +INFO:root:Average loss=1.2094545769691467 +INFO:root: +[Valid][2][UTT][ MSE = 1.0787 | LCC = 0.4843 | SRCC = 0.5010 ] +INFO:root:Average loss=1.0752092599868774 +INFO:root: +[Valid][3][UTT][ MSE = 1.0475 | LCC = 0.5302 | SRCC = 0.5233 ] +INFO:root:Average loss=1.0156070404052735 +INFO:root: +[Valid][4][UTT][ MSE = 0.9600 | LCC = 0.5666 | SRCC = 0.5411 ] +INFO:root:Average loss=0.9849668712615967 +INFO:root: +[Valid][5][UTT][ MSE = 1.0994 | LCC = 0.5061 | SRCC = 0.5369 ] +INFO:root:Average loss=0.999298602104187 +INFO:root: +[Valid][6][UTT][ MSE = 1.0211 | LCC = 0.5540 | SRCC = 0.5509 ] +INFO:root:Average loss=0.9399776520729065 +INFO:root: +[Valid][7][UTT][ MSE = 0.9617 | LCC = 0.5736 | SRCC = 0.5722 ] +INFO:root:Average loss=0.9784397664070129 +INFO:root: +[Valid][8][UTT][ MSE = 1.1506 | LCC = 0.5076 | SRCC = 0.5548 ] +INFO:root:Average loss=0.9473101840019226 +INFO:root: +[Valid][9][UTT][ MSE = 0.9395 | LCC = 0.5924 | SRCC = 0.5892 ] +INFO:root:Average loss=0.8985693416595459 +INFO:root: +[Valid][10][UTT][ MSE = 1.0796 | LCC = 0.5444 | SRCC = 0.5711 ] +INFO:root:Average loss=0.9150661301612854 +INFO:root: +[Valid][11][UTT][ MSE = 0.9617 | LCC = 0.5873 | SRCC = 0.5722 ] +INFO:root:Average loss=0.9139590497016907 +INFO:root: +[Valid][12][UTT][ MSE = 1.1158 | LCC = 0.5262 | SRCC = 0.5714 ] +INFO:root:Average loss=0.9358611030578613 +INFO:root: +[Valid][13][UTT][ MSE = 1.2096 | LCC = 0.4332 | SRCC = 0.3707 ] +INFO:root:Average loss=0.8878896789550781 +INFO:root: +[Valid][14][UTT][ MSE = 0.9113 | LCC = 0.6059 | SRCC = 0.6051 ] +INFO:root:Average loss=0.9019445605278015 +INFO:root: +[Valid][15][UTT][ MSE = 0.9326 | LCC = 0.5885 | SRCC = 0.5962 ] +INFO:root:Average loss=0.8482020072937012 +INFO:root: +[Valid][16][UTT][ MSE = 0.9076 | LCC = 0.6164 | SRCC = 0.6151 ] +INFO:root:Average loss=0.8622162408828735 +INFO:root: +[Valid][17][UTT][ MSE = 0.9166 | LCC = 0.6162 | SRCC = 0.6147 ] +INFO:root:Average loss=0.8521528110504151 +INFO:root: +[Valid][18][UTT][ MSE = 0.9565 | LCC = 0.6118 | SRCC = 0.6089 ] +INFO:root:Average loss=0.8448067750930787 +INFO:root: +[Valid][19][UTT][ MSE = 0.9469 | LCC = 0.5953 | SRCC = 0.6066 ] +INFO:root:Average loss=0.8656328310966491 +INFO:root: +[Valid][20][UTT][ MSE = 0.9086 | LCC = 0.6037 | SRCC = 0.6127 ] +INFO:root:Average loss=0.8235687294006347 +INFO:root: +[Valid][21][UTT][ MSE = 0.9038 | LCC = 0.6050 | SRCC = 0.6125 ] +INFO:root:Average loss=0.8329561674594879 +INFO:root: +[Valid][22][UTT][ MSE = 0.9117 | LCC = 0.6121 | SRCC = 0.6192 ] +INFO:root:Average loss=0.8170331144332885 +INFO:root: +[Valid][23][UTT][ MSE = 0.8916 | LCC = 0.6161 | SRCC = 0.6120 ] +INFO:root:Average loss=0.8542784533500671 +INFO:root: +[Valid][24][UTT][ MSE = 0.9062 | LCC = 0.6222 | SRCC = 0.6191 ] +INFO:root:Average loss=0.8383207111358643 +INFO:root: +[Valid][25][UTT][ MSE = 0.8887 | LCC = 0.6251 | SRCC = 0.6306 ] +INFO:root:Average loss=0.8084870781898499 +INFO:root: +[Valid][26][UTT][ MSE = 0.8734 | LCC = 0.6278 | SRCC = 0.6238 ] +INFO:root:Average loss=0.8049627108573914 +INFO:root: +[Valid][27][UTT][ MSE = 0.8859 | LCC = 0.6312 | SRCC = 0.6296 ] +INFO:root:Average loss=0.8142713389396667 +INFO:root: +[Valid][28][UTT][ MSE = 0.9437 | LCC = 0.5957 | SRCC = 0.6144 ] +INFO:root:Average loss=0.8651718735694885 +INFO:root: +[Valid][29][UTT][ MSE = 1.0143 | LCC = 0.5804 | SRCC = 0.5669 ] +INFO:root:Average loss=0.8287730617523194 +INFO:root: +[Valid][30][UTT][ MSE = 0.8514 | LCC = 0.6391 | SRCC = 0.6414 ] +INFO:root: +[Test][31][UTT][ MSE = 0.8454 | LCC = 0.6501 | SRCC = 0.6495 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run4/config.gin b/results_tencent/w2v2_large_xlsr53/layer22/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run4/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer22/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a9ae913443a5c9741aa6444da5c991d78d3ab6ff --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3765240e875917bb8d2d1c79176cf74f81d5592e58d2eeab9a1c30986eeeeaf6 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer22/run4/train.log b/results_tencent/w2v2_large_xlsr53/layer22/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..990e37c2fd8462acdc202349403e7bd5679f4b82 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer22/run4/train.log @@ -0,0 +1,94 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=1.356259642124176 +INFO:root: +[Valid][0][UTT][ MSE = 1.3401 | LCC = 0.2964 | SRCC = 0.3647 ] +INFO:root:Average loss=1.2133558416366577 +INFO:root: +[Valid][1][UTT][ MSE = 1.1484 | LCC = 0.4919 | SRCC = 0.4842 ] +INFO:root:Average loss=1.0919169664382935 +INFO:root: +[Valid][2][UTT][ MSE = 1.1631 | LCC = 0.5210 | SRCC = 0.5279 ] +INFO:root:Average loss=0.9957544240951538 +INFO:root: +[Valid][3][UTT][ MSE = 0.9691 | LCC = 0.5742 | SRCC = 0.5581 ] +INFO:root:Average loss=1.0152030334472657 +INFO:root: +[Valid][4][UTT][ MSE = 0.9617 | LCC = 0.5752 | SRCC = 0.5613 ] +INFO:root:Average loss=0.9667147655487061 +INFO:root: +[Valid][5][UTT][ MSE = 0.9905 | LCC = 0.5623 | SRCC = 0.5606 ] +INFO:root:Average loss=0.9736556425094605 +INFO:root: +[Valid][6][UTT][ MSE = 0.9488 | LCC = 0.5867 | SRCC = 0.5707 ] +INFO:root:Average loss=0.9562579870224 +INFO:root: +[Valid][7][UTT][ MSE = 1.0676 | LCC = 0.5756 | SRCC = 0.5759 ] +INFO:root:Average loss=0.9792323846817017 +INFO:root: +[Valid][8][UTT][ MSE = 1.0029 | LCC = 0.5657 | SRCC = 0.5499 ] +INFO:root:Average loss=0.9449162821769714 +INFO:root: +[Valid][9][UTT][ MSE = 0.9339 | LCC = 0.5961 | SRCC = 0.5890 ] +INFO:root:Average loss=0.9622143983840943 +INFO:root: +[Valid][10][UTT][ MSE = 0.9653 | LCC = 0.5867 | SRCC = 0.5850 ] +INFO:root:Average loss=0.8942998313903808 +INFO:root: +[Valid][11][UTT][ MSE = 0.9018 | LCC = 0.6080 | SRCC = 0.6052 ] +INFO:root:Average loss=0.9034388942718505 +INFO:root: +[Valid][12][UTT][ MSE = 0.9212 | LCC = 0.5909 | SRCC = 0.5947 ] +INFO:root:Average loss=0.8836018838882447 +INFO:root: +[Valid][13][UTT][ MSE = 1.0126 | LCC = 0.5724 | SRCC = 0.5695 ] +INFO:root:Average loss=0.8638240973949433 +INFO:root: +[Valid][14][UTT][ MSE = 0.9513 | LCC = 0.5917 | SRCC = 0.5791 ] +INFO:root:Average loss=0.872066520690918 +INFO:root: +[Valid][15][UTT][ MSE = 0.8882 | LCC = 0.6227 | SRCC = 0.6201 ] +INFO:root:Average loss=0.847031183719635 +INFO:root: +[Valid][16][UTT][ MSE = 0.8828 | LCC = 0.6229 | SRCC = 0.6183 ] +INFO:root:Average loss=0.8669022250175477 +INFO:root: +[Valid][17][UTT][ MSE = 0.9869 | LCC = 0.5878 | SRCC = 0.6028 ] +INFO:root:Average loss=0.8391982946395874 +INFO:root: +[Valid][18][UTT][ MSE = 0.9091 | LCC = 0.6172 | SRCC = 0.6214 ] +INFO:root:Average loss=0.8849582862854004 +INFO:root: +[Valid][19][UTT][ MSE = 0.8879 | LCC = 0.6194 | SRCC = 0.6143 ] +INFO:root:Average loss=0.8828304462432861 +INFO:root: +[Valid][20][UTT][ MSE = 0.8794 | LCC = 0.6239 | SRCC = 0.6201 ] +INFO:root:Average loss=0.8222908918857574 +INFO:root: +[Valid][21][UTT][ MSE = 0.9440 | LCC = 0.5924 | SRCC = 0.5778 ] +INFO:root:Average loss=0.8682467012405396 +INFO:root: +[Valid][22][UTT][ MSE = 0.8897 | LCC = 0.6306 | SRCC = 0.6288 ] +INFO:root:Average loss=0.8364812164306641 +INFO:root: +[Valid][23][UTT][ MSE = 1.1009 | LCC = 0.5260 | SRCC = 0.4970 ] +INFO:root:Average loss=0.8083230583667755 +INFO:root: +[Valid][24][UTT][ MSE = 0.8602 | LCC = 0.6315 | SRCC = 0.6284 ] +INFO:root:Average loss=0.8384731779098511 +INFO:root: +[Valid][25][UTT][ MSE = 0.9010 | LCC = 0.6159 | SRCC = 0.6103 ] +INFO:root:Average loss=0.8308447022438049 +INFO:root: +[Valid][26][UTT][ MSE = 0.9532 | LCC = 0.6023 | SRCC = 0.6120 ] +INFO:root:Average loss=0.8018397459983826 +INFO:root: +[Valid][27][UTT][ MSE = 0.9406 | LCC = 0.6036 | SRCC = 0.6170 ] +INFO:root:Average loss=0.8173521523475646 +INFO:root: +[Valid][28][UTT][ MSE = 0.8532 | LCC = 0.6380 | SRCC = 0.6351 ] +INFO:root:Average loss=0.8198311400413513 +INFO:root: +[Valid][29][UTT][ MSE = 0.9304 | LCC = 0.6119 | SRCC = 0.6187 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run0/config.gin b/results_tencent/w2v2_large_xlsr53/layer4/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run0/model.pt b/results_tencent/w2v2_large_xlsr53/layer4/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4b042067ef2f691de9e34f01748eac8dc981a498 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d53c47113cf8862857147a16263cc8120a5ea673c590079e9cda4aae85015e7d +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run0/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer4/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..1fa1690c6ec34795a88463e76b9daac3e1dbc963 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:671ccac5fe34e7d25bbf5febb0f90ae36041a596a5c89eb2da2bcaae57701b2a +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run0/train.log b/results_tencent/w2v2_large_xlsr53/layer4/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3cff5cab455ce0303b5325b49920aef7cdea0dc4 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.49568771278858187 +INFO:root: +[Valid][0][UTT][ MSE = 0.2891 | LCC = 0.9252 | SRCC = 0.9256 ] +INFO:root:Average loss=0.20665243685245513 +INFO:root: +[Valid][1][UTT][ MSE = 0.3106 | LCC = 0.9441 | SRCC = 0.9448 ] +INFO:root:Average loss=0.16773505944013595 +INFO:root: +[Valid][2][UTT][ MSE = 0.2674 | LCC = 0.9504 | SRCC = 0.9512 ] +INFO:root:Average loss=0.14203645437955856 +INFO:root: +[Valid][3][UTT][ MSE = 0.2615 | LCC = 0.9531 | SRCC = 0.9541 ] +INFO:root:Average loss=0.1304571964740753 +INFO:root: +[Valid][4][UTT][ MSE = 0.2066 | LCC = 0.9550 | SRCC = 0.9559 ] +INFO:root:Average loss=0.12236178946495056 +INFO:root: +[Valid][5][UTT][ MSE = 0.2514 | LCC = 0.9558 | SRCC = 0.9584 ] +INFO:root:Average loss=0.10491409486532212 +INFO:root: +[Valid][6][UTT][ MSE = 0.2358 | LCC = 0.9570 | SRCC = 0.9593 ] +INFO:root:Average loss=0.10390502947568893 +INFO:root: +[Valid][7][UTT][ MSE = 0.2561 | LCC = 0.9574 | SRCC = 0.9603 ] +INFO:root:Average loss=0.09268353062868118 +INFO:root: +[Valid][8][UTT][ MSE = 0.1842 | LCC = 0.9596 | SRCC = 0.9599 ] +INFO:root:Average loss=0.09018238738179207 +INFO:root: +[Valid][9][UTT][ MSE = 0.2193 | LCC = 0.9601 | SRCC = 0.9616 ] +INFO:root:Average loss=0.08421422469615936 +INFO:root: +[Valid][10][UTT][ MSE = 0.2042 | LCC = 0.9622 | SRCC = 0.9621 ] +INFO:root:Average loss=0.0819750099182129 +INFO:root: +[Valid][11][UTT][ MSE = 0.1814 | LCC = 0.9610 | SRCC = 0.9621 ] +INFO:root:Average loss=0.07878411772847176 +INFO:root: +[Valid][12][UTT][ MSE = 0.2114 | LCC = 0.9615 | SRCC = 0.9630 ] +INFO:root:Average loss=0.07244308656454086 +INFO:root: +[Valid][13][UTT][ MSE = 0.1892 | LCC = 0.9626 | SRCC = 0.9631 ] +INFO:root:Average loss=0.06986099311709404 +INFO:root: +[Valid][14][UTT][ MSE = 0.2319 | LCC = 0.9618 | SRCC = 0.9637 ] +INFO:root:Average loss=0.06782693353295326 +INFO:root: +[Valid][15][UTT][ MSE = 0.1873 | LCC = 0.9631 | SRCC = 0.9638 ] +INFO:root:Average loss=0.06399818402528763 +INFO:root: +[Valid][16][UTT][ MSE = 0.1666 | LCC = 0.9640 | SRCC = 0.9641 ] +INFO:root:Average loss=0.06087119287252426 +INFO:root: +[Valid][17][UTT][ MSE = 0.2052 | LCC = 0.9630 | SRCC = 0.9637 ] +INFO:root:Average loss=0.05885013017058373 +INFO:root: +[Valid][18][UTT][ MSE = 0.2225 | LCC = 0.9615 | SRCC = 0.9644 ] +INFO:root:Average loss=0.058068616032600405 +INFO:root: +[Valid][19][UTT][ MSE = 0.1938 | LCC = 0.9640 | SRCC = 0.9644 ] +INFO:root:Average loss=0.05432378536462784 +INFO:root: +[Valid][20][UTT][ MSE = 0.1856 | LCC = 0.9637 | SRCC = 0.9642 ] +INFO:root:Average loss=0.052037926599383354 +INFO:root: +[Valid][21][UTT][ MSE = 0.1732 | LCC = 0.9644 | SRCC = 0.9646 ] +INFO:root:Average loss=0.051944830656051634 +INFO:root: +[Valid][22][UTT][ MSE = 0.1760 | LCC = 0.9642 | SRCC = 0.9641 ] +INFO:root:Average loss=0.04712252528965473 +INFO:root: +[Valid][23][UTT][ MSE = 0.2016 | LCC = 0.9629 | SRCC = 0.9633 ] +INFO:root:Average loss=0.04501920002698898 +INFO:root: +[Valid][24][UTT][ MSE = 0.2044 | LCC = 0.9634 | SRCC = 0.9638 ] +INFO:root:Average loss=0.044788744404912 +INFO:root: +[Valid][25][UTT][ MSE = 0.1837 | LCC = 0.9643 | SRCC = 0.9644 ] +INFO:root:Average loss=0.042216482385993 +INFO:root: +[Valid][26][UTT][ MSE = 0.1909 | LCC = 0.9626 | SRCC = 0.9624 ] +INFO:root:Average loss=0.04065792453289032 +INFO:root: +[Valid][27][UTT][ MSE = 0.1920 | LCC = 0.9615 | SRCC = 0.9625 ] +INFO:root:Average loss=0.03906614589691162 +INFO:root: +[Valid][28][UTT][ MSE = 0.1874 | LCC = 0.9644 | SRCC = 0.9644 ] +INFO:root:Average loss=0.03863939885795117 +INFO:root: +[Valid][29][UTT][ MSE = 0.1678 | LCC = 0.9638 | SRCC = 0.9637 ] +INFO:root:Average loss=0.037019078463315966 +INFO:root: +[Valid][30][UTT][ MSE = 0.2081 | LCC = 0.9633 | SRCC = 0.9638 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1788 | LCC = 0.9644 | SRCC = 0.9648 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run1/config.gin b/results_tencent/w2v2_large_xlsr53/layer4/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run1/model.pt b/results_tencent/w2v2_large_xlsr53/layer4/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..11b863295d1e08f2a8713dabcad5c5b9457979a0 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d1f09fd3aa081f6697c5d44aa536003c0c6b1f47a2109cb4f9faca947494bbb7 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run1/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer4/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4e78f11b98f596e3911957b7a8570afa37bafba7 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a5bbdceb4a7ba7ac7aa056dc67b9039ed77044d37bbad694b50a68f1b8c14bb0 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run1/train.log b/results_tencent/w2v2_large_xlsr53/layer4/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..1337c7443c9816dbe171d39089b8fb046b5a2fdb --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44240096366405485 +INFO:root: +[Valid][0][UTT][ MSE = 0.3265 | LCC = 0.9215 | SRCC = 0.9239 ] +INFO:root:Average loss=0.20229079592227936 +INFO:root: +[Valid][1][UTT][ MSE = 0.2591 | LCC = 0.9405 | SRCC = 0.9410 ] +INFO:root:Average loss=0.17063200271129608 +INFO:root: +[Valid][2][UTT][ MSE = 0.2759 | LCC = 0.9438 | SRCC = 0.9473 ] +INFO:root:Average loss=0.14838811844587327 +INFO:root: +[Valid][3][UTT][ MSE = 0.2397 | LCC = 0.9501 | SRCC = 0.9517 ] +INFO:root:Average loss=0.13592338919639588 +INFO:root: +[Valid][4][UTT][ MSE = 0.2363 | LCC = 0.9525 | SRCC = 0.9537 ] +INFO:root:Average loss=0.11876079398393631 +INFO:root: +[Valid][5][UTT][ MSE = 0.2410 | LCC = 0.9550 | SRCC = 0.9567 ] +INFO:root:Average loss=0.11529283645749092 +INFO:root: +[Valid][6][UTT][ MSE = 0.2523 | LCC = 0.9568 | SRCC = 0.9585 ] +INFO:root:Average loss=0.10349653246998787 +INFO:root: +[Valid][7][UTT][ MSE = 0.2474 | LCC = 0.9560 | SRCC = 0.9566 ] +INFO:root:Average loss=0.10414392253756523 +INFO:root: +[Valid][8][UTT][ MSE = 0.2127 | LCC = 0.9586 | SRCC = 0.9588 ] +INFO:root:Average loss=0.09088124322891235 +INFO:root: +[Valid][9][UTT][ MSE = 0.2579 | LCC = 0.9601 | SRCC = 0.9625 ] +INFO:root:Average loss=0.08896429407596589 +INFO:root: +[Valid][10][UTT][ MSE = 0.2331 | LCC = 0.9609 | SRCC = 0.9618 ] +INFO:root:Average loss=0.08434439840912819 +INFO:root: +[Valid][11][UTT][ MSE = 0.2281 | LCC = 0.9600 | SRCC = 0.9617 ] +INFO:root:Average loss=0.07783893156051636 +INFO:root: +[Valid][12][UTT][ MSE = 0.1958 | LCC = 0.9604 | SRCC = 0.9613 ] +INFO:root:Average loss=0.0752840034365654 +INFO:root: +[Valid][13][UTT][ MSE = 0.2069 | LCC = 0.9609 | SRCC = 0.9611 ] +INFO:root:Average loss=0.07023538342118263 +INFO:root: +[Valid][14][UTT][ MSE = 0.1955 | LCC = 0.9629 | SRCC = 0.9632 ] +INFO:root:Average loss=0.07019751277565957 +INFO:root: +[Valid][15][UTT][ MSE = 0.1981 | LCC = 0.9634 | SRCC = 0.9641 ] +INFO:root:Average loss=0.06666027227044105 +INFO:root: +[Valid][16][UTT][ MSE = 0.1950 | LCC = 0.9635 | SRCC = 0.9641 ] +INFO:root:Average loss=0.06168326431512833 +INFO:root: +[Valid][17][UTT][ MSE = 0.2152 | LCC = 0.9629 | SRCC = 0.9632 ] +INFO:root:Average loss=0.058819687336683274 +INFO:root: +[Valid][18][UTT][ MSE = 0.2184 | LCC = 0.9626 | SRCC = 0.9646 ] +INFO:root:Average loss=0.05694060525298118 +INFO:root: +[Valid][19][UTT][ MSE = 0.1907 | LCC = 0.9644 | SRCC = 0.9647 ] +INFO:root:Average loss=0.05256781232357025 +INFO:root: +[Valid][20][UTT][ MSE = 0.1772 | LCC = 0.9637 | SRCC = 0.9636 ] +INFO:root:Average loss=0.05558414015173912 +INFO:root: +[Valid][21][UTT][ MSE = 0.1707 | LCC = 0.9647 | SRCC = 0.9643 ] +INFO:root:Average loss=0.049561049550771716 +INFO:root: +[Valid][22][UTT][ MSE = 0.1776 | LCC = 0.9654 | SRCC = 0.9651 ] +INFO:root:Average loss=0.049000617220997814 +INFO:root: +[Valid][23][UTT][ MSE = 0.1802 | LCC = 0.9642 | SRCC = 0.9638 ] +INFO:root:Average loss=0.04703338797390461 +INFO:root: +[Valid][24][UTT][ MSE = 0.1773 | LCC = 0.9652 | SRCC = 0.9651 ] +INFO:root:Average loss=0.04335244090855122 +INFO:root: +[Valid][25][UTT][ MSE = 0.1686 | LCC = 0.9646 | SRCC = 0.9646 ] +INFO:root:Average loss=0.04411008015275002 +INFO:root: +[Valid][26][UTT][ MSE = 0.1937 | LCC = 0.9642 | SRCC = 0.9647 ] +INFO:root:Average loss=0.04230431069433689 +INFO:root: +[Valid][27][UTT][ MSE = 0.1853 | LCC = 0.9647 | SRCC = 0.9646 ] +INFO:root:Average loss=0.04393478155136108 +INFO:root: +[Valid][28][UTT][ MSE = 0.1902 | LCC = 0.9626 | SRCC = 0.9626 ] +INFO:root:Average loss=0.040181911572813984 +INFO:root: +[Valid][29][UTT][ MSE = 0.1829 | LCC = 0.9648 | SRCC = 0.9650 ] +INFO:root:Average loss=0.03744427041709423 +INFO:root: +[Valid][30][UTT][ MSE = 0.2003 | LCC = 0.9648 | SRCC = 0.9656 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1846 | LCC = 0.9632 | SRCC = 0.9639 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run2/config.gin b/results_tencent/w2v2_large_xlsr53/layer4/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run2/model.pt b/results_tencent/w2v2_large_xlsr53/layer4/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1c081d1995c4d0a326393e4bdabbce332718ac9c --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ce0c81a94e1678448e4cab7e66c41d6a1c2ffa786a4e5c009d787ea59782808c +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run2/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer4/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a5d77476777afebfd1b673bf95cd9d1ab6660615 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d630e885e194e6b46f380e7a033207b91644d42ffbbeb9b6e2839f14e1a964ca +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run2/train.log b/results_tencent/w2v2_large_xlsr53/layer4/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..16b177de8819b2c6014aa60b3c9cace17e26da8b --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4316055310368538 +INFO:root: +[Valid][0][UTT][ MSE = 0.2896 | LCC = 0.9280 | SRCC = 0.9286 ] +INFO:root:Average loss=0.18723186254501342 +INFO:root: +[Valid][1][UTT][ MSE = 0.2692 | LCC = 0.9431 | SRCC = 0.9443 ] +INFO:root:Average loss=0.16085913902521135 +INFO:root: +[Valid][2][UTT][ MSE = 0.2227 | LCC = 0.9498 | SRCC = 0.9504 ] +INFO:root:Average loss=0.14694820156693458 +INFO:root: +[Valid][3][UTT][ MSE = 0.2231 | LCC = 0.9495 | SRCC = 0.9509 ] +INFO:root:Average loss=0.13272786569595338 +INFO:root: +[Valid][4][UTT][ MSE = 0.2481 | LCC = 0.9548 | SRCC = 0.9560 ] +INFO:root:Average loss=0.11466223323345184 +INFO:root: +[Valid][5][UTT][ MSE = 0.2023 | LCC = 0.9566 | SRCC = 0.9581 ] +INFO:root:Average loss=0.11551899361610413 +INFO:root: +[Valid][6][UTT][ MSE = 0.2154 | LCC = 0.9562 | SRCC = 0.9576 ] +INFO:root:Average loss=0.10499017107486724 +INFO:root: +[Valid][7][UTT][ MSE = 0.1905 | LCC = 0.9582 | SRCC = 0.9601 ] +INFO:root:Average loss=0.10011137548089027 +INFO:root: +[Valid][8][UTT][ MSE = 0.1830 | LCC = 0.9597 | SRCC = 0.9608 ] +INFO:root:Average loss=0.09029098671674729 +INFO:root: +[Valid][9][UTT][ MSE = 0.1994 | LCC = 0.9614 | SRCC = 0.9628 ] +INFO:root:Average loss=0.08588666477799416 +INFO:root: +[Valid][10][UTT][ MSE = 0.1828 | LCC = 0.9617 | SRCC = 0.9638 ] +INFO:root:Average loss=0.08650356969237327 +INFO:root: +[Valid][11][UTT][ MSE = 0.2015 | LCC = 0.9614 | SRCC = 0.9636 ] +INFO:root:Average loss=0.077980783700943 +INFO:root: +[Valid][12][UTT][ MSE = 0.1682 | LCC = 0.9616 | SRCC = 0.9627 ] +INFO:root:Average loss=0.0759276907145977 +INFO:root: +[Valid][13][UTT][ MSE = 0.2218 | LCC = 0.9624 | SRCC = 0.9641 ] +INFO:root:Average loss=0.07222943523526192 +INFO:root: +[Valid][14][UTT][ MSE = 0.1978 | LCC = 0.9627 | SRCC = 0.9638 ] +INFO:root:Average loss=0.06764454585313796 +INFO:root: +[Valid][15][UTT][ MSE = 0.1999 | LCC = 0.9639 | SRCC = 0.9645 ] +INFO:root:Average loss=0.06574479800462722 +INFO:root: +[Valid][16][UTT][ MSE = 0.1790 | LCC = 0.9624 | SRCC = 0.9641 ] +INFO:root:Average loss=0.06271636083722115 +INFO:root: +[Valid][17][UTT][ MSE = 0.1716 | LCC = 0.9632 | SRCC = 0.9648 ] +INFO:root:Average loss=0.0633160856962204 +INFO:root: +[Valid][18][UTT][ MSE = 0.1769 | LCC = 0.9640 | SRCC = 0.9647 ] +INFO:root:Average loss=0.05579391665756703 +INFO:root: +[Valid][19][UTT][ MSE = 0.1832 | LCC = 0.9660 | SRCC = 0.9665 ] +INFO:root:Average loss=0.055092721581459045 +INFO:root: +[Valid][20][UTT][ MSE = 0.1808 | LCC = 0.9654 | SRCC = 0.9659 ] +INFO:root:Average loss=0.05562936975061893 +INFO:root: +[Valid][21][UTT][ MSE = 0.1878 | LCC = 0.9645 | SRCC = 0.9652 ] +INFO:root:Average loss=0.050083819717168805 +INFO:root: +[Valid][22][UTT][ MSE = 0.1650 | LCC = 0.9660 | SRCC = 0.9662 ] +INFO:root:Average loss=0.04822753404080868 +INFO:root: +[Valid][23][UTT][ MSE = 0.1705 | LCC = 0.9643 | SRCC = 0.9648 ] +INFO:root:Average loss=0.045232426300644876 +INFO:root: +[Valid][24][UTT][ MSE = 0.1713 | LCC = 0.9638 | SRCC = 0.9651 ] +INFO:root:Average loss=0.04525612276792526 +INFO:root: +[Valid][25][UTT][ MSE = 0.1890 | LCC = 0.9644 | SRCC = 0.9650 ] +INFO:root:Average loss=0.04516527709364891 +INFO:root: +[Valid][26][UTT][ MSE = 0.1892 | LCC = 0.9644 | SRCC = 0.9648 ] +INFO:root:Average loss=0.0424241561293602 +INFO:root: +[Valid][27][UTT][ MSE = 0.1627 | LCC = 0.9631 | SRCC = 0.9642 ] +INFO:root:Average loss=0.041688225105404854 +INFO:root: +[Valid][28][UTT][ MSE = 0.1639 | LCC = 0.9606 | SRCC = 0.9639 ] +INFO:root:Average loss=0.043878531277179715 +INFO:root: +[Valid][29][UTT][ MSE = 0.1755 | LCC = 0.9632 | SRCC = 0.9640 ] +INFO:root:Average loss=0.03870697954297066 +INFO:root: +[Valid][30][UTT][ MSE = 0.1622 | LCC = 0.9660 | SRCC = 0.9659 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1689 | LCC = 0.9650 | SRCC = 0.9657 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run3/config.gin b/results_tencent/w2v2_large_xlsr53/layer4/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run3/model.pt b/results_tencent/w2v2_large_xlsr53/layer4/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f779b6e9b2c8131bc62377143d8ad164a91e1372 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d5b3cb6212cd4d3429e864ed55e789ef94a70d39d9a882bc69ac89503904e64 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run3/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer4/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..abcc4f633ad368682a495fd171b7c50a68d3b49c --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b9794a08d357383fb91c14fc3fd499a27a0e53abdcb5c9ba32b0a8f07b4405f7 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run3/train.log b/results_tencent/w2v2_large_xlsr53/layer4/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f2bc0931925e798d103e3f3ee641ba0483a57473 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.47695670294761655 +INFO:root: +[Valid][0][UTT][ MSE = 0.2821 | LCC = 0.9269 | SRCC = 0.9278 ] +INFO:root:Average loss=0.20700747966766359 +INFO:root: +[Valid][1][UTT][ MSE = 0.2315 | LCC = 0.9430 | SRCC = 0.9447 ] +INFO:root:Average loss=0.16725861978530884 +INFO:root: +[Valid][2][UTT][ MSE = 0.2102 | LCC = 0.9469 | SRCC = 0.9512 ] +INFO:root:Average loss=0.14007894986867905 +INFO:root: +[Valid][3][UTT][ MSE = 0.2298 | LCC = 0.9545 | SRCC = 0.9558 ] +INFO:root:Average loss=0.1328069130182266 +INFO:root: +[Valid][4][UTT][ MSE = 0.2017 | LCC = 0.9567 | SRCC = 0.9576 ] +INFO:root:Average loss=0.11379933914542198 +INFO:root: +[Valid][5][UTT][ MSE = 0.2606 | LCC = 0.9562 | SRCC = 0.9575 ] +INFO:root:Average loss=0.10772227916121482 +INFO:root: +[Valid][6][UTT][ MSE = 0.2203 | LCC = 0.9594 | SRCC = 0.9601 ] +INFO:root:Average loss=0.1007605536878109 +INFO:root: +[Valid][7][UTT][ MSE = 0.1964 | LCC = 0.9615 | SRCC = 0.9618 ] +INFO:root:Average loss=0.09526341050863266 +INFO:root: +[Valid][8][UTT][ MSE = 0.2030 | LCC = 0.9600 | SRCC = 0.9604 ] +INFO:root:Average loss=0.08849230068922043 +INFO:root: +[Valid][9][UTT][ MSE = 0.2090 | LCC = 0.9636 | SRCC = 0.9638 ] +INFO:root:Average loss=0.08360091787576675 +INFO:root: +[Valid][10][UTT][ MSE = 0.1868 | LCC = 0.9633 | SRCC = 0.9638 ] +INFO:root:Average loss=0.0826688843369484 +INFO:root: +[Valid][11][UTT][ MSE = 0.2196 | LCC = 0.9646 | SRCC = 0.9647 ] +INFO:root:Average loss=0.07369363018870353 +INFO:root: +[Valid][12][UTT][ MSE = 0.1792 | LCC = 0.9629 | SRCC = 0.9627 ] +INFO:root:Average loss=0.074721346616745 +INFO:root: +[Valid][13][UTT][ MSE = 0.2315 | LCC = 0.9630 | SRCC = 0.9638 ] +INFO:root:Average loss=0.07580056321620941 +INFO:root: +[Valid][14][UTT][ MSE = 0.1699 | LCC = 0.9624 | SRCC = 0.9629 ] +INFO:root:Average loss=0.06749401897192002 +INFO:root: +[Valid][15][UTT][ MSE = 0.1786 | LCC = 0.9645 | SRCC = 0.9647 ] +INFO:root:Average loss=0.0638028718829155 +INFO:root: +[Valid][16][UTT][ MSE = 0.2211 | LCC = 0.9626 | SRCC = 0.9631 ] +INFO:root:Average loss=0.06094066305458546 +INFO:root: +[Valid][17][UTT][ MSE = 0.1932 | LCC = 0.9643 | SRCC = 0.9652 ] +INFO:root:Average loss=0.05788260370492935 +INFO:root: +[Valid][18][UTT][ MSE = 0.1471 | LCC = 0.9642 | SRCC = 0.9654 ] +INFO:root:Average loss=0.05457210826873779 +INFO:root: +[Valid][19][UTT][ MSE = 0.1873 | LCC = 0.9659 | SRCC = 0.9664 ] +INFO:root:Average loss=0.05459091331064701 +INFO:root: +[Valid][20][UTT][ MSE = 0.2115 | LCC = 0.9639 | SRCC = 0.9642 ] +INFO:root:Average loss=0.053487000063061715 +INFO:root: +[Valid][21][UTT][ MSE = 0.2043 | LCC = 0.9644 | SRCC = 0.9651 ] +INFO:root:Average loss=0.05190041346848011 +INFO:root: +[Valid][22][UTT][ MSE = 0.1885 | LCC = 0.9656 | SRCC = 0.9661 ] +INFO:root:Average loss=0.04827850514650345 +INFO:root: +[Valid][23][UTT][ MSE = 0.2146 | LCC = 0.9656 | SRCC = 0.9655 ] +INFO:root:Average loss=0.047177909761667254 +INFO:root: +[Valid][24][UTT][ MSE = 0.1810 | LCC = 0.9659 | SRCC = 0.9662 ] +INFO:root:Average loss=0.04455318540334702 +INFO:root: +[Valid][25][UTT][ MSE = 0.1834 | LCC = 0.9662 | SRCC = 0.9665 ] +INFO:root:Average loss=0.04286661225557327 +INFO:root: +[Valid][26][UTT][ MSE = 0.1836 | LCC = 0.9662 | SRCC = 0.9668 ] +INFO:root:Average loss=0.04267376346886158 +INFO:root: +[Valid][27][UTT][ MSE = 0.1719 | LCC = 0.9659 | SRCC = 0.9663 ] +INFO:root:Average loss=0.03875344653427601 +INFO:root: +[Valid][28][UTT][ MSE = 0.1743 | LCC = 0.9667 | SRCC = 0.9663 ] +INFO:root:Average loss=0.03904908874630928 +INFO:root: +[Valid][29][UTT][ MSE = 0.1853 | LCC = 0.9664 | SRCC = 0.9661 ] +INFO:root:Average loss=0.036112372040748596 +INFO:root: +[Valid][30][UTT][ MSE = 0.1619 | LCC = 0.9640 | SRCC = 0.9642 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1806 | LCC = 0.9650 | SRCC = 0.9656 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run4/config.gin b/results_tencent/w2v2_large_xlsr53/layer4/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run4/model.pt b/results_tencent/w2v2_large_xlsr53/layer4/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..d20c72cd6e490230d1f3fe8533357dc997770331 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1717d392288ec74cebd55c0da84ade1906a079d69288763252ac73709078e4e9 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run4/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer4/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..acb4cd065701ea15e76dd66feaf9f39f7e8c3fbc --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20bfcb63868b2fc2da02d4e8c3d8e7a01a3c5000caefc4739002407751c4974b +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer4/run4/train.log b/results_tencent/w2v2_large_xlsr53/layer4/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..03197f39fe975a63b571dd8b2f4acb787879cefa --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer4/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3982408106923103 +INFO:root: +[Valid][0][UTT][ MSE = 0.3264 | LCC = 0.9263 | SRCC = 0.9289 ] +INFO:root:Average loss=0.1928852699995041 +INFO:root: +[Valid][1][UTT][ MSE = 0.2702 | LCC = 0.9424 | SRCC = 0.9455 ] +INFO:root:Average loss=0.1744468975663185 +INFO:root: +[Valid][2][UTT][ MSE = 0.2913 | LCC = 0.9490 | SRCC = 0.9511 ] +INFO:root:Average loss=0.14301978439092636 +INFO:root: +[Valid][3][UTT][ MSE = 0.2936 | LCC = 0.9510 | SRCC = 0.9536 ] +INFO:root:Average loss=0.12937333709001542 +INFO:root: +[Valid][4][UTT][ MSE = 0.2704 | LCC = 0.9536 | SRCC = 0.9552 ] +INFO:root:Average loss=0.12182447254657745 +INFO:root: +[Valid][5][UTT][ MSE = 0.2150 | LCC = 0.9537 | SRCC = 0.9563 ] +INFO:root:Average loss=0.11138657346367836 +INFO:root: +[Valid][6][UTT][ MSE = 0.2550 | LCC = 0.9576 | SRCC = 0.9596 ] +INFO:root:Average loss=0.10513779002428054 +INFO:root: +[Valid][7][UTT][ MSE = 0.2192 | LCC = 0.9543 | SRCC = 0.9550 ] +INFO:root:Average loss=0.0983644790649414 +INFO:root: +[Valid][8][UTT][ MSE = 0.2249 | LCC = 0.9590 | SRCC = 0.9605 ] +INFO:root:Average loss=0.09568399375677109 +INFO:root: +[Valid][9][UTT][ MSE = 0.2210 | LCC = 0.9602 | SRCC = 0.9611 ] +INFO:root:Average loss=0.0867559569478035 +INFO:root: +[Valid][10][UTT][ MSE = 0.1945 | LCC = 0.9614 | SRCC = 0.9618 ] +INFO:root:Average loss=0.08660943707823754 +INFO:root: +[Valid][11][UTT][ MSE = 0.2048 | LCC = 0.9619 | SRCC = 0.9626 ] +INFO:root:Average loss=0.08444343811273575 +INFO:root: +[Valid][12][UTT][ MSE = 0.2219 | LCC = 0.9610 | SRCC = 0.9619 ] +INFO:root:Average loss=0.07574411821365357 +INFO:root: +[Valid][13][UTT][ MSE = 0.2368 | LCC = 0.9612 | SRCC = 0.9633 ] +INFO:root:Average loss=0.0745062845647335 +INFO:root: +[Valid][14][UTT][ MSE = 0.2349 | LCC = 0.9611 | SRCC = 0.9626 ] +INFO:root:Average loss=0.07296382549405098 +INFO:root: +[Valid][15][UTT][ MSE = 0.2039 | LCC = 0.9633 | SRCC = 0.9644 ] +INFO:root:Average loss=0.07199696093797683 +INFO:root: +[Valid][16][UTT][ MSE = 0.2101 | LCC = 0.9642 | SRCC = 0.9646 ] +INFO:root:Average loss=0.0650825515985489 +INFO:root: +[Valid][17][UTT][ MSE = 0.1927 | LCC = 0.9638 | SRCC = 0.9642 ] +INFO:root:Average loss=0.06093997931480408 +INFO:root: +[Valid][18][UTT][ MSE = 0.1591 | LCC = 0.9627 | SRCC = 0.9633 ] +INFO:root:Average loss=0.059088253408670424 +INFO:root: +[Valid][19][UTT][ MSE = 0.2024 | LCC = 0.9638 | SRCC = 0.9642 ] +INFO:root:Average loss=0.053958691775798796 +INFO:root: +[Valid][20][UTT][ MSE = 0.2010 | LCC = 0.9632 | SRCC = 0.9640 ] +INFO:root:Average loss=0.05621403443813324 +INFO:root: +[Valid][21][UTT][ MSE = 0.1954 | LCC = 0.9633 | SRCC = 0.9639 ] +INFO:root:Average loss=0.05193243576586246 +INFO:root: +[Valid][22][UTT][ MSE = 0.1774 | LCC = 0.9644 | SRCC = 0.9648 ] +INFO:root:Average loss=0.04836908069252968 +INFO:root: +[Valid][23][UTT][ MSE = 0.1919 | LCC = 0.9653 | SRCC = 0.9657 ] +INFO:root:Average loss=0.05011013589799404 +INFO:root: +[Valid][24][UTT][ MSE = 0.2121 | LCC = 0.9634 | SRCC = 0.9650 ] +INFO:root:Average loss=0.04928064042329788 +INFO:root: +[Valid][25][UTT][ MSE = 0.1688 | LCC = 0.9645 | SRCC = 0.9649 ] +INFO:root:Average loss=0.048081866145133975 +INFO:root: +[Valid][26][UTT][ MSE = 0.2014 | LCC = 0.9633 | SRCC = 0.9636 ] +INFO:root:Average loss=0.044603990986943245 +INFO:root: +[Valid][27][UTT][ MSE = 0.1833 | LCC = 0.9647 | SRCC = 0.9651 ] +INFO:root:Average loss=0.043158241152763364 +INFO:root: +[Valid][28][UTT][ MSE = 0.1810 | LCC = 0.9655 | SRCC = 0.9654 ] +INFO:root:Average loss=0.039629221752285955 +INFO:root: +[Valid][29][UTT][ MSE = 0.1730 | LCC = 0.9646 | SRCC = 0.9648 ] +INFO:root:Average loss=0.04030624684691429 +INFO:root: +[Valid][30][UTT][ MSE = 0.2179 | LCC = 0.9642 | SRCC = 0.9652 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1891 | LCC = 0.9642 | SRCC = 0.9646 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run0/config.gin b/results_tencent/w2v2_large_xlsr53/layer6/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run0/model.pt b/results_tencent/w2v2_large_xlsr53/layer6/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0bc4052121cdfed9432e780b62167c1bef6c3083 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1c7c483a2757c426059e1de8c1cddbef27232ecc54e94581c60dcb07abe6ab16 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run0/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer6/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..3a1aa567c5372aaaf614a4d633c6825f52514045 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:327092c4f60b0a073aab9932cd1c55d9f88542813c7c6abd51fb8108319c5390 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run0/train.log b/results_tencent/w2v2_large_xlsr53/layer6/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..77b839bb63fd2c5bd26633a760d5317d3202e9e1 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42776052057743075 +INFO:root: +[Valid][0][UTT][ MSE = 0.2806 | LCC = 0.9243 | SRCC = 0.9248 ] +INFO:root:Average loss=0.17520554256439208 +INFO:root: +[Valid][1][UTT][ MSE = 0.2965 | LCC = 0.9392 | SRCC = 0.9446 ] +INFO:root:Average loss=0.15078195208311082 +INFO:root: +[Valid][2][UTT][ MSE = 0.2761 | LCC = 0.9509 | SRCC = 0.9526 ] +INFO:root:Average loss=0.12841786479949951 +INFO:root: +[Valid][3][UTT][ MSE = 0.1964 | LCC = 0.9537 | SRCC = 0.9562 ] +INFO:root:Average loss=0.12218274343013763 +INFO:root: +[Valid][4][UTT][ MSE = 0.2441 | LCC = 0.9548 | SRCC = 0.9573 ] +INFO:root:Average loss=0.11183650892972946 +INFO:root: +[Valid][5][UTT][ MSE = 0.1935 | LCC = 0.9596 | SRCC = 0.9604 ] +INFO:root:Average loss=0.10197863167524338 +INFO:root: +[Valid][6][UTT][ MSE = 0.2179 | LCC = 0.9591 | SRCC = 0.9610 ] +INFO:root:Average loss=0.09534309288859368 +INFO:root: +[Valid][7][UTT][ MSE = 0.2183 | LCC = 0.9606 | SRCC = 0.9614 ] +INFO:root:Average loss=0.09343912133574486 +INFO:root: +[Valid][8][UTT][ MSE = 0.2029 | LCC = 0.9619 | SRCC = 0.9629 ] +INFO:root:Average loss=0.08587315428256989 +INFO:root: +[Valid][9][UTT][ MSE = 0.1969 | LCC = 0.9630 | SRCC = 0.9635 ] +INFO:root:Average loss=0.0804237893819809 +INFO:root: +[Valid][10][UTT][ MSE = 0.1913 | LCC = 0.9621 | SRCC = 0.9631 ] +INFO:root:Average loss=0.0756698507964611 +INFO:root: +[Valid][11][UTT][ MSE = 0.1958 | LCC = 0.9624 | SRCC = 0.9630 ] +INFO:root:Average loss=0.07580278377234936 +INFO:root: +[Valid][12][UTT][ MSE = 0.1985 | LCC = 0.9630 | SRCC = 0.9642 ] +INFO:root:Average loss=0.07042074438929558 +INFO:root: +[Valid][13][UTT][ MSE = 0.1756 | LCC = 0.9639 | SRCC = 0.9651 ] +INFO:root:Average loss=0.06966361472010613 +INFO:root: +[Valid][14][UTT][ MSE = 0.1715 | LCC = 0.9633 | SRCC = 0.9643 ] +INFO:root:Average loss=0.06504571914672852 +INFO:root: +[Valid][15][UTT][ MSE = 0.1583 | LCC = 0.9638 | SRCC = 0.9644 ] +INFO:root:Average loss=0.06229233887791634 +INFO:root: +[Valid][16][UTT][ MSE = 0.2184 | LCC = 0.9635 | SRCC = 0.9649 ] +INFO:root:Average loss=0.05887003144621849 +INFO:root: +[Valid][17][UTT][ MSE = 0.1901 | LCC = 0.9646 | SRCC = 0.9651 ] +INFO:root:Average loss=0.0543861271739006 +INFO:root: +[Valid][18][UTT][ MSE = 0.1826 | LCC = 0.9653 | SRCC = 0.9650 ] +INFO:root:Average loss=0.05487941654026508 +INFO:root: +[Valid][19][UTT][ MSE = 0.2073 | LCC = 0.9641 | SRCC = 0.9637 ] +INFO:root:Average loss=0.05016188417375088 +INFO:root: +[Valid][20][UTT][ MSE = 0.1804 | LCC = 0.9645 | SRCC = 0.9649 ] +INFO:root:Average loss=0.05240056692063808 +INFO:root: +[Valid][21][UTT][ MSE = 0.1946 | LCC = 0.9641 | SRCC = 0.9643 ] +INFO:root:Average loss=0.04676387184858322 +INFO:root: +[Valid][22][UTT][ MSE = 0.1941 | LCC = 0.9627 | SRCC = 0.9632 ] +INFO:root:Average loss=0.049169985488057136 +INFO:root: +[Valid][23][UTT][ MSE = 0.1691 | LCC = 0.9632 | SRCC = 0.9640 ] +INFO:root:Average loss=0.04454577276110649 +INFO:root: +[Valid][24][UTT][ MSE = 0.1813 | LCC = 0.9638 | SRCC = 0.9637 ] +INFO:root:Average loss=0.04364217191934586 +INFO:root: +[Valid][25][UTT][ MSE = 0.1858 | LCC = 0.9643 | SRCC = 0.9644 ] +INFO:root:Average loss=0.039874628603458404 +INFO:root: +[Valid][26][UTT][ MSE = 0.1870 | LCC = 0.9645 | SRCC = 0.9646 ] +INFO:root:Average loss=0.03822584682703018 +INFO:root: +[Valid][27][UTT][ MSE = 0.1963 | LCC = 0.9633 | SRCC = 0.9635 ] +INFO:root:Average loss=0.03777239926159382 +INFO:root: +[Valid][28][UTT][ MSE = 0.1721 | LCC = 0.9645 | SRCC = 0.9644 ] +INFO:root:Average loss=0.04352941931784153 +INFO:root: +[Valid][29][UTT][ MSE = 0.1515 | LCC = 0.9649 | SRCC = 0.9648 ] +INFO:root:Average loss=0.03595091424882412 +INFO:root: +[Valid][30][UTT][ MSE = 0.1797 | LCC = 0.9638 | SRCC = 0.9637 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1872 | LCC = 0.9632 | SRCC = 0.9642 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run1/config.gin b/results_tencent/w2v2_large_xlsr53/layer6/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run1/model.pt b/results_tencent/w2v2_large_xlsr53/layer6/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..9e6b72924a2c11f5bff029e5c2d34b1495e14e98 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e0346dce47ade659a2143500f0fe65cecff4ebde42667e021850b65335f141a6 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run1/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer6/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b2ab272daab6cfefdfdf587697b6b6a8ac72845b --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f01ba47c27521acd40f86843749b63b10ded2f8857370b6b3b87c4129741769e +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run1/train.log b/results_tencent/w2v2_large_xlsr53/layer6/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0d5ea572f636b152fa8c81d44f8b4fa3abbc42bd --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4267706727981567 +INFO:root: +[Valid][0][UTT][ MSE = 0.3104 | LCC = 0.9297 | SRCC = 0.9306 ] +INFO:root:Average loss=0.1709184085726738 +INFO:root: +[Valid][1][UTT][ MSE = 0.2772 | LCC = 0.9433 | SRCC = 0.9455 ] +INFO:root:Average loss=0.1431897661089897 +INFO:root: +[Valid][2][UTT][ MSE = 0.2713 | LCC = 0.9495 | SRCC = 0.9519 ] +INFO:root:Average loss=0.12896151000261308 +INFO:root: +[Valid][3][UTT][ MSE = 0.2602 | LCC = 0.9529 | SRCC = 0.9550 ] +INFO:root:Average loss=0.11669549617171288 +INFO:root: +[Valid][4][UTT][ MSE = 0.2320 | LCC = 0.9577 | SRCC = 0.9587 ] +INFO:root:Average loss=0.11042605537176133 +INFO:root: +[Valid][5][UTT][ MSE = 0.1865 | LCC = 0.9580 | SRCC = 0.9593 ] +INFO:root:Average loss=0.10144192028045654 +INFO:root: +[Valid][6][UTT][ MSE = 0.2256 | LCC = 0.9589 | SRCC = 0.9603 ] +INFO:root:Average loss=0.09852203515172005 +INFO:root: +[Valid][7][UTT][ MSE = 0.2085 | LCC = 0.9608 | SRCC = 0.9613 ] +INFO:root:Average loss=0.08847050401568413 +INFO:root: +[Valid][8][UTT][ MSE = 0.1995 | LCC = 0.9623 | SRCC = 0.9624 ] +INFO:root:Average loss=0.08921562042832375 +INFO:root: +[Valid][9][UTT][ MSE = 0.1929 | LCC = 0.9636 | SRCC = 0.9636 ] +INFO:root:Average loss=0.0799767706990242 +INFO:root: +[Valid][10][UTT][ MSE = 0.1632 | LCC = 0.9630 | SRCC = 0.9630 ] +INFO:root:Average loss=0.07503049165010452 +INFO:root: +[Valid][11][UTT][ MSE = 0.1887 | LCC = 0.9612 | SRCC = 0.9617 ] +INFO:root:Average loss=0.07359297719597817 +INFO:root: +[Valid][12][UTT][ MSE = 0.1650 | LCC = 0.9624 | SRCC = 0.9630 ] +INFO:root:Average loss=0.06899314054846764 +INFO:root: +[Valid][13][UTT][ MSE = 0.2196 | LCC = 0.9614 | SRCC = 0.9628 ] +INFO:root:Average loss=0.06778476017713547 +INFO:root: +[Valid][14][UTT][ MSE = 0.2179 | LCC = 0.9628 | SRCC = 0.9634 ] +INFO:root:Average loss=0.06323072311282157 +INFO:root: +[Valid][15][UTT][ MSE = 0.1888 | LCC = 0.9656 | SRCC = 0.9657 ] +INFO:root:Average loss=0.06237652909755707 +INFO:root: +[Valid][16][UTT][ MSE = 0.1767 | LCC = 0.9650 | SRCC = 0.9646 ] +INFO:root:Average loss=0.05830976384878159 +INFO:root: +[Valid][17][UTT][ MSE = 0.2236 | LCC = 0.9637 | SRCC = 0.9646 ] +INFO:root:Average loss=0.055063088238239286 +INFO:root: +[Valid][18][UTT][ MSE = 0.1809 | LCC = 0.9645 | SRCC = 0.9645 ] +INFO:root:Average loss=0.053022084549069405 +INFO:root: +[Valid][19][UTT][ MSE = 0.2074 | LCC = 0.9638 | SRCC = 0.9647 ] +INFO:root:Average loss=0.05071144354343414 +INFO:root: +[Valid][20][UTT][ MSE = 0.1979 | LCC = 0.9636 | SRCC = 0.9638 ] +INFO:root:Average loss=0.04858694152534008 +INFO:root: +[Valid][21][UTT][ MSE = 0.2128 | LCC = 0.9647 | SRCC = 0.9650 ] +INFO:root:Average loss=0.04658244314789772 +INFO:root: +[Valid][22][UTT][ MSE = 0.1991 | LCC = 0.9654 | SRCC = 0.9658 ] +INFO:root:Average loss=0.044456400364637376 +INFO:root: +[Valid][23][UTT][ MSE = 0.2141 | LCC = 0.9629 | SRCC = 0.9636 ] +INFO:root:Average loss=0.04906674003601074 +INFO:root: +[Valid][24][UTT][ MSE = 0.2270 | LCC = 0.9622 | SRCC = 0.9628 ] +INFO:root:Average loss=0.04257615876197815 +INFO:root: +[Valid][25][UTT][ MSE = 0.1701 | LCC = 0.9651 | SRCC = 0.9650 ] +INFO:root:Average loss=0.04239001849293709 +INFO:root: +[Valid][26][UTT][ MSE = 0.1957 | LCC = 0.9647 | SRCC = 0.9648 ] +INFO:root:Average loss=0.03931380258500576 +INFO:root: +[Valid][27][UTT][ MSE = 0.1898 | LCC = 0.9645 | SRCC = 0.9646 ] +INFO:root:Average loss=0.037255605205893515 +INFO:root: +[Valid][28][UTT][ MSE = 0.2229 | LCC = 0.9627 | SRCC = 0.9637 ] +INFO:root:Average loss=0.03813868074864149 +INFO:root: +[Valid][29][UTT][ MSE = 0.1859 | LCC = 0.9645 | SRCC = 0.9645 ] +INFO:root:Average loss=0.03548190349340439 +INFO:root: +[Valid][30][UTT][ MSE = 0.1865 | LCC = 0.9643 | SRCC = 0.9644 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1955 | LCC = 0.9634 | SRCC = 0.9642 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run2/config.gin b/results_tencent/w2v2_large_xlsr53/layer6/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run2/model.pt b/results_tencent/w2v2_large_xlsr53/layer6/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..abd9019fbdd6e2687f0cd70d3d118e7d067beb0a --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f08e5145254d95fa761bd5cdf0d80f776b773e5276be13c074299d4c3285fc71 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run2/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer6/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a83f39d60787d40ab1f2ac97dbfde6618081beab --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1d79065c370e04a4ace7e2f6d94dcc483e76f157357130237c72d7d72f2b7727 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run2/train.log b/results_tencent/w2v2_large_xlsr53/layer6/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..625a80260ac7ec28f26cce2d079b46a3d74dda0b --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.49720319187641143 +INFO:root: +[Valid][0][UTT][ MSE = 0.2912 | LCC = 0.9274 | SRCC = 0.9285 ] +INFO:root:Average loss=0.185856114923954 +INFO:root: +[Valid][1][UTT][ MSE = 0.3082 | LCC = 0.9382 | SRCC = 0.9410 ] +INFO:root:Average loss=0.16646740561723708 +INFO:root: +[Valid][2][UTT][ MSE = 0.2460 | LCC = 0.9464 | SRCC = 0.9487 ] +INFO:root:Average loss=0.13637796980142594 +INFO:root: +[Valid][3][UTT][ MSE = 0.2716 | LCC = 0.9512 | SRCC = 0.9526 ] +INFO:root:Average loss=0.1268985649049282 +INFO:root: +[Valid][4][UTT][ MSE = 0.2568 | LCC = 0.9535 | SRCC = 0.9557 ] +INFO:root:Average loss=0.11599129623174667 +INFO:root: +[Valid][5][UTT][ MSE = 0.2646 | LCC = 0.9561 | SRCC = 0.9573 ] +INFO:root:Average loss=0.10622168663144112 +INFO:root: +[Valid][6][UTT][ MSE = 0.2405 | LCC = 0.9557 | SRCC = 0.9586 ] +INFO:root:Average loss=0.09998069539666175 +INFO:root: +[Valid][7][UTT][ MSE = 0.2090 | LCC = 0.9559 | SRCC = 0.9588 ] +INFO:root:Average loss=0.09433874985575676 +INFO:root: +[Valid][8][UTT][ MSE = 0.1845 | LCC = 0.9610 | SRCC = 0.9613 ] +INFO:root:Average loss=0.0961557714343071 +INFO:root: +[Valid][9][UTT][ MSE = 0.2205 | LCC = 0.9617 | SRCC = 0.9628 ] +INFO:root:Average loss=0.08750418755412102 +INFO:root: +[Valid][10][UTT][ MSE = 0.1971 | LCC = 0.9616 | SRCC = 0.9632 ] +INFO:root:Average loss=0.0810914036333561 +INFO:root: +[Valid][11][UTT][ MSE = 0.2004 | LCC = 0.9634 | SRCC = 0.9636 ] +INFO:root:Average loss=0.07662776774168015 +INFO:root: +[Valid][12][UTT][ MSE = 0.1972 | LCC = 0.9632 | SRCC = 0.9637 ] +INFO:root:Average loss=0.07397984543442726 +INFO:root: +[Valid][13][UTT][ MSE = 0.2245 | LCC = 0.9634 | SRCC = 0.9638 ] +INFO:root:Average loss=0.0699522742331028 +INFO:root: +[Valid][14][UTT][ MSE = 0.1933 | LCC = 0.9628 | SRCC = 0.9631 ] +INFO:root:Average loss=0.06798369100689888 +INFO:root: +[Valid][15][UTT][ MSE = 0.1939 | LCC = 0.9641 | SRCC = 0.9647 ] +INFO:root:Average loss=0.06506896197795868 +INFO:root: +[Valid][16][UTT][ MSE = 0.1836 | LCC = 0.9640 | SRCC = 0.9644 ] +INFO:root:Average loss=0.0619571325480938 +INFO:root: +[Valid][17][UTT][ MSE = 0.1774 | LCC = 0.9641 | SRCC = 0.9646 ] +INFO:root:Average loss=0.0604584933668375 +INFO:root: +[Valid][18][UTT][ MSE = 0.2005 | LCC = 0.9654 | SRCC = 0.9655 ] +INFO:root:Average loss=0.05958259251713753 +INFO:root: +[Valid][19][UTT][ MSE = 0.1992 | LCC = 0.9653 | SRCC = 0.9656 ] +INFO:root:Average loss=0.054145484149456025 +INFO:root: +[Valid][20][UTT][ MSE = 0.2538 | LCC = 0.9634 | SRCC = 0.9654 ] +INFO:root:Average loss=0.052279577165842056 +INFO:root: +[Valid][21][UTT][ MSE = 0.1865 | LCC = 0.9652 | SRCC = 0.9654 ] +INFO:root:Average loss=0.05046444070339203 +INFO:root: +[Valid][22][UTT][ MSE = 0.1852 | LCC = 0.9646 | SRCC = 0.9646 ] +INFO:root:Average loss=0.049110445946455 +INFO:root: +[Valid][23][UTT][ MSE = 0.1966 | LCC = 0.9642 | SRCC = 0.9652 ] +INFO:root:Average loss=0.0478690530359745 +INFO:root: +[Valid][24][UTT][ MSE = 0.1878 | LCC = 0.9657 | SRCC = 0.9656 ] +INFO:root:Average loss=0.04726023036241531 +INFO:root: +[Valid][25][UTT][ MSE = 0.1871 | LCC = 0.9658 | SRCC = 0.9656 ] +INFO:root:Average loss=0.04236625500023365 +INFO:root: +[Valid][26][UTT][ MSE = 0.1828 | LCC = 0.9644 | SRCC = 0.9645 ] +INFO:root:Average loss=0.041374436974525454 +INFO:root: +[Valid][27][UTT][ MSE = 0.1958 | LCC = 0.9649 | SRCC = 0.9649 ] +INFO:root:Average loss=0.04127892982959747 +INFO:root: +[Valid][28][UTT][ MSE = 0.1822 | LCC = 0.9649 | SRCC = 0.9646 ] +INFO:root:Average loss=0.040924326121807096 +INFO:root: +[Valid][29][UTT][ MSE = 0.1913 | LCC = 0.9655 | SRCC = 0.9653 ] +INFO:root:Average loss=0.03871473725140095 +INFO:root: +[Valid][30][UTT][ MSE = 0.2041 | LCC = 0.9647 | SRCC = 0.9653 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1971 | LCC = 0.9651 | SRCC = 0.9658 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run3/config.gin b/results_tencent/w2v2_large_xlsr53/layer6/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run3/model.pt b/results_tencent/w2v2_large_xlsr53/layer6/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..df14beb3c4f5c6b2543d3dee7a04e9509f2fe9db --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4df2a24a7f1445ded66e26a7f8579b14fab50227c64b0410eb873f56420de6f9 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run3/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer6/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..519a928759860cff023d8edebb43a7b0cdf2a3c0 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b16828fd121771e7083b266647d4cd5276a922e671c19b46f4b8ab01f45cd1c3 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run3/train.log b/results_tencent/w2v2_large_xlsr53/layer6/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2b1b47c1901eff606e91655dff0c2cc13a3ddd63 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4425136593580246 +INFO:root: +[Valid][0][UTT][ MSE = 0.3102 | LCC = 0.9238 | SRCC = 0.9263 ] +INFO:root:Average loss=0.18654133629798889 +INFO:root: +[Valid][1][UTT][ MSE = 0.2599 | LCC = 0.9428 | SRCC = 0.9446 ] +INFO:root:Average loss=0.14808468407392503 +INFO:root: +[Valid][2][UTT][ MSE = 0.2771 | LCC = 0.9490 | SRCC = 0.9500 ] +INFO:root:Average loss=0.13751288211345672 +INFO:root: +[Valid][3][UTT][ MSE = 0.2574 | LCC = 0.9502 | SRCC = 0.9515 ] +INFO:root:Average loss=0.12287648248672485 +INFO:root: +[Valid][4][UTT][ MSE = 0.1929 | LCC = 0.9566 | SRCC = 0.9578 ] +INFO:root:Average loss=0.10967026659846306 +INFO:root: +[Valid][5][UTT][ MSE = 0.2403 | LCC = 0.9602 | SRCC = 0.9610 ] +INFO:root:Average loss=0.0970302363038063 +INFO:root: +[Valid][6][UTT][ MSE = 0.2280 | LCC = 0.9604 | SRCC = 0.9613 ] +INFO:root:Average loss=0.09707640075683593 +INFO:root: +[Valid][7][UTT][ MSE = 0.2142 | LCC = 0.9606 | SRCC = 0.9606 ] +INFO:root:Average loss=0.08810893365740777 +INFO:root: +[Valid][8][UTT][ MSE = 0.2288 | LCC = 0.9612 | SRCC = 0.9617 ] +INFO:root:Average loss=0.0828043391406536 +INFO:root: +[Valid][9][UTT][ MSE = 0.1703 | LCC = 0.9596 | SRCC = 0.9619 ] +INFO:root:Average loss=0.08108965674042702 +INFO:root: +[Valid][10][UTT][ MSE = 0.2230 | LCC = 0.9619 | SRCC = 0.9623 ] +INFO:root:Average loss=0.0763518531024456 +INFO:root: +[Valid][11][UTT][ MSE = 0.2081 | LCC = 0.9631 | SRCC = 0.9640 ] +INFO:root:Average loss=0.07276627737283707 +INFO:root: +[Valid][12][UTT][ MSE = 0.1713 | LCC = 0.9639 | SRCC = 0.9638 ] +INFO:root:Average loss=0.07072417622804641 +INFO:root: +[Valid][13][UTT][ MSE = 0.1861 | LCC = 0.9649 | SRCC = 0.9651 ] +INFO:root:Average loss=0.06943746185302735 +INFO:root: +[Valid][14][UTT][ MSE = 0.2282 | LCC = 0.9633 | SRCC = 0.9639 ] +INFO:root:Average loss=0.0627463212609291 +INFO:root: +[Valid][15][UTT][ MSE = 0.1763 | LCC = 0.9647 | SRCC = 0.9650 ] +INFO:root:Average loss=0.06201254561543465 +INFO:root: +[Valid][16][UTT][ MSE = 0.1975 | LCC = 0.9635 | SRCC = 0.9642 ] +INFO:root:Average loss=0.05920196644961834 +INFO:root: +[Valid][17][UTT][ MSE = 0.1763 | LCC = 0.9645 | SRCC = 0.9645 ] +INFO:root:Average loss=0.057572549030184744 +INFO:root: +[Valid][18][UTT][ MSE = 0.2367 | LCC = 0.9645 | SRCC = 0.9651 ] +INFO:root:Average loss=0.05368003126978874 +INFO:root: +[Valid][19][UTT][ MSE = 0.1987 | LCC = 0.9647 | SRCC = 0.9653 ] +INFO:root:Average loss=0.0508114432990551 +INFO:root: +[Valid][20][UTT][ MSE = 0.1835 | LCC = 0.9653 | SRCC = 0.9655 ] +INFO:root:Average loss=0.04863399624824524 +INFO:root: +[Valid][21][UTT][ MSE = 0.1855 | LCC = 0.9642 | SRCC = 0.9651 ] +INFO:root:Average loss=0.04804688355326653 +INFO:root: +[Valid][22][UTT][ MSE = 0.1682 | LCC = 0.9649 | SRCC = 0.9655 ] +INFO:root:Average loss=0.04533439223468304 +INFO:root: +[Valid][23][UTT][ MSE = 0.1851 | LCC = 0.9644 | SRCC = 0.9643 ] +INFO:root:Average loss=0.04499140465259552 +INFO:root: +[Valid][24][UTT][ MSE = 0.1718 | LCC = 0.9653 | SRCC = 0.9653 ] +INFO:root:Average loss=0.04158440087735653 +INFO:root: +[Valid][25][UTT][ MSE = 0.1790 | LCC = 0.9653 | SRCC = 0.9655 ] +INFO:root:Average loss=0.04079334926605225 +INFO:root: +[Valid][26][UTT][ MSE = 0.1829 | LCC = 0.9643 | SRCC = 0.9645 ] +INFO:root:Average loss=0.04046682858467102 +INFO:root: +[Valid][27][UTT][ MSE = 0.1903 | LCC = 0.9654 | SRCC = 0.9652 ] +INFO:root:Average loss=0.04213738361001015 +INFO:root: +[Valid][28][UTT][ MSE = 0.1829 | LCC = 0.9660 | SRCC = 0.9663 ] +INFO:root:Average loss=0.03922955347597599 +INFO:root: +[Valid][29][UTT][ MSE = 0.1987 | LCC = 0.9651 | SRCC = 0.9658 ] +INFO:root:Average loss=0.03511569559574127 +INFO:root: +[Valid][30][UTT][ MSE = 0.1987 | LCC = 0.9648 | SRCC = 0.9653 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1844 | LCC = 0.9646 | SRCC = 0.9656 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run4/config.gin b/results_tencent/w2v2_large_xlsr53/layer6/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run4/model.pt b/results_tencent/w2v2_large_xlsr53/layer6/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f7d84f7c6e6ed6c8df45d08613d6e972b2c5ba86 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:67c88479aec09be0eb1b174b3157c81fd92d5186a4fd941cfabdf1cdc0510df5 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run4/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer6/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2baf7ee6e256cd00f82b12ef2769c261543af670 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8b77cc1447bdcfb86f8985c44329c2d27101265312e357b382065c30ca4fc12d +size 3404998 diff --git a/results_tencent/w2v2_large_xlsr53/layer6/run4/train.log b/results_tencent/w2v2_large_xlsr53/layer6/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..804e05e24a8d90fc004d5f9a5df49f88658b9bd9 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer6/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4168706448674202 +INFO:root: +[Valid][0][UTT][ MSE = 0.3089 | LCC = 0.9239 | SRCC = 0.9258 ] +INFO:root:Average loss=0.18376308995485305 +INFO:root: +[Valid][1][UTT][ MSE = 0.2933 | LCC = 0.9430 | SRCC = 0.9430 ] +INFO:root:Average loss=0.1562918059825897 +INFO:root: +[Valid][2][UTT][ MSE = 0.2417 | LCC = 0.9510 | SRCC = 0.9513 ] +INFO:root:Average loss=0.13808454644680024 +INFO:root: +[Valid][3][UTT][ MSE = 0.2312 | LCC = 0.9504 | SRCC = 0.9520 ] +INFO:root:Average loss=0.12244148379564285 +INFO:root: +[Valid][4][UTT][ MSE = 0.2425 | LCC = 0.9547 | SRCC = 0.9555 ] +INFO:root:Average loss=0.11423700997233391 +INFO:root: +[Valid][5][UTT][ MSE = 0.2019 | LCC = 0.9572 | SRCC = 0.9577 ] +INFO:root:Average loss=0.10781419053673744 +INFO:root: +[Valid][6][UTT][ MSE = 0.2143 | LCC = 0.9581 | SRCC = 0.9583 ] +INFO:root:Average loss=0.10122658163309098 +INFO:root: +[Valid][7][UTT][ MSE = 0.2357 | LCC = 0.9618 | SRCC = 0.9618 ] +INFO:root:Average loss=0.0936762369275093 +INFO:root: +[Valid][8][UTT][ MSE = 0.2797 | LCC = 0.9592 | SRCC = 0.9594 ] +INFO:root:Average loss=0.09148166707158088 +INFO:root: +[Valid][9][UTT][ MSE = 0.2205 | LCC = 0.9624 | SRCC = 0.9628 ] +INFO:root:Average loss=0.08480357158184051 +INFO:root: +[Valid][10][UTT][ MSE = 0.2068 | LCC = 0.9633 | SRCC = 0.9638 ] +INFO:root:Average loss=0.07521351408958435 +INFO:root: +[Valid][11][UTT][ MSE = 0.2047 | LCC = 0.9640 | SRCC = 0.9641 ] +INFO:root:Average loss=0.07611689239740371 +INFO:root: +[Valid][12][UTT][ MSE = 0.1952 | LCC = 0.9635 | SRCC = 0.9640 ] +INFO:root:Average loss=0.07020832672715187 +INFO:root: +[Valid][13][UTT][ MSE = 0.2463 | LCC = 0.9633 | SRCC = 0.9636 ] +INFO:root:Average loss=0.06983798217773438 +INFO:root: +[Valid][14][UTT][ MSE = 0.2356 | LCC = 0.9612 | SRCC = 0.9624 ] +INFO:root:Average loss=0.06839938658475876 +INFO:root: +[Valid][15][UTT][ MSE = 0.1808 | LCC = 0.9655 | SRCC = 0.9655 ] +INFO:root:Average loss=0.06394668501615525 +INFO:root: +[Valid][16][UTT][ MSE = 0.1715 | LCC = 0.9639 | SRCC = 0.9647 ] +INFO:root:Average loss=0.05791815236210823 +INFO:root: +[Valid][17][UTT][ MSE = 0.1947 | LCC = 0.9619 | SRCC = 0.9631 ] +INFO:root:Average loss=0.057920559138059616 +INFO:root: +[Valid][18][UTT][ MSE = 0.1797 | LCC = 0.9663 | SRCC = 0.9660 ] +INFO:root:Average loss=0.05448682303726673 +INFO:root: +[Valid][19][UTT][ MSE = 0.2174 | LCC = 0.9655 | SRCC = 0.9662 ] +INFO:root:Average loss=0.053316744938492776 +INFO:root: +[Valid][20][UTT][ MSE = 0.2000 | LCC = 0.9658 | SRCC = 0.9658 ] +INFO:root:Average loss=0.05052268616855145 +INFO:root: +[Valid][21][UTT][ MSE = 0.1759 | LCC = 0.9660 | SRCC = 0.9655 ] +INFO:root:Average loss=0.04909460756182671 +INFO:root: +[Valid][22][UTT][ MSE = 0.1973 | LCC = 0.9642 | SRCC = 0.9639 ] +INFO:root:Average loss=0.04771869860589504 +INFO:root: +[Valid][23][UTT][ MSE = 0.1843 | LCC = 0.9639 | SRCC = 0.9647 ] +INFO:root:Average loss=0.044019034147262576 +INFO:root: +[Valid][24][UTT][ MSE = 0.1794 | LCC = 0.9643 | SRCC = 0.9647 ] +INFO:root:Average loss=0.04239489178359508 +INFO:root: +[Valid][25][UTT][ MSE = 0.1859 | LCC = 0.9658 | SRCC = 0.9657 ] +INFO:root:Average loss=0.043688511088490485 +INFO:root: +[Valid][26][UTT][ MSE = 0.2075 | LCC = 0.9663 | SRCC = 0.9658 ] +INFO:root:Average loss=0.04041205382347107 +INFO:root: +[Valid][27][UTT][ MSE = 0.2063 | LCC = 0.9646 | SRCC = 0.9646 ] +INFO:root:Average loss=0.03689995972812176 +INFO:root: +[Valid][28][UTT][ MSE = 0.1835 | LCC = 0.9658 | SRCC = 0.9654 ] +INFO:root:Average loss=0.03872578375041485 +INFO:root: +[Valid][29][UTT][ MSE = 0.1909 | LCC = 0.9666 | SRCC = 0.9660 ] +INFO:root:Average loss=0.035447721406817435 +INFO:root: +[Valid][30][UTT][ MSE = 0.1577 | LCC = 0.9653 | SRCC = 0.9653 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1910 | LCC = 0.9670 | SRCC = 0.9675 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run0/config.gin b/results_tencent/w2v2_large_xlsr53/layer8/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run0/model.pt b/results_tencent/w2v2_large_xlsr53/layer8/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7b7eb7e123ed172f7c23e3fd97df99e7364bd0eb --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f2584cc3ad8e409048c6b6c6b12adb06c7f5ec0dd17177930a7d8f0016fbaa7 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run0/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer8/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2846e6dd2152691ca65094465c2032cc20a21b84 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:380fb69aa289049c366aa83186e26553123793bb90af18574c0356764fc90d65 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run0/train.log b/results_tencent/w2v2_large_xlsr53/layer8/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..892f4d7d4f33176b0bca5e18401ed0cd3013adc5 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.47408699643611907 +INFO:root: +[Valid][0][UTT][ MSE = 0.3474 | LCC = 0.9237 | SRCC = 0.9246 ] +INFO:root:Average loss=0.19977533251047133 +INFO:root: +[Valid][1][UTT][ MSE = 0.2925 | LCC = 0.9426 | SRCC = 0.9431 ] +INFO:root:Average loss=0.1621724375486374 +INFO:root: +[Valid][2][UTT][ MSE = 0.2579 | LCC = 0.9484 | SRCC = 0.9509 ] +INFO:root:Average loss=0.14374615699052812 +INFO:root: +[Valid][3][UTT][ MSE = 0.2507 | LCC = 0.9520 | SRCC = 0.9540 ] +INFO:root:Average loss=0.12760443520545958 +INFO:root: +[Valid][4][UTT][ MSE = 0.2589 | LCC = 0.9540 | SRCC = 0.9551 ] +INFO:root:Average loss=0.11495004603266716 +INFO:root: +[Valid][5][UTT][ MSE = 0.2545 | LCC = 0.9582 | SRCC = 0.9600 ] +INFO:root:Average loss=0.10709780836105347 +INFO:root: +[Valid][6][UTT][ MSE = 0.2423 | LCC = 0.9586 | SRCC = 0.9611 ] +INFO:root:Average loss=0.10120798614621163 +INFO:root: +[Valid][7][UTT][ MSE = 0.1998 | LCC = 0.9605 | SRCC = 0.9613 ] +INFO:root:Average loss=0.09231431332230568 +INFO:root: +[Valid][8][UTT][ MSE = 0.2024 | LCC = 0.9617 | SRCC = 0.9617 ] +INFO:root:Average loss=0.08526336190104485 +INFO:root: +[Valid][9][UTT][ MSE = 0.1844 | LCC = 0.9620 | SRCC = 0.9631 ] +INFO:root:Average loss=0.08118265688419342 +INFO:root: +[Valid][10][UTT][ MSE = 0.2108 | LCC = 0.9607 | SRCC = 0.9617 ] +INFO:root:Average loss=0.07915295228362083 +INFO:root: +[Valid][11][UTT][ MSE = 0.1922 | LCC = 0.9634 | SRCC = 0.9638 ] +INFO:root:Average loss=0.07579377040266991 +INFO:root: +[Valid][12][UTT][ MSE = 0.2301 | LCC = 0.9615 | SRCC = 0.9630 ] +INFO:root:Average loss=0.06985641968250275 +INFO:root: +[Valid][13][UTT][ MSE = 0.2047 | LCC = 0.9635 | SRCC = 0.9637 ] +INFO:root:Average loss=0.06768869236111641 +INFO:root: +[Valid][14][UTT][ MSE = 0.1908 | LCC = 0.9641 | SRCC = 0.9644 ] +INFO:root:Average loss=0.06539874917268754 +INFO:root: +[Valid][15][UTT][ MSE = 0.1980 | LCC = 0.9638 | SRCC = 0.9637 ] +INFO:root:Average loss=0.062354588210582734 +INFO:root: +[Valid][16][UTT][ MSE = 0.1873 | LCC = 0.9627 | SRCC = 0.9633 ] +INFO:root:Average loss=0.060592157930135726 +INFO:root: +[Valid][17][UTT][ MSE = 0.1783 | LCC = 0.9640 | SRCC = 0.9640 ] +INFO:root:Average loss=0.05719876065850258 +INFO:root: +[Valid][18][UTT][ MSE = 0.1939 | LCC = 0.9634 | SRCC = 0.9628 ] +INFO:root:Average loss=0.0526318152397871 +INFO:root: +[Valid][19][UTT][ MSE = 0.1936 | LCC = 0.9640 | SRCC = 0.9640 ] +INFO:root:Average loss=0.05414509123563766 +INFO:root: +[Valid][20][UTT][ MSE = 0.1949 | LCC = 0.9631 | SRCC = 0.9635 ] +INFO:root:Average loss=0.04854014439880848 +INFO:root: +[Valid][21][UTT][ MSE = 0.1975 | LCC = 0.9632 | SRCC = 0.9641 ] +INFO:root:Average loss=0.04653800724446774 +INFO:root: +[Valid][22][UTT][ MSE = 0.1710 | LCC = 0.9606 | SRCC = 0.9616 ] +INFO:root:Average loss=0.04391863812506199 +INFO:root: +[Valid][23][UTT][ MSE = 0.2136 | LCC = 0.9635 | SRCC = 0.9642 ] +INFO:root:Average loss=0.04457526835799217 +INFO:root: +[Valid][24][UTT][ MSE = 0.1876 | LCC = 0.9646 | SRCC = 0.9644 ] +INFO:root:Average loss=0.04059094746410847 +INFO:root: +[Valid][25][UTT][ MSE = 0.2134 | LCC = 0.9639 | SRCC = 0.9638 ] +INFO:root:Average loss=0.04084326629340649 +INFO:root: +[Valid][26][UTT][ MSE = 0.1872 | LCC = 0.9647 | SRCC = 0.9648 ] +INFO:root:Average loss=0.03725433307886124 +INFO:root: +[Valid][27][UTT][ MSE = 0.2072 | LCC = 0.9628 | SRCC = 0.9627 ] +INFO:root:Average loss=0.037680369779467586 +INFO:root: +[Valid][28][UTT][ MSE = 0.1860 | LCC = 0.9645 | SRCC = 0.9642 ] +INFO:root:Average loss=0.03599543708562851 +INFO:root: +[Valid][29][UTT][ MSE = 0.1978 | LCC = 0.9642 | SRCC = 0.9641 ] +INFO:root:Average loss=0.03388662537932396 +INFO:root: +[Valid][30][UTT][ MSE = 0.1827 | LCC = 0.9649 | SRCC = 0.9648 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1886 | LCC = 0.9644 | SRCC = 0.9656 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run1/config.gin b/results_tencent/w2v2_large_xlsr53/layer8/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run1/model.pt b/results_tencent/w2v2_large_xlsr53/layer8/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6d04b8c97aabd3e79897eb9eb463003889cc1d58 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d228e307fc2de6e88742e8bb504ffa5cd84a89e13ff745ea6352e75ffb7bf09c +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run1/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer8/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..22112241a9aadcb61b6b2c2d949c43d38ecefa79 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e6168f5da65a4c9fb5ada8c9ae9abf7486f4f981287704e5e04cec112cc90bd5 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run1/train.log b/results_tencent/w2v2_large_xlsr53/layer8/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..78d1304c245e8e4cb0debc7f03f8675b6f77227e --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4237660726308823 +INFO:root: +[Valid][0][UTT][ MSE = 0.2903 | LCC = 0.9233 | SRCC = 0.9280 ] +INFO:root:Average loss=0.1777423866391182 +INFO:root: +[Valid][1][UTT][ MSE = 0.3264 | LCC = 0.9397 | SRCC = 0.9425 ] +INFO:root:Average loss=0.15116585302352906 +INFO:root: +[Valid][2][UTT][ MSE = 0.2553 | LCC = 0.9502 | SRCC = 0.9516 ] +INFO:root:Average loss=0.13092392468452455 +INFO:root: +[Valid][3][UTT][ MSE = 0.2520 | LCC = 0.9509 | SRCC = 0.9528 ] +INFO:root:Average loss=0.12133956062793731 +INFO:root: +[Valid][4][UTT][ MSE = 0.2275 | LCC = 0.9550 | SRCC = 0.9570 ] +INFO:root:Average loss=0.11526175796985626 +INFO:root: +[Valid][5][UTT][ MSE = 0.2709 | LCC = 0.9562 | SRCC = 0.9580 ] +INFO:root:Average loss=0.10646723568439484 +INFO:root: +[Valid][6][UTT][ MSE = 0.2609 | LCC = 0.9570 | SRCC = 0.9589 ] +INFO:root:Average loss=0.09845127621293068 +INFO:root: +[Valid][7][UTT][ MSE = 0.2712 | LCC = 0.9582 | SRCC = 0.9602 ] +INFO:root:Average loss=0.09279178911447525 +INFO:root: +[Valid][8][UTT][ MSE = 0.2366 | LCC = 0.9589 | SRCC = 0.9610 ] +INFO:root:Average loss=0.08856643265485764 +INFO:root: +[Valid][9][UTT][ MSE = 0.1971 | LCC = 0.9598 | SRCC = 0.9607 ] +INFO:root:Average loss=0.08076558065414428 +INFO:root: +[Valid][10][UTT][ MSE = 0.1991 | LCC = 0.9631 | SRCC = 0.9639 ] +INFO:root:Average loss=0.07971940648555756 +INFO:root: +[Valid][11][UTT][ MSE = 0.1995 | LCC = 0.9618 | SRCC = 0.9636 ] +INFO:root:Average loss=0.07630566036701203 +INFO:root: +[Valid][12][UTT][ MSE = 0.1948 | LCC = 0.9625 | SRCC = 0.9638 ] +INFO:root:Average loss=0.07431469896435737 +INFO:root: +[Valid][13][UTT][ MSE = 0.1990 | LCC = 0.9615 | SRCC = 0.9618 ] +INFO:root:Average loss=0.07185326600074768 +INFO:root: +[Valid][14][UTT][ MSE = 0.2183 | LCC = 0.9630 | SRCC = 0.9635 ] +INFO:root:Average loss=0.0656389016211033 +INFO:root: +[Valid][15][UTT][ MSE = 0.1941 | LCC = 0.9627 | SRCC = 0.9636 ] +INFO:root:Average loss=0.06639042416214944 +INFO:root: +[Valid][16][UTT][ MSE = 0.2538 | LCC = 0.9627 | SRCC = 0.9643 ] +INFO:root:Average loss=0.060580303281545636 +INFO:root: +[Valid][17][UTT][ MSE = 0.1829 | LCC = 0.9628 | SRCC = 0.9636 ] +INFO:root:Average loss=0.0593016839325428 +INFO:root: +[Valid][18][UTT][ MSE = 0.2052 | LCC = 0.9634 | SRCC = 0.9636 ] +INFO:root:Average loss=0.0568260447382927 +INFO:root: +[Valid][19][UTT][ MSE = 0.2064 | LCC = 0.9632 | SRCC = 0.9639 ] +INFO:root:Average loss=0.05165979401767254 +INFO:root: +[Valid][20][UTT][ MSE = 0.1744 | LCC = 0.9632 | SRCC = 0.9638 ] +INFO:root:Average loss=0.05498641219735145 +INFO:root: +[Valid][21][UTT][ MSE = 0.2062 | LCC = 0.9643 | SRCC = 0.9642 ] +INFO:root:Average loss=0.04841242308914662 +INFO:root: +[Valid][22][UTT][ MSE = 0.2241 | LCC = 0.9622 | SRCC = 0.9633 ] +INFO:root:Average loss=0.04619619938731193 +INFO:root: +[Valid][23][UTT][ MSE = 0.1687 | LCC = 0.9641 | SRCC = 0.9643 ] +INFO:root:Average loss=0.04643401364982128 +INFO:root: +[Valid][24][UTT][ MSE = 0.2111 | LCC = 0.9611 | SRCC = 0.9625 ] +INFO:root:Average loss=0.04288865815103054 +INFO:root: +[Valid][25][UTT][ MSE = 0.1734 | LCC = 0.9641 | SRCC = 0.9645 ] +INFO:root:Average loss=0.042465521886944774 +INFO:root: +[Valid][26][UTT][ MSE = 0.2087 | LCC = 0.9636 | SRCC = 0.9645 ] +INFO:root:Average loss=0.04161148124933243 +INFO:root: +[Valid][27][UTT][ MSE = 0.2045 | LCC = 0.9639 | SRCC = 0.9645 ] +INFO:root:Average loss=0.042161126688122746 +INFO:root: +[Valid][28][UTT][ MSE = 0.2003 | LCC = 0.9639 | SRCC = 0.9643 ] +INFO:root:Average loss=0.038738575667142866 +INFO:root: +[Valid][29][UTT][ MSE = 0.1789 | LCC = 0.9615 | SRCC = 0.9623 ] +INFO:root:Average loss=0.03896158173680306 +INFO:root: +[Valid][30][UTT][ MSE = 0.1815 | LCC = 0.9637 | SRCC = 0.9637 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2107 | LCC = 0.9628 | SRCC = 0.9639 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run2/config.gin b/results_tencent/w2v2_large_xlsr53/layer8/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run2/model.pt b/results_tencent/w2v2_large_xlsr53/layer8/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e2df2ca4c0183d43485b888e2f661298f8f19cfd --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:68b7d7df58de3b61e65833d7b75f5eeb3139a9c9a553f351e5afb634d46520fd +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run2/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer8/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..586a6e7cdfb4a23243dd0363d17f7af05ef22dbb --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:110af7818153d662516b77d2bd332725749725d9938dd4e1b1239e375e2c552b +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run2/train.log b/results_tencent/w2v2_large_xlsr53/layer8/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..85dfd74547b94d945b9fc73e44ca0e55d3de59cc --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42698113656044007 +INFO:root: +[Valid][0][UTT][ MSE = 0.2812 | LCC = 0.9261 | SRCC = 0.9266 ] +INFO:root:Average loss=0.1933620324730873 +INFO:root: +[Valid][1][UTT][ MSE = 0.2454 | LCC = 0.9435 | SRCC = 0.9443 ] +INFO:root:Average loss=0.15762415808439253 +INFO:root: +[Valid][2][UTT][ MSE = 0.2696 | LCC = 0.9454 | SRCC = 0.9478 ] +INFO:root:Average loss=0.14015505015850066 +INFO:root: +[Valid][3][UTT][ MSE = 0.3091 | LCC = 0.9457 | SRCC = 0.9481 ] +INFO:root:Average loss=0.12960160011053085 +INFO:root: +[Valid][4][UTT][ MSE = 0.2098 | LCC = 0.9558 | SRCC = 0.9567 ] +INFO:root:Average loss=0.11226392528414726 +INFO:root: +[Valid][5][UTT][ MSE = 0.2151 | LCC = 0.9546 | SRCC = 0.9557 ] +INFO:root:Average loss=0.10195391857624055 +INFO:root: +[Valid][6][UTT][ MSE = 0.1896 | LCC = 0.9600 | SRCC = 0.9603 ] +INFO:root:Average loss=0.099230936139822 +INFO:root: +[Valid][7][UTT][ MSE = 0.2099 | LCC = 0.9572 | SRCC = 0.9592 ] +INFO:root:Average loss=0.08953744491934776 +INFO:root: +[Valid][8][UTT][ MSE = 0.2413 | LCC = 0.9595 | SRCC = 0.9602 ] +INFO:root:Average loss=0.08777485519647599 +INFO:root: +[Valid][9][UTT][ MSE = 0.2078 | LCC = 0.9603 | SRCC = 0.9604 ] +INFO:root:Average loss=0.08400351291894913 +INFO:root: +[Valid][10][UTT][ MSE = 0.2115 | LCC = 0.9619 | SRCC = 0.9621 ] +INFO:root:Average loss=0.0788278304040432 +INFO:root: +[Valid][11][UTT][ MSE = 0.1896 | LCC = 0.9634 | SRCC = 0.9633 ] +INFO:root:Average loss=0.07265683379769325 +INFO:root: +[Valid][12][UTT][ MSE = 0.1823 | LCC = 0.9619 | SRCC = 0.9622 ] +INFO:root:Average loss=0.07070838955044746 +INFO:root: +[Valid][13][UTT][ MSE = 0.2060 | LCC = 0.9630 | SRCC = 0.9633 ] +INFO:root:Average loss=0.06809229111671447 +INFO:root: +[Valid][14][UTT][ MSE = 0.1825 | LCC = 0.9636 | SRCC = 0.9634 ] +INFO:root:Average loss=0.06319330057501793 +INFO:root: +[Valid][15][UTT][ MSE = 0.2043 | LCC = 0.9602 | SRCC = 0.9600 ] +INFO:root:Average loss=0.0648521880954504 +INFO:root: +[Valid][16][UTT][ MSE = 0.1789 | LCC = 0.9636 | SRCC = 0.9633 ] +INFO:root:Average loss=0.05780076299607754 +INFO:root: +[Valid][17][UTT][ MSE = 0.1863 | LCC = 0.9649 | SRCC = 0.9645 ] +INFO:root:Average loss=0.0561670481711626 +INFO:root: +[Valid][18][UTT][ MSE = 0.1921 | LCC = 0.9650 | SRCC = 0.9645 ] +INFO:root:Average loss=0.05501030370593071 +INFO:root: +[Valid][19][UTT][ MSE = 0.1952 | LCC = 0.9630 | SRCC = 0.9628 ] +INFO:root:Average loss=0.050936109632253646 +INFO:root: +[Valid][20][UTT][ MSE = 0.1875 | LCC = 0.9617 | SRCC = 0.9616 ] +INFO:root:Average loss=0.048893599584698674 +INFO:root: +[Valid][21][UTT][ MSE = 0.2036 | LCC = 0.9636 | SRCC = 0.9625 ] +INFO:root:Average loss=0.04539628174901009 +INFO:root: +[Valid][22][UTT][ MSE = 0.2039 | LCC = 0.9625 | SRCC = 0.9624 ] +INFO:root:Average loss=0.04520884445309639 +INFO:root: +[Valid][23][UTT][ MSE = 0.1869 | LCC = 0.9627 | SRCC = 0.9627 ] +INFO:root:Average loss=0.04367102500796318 +INFO:root: +[Valid][24][UTT][ MSE = 0.1968 | LCC = 0.9640 | SRCC = 0.9635 ] +INFO:root:Average loss=0.04084760320186615 +INFO:root: +[Valid][25][UTT][ MSE = 0.1849 | LCC = 0.9635 | SRCC = 0.9634 ] +INFO:root:Average loss=0.04076986637711525 +INFO:root: +[Valid][26][UTT][ MSE = 0.1963 | LCC = 0.9647 | SRCC = 0.9642 ] +INFO:root:Average loss=0.03839383034408093 +INFO:root: +[Valid][27][UTT][ MSE = 0.1895 | LCC = 0.9641 | SRCC = 0.9636 ] +INFO:root:Average loss=0.03824831096827984 +INFO:root: +[Valid][28][UTT][ MSE = 0.1706 | LCC = 0.9652 | SRCC = 0.9649 ] +INFO:root:Average loss=0.0350327690243721 +INFO:root: +[Valid][29][UTT][ MSE = 0.1757 | LCC = 0.9636 | SRCC = 0.9632 ] +INFO:root:Average loss=0.03808481013774872 +INFO:root: +[Valid][30][UTT][ MSE = 0.1704 | LCC = 0.9648 | SRCC = 0.9644 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1730 | LCC = 0.9645 | SRCC = 0.9656 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run3/config.gin b/results_tencent/w2v2_large_xlsr53/layer8/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run3/model.pt b/results_tencent/w2v2_large_xlsr53/layer8/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a281cb6f3b5cfecd0a88cecbdc627873884ab18f --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:276ac57c38a485189203308f19b1257d2f57421b7a1ed072a95f2e995357dd55 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run3/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer8/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e3237b4bf4ef21e74250793ea13ccca733113b44 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:10aa5e03e5181b60084a336cbe7f4c4be331dc26b89ada6dab76c4db3ad3935c +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run3/train.log b/results_tencent/w2v2_large_xlsr53/layer8/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..23b3ba4a9a43a2eba280588f431ad050b9322de4 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.44926624310016633 +INFO:root: +[Valid][0][UTT][ MSE = 0.3859 | LCC = 0.9229 | SRCC = 0.9260 ] +INFO:root:Average loss=0.18402774846553802 +INFO:root: +[Valid][1][UTT][ MSE = 0.3036 | LCC = 0.9387 | SRCC = 0.9419 ] +INFO:root:Average loss=0.15069237017631532 +INFO:root: +[Valid][2][UTT][ MSE = 0.2815 | LCC = 0.9419 | SRCC = 0.9481 ] +INFO:root:Average loss=0.13259655213356017 +INFO:root: +[Valid][3][UTT][ MSE = 0.2244 | LCC = 0.9531 | SRCC = 0.9536 ] +INFO:root:Average loss=0.11909201163053512 +INFO:root: +[Valid][4][UTT][ MSE = 0.2188 | LCC = 0.9564 | SRCC = 0.9571 ] +INFO:root:Average loss=0.11136495423316956 +INFO:root: +[Valid][5][UTT][ MSE = 0.2154 | LCC = 0.9562 | SRCC = 0.9577 ] +INFO:root:Average loss=0.10479364022612572 +INFO:root: +[Valid][6][UTT][ MSE = 0.2001 | LCC = 0.9569 | SRCC = 0.9592 ] +INFO:root:Average loss=0.10000358900427818 +INFO:root: +[Valid][7][UTT][ MSE = 0.1987 | LCC = 0.9602 | SRCC = 0.9606 ] +INFO:root:Average loss=0.09203364977240562 +INFO:root: +[Valid][8][UTT][ MSE = 0.2284 | LCC = 0.9596 | SRCC = 0.9598 ] +INFO:root:Average loss=0.08924224472045898 +INFO:root: +[Valid][9][UTT][ MSE = 0.1923 | LCC = 0.9600 | SRCC = 0.9622 ] +INFO:root:Average loss=0.08539849793910981 +INFO:root: +[Valid][10][UTT][ MSE = 0.1884 | LCC = 0.9624 | SRCC = 0.9628 ] +INFO:root:Average loss=0.07877899104356766 +INFO:root: +[Valid][11][UTT][ MSE = 0.1873 | LCC = 0.9610 | SRCC = 0.9619 ] +INFO:root:Average loss=0.07496151575446129 +INFO:root: +[Valid][12][UTT][ MSE = 0.1592 | LCC = 0.9630 | SRCC = 0.9641 ] +INFO:root:Average loss=0.07107259657979012 +INFO:root: +[Valid][13][UTT][ MSE = 0.2004 | LCC = 0.9637 | SRCC = 0.9641 ] +INFO:root:Average loss=0.06975911980867386 +INFO:root: +[Valid][14][UTT][ MSE = 0.1789 | LCC = 0.9621 | SRCC = 0.9629 ] +INFO:root:Average loss=0.06290123756229878 +INFO:root: +[Valid][15][UTT][ MSE = 0.1542 | LCC = 0.9630 | SRCC = 0.9640 ] +INFO:root:Average loss=0.06186127465963363 +INFO:root: +[Valid][16][UTT][ MSE = 0.2121 | LCC = 0.9633 | SRCC = 0.9637 ] +INFO:root:Average loss=0.05965369728207588 +INFO:root: +[Valid][17][UTT][ MSE = 0.1834 | LCC = 0.9637 | SRCC = 0.9638 ] +INFO:root:Average loss=0.05784074349701405 +INFO:root: +[Valid][18][UTT][ MSE = 0.2131 | LCC = 0.9647 | SRCC = 0.9645 ] +INFO:root:Average loss=0.05726115719974041 +INFO:root: +[Valid][19][UTT][ MSE = 0.1880 | LCC = 0.9653 | SRCC = 0.9651 ] +INFO:root:Average loss=0.05260918296873569 +INFO:root: +[Valid][20][UTT][ MSE = 0.1723 | LCC = 0.9650 | SRCC = 0.9647 ] +INFO:root:Average loss=0.05007612366974354 +INFO:root: +[Valid][21][UTT][ MSE = 0.1954 | LCC = 0.9637 | SRCC = 0.9642 ] +INFO:root:Average loss=0.04789014318585396 +INFO:root: +[Valid][22][UTT][ MSE = 0.1726 | LCC = 0.9655 | SRCC = 0.9654 ] +INFO:root:Average loss=0.049636610612273216 +INFO:root: +[Valid][23][UTT][ MSE = 0.1896 | LCC = 0.9644 | SRCC = 0.9645 ] +INFO:root:Average loss=0.04457734052836895 +INFO:root: +[Valid][24][UTT][ MSE = 0.1857 | LCC = 0.9648 | SRCC = 0.9647 ] +INFO:root:Average loss=0.04355504608154297 +INFO:root: +[Valid][25][UTT][ MSE = 0.1902 | LCC = 0.9649 | SRCC = 0.9647 ] +INFO:root:Average loss=0.041129905447363856 +INFO:root: +[Valid][26][UTT][ MSE = 0.2051 | LCC = 0.9646 | SRCC = 0.9644 ] +INFO:root:Average loss=0.039434086576104165 +INFO:root: +[Valid][27][UTT][ MSE = 0.1684 | LCC = 0.9649 | SRCC = 0.9646 ] +INFO:root:Average loss=0.03998909735679627 +INFO:root: +[Valid][28][UTT][ MSE = 0.1872 | LCC = 0.9635 | SRCC = 0.9639 ] +INFO:root:Average loss=0.03745378929376602 +INFO:root: +[Valid][29][UTT][ MSE = 0.2013 | LCC = 0.9643 | SRCC = 0.9645 ] +INFO:root:Average loss=0.03771223025023937 +INFO:root: +[Valid][30][UTT][ MSE = 0.1928 | LCC = 0.9646 | SRCC = 0.9654 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1758 | LCC = 0.9651 | SRCC = 0.9660 ] diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run4/config.gin b/results_tencent/w2v2_large_xlsr53/layer8/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run4/model.pt b/results_tencent/w2v2_large_xlsr53/layer8/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bdf7ea97187cf5dbaa9197ff26134edb28ad65d6 --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c8739d46b8678245ffe34b3f098a210fc0f45df93aaa0b09a97434d3d379fad7 +size 3403339 diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run4/model_best.pt b/results_tencent/w2v2_large_xlsr53/layer8/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..cddeab5fb24579e44503cd4461ffecded0e5ff4b --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:47fbb993371553b3d641c8a4b53f772904bc1073e39ddd44c91f625b5a4e7cf8 +size 3405062 diff --git a/results_tencent/w2v2_large_xlsr53/layer8/run4/train.log b/results_tencent/w2v2_large_xlsr53/layer8/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3139576af233449b5a113ee7ba65f3c0083c3f3e --- /dev/null +++ b/results_tencent/w2v2_large_xlsr53/layer8/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4532853927612305 +INFO:root: +[Valid][0][UTT][ MSE = 0.3082 | LCC = 0.9211 | SRCC = 0.9209 ] +INFO:root:Average loss=0.20808307564258574 +INFO:root: +[Valid][1][UTT][ MSE = 0.2438 | LCC = 0.9410 | SRCC = 0.9416 ] +INFO:root:Average loss=0.168660764336586 +INFO:root: +[Valid][2][UTT][ MSE = 0.2308 | LCC = 0.9492 | SRCC = 0.9495 ] +INFO:root:Average loss=0.13941887682676316 +INFO:root: +[Valid][3][UTT][ MSE = 0.2604 | LCC = 0.9516 | SRCC = 0.9534 ] +INFO:root:Average loss=0.1264129865169525 +INFO:root: +[Valid][4][UTT][ MSE = 0.2248 | LCC = 0.9549 | SRCC = 0.9568 ] +INFO:root:Average loss=0.11134307801723481 +INFO:root: +[Valid][5][UTT][ MSE = 0.2250 | LCC = 0.9594 | SRCC = 0.9599 ] +INFO:root:Average loss=0.11241292241215706 +INFO:root: +[Valid][6][UTT][ MSE = 0.2367 | LCC = 0.9585 | SRCC = 0.9598 ] +INFO:root:Average loss=0.10293239718675613 +INFO:root: +[Valid][7][UTT][ MSE = 0.2005 | LCC = 0.9610 | SRCC = 0.9623 ] +INFO:root:Average loss=0.09783596616983414 +INFO:root: +[Valid][8][UTT][ MSE = 0.1716 | LCC = 0.9608 | SRCC = 0.9616 ] +INFO:root:Average loss=0.08903774407505989 +INFO:root: +[Valid][9][UTT][ MSE = 0.2081 | LCC = 0.9612 | SRCC = 0.9627 ] +INFO:root:Average loss=0.08248636743426323 +INFO:root: +[Valid][10][UTT][ MSE = 0.1906 | LCC = 0.9627 | SRCC = 0.9633 ] +INFO:root:Average loss=0.08118875631690026 +INFO:root: +[Valid][11][UTT][ MSE = 0.2157 | LCC = 0.9617 | SRCC = 0.9633 ] +INFO:root:Average loss=0.07581612464785575 +INFO:root: +[Valid][12][UTT][ MSE = 0.2113 | LCC = 0.9627 | SRCC = 0.9632 ] +INFO:root:Average loss=0.07358697259426117 +INFO:root: +[Valid][13][UTT][ MSE = 0.2086 | LCC = 0.9602 | SRCC = 0.9620 ] +INFO:root:Average loss=0.0716957856118679 +INFO:root: +[Valid][14][UTT][ MSE = 0.2171 | LCC = 0.9643 | SRCC = 0.9647 ] +INFO:root:Average loss=0.06504642188549042 +INFO:root: +[Valid][15][UTT][ MSE = 0.1864 | LCC = 0.9639 | SRCC = 0.9637 ] +INFO:root:Average loss=0.06623587653040885 +INFO:root: +[Valid][16][UTT][ MSE = 0.2239 | LCC = 0.9612 | SRCC = 0.9644 ] +INFO:root:Average loss=0.06200912757217884 +INFO:root: +[Valid][17][UTT][ MSE = 0.1895 | LCC = 0.9636 | SRCC = 0.9640 ] +INFO:root:Average loss=0.056418537080287934 +INFO:root: +[Valid][18][UTT][ MSE = 0.1758 | LCC = 0.9642 | SRCC = 0.9645 ] +INFO:root:Average loss=0.054264900237321854 +INFO:root: +[Valid][19][UTT][ MSE = 0.2062 | LCC = 0.9640 | SRCC = 0.9648 ] +INFO:root:Average loss=0.05440304237604141 +INFO:root: +[Valid][20][UTT][ MSE = 0.1717 | LCC = 0.9644 | SRCC = 0.9646 ] +INFO:root:Average loss=0.04995433093607426 +INFO:root: +[Valid][21][UTT][ MSE = 0.1953 | LCC = 0.9642 | SRCC = 0.9644 ] +INFO:root:Average loss=0.049628633931279184 +INFO:root: +[Valid][22][UTT][ MSE = 0.1762 | LCC = 0.9643 | SRCC = 0.9641 ] +INFO:root:Average loss=0.0486623710244894 +INFO:root: +[Valid][23][UTT][ MSE = 0.1807 | LCC = 0.9635 | SRCC = 0.9634 ] +INFO:root:Average loss=0.04289362668991089 +INFO:root: +[Valid][24][UTT][ MSE = 0.1885 | LCC = 0.9618 | SRCC = 0.9633 ] +INFO:root:Average loss=0.04503890559077263 +INFO:root: +[Valid][25][UTT][ MSE = 0.1759 | LCC = 0.9648 | SRCC = 0.9651 ] +INFO:root:Average loss=0.04018752071261406 +INFO:root: +[Valid][26][UTT][ MSE = 0.1669 | LCC = 0.9645 | SRCC = 0.9643 ] +INFO:root:Average loss=0.03996180734038353 +INFO:root: +[Valid][27][UTT][ MSE = 0.2006 | LCC = 0.9615 | SRCC = 0.9635 ] +INFO:root:Average loss=0.03936391177773475 +INFO:root: +[Valid][28][UTT][ MSE = 0.1798 | LCC = 0.9647 | SRCC = 0.9645 ] +INFO:root:Average loss=0.0379727758616209 +INFO:root: +[Valid][29][UTT][ MSE = 0.1607 | LCC = 0.9647 | SRCC = 0.9647 ] +INFO:root:Average loss=0.035420768022537234 +INFO:root: +[Valid][30][UTT][ MSE = 0.2040 | LCC = 0.9642 | SRCC = 0.9642 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1767 | LCC = 0.9637 | SRCC = 0.9653 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run0/config.gin b/results_tencent/w2v2_xlsr_1b/layer0/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run0/model.pt b/results_tencent/w2v2_xlsr_1b/layer0/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..db9964cf2b50a038a11b25fed367021d0a3eb47d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a240f01d14a94d0356f411f37b4372f962560b8963c30253c04443ccd0303cc5 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run0/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer0/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d55579937ff295805c3d7e61d0ffdd8da5e4e742 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8786798c88c78b3ddb154e3b163780d725124e992689b4107b342a63ac6a0764 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run0/train.log b/results_tencent/w2v2_xlsr_1b/layer0/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..634dc4249271e82e22c6df993b99fc701ef50bfb --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5977000353336335 +INFO:root: +[Valid][0][UTT][ MSE = 0.5587 | LCC = 0.8421 | SRCC = 0.8532 ] +INFO:root:Average loss=0.320256737947464 +INFO:root: +[Valid][1][UTT][ MSE = 0.3173 | LCC = 0.8991 | SRCC = 0.8992 ] +INFO:root:Average loss=0.26743631637096404 +INFO:root: +[Valid][2][UTT][ MSE = 0.3312 | LCC = 0.9038 | SRCC = 0.9051 ] +INFO:root:Average loss=0.23415940022468568 +INFO:root: +[Valid][3][UTT][ MSE = 0.3302 | LCC = 0.9112 | SRCC = 0.9135 ] +INFO:root:Average loss=0.21453991138935088 +INFO:root: +[Valid][4][UTT][ MSE = 0.2918 | LCC = 0.9208 | SRCC = 0.9223 ] +INFO:root:Average loss=0.20557955676317216 +INFO:root: +[Valid][5][UTT][ MSE = 0.3077 | LCC = 0.9089 | SRCC = 0.9118 ] +INFO:root:Average loss=0.19194276773929597 +INFO:root: +[Valid][6][UTT][ MSE = 0.2691 | LCC = 0.9227 | SRCC = 0.9237 ] +INFO:root:Average loss=0.1802665142416954 +INFO:root: +[Valid][7][UTT][ MSE = 0.3244 | LCC = 0.9075 | SRCC = 0.9069 ] +INFO:root:Average loss=0.17032058668136596 +INFO:root: +[Valid][8][UTT][ MSE = 0.2758 | LCC = 0.9255 | SRCC = 0.9258 ] +INFO:root:Average loss=0.16550731110572814 +INFO:root: +[Valid][9][UTT][ MSE = 0.2502 | LCC = 0.9299 | SRCC = 0.9304 ] +INFO:root:Average loss=0.15510898959636688 +INFO:root: +[Valid][10][UTT][ MSE = 0.2713 | LCC = 0.9234 | SRCC = 0.9253 ] +INFO:root:Average loss=0.1551574856042862 +INFO:root: +[Valid][11][UTT][ MSE = 0.2610 | LCC = 0.9304 | SRCC = 0.9313 ] +INFO:root:Average loss=0.143892664372921 +INFO:root: +[Valid][12][UTT][ MSE = 0.2413 | LCC = 0.9341 | SRCC = 0.9383 ] +INFO:root:Average loss=0.13639079868793488 +INFO:root: +[Valid][13][UTT][ MSE = 0.2606 | LCC = 0.9253 | SRCC = 0.9269 ] +INFO:root:Average loss=0.13784857660532 +INFO:root: +[Valid][14][UTT][ MSE = 0.2199 | LCC = 0.9372 | SRCC = 0.9380 ] +INFO:root:Average loss=0.12419737839698791 +INFO:root: +[Valid][15][UTT][ MSE = 0.2348 | LCC = 0.9405 | SRCC = 0.9412 ] +INFO:root:Average loss=0.11854326629638672 +INFO:root: +[Valid][16][UTT][ MSE = 0.2767 | LCC = 0.9288 | SRCC = 0.9293 ] +INFO:root:Average loss=0.11730419766902923 +INFO:root: +[Valid][17][UTT][ MSE = 0.2203 | LCC = 0.9361 | SRCC = 0.9369 ] +INFO:root:Average loss=0.11867867583036423 +INFO:root: +[Valid][18][UTT][ MSE = 0.2314 | LCC = 0.9395 | SRCC = 0.9401 ] +INFO:root:Average loss=0.10768940830230712 +INFO:root: +[Valid][19][UTT][ MSE = 0.2113 | LCC = 0.9391 | SRCC = 0.9405 ] +INFO:root:Average loss=0.10458001136779785 +INFO:root: +[Valid][20][UTT][ MSE = 0.2126 | LCC = 0.9370 | SRCC = 0.9387 ] +INFO:root:Average loss=0.10168933564424515 +INFO:root: +[Valid][21][UTT][ MSE = 0.2356 | LCC = 0.9395 | SRCC = 0.9395 ] +INFO:root:Average loss=0.1001146149635315 +INFO:root: +[Valid][22][UTT][ MSE = 0.2176 | LCC = 0.9402 | SRCC = 0.9407 ] +INFO:root:Average loss=0.09795925608277321 +INFO:root: +[Valid][23][UTT][ MSE = 0.2277 | LCC = 0.9406 | SRCC = 0.9410 ] +INFO:root:Average loss=0.09502106755971909 +INFO:root: +[Valid][24][UTT][ MSE = 0.2461 | LCC = 0.9345 | SRCC = 0.9362 ] +INFO:root:Average loss=0.09201507613062859 +INFO:root: +[Valid][25][UTT][ MSE = 0.2229 | LCC = 0.9415 | SRCC = 0.9426 ] +INFO:root:Average loss=0.09123263716697692 +INFO:root: +[Valid][26][UTT][ MSE = 0.2210 | LCC = 0.9396 | SRCC = 0.9411 ] +INFO:root:Average loss=0.08635641387104988 +INFO:root: +[Valid][27][UTT][ MSE = 0.2172 | LCC = 0.9403 | SRCC = 0.9397 ] +INFO:root:Average loss=0.08284286966919899 +INFO:root: +[Valid][28][UTT][ MSE = 0.2184 | LCC = 0.9430 | SRCC = 0.9430 ] +INFO:root:Average loss=0.0829875186085701 +INFO:root: +[Valid][29][UTT][ MSE = 0.2249 | LCC = 0.9373 | SRCC = 0.9387 ] +INFO:root:Average loss=0.07729158017039299 +INFO:root: +[Valid][30][UTT][ MSE = 0.2030 | LCC = 0.9411 | SRCC = 0.9411 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2250 | LCC = 0.9433 | SRCC = 0.9435 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run1/config.gin b/results_tencent/w2v2_xlsr_1b/layer0/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run1/model.pt b/results_tencent/w2v2_xlsr_1b/layer0/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bc205aec31e15d7fdc34de79d1964a346ba71ef6 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e6825597aee4dcfee8ce1130b5971b29a52b071a4ae518d3972a8e289c15e24d +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run1/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer0/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4b06e265de1e727d07664e3d77ceddf3d8631e2e --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:92fd34fb1a76392fcbc3a9997bb2f1bffa1dfde6b13931a1918d7538f42ff0b0 +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run1/train.log b/results_tencent/w2v2_xlsr_1b/layer0/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c92434eb6f2d41d48aebada982538474facd4e6b --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5625625503063202 +INFO:root: +[Valid][0][UTT][ MSE = 0.4549 | LCC = 0.8689 | SRCC = 0.8635 ] +INFO:root:Average loss=0.3226039061546326 +INFO:root: +[Valid][1][UTT][ MSE = 0.3311 | LCC = 0.9005 | SRCC = 0.9001 ] +INFO:root:Average loss=0.2796025764346123 +INFO:root: +[Valid][2][UTT][ MSE = 0.3341 | LCC = 0.9019 | SRCC = 0.9002 ] +INFO:root:Average loss=0.2545410225391388 +INFO:root: +[Valid][3][UTT][ MSE = 0.3420 | LCC = 0.9129 | SRCC = 0.9121 ] +INFO:root:Average loss=0.2210259292125702 +INFO:root: +[Valid][4][UTT][ MSE = 0.3081 | LCC = 0.9213 | SRCC = 0.9225 ] +INFO:root:Average loss=0.2035819092988968 +INFO:root: +[Valid][5][UTT][ MSE = 0.3151 | LCC = 0.9181 | SRCC = 0.9180 ] +INFO:root:Average loss=0.18960356712341309 +INFO:root: +[Valid][6][UTT][ MSE = 0.2911 | LCC = 0.9114 | SRCC = 0.9114 ] +INFO:root:Average loss=0.18918941533565523 +INFO:root: +[Valid][7][UTT][ MSE = 0.3167 | LCC = 0.9216 | SRCC = 0.9229 ] +INFO:root:Average loss=0.17717679625749588 +INFO:root: +[Valid][8][UTT][ MSE = 0.2689 | LCC = 0.9297 | SRCC = 0.9307 ] +INFO:root:Average loss=0.1636933110356331 +INFO:root: +[Valid][9][UTT][ MSE = 0.2769 | LCC = 0.9299 | SRCC = 0.9296 ] +INFO:root:Average loss=0.15621712577342986 +INFO:root: +[Valid][10][UTT][ MSE = 0.2714 | LCC = 0.9319 | SRCC = 0.9331 ] +INFO:root:Average loss=0.14770608586072922 +INFO:root: +[Valid][11][UTT][ MSE = 0.2731 | LCC = 0.9307 | SRCC = 0.9313 ] +INFO:root:Average loss=0.14727920913696288 +INFO:root: +[Valid][12][UTT][ MSE = 0.2438 | LCC = 0.9379 | SRCC = 0.9382 ] +INFO:root:Average loss=0.1386261332631111 +INFO:root: +[Valid][13][UTT][ MSE = 0.2465 | LCC = 0.9368 | SRCC = 0.9366 ] +INFO:root:Average loss=0.13319757807254792 +INFO:root: +[Valid][14][UTT][ MSE = 0.2602 | LCC = 0.9369 | SRCC = 0.9379 ] +INFO:root:Average loss=0.12551428055763245 +INFO:root: +[Valid][15][UTT][ MSE = 0.2450 | LCC = 0.9389 | SRCC = 0.9392 ] +INFO:root:Average loss=0.12412531170248985 +INFO:root: +[Valid][16][UTT][ MSE = 0.2668 | LCC = 0.9360 | SRCC = 0.9377 ] +INFO:root:Average loss=0.11730083489418029 +INFO:root: +[Valid][17][UTT][ MSE = 0.2411 | LCC = 0.9392 | SRCC = 0.9402 ] +INFO:root:Average loss=0.11449117630720139 +INFO:root: +[Valid][18][UTT][ MSE = 0.2487 | LCC = 0.9373 | SRCC = 0.9373 ] +INFO:root:Average loss=0.10626060351729393 +INFO:root: +[Valid][19][UTT][ MSE = 0.2328 | LCC = 0.9400 | SRCC = 0.9403 ] +INFO:root:Average loss=0.10432314437627792 +INFO:root: +[Valid][20][UTT][ MSE = 0.2727 | LCC = 0.9365 | SRCC = 0.9366 ] +INFO:root:Average loss=0.09959000548720359 +INFO:root: +[Valid][21][UTT][ MSE = 0.2385 | LCC = 0.9403 | SRCC = 0.9402 ] +INFO:root:Average loss=0.09918516007065772 +INFO:root: +[Valid][22][UTT][ MSE = 0.2277 | LCC = 0.9393 | SRCC = 0.9393 ] +INFO:root:Average loss=0.0930755877494812 +INFO:root: +[Valid][23][UTT][ MSE = 0.2218 | LCC = 0.9417 | SRCC = 0.9412 ] +INFO:root:Average loss=0.09026503226161003 +INFO:root: +[Valid][24][UTT][ MSE = 0.2187 | LCC = 0.9395 | SRCC = 0.9397 ] +INFO:root:Average loss=0.0856800297498703 +INFO:root: +[Valid][25][UTT][ MSE = 0.2509 | LCC = 0.9364 | SRCC = 0.9375 ] +INFO:root:Average loss=0.08366427919268608 +INFO:root: +[Valid][26][UTT][ MSE = 0.2251 | LCC = 0.9437 | SRCC = 0.9437 ] +INFO:root:Average loss=0.08727223020792008 +INFO:root: +[Valid][27][UTT][ MSE = 0.2206 | LCC = 0.9444 | SRCC = 0.9447 ] +INFO:root:Average loss=0.08036321014165879 +INFO:root: +[Valid][28][UTT][ MSE = 0.2237 | LCC = 0.9439 | SRCC = 0.9438 ] +INFO:root:Average loss=0.07954877570271492 +INFO:root: +[Valid][29][UTT][ MSE = 0.2125 | LCC = 0.9437 | SRCC = 0.9435 ] +INFO:root:Average loss=0.0771583261191845 +INFO:root: +[Valid][30][UTT][ MSE = 0.2335 | LCC = 0.9432 | SRCC = 0.9431 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2273 | LCC = 0.9436 | SRCC = 0.9440 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run2/config.gin b/results_tencent/w2v2_xlsr_1b/layer0/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run2/model.pt b/results_tencent/w2v2_xlsr_1b/layer0/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..38085f88d756df5252e1fc09fa94afdcce3932ee --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ce65451359eb42e536f49ce5a3ea5addf565d262e2aaa6444a7767a4b4d2360a +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run2/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer0/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..37ef500fe6c662d21600baf579f6749d8129ce8c --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:237f75783b4c9a2bf7f03abfb1e3db3624055de790e7e5bcb1779e9ec102d7e5 +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run2/train.log b/results_tencent/w2v2_xlsr_1b/layer0/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..57119b2ac8e95dac7662e6bc6e3709de02f6f1b1 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6154819980859757 +INFO:root: +[Valid][0][UTT][ MSE = 0.4125 | LCC = 0.8755 | SRCC = 0.8704 ] +INFO:root:Average loss=0.3077671866416931 +INFO:root: +[Valid][1][UTT][ MSE = 0.4118 | LCC = 0.8892 | SRCC = 0.8898 ] +INFO:root:Average loss=0.2684774111509323 +INFO:root: +[Valid][2][UTT][ MSE = 0.3391 | LCC = 0.9119 | SRCC = 0.9120 ] +INFO:root:Average loss=0.23913850313425064 +INFO:root: +[Valid][3][UTT][ MSE = 0.3493 | LCC = 0.9119 | SRCC = 0.9149 ] +INFO:root:Average loss=0.2197554315328598 +INFO:root: +[Valid][4][UTT][ MSE = 0.3604 | LCC = 0.9114 | SRCC = 0.9144 ] +INFO:root:Average loss=0.20174446111917496 +INFO:root: +[Valid][5][UTT][ MSE = 0.2841 | LCC = 0.9200 | SRCC = 0.9196 ] +INFO:root:Average loss=0.20357848501205444 +INFO:root: +[Valid][6][UTT][ MSE = 0.2890 | LCC = 0.9190 | SRCC = 0.9201 ] +INFO:root:Average loss=0.18638569480180742 +INFO:root: +[Valid][7][UTT][ MSE = 0.3248 | LCC = 0.9161 | SRCC = 0.9169 ] +INFO:root:Average loss=0.17652368277311326 +INFO:root: +[Valid][8][UTT][ MSE = 0.3250 | LCC = 0.9162 | SRCC = 0.9179 ] +INFO:root:Average loss=0.15745565724372865 +INFO:root: +[Valid][9][UTT][ MSE = 0.2949 | LCC = 0.9223 | SRCC = 0.9219 ] +INFO:root:Average loss=0.1558444114923477 +INFO:root: +[Valid][10][UTT][ MSE = 0.2906 | LCC = 0.9264 | SRCC = 0.9265 ] +INFO:root:Average loss=0.1521639266014099 +INFO:root: +[Valid][11][UTT][ MSE = 0.2942 | LCC = 0.9304 | SRCC = 0.9314 ] +INFO:root:Average loss=0.14649193543195724 +INFO:root: +[Valid][12][UTT][ MSE = 0.2536 | LCC = 0.9295 | SRCC = 0.9323 ] +INFO:root:Average loss=0.1342945144176483 +INFO:root: +[Valid][13][UTT][ MSE = 0.2706 | LCC = 0.9300 | SRCC = 0.9305 ] +INFO:root:Average loss=0.1337159252166748 +INFO:root: +[Valid][14][UTT][ MSE = 0.2606 | LCC = 0.9326 | SRCC = 0.9326 ] +INFO:root:Average loss=0.12801638543605803 +INFO:root: +[Valid][15][UTT][ MSE = 0.2504 | LCC = 0.9316 | SRCC = 0.9314 ] +INFO:root:Average loss=0.1212400039434433 +INFO:root: +[Valid][16][UTT][ MSE = 0.2509 | LCC = 0.9348 | SRCC = 0.9355 ] +INFO:root:Average loss=0.11667450961470605 +INFO:root: +[Valid][17][UTT][ MSE = 0.2481 | LCC = 0.9384 | SRCC = 0.9387 ] +INFO:root:Average loss=0.1185927295088768 +INFO:root: +[Valid][18][UTT][ MSE = 0.2569 | LCC = 0.9383 | SRCC = 0.9382 ] +INFO:root:Average loss=0.1035855839252472 +INFO:root: +[Valid][19][UTT][ MSE = 0.2678 | LCC = 0.9372 | SRCC = 0.9379 ] +INFO:root:Average loss=0.10412630984187127 +INFO:root: +[Valid][20][UTT][ MSE = 0.2261 | LCC = 0.9395 | SRCC = 0.9396 ] +INFO:root:Average loss=0.10025821051001549 +INFO:root: +[Valid][21][UTT][ MSE = 0.2419 | LCC = 0.9344 | SRCC = 0.9353 ] +INFO:root:Average loss=0.10060746151208877 +INFO:root: +[Valid][22][UTT][ MSE = 0.2334 | LCC = 0.9401 | SRCC = 0.9412 ] +INFO:root:Average loss=0.09488844057917595 +INFO:root: +[Valid][23][UTT][ MSE = 0.2612 | LCC = 0.9367 | SRCC = 0.9361 ] +INFO:root:Average loss=0.09355402553081513 +INFO:root: +[Valid][24][UTT][ MSE = 0.2433 | LCC = 0.9406 | SRCC = 0.9417 ] +INFO:root:Average loss=0.08714259898662567 +INFO:root: +[Valid][25][UTT][ MSE = 0.2205 | LCC = 0.9388 | SRCC = 0.9397 ] +INFO:root:Average loss=0.08673446053266526 +INFO:root: +[Valid][26][UTT][ MSE = 0.2435 | LCC = 0.9338 | SRCC = 0.9361 ] +INFO:root:Average loss=0.08388666173815727 +INFO:root: +[Valid][27][UTT][ MSE = 0.2539 | LCC = 0.9385 | SRCC = 0.9394 ] +INFO:root:Average loss=0.08105896025896073 +INFO:root: +[Valid][28][UTT][ MSE = 0.2328 | LCC = 0.9381 | SRCC = 0.9394 ] +INFO:root:Average loss=0.07738358035683632 +INFO:root: +[Valid][29][UTT][ MSE = 0.2286 | LCC = 0.9403 | SRCC = 0.9416 ] +INFO:root:Average loss=0.07960386046767234 +INFO:root: +[Valid][30][UTT][ MSE = 0.2387 | LCC = 0.9410 | SRCC = 0.9404 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2478 | LCC = 0.9377 | SRCC = 0.9393 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run3/config.gin b/results_tencent/w2v2_xlsr_1b/layer0/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run3/model.pt b/results_tencent/w2v2_xlsr_1b/layer0/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..176d253b035bb8d60e94407af05d1259e56b3405 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d0c47c1a3f73abd87bb512cce20d4ff05ff4da5b7a95c8f5f1c10aed1e552ba4 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run3/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer0/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..972d5d62d4bc376cf122e43a3bef3485a2cc7818 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9fc3e04ce18b5d8d4fd74903a608925672f5c7ff5e994e098fa0b84816477d14 +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run3/train.log b/results_tencent/w2v2_xlsr_1b/layer0/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..72fbc1ef87521914c4aa12666ad03f47c8cc20a7 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.6068713637590408 +INFO:root: +[Valid][0][UTT][ MSE = 0.4666 | LCC = 0.8770 | SRCC = 0.8744 ] +INFO:root:Average loss=0.31169996321201326 +INFO:root: +[Valid][1][UTT][ MSE = 0.3484 | LCC = 0.9010 | SRCC = 0.9015 ] +INFO:root:Average loss=0.2774031388759613 +INFO:root: +[Valid][2][UTT][ MSE = 0.3230 | LCC = 0.9090 | SRCC = 0.9090 ] +INFO:root:Average loss=0.24974327623844148 +INFO:root: +[Valid][3][UTT][ MSE = 0.2867 | LCC = 0.9156 | SRCC = 0.9161 ] +INFO:root:Average loss=0.22468057066202163 +INFO:root: +[Valid][4][UTT][ MSE = 0.3234 | LCC = 0.9076 | SRCC = 0.9113 ] +INFO:root:Average loss=0.22209899193048477 +INFO:root: +[Valid][5][UTT][ MSE = 0.2801 | LCC = 0.9257 | SRCC = 0.9255 ] +INFO:root:Average loss=0.1906546623110771 +INFO:root: +[Valid][6][UTT][ MSE = 0.2655 | LCC = 0.9227 | SRCC = 0.9221 ] +INFO:root:Average loss=0.1798563766479492 +INFO:root: +[Valid][7][UTT][ MSE = 0.2675 | LCC = 0.9273 | SRCC = 0.9273 ] +INFO:root:Average loss=0.18811516976356507 +INFO:root: +[Valid][8][UTT][ MSE = 0.3199 | LCC = 0.9175 | SRCC = 0.9193 ] +INFO:root:Average loss=0.16699067562818526 +INFO:root: +[Valid][9][UTT][ MSE = 0.2566 | LCC = 0.9332 | SRCC = 0.9327 ] +INFO:root:Average loss=0.15224583756923676 +INFO:root: +[Valid][10][UTT][ MSE = 0.2134 | LCC = 0.9346 | SRCC = 0.9351 ] +INFO:root:Average loss=0.14407925426959992 +INFO:root: +[Valid][11][UTT][ MSE = 0.2397 | LCC = 0.9358 | SRCC = 0.9352 ] +INFO:root:Average loss=0.14253511691093446 +INFO:root: +[Valid][12][UTT][ MSE = 0.2099 | LCC = 0.9347 | SRCC = 0.9351 ] +INFO:root:Average loss=0.1364249673485756 +INFO:root: +[Valid][13][UTT][ MSE = 0.2217 | LCC = 0.9373 | SRCC = 0.9368 ] +INFO:root:Average loss=0.1255361652970314 +INFO:root: +[Valid][14][UTT][ MSE = 0.2267 | LCC = 0.9357 | SRCC = 0.9369 ] +INFO:root:Average loss=0.12403583645820618 +INFO:root: +[Valid][15][UTT][ MSE = 0.2329 | LCC = 0.9394 | SRCC = 0.9390 ] +INFO:root:Average loss=0.11516309452056885 +INFO:root: +[Valid][16][UTT][ MSE = 0.2307 | LCC = 0.9391 | SRCC = 0.9392 ] +INFO:root:Average loss=0.11282758980989456 +INFO:root: +[Valid][17][UTT][ MSE = 0.2349 | LCC = 0.9425 | SRCC = 0.9422 ] +INFO:root:Average loss=0.11103299999237061 +INFO:root: +[Valid][18][UTT][ MSE = 0.2139 | LCC = 0.9394 | SRCC = 0.9388 ] +INFO:root:Average loss=0.11136972174048423 +INFO:root: +[Valid][19][UTT][ MSE = 0.2156 | LCC = 0.9401 | SRCC = 0.9399 ] +INFO:root:Average loss=0.09975002661347389 +INFO:root: +[Valid][20][UTT][ MSE = 0.2357 | LCC = 0.9416 | SRCC = 0.9419 ] +INFO:root:Average loss=0.09740802839398384 +INFO:root: +[Valid][21][UTT][ MSE = 0.2269 | LCC = 0.9405 | SRCC = 0.9401 ] +INFO:root:Average loss=0.09671939477324486 +INFO:root: +[Valid][22][UTT][ MSE = 0.2096 | LCC = 0.9420 | SRCC = 0.9412 ] +INFO:root:Average loss=0.0920209115743637 +INFO:root: +[Valid][23][UTT][ MSE = 0.1918 | LCC = 0.9420 | SRCC = 0.9410 ] +INFO:root:Average loss=0.08833928251266479 +INFO:root: +[Valid][24][UTT][ MSE = 0.2200 | LCC = 0.9409 | SRCC = 0.9416 ] +INFO:root:Average loss=0.08906935358047485 +INFO:root: +[Valid][25][UTT][ MSE = 0.2358 | LCC = 0.9393 | SRCC = 0.9388 ] +INFO:root:Average loss=0.08243948975205422 +INFO:root: +[Valid][26][UTT][ MSE = 0.2074 | LCC = 0.9429 | SRCC = 0.9428 ] +INFO:root:Average loss=0.08152765372395515 +INFO:root: +[Valid][27][UTT][ MSE = 0.2065 | LCC = 0.9430 | SRCC = 0.9424 ] +INFO:root:Average loss=0.08510865947604179 +INFO:root: +[Valid][28][UTT][ MSE = 0.2146 | LCC = 0.9438 | SRCC = 0.9431 ] +INFO:root:Average loss=0.08027066376805306 +INFO:root: +[Valid][29][UTT][ MSE = 0.2023 | LCC = 0.9432 | SRCC = 0.9432 ] +INFO:root:Average loss=0.0777191453576088 +INFO:root: +[Valid][30][UTT][ MSE = 0.2357 | LCC = 0.9411 | SRCC = 0.9409 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2197 | LCC = 0.9426 | SRCC = 0.9431 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run4/config.gin b/results_tencent/w2v2_xlsr_1b/layer0/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run4/model.pt b/results_tencent/w2v2_xlsr_1b/layer0/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..db5403752413baf82888a65b14c7ba03cb03f4a7 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ab646850dff19f57a999df1fe455001e0f3f2a8bef0a63f9162c415cf654614d +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run4/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer0/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..860f5cb240f598e97a400767717302e6100198c5 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6433f8410c74b30b13755194b7c51ccc14273d9004c4fa039fc4ae58b7087aca +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer0/run4/train.log b/results_tencent/w2v2_xlsr_1b/layer0/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e8f0176c4d3e24c89b8616364bc6b9c1c769cb5b --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer0/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5349561239480972 +INFO:root: +[Valid][0][UTT][ MSE = 0.3650 | LCC = 0.8857 | SRCC = 0.8836 ] +INFO:root:Average loss=0.31492416870594026 +INFO:root: +[Valid][1][UTT][ MSE = 0.3664 | LCC = 0.9012 | SRCC = 0.9003 ] +INFO:root:Average loss=0.2785230518579483 +INFO:root: +[Valid][2][UTT][ MSE = 0.3281 | LCC = 0.9068 | SRCC = 0.9081 ] +INFO:root:Average loss=0.24080859124660492 +INFO:root: +[Valid][3][UTT][ MSE = 0.3198 | LCC = 0.9105 | SRCC = 0.9098 ] +INFO:root:Average loss=0.22228272664546966 +INFO:root: +[Valid][4][UTT][ MSE = 0.3309 | LCC = 0.9142 | SRCC = 0.9175 ] +INFO:root:Average loss=0.21112481969594957 +INFO:root: +[Valid][5][UTT][ MSE = 0.2834 | LCC = 0.9108 | SRCC = 0.9116 ] +INFO:root:Average loss=0.19377190309762954 +INFO:root: +[Valid][6][UTT][ MSE = 0.2675 | LCC = 0.9231 | SRCC = 0.9214 ] +INFO:root:Average loss=0.17731251734495163 +INFO:root: +[Valid][7][UTT][ MSE = 0.2569 | LCC = 0.9277 | SRCC = 0.9275 ] +INFO:root:Average loss=0.1720678073167801 +INFO:root: +[Valid][8][UTT][ MSE = 0.2482 | LCC = 0.9287 | SRCC = 0.9280 ] +INFO:root:Average loss=0.162785777926445 +INFO:root: +[Valid][9][UTT][ MSE = 0.2545 | LCC = 0.9312 | SRCC = 0.9314 ] +INFO:root:Average loss=0.15759120577573776 +INFO:root: +[Valid][10][UTT][ MSE = 0.2943 | LCC = 0.9225 | SRCC = 0.9232 ] +INFO:root:Average loss=0.14641180276870727 +INFO:root: +[Valid][11][UTT][ MSE = 0.2384 | LCC = 0.9337 | SRCC = 0.9334 ] +INFO:root:Average loss=0.1441490847468376 +INFO:root: +[Valid][12][UTT][ MSE = 0.2350 | LCC = 0.9356 | SRCC = 0.9354 ] +INFO:root:Average loss=0.13973256444931031 +INFO:root: +[Valid][13][UTT][ MSE = 0.2211 | LCC = 0.9389 | SRCC = 0.9396 ] +INFO:root:Average loss=0.1280410808324814 +INFO:root: +[Valid][14][UTT][ MSE = 0.2408 | LCC = 0.9387 | SRCC = 0.9386 ] +INFO:root:Average loss=0.1207559962272644 +INFO:root: +[Valid][15][UTT][ MSE = 0.2159 | LCC = 0.9375 | SRCC = 0.9386 ] +INFO:root:Average loss=0.1261181117296219 +INFO:root: +[Valid][16][UTT][ MSE = 0.2529 | LCC = 0.9404 | SRCC = 0.9399 ] +INFO:root:Average loss=0.11660472506284714 +INFO:root: +[Valid][17][UTT][ MSE = 0.2302 | LCC = 0.9415 | SRCC = 0.9417 ] +INFO:root:Average loss=0.11482787120342254 +INFO:root: +[Valid][18][UTT][ MSE = 0.2249 | LCC = 0.9380 | SRCC = 0.9374 ] +INFO:root:Average loss=0.10892842146754265 +INFO:root: +[Valid][19][UTT][ MSE = 0.2112 | LCC = 0.9399 | SRCC = 0.9401 ] +INFO:root:Average loss=0.10864609450101853 +INFO:root: +[Valid][20][UTT][ MSE = 0.2407 | LCC = 0.9413 | SRCC = 0.9412 ] +INFO:root:Average loss=0.10043803679943085 +INFO:root: +[Valid][21][UTT][ MSE = 0.2161 | LCC = 0.9390 | SRCC = 0.9388 ] +INFO:root:Average loss=0.10733116045594215 +INFO:root: +[Valid][22][UTT][ MSE = 0.2122 | LCC = 0.9398 | SRCC = 0.9396 ] +INFO:root:Average loss=0.09278872272372246 +INFO:root: +[Valid][23][UTT][ MSE = 0.2037 | LCC = 0.9426 | SRCC = 0.9426 ] +INFO:root:Average loss=0.09084905824065208 +INFO:root: +[Valid][24][UTT][ MSE = 0.1980 | LCC = 0.9436 | SRCC = 0.9437 ] +INFO:root:Average loss=0.09212620869278908 +INFO:root: +[Valid][25][UTT][ MSE = 0.2127 | LCC = 0.9446 | SRCC = 0.9445 ] +INFO:root:Average loss=0.08599353593587876 +INFO:root: +[Valid][26][UTT][ MSE = 0.2163 | LCC = 0.9441 | SRCC = 0.9430 ] +INFO:root:Average loss=0.08607870477437973 +INFO:root: +[Valid][27][UTT][ MSE = 0.2077 | LCC = 0.9438 | SRCC = 0.9427 ] +INFO:root:Average loss=0.08076722538471222 +INFO:root: +[Valid][28][UTT][ MSE = 0.2188 | LCC = 0.9409 | SRCC = 0.9397 ] +INFO:root:Average loss=0.08181671094894409 +INFO:root: +[Valid][29][UTT][ MSE = 0.2161 | LCC = 0.9438 | SRCC = 0.9431 ] +INFO:root:Average loss=0.08122952085733413 +INFO:root: +[Valid][30][UTT][ MSE = 0.2092 | LCC = 0.9450 | SRCC = 0.9449 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2158 | LCC = 0.9441 | SRCC = 0.9443 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run0/config.gin b/results_tencent/w2v2_xlsr_1b/layer10/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run0/model.pt b/results_tencent/w2v2_xlsr_1b/layer10/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..96c74afb9deed46e64dd5d8b4c3f10ce7f09fd27 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2396441248c5a1138b20da70c2d90dabde3b9316d2f305997066a00a741e9869 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run0/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer10/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..865d0755f3a9c6c2553297113b366ad1ec7af0a5 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0099715df543dc19e3a1ea5b08f5abaa4af41fd691fb23d94816389944e50826 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run0/train.log b/results_tencent/w2v2_xlsr_1b/layer10/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..ee8ffaa8f326fe3f9035dac3f013ad1130d8620d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.336890491604805 +INFO:root: +[Valid][0][UTT][ MSE = 0.2087 | LCC = 0.9520 | SRCC = 0.9504 ] +INFO:root:Average loss=0.1259454755783081 +INFO:root: +[Valid][1][UTT][ MSE = 0.2116 | LCC = 0.9610 | SRCC = 0.9596 ] +INFO:root:Average loss=0.09819705653190612 +INFO:root: +[Valid][2][UTT][ MSE = 0.1852 | LCC = 0.9637 | SRCC = 0.9630 ] +INFO:root:Average loss=0.08965905353426934 +INFO:root: +[Valid][3][UTT][ MSE = 0.2002 | LCC = 0.9663 | SRCC = 0.9652 ] +INFO:root:Average loss=0.07846126902103424 +INFO:root: +[Valid][4][UTT][ MSE = 0.1938 | LCC = 0.9677 | SRCC = 0.9668 ] +INFO:root:Average loss=0.07369522723555565 +INFO:root: +[Valid][5][UTT][ MSE = 0.1706 | LCC = 0.9680 | SRCC = 0.9670 ] +INFO:root:Average loss=0.06736601281166077 +INFO:root: +[Valid][6][UTT][ MSE = 0.1857 | LCC = 0.9697 | SRCC = 0.9686 ] +INFO:root:Average loss=0.06256510174274445 +INFO:root: +[Valid][7][UTT][ MSE = 0.1841 | LCC = 0.9687 | SRCC = 0.9683 ] +INFO:root:Average loss=0.057178963392972945 +INFO:root: +[Valid][8][UTT][ MSE = 0.1662 | LCC = 0.9705 | SRCC = 0.9695 ] +INFO:root:Average loss=0.055677166670560836 +INFO:root: +[Valid][9][UTT][ MSE = 0.1803 | LCC = 0.9708 | SRCC = 0.9700 ] +INFO:root:Average loss=0.05127939222753048 +INFO:root: +[Valid][10][UTT][ MSE = 0.1862 | LCC = 0.9685 | SRCC = 0.9683 ] +INFO:root:Average loss=0.05114404490590096 +INFO:root: +[Valid][11][UTT][ MSE = 0.1704 | LCC = 0.9709 | SRCC = 0.9698 ] +INFO:root:Average loss=0.04638264718651772 +INFO:root: +[Valid][12][UTT][ MSE = 0.1547 | LCC = 0.9694 | SRCC = 0.9687 ] +INFO:root:Average loss=0.043258088424801826 +INFO:root: +[Valid][13][UTT][ MSE = 0.1559 | LCC = 0.9706 | SRCC = 0.9692 ] +INFO:root:Average loss=0.0386895831823349 +INFO:root: +[Valid][14][UTT][ MSE = 0.1725 | LCC = 0.9717 | SRCC = 0.9703 ] +INFO:root:Average loss=0.037947445198893544 +INFO:root: +[Valid][15][UTT][ MSE = 0.1603 | LCC = 0.9706 | SRCC = 0.9701 ] +INFO:root:Average loss=0.03552632574737072 +INFO:root: +[Valid][16][UTT][ MSE = 0.1715 | LCC = 0.9711 | SRCC = 0.9701 ] +INFO:root:Average loss=0.03767802760004997 +INFO:root: +[Valid][17][UTT][ MSE = 0.1625 | LCC = 0.9719 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03230235294997692 +INFO:root: +[Valid][18][UTT][ MSE = 0.1610 | LCC = 0.9716 | SRCC = 0.9706 ] +INFO:root:Average loss=0.02946484015882015 +INFO:root: +[Valid][19][UTT][ MSE = 0.1672 | LCC = 0.9709 | SRCC = 0.9699 ] +INFO:root:Average loss=0.0305739828646183 +INFO:root: +[Valid][20][UTT][ MSE = 0.1783 | LCC = 0.9721 | SRCC = 0.9713 ] +INFO:root:Average loss=0.026990802958607674 +INFO:root: +[Valid][21][UTT][ MSE = 0.1634 | LCC = 0.9709 | SRCC = 0.9695 ] +INFO:root:Average loss=0.0273009320050478 +INFO:root: +[Valid][22][UTT][ MSE = 0.1631 | LCC = 0.9719 | SRCC = 0.9705 ] +INFO:root:Average loss=0.02483071409165859 +INFO:root: +[Valid][23][UTT][ MSE = 0.1716 | LCC = 0.9713 | SRCC = 0.9698 ] +INFO:root:Average loss=0.025143111012876033 +INFO:root: +[Valid][24][UTT][ MSE = 0.1823 | LCC = 0.9716 | SRCC = 0.9705 ] +INFO:root:Average loss=0.024068965695798396 +INFO:root: +[Valid][25][UTT][ MSE = 0.1820 | LCC = 0.9700 | SRCC = 0.9687 ] +INFO:root:Average loss=0.023702622689306736 +INFO:root: +[Valid][26][UTT][ MSE = 0.1576 | LCC = 0.9705 | SRCC = 0.9688 ] +INFO:root:Average loss=0.02217792773991823 +INFO:root: +[Valid][27][UTT][ MSE = 0.1528 | LCC = 0.9711 | SRCC = 0.9694 ] +INFO:root:Average loss=0.01914349914342165 +INFO:root: +[Valid][28][UTT][ MSE = 0.1678 | LCC = 0.9711 | SRCC = 0.9695 ] +INFO:root:Average loss=0.01990319024026394 +INFO:root: +[Valid][29][UTT][ MSE = 0.1630 | LCC = 0.9710 | SRCC = 0.9702 ] +INFO:root:Average loss=0.01843558806926012 +INFO:root: +[Valid][30][UTT][ MSE = 0.1778 | LCC = 0.9710 | SRCC = 0.9695 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1782 | LCC = 0.9736 | SRCC = 0.9738 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run1/config.gin b/results_tencent/w2v2_xlsr_1b/layer10/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run1/model.pt b/results_tencent/w2v2_xlsr_1b/layer10/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..689feb6dc457c30b147ec14d92e69b782e83cfef --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7eea0e62a47f411bb430d64574e93f99219dda317b91007f49d8c090083536c5 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run1/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer10/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..461c6973a74a43ca091984c6c20ae22d168619f9 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5ec8ae6652c5d965c2cfc0d834ea0506e235ddc0485bc24259ace0e1ca283d4e +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run1/train.log b/results_tencent/w2v2_xlsr_1b/layer10/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9bb31686f5f25726a21710e8381b4eb8596d6ddb --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3034776659011841 +INFO:root: +[Valid][0][UTT][ MSE = 0.2369 | LCC = 0.9552 | SRCC = 0.9539 ] +INFO:root:Average loss=0.11140210217237473 +INFO:root: +[Valid][1][UTT][ MSE = 0.2338 | LCC = 0.9629 | SRCC = 0.9619 ] +INFO:root:Average loss=0.09449826470017433 +INFO:root: +[Valid][2][UTT][ MSE = 0.2100 | LCC = 0.9654 | SRCC = 0.9650 ] +INFO:root:Average loss=0.0853785012960434 +INFO:root: +[Valid][3][UTT][ MSE = 0.2022 | LCC = 0.9667 | SRCC = 0.9664 ] +INFO:root:Average loss=0.07314070501923561 +INFO:root: +[Valid][4][UTT][ MSE = 0.2024 | LCC = 0.9675 | SRCC = 0.9675 ] +INFO:root:Average loss=0.06660595592856407 +INFO:root: +[Valid][5][UTT][ MSE = 0.1696 | LCC = 0.9682 | SRCC = 0.9679 ] +INFO:root:Average loss=0.06069347958266735 +INFO:root: +[Valid][6][UTT][ MSE = 0.1560 | LCC = 0.9685 | SRCC = 0.9683 ] +INFO:root:Average loss=0.057793654263019564 +INFO:root: +[Valid][7][UTT][ MSE = 0.1777 | LCC = 0.9695 | SRCC = 0.9689 ] +INFO:root:Average loss=0.05186704158782959 +INFO:root: +[Valid][8][UTT][ MSE = 0.1877 | LCC = 0.9690 | SRCC = 0.9680 ] +INFO:root:Average loss=0.049789395943284034 +INFO:root: +[Valid][9][UTT][ MSE = 0.1742 | LCC = 0.9695 | SRCC = 0.9688 ] +INFO:root:Average loss=0.04743037037551403 +INFO:root: +[Valid][10][UTT][ MSE = 0.1809 | LCC = 0.9697 | SRCC = 0.9689 ] +INFO:root:Average loss=0.04348117962479591 +INFO:root: +[Valid][11][UTT][ MSE = 0.2009 | LCC = 0.9699 | SRCC = 0.9694 ] +INFO:root:Average loss=0.04380490492284298 +INFO:root: +[Valid][12][UTT][ MSE = 0.2151 | LCC = 0.9699 | SRCC = 0.9690 ] +INFO:root:Average loss=0.041585229367017744 +INFO:root: +[Valid][13][UTT][ MSE = 0.1945 | LCC = 0.9695 | SRCC = 0.9689 ] +INFO:root:Average loss=0.037729819104075434 +INFO:root: +[Valid][14][UTT][ MSE = 0.1779 | LCC = 0.9695 | SRCC = 0.9694 ] +INFO:root:Average loss=0.03572335395216942 +INFO:root: +[Valid][15][UTT][ MSE = 0.2059 | LCC = 0.9696 | SRCC = 0.9688 ] +INFO:root:Average loss=0.03362747560441494 +INFO:root: +[Valid][16][UTT][ MSE = 0.1552 | LCC = 0.9702 | SRCC = 0.9700 ] +INFO:root:Average loss=0.03154332999885082 +INFO:root: +[Valid][17][UTT][ MSE = 0.1808 | LCC = 0.9699 | SRCC = 0.9696 ] +INFO:root:Average loss=0.029907165586948396 +INFO:root: +[Valid][18][UTT][ MSE = 0.1972 | LCC = 0.9702 | SRCC = 0.9694 ] +INFO:root:Average loss=0.0304673965126276 +INFO:root: +[Valid][19][UTT][ MSE = 0.1670 | LCC = 0.9708 | SRCC = 0.9704 ] +INFO:root:Average loss=0.02730748454481363 +INFO:root: +[Valid][20][UTT][ MSE = 0.1766 | LCC = 0.9694 | SRCC = 0.9694 ] +INFO:root:Average loss=0.026533842198550702 +INFO:root: +[Valid][21][UTT][ MSE = 0.1582 | LCC = 0.9703 | SRCC = 0.9699 ] +INFO:root:Average loss=0.02349126345664263 +INFO:root: +[Valid][22][UTT][ MSE = 0.1728 | LCC = 0.9709 | SRCC = 0.9698 ] +INFO:root:Average loss=0.022217141792178156 +INFO:root: +[Valid][23][UTT][ MSE = 0.1931 | LCC = 0.9703 | SRCC = 0.9694 ] +INFO:root:Average loss=0.022631929434835912 +INFO:root: +[Valid][24][UTT][ MSE = 0.1656 | LCC = 0.9704 | SRCC = 0.9699 ] +INFO:root:Average loss=0.021372505009174345 +INFO:root: +[Valid][25][UTT][ MSE = 0.1665 | LCC = 0.9704 | SRCC = 0.9696 ] +INFO:root:Average loss=0.021069272249937057 +INFO:root: +[Valid][26][UTT][ MSE = 0.1751 | LCC = 0.9704 | SRCC = 0.9696 ] +INFO:root:Average loss=0.020828584395349025 +INFO:root: +[Valid][27][UTT][ MSE = 0.1647 | LCC = 0.9707 | SRCC = 0.9696 ] +INFO:root:Average loss=0.020596337288618087 +INFO:root: +[Valid][28][UTT][ MSE = 0.1730 | LCC = 0.9709 | SRCC = 0.9697 ] +INFO:root:Average loss=0.019751483879983427 +INFO:root: +[Valid][29][UTT][ MSE = 0.1904 | LCC = 0.9684 | SRCC = 0.9689 ] +INFO:root:Average loss=0.017993606075644493 +INFO:root: +[Valid][30][UTT][ MSE = 0.1747 | LCC = 0.9711 | SRCC = 0.9702 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1717 | LCC = 0.9727 | SRCC = 0.9728 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run2/config.gin b/results_tencent/w2v2_xlsr_1b/layer10/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run2/model.pt b/results_tencent/w2v2_xlsr_1b/layer10/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a45ae13f5cec8f61daeaa856b71e274577c17cd2 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b595b49c4315217c0bc3ba777ab30a8b0d6bbca6074af4f2241a3174ba2fff91 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run2/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer10/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..3351770872cc1948442a7b592fc04bec488e0c44 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:85d1d676199fb844b11cd36974862a7685142a3b60f8bcc0a9f7e849f05d1518 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run2/train.log b/results_tencent/w2v2_xlsr_1b/layer10/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..bf01d475cfccfe44a311e96c2ff61c682cb94108 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.34204108798503874 +INFO:root: +[Valid][0][UTT][ MSE = 0.2202 | LCC = 0.9549 | SRCC = 0.9542 ] +INFO:root:Average loss=0.11304460775852203 +INFO:root: +[Valid][1][UTT][ MSE = 0.2289 | LCC = 0.9642 | SRCC = 0.9630 ] +INFO:root:Average loss=0.09458033269643784 +INFO:root: +[Valid][2][UTT][ MSE = 0.1963 | LCC = 0.9666 | SRCC = 0.9660 ] +INFO:root:Average loss=0.08214647570252419 +INFO:root: +[Valid][3][UTT][ MSE = 0.1766 | LCC = 0.9686 | SRCC = 0.9679 ] +INFO:root:Average loss=0.07300080180168152 +INFO:root: +[Valid][4][UTT][ MSE = 0.2141 | LCC = 0.9681 | SRCC = 0.9681 ] +INFO:root:Average loss=0.06647084325551987 +INFO:root: +[Valid][5][UTT][ MSE = 0.1764 | LCC = 0.9700 | SRCC = 0.9691 ] +INFO:root:Average loss=0.062042710900306705 +INFO:root: +[Valid][6][UTT][ MSE = 0.1651 | LCC = 0.9712 | SRCC = 0.9700 ] +INFO:root:Average loss=0.05877923101186752 +INFO:root: +[Valid][7][UTT][ MSE = 0.1852 | LCC = 0.9709 | SRCC = 0.9701 ] +INFO:root:Average loss=0.05500476783514023 +INFO:root: +[Valid][8][UTT][ MSE = 0.1949 | LCC = 0.9711 | SRCC = 0.9709 ] +INFO:root:Average loss=0.052038171619176864 +INFO:root: +[Valid][9][UTT][ MSE = 0.2005 | LCC = 0.9712 | SRCC = 0.9703 ] +INFO:root:Average loss=0.04960429459810257 +INFO:root: +[Valid][10][UTT][ MSE = 0.1776 | LCC = 0.9719 | SRCC = 0.9709 ] +INFO:root:Average loss=0.04712525194883346 +INFO:root: +[Valid][11][UTT][ MSE = 0.1948 | LCC = 0.9718 | SRCC = 0.9708 ] +INFO:root:Average loss=0.04424616032838821 +INFO:root: +[Valid][12][UTT][ MSE = 0.2082 | LCC = 0.9712 | SRCC = 0.9708 ] +INFO:root:Average loss=0.04101126852631569 +INFO:root: +[Valid][13][UTT][ MSE = 0.1680 | LCC = 0.9722 | SRCC = 0.9710 ] +INFO:root:Average loss=0.03821557496488094 +INFO:root: +[Valid][14][UTT][ MSE = 0.1806 | LCC = 0.9723 | SRCC = 0.9713 ] +INFO:root:Average loss=0.0390521232932806 +INFO:root: +[Valid][15][UTT][ MSE = 0.1529 | LCC = 0.9724 | SRCC = 0.9714 ] +INFO:root:Average loss=0.033348818704485896 +INFO:root: +[Valid][16][UTT][ MSE = 0.1499 | LCC = 0.9722 | SRCC = 0.9707 ] +INFO:root:Average loss=0.03261998799443245 +INFO:root: +[Valid][17][UTT][ MSE = 0.1505 | LCC = 0.9726 | SRCC = 0.9712 ] +INFO:root:Average loss=0.030726552873849868 +INFO:root: +[Valid][18][UTT][ MSE = 0.1542 | LCC = 0.9725 | SRCC = 0.9714 ] +INFO:root:Average loss=0.02925019232928753 +INFO:root: +[Valid][19][UTT][ MSE = 0.1656 | LCC = 0.9716 | SRCC = 0.9707 ] +INFO:root:Average loss=0.027687566876411437 +INFO:root: +[Valid][20][UTT][ MSE = 0.1659 | LCC = 0.9727 | SRCC = 0.9715 ] +INFO:root:Average loss=0.025665413960814475 +INFO:root: +[Valid][21][UTT][ MSE = 0.1684 | LCC = 0.9715 | SRCC = 0.9704 ] +INFO:root:Average loss=0.026192927345633506 +INFO:root: +[Valid][22][UTT][ MSE = 0.1501 | LCC = 0.9717 | SRCC = 0.9707 ] +INFO:root:Average loss=0.02539995667338371 +INFO:root: +[Valid][23][UTT][ MSE = 0.1793 | LCC = 0.9723 | SRCC = 0.9711 ] +INFO:root:Average loss=0.025255893766880037 +INFO:root: +[Valid][24][UTT][ MSE = 0.1570 | LCC = 0.9727 | SRCC = 0.9716 ] +INFO:root:Average loss=0.021550463922321798 +INFO:root: +[Valid][25][UTT][ MSE = 0.1606 | LCC = 0.9724 | SRCC = 0.9716 ] +INFO:root:Average loss=0.021381960965692996 +INFO:root: +[Valid][26][UTT][ MSE = 0.1585 | LCC = 0.9728 | SRCC = 0.9716 ] +INFO:root:Average loss=0.0205046251937747 +INFO:root: +[Valid][27][UTT][ MSE = 0.1578 | LCC = 0.9727 | SRCC = 0.9714 ] +INFO:root:Average loss=0.019559857055544853 +INFO:root: +[Valid][28][UTT][ MSE = 0.1575 | LCC = 0.9731 | SRCC = 0.9716 ] +INFO:root:Average loss=0.01873190299421549 +INFO:root: +[Valid][29][UTT][ MSE = 0.1730 | LCC = 0.9725 | SRCC = 0.9711 ] +INFO:root:Average loss=0.01760769736021757 +INFO:root: +[Valid][30][UTT][ MSE = 0.1468 | LCC = 0.9732 | SRCC = 0.9719 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1460 | LCC = 0.9735 | SRCC = 0.9730 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run3/config.gin b/results_tencent/w2v2_xlsr_1b/layer10/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run3/model.pt b/results_tencent/w2v2_xlsr_1b/layer10/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..75396148af4fc4afd8e4ffc9d10532fa87920373 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:746ab857325ae3e972d8fbcd2d063cca8c21c4b97d7057acf92a2d69f39f158f +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run3/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer10/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d604f36c3f63001f7a8e8bb764421cbdfe714ddb --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd97e3d212f266e7a854a03d408d8eabebaaf012d3f1308deade85ff18bc4232 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run3/train.log b/results_tencent/w2v2_xlsr_1b/layer10/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..98d9555d26399d38affe7f3bbbb7eb13b1742ef2 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.39740081477165223 +INFO:root: +[Valid][0][UTT][ MSE = 0.2787 | LCC = 0.9532 | SRCC = 0.9515 ] +INFO:root:Average loss=0.11901652523875236 +INFO:root: +[Valid][1][UTT][ MSE = 0.2036 | LCC = 0.9637 | SRCC = 0.9630 ] +INFO:root:Average loss=0.09661181566119194 +INFO:root: +[Valid][2][UTT][ MSE = 0.2156 | LCC = 0.9666 | SRCC = 0.9660 ] +INFO:root:Average loss=0.08478138923645019 +INFO:root: +[Valid][3][UTT][ MSE = 0.1929 | LCC = 0.9684 | SRCC = 0.9685 ] +INFO:root:Average loss=0.07350452840328217 +INFO:root: +[Valid][4][UTT][ MSE = 0.1891 | LCC = 0.9695 | SRCC = 0.9688 ] +INFO:root:Average loss=0.06660157719254493 +INFO:root: +[Valid][5][UTT][ MSE = 0.1941 | LCC = 0.9696 | SRCC = 0.9692 ] +INFO:root:Average loss=0.06422205093502999 +INFO:root: +[Valid][6][UTT][ MSE = 0.1729 | LCC = 0.9700 | SRCC = 0.9693 ] +INFO:root:Average loss=0.05897677606344223 +INFO:root: +[Valid][7][UTT][ MSE = 0.1992 | LCC = 0.9680 | SRCC = 0.9687 ] +INFO:root:Average loss=0.05546419396996498 +INFO:root: +[Valid][8][UTT][ MSE = 0.2054 | LCC = 0.9708 | SRCC = 0.9707 ] +INFO:root:Average loss=0.052445594802498816 +INFO:root: +[Valid][9][UTT][ MSE = 0.1854 | LCC = 0.9713 | SRCC = 0.9702 ] +INFO:root:Average loss=0.04696131311357021 +INFO:root: +[Valid][10][UTT][ MSE = 0.1826 | LCC = 0.9697 | SRCC = 0.9683 ] +INFO:root:Average loss=0.04471487976610661 +INFO:root: +[Valid][11][UTT][ MSE = 0.1843 | LCC = 0.9714 | SRCC = 0.9707 ] +INFO:root:Average loss=0.04229327766597271 +INFO:root: +[Valid][12][UTT][ MSE = 0.1835 | LCC = 0.9710 | SRCC = 0.9706 ] +INFO:root:Average loss=0.041131781801581384 +INFO:root: +[Valid][13][UTT][ MSE = 0.1792 | LCC = 0.9710 | SRCC = 0.9702 ] +INFO:root:Average loss=0.03820484176278114 +INFO:root: +[Valid][14][UTT][ MSE = 0.1694 | LCC = 0.9710 | SRCC = 0.9703 ] +INFO:root:Average loss=0.03439083243906498 +INFO:root: +[Valid][15][UTT][ MSE = 0.1821 | LCC = 0.9715 | SRCC = 0.9704 ] +INFO:root:Average loss=0.031939811855554584 +INFO:root: +[Valid][16][UTT][ MSE = 0.1838 | LCC = 0.9701 | SRCC = 0.9690 ] +INFO:root:Average loss=0.03014252233505249 +INFO:root: +[Valid][17][UTT][ MSE = 0.1854 | LCC = 0.9714 | SRCC = 0.9701 ] +INFO:root:Average loss=0.030412751704454424 +INFO:root: +[Valid][18][UTT][ MSE = 0.1807 | LCC = 0.9713 | SRCC = 0.9700 ] +INFO:root:Average loss=0.02754251857101917 +INFO:root: +[Valid][19][UTT][ MSE = 0.1922 | LCC = 0.9708 | SRCC = 0.9695 ] +INFO:root:Average loss=0.025810330376029013 +INFO:root: +[Valid][20][UTT][ MSE = 0.1633 | LCC = 0.9711 | SRCC = 0.9700 ] +INFO:root:Average loss=0.026442264035344124 +INFO:root: +[Valid][21][UTT][ MSE = 0.1669 | LCC = 0.9712 | SRCC = 0.9702 ] +INFO:root:Average loss=0.023083708390593528 +INFO:root: +[Valid][22][UTT][ MSE = 0.1736 | LCC = 0.9708 | SRCC = 0.9697 ] +INFO:root:Average loss=0.023433920532464982 +INFO:root: +[Valid][23][UTT][ MSE = 0.1724 | LCC = 0.9711 | SRCC = 0.9697 ] +INFO:root:Average loss=0.022160602807998656 +INFO:root: +[Valid][24][UTT][ MSE = 0.1711 | LCC = 0.9703 | SRCC = 0.9687 ] +INFO:root:Average loss=0.020949866011738778 +INFO:root: +[Valid][25][UTT][ MSE = 0.1776 | LCC = 0.9708 | SRCC = 0.9696 ] +INFO:root:Average loss=0.021094712637364864 +INFO:root: +[Valid][26][UTT][ MSE = 0.2109 | LCC = 0.9700 | SRCC = 0.9690 ] +INFO:root:Average loss=0.019963913790881634 +INFO:root: +[Valid][27][UTT][ MSE = 0.1787 | LCC = 0.9721 | SRCC = 0.9708 ] +INFO:root:Average loss=0.019444172628223896 +INFO:root: +[Valid][28][UTT][ MSE = 0.1838 | LCC = 0.9716 | SRCC = 0.9703 ] +INFO:root:Average loss=0.017823887392878532 +INFO:root: +[Valid][29][UTT][ MSE = 0.1757 | LCC = 0.9712 | SRCC = 0.9698 ] +INFO:root:Average loss=0.017376337982714176 +INFO:root: +[Valid][30][UTT][ MSE = 0.1821 | LCC = 0.9710 | SRCC = 0.9697 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1758 | LCC = 0.9732 | SRCC = 0.9731 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run4/config.gin b/results_tencent/w2v2_xlsr_1b/layer10/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run4/model.pt b/results_tencent/w2v2_xlsr_1b/layer10/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..00b14e70bdd530421a5a6e078ee7339063f31d01 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aea3b85321d3455c985831a2a372ef751e9690bba4412d39e1efff1f5084213e +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run4/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer10/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..750c90fa5577cdb049f67613b44ee2c629496077 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a33bc24c2502c4b6aa515f930894960027361d85155fddc763f7e2b9e87e3dec +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer10/run4/train.log b/results_tencent/w2v2_xlsr_1b/layer10/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3145afa1413b7ac823015ff44426c29468bef0a9 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer10/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3254189197421074 +INFO:root: +[Valid][0][UTT][ MSE = 0.2328 | LCC = 0.9525 | SRCC = 0.9495 ] +INFO:root:Average loss=0.11622360116243362 +INFO:root: +[Valid][1][UTT][ MSE = 0.2439 | LCC = 0.9613 | SRCC = 0.9607 ] +INFO:root:Average loss=0.097457883477211 +INFO:root: +[Valid][2][UTT][ MSE = 0.1929 | LCC = 0.9648 | SRCC = 0.9639 ] +INFO:root:Average loss=0.0855853649675846 +INFO:root: +[Valid][3][UTT][ MSE = 0.1690 | LCC = 0.9670 | SRCC = 0.9664 ] +INFO:root:Average loss=0.07717177510261536 +INFO:root: +[Valid][4][UTT][ MSE = 0.1796 | LCC = 0.9684 | SRCC = 0.9677 ] +INFO:root:Average loss=0.06886540412902832 +INFO:root: +[Valid][5][UTT][ MSE = 0.1569 | LCC = 0.9689 | SRCC = 0.9686 ] +INFO:root:Average loss=0.06412403047084808 +INFO:root: +[Valid][6][UTT][ MSE = 0.1882 | LCC = 0.9695 | SRCC = 0.9696 ] +INFO:root:Average loss=0.061883780986070634 +INFO:root: +[Valid][7][UTT][ MSE = 0.2012 | LCC = 0.9701 | SRCC = 0.9698 ] +INFO:root:Average loss=0.05761380103230476 +INFO:root: +[Valid][8][UTT][ MSE = 0.1557 | LCC = 0.9694 | SRCC = 0.9691 ] +INFO:root:Average loss=0.05313557668030262 +INFO:root: +[Valid][9][UTT][ MSE = 0.1749 | LCC = 0.9693 | SRCC = 0.9685 ] +INFO:root:Average loss=0.05055846644937992 +INFO:root: +[Valid][10][UTT][ MSE = 0.1945 | LCC = 0.9697 | SRCC = 0.9689 ] +INFO:root:Average loss=0.04533326227962971 +INFO:root: +[Valid][11][UTT][ MSE = 0.1820 | LCC = 0.9707 | SRCC = 0.9698 ] +INFO:root:Average loss=0.042704883709549904 +INFO:root: +[Valid][12][UTT][ MSE = 0.1841 | LCC = 0.9713 | SRCC = 0.9701 ] +INFO:root:Average loss=0.04192052176594734 +INFO:root: +[Valid][13][UTT][ MSE = 0.1956 | LCC = 0.9708 | SRCC = 0.9696 ] +INFO:root:Average loss=0.03833007776737213 +INFO:root: +[Valid][14][UTT][ MSE = 0.1579 | LCC = 0.9695 | SRCC = 0.9690 ] +INFO:root:Average loss=0.037045301467180256 +INFO:root: +[Valid][15][UTT][ MSE = 0.1925 | LCC = 0.9708 | SRCC = 0.9700 ] +INFO:root:Average loss=0.033236658349633214 +INFO:root: +[Valid][16][UTT][ MSE = 0.1916 | LCC = 0.9702 | SRCC = 0.9697 ] +INFO:root:Average loss=0.03228645910322666 +INFO:root: +[Valid][17][UTT][ MSE = 0.1595 | LCC = 0.9704 | SRCC = 0.9692 ] +INFO:root:Average loss=0.030627907931804657 +INFO:root: +[Valid][18][UTT][ MSE = 0.1849 | LCC = 0.9702 | SRCC = 0.9691 ] +INFO:root:Average loss=0.028916697904467582 +INFO:root: +[Valid][19][UTT][ MSE = 0.1647 | LCC = 0.9704 | SRCC = 0.9693 ] +INFO:root:Average loss=0.027169602379202844 +INFO:root: +[Valid][20][UTT][ MSE = 0.1484 | LCC = 0.9702 | SRCC = 0.9694 ] +INFO:root:Average loss=0.026740327596664427 +INFO:root: +[Valid][21][UTT][ MSE = 0.1778 | LCC = 0.9710 | SRCC = 0.9698 ] +INFO:root:Average loss=0.026021186761558054 +INFO:root: +[Valid][22][UTT][ MSE = 0.1766 | LCC = 0.9701 | SRCC = 0.9693 ] +INFO:root:Average loss=0.02409725905954838 +INFO:root: +[Valid][23][UTT][ MSE = 0.1609 | LCC = 0.9703 | SRCC = 0.9691 ] +INFO:root:Average loss=0.02387845655530691 +INFO:root: +[Valid][24][UTT][ MSE = 0.1695 | LCC = 0.9705 | SRCC = 0.9696 ] +INFO:root:Average loss=0.023092839457094668 +INFO:root: +[Valid][25][UTT][ MSE = 0.1682 | LCC = 0.9706 | SRCC = 0.9694 ] +INFO:root:Average loss=0.020906803645193577 +INFO:root: +[Valid][26][UTT][ MSE = 0.1781 | LCC = 0.9710 | SRCC = 0.9699 ] +INFO:root:Average loss=0.019768258430063725 +INFO:root: +[Valid][27][UTT][ MSE = 0.1624 | LCC = 0.9705 | SRCC = 0.9699 ] +INFO:root:Average loss=0.020572416350245475 +INFO:root: +[Valid][28][UTT][ MSE = 0.1748 | LCC = 0.9710 | SRCC = 0.9700 ] +INFO:root:Average loss=0.020418645180761814 +INFO:root: +[Valid][29][UTT][ MSE = 0.1812 | LCC = 0.9712 | SRCC = 0.9702 ] +INFO:root:Average loss=0.017605670966207982 +INFO:root: +[Valid][30][UTT][ MSE = 0.1581 | LCC = 0.9707 | SRCC = 0.9696 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1801 | LCC = 0.9732 | SRCC = 0.9729 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run0/config.gin b/results_tencent/w2v2_xlsr_1b/layer12/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run0/model.pt b/results_tencent/w2v2_xlsr_1b/layer12/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..58fe6df3d142ae63118b25295adaae234bea6739 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:51c086fcc468bc8c05fc1735e6b0302e370c6ac5a6377d5ae166e585c903309f +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run0/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer12/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..884bcd7cc06b6043e619fc5fe48eab55aebad29f --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1dcdaeb30fd76f9bfe1d5d5f299d3126b3d73bc7e86c49776bc596affc3db151 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run0/train.log b/results_tencent/w2v2_xlsr_1b/layer12/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a27fe5f5661e07f02883851eafff43909c46bf1d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3913863769173622 +INFO:root: +[Valid][0][UTT][ MSE = 0.2293 | LCC = 0.9518 | SRCC = 0.9520 ] +INFO:root:Average loss=0.1199805645942688 +INFO:root: +[Valid][1][UTT][ MSE = 0.2001 | LCC = 0.9613 | SRCC = 0.9607 ] +INFO:root:Average loss=0.09551390075683594 +INFO:root: +[Valid][2][UTT][ MSE = 0.1809 | LCC = 0.9652 | SRCC = 0.9661 ] +INFO:root:Average loss=0.08739346927404404 +INFO:root: +[Valid][3][UTT][ MSE = 0.1697 | LCC = 0.9676 | SRCC = 0.9675 ] +INFO:root:Average loss=0.07606005719304085 +INFO:root: +[Valid][4][UTT][ MSE = 0.1848 | LCC = 0.9685 | SRCC = 0.9681 ] +INFO:root:Average loss=0.07147465178370475 +INFO:root: +[Valid][5][UTT][ MSE = 0.1736 | LCC = 0.9705 | SRCC = 0.9702 ] +INFO:root:Average loss=0.06407025992870331 +INFO:root: +[Valid][6][UTT][ MSE = 0.1833 | LCC = 0.9709 | SRCC = 0.9704 ] +INFO:root:Average loss=0.06346217229962349 +INFO:root: +[Valid][7][UTT][ MSE = 0.1580 | LCC = 0.9718 | SRCC = 0.9715 ] +INFO:root:Average loss=0.05562333595752716 +INFO:root: +[Valid][8][UTT][ MSE = 0.1867 | LCC = 0.9716 | SRCC = 0.9715 ] +INFO:root:Average loss=0.05358566081523895 +INFO:root: +[Valid][9][UTT][ MSE = 0.1699 | LCC = 0.9722 | SRCC = 0.9718 ] +INFO:root:Average loss=0.050236252173781394 +INFO:root: +[Valid][10][UTT][ MSE = 0.1903 | LCC = 0.9729 | SRCC = 0.9720 ] +INFO:root:Average loss=0.047602344170212745 +INFO:root: +[Valid][11][UTT][ MSE = 0.1664 | LCC = 0.9725 | SRCC = 0.9714 ] +INFO:root:Average loss=0.04492212434113026 +INFO:root: +[Valid][12][UTT][ MSE = 0.1652 | LCC = 0.9724 | SRCC = 0.9712 ] +INFO:root:Average loss=0.0411251454949379 +INFO:root: +[Valid][13][UTT][ MSE = 0.1594 | LCC = 0.9723 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03901994863152504 +INFO:root: +[Valid][14][UTT][ MSE = 0.1594 | LCC = 0.9718 | SRCC = 0.9713 ] +INFO:root:Average loss=0.038402720019221306 +INFO:root: +[Valid][15][UTT][ MSE = 0.1518 | LCC = 0.9720 | SRCC = 0.9711 ] +INFO:root:Average loss=0.03591972129046917 +INFO:root: +[Valid][16][UTT][ MSE = 0.1750 | LCC = 0.9723 | SRCC = 0.9718 ] +INFO:root:Average loss=0.035405205830931666 +INFO:root: +[Valid][17][UTT][ MSE = 0.1747 | LCC = 0.9722 | SRCC = 0.9709 ] +INFO:root:Average loss=0.030866321444511412 +INFO:root: +[Valid][18][UTT][ MSE = 0.1858 | LCC = 0.9724 | SRCC = 0.9716 ] +INFO:root:Average loss=0.031357436463236806 +INFO:root: +[Valid][19][UTT][ MSE = 0.1831 | LCC = 0.9726 | SRCC = 0.9714 ] +INFO:root:Average loss=0.02966436804831028 +INFO:root: +[Valid][20][UTT][ MSE = 0.1685 | LCC = 0.9726 | SRCC = 0.9717 ] +INFO:root:Average loss=0.028118609726428986 +INFO:root: +[Valid][21][UTT][ MSE = 0.1456 | LCC = 0.9709 | SRCC = 0.9710 ] +INFO:root:Average loss=0.026156491190195082 +INFO:root: +[Valid][22][UTT][ MSE = 0.1554 | LCC = 0.9719 | SRCC = 0.9715 ] +INFO:root:Average loss=0.02366402518749237 +INFO:root: +[Valid][23][UTT][ MSE = 0.1526 | LCC = 0.9718 | SRCC = 0.9711 ] +INFO:root:Average loss=0.02481226622313261 +INFO:root: +[Valid][24][UTT][ MSE = 0.1585 | LCC = 0.9727 | SRCC = 0.9717 ] +INFO:root:Average loss=0.023018458656966688 +INFO:root: +[Valid][25][UTT][ MSE = 0.1665 | LCC = 0.9729 | SRCC = 0.9715 ] +INFO:root:Average loss=0.02344856173545122 +INFO:root: +[Valid][26][UTT][ MSE = 0.1791 | LCC = 0.9722 | SRCC = 0.9708 ] +INFO:root:Average loss=0.02137614921480417 +INFO:root: +[Valid][27][UTT][ MSE = 0.1714 | LCC = 0.9726 | SRCC = 0.9714 ] +INFO:root:Average loss=0.021642418719828128 +INFO:root: +[Valid][28][UTT][ MSE = 0.1734 | LCC = 0.9727 | SRCC = 0.9713 ] +INFO:root:Average loss=0.01972642481327057 +INFO:root: +[Valid][29][UTT][ MSE = 0.1540 | LCC = 0.9720 | SRCC = 0.9709 ] +INFO:root:Average loss=0.020498004361987114 +INFO:root: +[Valid][30][UTT][ MSE = 0.1548 | LCC = 0.9718 | SRCC = 0.9709 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1886 | LCC = 0.9737 | SRCC = 0.9740 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run1/config.gin b/results_tencent/w2v2_xlsr_1b/layer12/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run1/model.pt b/results_tencent/w2v2_xlsr_1b/layer12/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..be99ab534dce66dcd8d74cc577e28b17b0a0f298 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3e5fdc1f44e0f988a137b17ac5b3b5d44159ebf9a2347f828007727ec2672fe2 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run1/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer12/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..24768c709658d32cd4389baf3e4d5ad6c83dae9f --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c320c940a8e8abbdeba2afd0ed2197cf65464ab808b924e7f7ee203d2b704e31 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run1/train.log b/results_tencent/w2v2_xlsr_1b/layer12/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..dba6982e53728e956181e7c16b3c24704ce53bd9 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3348625126481056 +INFO:root: +[Valid][0][UTT][ MSE = 0.2373 | LCC = 0.9543 | SRCC = 0.9526 ] +INFO:root:Average loss=0.11907947254180908 +INFO:root: +[Valid][1][UTT][ MSE = 0.2054 | LCC = 0.9640 | SRCC = 0.9631 ] +INFO:root:Average loss=0.09422345995903016 +INFO:root: +[Valid][2][UTT][ MSE = 0.1797 | LCC = 0.9662 | SRCC = 0.9655 ] +INFO:root:Average loss=0.08448675334453583 +INFO:root: +[Valid][3][UTT][ MSE = 0.2200 | LCC = 0.9687 | SRCC = 0.9681 ] +INFO:root:Average loss=0.07307063591480255 +INFO:root: +[Valid][4][UTT][ MSE = 0.1855 | LCC = 0.9698 | SRCC = 0.9690 ] +INFO:root:Average loss=0.06728783163428306 +INFO:root: +[Valid][5][UTT][ MSE = 0.1873 | LCC = 0.9702 | SRCC = 0.9698 ] +INFO:root:Average loss=0.06352852311730385 +INFO:root: +[Valid][6][UTT][ MSE = 0.1771 | LCC = 0.9700 | SRCC = 0.9697 ] +INFO:root:Average loss=0.06100239783525467 +INFO:root: +[Valid][7][UTT][ MSE = 0.2262 | LCC = 0.9709 | SRCC = 0.9703 ] +INFO:root:Average loss=0.05482727211713791 +INFO:root: +[Valid][8][UTT][ MSE = 0.1920 | LCC = 0.9708 | SRCC = 0.9705 ] +INFO:root:Average loss=0.05251459929347038 +INFO:root: +[Valid][9][UTT][ MSE = 0.2090 | LCC = 0.9713 | SRCC = 0.9711 ] +INFO:root:Average loss=0.051476554945111275 +INFO:root: +[Valid][10][UTT][ MSE = 0.1840 | LCC = 0.9723 | SRCC = 0.9714 ] +INFO:root:Average loss=0.048418269246816636 +INFO:root: +[Valid][11][UTT][ MSE = 0.1906 | LCC = 0.9726 | SRCC = 0.9711 ] +INFO:root:Average loss=0.042525299414992335 +INFO:root: +[Valid][12][UTT][ MSE = 0.1849 | LCC = 0.9730 | SRCC = 0.9721 ] +INFO:root:Average loss=0.040628058642148974 +INFO:root: +[Valid][13][UTT][ MSE = 0.1816 | LCC = 0.9727 | SRCC = 0.9715 ] +INFO:root:Average loss=0.03941426111757755 +INFO:root: +[Valid][14][UTT][ MSE = 0.1777 | LCC = 0.9724 | SRCC = 0.9715 ] +INFO:root:Average loss=0.038345316156744955 +INFO:root: +[Valid][15][UTT][ MSE = 0.1572 | LCC = 0.9709 | SRCC = 0.9702 ] +INFO:root:Average loss=0.03394390462338925 +INFO:root: +[Valid][16][UTT][ MSE = 0.1803 | LCC = 0.9714 | SRCC = 0.9701 ] +INFO:root:Average loss=0.03360617782175541 +INFO:root: +[Valid][17][UTT][ MSE = 0.1685 | LCC = 0.9720 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03335613293945789 +INFO:root: +[Valid][18][UTT][ MSE = 0.1852 | LCC = 0.9724 | SRCC = 0.9717 ] +INFO:root:Average loss=0.029015460774302483 +INFO:root: +[Valid][19][UTT][ MSE = 0.1662 | LCC = 0.9732 | SRCC = 0.9721 ] +INFO:root:Average loss=0.02798653718084097 +INFO:root: +[Valid][20][UTT][ MSE = 0.1986 | LCC = 0.9726 | SRCC = 0.9716 ] +INFO:root:Average loss=0.026351450189948084 +INFO:root: +[Valid][21][UTT][ MSE = 0.1723 | LCC = 0.9716 | SRCC = 0.9703 ] +INFO:root:Average loss=0.025744465306401253 +INFO:root: +[Valid][22][UTT][ MSE = 0.1697 | LCC = 0.9728 | SRCC = 0.9718 ] +INFO:root:Average loss=0.02374438074231148 +INFO:root: +[Valid][23][UTT][ MSE = 0.1857 | LCC = 0.9727 | SRCC = 0.9716 ] +INFO:root:Average loss=0.023448499098420145 +INFO:root: +[Valid][24][UTT][ MSE = 0.1511 | LCC = 0.9723 | SRCC = 0.9709 ] +INFO:root:Average loss=0.02465902818739414 +INFO:root: +[Valid][25][UTT][ MSE = 0.2050 | LCC = 0.9725 | SRCC = 0.9710 ] +INFO:root:Average loss=0.02392890863865614 +INFO:root: +[Valid][26][UTT][ MSE = 0.1643 | LCC = 0.9709 | SRCC = 0.9701 ] +INFO:root:Average loss=0.020069898307323454 +INFO:root: +[Valid][27][UTT][ MSE = 0.1995 | LCC = 0.9713 | SRCC = 0.9699 ] +INFO:root:Average loss=0.01928895681351423 +INFO:root: +[Valid][28][UTT][ MSE = 0.1784 | LCC = 0.9720 | SRCC = 0.9704 ] +INFO:root:Average loss=0.019911855436861515 +INFO:root: +[Valid][29][UTT][ MSE = 0.2002 | LCC = 0.9718 | SRCC = 0.9712 ] +INFO:root:Average loss=0.01944432993978262 +INFO:root: +[Valid][30][UTT][ MSE = 0.1801 | LCC = 0.9726 | SRCC = 0.9709 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1610 | LCC = 0.9746 | SRCC = 0.9745 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run2/config.gin b/results_tencent/w2v2_xlsr_1b/layer12/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run2/model.pt b/results_tencent/w2v2_xlsr_1b/layer12/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0aafae92b98091c7361c022b6523e6409ed96295 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0be065829f58cbb2783e0c72e65f8b087c528c6e35276f51b06fc9e5aea9a336 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run2/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer12/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b8a6506753ecfb017283608667c508bb749fe529 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6959ccf056b912975a7d2df50aa8342016623d7443c88f44e059e8d61331476 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run2/train.log b/results_tencent/w2v2_xlsr_1b/layer12/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..513a88bd4bddf88da865b4db3c3df27dd26d3c40 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3600151588320732 +INFO:root: +[Valid][0][UTT][ MSE = 0.2549 | LCC = 0.9496 | SRCC = 0.9471 ] +INFO:root:Average loss=0.11669719356298447 +INFO:root: +[Valid][1][UTT][ MSE = 0.1990 | LCC = 0.9619 | SRCC = 0.9611 ] +INFO:root:Average loss=0.09916962510347366 +INFO:root: +[Valid][2][UTT][ MSE = 0.2357 | LCC = 0.9663 | SRCC = 0.9658 ] +INFO:root:Average loss=0.0858173077404499 +INFO:root: +[Valid][3][UTT][ MSE = 0.2209 | LCC = 0.9673 | SRCC = 0.9673 ] +INFO:root:Average loss=0.07603137770295143 +INFO:root: +[Valid][4][UTT][ MSE = 0.2384 | LCC = 0.9680 | SRCC = 0.9677 ] +INFO:root:Average loss=0.07292900225520134 +INFO:root: +[Valid][5][UTT][ MSE = 0.2105 | LCC = 0.9689 | SRCC = 0.9684 ] +INFO:root:Average loss=0.0682813430428505 +INFO:root: +[Valid][6][UTT][ MSE = 0.2041 | LCC = 0.9703 | SRCC = 0.9693 ] +INFO:root:Average loss=0.06568436309695244 +INFO:root: +[Valid][7][UTT][ MSE = 0.1785 | LCC = 0.9702 | SRCC = 0.9694 ] +INFO:root:Average loss=0.05976380015909672 +INFO:root: +[Valid][8][UTT][ MSE = 0.1654 | LCC = 0.9711 | SRCC = 0.9706 ] +INFO:root:Average loss=0.055057814180850984 +INFO:root: +[Valid][9][UTT][ MSE = 0.2097 | LCC = 0.9712 | SRCC = 0.9708 ] +INFO:root:Average loss=0.05126457738876343 +INFO:root: +[Valid][10][UTT][ MSE = 0.1824 | LCC = 0.9704 | SRCC = 0.9698 ] +INFO:root:Average loss=0.04781511995196343 +INFO:root: +[Valid][11][UTT][ MSE = 0.1862 | LCC = 0.9697 | SRCC = 0.9691 ] +INFO:root:Average loss=0.045222869709134104 +INFO:root: +[Valid][12][UTT][ MSE = 0.2062 | LCC = 0.9707 | SRCC = 0.9701 ] +INFO:root:Average loss=0.043033515229821205 +INFO:root: +[Valid][13][UTT][ MSE = 0.1740 | LCC = 0.9716 | SRCC = 0.9708 ] +INFO:root:Average loss=0.04049786897003651 +INFO:root: +[Valid][14][UTT][ MSE = 0.1746 | LCC = 0.9714 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03779034799337387 +INFO:root: +[Valid][15][UTT][ MSE = 0.1699 | LCC = 0.9711 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03556737481057644 +INFO:root: +[Valid][16][UTT][ MSE = 0.1720 | LCC = 0.9718 | SRCC = 0.9710 ] +INFO:root:Average loss=0.03451529760658741 +INFO:root: +[Valid][17][UTT][ MSE = 0.1769 | LCC = 0.9719 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03423243919014931 +INFO:root: +[Valid][18][UTT][ MSE = 0.1654 | LCC = 0.9725 | SRCC = 0.9717 ] +INFO:root:Average loss=0.03102172437310219 +INFO:root: +[Valid][19][UTT][ MSE = 0.1983 | LCC = 0.9717 | SRCC = 0.9710 ] +INFO:root:Average loss=0.028976590864360333 +INFO:root: +[Valid][20][UTT][ MSE = 0.1928 | LCC = 0.9714 | SRCC = 0.9708 ] +INFO:root:Average loss=0.027848093539476395 +INFO:root: +[Valid][21][UTT][ MSE = 0.1806 | LCC = 0.9716 | SRCC = 0.9709 ] +INFO:root:Average loss=0.026132403247058393 +INFO:root: +[Valid][22][UTT][ MSE = 0.1875 | LCC = 0.9715 | SRCC = 0.9707 ] +INFO:root:Average loss=0.027240149565041064 +INFO:root: +[Valid][23][UTT][ MSE = 0.1997 | LCC = 0.9712 | SRCC = 0.9711 ] +INFO:root:Average loss=0.025335708312690258 +INFO:root: +[Valid][24][UTT][ MSE = 0.1532 | LCC = 0.9719 | SRCC = 0.9710 ] +INFO:root:Average loss=0.025059504114091396 +INFO:root: +[Valid][25][UTT][ MSE = 0.1846 | LCC = 0.9719 | SRCC = 0.9711 ] +INFO:root:Average loss=0.02314675182849169 +INFO:root: +[Valid][26][UTT][ MSE = 0.1644 | LCC = 0.9721 | SRCC = 0.9711 ] +INFO:root:Average loss=0.021322533398866653 +INFO:root: +[Valid][27][UTT][ MSE = 0.1732 | LCC = 0.9717 | SRCC = 0.9708 ] +INFO:root:Average loss=0.021072984896600245 +INFO:root: +[Valid][28][UTT][ MSE = 0.1757 | LCC = 0.9719 | SRCC = 0.9708 ] +INFO:root:Average loss=0.020112346909940244 +INFO:root: +[Valid][29][UTT][ MSE = 0.1815 | LCC = 0.9717 | SRCC = 0.9706 ] +INFO:root:Average loss=0.021243657067418098 +INFO:root: +[Valid][30][UTT][ MSE = 0.1729 | LCC = 0.9725 | SRCC = 0.9716 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1655 | LCC = 0.9737 | SRCC = 0.9740 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run3/config.gin b/results_tencent/w2v2_xlsr_1b/layer12/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run3/model.pt b/results_tencent/w2v2_xlsr_1b/layer12/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..26c4f727a9e780d74a3df1ccaf35aef32c4b8e28 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7ef3dc1f449d2816c49d7bd5f8f8e603d67839269088a4fcea3827be3f694e94 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run3/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer12/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a983d9dbdf068efa9c13f6f2b230e8220be0256a --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe748561a5a8dafd4487f6e6ed2a824f266f318b2d445f2ced6671475d944f4d +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run3/train.log b/results_tencent/w2v2_xlsr_1b/layer12/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a1d7969830049224cbdd51aed02fe471532feed9 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.33308699518442153 +INFO:root: +[Valid][0][UTT][ MSE = 0.2272 | LCC = 0.9543 | SRCC = 0.9526 ] +INFO:root:Average loss=0.12192967706918717 +INFO:root: +[Valid][1][UTT][ MSE = 0.1833 | LCC = 0.9594 | SRCC = 0.9576 ] +INFO:root:Average loss=0.096090646058321 +INFO:root: +[Valid][2][UTT][ MSE = 0.1745 | LCC = 0.9631 | SRCC = 0.9627 ] +INFO:root:Average loss=0.0844027968943119 +INFO:root: +[Valid][3][UTT][ MSE = 0.1712 | LCC = 0.9663 | SRCC = 0.9654 ] +INFO:root:Average loss=0.07752258706092835 +INFO:root: +[Valid][4][UTT][ MSE = 0.1777 | LCC = 0.9676 | SRCC = 0.9670 ] +INFO:root:Average loss=0.07018144589662552 +INFO:root: +[Valid][5][UTT][ MSE = 0.1416 | LCC = 0.9677 | SRCC = 0.9681 ] +INFO:root:Average loss=0.0652089504301548 +INFO:root: +[Valid][6][UTT][ MSE = 0.1764 | LCC = 0.9686 | SRCC = 0.9675 ] +INFO:root:Average loss=0.06465381525456905 +INFO:root: +[Valid][7][UTT][ MSE = 0.1659 | LCC = 0.9698 | SRCC = 0.9687 ] +INFO:root:Average loss=0.05623853003978729 +INFO:root: +[Valid][8][UTT][ MSE = 0.1470 | LCC = 0.9686 | SRCC = 0.9687 ] +INFO:root:Average loss=0.05547955045104027 +INFO:root: +[Valid][9][UTT][ MSE = 0.1418 | LCC = 0.9701 | SRCC = 0.9695 ] +INFO:root:Average loss=0.050350156426429746 +INFO:root: +[Valid][10][UTT][ MSE = 0.1588 | LCC = 0.9706 | SRCC = 0.9695 ] +INFO:root:Average loss=0.049288375243544576 +INFO:root: +[Valid][11][UTT][ MSE = 0.1698 | LCC = 0.9699 | SRCC = 0.9693 ] +INFO:root:Average loss=0.04378345946967602 +INFO:root: +[Valid][12][UTT][ MSE = 0.1593 | LCC = 0.9707 | SRCC = 0.9694 ] +INFO:root:Average loss=0.04247292163968086 +INFO:root: +[Valid][13][UTT][ MSE = 0.1391 | LCC = 0.9672 | SRCC = 0.9671 ] +INFO:root:Average loss=0.04209457015991211 +INFO:root: +[Valid][14][UTT][ MSE = 0.1584 | LCC = 0.9700 | SRCC = 0.9691 ] +INFO:root:Average loss=0.03757225903868675 +INFO:root: +[Valid][15][UTT][ MSE = 0.1520 | LCC = 0.9710 | SRCC = 0.9700 ] +INFO:root:Average loss=0.03553861626982689 +INFO:root: +[Valid][16][UTT][ MSE = 0.1532 | LCC = 0.9711 | SRCC = 0.9700 ] +INFO:root:Average loss=0.03415814504027367 +INFO:root: +[Valid][17][UTT][ MSE = 0.1490 | LCC = 0.9715 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03373126682639122 +INFO:root: +[Valid][18][UTT][ MSE = 0.1750 | LCC = 0.9703 | SRCC = 0.9693 ] +INFO:root:Average loss=0.030285804361104965 +INFO:root: +[Valid][19][UTT][ MSE = 0.1710 | LCC = 0.9714 | SRCC = 0.9704 ] +INFO:root:Average loss=0.02879110249131918 +INFO:root: +[Valid][20][UTT][ MSE = 0.1704 | LCC = 0.9709 | SRCC = 0.9698 ] +INFO:root:Average loss=0.02964004446566105 +INFO:root: +[Valid][21][UTT][ MSE = 0.1556 | LCC = 0.9705 | SRCC = 0.9694 ] +INFO:root:Average loss=0.029375225082039835 +INFO:root: +[Valid][22][UTT][ MSE = 0.1459 | LCC = 0.9708 | SRCC = 0.9699 ] +INFO:root:Average loss=0.025787038430571556 +INFO:root: +[Valid][23][UTT][ MSE = 0.1520 | LCC = 0.9713 | SRCC = 0.9702 ] +INFO:root:Average loss=0.023207168333232404 +INFO:root: +[Valid][24][UTT][ MSE = 0.1654 | LCC = 0.9711 | SRCC = 0.9701 ] +INFO:root:Average loss=0.021904333770275116 +INFO:root: +[Valid][25][UTT][ MSE = 0.1454 | LCC = 0.9709 | SRCC = 0.9695 ] +INFO:root:Average loss=0.02208697086572647 +INFO:root: +[Valid][26][UTT][ MSE = 0.1430 | LCC = 0.9714 | SRCC = 0.9704 ] +INFO:root:Average loss=0.020774987392127513 +INFO:root: +[Valid][27][UTT][ MSE = 0.1605 | LCC = 0.9709 | SRCC = 0.9697 ] +INFO:root:Average loss=0.020628864169120787 +INFO:root: +[Valid][28][UTT][ MSE = 0.1638 | LCC = 0.9703 | SRCC = 0.9694 ] +INFO:root:Average loss=0.021376277767121792 +INFO:root: +[Valid][29][UTT][ MSE = 0.1599 | LCC = 0.9703 | SRCC = 0.9695 ] +INFO:root:Average loss=0.018966404475271703 +INFO:root: +[Valid][30][UTT][ MSE = 0.1732 | LCC = 0.9708 | SRCC = 0.9698 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1491 | LCC = 0.9720 | SRCC = 0.9718 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run4/config.gin b/results_tencent/w2v2_xlsr_1b/layer12/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run4/model.pt b/results_tencent/w2v2_xlsr_1b/layer12/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..34b2b85b039e00917c9b76ddafb968f23330228d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ada32b3c5c28322eb46380b44d3a49b0419aceeb5864c8306398080b313af212 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run4/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer12/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..dde753ed1f53a6d72b9bf1a25ed59c560451dced --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c92338d567db29c8d619ee89bc1ab0e00eb4f1562486ca160876afd4726585a +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer12/run4/train.log b/results_tencent/w2v2_xlsr_1b/layer12/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..7a4ef5948ae01a41f5c1cbf47d54940fee53d511 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer12/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.29663215565681456 +INFO:root: +[Valid][0][UTT][ MSE = 0.2275 | LCC = 0.9548 | SRCC = 0.9532 ] +INFO:root:Average loss=0.11999895679950714 +INFO:root: +[Valid][1][UTT][ MSE = 0.2095 | LCC = 0.9617 | SRCC = 0.9604 ] +INFO:root:Average loss=0.09811235603690148 +INFO:root: +[Valid][2][UTT][ MSE = 0.1643 | LCC = 0.9655 | SRCC = 0.9646 ] +INFO:root:Average loss=0.08616778528690339 +INFO:root: +[Valid][3][UTT][ MSE = 0.2367 | LCC = 0.9656 | SRCC = 0.9650 ] +INFO:root:Average loss=0.07892942205071449 +INFO:root: +[Valid][4][UTT][ MSE = 0.1833 | LCC = 0.9659 | SRCC = 0.9640 ] +INFO:root:Average loss=0.07067632403969765 +INFO:root: +[Valid][5][UTT][ MSE = 0.1789 | LCC = 0.9688 | SRCC = 0.9680 ] +INFO:root:Average loss=0.0663063605427742 +INFO:root: +[Valid][6][UTT][ MSE = 0.1845 | LCC = 0.9692 | SRCC = 0.9690 ] +INFO:root:Average loss=0.06018025963008404 +INFO:root: +[Valid][7][UTT][ MSE = 0.1708 | LCC = 0.9694 | SRCC = 0.9693 ] +INFO:root:Average loss=0.057087264716625216 +INFO:root: +[Valid][8][UTT][ MSE = 0.1624 | LCC = 0.9708 | SRCC = 0.9698 ] +INFO:root:Average loss=0.05612816813588142 +INFO:root: +[Valid][9][UTT][ MSE = 0.1652 | LCC = 0.9710 | SRCC = 0.9703 ] +INFO:root:Average loss=0.05133951646089554 +INFO:root: +[Valid][10][UTT][ MSE = 0.1902 | LCC = 0.9678 | SRCC = 0.9695 ] +INFO:root:Average loss=0.04951666332781315 +INFO:root: +[Valid][11][UTT][ MSE = 0.1794 | LCC = 0.9683 | SRCC = 0.9684 ] +INFO:root:Average loss=0.04647061142325401 +INFO:root: +[Valid][12][UTT][ MSE = 0.1496 | LCC = 0.9711 | SRCC = 0.9700 ] +INFO:root:Average loss=0.041609065264463424 +INFO:root: +[Valid][13][UTT][ MSE = 0.1671 | LCC = 0.9708 | SRCC = 0.9701 ] +INFO:root:Average loss=0.04025442455708981 +INFO:root: +[Valid][14][UTT][ MSE = 0.1674 | LCC = 0.9712 | SRCC = 0.9710 ] +INFO:root:Average loss=0.03882538564503193 +INFO:root: +[Valid][15][UTT][ MSE = 0.1691 | LCC = 0.9699 | SRCC = 0.9694 ] +INFO:root:Average loss=0.03632990884780884 +INFO:root: +[Valid][16][UTT][ MSE = 0.1659 | LCC = 0.9708 | SRCC = 0.9699 ] +INFO:root:Average loss=0.03502399093657732 +INFO:root: +[Valid][17][UTT][ MSE = 0.1504 | LCC = 0.9705 | SRCC = 0.9696 ] +INFO:root:Average loss=0.03269609369337559 +INFO:root: +[Valid][18][UTT][ MSE = 0.1854 | LCC = 0.9703 | SRCC = 0.9693 ] +INFO:root:Average loss=0.02919468815624714 +INFO:root: +[Valid][19][UTT][ MSE = 0.1845 | LCC = 0.9705 | SRCC = 0.9699 ] +INFO:root:Average loss=0.02924080567806959 +INFO:root: +[Valid][20][UTT][ MSE = 0.1920 | LCC = 0.9693 | SRCC = 0.9687 ] +INFO:root:Average loss=0.028776648357510568 +INFO:root: +[Valid][21][UTT][ MSE = 0.1582 | LCC = 0.9706 | SRCC = 0.9699 ] +INFO:root:Average loss=0.0260695690959692 +INFO:root: +[Valid][22][UTT][ MSE = 0.1726 | LCC = 0.9700 | SRCC = 0.9706 ] +INFO:root:Average loss=0.0268720426261425 +INFO:root: +[Valid][23][UTT][ MSE = 0.1384 | LCC = 0.9700 | SRCC = 0.9699 ] +INFO:root:Average loss=0.024204219453036784 +INFO:root: +[Valid][24][UTT][ MSE = 0.1778 | LCC = 0.9709 | SRCC = 0.9709 ] +INFO:root:Average loss=0.024955698125064374 +INFO:root: +[Valid][25][UTT][ MSE = 0.1663 | LCC = 0.9713 | SRCC = 0.9705 ] +INFO:root:Average loss=0.022293042741715906 +INFO:root: +[Valid][26][UTT][ MSE = 0.1813 | LCC = 0.9702 | SRCC = 0.9697 ] +INFO:root:Average loss=0.024111538223922254 +INFO:root: +[Valid][27][UTT][ MSE = 0.1481 | LCC = 0.9713 | SRCC = 0.9701 ] +INFO:root:Average loss=0.022000816367566584 +INFO:root: +[Valid][28][UTT][ MSE = 0.1534 | LCC = 0.9711 | SRCC = 0.9699 ] +INFO:root:Average loss=0.020128659904003143 +INFO:root: +[Valid][29][UTT][ MSE = 0.1795 | LCC = 0.9711 | SRCC = 0.9701 ] +INFO:root:Average loss=0.019167229808866978 +INFO:root: +[Valid][30][UTT][ MSE = 0.1531 | LCC = 0.9711 | SRCC = 0.9701 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1392 | LCC = 0.9738 | SRCC = 0.9738 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run0/config.gin b/results_tencent/w2v2_xlsr_1b/layer14/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run0/model.pt b/results_tencent/w2v2_xlsr_1b/layer14/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f3b9705dfeaffec6ed7b4ee62108e429dedfd685 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:791e969f6f7b0a2bcc508991b156aa63c65256d83996ddd691d74a112a4c300b +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run0/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer14/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d603e16cece5ff2f060378b577518203f13cde03 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6379c480b2cfa952a2d0521e8898945d7008bcaa95d99a70aad46d25a6e8983a +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run0/train.log b/results_tencent/w2v2_xlsr_1b/layer14/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..72be788cbd22111ff14e96b9a483c3503f920290 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.35913520205020905 +INFO:root: +[Valid][0][UTT][ MSE = 0.2135 | LCC = 0.9531 | SRCC = 0.9513 ] +INFO:root:Average loss=0.12993512451648712 +INFO:root: +[Valid][1][UTT][ MSE = 0.2025 | LCC = 0.9628 | SRCC = 0.9624 ] +INFO:root:Average loss=0.10660694479942322 +INFO:root: +[Valid][2][UTT][ MSE = 0.2280 | LCC = 0.9662 | SRCC = 0.9657 ] +INFO:root:Average loss=0.09104528567194939 +INFO:root: +[Valid][3][UTT][ MSE = 0.1839 | LCC = 0.9676 | SRCC = 0.9672 ] +INFO:root:Average loss=0.08307678270339966 +INFO:root: +[Valid][4][UTT][ MSE = 0.1559 | LCC = 0.9689 | SRCC = 0.9685 ] +INFO:root:Average loss=0.08020758709311485 +INFO:root: +[Valid][5][UTT][ MSE = 0.1597 | LCC = 0.9695 | SRCC = 0.9691 ] +INFO:root:Average loss=0.07088887912034988 +INFO:root: +[Valid][6][UTT][ MSE = 0.1914 | LCC = 0.9711 | SRCC = 0.9704 ] +INFO:root:Average loss=0.06603987520933151 +INFO:root: +[Valid][7][UTT][ MSE = 0.1523 | LCC = 0.9720 | SRCC = 0.9715 ] +INFO:root:Average loss=0.06181770527362823 +INFO:root: +[Valid][8][UTT][ MSE = 0.1579 | LCC = 0.9719 | SRCC = 0.9709 ] +INFO:root:Average loss=0.05992174127697945 +INFO:root: +[Valid][9][UTT][ MSE = 0.1801 | LCC = 0.9720 | SRCC = 0.9708 ] +INFO:root:Average loss=0.05536811628937721 +INFO:root: +[Valid][10][UTT][ MSE = 0.1740 | LCC = 0.9721 | SRCC = 0.9711 ] +INFO:root:Average loss=0.05330888241529465 +INFO:root: +[Valid][11][UTT][ MSE = 0.1432 | LCC = 0.9733 | SRCC = 0.9725 ] +INFO:root:Average loss=0.04986370286345482 +INFO:root: +[Valid][12][UTT][ MSE = 0.1662 | LCC = 0.9723 | SRCC = 0.9721 ] +INFO:root:Average loss=0.04711728510260582 +INFO:root: +[Valid][13][UTT][ MSE = 0.1604 | LCC = 0.9725 | SRCC = 0.9718 ] +INFO:root:Average loss=0.04604448440670967 +INFO:root: +[Valid][14][UTT][ MSE = 0.1712 | LCC = 0.9731 | SRCC = 0.9726 ] +INFO:root:Average loss=0.04426930822432041 +INFO:root: +[Valid][15][UTT][ MSE = 0.1752 | LCC = 0.9725 | SRCC = 0.9713 ] +INFO:root:Average loss=0.04151702205836773 +INFO:root: +[Valid][16][UTT][ MSE = 0.1732 | LCC = 0.9728 | SRCC = 0.9719 ] +INFO:root:Average loss=0.03775803191959858 +INFO:root: +[Valid][17][UTT][ MSE = 0.1363 | LCC = 0.9724 | SRCC = 0.9714 ] +INFO:root:Average loss=0.03887156434357166 +INFO:root: +[Valid][18][UTT][ MSE = 0.1512 | LCC = 0.9736 | SRCC = 0.9722 ] +INFO:root:Average loss=0.035700261414051054 +INFO:root: +[Valid][19][UTT][ MSE = 0.1435 | LCC = 0.9733 | SRCC = 0.9721 ] +INFO:root:Average loss=0.03476524882018566 +INFO:root: +[Valid][20][UTT][ MSE = 0.1603 | LCC = 0.9723 | SRCC = 0.9720 ] +INFO:root:Average loss=0.03448873206973076 +INFO:root: +[Valid][21][UTT][ MSE = 0.1307 | LCC = 0.9734 | SRCC = 0.9722 ] +INFO:root:Average loss=0.03187331368029118 +INFO:root: +[Valid][22][UTT][ MSE = 0.1474 | LCC = 0.9736 | SRCC = 0.9724 ] +INFO:root:Average loss=0.030219587728381157 +INFO:root: +[Valid][23][UTT][ MSE = 0.1656 | LCC = 0.9731 | SRCC = 0.9719 ] +INFO:root:Average loss=0.02987503761053085 +INFO:root: +[Valid][24][UTT][ MSE = 0.1849 | LCC = 0.9727 | SRCC = 0.9716 ] +INFO:root:Average loss=0.02659613102674484 +INFO:root: +[Valid][25][UTT][ MSE = 0.1562 | LCC = 0.9729 | SRCC = 0.9716 ] +INFO:root:Average loss=0.02710712546110153 +INFO:root: +[Valid][26][UTT][ MSE = 0.1753 | LCC = 0.9735 | SRCC = 0.9727 ] +INFO:root:Average loss=0.024633434861898423 +INFO:root: +[Valid][27][UTT][ MSE = 0.1576 | LCC = 0.9721 | SRCC = 0.9714 ] +INFO:root:Average loss=0.025024041272699832 +INFO:root: +[Valid][28][UTT][ MSE = 0.1636 | LCC = 0.9731 | SRCC = 0.9724 ] +INFO:root:Average loss=0.023126162804663182 +INFO:root: +[Valid][29][UTT][ MSE = 0.1481 | LCC = 0.9733 | SRCC = 0.9720 ] +INFO:root:Average loss=0.02438951075077057 +INFO:root: +[Valid][30][UTT][ MSE = 0.1535 | LCC = 0.9728 | SRCC = 0.9718 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1449 | LCC = 0.9747 | SRCC = 0.9743 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run1/config.gin b/results_tencent/w2v2_xlsr_1b/layer14/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run1/model.pt b/results_tencent/w2v2_xlsr_1b/layer14/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..43e75fad16152f9e8cad092c2087828bf6280160 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:820f192f64499065655d1ecab5637b2fd9b4eec91cccb0563538a4907e08bbc1 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run1/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer14/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0da9c6dce79478b55093fef8397e94f8949976db --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:412c81ac873ed7f31b0656c763c5267b5ccc44b4558b819b7f97ed9b25530b88 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run1/train.log b/results_tencent/w2v2_xlsr_1b/layer14/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..42cff90ebde987e9d405947023adcfc0d5ff9f90 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46842340528965 +INFO:root: +[Valid][0][UTT][ MSE = 0.2312 | LCC = 0.9471 | SRCC = 0.9456 ] +INFO:root:Average loss=0.131332574903965 +INFO:root: +[Valid][1][UTT][ MSE = 0.2392 | LCC = 0.9605 | SRCC = 0.9611 ] +INFO:root:Average loss=0.10699339973926544 +INFO:root: +[Valid][2][UTT][ MSE = 0.2093 | LCC = 0.9652 | SRCC = 0.9640 ] +INFO:root:Average loss=0.09190927445888519 +INFO:root: +[Valid][3][UTT][ MSE = 0.2144 | LCC = 0.9647 | SRCC = 0.9640 ] +INFO:root:Average loss=0.0908402564227581 +INFO:root: +[Valid][4][UTT][ MSE = 0.1827 | LCC = 0.9679 | SRCC = 0.9678 ] +INFO:root:Average loss=0.07577893301844597 +INFO:root: +[Valid][5][UTT][ MSE = 0.2086 | LCC = 0.9691 | SRCC = 0.9683 ] +INFO:root:Average loss=0.07039803746342659 +INFO:root: +[Valid][6][UTT][ MSE = 0.2066 | LCC = 0.9699 | SRCC = 0.9693 ] +INFO:root:Average loss=0.0692503654062748 +INFO:root: +[Valid][7][UTT][ MSE = 0.2056 | LCC = 0.9698 | SRCC = 0.9695 ] +INFO:root:Average loss=0.06334439212083817 +INFO:root: +[Valid][8][UTT][ MSE = 0.1688 | LCC = 0.9703 | SRCC = 0.9688 ] +INFO:root:Average loss=0.05916948786377907 +INFO:root: +[Valid][9][UTT][ MSE = 0.1754 | LCC = 0.9714 | SRCC = 0.9705 ] +INFO:root:Average loss=0.05409295932948589 +INFO:root: +[Valid][10][UTT][ MSE = 0.1786 | LCC = 0.9716 | SRCC = 0.9707 ] +INFO:root:Average loss=0.05353528892993927 +INFO:root: +[Valid][11][UTT][ MSE = 0.2011 | LCC = 0.9715 | SRCC = 0.9707 ] +INFO:root:Average loss=0.050369425162672996 +INFO:root: +[Valid][12][UTT][ MSE = 0.1903 | LCC = 0.9723 | SRCC = 0.9713 ] +INFO:root:Average loss=0.048289632186293605 +INFO:root: +[Valid][13][UTT][ MSE = 0.1899 | LCC = 0.9718 | SRCC = 0.9707 ] +INFO:root:Average loss=0.04549264313280582 +INFO:root: +[Valid][14][UTT][ MSE = 0.1581 | LCC = 0.9721 | SRCC = 0.9711 ] +INFO:root:Average loss=0.043626859918236734 +INFO:root: +[Valid][15][UTT][ MSE = 0.1674 | LCC = 0.9717 | SRCC = 0.9707 ] +INFO:root:Average loss=0.04523476073145866 +INFO:root: +[Valid][16][UTT][ MSE = 0.2016 | LCC = 0.9716 | SRCC = 0.9708 ] +INFO:root:Average loss=0.04013829173147678 +INFO:root: +[Valid][17][UTT][ MSE = 0.1711 | LCC = 0.9724 | SRCC = 0.9711 ] +INFO:root:Average loss=0.036804634496569635 +INFO:root: +[Valid][18][UTT][ MSE = 0.1758 | LCC = 0.9729 | SRCC = 0.9717 ] +INFO:root:Average loss=0.03758644551038742 +INFO:root: +[Valid][19][UTT][ MSE = 0.1800 | LCC = 0.9726 | SRCC = 0.9719 ] +INFO:root:Average loss=0.03360430620610714 +INFO:root: +[Valid][20][UTT][ MSE = 0.1588 | LCC = 0.9725 | SRCC = 0.9713 ] +INFO:root:Average loss=0.032929503858089446 +INFO:root: +[Valid][21][UTT][ MSE = 0.1649 | LCC = 0.9718 | SRCC = 0.9705 ] +INFO:root:Average loss=0.031103705510497095 +INFO:root: +[Valid][22][UTT][ MSE = 0.1592 | LCC = 0.9708 | SRCC = 0.9707 ] +INFO:root:Average loss=0.031575690180063246 +INFO:root: +[Valid][23][UTT][ MSE = 0.1683 | LCC = 0.9719 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03138573814928532 +INFO:root: +[Valid][24][UTT][ MSE = 0.1661 | LCC = 0.9717 | SRCC = 0.9704 ] +INFO:root:Average loss=0.02788700868189335 +INFO:root: +[Valid][25][UTT][ MSE = 0.1630 | LCC = 0.9720 | SRCC = 0.9709 ] +INFO:root:Average loss=0.02672255340218544 +INFO:root: +[Valid][26][UTT][ MSE = 0.1710 | LCC = 0.9712 | SRCC = 0.9706 ] +INFO:root:Average loss=0.02550650602579117 +INFO:root: +[Valid][27][UTT][ MSE = 0.1591 | LCC = 0.9726 | SRCC = 0.9714 ] +INFO:root:Average loss=0.024903478138148784 +INFO:root: +[Valid][28][UTT][ MSE = 0.1644 | LCC = 0.9714 | SRCC = 0.9698 ] +INFO:root:Average loss=0.023085545137524605 +INFO:root: +[Valid][29][UTT][ MSE = 0.1736 | LCC = 0.9708 | SRCC = 0.9693 ] +INFO:root:Average loss=0.025171086363494397 +INFO:root: +[Valid][30][UTT][ MSE = 0.1808 | LCC = 0.9715 | SRCC = 0.9708 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1711 | LCC = 0.9741 | SRCC = 0.9738 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run2/config.gin b/results_tencent/w2v2_xlsr_1b/layer14/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run2/model.pt b/results_tencent/w2v2_xlsr_1b/layer14/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8b2fb98852a3a22650849c11803b03100a0d2f52 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:443ed5350ec0755b86dd26d2c850d39404bb551504d6d91d50bc15c65c7eda92 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run2/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer14/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..74dfc543dccd18038e225bd9c76d51c1d11918f0 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:064da0f9b4445dfc749d34de9993a289271732987baf070a108a27a196faaf69 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run2/train.log b/results_tencent/w2v2_xlsr_1b/layer14/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f04964d47fb3bca9953bb8b2b963013c536ce794 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4094330803155899 +INFO:root: +[Valid][0][UTT][ MSE = 0.2250 | LCC = 0.9541 | SRCC = 0.9529 ] +INFO:root:Average loss=0.12338552010059357 +INFO:root: +[Valid][1][UTT][ MSE = 0.2381 | LCC = 0.9628 | SRCC = 0.9620 ] +INFO:root:Average loss=0.100899487555027 +INFO:root: +[Valid][2][UTT][ MSE = 0.1826 | LCC = 0.9651 | SRCC = 0.9641 ] +INFO:root:Average loss=0.09398247587680816 +INFO:root: +[Valid][3][UTT][ MSE = 0.1992 | LCC = 0.9684 | SRCC = 0.9675 ] +INFO:root:Average loss=0.0790691710114479 +INFO:root: +[Valid][4][UTT][ MSE = 0.1794 | LCC = 0.9691 | SRCC = 0.9681 ] +INFO:root:Average loss=0.0768638219833374 +INFO:root: +[Valid][5][UTT][ MSE = 0.1962 | LCC = 0.9695 | SRCC = 0.9691 ] +INFO:root:Average loss=0.06978278830647469 +INFO:root: +[Valid][6][UTT][ MSE = 0.1753 | LCC = 0.9710 | SRCC = 0.9702 ] +INFO:root:Average loss=0.06701670476794243 +INFO:root: +[Valid][7][UTT][ MSE = 0.1564 | LCC = 0.9706 | SRCC = 0.9704 ] +INFO:root:Average loss=0.06056914708018303 +INFO:root: +[Valid][8][UTT][ MSE = 0.1828 | LCC = 0.9718 | SRCC = 0.9707 ] +INFO:root:Average loss=0.05702114236354828 +INFO:root: +[Valid][9][UTT][ MSE = 0.1833 | LCC = 0.9720 | SRCC = 0.9711 ] +INFO:root:Average loss=0.057015585467219354 +INFO:root: +[Valid][10][UTT][ MSE = 0.1828 | LCC = 0.9720 | SRCC = 0.9712 ] +INFO:root:Average loss=0.05095863002538681 +INFO:root: +[Valid][11][UTT][ MSE = 0.1597 | LCC = 0.9721 | SRCC = 0.9712 ] +INFO:root:Average loss=0.04993917168676853 +INFO:root: +[Valid][12][UTT][ MSE = 0.1744 | LCC = 0.9721 | SRCC = 0.9712 ] +INFO:root:Average loss=0.04709270837903023 +INFO:root: +[Valid][13][UTT][ MSE = 0.1811 | LCC = 0.9729 | SRCC = 0.9714 ] +INFO:root:Average loss=0.04403570774197579 +INFO:root: +[Valid][14][UTT][ MSE = 0.2101 | LCC = 0.9718 | SRCC = 0.9717 ] +INFO:root:Average loss=0.042721274822950364 +INFO:root: +[Valid][15][UTT][ MSE = 0.1670 | LCC = 0.9727 | SRCC = 0.9717 ] +INFO:root:Average loss=0.04148718075454235 +INFO:root: +[Valid][16][UTT][ MSE = 0.1474 | LCC = 0.9713 | SRCC = 0.9717 ] +INFO:root:Average loss=0.039395705997943876 +INFO:root: +[Valid][17][UTT][ MSE = 0.1753 | LCC = 0.9725 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03698338621854782 +INFO:root: +[Valid][18][UTT][ MSE = 0.1892 | LCC = 0.9729 | SRCC = 0.9723 ] +INFO:root:Average loss=0.03425286029279232 +INFO:root: +[Valid][19][UTT][ MSE = 0.1905 | LCC = 0.9721 | SRCC = 0.9714 ] +INFO:root:Average loss=0.031680950209498404 +INFO:root: +[Valid][20][UTT][ MSE = 0.1621 | LCC = 0.9733 | SRCC = 0.9723 ] +INFO:root:Average loss=0.031147871360182763 +INFO:root: +[Valid][21][UTT][ MSE = 0.1665 | LCC = 0.9724 | SRCC = 0.9713 ] +INFO:root:Average loss=0.031139028280973435 +INFO:root: +[Valid][22][UTT][ MSE = 0.1640 | LCC = 0.9724 | SRCC = 0.9717 ] +INFO:root:Average loss=0.02952673776447773 +INFO:root: +[Valid][23][UTT][ MSE = 0.1550 | LCC = 0.9731 | SRCC = 0.9724 ] +INFO:root:Average loss=0.02764656598865986 +INFO:root: +[Valid][24][UTT][ MSE = 0.1649 | LCC = 0.9732 | SRCC = 0.9714 ] +INFO:root:Average loss=0.027435823313891886 +INFO:root: +[Valid][25][UTT][ MSE = 0.1706 | LCC = 0.9736 | SRCC = 0.9722 ] +INFO:root:Average loss=0.02766454228758812 +INFO:root: +[Valid][26][UTT][ MSE = 0.1719 | LCC = 0.9729 | SRCC = 0.9716 ] +INFO:root:Average loss=0.024470788531005382 +INFO:root: +[Valid][27][UTT][ MSE = 0.1999 | LCC = 0.9729 | SRCC = 0.9714 ] +INFO:root:Average loss=0.02447350437939167 +INFO:root: +[Valid][28][UTT][ MSE = 0.1770 | LCC = 0.9734 | SRCC = 0.9718 ] +INFO:root:Average loss=0.02410015068948269 +INFO:root: +[Valid][29][UTT][ MSE = 0.1765 | LCC = 0.9717 | SRCC = 0.9702 ] +INFO:root:Average loss=0.022663372933864594 +INFO:root: +[Valid][30][UTT][ MSE = 0.1622 | LCC = 0.9730 | SRCC = 0.9717 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1658 | LCC = 0.9747 | SRCC = 0.9745 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run3/config.gin b/results_tencent/w2v2_xlsr_1b/layer14/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run3/model.pt b/results_tencent/w2v2_xlsr_1b/layer14/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..389637fb541e42013c470a44803a01a83ea08927 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c6cc0a551e0fbf7194c8ebd8e01fab1b46a7c86ab5c75f49260b5dd87b685954 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run3/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer14/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..474df0eb02a4a175c2a13d3a2553662c2df543df --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:704d9e0154c2413f88204e48cdccdf8025f67b31fe92426d6c25989bcaadc534 +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run3/train.log b/results_tencent/w2v2_xlsr_1b/layer14/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9d289a5f121e673ab562887f5ceb9af5fe500b2a --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3498149796128273 +INFO:root: +[Valid][0][UTT][ MSE = 0.2541 | LCC = 0.9513 | SRCC = 0.9511 ] +INFO:root:Average loss=0.12520793694257737 +INFO:root: +[Valid][1][UTT][ MSE = 0.2516 | LCC = 0.9595 | SRCC = 0.9583 ] +INFO:root:Average loss=0.10679903656244277 +INFO:root: +[Valid][2][UTT][ MSE = 0.2126 | LCC = 0.9654 | SRCC = 0.9652 ] +INFO:root:Average loss=0.0899157634973526 +INFO:root: +[Valid][3][UTT][ MSE = 0.2250 | LCC = 0.9658 | SRCC = 0.9667 ] +INFO:root:Average loss=0.08075604504346848 +INFO:root: +[Valid][4][UTT][ MSE = 0.2094 | LCC = 0.9691 | SRCC = 0.9685 ] +INFO:root:Average loss=0.07675342950224877 +INFO:root: +[Valid][5][UTT][ MSE = 0.2521 | LCC = 0.9696 | SRCC = 0.9694 ] +INFO:root:Average loss=0.06968105363845825 +INFO:root: +[Valid][6][UTT][ MSE = 0.2195 | LCC = 0.9690 | SRCC = 0.9694 ] +INFO:root:Average loss=0.06533764672279357 +INFO:root: +[Valid][7][UTT][ MSE = 0.1868 | LCC = 0.9711 | SRCC = 0.9703 ] +INFO:root:Average loss=0.062253045052289964 +INFO:root: +[Valid][8][UTT][ MSE = 0.2011 | LCC = 0.9717 | SRCC = 0.9709 ] +INFO:root:Average loss=0.06472649559378624 +INFO:root: +[Valid][9][UTT][ MSE = 0.2016 | LCC = 0.9722 | SRCC = 0.9716 ] +INFO:root:Average loss=0.059630883872509005 +INFO:root: +[Valid][10][UTT][ MSE = 0.1761 | LCC = 0.9725 | SRCC = 0.9717 ] +INFO:root:Average loss=0.05656991240382195 +INFO:root: +[Valid][11][UTT][ MSE = 0.1969 | LCC = 0.9729 | SRCC = 0.9723 ] +INFO:root:Average loss=0.051876720547676086 +INFO:root: +[Valid][12][UTT][ MSE = 0.1788 | LCC = 0.9728 | SRCC = 0.9721 ] +INFO:root:Average loss=0.04794550774991512 +INFO:root: +[Valid][13][UTT][ MSE = 0.1999 | LCC = 0.9731 | SRCC = 0.9723 ] +INFO:root:Average loss=0.04614357663691044 +INFO:root: +[Valid][14][UTT][ MSE = 0.1760 | LCC = 0.9717 | SRCC = 0.9711 ] +INFO:root:Average loss=0.045469377428293226 +INFO:root: +[Valid][15][UTT][ MSE = 0.1797 | LCC = 0.9726 | SRCC = 0.9718 ] +INFO:root:Average loss=0.03945237232744694 +INFO:root: +[Valid][16][UTT][ MSE = 0.1949 | LCC = 0.9723 | SRCC = 0.9712 ] +INFO:root:Average loss=0.03816886077821255 +INFO:root: +[Valid][17][UTT][ MSE = 0.1953 | LCC = 0.9726 | SRCC = 0.9717 ] +INFO:root:Average loss=0.03755255664885044 +INFO:root: +[Valid][18][UTT][ MSE = 0.1939 | LCC = 0.9725 | SRCC = 0.9718 ] +INFO:root:Average loss=0.03607684703171253 +INFO:root: +[Valid][19][UTT][ MSE = 0.1827 | LCC = 0.9723 | SRCC = 0.9716 ] +INFO:root:Average loss=0.03535384340584278 +INFO:root: +[Valid][20][UTT][ MSE = 0.1609 | LCC = 0.9727 | SRCC = 0.9718 ] +INFO:root:Average loss=0.03283886149525642 +INFO:root: +[Valid][21][UTT][ MSE = 0.1880 | LCC = 0.9725 | SRCC = 0.9717 ] +INFO:root:Average loss=0.031010922014713286 +INFO:root: +[Valid][22][UTT][ MSE = 0.1891 | LCC = 0.9723 | SRCC = 0.9710 ] +INFO:root:Average loss=0.029217113301157952 +INFO:root: +[Valid][23][UTT][ MSE = 0.1830 | LCC = 0.9729 | SRCC = 0.9715 ] +INFO:root:Average loss=0.03060842104256153 +INFO:root: +[Valid][24][UTT][ MSE = 0.1855 | LCC = 0.9723 | SRCC = 0.9711 ] +INFO:root:Average loss=0.026756412640213965 +INFO:root: +[Valid][25][UTT][ MSE = 0.1788 | LCC = 0.9721 | SRCC = 0.9710 ] +INFO:root:Average loss=0.027413683444261552 +INFO:root: +[Valid][26][UTT][ MSE = 0.1849 | LCC = 0.9727 | SRCC = 0.9713 ] +INFO:root:Average loss=0.02726874902844429 +INFO:root: +[Valid][27][UTT][ MSE = 0.1877 | LCC = 0.9717 | SRCC = 0.9707 ] +INFO:root:Average loss=0.02572280989587307 +INFO:root: +[Valid][28][UTT][ MSE = 0.1765 | LCC = 0.9719 | SRCC = 0.9710 ] +INFO:root:Average loss=0.023905714556574822 +INFO:root: +[Valid][29][UTT][ MSE = 0.1651 | LCC = 0.9720 | SRCC = 0.9711 ] +INFO:root:Average loss=0.02475802854448557 +INFO:root: +[Valid][30][UTT][ MSE = 0.1775 | LCC = 0.9728 | SRCC = 0.9716 ] +INFO:root: +[Test][31][UTT][ MSE = 0.2002 | LCC = 0.9729 | SRCC = 0.9730 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run4/config.gin b/results_tencent/w2v2_xlsr_1b/layer14/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run4/model.pt b/results_tencent/w2v2_xlsr_1b/layer14/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..31bfc55befefdbcfb44c2e53817693f436e0ff22 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7ba8e3497239a51e2c6438d103930af154714ba11dcd8bb16ac0f1ff2f1ffb63 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run4/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer14/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d18a215f6c4ee0565c07c1f317224d2c81787c20 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:14d4ae939b0e911146f6c3e7749f2268cb7f52985fd83f040c79912d3749e24b +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer14/run4/train.log b/results_tencent/w2v2_xlsr_1b/layer14/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..bfc6a2de1b2ccafa55a6e5d3b191d9e85bfe3af1 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer14/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.30048230546712873 +INFO:root: +[Valid][0][UTT][ MSE = 0.2354 | LCC = 0.9541 | SRCC = 0.9532 ] +INFO:root:Average loss=0.12669492787122727 +INFO:root: +[Valid][1][UTT][ MSE = 0.2265 | LCC = 0.9598 | SRCC = 0.9597 ] +INFO:root:Average loss=0.10164639735221863 +INFO:root: +[Valid][2][UTT][ MSE = 0.1582 | LCC = 0.9656 | SRCC = 0.9643 ] +INFO:root:Average loss=0.08892551997303963 +INFO:root: +[Valid][3][UTT][ MSE = 0.2097 | LCC = 0.9670 | SRCC = 0.9662 ] +INFO:root:Average loss=0.08307573947310448 +INFO:root: +[Valid][4][UTT][ MSE = 0.1850 | LCC = 0.9681 | SRCC = 0.9669 ] +INFO:root:Average loss=0.07276041358709336 +INFO:root: +[Valid][5][UTT][ MSE = 0.1785 | LCC = 0.9688 | SRCC = 0.9677 ] +INFO:root:Average loss=0.06820533552765846 +INFO:root: +[Valid][6][UTT][ MSE = 0.1787 | LCC = 0.9703 | SRCC = 0.9690 ] +INFO:root:Average loss=0.06666072013974189 +INFO:root: +[Valid][7][UTT][ MSE = 0.1801 | LCC = 0.9703 | SRCC = 0.9700 ] +INFO:root:Average loss=0.060466815263032915 +INFO:root: +[Valid][8][UTT][ MSE = 0.1935 | LCC = 0.9703 | SRCC = 0.9695 ] +INFO:root:Average loss=0.06393665878474712 +INFO:root: +[Valid][9][UTT][ MSE = 0.1768 | LCC = 0.9703 | SRCC = 0.9699 ] +INFO:root:Average loss=0.05590659603476524 +INFO:root: +[Valid][10][UTT][ MSE = 0.1568 | LCC = 0.9701 | SRCC = 0.9694 ] +INFO:root:Average loss=0.05479471608996391 +INFO:root: +[Valid][11][UTT][ MSE = 0.1677 | LCC = 0.9709 | SRCC = 0.9697 ] +INFO:root:Average loss=0.047553910151124 +INFO:root: +[Valid][12][UTT][ MSE = 0.1786 | LCC = 0.9702 | SRCC = 0.9693 ] +INFO:root:Average loss=0.04834937435388565 +INFO:root: +[Valid][13][UTT][ MSE = 0.1840 | LCC = 0.9702 | SRCC = 0.9694 ] +INFO:root:Average loss=0.047334659710526465 +INFO:root: +[Valid][14][UTT][ MSE = 0.1661 | LCC = 0.9708 | SRCC = 0.9696 ] +INFO:root:Average loss=0.04262419740855694 +INFO:root: +[Valid][15][UTT][ MSE = 0.1741 | LCC = 0.9710 | SRCC = 0.9698 ] +INFO:root:Average loss=0.043102706044912335 +INFO:root: +[Valid][16][UTT][ MSE = 0.1600 | LCC = 0.9705 | SRCC = 0.9698 ] +INFO:root:Average loss=0.039748920768499375 +INFO:root: +[Valid][17][UTT][ MSE = 0.1602 | LCC = 0.9709 | SRCC = 0.9698 ] +INFO:root:Average loss=0.0367807492762804 +INFO:root: +[Valid][18][UTT][ MSE = 0.1443 | LCC = 0.9711 | SRCC = 0.9700 ] +INFO:root:Average loss=0.0351359471976757 +INFO:root: +[Valid][19][UTT][ MSE = 0.1462 | LCC = 0.9710 | SRCC = 0.9696 ] +INFO:root:Average loss=0.035489044636487964 +INFO:root: +[Valid][20][UTT][ MSE = 0.1700 | LCC = 0.9706 | SRCC = 0.9696 ] +INFO:root:Average loss=0.03468761394917965 +INFO:root: +[Valid][21][UTT][ MSE = 0.1719 | LCC = 0.9710 | SRCC = 0.9696 ] +INFO:root:Average loss=0.03210029791295528 +INFO:root: +[Valid][22][UTT][ MSE = 0.1655 | LCC = 0.9706 | SRCC = 0.9694 ] +INFO:root:Average loss=0.03042221274971962 +INFO:root: +[Valid][23][UTT][ MSE = 0.1534 | LCC = 0.9716 | SRCC = 0.9702 ] +INFO:root:Average loss=0.028675641655921937 +INFO:root: +[Valid][24][UTT][ MSE = 0.1530 | LCC = 0.9713 | SRCC = 0.9700 ] +INFO:root:Average loss=0.02914580476284027 +INFO:root: +[Valid][25][UTT][ MSE = 0.1529 | LCC = 0.9709 | SRCC = 0.9694 ] +INFO:root:Average loss=0.027083216354250907 +INFO:root: +[Valid][26][UTT][ MSE = 0.1589 | LCC = 0.9714 | SRCC = 0.9700 ] +INFO:root:Average loss=0.024772707387804984 +INFO:root: +[Valid][27][UTT][ MSE = 0.1630 | LCC = 0.9707 | SRCC = 0.9691 ] +INFO:root:Average loss=0.025440836630761625 +INFO:root: +[Valid][28][UTT][ MSE = 0.1573 | LCC = 0.9715 | SRCC = 0.9700 ] +INFO:root:Average loss=0.024055237255990507 +INFO:root: +[Valid][29][UTT][ MSE = 0.1643 | LCC = 0.9711 | SRCC = 0.9696 ] +INFO:root:Average loss=0.025066674910485743 +INFO:root: +[Valid][30][UTT][ MSE = 0.1745 | LCC = 0.9716 | SRCC = 0.9707 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1459 | LCC = 0.9738 | SRCC = 0.9740 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run0/config.gin b/results_tencent/w2v2_xlsr_1b/layer16/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run0/model.pt b/results_tencent/w2v2_xlsr_1b/layer16/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..140a96823829285ecca4a6d9063aa2c89ddc2cdb --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:96f9921f5f1fe9d902686d252410fb976831936c09f10b9eb0f64761a7a2c8bc +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run0/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer16/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..aee52d3b6a282989ddeb78c9aef35e4738aa951e --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e7a558021c16791d7320dca370a302cdbf0f1471d8dbe012dd68cd4a45cf2b48 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run0/train.log b/results_tencent/w2v2_xlsr_1b/layer16/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..dd44e17c24aea129d4af4766357ca0aaf5c376be --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.32978833848237993 +INFO:root: +[Valid][0][UTT][ MSE = 0.2501 | LCC = 0.9509 | SRCC = 0.9491 ] +INFO:root:Average loss=0.135498206615448 +INFO:root: +[Valid][1][UTT][ MSE = 0.2139 | LCC = 0.9605 | SRCC = 0.9602 ] +INFO:root:Average loss=0.1082686984539032 +INFO:root: +[Valid][2][UTT][ MSE = 0.1847 | LCC = 0.9647 | SRCC = 0.9644 ] +INFO:root:Average loss=0.09570474877953529 +INFO:root: +[Valid][3][UTT][ MSE = 0.1649 | LCC = 0.9656 | SRCC = 0.9667 ] +INFO:root:Average loss=0.08943065494298935 +INFO:root: +[Valid][4][UTT][ MSE = 0.1883 | LCC = 0.9681 | SRCC = 0.9676 ] +INFO:root:Average loss=0.07785120531916619 +INFO:root: +[Valid][5][UTT][ MSE = 0.1937 | LCC = 0.9702 | SRCC = 0.9697 ] +INFO:root:Average loss=0.0775277078151703 +INFO:root: +[Valid][6][UTT][ MSE = 0.2177 | LCC = 0.9695 | SRCC = 0.9702 ] +INFO:root:Average loss=0.07034219750761986 +INFO:root: +[Valid][7][UTT][ MSE = 0.1876 | LCC = 0.9704 | SRCC = 0.9701 ] +INFO:root:Average loss=0.06430863370001316 +INFO:root: +[Valid][8][UTT][ MSE = 0.1954 | LCC = 0.9716 | SRCC = 0.9710 ] +INFO:root:Average loss=0.060873357087373736 +INFO:root: +[Valid][9][UTT][ MSE = 0.1744 | LCC = 0.9712 | SRCC = 0.9705 ] +INFO:root:Average loss=0.05888403931260109 +INFO:root: +[Valid][10][UTT][ MSE = 0.1895 | LCC = 0.9713 | SRCC = 0.9708 ] +INFO:root:Average loss=0.05431498481333256 +INFO:root: +[Valid][11][UTT][ MSE = 0.1661 | LCC = 0.9725 | SRCC = 0.9716 ] +INFO:root:Average loss=0.053655062705278395 +INFO:root: +[Valid][12][UTT][ MSE = 0.1494 | LCC = 0.9716 | SRCC = 0.9715 ] +INFO:root:Average loss=0.05127254186570644 +INFO:root: +[Valid][13][UTT][ MSE = 0.1567 | LCC = 0.9724 | SRCC = 0.9719 ] +INFO:root:Average loss=0.04778581410646439 +INFO:root: +[Valid][14][UTT][ MSE = 0.1729 | LCC = 0.9726 | SRCC = 0.9718 ] +INFO:root:Average loss=0.04501250927150249 +INFO:root: +[Valid][15][UTT][ MSE = 0.1775 | LCC = 0.9726 | SRCC = 0.9714 ] +INFO:root:Average loss=0.043521199718117715 +INFO:root: +[Valid][16][UTT][ MSE = 0.1883 | LCC = 0.9725 | SRCC = 0.9716 ] +INFO:root:Average loss=0.041197728529572486 +INFO:root: +[Valid][17][UTT][ MSE = 0.1625 | LCC = 0.9734 | SRCC = 0.9726 ] +INFO:root:Average loss=0.04158296571671963 +INFO:root: +[Valid][18][UTT][ MSE = 0.1890 | LCC = 0.9724 | SRCC = 0.9710 ] +INFO:root:Average loss=0.0390112311989069 +INFO:root: +[Valid][19][UTT][ MSE = 0.1745 | LCC = 0.9731 | SRCC = 0.9719 ] +INFO:root:Average loss=0.036444891929626466 +INFO:root: +[Valid][20][UTT][ MSE = 0.1609 | LCC = 0.9730 | SRCC = 0.9722 ] +INFO:root:Average loss=0.03500626638531685 +INFO:root: +[Valid][21][UTT][ MSE = 0.1653 | LCC = 0.9723 | SRCC = 0.9725 ] +INFO:root:Average loss=0.03436871945858002 +INFO:root: +[Valid][22][UTT][ MSE = 0.1714 | LCC = 0.9728 | SRCC = 0.9717 ] +INFO:root:Average loss=0.03348481321334839 +INFO:root: +[Valid][23][UTT][ MSE = 0.1687 | LCC = 0.9727 | SRCC = 0.9722 ] +INFO:root:Average loss=0.031027580931782723 +INFO:root: +[Valid][24][UTT][ MSE = 0.1665 | LCC = 0.9734 | SRCC = 0.9722 ] +INFO:root:Average loss=0.03141454090178013 +INFO:root: +[Valid][25][UTT][ MSE = 0.1782 | LCC = 0.9731 | SRCC = 0.9720 ] +INFO:root:Average loss=0.028503124147653578 +INFO:root: +[Valid][26][UTT][ MSE = 0.1661 | LCC = 0.9738 | SRCC = 0.9720 ] +INFO:root:Average loss=0.028829143896698953 +INFO:root: +[Valid][27][UTT][ MSE = 0.1794 | LCC = 0.9729 | SRCC = 0.9718 ] +INFO:root:Average loss=0.028518202155828475 +INFO:root: +[Valid][28][UTT][ MSE = 0.1486 | LCC = 0.9728 | SRCC = 0.9726 ] +INFO:root:Average loss=0.02543006057292223 +INFO:root: +[Valid][29][UTT][ MSE = 0.1700 | LCC = 0.9738 | SRCC = 0.9723 ] +INFO:root:Average loss=0.0269305290132761 +INFO:root: +[Valid][30][UTT][ MSE = 0.1651 | LCC = 0.9734 | SRCC = 0.9721 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1674 | LCC = 0.9739 | SRCC = 0.9736 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run1/config.gin b/results_tencent/w2v2_xlsr_1b/layer16/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run1/model.pt b/results_tencent/w2v2_xlsr_1b/layer16/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..4cf220a215b88a202e6d24bafc087a6998d5c4a3 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3cf00187c27bb8825ce51e02969ba06bae98af45050feb8eab68ec7e6d7573ec +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run1/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer16/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4f6a8fd6ba8e9ab2eb9ba7ff01253e3700b3a3f5 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:27d4d10b2da4ed4dea639e8ab4fb8f07271cbabc8ee7bc7dc7d6b74bfdb38c90 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run1/train.log b/results_tencent/w2v2_xlsr_1b/layer16/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..a75cb5ac172aa48b58b32e0644c3e0bd26dc1c63 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.39289743030071256 +INFO:root: +[Valid][0][UTT][ MSE = 0.2226 | LCC = 0.9473 | SRCC = 0.9466 ] +INFO:root:Average loss=0.1409991638958454 +INFO:root: +[Valid][1][UTT][ MSE = 0.2166 | LCC = 0.9602 | SRCC = 0.9609 ] +INFO:root:Average loss=0.1142392874956131 +INFO:root: +[Valid][2][UTT][ MSE = 0.1827 | LCC = 0.9640 | SRCC = 0.9638 ] +INFO:root:Average loss=0.0976745948791504 +INFO:root: +[Valid][3][UTT][ MSE = 0.1864 | LCC = 0.9677 | SRCC = 0.9674 ] +INFO:root:Average loss=0.08410291373729706 +INFO:root: +[Valid][4][UTT][ MSE = 0.1910 | LCC = 0.9692 | SRCC = 0.9687 ] +INFO:root:Average loss=0.08050060567259788 +INFO:root: +[Valid][5][UTT][ MSE = 0.1861 | LCC = 0.9705 | SRCC = 0.9698 ] +INFO:root:Average loss=0.07455694219470024 +INFO:root: +[Valid][6][UTT][ MSE = 0.1886 | LCC = 0.9712 | SRCC = 0.9707 ] +INFO:root:Average loss=0.07212529742717742 +INFO:root: +[Valid][7][UTT][ MSE = 0.1888 | LCC = 0.9716 | SRCC = 0.9707 ] +INFO:root:Average loss=0.06621829041838646 +INFO:root: +[Valid][8][UTT][ MSE = 0.2298 | LCC = 0.9712 | SRCC = 0.9710 ] +INFO:root:Average loss=0.06296955907344819 +INFO:root: +[Valid][9][UTT][ MSE = 0.1616 | LCC = 0.9700 | SRCC = 0.9708 ] +INFO:root:Average loss=0.058941806137561795 +INFO:root: +[Valid][10][UTT][ MSE = 0.1677 | LCC = 0.9725 | SRCC = 0.9717 ] +INFO:root:Average loss=0.057044098407030105 +INFO:root: +[Valid][11][UTT][ MSE = 0.1807 | LCC = 0.9731 | SRCC = 0.9724 ] +INFO:root:Average loss=0.05224733059108257 +INFO:root: +[Valid][12][UTT][ MSE = 0.1741 | LCC = 0.9727 | SRCC = 0.9719 ] +INFO:root:Average loss=0.050265861690044406 +INFO:root: +[Valid][13][UTT][ MSE = 0.1873 | LCC = 0.9723 | SRCC = 0.9711 ] +INFO:root:Average loss=0.04813371841609478 +INFO:root: +[Valid][14][UTT][ MSE = 0.1915 | LCC = 0.9717 | SRCC = 0.9711 ] +INFO:root:Average loss=0.04533713187277317 +INFO:root: +[Valid][15][UTT][ MSE = 0.1726 | LCC = 0.9732 | SRCC = 0.9720 ] +INFO:root:Average loss=0.04339116528630257 +INFO:root: +[Valid][16][UTT][ MSE = 0.1756 | LCC = 0.9731 | SRCC = 0.9719 ] +INFO:root:Average loss=0.04116500888764858 +INFO:root: +[Valid][17][UTT][ MSE = 0.1712 | LCC = 0.9731 | SRCC = 0.9721 ] +INFO:root:Average loss=0.03894667837023735 +INFO:root: +[Valid][18][UTT][ MSE = 0.1731 | LCC = 0.9731 | SRCC = 0.9720 ] +INFO:root:Average loss=0.036711257144808766 +INFO:root: +[Valid][19][UTT][ MSE = 0.1769 | LCC = 0.9730 | SRCC = 0.9719 ] +INFO:root:Average loss=0.03705929772555828 +INFO:root: +[Valid][20][UTT][ MSE = 0.1967 | LCC = 0.9733 | SRCC = 0.9724 ] +INFO:root:Average loss=0.034721498191356656 +INFO:root: +[Valid][21][UTT][ MSE = 0.2245 | LCC = 0.9714 | SRCC = 0.9712 ] +INFO:root:Average loss=0.03270454277098179 +INFO:root: +[Valid][22][UTT][ MSE = 0.1851 | LCC = 0.9723 | SRCC = 0.9716 ] +INFO:root:Average loss=0.031208634853363035 +INFO:root: +[Valid][23][UTT][ MSE = 0.1996 | LCC = 0.9725 | SRCC = 0.9716 ] +INFO:root:Average loss=0.030608485221862793 +INFO:root: +[Valid][24][UTT][ MSE = 0.1645 | LCC = 0.9728 | SRCC = 0.9717 ] +INFO:root:Average loss=0.029035234481096267 +INFO:root: +[Valid][25][UTT][ MSE = 0.1695 | LCC = 0.9727 | SRCC = 0.9713 ] +INFO:root:Average loss=0.027976490922272206 +INFO:root: +[Valid][26][UTT][ MSE = 0.1984 | LCC = 0.9721 | SRCC = 0.9710 ] +INFO:root:Average loss=0.027334371566772462 +INFO:root: +[Valid][27][UTT][ MSE = 0.1692 | LCC = 0.9727 | SRCC = 0.9716 ] +INFO:root:Average loss=0.02619895838201046 +INFO:root: +[Valid][28][UTT][ MSE = 0.1941 | LCC = 0.9727 | SRCC = 0.9718 ] +INFO:root:Average loss=0.02569424569606781 +INFO:root: +[Valid][29][UTT][ MSE = 0.1621 | LCC = 0.9730 | SRCC = 0.9714 ] +INFO:root:Average loss=0.024380564078688622 +INFO:root: +[Valid][30][UTT][ MSE = 0.1906 | LCC = 0.9726 | SRCC = 0.9715 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1934 | LCC = 0.9739 | SRCC = 0.9740 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run2/config.gin b/results_tencent/w2v2_xlsr_1b/layer16/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run2/model.pt b/results_tencent/w2v2_xlsr_1b/layer16/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3bc66e2de3af9bded67683e94c4f941f2ea0e2d2 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b894ee7773ab984578c4a172af2d0d8e34bd0f7c8d0bafd9a6f374e666d79ed8 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run2/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer16/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ddf5d5320a4a34a389fb8dca2f31d4cd7b5e6280 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8de5ebb0891762b9484382f41fdae3cdd6da2babe2529a665c33e8e47ba4b596 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run2/train.log b/results_tencent/w2v2_xlsr_1b/layer16/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..53855602b0c2032a248645bd7a2cb7115a13171f --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3519496709108353 +INFO:root: +[Valid][0][UTT][ MSE = 0.2628 | LCC = 0.9530 | SRCC = 0.9521 ] +INFO:root:Average loss=0.1381306784749031 +INFO:root: +[Valid][1][UTT][ MSE = 0.2904 | LCC = 0.9592 | SRCC = 0.9576 ] +INFO:root:Average loss=0.11325771272182465 +INFO:root: +[Valid][2][UTT][ MSE = 0.2275 | LCC = 0.9644 | SRCC = 0.9643 ] +INFO:root:Average loss=0.10248771142959595 +INFO:root: +[Valid][3][UTT][ MSE = 0.2703 | LCC = 0.9658 | SRCC = 0.9655 ] +INFO:root:Average loss=0.0882539504468441 +INFO:root: +[Valid][4][UTT][ MSE = 0.2350 | LCC = 0.9676 | SRCC = 0.9668 ] +INFO:root:Average loss=0.07979561686515808 +INFO:root: +[Valid][5][UTT][ MSE = 0.1707 | LCC = 0.9690 | SRCC = 0.9679 ] +INFO:root:Average loss=0.0760783205628395 +INFO:root: +[Valid][6][UTT][ MSE = 0.1971 | LCC = 0.9699 | SRCC = 0.9688 ] +INFO:root:Average loss=0.06910501489043236 +INFO:root: +[Valid][7][UTT][ MSE = 0.1951 | LCC = 0.9709 | SRCC = 0.9701 ] +INFO:root:Average loss=0.0668338846564293 +INFO:root: +[Valid][8][UTT][ MSE = 0.1859 | LCC = 0.9712 | SRCC = 0.9701 ] +INFO:root:Average loss=0.06476246473193169 +INFO:root: +[Valid][9][UTT][ MSE = 0.2021 | LCC = 0.9695 | SRCC = 0.9701 ] +INFO:root:Average loss=0.06230820921063423 +INFO:root: +[Valid][10][UTT][ MSE = 0.1884 | LCC = 0.9700 | SRCC = 0.9701 ] +INFO:root:Average loss=0.05767055383324623 +INFO:root: +[Valid][11][UTT][ MSE = 0.1660 | LCC = 0.9719 | SRCC = 0.9710 ] +INFO:root:Average loss=0.055034693092107775 +INFO:root: +[Valid][12][UTT][ MSE = 0.1845 | LCC = 0.9718 | SRCC = 0.9709 ] +INFO:root:Average loss=0.05187597268819809 +INFO:root: +[Valid][13][UTT][ MSE = 0.1945 | LCC = 0.9725 | SRCC = 0.9716 ] +INFO:root:Average loss=0.049474487975239755 +INFO:root: +[Valid][14][UTT][ MSE = 0.1857 | LCC = 0.9717 | SRCC = 0.9713 ] +INFO:root:Average loss=0.04784599708020687 +INFO:root: +[Valid][15][UTT][ MSE = 0.1716 | LCC = 0.9714 | SRCC = 0.9707 ] +INFO:root:Average loss=0.05009053386747837 +INFO:root: +[Valid][16][UTT][ MSE = 0.1719 | LCC = 0.9730 | SRCC = 0.9718 ] +INFO:root:Average loss=0.043182773992419245 +INFO:root: +[Valid][17][UTT][ MSE = 0.1972 | LCC = 0.9723 | SRCC = 0.9714 ] +INFO:root:Average loss=0.0411893398463726 +INFO:root: +[Valid][18][UTT][ MSE = 0.1788 | LCC = 0.9729 | SRCC = 0.9721 ] +INFO:root:Average loss=0.039105871632695195 +INFO:root: +[Valid][19][UTT][ MSE = 0.1899 | LCC = 0.9724 | SRCC = 0.9714 ] +INFO:root:Average loss=0.03808129104971886 +INFO:root: +[Valid][20][UTT][ MSE = 0.2031 | LCC = 0.9720 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03920415551960468 +INFO:root: +[Valid][21][UTT][ MSE = 0.2221 | LCC = 0.9719 | SRCC = 0.9717 ] +INFO:root:Average loss=0.03606594680249691 +INFO:root: +[Valid][22][UTT][ MSE = 0.1821 | LCC = 0.9726 | SRCC = 0.9715 ] +INFO:root:Average loss=0.03417611461877823 +INFO:root: +[Valid][23][UTT][ MSE = 0.1919 | LCC = 0.9722 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03206689038872719 +INFO:root: +[Valid][24][UTT][ MSE = 0.1867 | LCC = 0.9718 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03241590923070908 +INFO:root: +[Valid][25][UTT][ MSE = 0.1802 | LCC = 0.9728 | SRCC = 0.9715 ] +INFO:root:Average loss=0.029256480023264884 +INFO:root: +[Valid][26][UTT][ MSE = 0.1781 | LCC = 0.9725 | SRCC = 0.9715 ] +INFO:root:Average loss=0.02850415512919426 +INFO:root: +[Valid][27][UTT][ MSE = 0.1499 | LCC = 0.9723 | SRCC = 0.9712 ] +INFO:root:Average loss=0.0298829288110137 +INFO:root: +[Valid][28][UTT][ MSE = 0.1978 | LCC = 0.9717 | SRCC = 0.9711 ] +INFO:root:Average loss=0.027851839184761047 +INFO:root: +[Valid][29][UTT][ MSE = 0.1782 | LCC = 0.9718 | SRCC = 0.9710 ] +INFO:root:Average loss=0.026445484638214112 +INFO:root: +[Valid][30][UTT][ MSE = 0.2009 | LCC = 0.9721 | SRCC = 0.9710 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1675 | LCC = 0.9742 | SRCC = 0.9740 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run3/config.gin b/results_tencent/w2v2_xlsr_1b/layer16/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run3/model.pt b/results_tencent/w2v2_xlsr_1b/layer16/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..1e64ba766f4e92fbfb8a0cfe00bf105d418bc35b --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8258c0d2f28e9475483f9a96d1acda6efb97f10100872dc1326a04dc21efd0a3 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run3/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer16/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..cf83a7b47be1212cb0be9cc138dfd37d3688109c --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6d4fd7cfe84dd1efe7ac9f783b8287f08d4f03a416e01d11c0bd2afa133ff2de +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run3/train.log b/results_tencent/w2v2_xlsr_1b/layer16/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..89c0efc02d0fac1713ca81f7a6cfd74546d32638 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3903228822350502 +INFO:root: +[Valid][0][UTT][ MSE = 0.2336 | LCC = 0.9484 | SRCC = 0.9483 ] +INFO:root:Average loss=0.1371130431294441 +INFO:root: +[Valid][1][UTT][ MSE = 0.1896 | LCC = 0.9607 | SRCC = 0.9602 ] +INFO:root:Average loss=0.1177017804980278 +INFO:root: +[Valid][2][UTT][ MSE = 0.1938 | LCC = 0.9639 | SRCC = 0.9640 ] +INFO:root:Average loss=0.09743786326050759 +INFO:root: +[Valid][3][UTT][ MSE = 0.1933 | LCC = 0.9648 | SRCC = 0.9655 ] +INFO:root:Average loss=0.08984773552417755 +INFO:root: +[Valid][4][UTT][ MSE = 0.2093 | LCC = 0.9659 | SRCC = 0.9649 ] +INFO:root:Average loss=0.08339403593540191 +INFO:root: +[Valid][5][UTT][ MSE = 0.1800 | LCC = 0.9637 | SRCC = 0.9671 ] +INFO:root:Average loss=0.07650478914380074 +INFO:root: +[Valid][6][UTT][ MSE = 0.1634 | LCC = 0.9689 | SRCC = 0.9689 ] +INFO:root:Average loss=0.07295277896523476 +INFO:root: +[Valid][7][UTT][ MSE = 0.1539 | LCC = 0.9694 | SRCC = 0.9695 ] +INFO:root:Average loss=0.0687447170317173 +INFO:root: +[Valid][8][UTT][ MSE = 0.1583 | LCC = 0.9711 | SRCC = 0.9709 ] +INFO:root:Average loss=0.06170398110151291 +INFO:root: +[Valid][9][UTT][ MSE = 0.1603 | LCC = 0.9710 | SRCC = 0.9702 ] +INFO:root:Average loss=0.05993828511238098 +INFO:root: +[Valid][10][UTT][ MSE = 0.1567 | LCC = 0.9710 | SRCC = 0.9710 ] +INFO:root:Average loss=0.05586001719534397 +INFO:root: +[Valid][11][UTT][ MSE = 0.1592 | LCC = 0.9709 | SRCC = 0.9708 ] +INFO:root:Average loss=0.052456519290804866 +INFO:root: +[Valid][12][UTT][ MSE = 0.1823 | LCC = 0.9711 | SRCC = 0.9704 ] +INFO:root:Average loss=0.051123328983783724 +INFO:root: +[Valid][13][UTT][ MSE = 0.1626 | LCC = 0.9718 | SRCC = 0.9709 ] +INFO:root:Average loss=0.04962517528235912 +INFO:root: +[Valid][14][UTT][ MSE = 0.1441 | LCC = 0.9714 | SRCC = 0.9707 ] +INFO:root:Average loss=0.04545102322101593 +INFO:root: +[Valid][15][UTT][ MSE = 0.1629 | LCC = 0.9715 | SRCC = 0.9701 ] +INFO:root:Average loss=0.04425671137869358 +INFO:root: +[Valid][16][UTT][ MSE = 0.1583 | LCC = 0.9719 | SRCC = 0.9708 ] +INFO:root:Average loss=0.041817469075322154 +INFO:root: +[Valid][17][UTT][ MSE = 0.1560 | LCC = 0.9717 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03920710581541061 +INFO:root: +[Valid][18][UTT][ MSE = 0.1464 | LCC = 0.9701 | SRCC = 0.9697 ] +INFO:root:Average loss=0.04085639514029026 +INFO:root: +[Valid][19][UTT][ MSE = 0.1556 | LCC = 0.9725 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03738797891139984 +INFO:root: +[Valid][20][UTT][ MSE = 0.1517 | LCC = 0.9722 | SRCC = 0.9710 ] +INFO:root:Average loss=0.03552234362065792 +INFO:root: +[Valid][21][UTT][ MSE = 0.1617 | LCC = 0.9717 | SRCC = 0.9706 ] +INFO:root:Average loss=0.035254885077476504 +INFO:root: +[Valid][22][UTT][ MSE = 0.1775 | LCC = 0.9724 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03198880012333393 +INFO:root: +[Valid][23][UTT][ MSE = 0.1807 | LCC = 0.9723 | SRCC = 0.9709 ] +INFO:root:Average loss=0.030698479428887367 +INFO:root: +[Valid][24][UTT][ MSE = 0.1549 | LCC = 0.9716 | SRCC = 0.9702 ] +INFO:root:Average loss=0.031150805845856666 +INFO:root: +[Valid][25][UTT][ MSE = 0.1655 | LCC = 0.9714 | SRCC = 0.9703 ] +INFO:root:Average loss=0.0289570282548666 +INFO:root: +[Valid][26][UTT][ MSE = 0.1631 | LCC = 0.9713 | SRCC = 0.9699 ] +INFO:root:Average loss=0.02951703941822052 +INFO:root: +[Valid][27][UTT][ MSE = 0.1528 | LCC = 0.9712 | SRCC = 0.9706 ] +INFO:root:Average loss=0.026915485724806787 +INFO:root: +[Valid][28][UTT][ MSE = 0.1649 | LCC = 0.9697 | SRCC = 0.9699 ] +INFO:root:Average loss=0.02617525954544544 +INFO:root: +[Valid][29][UTT][ MSE = 0.1598 | LCC = 0.9713 | SRCC = 0.9704 ] +INFO:root:Average loss=0.02651814027130604 +INFO:root: +[Valid][30][UTT][ MSE = 0.1573 | LCC = 0.9714 | SRCC = 0.9707 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1517 | LCC = 0.9736 | SRCC = 0.9733 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run4/config.gin b/results_tencent/w2v2_xlsr_1b/layer16/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run4/model.pt b/results_tencent/w2v2_xlsr_1b/layer16/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6931de900709f68e3e17e2dbba4dc0952b86c280 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:523a067d701c73f2f854a38b1965baea6651afaed05d1aa9a22bd2a085331d37 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run4/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer16/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e886537046e5b1d335ca82ac6123789e28dbb3f5 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:447cfe8407a96c39b61e97cf823f424f5ec910c3b0c0ec692728874e2da88ec8 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer16/run4/train.log b/results_tencent/w2v2_xlsr_1b/layer16/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b5e87dc338476f1fe79f3f5ef6b5428c81790ddf --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer16/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3919012569785118 +INFO:root: +[Valid][0][UTT][ MSE = 0.2958 | LCC = 0.9417 | SRCC = 0.9437 ] +INFO:root:Average loss=0.1558797020316124 +INFO:root: +[Valid][1][UTT][ MSE = 0.1893 | LCC = 0.9574 | SRCC = 0.9564 ] +INFO:root:Average loss=0.11004917353391647 +INFO:root: +[Valid][2][UTT][ MSE = 0.2125 | LCC = 0.9631 | SRCC = 0.9626 ] +INFO:root:Average loss=0.09491517975926399 +INFO:root: +[Valid][3][UTT][ MSE = 0.2443 | LCC = 0.9660 | SRCC = 0.9654 ] +INFO:root:Average loss=0.08747583225369454 +INFO:root: +[Valid][4][UTT][ MSE = 0.1960 | LCC = 0.9669 | SRCC = 0.9667 ] +INFO:root:Average loss=0.07929211413860321 +INFO:root: +[Valid][5][UTT][ MSE = 0.2191 | LCC = 0.9684 | SRCC = 0.9674 ] +INFO:root:Average loss=0.07363135686516761 +INFO:root: +[Valid][6][UTT][ MSE = 0.2098 | LCC = 0.9697 | SRCC = 0.9688 ] +INFO:root:Average loss=0.07083866757154465 +INFO:root: +[Valid][7][UTT][ MSE = 0.1739 | LCC = 0.9692 | SRCC = 0.9693 ] +INFO:root:Average loss=0.06486295643448829 +INFO:root: +[Valid][8][UTT][ MSE = 0.1647 | LCC = 0.9694 | SRCC = 0.9691 ] +INFO:root:Average loss=0.06338139042258263 +INFO:root: +[Valid][9][UTT][ MSE = 0.1979 | LCC = 0.9712 | SRCC = 0.9703 ] +INFO:root:Average loss=0.05810117855668068 +INFO:root: +[Valid][10][UTT][ MSE = 0.1748 | LCC = 0.9713 | SRCC = 0.9702 ] +INFO:root:Average loss=0.054110854044556615 +INFO:root: +[Valid][11][UTT][ MSE = 0.1790 | LCC = 0.9711 | SRCC = 0.9699 ] +INFO:root:Average loss=0.05444037607312203 +INFO:root: +[Valid][12][UTT][ MSE = 0.1635 | LCC = 0.9716 | SRCC = 0.9708 ] +INFO:root:Average loss=0.04965738090872764 +INFO:root: +[Valid][13][UTT][ MSE = 0.1603 | LCC = 0.9711 | SRCC = 0.9702 ] +INFO:root:Average loss=0.04518763367831707 +INFO:root: +[Valid][14][UTT][ MSE = 0.1871 | LCC = 0.9718 | SRCC = 0.9704 ] +INFO:root:Average loss=0.04492381440103054 +INFO:root: +[Valid][15][UTT][ MSE = 0.1754 | LCC = 0.9716 | SRCC = 0.9707 ] +INFO:root:Average loss=0.043857019364833834 +INFO:root: +[Valid][16][UTT][ MSE = 0.1551 | LCC = 0.9722 | SRCC = 0.9712 ] +INFO:root:Average loss=0.041066671013832094 +INFO:root: +[Valid][17][UTT][ MSE = 0.1859 | LCC = 0.9715 | SRCC = 0.9703 ] +INFO:root:Average loss=0.04018837851285934 +INFO:root: +[Valid][18][UTT][ MSE = 0.1666 | LCC = 0.9722 | SRCC = 0.9710 ] +INFO:root:Average loss=0.039408706903457644 +INFO:root: +[Valid][19][UTT][ MSE = 0.1938 | LCC = 0.9714 | SRCC = 0.9700 ] +INFO:root:Average loss=0.03469209612905979 +INFO:root: +[Valid][20][UTT][ MSE = 0.1624 | LCC = 0.9720 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03495200857520103 +INFO:root: +[Valid][21][UTT][ MSE = 0.1789 | LCC = 0.9720 | SRCC = 0.9705 ] +INFO:root:Average loss=0.033640964329242705 +INFO:root: +[Valid][22][UTT][ MSE = 0.1816 | LCC = 0.9718 | SRCC = 0.9705 ] +INFO:root:Average loss=0.03237354877591133 +INFO:root: +[Valid][23][UTT][ MSE = 0.1658 | LCC = 0.9719 | SRCC = 0.9705 ] +INFO:root:Average loss=0.030855123817920684 +INFO:root: +[Valid][24][UTT][ MSE = 0.1934 | LCC = 0.9720 | SRCC = 0.9704 ] +INFO:root:Average loss=0.028291214153170586 +INFO:root: +[Valid][25][UTT][ MSE = 0.1564 | LCC = 0.9715 | SRCC = 0.9702 ] +INFO:root:Average loss=0.02874122370779514 +INFO:root: +[Valid][26][UTT][ MSE = 0.1717 | LCC = 0.9722 | SRCC = 0.9709 ] +INFO:root:Average loss=0.027497541189193727 +INFO:root: +[Valid][27][UTT][ MSE = 0.1622 | LCC = 0.9713 | SRCC = 0.9698 ] +INFO:root:Average loss=0.02612943008542061 +INFO:root: +[Valid][28][UTT][ MSE = 0.1658 | LCC = 0.9719 | SRCC = 0.9703 ] +INFO:root:Average loss=0.024404166489839552 +INFO:root: +[Valid][29][UTT][ MSE = 0.1453 | LCC = 0.9725 | SRCC = 0.9711 ] +INFO:root:Average loss=0.024349979668855666 +INFO:root: +[Valid][30][UTT][ MSE = 0.1782 | LCC = 0.9723 | SRCC = 0.9704 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1390 | LCC = 0.9742 | SRCC = 0.9740 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run0/config.gin b/results_tencent/w2v2_xlsr_1b/layer18/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run0/model.pt b/results_tencent/w2v2_xlsr_1b/layer18/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..67aa6fe13662d5d1661d17a1620f769b69069493 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b0a7032dceeb68a3291437b487ca5be631a9f80fc7e1ac8860f8838fed745ef +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run0/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer18/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..51eb846bed7354981c415503136cd4119c6e8d83 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b0c641bace44c2adc961ea12877d599485cbfbda46b443d3f82a3f6e95442b5 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run0/train.log b/results_tencent/w2v2_xlsr_1b/layer18/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c636f00ea86fb8aff4d6451450fd01fcd711ff54 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.39884474098682404 +INFO:root: +[Valid][0][UTT][ MSE = 0.2834 | LCC = 0.9493 | SRCC = 0.9485 ] +INFO:root:Average loss=0.1412142089009285 +INFO:root: +[Valid][1][UTT][ MSE = 0.1923 | LCC = 0.9581 | SRCC = 0.9579 ] +INFO:root:Average loss=0.11518529185652733 +INFO:root: +[Valid][2][UTT][ MSE = 0.1971 | LCC = 0.9619 | SRCC = 0.9623 ] +INFO:root:Average loss=0.10273287037014961 +INFO:root: +[Valid][3][UTT][ MSE = 0.1781 | LCC = 0.9664 | SRCC = 0.9658 ] +INFO:root:Average loss=0.09212962293624878 +INFO:root: +[Valid][4][UTT][ MSE = 0.1680 | LCC = 0.9669 | SRCC = 0.9669 ] +INFO:root:Average loss=0.08315143415331841 +INFO:root: +[Valid][5][UTT][ MSE = 0.2462 | LCC = 0.9656 | SRCC = 0.9673 ] +INFO:root:Average loss=0.08015126159787178 +INFO:root: +[Valid][6][UTT][ MSE = 0.1590 | LCC = 0.9697 | SRCC = 0.9689 ] +INFO:root:Average loss=0.07553305038809777 +INFO:root: +[Valid][7][UTT][ MSE = 0.2043 | LCC = 0.9701 | SRCC = 0.9696 ] +INFO:root:Average loss=0.07259776949882507 +INFO:root: +[Valid][8][UTT][ MSE = 0.1789 | LCC = 0.9696 | SRCC = 0.9697 ] +INFO:root:Average loss=0.06474570015072822 +INFO:root: +[Valid][9][UTT][ MSE = 0.1797 | LCC = 0.9699 | SRCC = 0.9698 ] +INFO:root:Average loss=0.06492722517251968 +INFO:root: +[Valid][10][UTT][ MSE = 0.1746 | LCC = 0.9715 | SRCC = 0.9709 ] +INFO:root:Average loss=0.05773082441091538 +INFO:root: +[Valid][11][UTT][ MSE = 0.1780 | LCC = 0.9714 | SRCC = 0.9711 ] +INFO:root:Average loss=0.05734859624505043 +INFO:root: +[Valid][12][UTT][ MSE = 0.1930 | LCC = 0.9714 | SRCC = 0.9711 ] +INFO:root:Average loss=0.052310083270072937 +INFO:root: +[Valid][13][UTT][ MSE = 0.1535 | LCC = 0.9716 | SRCC = 0.9708 ] +INFO:root:Average loss=0.05217196674644947 +INFO:root: +[Valid][14][UTT][ MSE = 0.1881 | LCC = 0.9716 | SRCC = 0.9709 ] +INFO:root:Average loss=0.047976942673325536 +INFO:root: +[Valid][15][UTT][ MSE = 0.1620 | LCC = 0.9707 | SRCC = 0.9711 ] +INFO:root:Average loss=0.046127408340573314 +INFO:root: +[Valid][16][UTT][ MSE = 0.1742 | LCC = 0.9723 | SRCC = 0.9716 ] +INFO:root:Average loss=0.043036157503724096 +INFO:root: +[Valid][17][UTT][ MSE = 0.1762 | LCC = 0.9717 | SRCC = 0.9702 ] +INFO:root:Average loss=0.04131601446866989 +INFO:root: +[Valid][18][UTT][ MSE = 0.1544 | LCC = 0.9715 | SRCC = 0.9708 ] +INFO:root:Average loss=0.043526227965950966 +INFO:root: +[Valid][19][UTT][ MSE = 0.1596 | LCC = 0.9720 | SRCC = 0.9712 ] +INFO:root:Average loss=0.03880362224578857 +INFO:root: +[Valid][20][UTT][ MSE = 0.1460 | LCC = 0.9713 | SRCC = 0.9708 ] +INFO:root:Average loss=0.0370343055576086 +INFO:root: +[Valid][21][UTT][ MSE = 0.1373 | LCC = 0.9713 | SRCC = 0.9704 ] +INFO:root:Average loss=0.035548977494239804 +INFO:root: +[Valid][22][UTT][ MSE = 0.1797 | LCC = 0.9712 | SRCC = 0.9705 ] +INFO:root:Average loss=0.033936493933200834 +INFO:root: +[Valid][23][UTT][ MSE = 0.1638 | LCC = 0.9721 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03137627817690372 +INFO:root: +[Valid][24][UTT][ MSE = 0.1641 | LCC = 0.9722 | SRCC = 0.9707 ] +INFO:root:Average loss=0.029639830872416496 +INFO:root: +[Valid][25][UTT][ MSE = 0.1610 | LCC = 0.9722 | SRCC = 0.9709 ] +INFO:root:Average loss=0.02813432654738426 +INFO:root: +[Valid][26][UTT][ MSE = 0.1726 | LCC = 0.9719 | SRCC = 0.9708 ] +INFO:root:Average loss=0.029613826662302018 +INFO:root: +[Valid][27][UTT][ MSE = 0.1477 | LCC = 0.9712 | SRCC = 0.9706 ] +INFO:root:Average loss=0.02752702385932207 +INFO:root: +[Valid][28][UTT][ MSE = 0.1499 | LCC = 0.9718 | SRCC = 0.9704 ] +INFO:root:Average loss=0.02626193302869797 +INFO:root: +[Valid][29][UTT][ MSE = 0.1599 | LCC = 0.9725 | SRCC = 0.9710 ] +INFO:root:Average loss=0.02522966130077839 +INFO:root: +[Valid][30][UTT][ MSE = 0.1517 | LCC = 0.9723 | SRCC = 0.9710 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1566 | LCC = 0.9734 | SRCC = 0.9732 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run1/config.gin b/results_tencent/w2v2_xlsr_1b/layer18/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run1/model.pt b/results_tencent/w2v2_xlsr_1b/layer18/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2e06a8c38e96b61768126a4eb60c23de30f1e52f --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6b39b4b597d71dc9d21f5195657bf5958eaea7e443ae548ab8393605b18651bc +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run1/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer18/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..83c6a12126ec9511964447e12e959196929f9067 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f25ec075aedbfc8aabe81a3083ae5d1786590c668c123b4aaf8d1be93390a311 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run1/train.log b/results_tencent/w2v2_xlsr_1b/layer18/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..26fab54372d2cc04a8fd7348ed4949d4417bb3ba --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.37989562457799914 +INFO:root: +[Valid][0][UTT][ MSE = 0.2830 | LCC = 0.9506 | SRCC = 0.9516 ] +INFO:root:Average loss=0.1419768278002739 +INFO:root: +[Valid][1][UTT][ MSE = 0.2316 | LCC = 0.9599 | SRCC = 0.9589 ] +INFO:root:Average loss=0.11243384492397308 +INFO:root: +[Valid][2][UTT][ MSE = 0.2461 | LCC = 0.9636 | SRCC = 0.9632 ] +INFO:root:Average loss=0.10229746532440186 +INFO:root: +[Valid][3][UTT][ MSE = 0.1662 | LCC = 0.9681 | SRCC = 0.9675 ] +INFO:root:Average loss=0.08725090724229813 +INFO:root: +[Valid][4][UTT][ MSE = 0.1954 | LCC = 0.9695 | SRCC = 0.9693 ] +INFO:root:Average loss=0.08081582367420197 +INFO:root: +[Valid][5][UTT][ MSE = 0.1797 | LCC = 0.9686 | SRCC = 0.9680 ] +INFO:root:Average loss=0.07458240932226182 +INFO:root: +[Valid][6][UTT][ MSE = 0.1478 | LCC = 0.9708 | SRCC = 0.9700 ] +INFO:root:Average loss=0.06936443981528283 +INFO:root: +[Valid][7][UTT][ MSE = 0.1907 | LCC = 0.9714 | SRCC = 0.9709 ] +INFO:root:Average loss=0.06694496464729309 +INFO:root: +[Valid][8][UTT][ MSE = 0.1788 | LCC = 0.9705 | SRCC = 0.9704 ] +INFO:root:Average loss=0.06236984372138977 +INFO:root: +[Valid][9][UTT][ MSE = 0.2051 | LCC = 0.9694 | SRCC = 0.9698 ] +INFO:root:Average loss=0.061299318581819535 +INFO:root: +[Valid][10][UTT][ MSE = 0.1713 | LCC = 0.9712 | SRCC = 0.9700 ] +INFO:root:Average loss=0.057250167042016985 +INFO:root: +[Valid][11][UTT][ MSE = 0.1906 | LCC = 0.9704 | SRCC = 0.9707 ] +INFO:root:Average loss=0.055491302341222765 +INFO:root: +[Valid][12][UTT][ MSE = 0.1496 | LCC = 0.9709 | SRCC = 0.9700 ] +INFO:root:Average loss=0.05156465290486813 +INFO:root: +[Valid][13][UTT][ MSE = 0.2022 | LCC = 0.9706 | SRCC = 0.9696 ] +INFO:root:Average loss=0.04940430757403374 +INFO:root: +[Valid][14][UTT][ MSE = 0.1655 | LCC = 0.9715 | SRCC = 0.9707 ] +INFO:root:Average loss=0.0483424926251173 +INFO:root: +[Valid][15][UTT][ MSE = 0.1638 | LCC = 0.9698 | SRCC = 0.9689 ] +INFO:root:Average loss=0.04486964657902717 +INFO:root: +[Valid][16][UTT][ MSE = 0.1720 | LCC = 0.9717 | SRCC = 0.9704 ] +INFO:root:Average loss=0.04500946176052094 +INFO:root: +[Valid][17][UTT][ MSE = 0.1766 | LCC = 0.9717 | SRCC = 0.9706 ] +INFO:root:Average loss=0.0405761690735817 +INFO:root: +[Valid][18][UTT][ MSE = 0.1720 | LCC = 0.9714 | SRCC = 0.9704 ] +INFO:root:Average loss=0.03992029596865177 +INFO:root: +[Valid][19][UTT][ MSE = 0.1621 | LCC = 0.9713 | SRCC = 0.9705 ] +INFO:root:Average loss=0.037100955814123156 +INFO:root: +[Valid][20][UTT][ MSE = 0.1585 | LCC = 0.9717 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03560165393352509 +INFO:root: +[Valid][21][UTT][ MSE = 0.1490 | LCC = 0.9722 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03352496050298214 +INFO:root: +[Valid][22][UTT][ MSE = 0.1581 | LCC = 0.9719 | SRCC = 0.9704 ] +INFO:root:Average loss=0.0337829060703516 +INFO:root: +[Valid][23][UTT][ MSE = 0.1550 | LCC = 0.9715 | SRCC = 0.9697 ] +INFO:root:Average loss=0.03194542199373245 +INFO:root: +[Valid][24][UTT][ MSE = 0.1421 | LCC = 0.9719 | SRCC = 0.9707 ] +INFO:root:Average loss=0.02971413537859917 +INFO:root: +[Valid][25][UTT][ MSE = 0.1654 | LCC = 0.9721 | SRCC = 0.9705 ] +INFO:root:Average loss=0.02888640612363815 +INFO:root: +[Valid][26][UTT][ MSE = 0.1386 | LCC = 0.9721 | SRCC = 0.9707 ] +INFO:root:Average loss=0.03018641585111618 +INFO:root: +[Valid][27][UTT][ MSE = 0.1504 | LCC = 0.9721 | SRCC = 0.9708 ] +INFO:root:Average loss=0.027414528593420983 +INFO:root: +[Valid][28][UTT][ MSE = 0.1563 | LCC = 0.9721 | SRCC = 0.9704 ] +INFO:root:Average loss=0.029037168703973295 +INFO:root: +[Valid][29][UTT][ MSE = 0.1586 | LCC = 0.9719 | SRCC = 0.9706 ] +INFO:root:Average loss=0.025471188083291053 +INFO:root: +[Valid][30][UTT][ MSE = 0.1492 | LCC = 0.9725 | SRCC = 0.9710 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1448 | LCC = 0.9745 | SRCC = 0.9741 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run2/config.gin b/results_tencent/w2v2_xlsr_1b/layer18/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run2/model.pt b/results_tencent/w2v2_xlsr_1b/layer18/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a601aea1d388063a11f5d674b042f9a7c7d19779 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:42bc0a561df81b51e109aa7207ae06b155ff58e0d44ba6b7f53149ba7334d2ee +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run2/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer18/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c9cc8b1acb77b1208083a61e768ddf31d27afa3a --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:080c5f9fabba933814b2d49d0374a947d3caddb2ed1abf38431170343e8df72e +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run2/train.log b/results_tencent/w2v2_xlsr_1b/layer18/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b01a85dc6fcf1f06b824edebc84df7297a491253 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3925053665041924 +INFO:root: +[Valid][0][UTT][ MSE = 0.2451 | LCC = 0.9471 | SRCC = 0.9475 ] +INFO:root:Average loss=0.1413901834487915 +INFO:root: +[Valid][1][UTT][ MSE = 0.1952 | LCC = 0.9582 | SRCC = 0.9578 ] +INFO:root:Average loss=0.11275727677345276 +INFO:root: +[Valid][2][UTT][ MSE = 0.2157 | LCC = 0.9610 | SRCC = 0.9613 ] +INFO:root:Average loss=0.10121078330278396 +INFO:root: +[Valid][3][UTT][ MSE = 0.1482 | LCC = 0.9677 | SRCC = 0.9672 ] +INFO:root:Average loss=0.09649816635251045 +INFO:root: +[Valid][4][UTT][ MSE = 0.1795 | LCC = 0.9671 | SRCC = 0.9676 ] +INFO:root:Average loss=0.08229172486066819 +INFO:root: +[Valid][5][UTT][ MSE = 0.1606 | LCC = 0.9705 | SRCC = 0.9700 ] +INFO:root:Average loss=0.08299995198845864 +INFO:root: +[Valid][6][UTT][ MSE = 0.1747 | LCC = 0.9692 | SRCC = 0.9679 ] +INFO:root:Average loss=0.07073028150200844 +INFO:root: +[Valid][7][UTT][ MSE = 0.2261 | LCC = 0.9696 | SRCC = 0.9691 ] +INFO:root:Average loss=0.06983058241009713 +INFO:root: +[Valid][8][UTT][ MSE = 0.1327 | LCC = 0.9702 | SRCC = 0.9699 ] +INFO:root:Average loss=0.06414890801906585 +INFO:root: +[Valid][9][UTT][ MSE = 0.1775 | LCC = 0.9717 | SRCC = 0.9708 ] +INFO:root:Average loss=0.05832006028294563 +INFO:root: +[Valid][10][UTT][ MSE = 0.1478 | LCC = 0.9708 | SRCC = 0.9698 ] +INFO:root:Average loss=0.058950645238161084 +INFO:root: +[Valid][11][UTT][ MSE = 0.1482 | LCC = 0.9715 | SRCC = 0.9708 ] +INFO:root:Average loss=0.060073359727859495 +INFO:root: +[Valid][12][UTT][ MSE = 0.1723 | LCC = 0.9723 | SRCC = 0.9714 ] +INFO:root:Average loss=0.052542269572615624 +INFO:root: +[Valid][13][UTT][ MSE = 0.1724 | LCC = 0.9711 | SRCC = 0.9708 ] +INFO:root:Average loss=0.05047399753332138 +INFO:root: +[Valid][14][UTT][ MSE = 0.1581 | LCC = 0.9720 | SRCC = 0.9712 ] +INFO:root:Average loss=0.04902419422566891 +INFO:root: +[Valid][15][UTT][ MSE = 0.1318 | LCC = 0.9717 | SRCC = 0.9702 ] +INFO:root:Average loss=0.04758783778548241 +INFO:root: +[Valid][16][UTT][ MSE = 0.1584 | LCC = 0.9717 | SRCC = 0.9709 ] +INFO:root:Average loss=0.04241194517910481 +INFO:root: +[Valid][17][UTT][ MSE = 0.1512 | LCC = 0.9727 | SRCC = 0.9716 ] +INFO:root:Average loss=0.040653066724538806 +INFO:root: +[Valid][18][UTT][ MSE = 0.1829 | LCC = 0.9726 | SRCC = 0.9717 ] +INFO:root:Average loss=0.04087791542708874 +INFO:root: +[Valid][19][UTT][ MSE = 0.1342 | LCC = 0.9720 | SRCC = 0.9711 ] +INFO:root:Average loss=0.037849420785903934 +INFO:root: +[Valid][20][UTT][ MSE = 0.1615 | LCC = 0.9724 | SRCC = 0.9714 ] +INFO:root:Average loss=0.036635195642709735 +INFO:root: +[Valid][21][UTT][ MSE = 0.1299 | LCC = 0.9722 | SRCC = 0.9714 ] +INFO:root:Average loss=0.03437546981871128 +INFO:root: +[Valid][22][UTT][ MSE = 0.1451 | LCC = 0.9722 | SRCC = 0.9711 ] +INFO:root:Average loss=0.03613059991598129 +INFO:root: +[Valid][23][UTT][ MSE = 0.1559 | LCC = 0.9711 | SRCC = 0.9700 ] +INFO:root:Average loss=0.032539474725723265 +INFO:root: +[Valid][24][UTT][ MSE = 0.1566 | LCC = 0.9724 | SRCC = 0.9713 ] +INFO:root:Average loss=0.032748414158821106 +INFO:root: +[Valid][25][UTT][ MSE = 0.1595 | LCC = 0.9723 | SRCC = 0.9706 ] +INFO:root:Average loss=0.0297414980083704 +INFO:root: +[Valid][26][UTT][ MSE = 0.1542 | LCC = 0.9714 | SRCC = 0.9709 ] +INFO:root:Average loss=0.02845164756476879 +INFO:root: +[Valid][27][UTT][ MSE = 0.1569 | LCC = 0.9721 | SRCC = 0.9711 ] +INFO:root:Average loss=0.028726099088788034 +INFO:root: +[Valid][28][UTT][ MSE = 0.1558 | LCC = 0.9718 | SRCC = 0.9703 ] +INFO:root:Average loss=0.027957935251295567 +INFO:root: +[Valid][29][UTT][ MSE = 0.1398 | LCC = 0.9708 | SRCC = 0.9697 ] +INFO:root:Average loss=0.02747870060801506 +INFO:root: +[Valid][30][UTT][ MSE = 0.1713 | LCC = 0.9712 | SRCC = 0.9698 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1492 | LCC = 0.9731 | SRCC = 0.9729 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run3/config.gin b/results_tencent/w2v2_xlsr_1b/layer18/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run3/model.pt b/results_tencent/w2v2_xlsr_1b/layer18/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..20f2488a753237da38efb5c00fa9ec277bb03c42 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:88f47e4d68e9edc0088698c5b3a7bd8be6b28689dcfb2b9ab481098cb05eec17 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run3/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer18/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..1291ec5ef07016b76d9f900b0ecf31ded1c41e95 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c9b842cf61adaf7b78def652295cf74cc4091904396a8d95cf1dcc3c11edfd3a +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run3/train.log b/results_tencent/w2v2_xlsr_1b/layer18/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5c062b6a9bd13ce4d95dec9ae078fce78e551a35 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.36170930057764056 +INFO:root: +[Valid][0][UTT][ MSE = 0.2398 | LCC = 0.9507 | SRCC = 0.9494 ] +INFO:root:Average loss=0.13917689377069473 +INFO:root: +[Valid][1][UTT][ MSE = 0.2375 | LCC = 0.9599 | SRCC = 0.9606 ] +INFO:root:Average loss=0.11460282504558564 +INFO:root: +[Valid][2][UTT][ MSE = 0.2171 | LCC = 0.9629 | SRCC = 0.9613 ] +INFO:root:Average loss=0.10185482794046402 +INFO:root: +[Valid][3][UTT][ MSE = 0.1804 | LCC = 0.9662 | SRCC = 0.9648 ] +INFO:root:Average loss=0.09181802582740783 +INFO:root: +[Valid][4][UTT][ MSE = 0.2317 | LCC = 0.9659 | SRCC = 0.9663 ] +INFO:root:Average loss=0.08854944729804992 +INFO:root: +[Valid][5][UTT][ MSE = 0.2329 | LCC = 0.9676 | SRCC = 0.9669 ] +INFO:root:Average loss=0.07867704138159752 +INFO:root: +[Valid][6][UTT][ MSE = 0.1619 | LCC = 0.9700 | SRCC = 0.9693 ] +INFO:root:Average loss=0.07154108920693397 +INFO:root: +[Valid][7][UTT][ MSE = 0.1690 | LCC = 0.9705 | SRCC = 0.9696 ] +INFO:root:Average loss=0.06829545211791992 +INFO:root: +[Valid][8][UTT][ MSE = 0.1898 | LCC = 0.9707 | SRCC = 0.9704 ] +INFO:root:Average loss=0.06807989576458931 +INFO:root: +[Valid][9][UTT][ MSE = 0.2188 | LCC = 0.9711 | SRCC = 0.9701 ] +INFO:root:Average loss=0.060547535449266436 +INFO:root: +[Valid][10][UTT][ MSE = 0.1750 | LCC = 0.9711 | SRCC = 0.9703 ] +INFO:root:Average loss=0.05845960062742233 +INFO:root: +[Valid][11][UTT][ MSE = 0.1818 | LCC = 0.9722 | SRCC = 0.9712 ] +INFO:root:Average loss=0.05519457820057869 +INFO:root: +[Valid][12][UTT][ MSE = 0.1419 | LCC = 0.9711 | SRCC = 0.9703 ] +INFO:root:Average loss=0.053685204684734346 +INFO:root: +[Valid][13][UTT][ MSE = 0.1914 | LCC = 0.9714 | SRCC = 0.9707 ] +INFO:root:Average loss=0.05093290218710899 +INFO:root: +[Valid][14][UTT][ MSE = 0.1745 | LCC = 0.9709 | SRCC = 0.9702 ] +INFO:root:Average loss=0.046426570445299145 +INFO:root: +[Valid][15][UTT][ MSE = 0.1723 | LCC = 0.9720 | SRCC = 0.9709 ] +INFO:root:Average loss=0.04521692016720772 +INFO:root: +[Valid][16][UTT][ MSE = 0.1776 | LCC = 0.9706 | SRCC = 0.9712 ] +INFO:root:Average loss=0.04467504341900349 +INFO:root: +[Valid][17][UTT][ MSE = 0.1642 | LCC = 0.9718 | SRCC = 0.9706 ] +INFO:root:Average loss=0.04253569138050079 +INFO:root: +[Valid][18][UTT][ MSE = 0.1589 | LCC = 0.9726 | SRCC = 0.9710 ] +INFO:root:Average loss=0.04046110601723194 +INFO:root: +[Valid][19][UTT][ MSE = 0.1757 | LCC = 0.9724 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03958019514381886 +INFO:root: +[Valid][20][UTT][ MSE = 0.1487 | LCC = 0.9721 | SRCC = 0.9706 ] +INFO:root:Average loss=0.037114737272262575 +INFO:root: +[Valid][21][UTT][ MSE = 0.1762 | LCC = 0.9719 | SRCC = 0.9706 ] +INFO:root:Average loss=0.037654796421527865 +INFO:root: +[Valid][22][UTT][ MSE = 0.1469 | LCC = 0.9720 | SRCC = 0.9705 ] +INFO:root:Average loss=0.03376363954693079 +INFO:root: +[Valid][23][UTT][ MSE = 0.1506 | LCC = 0.9718 | SRCC = 0.9703 ] +INFO:root:Average loss=0.0315698407292366 +INFO:root: +[Valid][24][UTT][ MSE = 0.1691 | LCC = 0.9717 | SRCC = 0.9700 ] +INFO:root:Average loss=0.029160205744206906 +INFO:root: +[Valid][25][UTT][ MSE = 0.1800 | LCC = 0.9711 | SRCC = 0.9694 ] +INFO:root:Average loss=0.028556211724877358 +INFO:root: +[Valid][26][UTT][ MSE = 0.1613 | LCC = 0.9719 | SRCC = 0.9705 ] +INFO:root:Average loss=0.027459448739886285 +INFO:root: +[Valid][27][UTT][ MSE = 0.1554 | LCC = 0.9718 | SRCC = 0.9701 ] +INFO:root:Average loss=0.028691038891673088 +INFO:root: +[Valid][28][UTT][ MSE = 0.1489 | LCC = 0.9720 | SRCC = 0.9707 ] +INFO:root:Average loss=0.02903450771421194 +INFO:root: +[Valid][29][UTT][ MSE = 0.1497 | LCC = 0.9712 | SRCC = 0.9702 ] +INFO:root:Average loss=0.02512928257137537 +INFO:root: +[Valid][30][UTT][ MSE = 0.1578 | LCC = 0.9719 | SRCC = 0.9703 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1564 | LCC = 0.9735 | SRCC = 0.9734 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run4/config.gin b/results_tencent/w2v2_xlsr_1b/layer18/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run4/model.pt b/results_tencent/w2v2_xlsr_1b/layer18/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..be331202adda5b47b1b1c63a516b94849079ccd6 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3bb1873f4f4eec039cef6bc6a6acb6ab61113a6725922b2da7b89adf164e64b8 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run4/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer18/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0da18ce3869d7cfff55ffc9859baf96b69dc2d14 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:abc5230fc7a5bb3ee8a8fe8b9deb00562abf735ede9e9f95558d0c150ed0a84c +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer18/run4/train.log b/results_tencent/w2v2_xlsr_1b/layer18/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..62a0a00c811bcb1e69030129a0e62d40fefbe3d9 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer18/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3476029076576233 +INFO:root: +[Valid][0][UTT][ MSE = 0.2493 | LCC = 0.9504 | SRCC = 0.9490 ] +INFO:root:Average loss=0.138854130089283 +INFO:root: +[Valid][1][UTT][ MSE = 0.2696 | LCC = 0.9565 | SRCC = 0.9575 ] +INFO:root:Average loss=0.13467363893985748 +INFO:root: +[Valid][2][UTT][ MSE = 0.2549 | LCC = 0.9601 | SRCC = 0.9606 ] +INFO:root:Average loss=0.10434150436520577 +INFO:root: +[Valid][3][UTT][ MSE = 0.2214 | LCC = 0.9639 | SRCC = 0.9635 ] +INFO:root:Average loss=0.09375260823965073 +INFO:root: +[Valid][4][UTT][ MSE = 0.2004 | LCC = 0.9661 | SRCC = 0.9655 ] +INFO:root:Average loss=0.08610571762919426 +INFO:root: +[Valid][5][UTT][ MSE = 0.2023 | LCC = 0.9679 | SRCC = 0.9668 ] +INFO:root:Average loss=0.07764945283532143 +INFO:root: +[Valid][6][UTT][ MSE = 0.2076 | LCC = 0.9681 | SRCC = 0.9674 ] +INFO:root:Average loss=0.07534728446602822 +INFO:root: +[Valid][7][UTT][ MSE = 0.1889 | LCC = 0.9676 | SRCC = 0.9668 ] +INFO:root:Average loss=0.06992820557951927 +INFO:root: +[Valid][8][UTT][ MSE = 0.1887 | LCC = 0.9690 | SRCC = 0.9683 ] +INFO:root:Average loss=0.06535612845420838 +INFO:root: +[Valid][9][UTT][ MSE = 0.2194 | LCC = 0.9689 | SRCC = 0.9682 ] +INFO:root:Average loss=0.06422428727149963 +INFO:root: +[Valid][10][UTT][ MSE = 0.2036 | LCC = 0.9700 | SRCC = 0.9692 ] +INFO:root:Average loss=0.0633074533045292 +INFO:root: +[Valid][11][UTT][ MSE = 0.1964 | LCC = 0.9700 | SRCC = 0.9694 ] +INFO:root:Average loss=0.058515494644641876 +INFO:root: +[Valid][12][UTT][ MSE = 0.1799 | LCC = 0.9704 | SRCC = 0.9697 ] +INFO:root:Average loss=0.053463618829846384 +INFO:root: +[Valid][13][UTT][ MSE = 0.2046 | LCC = 0.9695 | SRCC = 0.9690 ] +INFO:root:Average loss=0.05078811398148537 +INFO:root: +[Valid][14][UTT][ MSE = 0.1891 | LCC = 0.9705 | SRCC = 0.9692 ] +INFO:root:Average loss=0.047167347714304926 +INFO:root: +[Valid][15][UTT][ MSE = 0.2050 | LCC = 0.9700 | SRCC = 0.9695 ] +INFO:root:Average loss=0.04693391472101212 +INFO:root: +[Valid][16][UTT][ MSE = 0.1883 | LCC = 0.9709 | SRCC = 0.9700 ] +INFO:root:Average loss=0.044877503588795664 +INFO:root: +[Valid][17][UTT][ MSE = 0.1934 | LCC = 0.9712 | SRCC = 0.9704 ] +INFO:root:Average loss=0.04435712669789791 +INFO:root: +[Valid][18][UTT][ MSE = 0.2003 | LCC = 0.9706 | SRCC = 0.9699 ] +INFO:root:Average loss=0.040122104972600936 +INFO:root: +[Valid][19][UTT][ MSE = 0.1891 | LCC = 0.9707 | SRCC = 0.9694 ] +INFO:root:Average loss=0.039355655729770664 +INFO:root: +[Valid][20][UTT][ MSE = 0.1934 | LCC = 0.9706 | SRCC = 0.9698 ] +INFO:root:Average loss=0.03678679986298084 +INFO:root: +[Valid][21][UTT][ MSE = 0.1845 | LCC = 0.9695 | SRCC = 0.9687 ] +INFO:root:Average loss=0.037066786736249926 +INFO:root: +[Valid][22][UTT][ MSE = 0.1829 | LCC = 0.9711 | SRCC = 0.9702 ] +INFO:root:Average loss=0.03467199955880642 +INFO:root: +[Valid][23][UTT][ MSE = 0.1853 | LCC = 0.9713 | SRCC = 0.9701 ] +INFO:root:Average loss=0.036950561493635174 +INFO:root: +[Valid][24][UTT][ MSE = 0.1952 | LCC = 0.9714 | SRCC = 0.9703 ] +INFO:root:Average loss=0.03451171332597733 +INFO:root: +[Valid][25][UTT][ MSE = 0.2102 | LCC = 0.9707 | SRCC = 0.9700 ] +INFO:root:Average loss=0.03082791446149349 +INFO:root: +[Valid][26][UTT][ MSE = 0.2024 | LCC = 0.9713 | SRCC = 0.9700 ] +INFO:root:Average loss=0.029758317828178407 +INFO:root: +[Valid][27][UTT][ MSE = 0.1852 | LCC = 0.9712 | SRCC = 0.9702 ] +INFO:root:Average loss=0.02746470119059086 +INFO:root: +[Valid][28][UTT][ MSE = 0.1792 | LCC = 0.9707 | SRCC = 0.9694 ] +INFO:root:Average loss=0.026415343403816224 +INFO:root: +[Valid][29][UTT][ MSE = 0.1712 | LCC = 0.9712 | SRCC = 0.9704 ] +INFO:root:Average loss=0.02778579926490784 +INFO:root: +[Valid][30][UTT][ MSE = 0.1727 | LCC = 0.9710 | SRCC = 0.9700 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1907 | LCC = 0.9735 | SRCC = 0.9732 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run0/config.gin b/results_tencent/w2v2_xlsr_1b/layer2/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run0/model.pt b/results_tencent/w2v2_xlsr_1b/layer2/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8c49c40c8398bee61e409775b6459ec02519405f --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ac51012fa6daa4e5b18bddc3027e1d6af77c9df7dfe38a7d1977df8ec5ab1e3 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run0/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer2/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..312bc19b4b78161e0b163d34f1bf87dcf52ad3f4 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5977d261c45d8a492940277358a2a32d73d8523d4d1da50e15dddb681a7cbb05 +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run0/train.log b/results_tencent/w2v2_xlsr_1b/layer2/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..8e4ef0dc52bfac2c1a739d8880e2fc2c674229cd --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46757873630523683 +INFO:root: +[Valid][0][UTT][ MSE = 0.2790 | LCC = 0.9223 | SRCC = 0.9217 ] +INFO:root:Average loss=0.19927651447057723 +INFO:root: +[Valid][1][UTT][ MSE = 0.2452 | LCC = 0.9358 | SRCC = 0.9355 ] +INFO:root:Average loss=0.1594378947019577 +INFO:root: +[Valid][2][UTT][ MSE = 0.2294 | LCC = 0.9437 | SRCC = 0.9434 ] +INFO:root:Average loss=0.14650482177734375 +INFO:root: +[Valid][3][UTT][ MSE = 0.2092 | LCC = 0.9446 | SRCC = 0.9451 ] +INFO:root:Average loss=0.13430297699570656 +INFO:root: +[Valid][4][UTT][ MSE = 0.2261 | LCC = 0.9484 | SRCC = 0.9486 ] +INFO:root:Average loss=0.11894136869907379 +INFO:root: +[Valid][5][UTT][ MSE = 0.1863 | LCC = 0.9498 | SRCC = 0.9508 ] +INFO:root:Average loss=0.11233006352186203 +INFO:root: +[Valid][6][UTT][ MSE = 0.2369 | LCC = 0.9524 | SRCC = 0.9530 ] +INFO:root:Average loss=0.10018194127082825 +INFO:root: +[Valid][7][UTT][ MSE = 0.2678 | LCC = 0.9543 | SRCC = 0.9546 ] +INFO:root:Average loss=0.09744840401411056 +INFO:root: +[Valid][8][UTT][ MSE = 0.2060 | LCC = 0.9558 | SRCC = 0.9562 ] +INFO:root:Average loss=0.08991459280252456 +INFO:root: +[Valid][9][UTT][ MSE = 0.1969 | LCC = 0.9558 | SRCC = 0.9556 ] +INFO:root:Average loss=0.08285338708758354 +INFO:root: +[Valid][10][UTT][ MSE = 0.2144 | LCC = 0.9562 | SRCC = 0.9562 ] +INFO:root:Average loss=0.07615451729297638 +INFO:root: +[Valid][11][UTT][ MSE = 0.2025 | LCC = 0.9554 | SRCC = 0.9565 ] +INFO:root:Average loss=0.07463958790898323 +INFO:root: +[Valid][12][UTT][ MSE = 0.2000 | LCC = 0.9564 | SRCC = 0.9566 ] +INFO:root:Average loss=0.07050520080327988 +INFO:root: +[Valid][13][UTT][ MSE = 0.1582 | LCC = 0.9569 | SRCC = 0.9585 ] +INFO:root:Average loss=0.06961790752410889 +INFO:root: +[Valid][14][UTT][ MSE = 0.1820 | LCC = 0.9580 | SRCC = 0.9582 ] +INFO:root:Average loss=0.06374165971577167 +INFO:root: +[Valid][15][UTT][ MSE = 0.1762 | LCC = 0.9594 | SRCC = 0.9597 ] +INFO:root:Average loss=0.06397368201613426 +INFO:root: +[Valid][16][UTT][ MSE = 0.1920 | LCC = 0.9608 | SRCC = 0.9610 ] +INFO:root:Average loss=0.05736108380556106 +INFO:root: +[Valid][17][UTT][ MSE = 0.2217 | LCC = 0.9593 | SRCC = 0.9591 ] +INFO:root:Average loss=0.05953851434588432 +INFO:root: +[Valid][18][UTT][ MSE = 0.1801 | LCC = 0.9588 | SRCC = 0.9593 ] +INFO:root:Average loss=0.05171754205226898 +INFO:root: +[Valid][19][UTT][ MSE = 0.1947 | LCC = 0.9582 | SRCC = 0.9591 ] +INFO:root:Average loss=0.050094123095273974 +INFO:root: +[Valid][20][UTT][ MSE = 0.1644 | LCC = 0.9600 | SRCC = 0.9596 ] +INFO:root:Average loss=0.04698943684995174 +INFO:root: +[Valid][21][UTT][ MSE = 0.1573 | LCC = 0.9602 | SRCC = 0.9606 ] +INFO:root:Average loss=0.04766672253608704 +INFO:root: +[Valid][22][UTT][ MSE = 0.1772 | LCC = 0.9603 | SRCC = 0.9601 ] +INFO:root:Average loss=0.045652084663510326 +INFO:root: +[Valid][23][UTT][ MSE = 0.1676 | LCC = 0.9614 | SRCC = 0.9616 ] +INFO:root:Average loss=0.04586703182756901 +INFO:root: +[Valid][24][UTT][ MSE = 0.1678 | LCC = 0.9620 | SRCC = 0.9619 ] +INFO:root:Average loss=0.04136896887421608 +INFO:root: +[Valid][25][UTT][ MSE = 0.1776 | LCC = 0.9606 | SRCC = 0.9610 ] +INFO:root:Average loss=0.04093778370320797 +INFO:root: +[Valid][26][UTT][ MSE = 0.1773 | LCC = 0.9615 | SRCC = 0.9615 ] +INFO:root:Average loss=0.03912183478474617 +INFO:root: +[Valid][27][UTT][ MSE = 0.1740 | LCC = 0.9609 | SRCC = 0.9608 ] +INFO:root:Average loss=0.038100638419389724 +INFO:root: +[Valid][28][UTT][ MSE = 0.1739 | LCC = 0.9613 | SRCC = 0.9615 ] +INFO:root:Average loss=0.03632710340619087 +INFO:root: +[Valid][29][UTT][ MSE = 0.1896 | LCC = 0.9598 | SRCC = 0.9602 ] +INFO:root:Average loss=0.03590711899101734 +INFO:root: +[Valid][30][UTT][ MSE = 0.1798 | LCC = 0.9617 | SRCC = 0.9621 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1714 | LCC = 0.9617 | SRCC = 0.9615 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run1/config.gin b/results_tencent/w2v2_xlsr_1b/layer2/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run1/model.pt b/results_tencent/w2v2_xlsr_1b/layer2/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a4da969fc00e1f67e051afa0f0b7626101b28d6e --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:71573b9a5b1ec5bfa8396a51777a64d047681eb9ad2976cfe7eabd92a7846d0f +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run1/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer2/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..21505bbbd4ccc645210c67db253950f9d19c47b6 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:86ffaa80f1be4e020eeaee5065d174c0eedf337a36b4db3bd15fd8f65ebdd93b +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run1/train.log b/results_tencent/w2v2_xlsr_1b/layer2/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..45302e796545a78358b4f7e0dd8e74caec868089 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46467202043533323 +INFO:root: +[Valid][0][UTT][ MSE = 0.3291 | LCC = 0.9203 | SRCC = 0.9202 ] +INFO:root:Average loss=0.20145255649089813 +INFO:root: +[Valid][1][UTT][ MSE = 0.3023 | LCC = 0.9266 | SRCC = 0.9256 ] +INFO:root:Average loss=0.1704952263236046 +INFO:root: +[Valid][2][UTT][ MSE = 0.2503 | LCC = 0.9448 | SRCC = 0.9440 ] +INFO:root:Average loss=0.14258367162942887 +INFO:root: +[Valid][3][UTT][ MSE = 0.2524 | LCC = 0.9419 | SRCC = 0.9414 ] +INFO:root:Average loss=0.12905046075582505 +INFO:root: +[Valid][4][UTT][ MSE = 0.2269 | LCC = 0.9481 | SRCC = 0.9476 ] +INFO:root:Average loss=0.12368779355287551 +INFO:root: +[Valid][5][UTT][ MSE = 0.2138 | LCC = 0.9500 | SRCC = 0.9501 ] +INFO:root:Average loss=0.10810386854410171 +INFO:root: +[Valid][6][UTT][ MSE = 0.2353 | LCC = 0.9515 | SRCC = 0.9510 ] +INFO:root:Average loss=0.10116704866290092 +INFO:root: +[Valid][7][UTT][ MSE = 0.2235 | LCC = 0.9530 | SRCC = 0.9529 ] +INFO:root:Average loss=0.0971791820526123 +INFO:root: +[Valid][8][UTT][ MSE = 0.2170 | LCC = 0.9538 | SRCC = 0.9533 ] +INFO:root:Average loss=0.08843359971046448 +INFO:root: +[Valid][9][UTT][ MSE = 0.1984 | LCC = 0.9569 | SRCC = 0.9563 ] +INFO:root:Average loss=0.08270053434371948 +INFO:root: +[Valid][10][UTT][ MSE = 0.2017 | LCC = 0.9575 | SRCC = 0.9568 ] +INFO:root:Average loss=0.0768746231496334 +INFO:root: +[Valid][11][UTT][ MSE = 0.2332 | LCC = 0.9556 | SRCC = 0.9551 ] +INFO:root:Average loss=0.07488982114195823 +INFO:root: +[Valid][12][UTT][ MSE = 0.2128 | LCC = 0.9568 | SRCC = 0.9560 ] +INFO:root:Average loss=0.0699715566933155 +INFO:root: +[Valid][13][UTT][ MSE = 0.1887 | LCC = 0.9576 | SRCC = 0.9570 ] +INFO:root:Average loss=0.06551158154010772 +INFO:root: +[Valid][14][UTT][ MSE = 0.2070 | LCC = 0.9586 | SRCC = 0.9576 ] +INFO:root:Average loss=0.06390712484717369 +INFO:root: +[Valid][15][UTT][ MSE = 0.2083 | LCC = 0.9589 | SRCC = 0.9588 ] +INFO:root:Average loss=0.05955391001701355 +INFO:root: +[Valid][16][UTT][ MSE = 0.1900 | LCC = 0.9573 | SRCC = 0.9566 ] +INFO:root:Average loss=0.058385113924741744 +INFO:root: +[Valid][17][UTT][ MSE = 0.1850 | LCC = 0.9592 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05406233386695385 +INFO:root: +[Valid][18][UTT][ MSE = 0.1757 | LCC = 0.9591 | SRCC = 0.9586 ] +INFO:root:Average loss=0.05281825104355812 +INFO:root: +[Valid][19][UTT][ MSE = 0.1964 | LCC = 0.9597 | SRCC = 0.9593 ] +INFO:root:Average loss=0.049773103833198544 +INFO:root: +[Valid][20][UTT][ MSE = 0.2085 | LCC = 0.9601 | SRCC = 0.9603 ] +INFO:root:Average loss=0.04669042561948299 +INFO:root: +[Valid][21][UTT][ MSE = 0.1881 | LCC = 0.9587 | SRCC = 0.9575 ] +INFO:root:Average loss=0.04413036198914051 +INFO:root: +[Valid][22][UTT][ MSE = 0.1787 | LCC = 0.9594 | SRCC = 0.9581 ] +INFO:root:Average loss=0.04477673824131489 +INFO:root: +[Valid][23][UTT][ MSE = 0.1877 | LCC = 0.9595 | SRCC = 0.9587 ] +INFO:root:Average loss=0.04350121873617172 +INFO:root: +[Valid][24][UTT][ MSE = 0.1978 | LCC = 0.9601 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04319514824450016 +INFO:root: +[Valid][25][UTT][ MSE = 0.1801 | LCC = 0.9593 | SRCC = 0.9581 ] +INFO:root:Average loss=0.03794980843365192 +INFO:root: +[Valid][26][UTT][ MSE = 0.1930 | LCC = 0.9593 | SRCC = 0.9587 ] +INFO:root:Average loss=0.03668283255398273 +INFO:root: +[Valid][27][UTT][ MSE = 0.2175 | LCC = 0.9612 | SRCC = 0.9609 ] +INFO:root:Average loss=0.036286788150668145 +INFO:root: +[Valid][28][UTT][ MSE = 0.1880 | LCC = 0.9613 | SRCC = 0.9602 ] +INFO:root:Average loss=0.036002527356147764 +INFO:root: +[Valid][29][UTT][ MSE = 0.1866 | LCC = 0.9597 | SRCC = 0.9593 ] +INFO:root:Average loss=0.03426211403310299 +INFO:root: +[Valid][30][UTT][ MSE = 0.2060 | LCC = 0.9599 | SRCC = 0.9599 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1916 | LCC = 0.9627 | SRCC = 0.9626 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run2/config.gin b/results_tencent/w2v2_xlsr_1b/layer2/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run2/model.pt b/results_tencent/w2v2_xlsr_1b/layer2/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..db48485ce2a15dea2bc1bb9e25235cab2f8ec146 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:463678e21eb13945751ea9cc81e45c13f3a4cdbc7dfed2f12d289db7d99b5205 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run2/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer2/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a6952a1846deeec0853ebee269c5de9123a020f6 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:075bd284dc197560e0a0ac05a3f47679ae597f9789d8ea0ae92082cc37abccb2 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run2/train.log b/results_tencent/w2v2_xlsr_1b/layer2/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..7229c2c69497e0614f5f5f5936ed373f1ed93bae --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4361884158849716 +INFO:root: +[Valid][0][UTT][ MSE = 0.3118 | LCC = 0.9217 | SRCC = 0.9202 ] +INFO:root:Average loss=0.1956586617231369 +INFO:root: +[Valid][1][UTT][ MSE = 0.2780 | LCC = 0.9331 | SRCC = 0.9330 ] +INFO:root:Average loss=0.16693655437231064 +INFO:root: +[Valid][2][UTT][ MSE = 0.2350 | LCC = 0.9417 | SRCC = 0.9416 ] +INFO:root:Average loss=0.15423017942905426 +INFO:root: +[Valid][3][UTT][ MSE = 0.2359 | LCC = 0.9459 | SRCC = 0.9449 ] +INFO:root:Average loss=0.1342138198018074 +INFO:root: +[Valid][4][UTT][ MSE = 0.2297 | LCC = 0.9481 | SRCC = 0.9475 ] +INFO:root:Average loss=0.1227210379242897 +INFO:root: +[Valid][5][UTT][ MSE = 0.2096 | LCC = 0.9498 | SRCC = 0.9497 ] +INFO:root:Average loss=0.11509795036911964 +INFO:root: +[Valid][6][UTT][ MSE = 0.2318 | LCC = 0.9532 | SRCC = 0.9529 ] +INFO:root:Average loss=0.10660554832220077 +INFO:root: +[Valid][7][UTT][ MSE = 0.2612 | LCC = 0.9542 | SRCC = 0.9542 ] +INFO:root:Average loss=0.1028058714568615 +INFO:root: +[Valid][8][UTT][ MSE = 0.1853 | LCC = 0.9547 | SRCC = 0.9550 ] +INFO:root:Average loss=0.09098786070942878 +INFO:root: +[Valid][9][UTT][ MSE = 0.2093 | LCC = 0.9542 | SRCC = 0.9543 ] +INFO:root:Average loss=0.08671694675087929 +INFO:root: +[Valid][10][UTT][ MSE = 0.2152 | LCC = 0.9553 | SRCC = 0.9561 ] +INFO:root:Average loss=0.08428066664934158 +INFO:root: +[Valid][11][UTT][ MSE = 0.2206 | LCC = 0.9565 | SRCC = 0.9568 ] +INFO:root:Average loss=0.0763154656291008 +INFO:root: +[Valid][12][UTT][ MSE = 0.2019 | LCC = 0.9582 | SRCC = 0.9579 ] +INFO:root:Average loss=0.07423164683580398 +INFO:root: +[Valid][13][UTT][ MSE = 0.1956 | LCC = 0.9578 | SRCC = 0.9578 ] +INFO:root:Average loss=0.06879951748251915 +INFO:root: +[Valid][14][UTT][ MSE = 0.1965 | LCC = 0.9585 | SRCC = 0.9581 ] +INFO:root:Average loss=0.064233831346035 +INFO:root: +[Valid][15][UTT][ MSE = 0.2144 | LCC = 0.9574 | SRCC = 0.9570 ] +INFO:root:Average loss=0.06352647668123246 +INFO:root: +[Valid][16][UTT][ MSE = 0.1654 | LCC = 0.9573 | SRCC = 0.9571 ] +INFO:root:Average loss=0.06174310737848282 +INFO:root: +[Valid][17][UTT][ MSE = 0.2116 | LCC = 0.9587 | SRCC = 0.9588 ] +INFO:root:Average loss=0.055581676572561266 +INFO:root: +[Valid][18][UTT][ MSE = 0.1910 | LCC = 0.9583 | SRCC = 0.9585 ] +INFO:root:Average loss=0.057142119601368904 +INFO:root: +[Valid][19][UTT][ MSE = 0.1781 | LCC = 0.9594 | SRCC = 0.9590 ] +INFO:root:Average loss=0.052096551075577736 +INFO:root: +[Valid][20][UTT][ MSE = 0.1754 | LCC = 0.9605 | SRCC = 0.9601 ] +INFO:root:Average loss=0.05040637297928333 +INFO:root: +[Valid][21][UTT][ MSE = 0.1916 | LCC = 0.9591 | SRCC = 0.9587 ] +INFO:root:Average loss=0.04811551010608673 +INFO:root: +[Valid][22][UTT][ MSE = 0.1911 | LCC = 0.9603 | SRCC = 0.9599 ] +INFO:root:Average loss=0.045403084248304366 +INFO:root: +[Valid][23][UTT][ MSE = 0.1908 | LCC = 0.9587 | SRCC = 0.9583 ] +INFO:root:Average loss=0.04550700742006302 +INFO:root: +[Valid][24][UTT][ MSE = 0.1904 | LCC = 0.9594 | SRCC = 0.9586 ] +INFO:root:Average loss=0.04357918478548527 +INFO:root: +[Valid][25][UTT][ MSE = 0.2143 | LCC = 0.9588 | SRCC = 0.9585 ] +INFO:root:Average loss=0.04054701517522335 +INFO:root: +[Valid][26][UTT][ MSE = 0.1939 | LCC = 0.9610 | SRCC = 0.9609 ] +INFO:root:Average loss=0.03908897092938423 +INFO:root: +[Valid][27][UTT][ MSE = 0.1817 | LCC = 0.9609 | SRCC = 0.9603 ] +INFO:root:Average loss=0.03698319087922573 +INFO:root: +[Valid][28][UTT][ MSE = 0.1904 | LCC = 0.9593 | SRCC = 0.9591 ] +INFO:root:Average loss=0.035035135567188264 +INFO:root: +[Valid][29][UTT][ MSE = 0.2051 | LCC = 0.9597 | SRCC = 0.9592 ] +INFO:root:Average loss=0.03661058050394058 +INFO:root: +[Valid][30][UTT][ MSE = 0.1936 | LCC = 0.9588 | SRCC = 0.9586 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1995 | LCC = 0.9610 | SRCC = 0.9610 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run3/config.gin b/results_tencent/w2v2_xlsr_1b/layer2/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run3/model.pt b/results_tencent/w2v2_xlsr_1b/layer2/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..6ead8e48161c0579606ee71403ff77ab8a929587 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8e2a31231183e95d48eb65e4630f0ecf42fb2bfc1ee6903193204f4c3cbb5e91 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run3/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer2/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..66c678a9ec3a9e82513780332d325c1c34d23d6b --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f2522c68e99a1ba715c6776b5000f226e31a4e975a187da6209e9fd5613da675 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run3/train.log b/results_tencent/w2v2_xlsr_1b/layer2/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b2adc55a1f2648d9dbf67c2b20bbb1f91e54f2a6 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4743221793174744 +INFO:root: +[Valid][0][UTT][ MSE = 0.3614 | LCC = 0.9199 | SRCC = 0.9235 ] +INFO:root:Average loss=0.2084974582195282 +INFO:root: +[Valid][1][UTT][ MSE = 0.3101 | LCC = 0.9314 | SRCC = 0.9311 ] +INFO:root:Average loss=0.1688889299631119 +INFO:root: +[Valid][2][UTT][ MSE = 0.3061 | LCC = 0.9362 | SRCC = 0.9369 ] +INFO:root:Average loss=0.149858909368515 +INFO:root: +[Valid][3][UTT][ MSE = 0.2665 | LCC = 0.9467 | SRCC = 0.9466 ] +INFO:root:Average loss=0.13150044775009156 +INFO:root: +[Valid][4][UTT][ MSE = 0.2429 | LCC = 0.9497 | SRCC = 0.9502 ] +INFO:root:Average loss=0.12153983455896378 +INFO:root: +[Valid][5][UTT][ MSE = 0.2311 | LCC = 0.9512 | SRCC = 0.9526 ] +INFO:root:Average loss=0.11104896718263627 +INFO:root: +[Valid][6][UTT][ MSE = 0.2199 | LCC = 0.9545 | SRCC = 0.9548 ] +INFO:root:Average loss=0.1048041446506977 +INFO:root: +[Valid][7][UTT][ MSE = 0.2311 | LCC = 0.9553 | SRCC = 0.9551 ] +INFO:root:Average loss=0.09682611674070359 +INFO:root: +[Valid][8][UTT][ MSE = 0.2328 | LCC = 0.9566 | SRCC = 0.9566 ] +INFO:root:Average loss=0.08990117165446282 +INFO:root: +[Valid][9][UTT][ MSE = 0.2061 | LCC = 0.9552 | SRCC = 0.9557 ] +INFO:root:Average loss=0.08173113980889321 +INFO:root: +[Valid][10][UTT][ MSE = 0.2035 | LCC = 0.9558 | SRCC = 0.9552 ] +INFO:root:Average loss=0.07980290514230728 +INFO:root: +[Valid][11][UTT][ MSE = 0.2084 | LCC = 0.9577 | SRCC = 0.9581 ] +INFO:root:Average loss=0.07702980408072471 +INFO:root: +[Valid][12][UTT][ MSE = 0.1982 | LCC = 0.9562 | SRCC = 0.9567 ] +INFO:root:Average loss=0.07428418910503387 +INFO:root: +[Valid][13][UTT][ MSE = 0.2251 | LCC = 0.9575 | SRCC = 0.9581 ] +INFO:root:Average loss=0.06761265182495117 +INFO:root: +[Valid][14][UTT][ MSE = 0.1916 | LCC = 0.9573 | SRCC = 0.9584 ] +INFO:root:Average loss=0.06295046374201775 +INFO:root: +[Valid][15][UTT][ MSE = 0.1831 | LCC = 0.9540 | SRCC = 0.9549 ] +INFO:root:Average loss=0.061833007574081424 +INFO:root: +[Valid][16][UTT][ MSE = 0.1955 | LCC = 0.9576 | SRCC = 0.9579 ] +INFO:root:Average loss=0.057707219272851946 +INFO:root: +[Valid][17][UTT][ MSE = 0.2053 | LCC = 0.9602 | SRCC = 0.9602 ] +INFO:root:Average loss=0.05592323446273804 +INFO:root: +[Valid][18][UTT][ MSE = 0.1980 | LCC = 0.9585 | SRCC = 0.9576 ] +INFO:root:Average loss=0.05233854971826077 +INFO:root: +[Valid][19][UTT][ MSE = 0.1928 | LCC = 0.9607 | SRCC = 0.9597 ] +INFO:root:Average loss=0.04889902852475643 +INFO:root: +[Valid][20][UTT][ MSE = 0.2154 | LCC = 0.9596 | SRCC = 0.9603 ] +INFO:root:Average loss=0.05284094434976578 +INFO:root: +[Valid][21][UTT][ MSE = 0.1898 | LCC = 0.9599 | SRCC = 0.9595 ] +INFO:root:Average loss=0.04698914009332657 +INFO:root: +[Valid][22][UTT][ MSE = 0.1838 | LCC = 0.9585 | SRCC = 0.9584 ] +INFO:root:Average loss=0.04428050637245178 +INFO:root: +[Valid][23][UTT][ MSE = 0.2186 | LCC = 0.9598 | SRCC = 0.9593 ] +INFO:root:Average loss=0.04155408722162247 +INFO:root: +[Valid][24][UTT][ MSE = 0.1747 | LCC = 0.9611 | SRCC = 0.9610 ] +INFO:root:Average loss=0.041107996389269826 +INFO:root: +[Valid][25][UTT][ MSE = 0.1982 | LCC = 0.9589 | SRCC = 0.9587 ] +INFO:root:Average loss=0.04015895800292492 +INFO:root: +[Valid][26][UTT][ MSE = 0.1883 | LCC = 0.9577 | SRCC = 0.9577 ] +INFO:root:Average loss=0.037244582951068875 +INFO:root: +[Valid][27][UTT][ MSE = 0.1775 | LCC = 0.9614 | SRCC = 0.9607 ] +INFO:root:Average loss=0.03525900030136108 +INFO:root: +[Valid][28][UTT][ MSE = 0.2005 | LCC = 0.9609 | SRCC = 0.9601 ] +INFO:root:Average loss=0.03601951266825199 +INFO:root: +[Valid][29][UTT][ MSE = 0.2064 | LCC = 0.9600 | SRCC = 0.9602 ] +INFO:root:Average loss=0.033959459409117695 +INFO:root: +[Valid][30][UTT][ MSE = 0.1950 | LCC = 0.9605 | SRCC = 0.9597 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1775 | LCC = 0.9638 | SRCC = 0.9637 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run4/config.gin b/results_tencent/w2v2_xlsr_1b/layer2/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run4/model.pt b/results_tencent/w2v2_xlsr_1b/layer2/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2784fbde72a8a8e66c37327ebd4a742e83a89b45 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:224b8b3f283f100c5922be52a72fe6bcfbc1b6c0b7a3438da0739b2c0e1aa29e +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run4/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer2/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4a8c53bcfd519b6375a28afdb80876ad7e5570e3 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bf433ccd19f6e3748fadc127389688104170eeebf3c6c7a1bb75f503df097910 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer2/run4/train.log b/results_tencent/w2v2_xlsr_1b/layer2/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..770f34083757a13e04f4a5f347845d918c5a9e64 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer2/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3935891973972321 +INFO:root: +[Valid][0][UTT][ MSE = 0.3017 | LCC = 0.9234 | SRCC = 0.9253 ] +INFO:root:Average loss=0.19774833750724793 +INFO:root: +[Valid][1][UTT][ MSE = 0.2678 | LCC = 0.9341 | SRCC = 0.9336 ] +INFO:root:Average loss=0.1652243873476982 +INFO:root: +[Valid][2][UTT][ MSE = 0.2414 | LCC = 0.9411 | SRCC = 0.9392 ] +INFO:root:Average loss=0.14866545230150222 +INFO:root: +[Valid][3][UTT][ MSE = 0.2374 | LCC = 0.9450 | SRCC = 0.9449 ] +INFO:root:Average loss=0.13730909025669097 +INFO:root: +[Valid][4][UTT][ MSE = 0.2168 | LCC = 0.9482 | SRCC = 0.9481 ] +INFO:root:Average loss=0.1251459801197052 +INFO:root: +[Valid][5][UTT][ MSE = 0.2420 | LCC = 0.9500 | SRCC = 0.9492 ] +INFO:root:Average loss=0.11075344502925873 +INFO:root: +[Valid][6][UTT][ MSE = 0.2427 | LCC = 0.9505 | SRCC = 0.9503 ] +INFO:root:Average loss=0.10644187876582145 +INFO:root: +[Valid][7][UTT][ MSE = 0.2430 | LCC = 0.9545 | SRCC = 0.9549 ] +INFO:root:Average loss=0.10099264889955521 +INFO:root: +[Valid][8][UTT][ MSE = 0.2186 | LCC = 0.9553 | SRCC = 0.9550 ] +INFO:root:Average loss=0.0915990798175335 +INFO:root: +[Valid][9][UTT][ MSE = 0.2454 | LCC = 0.9561 | SRCC = 0.9564 ] +INFO:root:Average loss=0.0899885538816452 +INFO:root: +[Valid][10][UTT][ MSE = 0.2055 | LCC = 0.9560 | SRCC = 0.9556 ] +INFO:root:Average loss=0.0799311265051365 +INFO:root: +[Valid][11][UTT][ MSE = 0.2263 | LCC = 0.9570 | SRCC = 0.9563 ] +INFO:root:Average loss=0.07846022135019302 +INFO:root: +[Valid][12][UTT][ MSE = 0.2287 | LCC = 0.9569 | SRCC = 0.9570 ] +INFO:root:Average loss=0.07434172317385673 +INFO:root: +[Valid][13][UTT][ MSE = 0.2030 | LCC = 0.9563 | SRCC = 0.9561 ] +INFO:root:Average loss=0.06937556836009026 +INFO:root: +[Valid][14][UTT][ MSE = 0.1820 | LCC = 0.9590 | SRCC = 0.9583 ] +INFO:root:Average loss=0.0723463013768196 +INFO:root: +[Valid][15][UTT][ MSE = 0.2286 | LCC = 0.9579 | SRCC = 0.9578 ] +INFO:root:Average loss=0.06427243787050248 +INFO:root: +[Valid][16][UTT][ MSE = 0.1841 | LCC = 0.9585 | SRCC = 0.9575 ] +INFO:root:Average loss=0.06021723100543022 +INFO:root: +[Valid][17][UTT][ MSE = 0.1924 | LCC = 0.9600 | SRCC = 0.9595 ] +INFO:root:Average loss=0.05641474783420563 +INFO:root: +[Valid][18][UTT][ MSE = 0.1972 | LCC = 0.9589 | SRCC = 0.9588 ] +INFO:root:Average loss=0.05563181327283383 +INFO:root: +[Valid][19][UTT][ MSE = 0.1753 | LCC = 0.9581 | SRCC = 0.9572 ] +INFO:root:Average loss=0.0558266923725605 +INFO:root: +[Valid][20][UTT][ MSE = 0.2185 | LCC = 0.9605 | SRCC = 0.9601 ] +INFO:root:Average loss=0.04940615989267826 +INFO:root: +[Valid][21][UTT][ MSE = 0.1982 | LCC = 0.9599 | SRCC = 0.9596 ] +INFO:root:Average loss=0.049416433468461037 +INFO:root: +[Valid][22][UTT][ MSE = 0.1912 | LCC = 0.9596 | SRCC = 0.9597 ] +INFO:root:Average loss=0.04722319348156452 +INFO:root: +[Valid][23][UTT][ MSE = 0.2198 | LCC = 0.9595 | SRCC = 0.9590 ] +INFO:root:Average loss=0.04528326161205769 +INFO:root: +[Valid][24][UTT][ MSE = 0.2310 | LCC = 0.9606 | SRCC = 0.9605 ] +INFO:root:Average loss=0.04519787840545177 +INFO:root: +[Valid][25][UTT][ MSE = 0.2257 | LCC = 0.9584 | SRCC = 0.9579 ] +INFO:root:Average loss=0.04508697259426117 +INFO:root: +[Valid][26][UTT][ MSE = 0.2021 | LCC = 0.9602 | SRCC = 0.9597 ] +INFO:root:Average loss=0.04243042053282261 +INFO:root: +[Valid][27][UTT][ MSE = 0.2054 | LCC = 0.9597 | SRCC = 0.9593 ] +INFO:root:Average loss=0.04191453276574612 +INFO:root: +[Valid][28][UTT][ MSE = 0.2238 | LCC = 0.9595 | SRCC = 0.9592 ] +INFO:root:Average loss=0.037985398307442665 +INFO:root: +[Valid][29][UTT][ MSE = 0.2135 | LCC = 0.9599 | SRCC = 0.9585 ] +INFO:root:Average loss=0.0364434319883585 +INFO:root: +[Valid][30][UTT][ MSE = 0.1983 | LCC = 0.9608 | SRCC = 0.9603 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1979 | LCC = 0.9613 | SRCC = 0.9611 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run0/config.gin b/results_tencent/w2v2_xlsr_1b/layer20/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run0/model.pt b/results_tencent/w2v2_xlsr_1b/layer20/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7d4086a70d00e58b677fff4a810e7d7b3fbd63bd --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:33e655fa4a9e214751991047e26f92baed31480a20d0584961ef91d3ac11b5da +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run0/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer20/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..587a4e4550979e06d68fcb4e644ea8366a94c3c0 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:897bcd8f69c891e26028472e5eb801962f4d55911b5256ffb81dbab6f9078770 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run0/train.log b/results_tencent/w2v2_xlsr_1b/layer20/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..eee1b241f6205ec73798b2f0b9d1a7a48f8e109f --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4207360532283783 +INFO:root: +[Valid][0][UTT][ MSE = 0.2092 | LCC = 0.9491 | SRCC = 0.9491 ] +INFO:root:Average loss=0.1395231836438179 +INFO:root: +[Valid][1][UTT][ MSE = 0.2457 | LCC = 0.9578 | SRCC = 0.9586 ] +INFO:root:Average loss=0.11850662523508072 +INFO:root: +[Valid][2][UTT][ MSE = 0.2289 | LCC = 0.9644 | SRCC = 0.9636 ] +INFO:root:Average loss=0.11107894483208657 +INFO:root: +[Valid][3][UTT][ MSE = 0.2079 | LCC = 0.9649 | SRCC = 0.9642 ] +INFO:root:Average loss=0.09660373172163964 +INFO:root: +[Valid][4][UTT][ MSE = 0.1782 | LCC = 0.9669 | SRCC = 0.9669 ] +INFO:root:Average loss=0.09024651038646699 +INFO:root: +[Valid][5][UTT][ MSE = 0.2051 | LCC = 0.9690 | SRCC = 0.9682 ] +INFO:root:Average loss=0.08359943541884422 +INFO:root: +[Valid][6][UTT][ MSE = 0.1952 | LCC = 0.9693 | SRCC = 0.9689 ] +INFO:root:Average loss=0.07741576522588729 +INFO:root: +[Valid][7][UTT][ MSE = 0.1800 | LCC = 0.9703 | SRCC = 0.9695 ] +INFO:root:Average loss=0.06998672118782998 +INFO:root: +[Valid][8][UTT][ MSE = 0.1739 | LCC = 0.9704 | SRCC = 0.9697 ] +INFO:root:Average loss=0.07050219109654426 +INFO:root: +[Valid][9][UTT][ MSE = 0.1533 | LCC = 0.9705 | SRCC = 0.9697 ] +INFO:root:Average loss=0.06605657714605331 +INFO:root: +[Valid][10][UTT][ MSE = 0.1674 | LCC = 0.9708 | SRCC = 0.9700 ] +INFO:root:Average loss=0.05942569702863693 +INFO:root: +[Valid][11][UTT][ MSE = 0.1518 | LCC = 0.9715 | SRCC = 0.9706 ] +INFO:root:Average loss=0.06078849878907204 +INFO:root: +[Valid][12][UTT][ MSE = 0.1706 | LCC = 0.9715 | SRCC = 0.9710 ] +INFO:root:Average loss=0.05826843753457069 +INFO:root: +[Valid][13][UTT][ MSE = 0.1910 | LCC = 0.9716 | SRCC = 0.9711 ] +INFO:root:Average loss=0.05352363894879818 +INFO:root: +[Valid][14][UTT][ MSE = 0.1884 | LCC = 0.9712 | SRCC = 0.9704 ] +INFO:root:Average loss=0.05081995637714863 +INFO:root: +[Valid][15][UTT][ MSE = 0.1567 | LCC = 0.9715 | SRCC = 0.9702 ] +INFO:root:Average loss=0.04830490072071552 +INFO:root: +[Valid][16][UTT][ MSE = 0.1879 | LCC = 0.9721 | SRCC = 0.9709 ] +INFO:root:Average loss=0.047237848982214926 +INFO:root: +[Valid][17][UTT][ MSE = 0.2096 | LCC = 0.9727 | SRCC = 0.9716 ] +INFO:root:Average loss=0.04660056491196156 +INFO:root: +[Valid][18][UTT][ MSE = 0.1769 | LCC = 0.9697 | SRCC = 0.9693 ] +INFO:root:Average loss=0.044620296463370325 +INFO:root: +[Valid][19][UTT][ MSE = 0.1828 | LCC = 0.9720 | SRCC = 0.9710 ] +INFO:root:Average loss=0.0426322613209486 +INFO:root: +[Valid][20][UTT][ MSE = 0.1682 | LCC = 0.9730 | SRCC = 0.9713 ] +INFO:root:Average loss=0.04087795500457287 +INFO:root: +[Valid][21][UTT][ MSE = 0.1830 | LCC = 0.9723 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03954802891612053 +INFO:root: +[Valid][22][UTT][ MSE = 0.1863 | LCC = 0.9724 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03564738018810749 +INFO:root: +[Valid][23][UTT][ MSE = 0.1739 | LCC = 0.9720 | SRCC = 0.9705 ] +INFO:root:Average loss=0.03446473383903503 +INFO:root: +[Valid][24][UTT][ MSE = 0.1645 | LCC = 0.9724 | SRCC = 0.9711 ] +INFO:root:Average loss=0.0339380616992712 +INFO:root: +[Valid][25][UTT][ MSE = 0.1956 | LCC = 0.9719 | SRCC = 0.9709 ] +INFO:root:Average loss=0.0326080259680748 +INFO:root: +[Valid][26][UTT][ MSE = 0.1731 | LCC = 0.9721 | SRCC = 0.9710 ] +INFO:root:Average loss=0.029818164601922034 +INFO:root: +[Valid][27][UTT][ MSE = 0.1607 | LCC = 0.9720 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03204004164040088 +INFO:root: +[Valid][28][UTT][ MSE = 0.1627 | LCC = 0.9724 | SRCC = 0.9711 ] +INFO:root:Average loss=0.0285061496719718 +INFO:root: +[Valid][29][UTT][ MSE = 0.1675 | LCC = 0.9721 | SRCC = 0.9709 ] +INFO:root:Average loss=0.027298770293593407 +INFO:root: +[Valid][30][UTT][ MSE = 0.1605 | LCC = 0.9717 | SRCC = 0.9704 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1638 | LCC = 0.9740 | SRCC = 0.9734 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run1/config.gin b/results_tencent/w2v2_xlsr_1b/layer20/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run1/model.pt b/results_tencent/w2v2_xlsr_1b/layer20/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b09ba3443ce795aab0c397729a822277f288d348 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6aa7d718c50990257b0a4368f793ed10aed4b63707fb3f305a93bd7ea147def3 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run1/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer20/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a1628e979582ac1577ae22e86f4acef113f4eed6 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:36d72d19f3cbe14c05b8b5a5fe8e766deed4a17bc73e956646f59ac2ce36c0c4 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run1/train.log b/results_tencent/w2v2_xlsr_1b/layer20/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6b482e85f448b03d079026db461a9ae40ce63c38 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.38520575600862506 +INFO:root: +[Valid][0][UTT][ MSE = 0.2830 | LCC = 0.9471 | SRCC = 0.9458 ] +INFO:root:Average loss=0.15555082416534424 +INFO:root: +[Valid][1][UTT][ MSE = 0.2440 | LCC = 0.9588 | SRCC = 0.9583 ] +INFO:root:Average loss=0.12140778398513793 +INFO:root: +[Valid][2][UTT][ MSE = 0.2245 | LCC = 0.9625 | SRCC = 0.9629 ] +INFO:root:Average loss=0.10630846628546715 +INFO:root: +[Valid][3][UTT][ MSE = 0.2111 | LCC = 0.9614 | SRCC = 0.9642 ] +INFO:root:Average loss=0.09878227862715722 +INFO:root: +[Valid][4][UTT][ MSE = 0.1717 | LCC = 0.9657 | SRCC = 0.9659 ] +INFO:root:Average loss=0.08314048945903778 +INFO:root: +[Valid][5][UTT][ MSE = 0.2051 | LCC = 0.9689 | SRCC = 0.9680 ] +INFO:root:Average loss=0.08304547196626663 +INFO:root: +[Valid][6][UTT][ MSE = 0.2115 | LCC = 0.9686 | SRCC = 0.9677 ] +INFO:root:Average loss=0.07710136079788207 +INFO:root: +[Valid][7][UTT][ MSE = 0.2067 | LCC = 0.9707 | SRCC = 0.9698 ] +INFO:root:Average loss=0.06955500382184983 +INFO:root: +[Valid][8][UTT][ MSE = 0.1944 | LCC = 0.9705 | SRCC = 0.9698 ] +INFO:root:Average loss=0.07058799043297767 +INFO:root: +[Valid][9][UTT][ MSE = 0.1606 | LCC = 0.9689 | SRCC = 0.9686 ] +INFO:root:Average loss=0.06750756329298019 +INFO:root: +[Valid][10][UTT][ MSE = 0.2118 | LCC = 0.9704 | SRCC = 0.9704 ] +INFO:root:Average loss=0.06017570541799069 +INFO:root: +[Valid][11][UTT][ MSE = 0.1655 | LCC = 0.9702 | SRCC = 0.9703 ] +INFO:root:Average loss=0.0574020572155714 +INFO:root: +[Valid][12][UTT][ MSE = 0.1920 | LCC = 0.9707 | SRCC = 0.9702 ] +INFO:root:Average loss=0.05802043183147907 +INFO:root: +[Valid][13][UTT][ MSE = 0.1869 | LCC = 0.9705 | SRCC = 0.9695 ] +INFO:root:Average loss=0.05137535643577576 +INFO:root: +[Valid][14][UTT][ MSE = 0.1753 | LCC = 0.9714 | SRCC = 0.9702 ] +INFO:root:Average loss=0.049361525654792786 +INFO:root: +[Valid][15][UTT][ MSE = 0.1778 | LCC = 0.9702 | SRCC = 0.9697 ] +INFO:root:Average loss=0.0463139036744833 +INFO:root: +[Valid][16][UTT][ MSE = 0.1689 | LCC = 0.9693 | SRCC = 0.9693 ] +INFO:root:Average loss=0.04631818605959415 +INFO:root: +[Valid][17][UTT][ MSE = 0.1701 | LCC = 0.9713 | SRCC = 0.9702 ] +INFO:root:Average loss=0.04440351288020611 +INFO:root: +[Valid][18][UTT][ MSE = 0.1825 | LCC = 0.9715 | SRCC = 0.9707 ] +INFO:root:Average loss=0.04149155879020691 +INFO:root: +[Valid][19][UTT][ MSE = 0.1758 | LCC = 0.9713 | SRCC = 0.9704 ] +INFO:root:Average loss=0.04012129013240338 +INFO:root: +[Valid][20][UTT][ MSE = 0.1826 | LCC = 0.9721 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03942212146520615 +INFO:root: +[Valid][21][UTT][ MSE = 0.1675 | LCC = 0.9720 | SRCC = 0.9710 ] +INFO:root:Average loss=0.035590427562594416 +INFO:root: +[Valid][22][UTT][ MSE = 0.1943 | LCC = 0.9708 | SRCC = 0.9703 ] +INFO:root:Average loss=0.03698373386263847 +INFO:root: +[Valid][23][UTT][ MSE = 0.1659 | LCC = 0.9713 | SRCC = 0.9707 ] +INFO:root:Average loss=0.03216133077442646 +INFO:root: +[Valid][24][UTT][ MSE = 0.1590 | LCC = 0.9718 | SRCC = 0.9709 ] +INFO:root:Average loss=0.031017222464084625 +INFO:root: +[Valid][25][UTT][ MSE = 0.1689 | LCC = 0.9717 | SRCC = 0.9710 ] +INFO:root:Average loss=0.029839515641331672 +INFO:root: +[Valid][26][UTT][ MSE = 0.1715 | LCC = 0.9720 | SRCC = 0.9708 ] +INFO:root:Average loss=0.030062060102820396 +INFO:root: +[Valid][27][UTT][ MSE = 0.1761 | LCC = 0.9719 | SRCC = 0.9705 ] +INFO:root:Average loss=0.028475461512804032 +INFO:root: +[Valid][28][UTT][ MSE = 0.1761 | LCC = 0.9701 | SRCC = 0.9696 ] +INFO:root:Average loss=0.02671913269162178 +INFO:root: +[Valid][29][UTT][ MSE = 0.1670 | LCC = 0.9715 | SRCC = 0.9704 ] +INFO:root:Average loss=0.026331167474389076 +INFO:root: +[Valid][30][UTT][ MSE = 0.1757 | LCC = 0.9719 | SRCC = 0.9705 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1805 | LCC = 0.9731 | SRCC = 0.9731 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run2/config.gin b/results_tencent/w2v2_xlsr_1b/layer20/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run2/model.pt b/results_tencent/w2v2_xlsr_1b/layer20/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..2b7c3067cff7b78f79ae3b746a197eefbefe2084 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d8f6c91dc595e9c45dcf2a2161899b4b467769cf430e8c15a09da0312051dc95 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run2/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer20/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..80ae6dc9c9cd8f91f94b71bac87606b556d3652b --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5f24550cd78b95c529a787efdf9e67ed30de2e021742f1c7851f9e58738dbd6c +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run2/train.log b/results_tencent/w2v2_xlsr_1b/layer20/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..665ff1d3474833a7429868e7e78b7a5df6286039 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4505314869880676 +INFO:root: +[Valid][0][UTT][ MSE = 0.2304 | LCC = 0.9438 | SRCC = 0.9411 ] +INFO:root:Average loss=0.14529948061704637 +INFO:root: +[Valid][1][UTT][ MSE = 0.2239 | LCC = 0.9559 | SRCC = 0.9570 ] +INFO:root:Average loss=0.12144883722066879 +INFO:root: +[Valid][2][UTT][ MSE = 0.1927 | LCC = 0.9617 | SRCC = 0.9618 ] +INFO:root:Average loss=0.11193197298049927 +INFO:root: +[Valid][3][UTT][ MSE = 0.1723 | LCC = 0.9642 | SRCC = 0.9632 ] +INFO:root:Average loss=0.09597578650712967 +INFO:root: +[Valid][4][UTT][ MSE = 0.1950 | LCC = 0.9675 | SRCC = 0.9669 ] +INFO:root:Average loss=0.08988733053207397 +INFO:root: +[Valid][5][UTT][ MSE = 0.1717 | LCC = 0.9682 | SRCC = 0.9671 ] +INFO:root:Average loss=0.07973812776803971 +INFO:root: +[Valid][6][UTT][ MSE = 0.1571 | LCC = 0.9688 | SRCC = 0.9685 ] +INFO:root:Average loss=0.07390644934773445 +INFO:root: +[Valid][7][UTT][ MSE = 0.1365 | LCC = 0.9692 | SRCC = 0.9688 ] +INFO:root:Average loss=0.07523500692844391 +INFO:root: +[Valid][8][UTT][ MSE = 0.1354 | LCC = 0.9690 | SRCC = 0.9686 ] +INFO:root:Average loss=0.0691886691749096 +INFO:root: +[Valid][9][UTT][ MSE = 0.1645 | LCC = 0.9710 | SRCC = 0.9699 ] +INFO:root:Average loss=0.06531089879572391 +INFO:root: +[Valid][10][UTT][ MSE = 0.1734 | LCC = 0.9712 | SRCC = 0.9703 ] +INFO:root:Average loss=0.060701041013002395 +INFO:root: +[Valid][11][UTT][ MSE = 0.1595 | LCC = 0.9713 | SRCC = 0.9703 ] +INFO:root:Average loss=0.057034149616956714 +INFO:root: +[Valid][12][UTT][ MSE = 0.1544 | LCC = 0.9719 | SRCC = 0.9708 ] +INFO:root:Average loss=0.05368835589289665 +INFO:root: +[Valid][13][UTT][ MSE = 0.1315 | LCC = 0.9711 | SRCC = 0.9707 ] +INFO:root:Average loss=0.05183392743766308 +INFO:root: +[Valid][14][UTT][ MSE = 0.1486 | LCC = 0.9708 | SRCC = 0.9694 ] +INFO:root:Average loss=0.049498703271150586 +INFO:root: +[Valid][15][UTT][ MSE = 0.1341 | LCC = 0.9704 | SRCC = 0.9702 ] +INFO:root:Average loss=0.046887609764933585 +INFO:root: +[Valid][16][UTT][ MSE = 0.1665 | LCC = 0.9719 | SRCC = 0.9707 ] +INFO:root:Average loss=0.04843592543900013 +INFO:root: +[Valid][17][UTT][ MSE = 0.1661 | LCC = 0.9718 | SRCC = 0.9713 ] +INFO:root:Average loss=0.0427311138510704 +INFO:root: +[Valid][18][UTT][ MSE = 0.1517 | LCC = 0.9709 | SRCC = 0.9707 ] +INFO:root:Average loss=0.04303415401279926 +INFO:root: +[Valid][19][UTT][ MSE = 0.1416 | LCC = 0.9729 | SRCC = 0.9713 ] +INFO:root:Average loss=0.0400495917648077 +INFO:root: +[Valid][20][UTT][ MSE = 0.1450 | LCC = 0.9722 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03706507551670075 +INFO:root: +[Valid][21][UTT][ MSE = 0.1377 | LCC = 0.9720 | SRCC = 0.9707 ] +INFO:root:Average loss=0.035748070910573006 +INFO:root: +[Valid][22][UTT][ MSE = 0.1469 | LCC = 0.9712 | SRCC = 0.9696 ] +INFO:root:Average loss=0.03510888759791851 +INFO:root: +[Valid][23][UTT][ MSE = 0.1525 | LCC = 0.9719 | SRCC = 0.9704 ] +INFO:root:Average loss=0.033113690465688704 +INFO:root: +[Valid][24][UTT][ MSE = 0.1559 | LCC = 0.9721 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03219167874753475 +INFO:root: +[Valid][25][UTT][ MSE = 0.1481 | LCC = 0.9719 | SRCC = 0.9704 ] +INFO:root:Average loss=0.03149093781411648 +INFO:root: +[Valid][26][UTT][ MSE = 0.1572 | LCC = 0.9724 | SRCC = 0.9709 ] +INFO:root:Average loss=0.029402897283434867 +INFO:root: +[Valid][27][UTT][ MSE = 0.1587 | LCC = 0.9713 | SRCC = 0.9704 ] +INFO:root:Average loss=0.029810405597090722 +INFO:root: +[Valid][28][UTT][ MSE = 0.1634 | LCC = 0.9721 | SRCC = 0.9705 ] +INFO:root:Average loss=0.02675748450309038 +INFO:root: +[Valid][29][UTT][ MSE = 0.1499 | LCC = 0.9729 | SRCC = 0.9710 ] +INFO:root:Average loss=0.025759999796748163 +INFO:root: +[Valid][30][UTT][ MSE = 0.1667 | LCC = 0.9720 | SRCC = 0.9707 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1424 | LCC = 0.9729 | SRCC = 0.9729 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run3/config.gin b/results_tencent/w2v2_xlsr_1b/layer20/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run3/model.pt b/results_tencent/w2v2_xlsr_1b/layer20/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e76fa88efc60e0315a70ccac67178b6325f6de0e --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df85c901b647bc18614526403a5f69ccdfdb762ba7e9d79e8395075ad11545b6 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run3/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer20/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b31834f5a9197a5e383ed4058f3c93f6e0123863 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fb1bba981153c5f422277192aca04497a123aea8a841733599a1d7f281d17e60 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run3/train.log b/results_tencent/w2v2_xlsr_1b/layer20/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..8eebaa0fd51e49e622bffe460875df8cdd7c6a01 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3749356597661972 +INFO:root: +[Valid][0][UTT][ MSE = 0.2363 | LCC = 0.9484 | SRCC = 0.9485 ] +INFO:root:Average loss=0.13406599259376525 +INFO:root: +[Valid][1][UTT][ MSE = 0.2310 | LCC = 0.9607 | SRCC = 0.9602 ] +INFO:root:Average loss=0.11430383086204529 +INFO:root: +[Valid][2][UTT][ MSE = 0.1891 | LCC = 0.9625 | SRCC = 0.9621 ] +INFO:root:Average loss=0.09949648278951645 +INFO:root: +[Valid][3][UTT][ MSE = 0.1795 | LCC = 0.9664 | SRCC = 0.9661 ] +INFO:root:Average loss=0.09468415749073028 +INFO:root: +[Valid][4][UTT][ MSE = 0.2284 | LCC = 0.9678 | SRCC = 0.9685 ] +INFO:root:Average loss=0.08489186203479766 +INFO:root: +[Valid][5][UTT][ MSE = 0.1724 | LCC = 0.9693 | SRCC = 0.9687 ] +INFO:root:Average loss=0.07765079134702682 +INFO:root: +[Valid][6][UTT][ MSE = 0.2261 | LCC = 0.9690 | SRCC = 0.9693 ] +INFO:root:Average loss=0.07239574009180069 +INFO:root: +[Valid][7][UTT][ MSE = 0.2200 | LCC = 0.9703 | SRCC = 0.9704 ] +INFO:root:Average loss=0.07035252353549004 +INFO:root: +[Valid][8][UTT][ MSE = 0.1875 | LCC = 0.9709 | SRCC = 0.9698 ] +INFO:root:Average loss=0.06268994754552841 +INFO:root: +[Valid][9][UTT][ MSE = 0.1854 | LCC = 0.9720 | SRCC = 0.9713 ] +INFO:root:Average loss=0.06612324362993241 +INFO:root: +[Valid][10][UTT][ MSE = 0.2007 | LCC = 0.9701 | SRCC = 0.9711 ] +INFO:root:Average loss=0.05790847977995873 +INFO:root: +[Valid][11][UTT][ MSE = 0.1739 | LCC = 0.9714 | SRCC = 0.9705 ] +INFO:root:Average loss=0.054820761382579804 +INFO:root: +[Valid][12][UTT][ MSE = 0.1737 | LCC = 0.9707 | SRCC = 0.9707 ] +INFO:root:Average loss=0.05091793760657311 +INFO:root: +[Valid][13][UTT][ MSE = 0.1577 | LCC = 0.9729 | SRCC = 0.9714 ] +INFO:root:Average loss=0.04986516723036766 +INFO:root: +[Valid][14][UTT][ MSE = 0.1710 | LCC = 0.9724 | SRCC = 0.9715 ] +INFO:root:Average loss=0.048270538732409475 +INFO:root: +[Valid][15][UTT][ MSE = 0.1742 | LCC = 0.9726 | SRCC = 0.9718 ] +INFO:root:Average loss=0.04452754652500152 +INFO:root: +[Valid][16][UTT][ MSE = 0.1531 | LCC = 0.9729 | SRCC = 0.9717 ] +INFO:root:Average loss=0.04319024841487408 +INFO:root: +[Valid][17][UTT][ MSE = 0.1652 | LCC = 0.9718 | SRCC = 0.9705 ] +INFO:root:Average loss=0.03964598801732063 +INFO:root: +[Valid][18][UTT][ MSE = 0.1751 | LCC = 0.9723 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03832850779592991 +INFO:root: +[Valid][19][UTT][ MSE = 0.1762 | LCC = 0.9713 | SRCC = 0.9698 ] +INFO:root:Average loss=0.03662756796181202 +INFO:root: +[Valid][20][UTT][ MSE = 0.1794 | LCC = 0.9720 | SRCC = 0.9707 ] +INFO:root:Average loss=0.035830851271748544 +INFO:root: +[Valid][21][UTT][ MSE = 0.1500 | LCC = 0.9715 | SRCC = 0.9704 ] +INFO:root:Average loss=0.03440618290007114 +INFO:root: +[Valid][22][UTT][ MSE = 0.1841 | LCC = 0.9714 | SRCC = 0.9708 ] +INFO:root:Average loss=0.034502650558948515 +INFO:root: +[Valid][23][UTT][ MSE = 0.1726 | LCC = 0.9719 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03230590549111366 +INFO:root: +[Valid][24][UTT][ MSE = 0.1857 | LCC = 0.9711 | SRCC = 0.9699 ] +INFO:root:Average loss=0.030334644496440886 +INFO:root: +[Valid][25][UTT][ MSE = 0.1767 | LCC = 0.9708 | SRCC = 0.9703 ] +INFO:root:Average loss=0.029962536200881004 +INFO:root: +[Valid][26][UTT][ MSE = 0.1857 | LCC = 0.9713 | SRCC = 0.9702 ] +INFO:root:Average loss=0.027931828781962396 +INFO:root: +[Valid][27][UTT][ MSE = 0.1928 | LCC = 0.9710 | SRCC = 0.9698 ] +INFO:root:Average loss=0.029725330844521523 +INFO:root: +[Valid][28][UTT][ MSE = 0.1753 | LCC = 0.9713 | SRCC = 0.9699 ] +INFO:root:Average loss=0.029308917611837387 +INFO:root: +[Valid][29][UTT][ MSE = 0.1560 | LCC = 0.9722 | SRCC = 0.9708 ] +INFO:root:Average loss=0.02541006547957659 +INFO:root: +[Valid][30][UTT][ MSE = 0.1600 | LCC = 0.9719 | SRCC = 0.9706 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1563 | LCC = 0.9739 | SRCC = 0.9735 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run4/config.gin b/results_tencent/w2v2_xlsr_1b/layer20/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run4/model.pt b/results_tencent/w2v2_xlsr_1b/layer20/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..0e3adc4be6f39f8462184b5905526efdce9feef6 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a2d666bfe9879cd0d36972f42f2cd032224a64ea510bbfbefe8527c4abbf7cef +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run4/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer20/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..9e9e486c74aac5a1a7c0f10b54c8de4e4147dee1 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7916226dd0dd86d4c7ed5c0b93239ff89b972844f2a62c3a92f93109d0de3e2c +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer20/run4/train.log b/results_tencent/w2v2_xlsr_1b/layer20/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f02c3188540f87c333a67bb5d1d334054fbd794b --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer20/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.3665449292659759 +INFO:root: +[Valid][0][UTT][ MSE = 0.3387 | LCC = 0.9474 | SRCC = 0.9477 ] +INFO:root:Average loss=0.14064452868700028 +INFO:root: +[Valid][1][UTT][ MSE = 0.2174 | LCC = 0.9590 | SRCC = 0.9598 ] +INFO:root:Average loss=0.11731572264432907 +INFO:root: +[Valid][2][UTT][ MSE = 0.1992 | LCC = 0.9645 | SRCC = 0.9654 ] +INFO:root:Average loss=0.10261135375499725 +INFO:root: +[Valid][3][UTT][ MSE = 0.2027 | LCC = 0.9682 | SRCC = 0.9677 ] +INFO:root:Average loss=0.09151561275124549 +INFO:root: +[Valid][4][UTT][ MSE = 0.1920 | LCC = 0.9693 | SRCC = 0.9689 ] +INFO:root:Average loss=0.08618325713276863 +INFO:root: +[Valid][5][UTT][ MSE = 0.1819 | LCC = 0.9691 | SRCC = 0.9692 ] +INFO:root:Average loss=0.08129747307300568 +INFO:root: +[Valid][6][UTT][ MSE = 0.1567 | LCC = 0.9695 | SRCC = 0.9698 ] +INFO:root:Average loss=0.07520125836133956 +INFO:root: +[Valid][7][UTT][ MSE = 0.1635 | LCC = 0.9712 | SRCC = 0.9705 ] +INFO:root:Average loss=0.0688714887201786 +INFO:root: +[Valid][8][UTT][ MSE = 0.1662 | LCC = 0.9713 | SRCC = 0.9710 ] +INFO:root:Average loss=0.06715723139047623 +INFO:root: +[Valid][9][UTT][ MSE = 0.1621 | LCC = 0.9720 | SRCC = 0.9706 ] +INFO:root:Average loss=0.06251087817549705 +INFO:root: +[Valid][10][UTT][ MSE = 0.1612 | LCC = 0.9725 | SRCC = 0.9717 ] +INFO:root:Average loss=0.05781457154452801 +INFO:root: +[Valid][11][UTT][ MSE = 0.1762 | LCC = 0.9710 | SRCC = 0.9704 ] +INFO:root:Average loss=0.056437445476651193 +INFO:root: +[Valid][12][UTT][ MSE = 0.1598 | LCC = 0.9725 | SRCC = 0.9712 ] +INFO:root:Average loss=0.052538169354200366 +INFO:root: +[Valid][13][UTT][ MSE = 0.1706 | LCC = 0.9724 | SRCC = 0.9715 ] +INFO:root:Average loss=0.05013393799960613 +INFO:root: +[Valid][14][UTT][ MSE = 0.1749 | LCC = 0.9723 | SRCC = 0.9714 ] +INFO:root:Average loss=0.050324108138680455 +INFO:root: +[Valid][15][UTT][ MSE = 0.1389 | LCC = 0.9721 | SRCC = 0.9712 ] +INFO:root:Average loss=0.04762813414633274 +INFO:root: +[Valid][16][UTT][ MSE = 0.1516 | LCC = 0.9731 | SRCC = 0.9719 ] +INFO:root:Average loss=0.04494772656261921 +INFO:root: +[Valid][17][UTT][ MSE = 0.1911 | LCC = 0.9720 | SRCC = 0.9719 ] +INFO:root:Average loss=0.0452342017441988 +INFO:root: +[Valid][18][UTT][ MSE = 0.1658 | LCC = 0.9725 | SRCC = 0.9712 ] +INFO:root:Average loss=0.040140078708529474 +INFO:root: +[Valid][19][UTT][ MSE = 0.1853 | LCC = 0.9700 | SRCC = 0.9701 ] +INFO:root:Average loss=0.03843080206215382 +INFO:root: +[Valid][20][UTT][ MSE = 0.1608 | LCC = 0.9725 | SRCC = 0.9716 ] +INFO:root:Average loss=0.037547754764556886 +INFO:root: +[Valid][21][UTT][ MSE = 0.1657 | LCC = 0.9724 | SRCC = 0.9714 ] +INFO:root:Average loss=0.035411317586898805 +INFO:root: +[Valid][22][UTT][ MSE = 0.1610 | LCC = 0.9721 | SRCC = 0.9710 ] +INFO:root:Average loss=0.03393556836247444 +INFO:root: +[Valid][23][UTT][ MSE = 0.1694 | LCC = 0.9723 | SRCC = 0.9709 ] +INFO:root:Average loss=0.033077495351433756 +INFO:root: +[Valid][24][UTT][ MSE = 0.1740 | LCC = 0.9716 | SRCC = 0.9700 ] +INFO:root:Average loss=0.03337259115278721 +INFO:root: +[Valid][25][UTT][ MSE = 0.1756 | LCC = 0.9715 | SRCC = 0.9705 ] +INFO:root:Average loss=0.03133353751897812 +INFO:root: +[Valid][26][UTT][ MSE = 0.1431 | LCC = 0.9721 | SRCC = 0.9714 ] +INFO:root:Average loss=0.02996137209236622 +INFO:root: +[Valid][27][UTT][ MSE = 0.1524 | LCC = 0.9722 | SRCC = 0.9710 ] +INFO:root:Average loss=0.02788070422410965 +INFO:root: +[Valid][28][UTT][ MSE = 0.1652 | LCC = 0.9712 | SRCC = 0.9701 ] +INFO:root:Average loss=0.02672907330095768 +INFO:root: +[Valid][29][UTT][ MSE = 0.1490 | LCC = 0.9717 | SRCC = 0.9711 ] +INFO:root:Average loss=0.025625803746283055 +INFO:root: +[Valid][30][UTT][ MSE = 0.1659 | LCC = 0.9716 | SRCC = 0.9706 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1494 | LCC = 0.9736 | SRCC = 0.9728 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run0/config.gin b/results_tencent/w2v2_xlsr_1b/layer22/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run0/model.pt b/results_tencent/w2v2_xlsr_1b/layer22/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f8c4ef8b0e4546a9cd3c1bc4ea3c2c36be688153 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bd8f4ac3382d55931ecc45edcce4f00cda9f278d8b00e657209b474741d0d381 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run0/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer22/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..2c3cfd2a5b3b6c3bd575aec68b5714e71e9828b6 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:339a33b1e566bf48e3024a52e070d8a41b6ea4860d2d825c304e01b537153168 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run0/train.log b/results_tencent/w2v2_xlsr_1b/layer22/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d9b06a9dcfa362bd685a0256e9101d2c12d5885a --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.37937953919172285 +INFO:root: +[Valid][0][UTT][ MSE = 0.2645 | LCC = 0.9507 | SRCC = 0.9498 ] +INFO:root:Average loss=0.14574682259559632 +INFO:root: +[Valid][1][UTT][ MSE = 0.2206 | LCC = 0.9581 | SRCC = 0.9585 ] +INFO:root:Average loss=0.12113588199019432 +INFO:root: +[Valid][2][UTT][ MSE = 0.2198 | LCC = 0.9636 | SRCC = 0.9634 ] +INFO:root:Average loss=0.10187250745296478 +INFO:root: +[Valid][3][UTT][ MSE = 0.1569 | LCC = 0.9657 | SRCC = 0.9655 ] +INFO:root:Average loss=0.09977854549884796 +INFO:root: +[Valid][4][UTT][ MSE = 0.2250 | LCC = 0.9668 | SRCC = 0.9666 ] +INFO:root:Average loss=0.08860442474484444 +INFO:root: +[Valid][5][UTT][ MSE = 0.1965 | LCC = 0.9681 | SRCC = 0.9677 ] +INFO:root:Average loss=0.08204916152358055 +INFO:root: +[Valid][6][UTT][ MSE = 0.1888 | LCC = 0.9692 | SRCC = 0.9686 ] +INFO:root:Average loss=0.07492547330260277 +INFO:root: +[Valid][7][UTT][ MSE = 0.1866 | LCC = 0.9698 | SRCC = 0.9692 ] +INFO:root:Average loss=0.0712622489631176 +INFO:root: +[Valid][8][UTT][ MSE = 0.1901 | LCC = 0.9696 | SRCC = 0.9689 ] +INFO:root:Average loss=0.06911715149879455 +INFO:root: +[Valid][9][UTT][ MSE = 0.1981 | LCC = 0.9706 | SRCC = 0.9695 ] +INFO:root:Average loss=0.06599970555305482 +INFO:root: +[Valid][10][UTT][ MSE = 0.1794 | LCC = 0.9705 | SRCC = 0.9696 ] +INFO:root:Average loss=0.06105960711836815 +INFO:root: +[Valid][11][UTT][ MSE = 0.1922 | LCC = 0.9713 | SRCC = 0.9702 ] +INFO:root:Average loss=0.058188834637403486 +INFO:root: +[Valid][12][UTT][ MSE = 0.1753 | LCC = 0.9714 | SRCC = 0.9705 ] +INFO:root:Average loss=0.05495348110795021 +INFO:root: +[Valid][13][UTT][ MSE = 0.1653 | LCC = 0.9720 | SRCC = 0.9707 ] +INFO:root:Average loss=0.053225999176502226 +INFO:root: +[Valid][14][UTT][ MSE = 0.1651 | LCC = 0.9718 | SRCC = 0.9710 ] +INFO:root:Average loss=0.05175414389371872 +INFO:root: +[Valid][15][UTT][ MSE = 0.1960 | LCC = 0.9702 | SRCC = 0.9703 ] +INFO:root:Average loss=0.0496994489133358 +INFO:root: +[Valid][16][UTT][ MSE = 0.2070 | LCC = 0.9713 | SRCC = 0.9708 ] +INFO:root:Average loss=0.04556361773610115 +INFO:root: +[Valid][17][UTT][ MSE = 0.1867 | LCC = 0.9715 | SRCC = 0.9712 ] +INFO:root:Average loss=0.04553603242337704 +INFO:root: +[Valid][18][UTT][ MSE = 0.1434 | LCC = 0.9715 | SRCC = 0.9712 ] +INFO:root:Average loss=0.04333217464387417 +INFO:root: +[Valid][19][UTT][ MSE = 0.1764 | LCC = 0.9722 | SRCC = 0.9714 ] +INFO:root:Average loss=0.04062513636052609 +INFO:root: +[Valid][20][UTT][ MSE = 0.1688 | LCC = 0.9716 | SRCC = 0.9705 ] +INFO:root:Average loss=0.0401006615459919 +INFO:root: +[Valid][21][UTT][ MSE = 0.1617 | LCC = 0.9718 | SRCC = 0.9705 ] +INFO:root:Average loss=0.036852670192718505 +INFO:root: +[Valid][22][UTT][ MSE = 0.1737 | LCC = 0.9719 | SRCC = 0.9712 ] +INFO:root:Average loss=0.035826219469308855 +INFO:root: +[Valid][23][UTT][ MSE = 0.1540 | LCC = 0.9719 | SRCC = 0.9707 ] +INFO:root:Average loss=0.033566534385085105 +INFO:root: +[Valid][24][UTT][ MSE = 0.1570 | LCC = 0.9700 | SRCC = 0.9703 ] +INFO:root:Average loss=0.03447055362164974 +INFO:root: +[Valid][25][UTT][ MSE = 0.1772 | LCC = 0.9719 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03611612038314343 +INFO:root: +[Valid][26][UTT][ MSE = 0.1815 | LCC = 0.9712 | SRCC = 0.9705 ] +INFO:root:Average loss=0.03186299842596054 +INFO:root: +[Valid][27][UTT][ MSE = 0.1662 | LCC = 0.9720 | SRCC = 0.9707 ] +INFO:root:Average loss=0.028734208680689335 +INFO:root: +[Valid][28][UTT][ MSE = 0.1959 | LCC = 0.9716 | SRCC = 0.9707 ] +INFO:root:Average loss=0.02781691065430641 +INFO:root: +[Valid][29][UTT][ MSE = 0.1546 | LCC = 0.9709 | SRCC = 0.9695 ] +INFO:root:Average loss=0.02941848750412464 +INFO:root: +[Valid][30][UTT][ MSE = 0.1626 | LCC = 0.9721 | SRCC = 0.9708 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1746 | LCC = 0.9727 | SRCC = 0.9724 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run1/config.gin b/results_tencent/w2v2_xlsr_1b/layer22/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run1/model.pt b/results_tencent/w2v2_xlsr_1b/layer22/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8304c166600d89b52956c54a73b1a5f7a4c647e8 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1814a2b461cd684a03bd5499fd09cdc08075b747012ca90de7406e745f4e4a9c +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run1/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer22/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..bb88a9e12abb3c408b22a068793695495151e801 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4e07d4c8980c830f72a18cce1bd2aeabb09ae13e89f53a4c981563da9faaa087 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run1/train.log b/results_tencent/w2v2_xlsr_1b/layer22/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..cf08409a9928fa09cc0d37a6936ae52eef12fc7c --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4246047520637512 +INFO:root: +[Valid][0][UTT][ MSE = 0.2482 | LCC = 0.9491 | SRCC = 0.9478 ] +INFO:root:Average loss=0.14683588683605195 +INFO:root: +[Valid][1][UTT][ MSE = 0.2368 | LCC = 0.9591 | SRCC = 0.9589 ] +INFO:root:Average loss=0.11941920477151871 +INFO:root: +[Valid][2][UTT][ MSE = 0.2191 | LCC = 0.9628 | SRCC = 0.9626 ] +INFO:root:Average loss=0.10759582251310348 +INFO:root: +[Valid][3][UTT][ MSE = 0.1915 | LCC = 0.9659 | SRCC = 0.9649 ] +INFO:root:Average loss=0.09334513467550277 +INFO:root: +[Valid][4][UTT][ MSE = 0.1837 | LCC = 0.9681 | SRCC = 0.9673 ] +INFO:root:Average loss=0.08827109098434448 +INFO:root: +[Valid][5][UTT][ MSE = 0.2022 | LCC = 0.9688 | SRCC = 0.9679 ] +INFO:root:Average loss=0.08352969825267792 +INFO:root: +[Valid][6][UTT][ MSE = 0.2107 | LCC = 0.9678 | SRCC = 0.9688 ] +INFO:root:Average loss=0.07608207947015762 +INFO:root: +[Valid][7][UTT][ MSE = 0.2241 | LCC = 0.9678 | SRCC = 0.9689 ] +INFO:root:Average loss=0.0745347059071064 +INFO:root: +[Valid][8][UTT][ MSE = 0.1779 | LCC = 0.9711 | SRCC = 0.9700 ] +INFO:root:Average loss=0.06859187203645706 +INFO:root: +[Valid][9][UTT][ MSE = 0.1770 | LCC = 0.9711 | SRCC = 0.9704 ] +INFO:root:Average loss=0.062314319282770154 +INFO:root: +[Valid][10][UTT][ MSE = 0.1838 | LCC = 0.9706 | SRCC = 0.9700 ] +INFO:root:Average loss=0.060152841567993165 +INFO:root: +[Valid][11][UTT][ MSE = 0.1720 | LCC = 0.9711 | SRCC = 0.9702 ] +INFO:root:Average loss=0.06470645758509636 +INFO:root: +[Valid][12][UTT][ MSE = 0.1808 | LCC = 0.9719 | SRCC = 0.9705 ] +INFO:root:Average loss=0.05321415974199772 +INFO:root: +[Valid][13][UTT][ MSE = 0.1637 | LCC = 0.9712 | SRCC = 0.9706 ] +INFO:root:Average loss=0.050584160804748535 +INFO:root: +[Valid][14][UTT][ MSE = 0.1892 | LCC = 0.9713 | SRCC = 0.9706 ] +INFO:root:Average loss=0.0492957464903593 +INFO:root: +[Valid][15][UTT][ MSE = 0.1612 | LCC = 0.9712 | SRCC = 0.9702 ] +INFO:root:Average loss=0.046301457822322846 +INFO:root: +[Valid][16][UTT][ MSE = 0.1624 | LCC = 0.9716 | SRCC = 0.9705 ] +INFO:root:Average loss=0.05063492216169834 +INFO:root: +[Valid][17][UTT][ MSE = 0.1499 | LCC = 0.9722 | SRCC = 0.9711 ] +INFO:root:Average loss=0.04491191278398037 +INFO:root: +[Valid][18][UTT][ MSE = 0.1617 | LCC = 0.9710 | SRCC = 0.9700 ] +INFO:root:Average loss=0.04290108174085617 +INFO:root: +[Valid][19][UTT][ MSE = 0.1575 | LCC = 0.9717 | SRCC = 0.9708 ] +INFO:root:Average loss=0.04315167415142059 +INFO:root: +[Valid][20][UTT][ MSE = 0.1718 | LCC = 0.9721 | SRCC = 0.9710 ] +INFO:root:Average loss=0.04085550820827484 +INFO:root: +[Valid][21][UTT][ MSE = 0.1505 | LCC = 0.9722 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03663671097159386 +INFO:root: +[Valid][22][UTT][ MSE = 0.1904 | LCC = 0.9702 | SRCC = 0.9700 ] +INFO:root:Average loss=0.0381114282310009 +INFO:root: +[Valid][23][UTT][ MSE = 0.1488 | LCC = 0.9727 | SRCC = 0.9712 ] +INFO:root:Average loss=0.03412944926321507 +INFO:root: +[Valid][24][UTT][ MSE = 0.1619 | LCC = 0.9721 | SRCC = 0.9707 ] +INFO:root:Average loss=0.034592442244291303 +INFO:root: +[Valid][25][UTT][ MSE = 0.1751 | LCC = 0.9709 | SRCC = 0.9705 ] +INFO:root:Average loss=0.032631010681390765 +INFO:root: +[Valid][26][UTT][ MSE = 0.1390 | LCC = 0.9722 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03042756673693657 +INFO:root: +[Valid][27][UTT][ MSE = 0.1507 | LCC = 0.9718 | SRCC = 0.9703 ] +INFO:root:Average loss=0.030036949530243873 +INFO:root: +[Valid][28][UTT][ MSE = 0.1598 | LCC = 0.9695 | SRCC = 0.9704 ] +INFO:root:Average loss=0.030025996416807175 +INFO:root: +[Valid][29][UTT][ MSE = 0.1713 | LCC = 0.9723 | SRCC = 0.9706 ] +INFO:root:Average loss=0.027538332968950273 +INFO:root: +[Valid][30][UTT][ MSE = 0.1614 | LCC = 0.9731 | SRCC = 0.9716 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1590 | LCC = 0.9733 | SRCC = 0.9727 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run2/config.gin b/results_tencent/w2v2_xlsr_1b/layer22/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run2/model.pt b/results_tencent/w2v2_xlsr_1b/layer22/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7689ac978afbe3cd041291b0692c1771f7bbc2a2 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fc0c09484c1db396d6e729304195615985be85dfb05a30f3283a3e39d1c0d011 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run2/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer22/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..db61f78eb81102f9366c879b74777544f0f52652 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f7406a0ffbbe48dde2c44efea093cce3f3288da1200370585874f628628129a +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run2/train.log b/results_tencent/w2v2_xlsr_1b/layer22/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..4a707d4d95615b8f4536505526813ae83699c16e --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4043537237644196 +INFO:root: +[Valid][0][UTT][ MSE = 0.3013 | LCC = 0.9419 | SRCC = 0.9388 ] +INFO:root:Average loss=0.15517138469219208 +INFO:root: +[Valid][1][UTT][ MSE = 0.2188 | LCC = 0.9576 | SRCC = 0.9563 ] +INFO:root:Average loss=0.12454044950008393 +INFO:root: +[Valid][2][UTT][ MSE = 0.2139 | LCC = 0.9629 | SRCC = 0.9614 ] +INFO:root:Average loss=0.10637059581279755 +INFO:root: +[Valid][3][UTT][ MSE = 0.2081 | LCC = 0.9649 | SRCC = 0.9632 ] +INFO:root:Average loss=0.09684810376167298 +INFO:root: +[Valid][4][UTT][ MSE = 0.2068 | LCC = 0.9659 | SRCC = 0.9654 ] +INFO:root:Average loss=0.09142204508185386 +INFO:root: +[Valid][5][UTT][ MSE = 0.1815 | LCC = 0.9679 | SRCC = 0.9669 ] +INFO:root:Average loss=0.08728616032004356 +INFO:root: +[Valid][6][UTT][ MSE = 0.2047 | LCC = 0.9684 | SRCC = 0.9669 ] +INFO:root:Average loss=0.07741537895798684 +INFO:root: +[Valid][7][UTT][ MSE = 0.1898 | LCC = 0.9695 | SRCC = 0.9689 ] +INFO:root:Average loss=0.07771297198534012 +INFO:root: +[Valid][8][UTT][ MSE = 0.1906 | LCC = 0.9680 | SRCC = 0.9680 ] +INFO:root:Average loss=0.07779032796621323 +INFO:root: +[Valid][9][UTT][ MSE = 0.2205 | LCC = 0.9704 | SRCC = 0.9699 ] +INFO:root:Average loss=0.06538223922252655 +INFO:root: +[Valid][10][UTT][ MSE = 0.1570 | LCC = 0.9711 | SRCC = 0.9700 ] +INFO:root:Average loss=0.06283322477340698 +INFO:root: +[Valid][11][UTT][ MSE = 0.1597 | LCC = 0.9694 | SRCC = 0.9693 ] +INFO:root:Average loss=0.062115086793899536 +INFO:root: +[Valid][12][UTT][ MSE = 0.1758 | LCC = 0.9712 | SRCC = 0.9696 ] +INFO:root:Average loss=0.05739867505431175 +INFO:root: +[Valid][13][UTT][ MSE = 0.1512 | LCC = 0.9722 | SRCC = 0.9709 ] +INFO:root:Average loss=0.053644079342484476 +INFO:root: +[Valid][14][UTT][ MSE = 0.1955 | LCC = 0.9713 | SRCC = 0.9708 ] +INFO:root:Average loss=0.052001552879810335 +INFO:root: +[Valid][15][UTT][ MSE = 0.1669 | LCC = 0.9712 | SRCC = 0.9699 ] +INFO:root:Average loss=0.053126089930534366 +INFO:root: +[Valid][16][UTT][ MSE = 0.1982 | LCC = 0.9722 | SRCC = 0.9714 ] +INFO:root:Average loss=0.04671275283396244 +INFO:root: +[Valid][17][UTT][ MSE = 0.1485 | LCC = 0.9713 | SRCC = 0.9712 ] +INFO:root:Average loss=0.0453993948251009 +INFO:root: +[Valid][18][UTT][ MSE = 0.1858 | LCC = 0.9724 | SRCC = 0.9709 ] +INFO:root:Average loss=0.044282516866922375 +INFO:root: +[Valid][19][UTT][ MSE = 0.1883 | LCC = 0.9723 | SRCC = 0.9711 ] +INFO:root:Average loss=0.04278640478849411 +INFO:root: +[Valid][20][UTT][ MSE = 0.1809 | LCC = 0.9720 | SRCC = 0.9717 ] +INFO:root:Average loss=0.04182033751904964 +INFO:root: +[Valid][21][UTT][ MSE = 0.1609 | LCC = 0.9728 | SRCC = 0.9719 ] +INFO:root:Average loss=0.038518547222018244 +INFO:root: +[Valid][22][UTT][ MSE = 0.1666 | LCC = 0.9725 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03766340982913971 +INFO:root: +[Valid][23][UTT][ MSE = 0.1824 | LCC = 0.9718 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03426871745288372 +INFO:root: +[Valid][24][UTT][ MSE = 0.1590 | LCC = 0.9728 | SRCC = 0.9718 ] +INFO:root:Average loss=0.034008217565715315 +INFO:root: +[Valid][25][UTT][ MSE = 0.1731 | LCC = 0.9721 | SRCC = 0.9715 ] +INFO:root:Average loss=0.033000023752450945 +INFO:root: +[Valid][26][UTT][ MSE = 0.1671 | LCC = 0.9728 | SRCC = 0.9716 ] +INFO:root:Average loss=0.033417309358716014 +INFO:root: +[Valid][27][UTT][ MSE = 0.1650 | LCC = 0.9724 | SRCC = 0.9710 ] +INFO:root:Average loss=0.0298299660384655 +INFO:root: +[Valid][28][UTT][ MSE = 0.1538 | LCC = 0.9720 | SRCC = 0.9710 ] +INFO:root:Average loss=0.027627335518598556 +INFO:root: +[Valid][29][UTT][ MSE = 0.1596 | LCC = 0.9725 | SRCC = 0.9710 ] +INFO:root:Average loss=0.02677767913043499 +INFO:root: +[Valid][30][UTT][ MSE = 0.1784 | LCC = 0.9718 | SRCC = 0.9708 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1578 | LCC = 0.9734 | SRCC = 0.9730 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run3/config.gin b/results_tencent/w2v2_xlsr_1b/layer22/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run3/model.pt b/results_tencent/w2v2_xlsr_1b/layer22/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3fe6b16449ba17355d6576e5f0d78ac3e5499303 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:075847156349c5f168df46ecfe2ba3293a170d1fa3f61aa6fa783ee14a6b046b +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run3/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer22/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4a0e5e92aa591ac733f9371a9871384d653ce44a --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df621e15657e9d504584ebcd9f780ac953ada1f68bbd17d59ca4347fcea16efd +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run3/train.log b/results_tencent/w2v2_xlsr_1b/layer22/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d347db5825af98a696a37a9278190981ccda3c0a --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.39150518345832824 +INFO:root: +[Valid][0][UTT][ MSE = 0.2285 | LCC = 0.9486 | SRCC = 0.9481 ] +INFO:root:Average loss=0.15502199351787568 +INFO:root: +[Valid][1][UTT][ MSE = 0.2637 | LCC = 0.9578 | SRCC = 0.9568 ] +INFO:root:Average loss=0.1187450429201126 +INFO:root: +[Valid][2][UTT][ MSE = 0.2050 | LCC = 0.9624 | SRCC = 0.9619 ] +INFO:root:Average loss=0.11177939867973327 +INFO:root: +[Valid][3][UTT][ MSE = 0.1636 | LCC = 0.9657 | SRCC = 0.9656 ] +INFO:root:Average loss=0.09912158933281899 +INFO:root: +[Valid][4][UTT][ MSE = 0.1763 | LCC = 0.9661 | SRCC = 0.9664 ] +INFO:root:Average loss=0.09302311715483666 +INFO:root: +[Valid][5][UTT][ MSE = 0.2061 | LCC = 0.9684 | SRCC = 0.9685 ] +INFO:root:Average loss=0.08444374319911004 +INFO:root: +[Valid][6][UTT][ MSE = 0.1614 | LCC = 0.9689 | SRCC = 0.9685 ] +INFO:root:Average loss=0.08502554014325142 +INFO:root: +[Valid][7][UTT][ MSE = 0.1886 | LCC = 0.9696 | SRCC = 0.9697 ] +INFO:root:Average loss=0.07555209478735923 +INFO:root: +[Valid][8][UTT][ MSE = 0.1689 | LCC = 0.9708 | SRCC = 0.9700 ] +INFO:root:Average loss=0.07087563818693161 +INFO:root: +[Valid][9][UTT][ MSE = 0.2106 | LCC = 0.9706 | SRCC = 0.9698 ] +INFO:root:Average loss=0.07365060272812843 +INFO:root: +[Valid][10][UTT][ MSE = 0.1665 | LCC = 0.9713 | SRCC = 0.9705 ] +INFO:root:Average loss=0.06432029271125793 +INFO:root: +[Valid][11][UTT][ MSE = 0.2145 | LCC = 0.9710 | SRCC = 0.9715 ] +INFO:root:Average loss=0.06373470759391785 +INFO:root: +[Valid][12][UTT][ MSE = 0.1590 | LCC = 0.9727 | SRCC = 0.9719 ] +INFO:root:Average loss=0.05537628436088562 +INFO:root: +[Valid][13][UTT][ MSE = 0.1585 | LCC = 0.9719 | SRCC = 0.9716 ] +INFO:root:Average loss=0.052565513387322425 +INFO:root: +[Valid][14][UTT][ MSE = 0.1474 | LCC = 0.9686 | SRCC = 0.9701 ] +INFO:root:Average loss=0.05180085152387619 +INFO:root: +[Valid][15][UTT][ MSE = 0.1700 | LCC = 0.9728 | SRCC = 0.9720 ] +INFO:root:Average loss=0.05086081789433956 +INFO:root: +[Valid][16][UTT][ MSE = 0.1785 | LCC = 0.9723 | SRCC = 0.9718 ] +INFO:root:Average loss=0.048337286964058875 +INFO:root: +[Valid][17][UTT][ MSE = 0.1503 | LCC = 0.9716 | SRCC = 0.9705 ] +INFO:root:Average loss=0.04538563098013401 +INFO:root: +[Valid][18][UTT][ MSE = 0.1792 | LCC = 0.9725 | SRCC = 0.9721 ] +INFO:root:Average loss=0.041546905130147936 +INFO:root: +[Valid][19][UTT][ MSE = 0.1820 | LCC = 0.9725 | SRCC = 0.9717 ] +INFO:root:Average loss=0.04141782931983471 +INFO:root: +[Valid][20][UTT][ MSE = 0.1550 | LCC = 0.9725 | SRCC = 0.9715 ] +INFO:root:Average loss=0.03868637517094612 +INFO:root: +[Valid][21][UTT][ MSE = 0.1676 | LCC = 0.9722 | SRCC = 0.9714 ] +INFO:root:Average loss=0.03750641056895256 +INFO:root: +[Valid][22][UTT][ MSE = 0.1613 | LCC = 0.9723 | SRCC = 0.9715 ] +INFO:root:Average loss=0.03844325639307499 +INFO:root: +[Valid][23][UTT][ MSE = 0.1570 | LCC = 0.9726 | SRCC = 0.9718 ] +INFO:root:Average loss=0.0369047294408083 +INFO:root: +[Valid][24][UTT][ MSE = 0.1665 | LCC = 0.9719 | SRCC = 0.9711 ] +INFO:root:Average loss=0.0360407231003046 +INFO:root: +[Valid][25][UTT][ MSE = 0.1639 | LCC = 0.9728 | SRCC = 0.9716 ] +INFO:root:Average loss=0.03275169218331576 +INFO:root: +[Valid][26][UTT][ MSE = 0.1662 | LCC = 0.9725 | SRCC = 0.9715 ] +INFO:root:Average loss=0.033341304682195184 +INFO:root: +[Valid][27][UTT][ MSE = 0.1784 | LCC = 0.9715 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03355401846766472 +INFO:root: +[Valid][28][UTT][ MSE = 0.1649 | LCC = 0.9730 | SRCC = 0.9721 ] +INFO:root:Average loss=0.030332424744963647 +INFO:root: +[Valid][29][UTT][ MSE = 0.1577 | LCC = 0.9719 | SRCC = 0.9713 ] +INFO:root:Average loss=0.02867438840866089 +INFO:root: +[Valid][30][UTT][ MSE = 0.1574 | LCC = 0.9731 | SRCC = 0.9724 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1546 | LCC = 0.9737 | SRCC = 0.9734 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run4/config.gin b/results_tencent/w2v2_xlsr_1b/layer22/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run4/model.pt b/results_tencent/w2v2_xlsr_1b/layer22/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..7fc9e660caaf860f6744ee23a12e24dd97cafcfa --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a3f76b0736ac5065e2527c297303c8db22c4ad63920d22cc4c804291c961230a +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run4/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer22/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ba43caf7a98a393aa7e5196849dea30b13c2135f --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:21723e6151a5890b2372cb53f23207b7adc665e0cfbc3c08a317722429c0cc00 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer22/run4/train.log b/results_tencent/w2v2_xlsr_1b/layer22/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..55c9fc7d8fd1b50fe4d68ee1f5bd0303d75b68cf --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer22/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.43030133765935896 +INFO:root: +[Valid][0][UTT][ MSE = 0.2446 | LCC = 0.9462 | SRCC = 0.9454 ] +INFO:root:Average loss=0.14642684072256087 +INFO:root: +[Valid][1][UTT][ MSE = 0.2393 | LCC = 0.9567 | SRCC = 0.9552 ] +INFO:root:Average loss=0.11600806373357773 +INFO:root: +[Valid][2][UTT][ MSE = 0.2496 | LCC = 0.9627 | SRCC = 0.9619 ] +INFO:root:Average loss=0.10234499624371529 +INFO:root: +[Valid][3][UTT][ MSE = 0.2309 | LCC = 0.9637 | SRCC = 0.9643 ] +INFO:root:Average loss=0.09377273899316788 +INFO:root: +[Valid][4][UTT][ MSE = 0.2199 | LCC = 0.9668 | SRCC = 0.9665 ] +INFO:root:Average loss=0.08724287483096123 +INFO:root: +[Valid][5][UTT][ MSE = 0.2361 | LCC = 0.9675 | SRCC = 0.9662 ] +INFO:root:Average loss=0.0830313409268856 +INFO:root: +[Valid][6][UTT][ MSE = 0.1907 | LCC = 0.9681 | SRCC = 0.9674 ] +INFO:root:Average loss=0.0759401403516531 +INFO:root: +[Valid][7][UTT][ MSE = 0.1775 | LCC = 0.9692 | SRCC = 0.9684 ] +INFO:root:Average loss=0.0732645339667797 +INFO:root: +[Valid][8][UTT][ MSE = 0.1786 | LCC = 0.9694 | SRCC = 0.9681 ] +INFO:root:Average loss=0.06715850737690926 +INFO:root: +[Valid][9][UTT][ MSE = 0.1830 | LCC = 0.9682 | SRCC = 0.9673 ] +INFO:root:Average loss=0.06413132463395596 +INFO:root: +[Valid][10][UTT][ MSE = 0.1940 | LCC = 0.9700 | SRCC = 0.9692 ] +INFO:root:Average loss=0.06143205624818802 +INFO:root: +[Valid][11][UTT][ MSE = 0.2149 | LCC = 0.9698 | SRCC = 0.9692 ] +INFO:root:Average loss=0.0585558924973011 +INFO:root: +[Valid][12][UTT][ MSE = 0.1597 | LCC = 0.9688 | SRCC = 0.9676 ] +INFO:root:Average loss=0.05476798582077026 +INFO:root: +[Valid][13][UTT][ MSE = 0.1754 | LCC = 0.9708 | SRCC = 0.9699 ] +INFO:root:Average loss=0.05084435683488846 +INFO:root: +[Valid][14][UTT][ MSE = 0.1762 | LCC = 0.9692 | SRCC = 0.9685 ] +INFO:root:Average loss=0.05081777337193489 +INFO:root: +[Valid][15][UTT][ MSE = 0.1905 | LCC = 0.9696 | SRCC = 0.9692 ] +INFO:root:Average loss=0.047188512593507766 +INFO:root: +[Valid][16][UTT][ MSE = 0.1843 | LCC = 0.9704 | SRCC = 0.9698 ] +INFO:root:Average loss=0.043818410590291026 +INFO:root: +[Valid][17][UTT][ MSE = 0.1707 | LCC = 0.9709 | SRCC = 0.9700 ] +INFO:root:Average loss=0.045545006960630414 +INFO:root: +[Valid][18][UTT][ MSE = 0.1578 | LCC = 0.9685 | SRCC = 0.9684 ] +INFO:root:Average loss=0.04105765233933926 +INFO:root: +[Valid][19][UTT][ MSE = 0.1499 | LCC = 0.9704 | SRCC = 0.9697 ] +INFO:root:Average loss=0.03855113135278225 +INFO:root: +[Valid][20][UTT][ MSE = 0.1667 | LCC = 0.9706 | SRCC = 0.9694 ] +INFO:root:Average loss=0.03751187726855278 +INFO:root: +[Valid][21][UTT][ MSE = 0.1773 | LCC = 0.9700 | SRCC = 0.9695 ] +INFO:root:Average loss=0.03872545289993286 +INFO:root: +[Valid][22][UTT][ MSE = 0.1708 | LCC = 0.9702 | SRCC = 0.9693 ] +INFO:root:Average loss=0.03433785055577755 +INFO:root: +[Valid][23][UTT][ MSE = 0.1491 | LCC = 0.9698 | SRCC = 0.9687 ] +INFO:root:Average loss=0.03374849607050419 +INFO:root: +[Valid][24][UTT][ MSE = 0.1801 | LCC = 0.9705 | SRCC = 0.9695 ] +INFO:root:Average loss=0.03141722223162651 +INFO:root: +[Valid][25][UTT][ MSE = 0.1761 | LCC = 0.9699 | SRCC = 0.9691 ] +INFO:root:Average loss=0.030449796795845033 +INFO:root: +[Valid][26][UTT][ MSE = 0.1564 | LCC = 0.9695 | SRCC = 0.9687 ] +INFO:root:Average loss=0.028640971899032593 +INFO:root: +[Valid][27][UTT][ MSE = 0.1893 | LCC = 0.9694 | SRCC = 0.9690 ] +INFO:root:Average loss=0.03004670822620392 +INFO:root: +[Valid][28][UTT][ MSE = 0.1452 | LCC = 0.9698 | SRCC = 0.9692 ] +INFO:root:Average loss=0.028374969258904457 +INFO:root: +[Valid][29][UTT][ MSE = 0.1586 | LCC = 0.9708 | SRCC = 0.9698 ] +INFO:root:Average loss=0.02548914574086666 +INFO:root: +[Valid][30][UTT][ MSE = 0.1502 | LCC = 0.9700 | SRCC = 0.9690 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1657 | LCC = 0.9735 | SRCC = 0.9733 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run0/config.gin b/results_tencent/w2v2_xlsr_1b/layer24/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run0/model.pt b/results_tencent/w2v2_xlsr_1b/layer24/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8c88286be4c3d9b3ad2e7281c12c3f004ae40691 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:309fd922c8f24ccfc5b88ed6e5157df296be173f92e205d6141a618f081d36cd +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run0/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer24/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..762b05ed98c0800cf8b2d050001eddc2a79e3f09 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ef91a4ec25ab9771629da00ec08beb5d7ef7fdabbb29b53e2eac964ef8e937b7 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run0/train.log b/results_tencent/w2v2_xlsr_1b/layer24/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..417b593aa345ff09a85c8cc3e8db2a23bacddf00 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5355126326084136 +INFO:root: +[Valid][0][UTT][ MSE = 0.2670 | LCC = 0.9425 | SRCC = 0.9386 ] +INFO:root:Average loss=0.15569925767183304 +INFO:root: +[Valid][1][UTT][ MSE = 0.2265 | LCC = 0.9565 | SRCC = 0.9545 ] +INFO:root:Average loss=0.13197693702578545 +INFO:root: +[Valid][2][UTT][ MSE = 0.1892 | LCC = 0.9604 | SRCC = 0.9593 ] +INFO:root:Average loss=0.11201806634664535 +INFO:root: +[Valid][3][UTT][ MSE = 0.2184 | LCC = 0.9637 | SRCC = 0.9641 ] +INFO:root:Average loss=0.10981024634838105 +INFO:root: +[Valid][4][UTT][ MSE = 0.2058 | LCC = 0.9662 | SRCC = 0.9647 ] +INFO:root:Average loss=0.09812929919362068 +INFO:root: +[Valid][5][UTT][ MSE = 0.2130 | LCC = 0.9669 | SRCC = 0.9666 ] +INFO:root:Average loss=0.08794527116417884 +INFO:root: +[Valid][6][UTT][ MSE = 0.1675 | LCC = 0.9684 | SRCC = 0.9682 ] +INFO:root:Average loss=0.08128042095899582 +INFO:root: +[Valid][7][UTT][ MSE = 0.1621 | LCC = 0.9701 | SRCC = 0.9692 ] +INFO:root:Average loss=0.07911027926206589 +INFO:root: +[Valid][8][UTT][ MSE = 0.1838 | LCC = 0.9697 | SRCC = 0.9690 ] +INFO:root:Average loss=0.07580164450407029 +INFO:root: +[Valid][9][UTT][ MSE = 0.1950 | LCC = 0.9701 | SRCC = 0.9693 ] +INFO:root:Average loss=0.07492637515068054 +INFO:root: +[Valid][10][UTT][ MSE = 0.1585 | LCC = 0.9713 | SRCC = 0.9705 ] +INFO:root:Average loss=0.06465820562839508 +INFO:root: +[Valid][11][UTT][ MSE = 0.1786 | LCC = 0.9713 | SRCC = 0.9705 ] +INFO:root:Average loss=0.06105948704481125 +INFO:root: +[Valid][12][UTT][ MSE = 0.1734 | LCC = 0.9701 | SRCC = 0.9690 ] +INFO:root:Average loss=0.06161331921815872 +INFO:root: +[Valid][13][UTT][ MSE = 0.1647 | LCC = 0.9706 | SRCC = 0.9699 ] +INFO:root:Average loss=0.05767187532782555 +INFO:root: +[Valid][14][UTT][ MSE = 0.1698 | LCC = 0.9719 | SRCC = 0.9707 ] +INFO:root:Average loss=0.05845670476555824 +INFO:root: +[Valid][15][UTT][ MSE = 0.1728 | LCC = 0.9718 | SRCC = 0.9707 ] +INFO:root:Average loss=0.05168749587237835 +INFO:root: +[Valid][16][UTT][ MSE = 0.1471 | LCC = 0.9721 | SRCC = 0.9711 ] +INFO:root:Average loss=0.049812612950801846 +INFO:root: +[Valid][17][UTT][ MSE = 0.1605 | LCC = 0.9721 | SRCC = 0.9710 ] +INFO:root:Average loss=0.046987386152148244 +INFO:root: +[Valid][18][UTT][ MSE = 0.1829 | LCC = 0.9709 | SRCC = 0.9703 ] +INFO:root:Average loss=0.04464157697558403 +INFO:root: +[Valid][19][UTT][ MSE = 0.1638 | LCC = 0.9716 | SRCC = 0.9704 ] +INFO:root:Average loss=0.04567965476214886 +INFO:root: +[Valid][20][UTT][ MSE = 0.1525 | LCC = 0.9705 | SRCC = 0.9701 ] +INFO:root:Average loss=0.04162967924773693 +INFO:root: +[Valid][21][UTT][ MSE = 0.1921 | LCC = 0.9717 | SRCC = 0.9706 ] +INFO:root:Average loss=0.0385126746147871 +INFO:root: +[Valid][22][UTT][ MSE = 0.1554 | LCC = 0.9727 | SRCC = 0.9716 ] +INFO:root:Average loss=0.04024072256684303 +INFO:root: +[Valid][23][UTT][ MSE = 0.1470 | LCC = 0.9719 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03725295488536358 +INFO:root: +[Valid][24][UTT][ MSE = 0.1732 | LCC = 0.9717 | SRCC = 0.9710 ] +INFO:root:Average loss=0.035517090067267415 +INFO:root: +[Valid][25][UTT][ MSE = 0.1722 | LCC = 0.9714 | SRCC = 0.9707 ] +INFO:root:Average loss=0.0357254538834095 +INFO:root: +[Valid][26][UTT][ MSE = 0.1577 | LCC = 0.9722 | SRCC = 0.9712 ] +INFO:root:Average loss=0.03453656834363937 +INFO:root: +[Valid][27][UTT][ MSE = 0.1793 | LCC = 0.9724 | SRCC = 0.9711 ] +INFO:root:Average loss=0.03137255042791367 +INFO:root: +[Valid][28][UTT][ MSE = 0.1540 | LCC = 0.9713 | SRCC = 0.9707 ] +INFO:root:Average loss=0.030802138105034828 +INFO:root: +[Valid][29][UTT][ MSE = 0.1588 | LCC = 0.9723 | SRCC = 0.9712 ] +INFO:root:Average loss=0.02905023954808712 +INFO:root: +[Valid][30][UTT][ MSE = 0.1596 | LCC = 0.9716 | SRCC = 0.9702 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1534 | LCC = 0.9736 | SRCC = 0.9732 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run1/config.gin b/results_tencent/w2v2_xlsr_1b/layer24/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run1/model.pt b/results_tencent/w2v2_xlsr_1b/layer24/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..80cbce5e472bcf5b19fb4cd3d88a838a5ae8a981 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:977b2d9ff1a616cf10fd697ebd34a3f7dd4656aca71e6c82c5ad9bdc79942af3 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run1/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer24/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4501a1006cd14674f4e7ea546c529c57c3c6a65e --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc409d7fbaae093482ce1004a70979197babfaab4a2b7d0de04bb89c4acf319d +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run1/train.log b/results_tencent/w2v2_xlsr_1b/layer24/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..b2f191a519e1778e2623dcd8b8b2b034015cea52 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.376969289124012 +INFO:root: +[Valid][0][UTT][ MSE = 0.2066 | LCC = 0.9435 | SRCC = 0.9435 ] +INFO:root:Average loss=0.15562599021196366 +INFO:root: +[Valid][1][UTT][ MSE = 0.2435 | LCC = 0.9572 | SRCC = 0.9558 ] +INFO:root:Average loss=0.12761595928668976 +INFO:root: +[Valid][2][UTT][ MSE = 0.1604 | LCC = 0.9621 | SRCC = 0.9622 ] +INFO:root:Average loss=0.11458987614512443 +INFO:root: +[Valid][3][UTT][ MSE = 0.1707 | LCC = 0.9647 | SRCC = 0.9646 ] +INFO:root:Average loss=0.10026471227407456 +INFO:root: +[Valid][4][UTT][ MSE = 0.1532 | LCC = 0.9663 | SRCC = 0.9664 ] +INFO:root:Average loss=0.09011134243011475 +INFO:root: +[Valid][5][UTT][ MSE = 0.1556 | LCC = 0.9682 | SRCC = 0.9667 ] +INFO:root:Average loss=0.08452015307545661 +INFO:root: +[Valid][6][UTT][ MSE = 0.1652 | LCC = 0.9694 | SRCC = 0.9686 ] +INFO:root:Average loss=0.08103491216897965 +INFO:root: +[Valid][7][UTT][ MSE = 0.1733 | LCC = 0.9675 | SRCC = 0.9674 ] +INFO:root:Average loss=0.07772875413298606 +INFO:root: +[Valid][8][UTT][ MSE = 0.1629 | LCC = 0.9706 | SRCC = 0.9699 ] +INFO:root:Average loss=0.07042174053192139 +INFO:root: +[Valid][9][UTT][ MSE = 0.2050 | LCC = 0.9635 | SRCC = 0.9657 ] +INFO:root:Average loss=0.07247468128800393 +INFO:root: +[Valid][10][UTT][ MSE = 0.1734 | LCC = 0.9707 | SRCC = 0.9696 ] +INFO:root:Average loss=0.06813131830096245 +INFO:root: +[Valid][11][UTT][ MSE = 0.1619 | LCC = 0.9687 | SRCC = 0.9692 ] +INFO:root:Average loss=0.061732780307531356 +INFO:root: +[Valid][12][UTT][ MSE = 0.1533 | LCC = 0.9711 | SRCC = 0.9701 ] +INFO:root:Average loss=0.0594680991768837 +INFO:root: +[Valid][13][UTT][ MSE = 0.1530 | LCC = 0.9714 | SRCC = 0.9706 ] +INFO:root:Average loss=0.05544850179553032 +INFO:root: +[Valid][14][UTT][ MSE = 0.1773 | LCC = 0.9720 | SRCC = 0.9707 ] +INFO:root:Average loss=0.05405823548138142 +INFO:root: +[Valid][15][UTT][ MSE = 0.1614 | LCC = 0.9709 | SRCC = 0.9702 ] +INFO:root:Average loss=0.05121202683448792 +INFO:root: +[Valid][16][UTT][ MSE = 0.1611 | LCC = 0.9722 | SRCC = 0.9711 ] +INFO:root:Average loss=0.049384115993976595 +INFO:root: +[Valid][17][UTT][ MSE = 0.1575 | LCC = 0.9719 | SRCC = 0.9708 ] +INFO:root:Average loss=0.04913942894339562 +INFO:root: +[Valid][18][UTT][ MSE = 0.1473 | LCC = 0.9717 | SRCC = 0.9708 ] +INFO:root:Average loss=0.044412645772099496 +INFO:root: +[Valid][19][UTT][ MSE = 0.1639 | LCC = 0.9720 | SRCC = 0.9708 ] +INFO:root:Average loss=0.04390525847673416 +INFO:root: +[Valid][20][UTT][ MSE = 0.1397 | LCC = 0.9720 | SRCC = 0.9709 ] +INFO:root:Average loss=0.04343072861433029 +INFO:root: +[Valid][21][UTT][ MSE = 0.1798 | LCC = 0.9700 | SRCC = 0.9702 ] +INFO:root:Average loss=0.041303184241056445 +INFO:root: +[Valid][22][UTT][ MSE = 0.1543 | LCC = 0.9716 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03793208423256874 +INFO:root: +[Valid][23][UTT][ MSE = 0.1616 | LCC = 0.9714 | SRCC = 0.9702 ] +INFO:root:Average loss=0.03853663863241673 +INFO:root: +[Valid][24][UTT][ MSE = 0.1629 | LCC = 0.9714 | SRCC = 0.9702 ] +INFO:root:Average loss=0.037508022889494896 +INFO:root: +[Valid][25][UTT][ MSE = 0.1622 | LCC = 0.9718 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03472378431260586 +INFO:root: +[Valid][26][UTT][ MSE = 0.1629 | LCC = 0.9720 | SRCC = 0.9707 ] +INFO:root:Average loss=0.03301169176399708 +INFO:root: +[Valid][27][UTT][ MSE = 0.1850 | LCC = 0.9714 | SRCC = 0.9707 ] +INFO:root:Average loss=0.03225054475665092 +INFO:root: +[Valid][28][UTT][ MSE = 0.1660 | LCC = 0.9711 | SRCC = 0.9701 ] +INFO:root:Average loss=0.02991313521564007 +INFO:root: +[Valid][29][UTT][ MSE = 0.1477 | LCC = 0.9718 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03144907458126545 +INFO:root: +[Valid][30][UTT][ MSE = 0.1524 | LCC = 0.9712 | SRCC = 0.9703 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1602 | LCC = 0.9726 | SRCC = 0.9718 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run2/config.gin b/results_tencent/w2v2_xlsr_1b/layer24/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run2/model.pt b/results_tencent/w2v2_xlsr_1b/layer24/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..39b8a6e17eb1859981b0c2300698b20185dea382 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d0530195269b9ad0ca46a5f4d37d4c8576a07f821b1b12e498d8a60ad9a2cf70 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run2/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer24/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c23461a7e272fc9cfa49b5f9bb298ebf43681374 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e5848599b41e012c2d7edcec101a7456f0f3f838263784912393a4681b89add9 +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run2/train.log b/results_tencent/w2v2_xlsr_1b/layer24/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9e66c202d8f4d445f699cc78839117e53ffcd37e --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4062404305934906 +INFO:root: +[Valid][0][UTT][ MSE = 0.2167 | LCC = 0.9366 | SRCC = 0.9337 ] +INFO:root:Average loss=0.15511678439378737 +INFO:root: +[Valid][1][UTT][ MSE = 0.2148 | LCC = 0.9553 | SRCC = 0.9538 ] +INFO:root:Average loss=0.131702074944973 +INFO:root: +[Valid][2][UTT][ MSE = 0.1794 | LCC = 0.9624 | SRCC = 0.9616 ] +INFO:root:Average loss=0.10857149785757064 +INFO:root: +[Valid][3][UTT][ MSE = 0.2107 | LCC = 0.9646 | SRCC = 0.9640 ] +INFO:root:Average loss=0.10048070931434631 +INFO:root: +[Valid][4][UTT][ MSE = 0.1801 | LCC = 0.9653 | SRCC = 0.9651 ] +INFO:root:Average loss=0.09157172659039498 +INFO:root: +[Valid][5][UTT][ MSE = 0.1655 | LCC = 0.9683 | SRCC = 0.9673 ] +INFO:root:Average loss=0.08649664866924286 +INFO:root: +[Valid][6][UTT][ MSE = 0.1800 | LCC = 0.9679 | SRCC = 0.9665 ] +INFO:root:Average loss=0.08091924133896827 +INFO:root: +[Valid][7][UTT][ MSE = 0.1296 | LCC = 0.9691 | SRCC = 0.9685 ] +INFO:root:Average loss=0.07464925503730774 +INFO:root: +[Valid][8][UTT][ MSE = 0.1826 | LCC = 0.9695 | SRCC = 0.9687 ] +INFO:root:Average loss=0.07302123457193374 +INFO:root: +[Valid][9][UTT][ MSE = 0.1381 | LCC = 0.9706 | SRCC = 0.9698 ] +INFO:root:Average loss=0.06444566076993942 +INFO:root: +[Valid][10][UTT][ MSE = 0.1676 | LCC = 0.9694 | SRCC = 0.9693 ] +INFO:root:Average loss=0.06917715311050415 +INFO:root: +[Valid][11][UTT][ MSE = 0.1344 | LCC = 0.9700 | SRCC = 0.9702 ] +INFO:root:Average loss=0.06233011856675148 +INFO:root: +[Valid][12][UTT][ MSE = 0.1748 | LCC = 0.9679 | SRCC = 0.9682 ] +INFO:root:Average loss=0.05996262753009796 +INFO:root: +[Valid][13][UTT][ MSE = 0.1480 | LCC = 0.9708 | SRCC = 0.9702 ] +INFO:root:Average loss=0.057983679831027986 +INFO:root: +[Valid][14][UTT][ MSE = 0.1504 | LCC = 0.9716 | SRCC = 0.9708 ] +INFO:root:Average loss=0.055208091855049136 +INFO:root: +[Valid][15][UTT][ MSE = 0.1333 | LCC = 0.9714 | SRCC = 0.9707 ] +INFO:root:Average loss=0.05088571935892105 +INFO:root: +[Valid][16][UTT][ MSE = 0.1524 | LCC = 0.9721 | SRCC = 0.9710 ] +INFO:root:Average loss=0.05651950877904892 +INFO:root: +[Valid][17][UTT][ MSE = 0.1502 | LCC = 0.9720 | SRCC = 0.9708 ] +INFO:root:Average loss=0.04638941799104214 +INFO:root: +[Valid][18][UTT][ MSE = 0.1421 | LCC = 0.9726 | SRCC = 0.9715 ] +INFO:root:Average loss=0.04475841237604618 +INFO:root: +[Valid][19][UTT][ MSE = 0.1455 | LCC = 0.9724 | SRCC = 0.9710 ] +INFO:root:Average loss=0.04292649786174298 +INFO:root: +[Valid][20][UTT][ MSE = 0.1444 | LCC = 0.9722 | SRCC = 0.9710 ] +INFO:root:Average loss=0.04401890110969543 +INFO:root: +[Valid][21][UTT][ MSE = 0.1336 | LCC = 0.9719 | SRCC = 0.9710 ] +INFO:root:Average loss=0.04021459600329399 +INFO:root: +[Valid][22][UTT][ MSE = 0.1359 | LCC = 0.9728 | SRCC = 0.9714 ] +INFO:root:Average loss=0.039252900242805484 +INFO:root: +[Valid][23][UTT][ MSE = 0.1507 | LCC = 0.9718 | SRCC = 0.9707 ] +INFO:root:Average loss=0.038331442400813105 +INFO:root: +[Valid][24][UTT][ MSE = 0.1377 | LCC = 0.9716 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03513853980600834 +INFO:root: +[Valid][25][UTT][ MSE = 0.1709 | LCC = 0.9720 | SRCC = 0.9711 ] +INFO:root:Average loss=0.03683273349702358 +INFO:root: +[Valid][26][UTT][ MSE = 0.1436 | LCC = 0.9704 | SRCC = 0.9690 ] +INFO:root:Average loss=0.03324981920421124 +INFO:root: +[Valid][27][UTT][ MSE = 0.1430 | LCC = 0.9717 | SRCC = 0.9707 ] +INFO:root:Average loss=0.031180297672748564 +INFO:root: +[Valid][28][UTT][ MSE = 0.1548 | LCC = 0.9724 | SRCC = 0.9714 ] +INFO:root:Average loss=0.030202660664916038 +INFO:root: +[Valid][29][UTT][ MSE = 0.1332 | LCC = 0.9706 | SRCC = 0.9705 ] +INFO:root:Average loss=0.031098549142479896 +INFO:root: +[Valid][30][UTT][ MSE = 0.1644 | LCC = 0.9713 | SRCC = 0.9698 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1348 | LCC = 0.9721 | SRCC = 0.9716 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run3/config.gin b/results_tencent/w2v2_xlsr_1b/layer24/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run3/model.pt b/results_tencent/w2v2_xlsr_1b/layer24/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..91df12e59a291662cf3f71074c60b40974c50b84 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:37417e26fbe2717eebbc910eabf9628220201269ebdd7bacda8328e70b7cbd4f +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run3/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer24/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b4134edc39eef2a825dcfc09a29669d4bc5d5d17 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1284b4c46719a439936dc81b7b424660e7328bd962e69840de02955770729d42 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run3/train.log b/results_tencent/w2v2_xlsr_1b/layer24/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..16ab6c6cc2da348cf0abf12a856fa53bed4d9f3d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4324174091815948 +INFO:root: +[Valid][0][UTT][ MSE = 0.2199 | LCC = 0.9446 | SRCC = 0.9431 ] +INFO:root:Average loss=0.15777827268838882 +INFO:root: +[Valid][1][UTT][ MSE = 0.1860 | LCC = 0.9545 | SRCC = 0.9556 ] +INFO:root:Average loss=0.12931651842594147 +INFO:root: +[Valid][2][UTT][ MSE = 0.2053 | LCC = 0.9600 | SRCC = 0.9611 ] +INFO:root:Average loss=0.11765220320224762 +INFO:root: +[Valid][3][UTT][ MSE = 0.1970 | LCC = 0.9633 | SRCC = 0.9628 ] +INFO:root:Average loss=0.09896108409762383 +INFO:root: +[Valid][4][UTT][ MSE = 0.1552 | LCC = 0.9653 | SRCC = 0.9660 ] +INFO:root:Average loss=0.09338220176100731 +INFO:root: +[Valid][5][UTT][ MSE = 0.1739 | LCC = 0.9669 | SRCC = 0.9658 ] +INFO:root:Average loss=0.08650528478622437 +INFO:root: +[Valid][6][UTT][ MSE = 0.1509 | LCC = 0.9663 | SRCC = 0.9672 ] +INFO:root:Average loss=0.07917836761474609 +INFO:root: +[Valid][7][UTT][ MSE = 0.1696 | LCC = 0.9693 | SRCC = 0.9679 ] +INFO:root:Average loss=0.07776518705487251 +INFO:root: +[Valid][8][UTT][ MSE = 0.1494 | LCC = 0.9693 | SRCC = 0.9692 ] +INFO:root:Average loss=0.06925571119785309 +INFO:root: +[Valid][9][UTT][ MSE = 0.1486 | LCC = 0.9684 | SRCC = 0.9691 ] +INFO:root:Average loss=0.07142697562277317 +INFO:root: +[Valid][10][UTT][ MSE = 0.2001 | LCC = 0.9703 | SRCC = 0.9694 ] +INFO:root:Average loss=0.06574829778075218 +INFO:root: +[Valid][11][UTT][ MSE = 0.1761 | LCC = 0.9709 | SRCC = 0.9700 ] +INFO:root:Average loss=0.06408440175652504 +INFO:root: +[Valid][12][UTT][ MSE = 0.1380 | LCC = 0.9715 | SRCC = 0.9702 ] +INFO:root:Average loss=0.05707915702462196 +INFO:root: +[Valid][13][UTT][ MSE = 0.1509 | LCC = 0.9715 | SRCC = 0.9705 ] +INFO:root:Average loss=0.05898865503072739 +INFO:root: +[Valid][14][UTT][ MSE = 0.1422 | LCC = 0.9715 | SRCC = 0.9709 ] +INFO:root:Average loss=0.05813922542333603 +INFO:root: +[Valid][15][UTT][ MSE = 0.1841 | LCC = 0.9696 | SRCC = 0.9702 ] +INFO:root:Average loss=0.05234932953119278 +INFO:root: +[Valid][16][UTT][ MSE = 0.1516 | LCC = 0.9711 | SRCC = 0.9704 ] +INFO:root:Average loss=0.05170353329181671 +INFO:root: +[Valid][17][UTT][ MSE = 0.1497 | LCC = 0.9711 | SRCC = 0.9705 ] +INFO:root:Average loss=0.04971189989149571 +INFO:root: +[Valid][18][UTT][ MSE = 0.1564 | LCC = 0.9707 | SRCC = 0.9699 ] +INFO:root:Average loss=0.04859216845035553 +INFO:root: +[Valid][19][UTT][ MSE = 0.1516 | LCC = 0.9716 | SRCC = 0.9706 ] +INFO:root:Average loss=0.045402992069721225 +INFO:root: +[Valid][20][UTT][ MSE = 0.1666 | LCC = 0.9716 | SRCC = 0.9709 ] +INFO:root:Average loss=0.040977038517594336 +INFO:root: +[Valid][21][UTT][ MSE = 0.1604 | LCC = 0.9713 | SRCC = 0.9701 ] +INFO:root:Average loss=0.04230218704044819 +INFO:root: +[Valid][22][UTT][ MSE = 0.1402 | LCC = 0.9719 | SRCC = 0.9713 ] +INFO:root:Average loss=0.04046142888069153 +INFO:root: +[Valid][23][UTT][ MSE = 0.1449 | LCC = 0.9701 | SRCC = 0.9699 ] +INFO:root:Average loss=0.03986236903071404 +INFO:root: +[Valid][24][UTT][ MSE = 0.1562 | LCC = 0.9721 | SRCC = 0.9710 ] +INFO:root:Average loss=0.03740566043555737 +INFO:root: +[Valid][25][UTT][ MSE = 0.1263 | LCC = 0.9707 | SRCC = 0.9702 ] +INFO:root:Average loss=0.03732455025613308 +INFO:root: +[Valid][26][UTT][ MSE = 0.1341 | LCC = 0.9718 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03437468795478344 +INFO:root: +[Valid][27][UTT][ MSE = 0.1434 | LCC = 0.9711 | SRCC = 0.9701 ] +INFO:root:Average loss=0.031646317139267924 +INFO:root: +[Valid][28][UTT][ MSE = 0.1390 | LCC = 0.9709 | SRCC = 0.9703 ] +INFO:root:Average loss=0.03260174319148064 +INFO:root: +[Valid][29][UTT][ MSE = 0.1705 | LCC = 0.9696 | SRCC = 0.9704 ] +INFO:root:Average loss=0.03056065033376217 +INFO:root: +[Valid][30][UTT][ MSE = 0.1527 | LCC = 0.9701 | SRCC = 0.9694 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1532 | LCC = 0.9725 | SRCC = 0.9720 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run4/config.gin b/results_tencent/w2v2_xlsr_1b/layer24/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run4/model.pt b/results_tencent/w2v2_xlsr_1b/layer24/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..f398c94c49890af3bfa860c91d39524e5a59c2c2 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be5cb391ec118666905d9eb94b2792b22d9a2cfc65614cfd4cb0d46b08f53b70 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run4/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer24/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..a26ca6d392a0119e112a97c16956ef1eb54db508 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4ab7e96232cf232d11c7fa247375fefaccf8bc7394ded6b9ed062036ffe793bb +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer24/run4/train.log b/results_tencent/w2v2_xlsr_1b/layer24/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..73699f666f42f6ac92447c44458f6574fe36997f --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer24/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42346013724803927 +INFO:root: +[Valid][0][UTT][ MSE = 0.2861 | LCC = 0.9445 | SRCC = 0.9423 ] +INFO:root:Average loss=0.14959777122735976 +INFO:root: +[Valid][1][UTT][ MSE = 0.2427 | LCC = 0.9551 | SRCC = 0.9552 ] +INFO:root:Average loss=0.12268841832876205 +INFO:root: +[Valid][2][UTT][ MSE = 0.2265 | LCC = 0.9605 | SRCC = 0.9592 ] +INFO:root:Average loss=0.1114355681836605 +INFO:root: +[Valid][3][UTT][ MSE = 0.2278 | LCC = 0.9620 | SRCC = 0.9622 ] +INFO:root:Average loss=0.10870931017398834 +INFO:root: +[Valid][4][UTT][ MSE = 0.1810 | LCC = 0.9645 | SRCC = 0.9644 ] +INFO:root:Average loss=0.09473031738400459 +INFO:root: +[Valid][5][UTT][ MSE = 0.2297 | LCC = 0.9660 | SRCC = 0.9660 ] +INFO:root:Average loss=0.0887687548995018 +INFO:root: +[Valid][6][UTT][ MSE = 0.2148 | LCC = 0.9656 | SRCC = 0.9652 ] +INFO:root:Average loss=0.0815203407406807 +INFO:root: +[Valid][7][UTT][ MSE = 0.1703 | LCC = 0.9668 | SRCC = 0.9669 ] +INFO:root:Average loss=0.07641404062509537 +INFO:root: +[Valid][8][UTT][ MSE = 0.1834 | LCC = 0.9688 | SRCC = 0.9679 ] +INFO:root:Average loss=0.06861145681142806 +INFO:root: +[Valid][9][UTT][ MSE = 0.1998 | LCC = 0.9680 | SRCC = 0.9671 ] +INFO:root:Average loss=0.06876601558923721 +INFO:root: +[Valid][10][UTT][ MSE = 0.2027 | LCC = 0.9681 | SRCC = 0.9673 ] +INFO:root:Average loss=0.061939425244927404 +INFO:root: +[Valid][11][UTT][ MSE = 0.1813 | LCC = 0.9689 | SRCC = 0.9680 ] +INFO:root:Average loss=0.062322102725505826 +INFO:root: +[Valid][12][UTT][ MSE = 0.2044 | LCC = 0.9686 | SRCC = 0.9683 ] +INFO:root:Average loss=0.058351669192314146 +INFO:root: +[Valid][13][UTT][ MSE = 0.1597 | LCC = 0.9684 | SRCC = 0.9677 ] +INFO:root:Average loss=0.05916170585155487 +INFO:root: +[Valid][14][UTT][ MSE = 0.2251 | LCC = 0.9679 | SRCC = 0.9679 ] +INFO:root:Average loss=0.05456084716320038 +INFO:root: +[Valid][15][UTT][ MSE = 0.2038 | LCC = 0.9690 | SRCC = 0.9688 ] +INFO:root:Average loss=0.04883653122186661 +INFO:root: +[Valid][16][UTT][ MSE = 0.1916 | LCC = 0.9697 | SRCC = 0.9687 ] +INFO:root:Average loss=0.049614609509706496 +INFO:root: +[Valid][17][UTT][ MSE = 0.2340 | LCC = 0.9688 | SRCC = 0.9688 ] +INFO:root:Average loss=0.04641906286776066 +INFO:root: +[Valid][18][UTT][ MSE = 0.1948 | LCC = 0.9702 | SRCC = 0.9696 ] +INFO:root:Average loss=0.04693734969198704 +INFO:root: +[Valid][19][UTT][ MSE = 0.1628 | LCC = 0.9692 | SRCC = 0.9692 ] +INFO:root:Average loss=0.04456608389317989 +INFO:root: +[Valid][20][UTT][ MSE = 0.1731 | LCC = 0.9707 | SRCC = 0.9698 ] +INFO:root:Average loss=0.041192215234041214 +INFO:root: +[Valid][21][UTT][ MSE = 0.2133 | LCC = 0.9686 | SRCC = 0.9688 ] +INFO:root:Average loss=0.040590289294719695 +INFO:root: +[Valid][22][UTT][ MSE = 0.1676 | LCC = 0.9705 | SRCC = 0.9691 ] +INFO:root:Average loss=0.03612886236608028 +INFO:root: +[Valid][23][UTT][ MSE = 0.1950 | LCC = 0.9697 | SRCC = 0.9686 ] +INFO:root:Average loss=0.03554595571756363 +INFO:root: +[Valid][24][UTT][ MSE = 0.1580 | LCC = 0.9706 | SRCC = 0.9697 ] +INFO:root:Average loss=0.034882064759731296 +INFO:root: +[Valid][25][UTT][ MSE = 0.1643 | LCC = 0.9709 | SRCC = 0.9693 ] +INFO:root:Average loss=0.034705131992697716 +INFO:root: +[Valid][26][UTT][ MSE = 0.1482 | LCC = 0.9691 | SRCC = 0.9682 ] +INFO:root:Average loss=0.032070704132318495 +INFO:root: +[Valid][27][UTT][ MSE = 0.1459 | LCC = 0.9697 | SRCC = 0.9689 ] +INFO:root:Average loss=0.030858234107494355 +INFO:root: +[Valid][28][UTT][ MSE = 0.1699 | LCC = 0.9701 | SRCC = 0.9687 ] +INFO:root:Average loss=0.029556917548179625 +INFO:root: +[Valid][29][UTT][ MSE = 0.1717 | LCC = 0.9702 | SRCC = 0.9686 ] +INFO:root:Average loss=0.03227169591188431 +INFO:root: +[Valid][30][UTT][ MSE = 0.1939 | LCC = 0.9704 | SRCC = 0.9693 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1626 | LCC = 0.9720 | SRCC = 0.9720 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run0/config.gin b/results_tencent/w2v2_xlsr_1b/layer26/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run0/model.pt b/results_tencent/w2v2_xlsr_1b/layer26/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..b2f9e4cc021b02561b12cbcf1020fe5e8568962f --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2643d57174467d6e128266a2f11a94f78f4e318f779d601d4354a7d8711a2f4d +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run0/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer26/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..511fd858c934bde6728519ef6e7b10f7fb6af8e8 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7c1c70dd56fed3bd80664c9774f4d1c13b624e4b346aab39f3c4035c4afa899f +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run0/train.log b/results_tencent/w2v2_xlsr_1b/layer26/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c32e4acbfab1a60a60085d0c858ba9ef3db05c48 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4364948003292084 +INFO:root: +[Valid][0][UTT][ MSE = 0.2139 | LCC = 0.9441 | SRCC = 0.9414 ] +INFO:root:Average loss=0.16022294163703918 +INFO:root: +[Valid][1][UTT][ MSE = 0.2142 | LCC = 0.9556 | SRCC = 0.9534 ] +INFO:root:Average loss=0.1306026474237442 +INFO:root: +[Valid][2][UTT][ MSE = 0.1801 | LCC = 0.9605 | SRCC = 0.9600 ] +INFO:root:Average loss=0.10845386213064194 +INFO:root: +[Valid][3][UTT][ MSE = 0.2028 | LCC = 0.9651 | SRCC = 0.9639 ] +INFO:root:Average loss=0.10341848236322403 +INFO:root: +[Valid][4][UTT][ MSE = 0.1628 | LCC = 0.9667 | SRCC = 0.9663 ] +INFO:root:Average loss=0.09358523204922677 +INFO:root: +[Valid][5][UTT][ MSE = 0.2085 | LCC = 0.9676 | SRCC = 0.9670 ] +INFO:root:Average loss=0.08909978291392326 +INFO:root: +[Valid][6][UTT][ MSE = 0.1715 | LCC = 0.9679 | SRCC = 0.9667 ] +INFO:root:Average loss=0.07806793445348739 +INFO:root: +[Valid][7][UTT][ MSE = 0.1921 | LCC = 0.9696 | SRCC = 0.9684 ] +INFO:root:Average loss=0.07660788089036942 +INFO:root: +[Valid][8][UTT][ MSE = 0.1954 | LCC = 0.9705 | SRCC = 0.9699 ] +INFO:root:Average loss=0.07308598792552948 +INFO:root: +[Valid][9][UTT][ MSE = 0.1753 | LCC = 0.9712 | SRCC = 0.9703 ] +INFO:root:Average loss=0.07317063257098198 +INFO:root: +[Valid][10][UTT][ MSE = 0.1753 | LCC = 0.9708 | SRCC = 0.9694 ] +INFO:root:Average loss=0.06654537388682366 +INFO:root: +[Valid][11][UTT][ MSE = 0.1896 | LCC = 0.9714 | SRCC = 0.9706 ] +INFO:root:Average loss=0.062310593903064725 +INFO:root: +[Valid][12][UTT][ MSE = 0.1693 | LCC = 0.9712 | SRCC = 0.9698 ] +INFO:root:Average loss=0.05959021735191345 +INFO:root: +[Valid][13][UTT][ MSE = 0.1651 | LCC = 0.9716 | SRCC = 0.9704 ] +INFO:root:Average loss=0.05659789362549782 +INFO:root: +[Valid][14][UTT][ MSE = 0.1449 | LCC = 0.9717 | SRCC = 0.9704 ] +INFO:root:Average loss=0.05436706292629242 +INFO:root: +[Valid][15][UTT][ MSE = 0.1696 | LCC = 0.9715 | SRCC = 0.9706 ] +INFO:root:Average loss=0.0514833933711052 +INFO:root: +[Valid][16][UTT][ MSE = 0.1443 | LCC = 0.9714 | SRCC = 0.9706 ] +INFO:root:Average loss=0.04909857796132565 +INFO:root: +[Valid][17][UTT][ MSE = 0.1326 | LCC = 0.9721 | SRCC = 0.9710 ] +INFO:root:Average loss=0.04887544959783554 +INFO:root: +[Valid][18][UTT][ MSE = 0.1856 | LCC = 0.9713 | SRCC = 0.9701 ] +INFO:root:Average loss=0.04534586669504642 +INFO:root: +[Valid][19][UTT][ MSE = 0.1839 | LCC = 0.9718 | SRCC = 0.9712 ] +INFO:root:Average loss=0.042032768324017526 +INFO:root: +[Valid][20][UTT][ MSE = 0.1603 | LCC = 0.9714 | SRCC = 0.9712 ] +INFO:root:Average loss=0.043106306657195094 +INFO:root: +[Valid][21][UTT][ MSE = 0.1509 | LCC = 0.9725 | SRCC = 0.9715 ] +INFO:root:Average loss=0.03903661021590233 +INFO:root: +[Valid][22][UTT][ MSE = 0.1642 | LCC = 0.9708 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03960152427852154 +INFO:root: +[Valid][23][UTT][ MSE = 0.1900 | LCC = 0.9711 | SRCC = 0.9711 ] +INFO:root:Average loss=0.03954943858087063 +INFO:root: +[Valid][24][UTT][ MSE = 0.1780 | LCC = 0.9715 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03595333461463451 +INFO:root: +[Valid][25][UTT][ MSE = 0.1517 | LCC = 0.9715 | SRCC = 0.9710 ] +INFO:root:Average loss=0.0371765204668045 +INFO:root: +[Valid][26][UTT][ MSE = 0.1495 | LCC = 0.9723 | SRCC = 0.9711 ] +INFO:root:Average loss=0.03353607232868672 +INFO:root: +[Valid][27][UTT][ MSE = 0.1568 | LCC = 0.9716 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03355395695567131 +INFO:root: +[Valid][28][UTT][ MSE = 0.1303 | LCC = 0.9724 | SRCC = 0.9712 ] +INFO:root:Average loss=0.03216102574765682 +INFO:root: +[Valid][29][UTT][ MSE = 0.1459 | LCC = 0.9707 | SRCC = 0.9701 ] +INFO:root:Average loss=0.03019101783633232 +INFO:root: +[Valid][30][UTT][ MSE = 0.1555 | LCC = 0.9715 | SRCC = 0.9707 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1491 | LCC = 0.9725 | SRCC = 0.9722 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run1/config.gin b/results_tencent/w2v2_xlsr_1b/layer26/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run1/model.pt b/results_tencent/w2v2_xlsr_1b/layer26/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e75bc6aca91ee9fb8433c5d0d6db1513fb702dec --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ca913ea8e76789d72965036b3e55a901ddd51c9b87127a5c0db45fb3f9381cb8 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run1/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer26/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..ba3eb602d73917c3b0c76cd03c1f0e4e7e8e25a5 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aaea3b8a678c1e840c3b28b51b075ac6d764ebb05385ec7dd68bce2a01b51245 +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run1/train.log b/results_tencent/w2v2_xlsr_1b/layer26/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..6e9296d3852d818ae912797a4f6d9ef3f88dd766 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.39573390579223633 +INFO:root: +[Valid][0][UTT][ MSE = 0.2619 | LCC = 0.9405 | SRCC = 0.9369 ] +INFO:root:Average loss=0.16668039625883102 +INFO:root: +[Valid][1][UTT][ MSE = 0.1917 | LCC = 0.9555 | SRCC = 0.9554 ] +INFO:root:Average loss=0.12764196288585664 +INFO:root: +[Valid][2][UTT][ MSE = 0.2017 | LCC = 0.9609 | SRCC = 0.9606 ] +INFO:root:Average loss=0.11908436587452889 +INFO:root: +[Valid][3][UTT][ MSE = 0.2019 | LCC = 0.9635 | SRCC = 0.9634 ] +INFO:root:Average loss=0.10420277708768845 +INFO:root: +[Valid][4][UTT][ MSE = 0.2051 | LCC = 0.9641 | SRCC = 0.9652 ] +INFO:root:Average loss=0.09715469342470169 +INFO:root: +[Valid][5][UTT][ MSE = 0.2130 | LCC = 0.9660 | SRCC = 0.9659 ] +INFO:root:Average loss=0.09032490712404251 +INFO:root: +[Valid][6][UTT][ MSE = 0.1895 | LCC = 0.9672 | SRCC = 0.9676 ] +INFO:root:Average loss=0.08322419190406799 +INFO:root: +[Valid][7][UTT][ MSE = 0.1749 | LCC = 0.9693 | SRCC = 0.9690 ] +INFO:root:Average loss=0.07866777217388153 +INFO:root: +[Valid][8][UTT][ MSE = 0.1711 | LCC = 0.9692 | SRCC = 0.9691 ] +INFO:root:Average loss=0.07107583183050156 +INFO:root: +[Valid][9][UTT][ MSE = 0.1747 | LCC = 0.9700 | SRCC = 0.9691 ] +INFO:root:Average loss=0.07112545126676559 +INFO:root: +[Valid][10][UTT][ MSE = 0.1818 | LCC = 0.9671 | SRCC = 0.9682 ] +INFO:root:Average loss=0.0652295007109642 +INFO:root: +[Valid][11][UTT][ MSE = 0.1543 | LCC = 0.9697 | SRCC = 0.9700 ] +INFO:root:Average loss=0.06358333164453507 +INFO:root: +[Valid][12][UTT][ MSE = 0.1845 | LCC = 0.9706 | SRCC = 0.9704 ] +INFO:root:Average loss=0.06004732005298138 +INFO:root: +[Valid][13][UTT][ MSE = 0.1724 | LCC = 0.9708 | SRCC = 0.9704 ] +INFO:root:Average loss=0.059352685809135435 +INFO:root: +[Valid][14][UTT][ MSE = 0.1713 | LCC = 0.9712 | SRCC = 0.9704 ] +INFO:root:Average loss=0.054678775221109394 +INFO:root: +[Valid][15][UTT][ MSE = 0.1755 | LCC = 0.9712 | SRCC = 0.9708 ] +INFO:root:Average loss=0.05476911896467209 +INFO:root: +[Valid][16][UTT][ MSE = 0.1671 | LCC = 0.9710 | SRCC = 0.9710 ] +INFO:root:Average loss=0.05339259181916714 +INFO:root: +[Valid][17][UTT][ MSE = 0.1751 | LCC = 0.9687 | SRCC = 0.9696 ] +INFO:root:Average loss=0.04829356625676155 +INFO:root: +[Valid][18][UTT][ MSE = 0.1645 | LCC = 0.9703 | SRCC = 0.9705 ] +INFO:root:Average loss=0.04647945080697537 +INFO:root: +[Valid][19][UTT][ MSE = 0.1685 | LCC = 0.9705 | SRCC = 0.9699 ] +INFO:root:Average loss=0.04515993766486645 +INFO:root: +[Valid][20][UTT][ MSE = 0.1791 | LCC = 0.9718 | SRCC = 0.9709 ] +INFO:root:Average loss=0.04191035099327564 +INFO:root: +[Valid][21][UTT][ MSE = 0.1714 | LCC = 0.9713 | SRCC = 0.9708 ] +INFO:root:Average loss=0.04080983065068722 +INFO:root: +[Valid][22][UTT][ MSE = 0.1657 | LCC = 0.9717 | SRCC = 0.9710 ] +INFO:root:Average loss=0.03821597498655319 +INFO:root: +[Valid][23][UTT][ MSE = 0.1513 | LCC = 0.9719 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03729225125908851 +INFO:root: +[Valid][24][UTT][ MSE = 0.1380 | LCC = 0.9720 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03651246558129787 +INFO:root: +[Valid][25][UTT][ MSE = 0.1570 | LCC = 0.9712 | SRCC = 0.9705 ] +INFO:root:Average loss=0.034465343251824376 +INFO:root: +[Valid][26][UTT][ MSE = 0.1555 | LCC = 0.9715 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03538453520834446 +INFO:root: +[Valid][27][UTT][ MSE = 0.1612 | LCC = 0.9700 | SRCC = 0.9700 ] +INFO:root:Average loss=0.03217383563518524 +INFO:root: +[Valid][28][UTT][ MSE = 0.1515 | LCC = 0.9706 | SRCC = 0.9700 ] +INFO:root:Average loss=0.03126202166825533 +INFO:root: +[Valid][29][UTT][ MSE = 0.1448 | LCC = 0.9716 | SRCC = 0.9711 ] +INFO:root:Average loss=0.030788475707173347 +INFO:root: +[Valid][30][UTT][ MSE = 0.1583 | LCC = 0.9711 | SRCC = 0.9706 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1363 | LCC = 0.9712 | SRCC = 0.9711 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run2/config.gin b/results_tencent/w2v2_xlsr_1b/layer26/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run2/model.pt b/results_tencent/w2v2_xlsr_1b/layer26/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..50a1fa0a341db680fde6e2f2f9b690a7b245b85a --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:da7e931d57aa6d2621a020f9d65a77cfeef75a5215575a42cab7e04485d9101a +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run2/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer26/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..36bc659847886db7216dc1c6a291cf8d59f375d6 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:458ceaebf23b291649423c17752e20b1c7ccc5fa0e6bdfec9c25d642c2835af7 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run2/train.log b/results_tencent/w2v2_xlsr_1b/layer26/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e5911d1dd04fdf7dc1fb952c39e040db6b3271cd --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.40294132763147356 +INFO:root: +[Valid][0][UTT][ MSE = 0.2184 | LCC = 0.9436 | SRCC = 0.9405 ] +INFO:root:Average loss=0.15953411918878554 +INFO:root: +[Valid][1][UTT][ MSE = 0.2066 | LCC = 0.9551 | SRCC = 0.9573 ] +INFO:root:Average loss=0.1243051775097847 +INFO:root: +[Valid][2][UTT][ MSE = 0.1778 | LCC = 0.9612 | SRCC = 0.9613 ] +INFO:root:Average loss=0.10701795947551727 +INFO:root: +[Valid][3][UTT][ MSE = 0.1897 | LCC = 0.9664 | SRCC = 0.9664 ] +INFO:root:Average loss=0.09953897759318352 +INFO:root: +[Valid][4][UTT][ MSE = 0.1474 | LCC = 0.9664 | SRCC = 0.9662 ] +INFO:root:Average loss=0.09026239496469497 +INFO:root: +[Valid][5][UTT][ MSE = 0.1665 | LCC = 0.9676 | SRCC = 0.9680 ] +INFO:root:Average loss=0.08471592870354652 +INFO:root: +[Valid][6][UTT][ MSE = 0.1526 | LCC = 0.9698 | SRCC = 0.9694 ] +INFO:root:Average loss=0.07919338971376419 +INFO:root: +[Valid][7][UTT][ MSE = 0.1473 | LCC = 0.9695 | SRCC = 0.9693 ] +INFO:root:Average loss=0.07303610071539879 +INFO:root: +[Valid][8][UTT][ MSE = 0.1571 | LCC = 0.9704 | SRCC = 0.9701 ] +INFO:root:Average loss=0.07182381215691566 +INFO:root: +[Valid][9][UTT][ MSE = 0.1583 | LCC = 0.9701 | SRCC = 0.9691 ] +INFO:root:Average loss=0.0652364310324192 +INFO:root: +[Valid][10][UTT][ MSE = 0.1729 | LCC = 0.9719 | SRCC = 0.9709 ] +INFO:root:Average loss=0.061526652306318284 +INFO:root: +[Valid][11][UTT][ MSE = 0.1909 | LCC = 0.9710 | SRCC = 0.9704 ] +INFO:root:Average loss=0.05973244948685169 +INFO:root: +[Valid][12][UTT][ MSE = 0.1567 | LCC = 0.9712 | SRCC = 0.9709 ] +INFO:root:Average loss=0.05842477807402611 +INFO:root: +[Valid][13][UTT][ MSE = 0.1533 | LCC = 0.9723 | SRCC = 0.9715 ] +INFO:root:Average loss=0.054775118485093116 +INFO:root: +[Valid][14][UTT][ MSE = 0.1840 | LCC = 0.9722 | SRCC = 0.9717 ] +INFO:root:Average loss=0.0544399031996727 +INFO:root: +[Valid][15][UTT][ MSE = 0.1440 | LCC = 0.9713 | SRCC = 0.9709 ] +INFO:root:Average loss=0.050194256097078324 +INFO:root: +[Valid][16][UTT][ MSE = 0.1315 | LCC = 0.9716 | SRCC = 0.9712 ] +INFO:root:Average loss=0.04909486398100853 +INFO:root: +[Valid][17][UTT][ MSE = 0.1536 | LCC = 0.9725 | SRCC = 0.9717 ] +INFO:root:Average loss=0.0428465089648962 +INFO:root: +[Valid][18][UTT][ MSE = 0.1454 | LCC = 0.9714 | SRCC = 0.9712 ] +INFO:root:Average loss=0.0436489817649126 +INFO:root: +[Valid][19][UTT][ MSE = 0.1403 | LCC = 0.9728 | SRCC = 0.9721 ] +INFO:root:Average loss=0.0418948754966259 +INFO:root: +[Valid][20][UTT][ MSE = 0.1632 | LCC = 0.9730 | SRCC = 0.9722 ] +INFO:root:Average loss=0.040312297120690344 +INFO:root: +[Valid][21][UTT][ MSE = 0.1565 | LCC = 0.9723 | SRCC = 0.9716 ] +INFO:root:Average loss=0.038594195812940595 +INFO:root: +[Valid][22][UTT][ MSE = 0.1575 | LCC = 0.9726 | SRCC = 0.9717 ] +INFO:root:Average loss=0.037484331235289575 +INFO:root: +[Valid][23][UTT][ MSE = 0.1600 | LCC = 0.9724 | SRCC = 0.9716 ] +INFO:root:Average loss=0.03653448586165905 +INFO:root: +[Valid][24][UTT][ MSE = 0.1538 | LCC = 0.9727 | SRCC = 0.9722 ] +INFO:root:Average loss=0.03403429640829563 +INFO:root: +[Valid][25][UTT][ MSE = 0.1423 | LCC = 0.9723 | SRCC = 0.9712 ] +INFO:root:Average loss=0.03272468443214893 +INFO:root: +[Valid][26][UTT][ MSE = 0.1625 | LCC = 0.9731 | SRCC = 0.9722 ] +INFO:root:Average loss=0.031870307981967924 +INFO:root: +[Valid][27][UTT][ MSE = 0.1449 | LCC = 0.9726 | SRCC = 0.9718 ] +INFO:root:Average loss=0.030341023474931717 +INFO:root: +[Valid][28][UTT][ MSE = 0.1795 | LCC = 0.9732 | SRCC = 0.9722 ] +INFO:root:Average loss=0.030901411369442938 +INFO:root: +[Valid][29][UTT][ MSE = 0.1691 | LCC = 0.9727 | SRCC = 0.9725 ] +INFO:root:Average loss=0.028855195820331574 +INFO:root: +[Valid][30][UTT][ MSE = 0.1537 | LCC = 0.9722 | SRCC = 0.9717 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1791 | LCC = 0.9736 | SRCC = 0.9733 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run3/config.gin b/results_tencent/w2v2_xlsr_1b/layer26/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run3/model.pt b/results_tencent/w2v2_xlsr_1b/layer26/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..687628f6e47af2c71141d00774a26d284db38589 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:46be794b34173cb5e53d4d0c9eaacff73992a6f0809bc0ff30c37fb3455dbaef +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run3/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer26/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..660ed4d6bffc07f38ac8557ba3b48de3ae9e3a8e --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:17684b2203adba37cf096a552dc8b0171527125839299990384b96208db46832 +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run3/train.log b/results_tencent/w2v2_xlsr_1b/layer26/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f4840ef73edbcaaaa16db08fe07928b7106af7da --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4269409849643707 +INFO:root: +[Valid][0][UTT][ MSE = 0.2443 | LCC = 0.9471 | SRCC = 0.9457 ] +INFO:root:Average loss=0.15509869879484176 +INFO:root: +[Valid][1][UTT][ MSE = 0.2121 | LCC = 0.9585 | SRCC = 0.9580 ] +INFO:root:Average loss=0.1328282403945923 +INFO:root: +[Valid][2][UTT][ MSE = 0.1991 | LCC = 0.9607 | SRCC = 0.9606 ] +INFO:root:Average loss=0.11296733367443085 +INFO:root: +[Valid][3][UTT][ MSE = 0.2199 | LCC = 0.9639 | SRCC = 0.9636 ] +INFO:root:Average loss=0.10114970624446869 +INFO:root: +[Valid][4][UTT][ MSE = 0.2175 | LCC = 0.9664 | SRCC = 0.9663 ] +INFO:root:Average loss=0.09622769269347191 +INFO:root: +[Valid][5][UTT][ MSE = 0.2143 | LCC = 0.9680 | SRCC = 0.9674 ] +INFO:root:Average loss=0.09300159609317779 +INFO:root: +[Valid][6][UTT][ MSE = 0.2096 | LCC = 0.9679 | SRCC = 0.9680 ] +INFO:root:Average loss=0.08500400871038437 +INFO:root: +[Valid][7][UTT][ MSE = 0.1633 | LCC = 0.9691 | SRCC = 0.9685 ] +INFO:root:Average loss=0.07703800716996193 +INFO:root: +[Valid][8][UTT][ MSE = 0.1694 | LCC = 0.9681 | SRCC = 0.9681 ] +INFO:root:Average loss=0.07212945690751076 +INFO:root: +[Valid][9][UTT][ MSE = 0.1828 | LCC = 0.9698 | SRCC = 0.9694 ] +INFO:root:Average loss=0.07022044330835342 +INFO:root: +[Valid][10][UTT][ MSE = 0.2093 | LCC = 0.9692 | SRCC = 0.9694 ] +INFO:root:Average loss=0.06687699064612389 +INFO:root: +[Valid][11][UTT][ MSE = 0.1644 | LCC = 0.9699 | SRCC = 0.9697 ] +INFO:root:Average loss=0.06458845672011375 +INFO:root: +[Valid][12][UTT][ MSE = 0.1492 | LCC = 0.9714 | SRCC = 0.9709 ] +INFO:root:Average loss=0.05912389695644379 +INFO:root: +[Valid][13][UTT][ MSE = 0.1920 | LCC = 0.9699 | SRCC = 0.9694 ] +INFO:root:Average loss=0.05649955940246582 +INFO:root: +[Valid][14][UTT][ MSE = 0.1684 | LCC = 0.9720 | SRCC = 0.9712 ] +INFO:root:Average loss=0.055477713465690616 +INFO:root: +[Valid][15][UTT][ MSE = 0.1785 | LCC = 0.9717 | SRCC = 0.9710 ] +INFO:root:Average loss=0.053242841601371764 +INFO:root: +[Valid][16][UTT][ MSE = 0.1853 | LCC = 0.9683 | SRCC = 0.9694 ] +INFO:root:Average loss=0.05080889710783958 +INFO:root: +[Valid][17][UTT][ MSE = 0.1686 | LCC = 0.9712 | SRCC = 0.9709 ] +INFO:root:Average loss=0.044614056825637814 +INFO:root: +[Valid][18][UTT][ MSE = 0.1867 | LCC = 0.9719 | SRCC = 0.9714 ] +INFO:root:Average loss=0.045416024580597875 +INFO:root: +[Valid][19][UTT][ MSE = 0.1626 | LCC = 0.9718 | SRCC = 0.9710 ] +INFO:root:Average loss=0.04301246413588524 +INFO:root: +[Valid][20][UTT][ MSE = 0.1491 | LCC = 0.9731 | SRCC = 0.9717 ] +INFO:root:Average loss=0.04244551372528076 +INFO:root: +[Valid][21][UTT][ MSE = 0.1492 | LCC = 0.9716 | SRCC = 0.9716 ] +INFO:root:Average loss=0.04088009771704674 +INFO:root: +[Valid][22][UTT][ MSE = 0.1793 | LCC = 0.9723 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03775584964454174 +INFO:root: +[Valid][23][UTT][ MSE = 0.1656 | LCC = 0.9725 | SRCC = 0.9716 ] +INFO:root:Average loss=0.03540808254480362 +INFO:root: +[Valid][24][UTT][ MSE = 0.1727 | LCC = 0.9724 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03454203879833222 +INFO:root: +[Valid][25][UTT][ MSE = 0.1675 | LCC = 0.9708 | SRCC = 0.9705 ] +INFO:root:Average loss=0.034668474301695826 +INFO:root: +[Valid][26][UTT][ MSE = 0.1900 | LCC = 0.9697 | SRCC = 0.9692 ] +INFO:root:Average loss=0.03470615521073341 +INFO:root: +[Valid][27][UTT][ MSE = 0.1781 | LCC = 0.9722 | SRCC = 0.9713 ] +INFO:root:Average loss=0.030428285032510757 +INFO:root: +[Valid][28][UTT][ MSE = 0.1502 | LCC = 0.9723 | SRCC = 0.9713 ] +INFO:root:Average loss=0.030526206374168394 +INFO:root: +[Valid][29][UTT][ MSE = 0.1621 | LCC = 0.9725 | SRCC = 0.9715 ] +INFO:root:Average loss=0.029108927860856056 +INFO:root: +[Valid][30][UTT][ MSE = 0.1638 | LCC = 0.9726 | SRCC = 0.9718 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1488 | LCC = 0.9724 | SRCC = 0.9718 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run4/config.gin b/results_tencent/w2v2_xlsr_1b/layer26/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run4/model.pt b/results_tencent/w2v2_xlsr_1b/layer26/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..57066174f11a2907501d5343fc7d27332f600a46 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7898f342531495d4a49934e8aeb58f4af31e7b52b16fdf53a167f5b1667670c5 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run4/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer26/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..89507b01c5b7e4e95a686fc2aeb9fd312005ba5f --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e3d629ffad282e7805db400b906f21fc9ed761b2730c009fba4ab497bfba4f46 +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer26/run4/train.log b/results_tencent/w2v2_xlsr_1b/layer26/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..2ea565b8bca69e6de8b09516b9c48fddb43d263f --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer26/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46443319916725156 +INFO:root: +[Valid][0][UTT][ MSE = 0.2709 | LCC = 0.9432 | SRCC = 0.9405 ] +INFO:root:Average loss=0.1527217093706131 +INFO:root: +[Valid][1][UTT][ MSE = 0.2337 | LCC = 0.9567 | SRCC = 0.9558 ] +INFO:root:Average loss=0.13648396271467209 +INFO:root: +[Valid][2][UTT][ MSE = 0.2219 | LCC = 0.9617 | SRCC = 0.9608 ] +INFO:root:Average loss=0.11298325616121292 +INFO:root: +[Valid][3][UTT][ MSE = 0.2122 | LCC = 0.9636 | SRCC = 0.9626 ] +INFO:root:Average loss=0.10518078488111496 +INFO:root: +[Valid][4][UTT][ MSE = 0.2629 | LCC = 0.9630 | SRCC = 0.9650 ] +INFO:root:Average loss=0.10009078595042228 +INFO:root: +[Valid][5][UTT][ MSE = 0.2075 | LCC = 0.9674 | SRCC = 0.9670 ] +INFO:root:Average loss=0.09136433959007263 +INFO:root: +[Valid][6][UTT][ MSE = 0.1998 | LCC = 0.9682 | SRCC = 0.9671 ] +INFO:root:Average loss=0.08767670184373856 +INFO:root: +[Valid][7][UTT][ MSE = 0.2140 | LCC = 0.9669 | SRCC = 0.9670 ] +INFO:root:Average loss=0.0821607993543148 +INFO:root: +[Valid][8][UTT][ MSE = 0.2126 | LCC = 0.9685 | SRCC = 0.9691 ] +INFO:root:Average loss=0.07900611898303032 +INFO:root: +[Valid][9][UTT][ MSE = 0.2159 | LCC = 0.9697 | SRCC = 0.9698 ] +INFO:root:Average loss=0.07056331011652947 +INFO:root: +[Valid][10][UTT][ MSE = 0.2364 | LCC = 0.9695 | SRCC = 0.9704 ] +INFO:root:Average loss=0.06918503826856613 +INFO:root: +[Valid][11][UTT][ MSE = 0.2145 | LCC = 0.9692 | SRCC = 0.9699 ] +INFO:root:Average loss=0.06521756330132485 +INFO:root: +[Valid][12][UTT][ MSE = 0.1681 | LCC = 0.9705 | SRCC = 0.9694 ] +INFO:root:Average loss=0.062516385987401 +INFO:root: +[Valid][13][UTT][ MSE = 0.1869 | LCC = 0.9710 | SRCC = 0.9707 ] +INFO:root:Average loss=0.06084989914298058 +INFO:root: +[Valid][14][UTT][ MSE = 0.1699 | LCC = 0.9716 | SRCC = 0.9707 ] +INFO:root:Average loss=0.058928846895694734 +INFO:root: +[Valid][15][UTT][ MSE = 0.1587 | LCC = 0.9714 | SRCC = 0.9710 ] +INFO:root:Average loss=0.053427678018808364 +INFO:root: +[Valid][16][UTT][ MSE = 0.2044 | LCC = 0.9712 | SRCC = 0.9709 ] +INFO:root:Average loss=0.05349933432042599 +INFO:root: +[Valid][17][UTT][ MSE = 0.1930 | LCC = 0.9703 | SRCC = 0.9703 ] +INFO:root:Average loss=0.04933525040745735 +INFO:root: +[Valid][18][UTT][ MSE = 0.1746 | LCC = 0.9723 | SRCC = 0.9711 ] +INFO:root:Average loss=0.045228615313768386 +INFO:root: +[Valid][19][UTT][ MSE = 0.1711 | LCC = 0.9724 | SRCC = 0.9712 ] +INFO:root:Average loss=0.04363231483101845 +INFO:root: +[Valid][20][UTT][ MSE = 0.2072 | LCC = 0.9705 | SRCC = 0.9708 ] +INFO:root:Average loss=0.043517185226082805 +INFO:root: +[Valid][21][UTT][ MSE = 0.1881 | LCC = 0.9717 | SRCC = 0.9705 ] +INFO:root:Average loss=0.04396527580916881 +INFO:root: +[Valid][22][UTT][ MSE = 0.1863 | LCC = 0.9719 | SRCC = 0.9711 ] +INFO:root:Average loss=0.039572258666157725 +INFO:root: +[Valid][23][UTT][ MSE = 0.1708 | LCC = 0.9720 | SRCC = 0.9711 ] +INFO:root:Average loss=0.03951666654646396 +INFO:root: +[Valid][24][UTT][ MSE = 0.2179 | LCC = 0.9707 | SRCC = 0.9711 ] +INFO:root:Average loss=0.0375284054428339 +INFO:root: +[Valid][25][UTT][ MSE = 0.1756 | LCC = 0.9723 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03425534307956696 +INFO:root: +[Valid][26][UTT][ MSE = 0.1946 | LCC = 0.9702 | SRCC = 0.9705 ] +INFO:root:Average loss=0.033699871465563774 +INFO:root: +[Valid][27][UTT][ MSE = 0.1749 | LCC = 0.9721 | SRCC = 0.9715 ] +INFO:root:Average loss=0.03194917154312134 +INFO:root: +[Valid][28][UTT][ MSE = 0.2008 | LCC = 0.9707 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03301992149651051 +INFO:root: +[Valid][29][UTT][ MSE = 0.1799 | LCC = 0.9717 | SRCC = 0.9707 ] +INFO:root:Average loss=0.031021794229745867 +INFO:root: +[Valid][30][UTT][ MSE = 0.1804 | LCC = 0.9714 | SRCC = 0.9708 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1723 | LCC = 0.9719 | SRCC = 0.9717 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run0/config.gin b/results_tencent/w2v2_xlsr_1b/layer28/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run0/model.pt b/results_tencent/w2v2_xlsr_1b/layer28/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..a15bb4960245ac2e92cc36789109c9c0ceeb443e --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:72e93b7210f4778f458be25bb93d5484a07fd9ddb0719f13ee7223f1bda209fb +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run0/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer28/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..6533fa940fd52d928465209da14e312e8e9c12fa --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b820ab09b89e93eb45122b1de76c68065acfb6dffba354ecaf7ae5a046891d19 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run0/train.log b/results_tencent/w2v2_xlsr_1b/layer28/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..80a547e3f42180eb1403e110b5ece15c56a24b84 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.40992436414957045 +INFO:root: +[Valid][0][UTT][ MSE = 0.2606 | LCC = 0.9459 | SRCC = 0.9443 ] +INFO:root:Average loss=0.1617715703845024 +INFO:root: +[Valid][1][UTT][ MSE = 0.2143 | LCC = 0.9571 | SRCC = 0.9576 ] +INFO:root:Average loss=0.13120559722185135 +INFO:root: +[Valid][2][UTT][ MSE = 0.1914 | LCC = 0.9615 | SRCC = 0.9611 ] +INFO:root:Average loss=0.12011741137504578 +INFO:root: +[Valid][3][UTT][ MSE = 0.2494 | LCC = 0.9629 | SRCC = 0.9624 ] +INFO:root:Average loss=0.10142858815193176 +INFO:root: +[Valid][4][UTT][ MSE = 0.2182 | LCC = 0.9663 | SRCC = 0.9654 ] +INFO:root:Average loss=0.10299552091956139 +INFO:root: +[Valid][5][UTT][ MSE = 0.1913 | LCC = 0.9672 | SRCC = 0.9665 ] +INFO:root:Average loss=0.08831448641419411 +INFO:root: +[Valid][6][UTT][ MSE = 0.1929 | LCC = 0.9688 | SRCC = 0.9683 ] +INFO:root:Average loss=0.08383779233694076 +INFO:root: +[Valid][7][UTT][ MSE = 0.1976 | LCC = 0.9687 | SRCC = 0.9682 ] +INFO:root:Average loss=0.07751929399371148 +INFO:root: +[Valid][8][UTT][ MSE = 0.2150 | LCC = 0.9690 | SRCC = 0.9696 ] +INFO:root:Average loss=0.07503452563285827 +INFO:root: +[Valid][9][UTT][ MSE = 0.2279 | LCC = 0.9697 | SRCC = 0.9690 ] +INFO:root:Average loss=0.07387803143262864 +INFO:root: +[Valid][10][UTT][ MSE = 0.2039 | LCC = 0.9698 | SRCC = 0.9688 ] +INFO:root:Average loss=0.06706907144188881 +INFO:root: +[Valid][11][UTT][ MSE = 0.2131 | LCC = 0.9705 | SRCC = 0.9704 ] +INFO:root:Average loss=0.0623259691298008 +INFO:root: +[Valid][12][UTT][ MSE = 0.1698 | LCC = 0.9714 | SRCC = 0.9707 ] +INFO:root:Average loss=0.058300529211759565 +INFO:root: +[Valid][13][UTT][ MSE = 0.1844 | LCC = 0.9717 | SRCC = 0.9709 ] +INFO:root:Average loss=0.05690045063197613 +INFO:root: +[Valid][14][UTT][ MSE = 0.1919 | LCC = 0.9716 | SRCC = 0.9709 ] +INFO:root:Average loss=0.05508420366048813 +INFO:root: +[Valid][15][UTT][ MSE = 0.1689 | LCC = 0.9713 | SRCC = 0.9701 ] +INFO:root:Average loss=0.05105244129896164 +INFO:root: +[Valid][16][UTT][ MSE = 0.1802 | LCC = 0.9714 | SRCC = 0.9705 ] +INFO:root:Average loss=0.05080498577654362 +INFO:root: +[Valid][17][UTT][ MSE = 0.1567 | LCC = 0.9711 | SRCC = 0.9701 ] +INFO:root:Average loss=0.04562297323346138 +INFO:root: +[Valid][18][UTT][ MSE = 0.1796 | LCC = 0.9707 | SRCC = 0.9706 ] +INFO:root:Average loss=0.048393755063414576 +INFO:root: +[Valid][19][UTT][ MSE = 0.1707 | LCC = 0.9689 | SRCC = 0.9688 ] +INFO:root:Average loss=0.042039211049675944 +INFO:root: +[Valid][20][UTT][ MSE = 0.1487 | LCC = 0.9715 | SRCC = 0.9709 ] +INFO:root:Average loss=0.04353871689736843 +INFO:root: +[Valid][21][UTT][ MSE = 0.1816 | LCC = 0.9721 | SRCC = 0.9712 ] +INFO:root:Average loss=0.03835683164000511 +INFO:root: +[Valid][22][UTT][ MSE = 0.1658 | LCC = 0.9722 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03911817562580109 +INFO:root: +[Valid][23][UTT][ MSE = 0.1744 | LCC = 0.9718 | SRCC = 0.9706 ] +INFO:root:Average loss=0.036172188416123394 +INFO:root: +[Valid][24][UTT][ MSE = 0.1878 | LCC = 0.9718 | SRCC = 0.9708 ] +INFO:root:Average loss=0.035925492599606514 +INFO:root: +[Valid][25][UTT][ MSE = 0.1803 | LCC = 0.9709 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03421226956695318 +INFO:root: +[Valid][26][UTT][ MSE = 0.1787 | LCC = 0.9711 | SRCC = 0.9699 ] +INFO:root:Average loss=0.03388920994102955 +INFO:root: +[Valid][27][UTT][ MSE = 0.1492 | LCC = 0.9715 | SRCC = 0.9698 ] +INFO:root:Average loss=0.03172731824219227 +INFO:root: +[Valid][28][UTT][ MSE = 0.1730 | LCC = 0.9714 | SRCC = 0.9700 ] +INFO:root:Average loss=0.032832013979554174 +INFO:root: +[Valid][29][UTT][ MSE = 0.1728 | LCC = 0.9713 | SRCC = 0.9703 ] +INFO:root:Average loss=0.0299021430760622 +INFO:root: +[Valid][30][UTT][ MSE = 0.1407 | LCC = 0.9717 | SRCC = 0.9708 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1648 | LCC = 0.9725 | SRCC = 0.9721 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run1/config.gin b/results_tencent/w2v2_xlsr_1b/layer28/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run1/model.pt b/results_tencent/w2v2_xlsr_1b/layer28/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..67847919918ee68112a99c7d2d8067ba696b4261 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20e30232b893e6f072466650bff16e0205b5d945998c01b6443791c462f09499 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run1/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer28/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..e69ead50c3b790c66b961c1c0b99ab6e49dbcd43 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a893444bc9c1075156c626df0ab445d81c835efcdbb7335c620df950cb7df56e +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run1/train.log b/results_tencent/w2v2_xlsr_1b/layer28/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d11fcec74dad386813e71a5bb1052d1f50fabb8d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.43436520648002624 +INFO:root: +[Valid][0][UTT][ MSE = 0.2634 | LCC = 0.9446 | SRCC = 0.9416 ] +INFO:root:Average loss=0.16784337717294692 +INFO:root: +[Valid][1][UTT][ MSE = 0.2166 | LCC = 0.9556 | SRCC = 0.9534 ] +INFO:root:Average loss=0.1268920556306839 +INFO:root: +[Valid][2][UTT][ MSE = 0.1993 | LCC = 0.9617 | SRCC = 0.9605 ] +INFO:root:Average loss=0.12599636870622635 +INFO:root: +[Valid][3][UTT][ MSE = 0.1686 | LCC = 0.9636 | SRCC = 0.9621 ] +INFO:root:Average loss=0.11060533779859542 +INFO:root: +[Valid][4][UTT][ MSE = 0.1916 | LCC = 0.9664 | SRCC = 0.9648 ] +INFO:root:Average loss=0.09824752694368362 +INFO:root: +[Valid][5][UTT][ MSE = 0.1711 | LCC = 0.9658 | SRCC = 0.9652 ] +INFO:root:Average loss=0.09001963776350021 +INFO:root: +[Valid][6][UTT][ MSE = 0.2105 | LCC = 0.9679 | SRCC = 0.9668 ] +INFO:root:Average loss=0.0839187841117382 +INFO:root: +[Valid][7][UTT][ MSE = 0.2406 | LCC = 0.9679 | SRCC = 0.9666 ] +INFO:root:Average loss=0.085706000238657 +INFO:root: +[Valid][8][UTT][ MSE = 0.1756 | LCC = 0.9706 | SRCC = 0.9695 ] +INFO:root:Average loss=0.07496621346473693 +INFO:root: +[Valid][9][UTT][ MSE = 0.1870 | LCC = 0.9706 | SRCC = 0.9695 ] +INFO:root:Average loss=0.07240891191363334 +INFO:root: +[Valid][10][UTT][ MSE = 0.1993 | LCC = 0.9682 | SRCC = 0.9683 ] +INFO:root:Average loss=0.06787023165822029 +INFO:root: +[Valid][11][UTT][ MSE = 0.1795 | LCC = 0.9700 | SRCC = 0.9693 ] +INFO:root:Average loss=0.06193009042739868 +INFO:root: +[Valid][12][UTT][ MSE = 0.1939 | LCC = 0.9710 | SRCC = 0.9698 ] +INFO:root:Average loss=0.06089767247438431 +INFO:root: +[Valid][13][UTT][ MSE = 0.1559 | LCC = 0.9712 | SRCC = 0.9701 ] +INFO:root:Average loss=0.06032230058312416 +INFO:root: +[Valid][14][UTT][ MSE = 0.2027 | LCC = 0.9708 | SRCC = 0.9704 ] +INFO:root:Average loss=0.05299304236471653 +INFO:root: +[Valid][15][UTT][ MSE = 0.1728 | LCC = 0.9721 | SRCC = 0.9708 ] +INFO:root:Average loss=0.05215273910760879 +INFO:root: +[Valid][16][UTT][ MSE = 0.2106 | LCC = 0.9673 | SRCC = 0.9698 ] +INFO:root:Average loss=0.054256389260292054 +INFO:root: +[Valid][17][UTT][ MSE = 0.1702 | LCC = 0.9721 | SRCC = 0.9708 ] +INFO:root:Average loss=0.04950376665592194 +INFO:root: +[Valid][18][UTT][ MSE = 0.1765 | LCC = 0.9713 | SRCC = 0.9698 ] +INFO:root:Average loss=0.04630725379288197 +INFO:root: +[Valid][19][UTT][ MSE = 0.1849 | LCC = 0.9720 | SRCC = 0.9709 ] +INFO:root:Average loss=0.04474904543161392 +INFO:root: +[Valid][20][UTT][ MSE = 0.1970 | LCC = 0.9713 | SRCC = 0.9703 ] +INFO:root:Average loss=0.04204313734173775 +INFO:root: +[Valid][21][UTT][ MSE = 0.1675 | LCC = 0.9725 | SRCC = 0.9715 ] +INFO:root:Average loss=0.04047057408094406 +INFO:root: +[Valid][22][UTT][ MSE = 0.1607 | LCC = 0.9709 | SRCC = 0.9708 ] +INFO:root:Average loss=0.04052626420557499 +INFO:root: +[Valid][23][UTT][ MSE = 0.1945 | LCC = 0.9721 | SRCC = 0.9711 ] +INFO:root:Average loss=0.035219163969159126 +INFO:root: +[Valid][24][UTT][ MSE = 0.1653 | LCC = 0.9728 | SRCC = 0.9716 ] +INFO:root:Average loss=0.03510926152765751 +INFO:root: +[Valid][25][UTT][ MSE = 0.1916 | LCC = 0.9722 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03494519618153572 +INFO:root: +[Valid][26][UTT][ MSE = 0.1787 | LCC = 0.9728 | SRCC = 0.9714 ] +INFO:root:Average loss=0.034681648209691045 +INFO:root: +[Valid][27][UTT][ MSE = 0.1782 | LCC = 0.9718 | SRCC = 0.9703 ] +INFO:root:Average loss=0.032476795330643654 +INFO:root: +[Valid][28][UTT][ MSE = 0.1612 | LCC = 0.9716 | SRCC = 0.9710 ] +INFO:root:Average loss=0.03362356300652027 +INFO:root: +[Valid][29][UTT][ MSE = 0.1665 | LCC = 0.9728 | SRCC = 0.9717 ] +INFO:root:Average loss=0.035132650673389436 +INFO:root: +[Valid][30][UTT][ MSE = 0.1681 | LCC = 0.9732 | SRCC = 0.9717 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1653 | LCC = 0.9732 | SRCC = 0.9725 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run2/config.gin b/results_tencent/w2v2_xlsr_1b/layer28/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run2/model.pt b/results_tencent/w2v2_xlsr_1b/layer28/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..3d89d7a11baebcb174e845c06c61448d8f5b31b4 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:55f4697e244bbc8866117a9b77ef04b3d2d67f91999fc4b9277ab547aefda0a1 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run2/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer28/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4c2fe4da327d15c575d39ff11b2fa37301d2b483 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cdd22bcc8dfe69cf0c124b16c6bcff6a146434fe3e6d684805e9de48e412204a +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run2/train.log b/results_tencent/w2v2_xlsr_1b/layer28/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5ea05e3ef888803f954253ec80703a5196727675 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4292513455748558 +INFO:root: +[Valid][0][UTT][ MSE = 0.2331 | LCC = 0.9479 | SRCC = 0.9461 ] +INFO:root:Average loss=0.15263043808937074 +INFO:root: +[Valid][1][UTT][ MSE = 0.2184 | LCC = 0.9569 | SRCC = 0.9561 ] +INFO:root:Average loss=0.12652881544828415 +INFO:root: +[Valid][2][UTT][ MSE = 0.1727 | LCC = 0.9616 | SRCC = 0.9624 ] +INFO:root:Average loss=0.10891009867191315 +INFO:root: +[Valid][3][UTT][ MSE = 0.2580 | LCC = 0.9648 | SRCC = 0.9653 ] +INFO:root:Average loss=0.10195451146364212 +INFO:root: +[Valid][4][UTT][ MSE = 0.1769 | LCC = 0.9671 | SRCC = 0.9664 ] +INFO:root:Average loss=0.09393549999594689 +INFO:root: +[Valid][5][UTT][ MSE = 0.2003 | LCC = 0.9684 | SRCC = 0.9681 ] +INFO:root:Average loss=0.08801340937614441 +INFO:root: +[Valid][6][UTT][ MSE = 0.2236 | LCC = 0.9678 | SRCC = 0.9686 ] +INFO:root:Average loss=0.08528462001681328 +INFO:root: +[Valid][7][UTT][ MSE = 0.1714 | LCC = 0.9699 | SRCC = 0.9697 ] +INFO:root:Average loss=0.07911811527609826 +INFO:root: +[Valid][8][UTT][ MSE = 0.1463 | LCC = 0.9697 | SRCC = 0.9690 ] +INFO:root:Average loss=0.07635987541079521 +INFO:root: +[Valid][9][UTT][ MSE = 0.2358 | LCC = 0.9674 | SRCC = 0.9690 ] +INFO:root:Average loss=0.07250139260292053 +INFO:root: +[Valid][10][UTT][ MSE = 0.1513 | LCC = 0.9704 | SRCC = 0.9701 ] +INFO:root:Average loss=0.064033173173666 +INFO:root: +[Valid][11][UTT][ MSE = 0.1655 | LCC = 0.9709 | SRCC = 0.9697 ] +INFO:root:Average loss=0.06394339534640313 +INFO:root: +[Valid][12][UTT][ MSE = 0.1887 | LCC = 0.9714 | SRCC = 0.9704 ] +INFO:root:Average loss=0.06061153873801231 +INFO:root: +[Valid][13][UTT][ MSE = 0.1648 | LCC = 0.9714 | SRCC = 0.9706 ] +INFO:root:Average loss=0.059300130754709246 +INFO:root: +[Valid][14][UTT][ MSE = 0.1525 | LCC = 0.9711 | SRCC = 0.9697 ] +INFO:root:Average loss=0.052904400423169136 +INFO:root: +[Valid][15][UTT][ MSE = 0.1817 | LCC = 0.9712 | SRCC = 0.9702 ] +INFO:root:Average loss=0.05055258302390576 +INFO:root: +[Valid][16][UTT][ MSE = 0.1448 | LCC = 0.9712 | SRCC = 0.9701 ] +INFO:root:Average loss=0.04944312982261181 +INFO:root: +[Valid][17][UTT][ MSE = 0.1654 | LCC = 0.9719 | SRCC = 0.9706 ] +INFO:root:Average loss=0.0488998995423317 +INFO:root: +[Valid][18][UTT][ MSE = 0.1766 | LCC = 0.9709 | SRCC = 0.9694 ] +INFO:root:Average loss=0.045883558541536335 +INFO:root: +[Valid][19][UTT][ MSE = 0.1777 | LCC = 0.9714 | SRCC = 0.9704 ] +INFO:root:Average loss=0.04424264730513096 +INFO:root: +[Valid][20][UTT][ MSE = 0.1756 | LCC = 0.9716 | SRCC = 0.9706 ] +INFO:root:Average loss=0.04305603565275669 +INFO:root: +[Valid][21][UTT][ MSE = 0.1453 | LCC = 0.9712 | SRCC = 0.9698 ] +INFO:root:Average loss=0.040680935725569724 +INFO:root: +[Valid][22][UTT][ MSE = 0.1660 | LCC = 0.9714 | SRCC = 0.9704 ] +INFO:root:Average loss=0.03801097497344017 +INFO:root: +[Valid][23][UTT][ MSE = 0.1363 | LCC = 0.9706 | SRCC = 0.9701 ] +INFO:root:Average loss=0.03707802632451057 +INFO:root: +[Valid][24][UTT][ MSE = 0.1547 | LCC = 0.9720 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03492611868679524 +INFO:root: +[Valid][25][UTT][ MSE = 0.1791 | LCC = 0.9714 | SRCC = 0.9705 ] +INFO:root:Average loss=0.038130509808659556 +INFO:root: +[Valid][26][UTT][ MSE = 0.1598 | LCC = 0.9709 | SRCC = 0.9692 ] +INFO:root:Average loss=0.034468119844794275 +INFO:root: +[Valid][27][UTT][ MSE = 0.1507 | LCC = 0.9713 | SRCC = 0.9700 ] +INFO:root:Average loss=0.030540265172719957 +INFO:root: +[Valid][28][UTT][ MSE = 0.1580 | LCC = 0.9718 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03046226154267788 +INFO:root: +[Valid][29][UTT][ MSE = 0.1706 | LCC = 0.9722 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03055321532487869 +INFO:root: +[Valid][30][UTT][ MSE = 0.1622 | LCC = 0.9718 | SRCC = 0.9704 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1707 | LCC = 0.9718 | SRCC = 0.9712 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run3/config.gin b/results_tencent/w2v2_xlsr_1b/layer28/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run3/model.pt b/results_tencent/w2v2_xlsr_1b/layer28/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..ef3b578f538c2154153b5bb9d6e2daf33328423d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6fe35d8224d0bca0f7a280c5fb419e0d079f7ca7d173e9e8af50a427094c9f66 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run3/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer28/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..c52362c5dfdcb375f468a13b67fd95bf1924781c --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:48fd72329bf9694706a3196d3e14f0e35d09f55a9660629965e02fbc5f941e32 +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run3/train.log b/results_tencent/w2v2_xlsr_1b/layer28/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..3d03da6cddfb760d5f3e30dbdf900385631b0d69 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.440031523168087 +INFO:root: +[Valid][0][UTT][ MSE = 0.2606 | LCC = 0.9447 | SRCC = 0.9413 ] +INFO:root:Average loss=0.1510029149055481 +INFO:root: +[Valid][1][UTT][ MSE = 0.2277 | LCC = 0.9564 | SRCC = 0.9555 ] +INFO:root:Average loss=0.12951346534490585 +INFO:root: +[Valid][2][UTT][ MSE = 0.2121 | LCC = 0.9618 | SRCC = 0.9618 ] +INFO:root:Average loss=0.1164106184244156 +INFO:root: +[Valid][3][UTT][ MSE = 0.2149 | LCC = 0.9649 | SRCC = 0.9647 ] +INFO:root:Average loss=0.11412224125862122 +INFO:root: +[Valid][4][UTT][ MSE = 0.2068 | LCC = 0.9667 | SRCC = 0.9662 ] +INFO:root:Average loss=0.10684178051352501 +INFO:root: +[Valid][5][UTT][ MSE = 0.1820 | LCC = 0.9675 | SRCC = 0.9671 ] +INFO:root:Average loss=0.09372384083271026 +INFO:root: +[Valid][6][UTT][ MSE = 0.2091 | LCC = 0.9686 | SRCC = 0.9681 ] +INFO:root:Average loss=0.08466502675414085 +INFO:root: +[Valid][7][UTT][ MSE = 0.1867 | LCC = 0.9686 | SRCC = 0.9676 ] +INFO:root:Average loss=0.08651957088708878 +INFO:root: +[Valid][8][UTT][ MSE = 0.1985 | LCC = 0.9703 | SRCC = 0.9697 ] +INFO:root:Average loss=0.0773718546628952 +INFO:root: +[Valid][9][UTT][ MSE = 0.2203 | LCC = 0.9679 | SRCC = 0.9672 ] +INFO:root:Average loss=0.07718141084909438 +INFO:root: +[Valid][10][UTT][ MSE = 0.1960 | LCC = 0.9694 | SRCC = 0.9691 ] +INFO:root:Average loss=0.0699727030992508 +INFO:root: +[Valid][11][UTT][ MSE = 0.1605 | LCC = 0.9706 | SRCC = 0.9696 ] +INFO:root:Average loss=0.06372846475243568 +INFO:root: +[Valid][12][UTT][ MSE = 0.1913 | LCC = 0.9707 | SRCC = 0.9698 ] +INFO:root:Average loss=0.06211354053020477 +INFO:root: +[Valid][13][UTT][ MSE = 0.1786 | LCC = 0.9707 | SRCC = 0.9695 ] +INFO:root:Average loss=0.06097686414420605 +INFO:root: +[Valid][14][UTT][ MSE = 0.1859 | LCC = 0.9706 | SRCC = 0.9701 ] +INFO:root:Average loss=0.05755674448609352 +INFO:root: +[Valid][15][UTT][ MSE = 0.2055 | LCC = 0.9709 | SRCC = 0.9703 ] +INFO:root:Average loss=0.05522987197339535 +INFO:root: +[Valid][16][UTT][ MSE = 0.1963 | LCC = 0.9704 | SRCC = 0.9696 ] +INFO:root:Average loss=0.05007884892821312 +INFO:root: +[Valid][17][UTT][ MSE = 0.1711 | LCC = 0.9705 | SRCC = 0.9699 ] +INFO:root:Average loss=0.048262220531702045 +INFO:root: +[Valid][18][UTT][ MSE = 0.1782 | LCC = 0.9712 | SRCC = 0.9700 ] +INFO:root:Average loss=0.045916961193084714 +INFO:root: +[Valid][19][UTT][ MSE = 0.2125 | LCC = 0.9703 | SRCC = 0.9697 ] +INFO:root:Average loss=0.045268656119704244 +INFO:root: +[Valid][20][UTT][ MSE = 0.2065 | LCC = 0.9710 | SRCC = 0.9703 ] +INFO:root:Average loss=0.04502574908733368 +INFO:root: +[Valid][21][UTT][ MSE = 0.1781 | LCC = 0.9702 | SRCC = 0.9699 ] +INFO:root:Average loss=0.04432175059616566 +INFO:root: +[Valid][22][UTT][ MSE = 0.1924 | LCC = 0.9713 | SRCC = 0.9707 ] +INFO:root:Average loss=0.039326834768056866 +INFO:root: +[Valid][23][UTT][ MSE = 0.1723 | LCC = 0.9709 | SRCC = 0.9696 ] +INFO:root:Average loss=0.03819737124443054 +INFO:root: +[Valid][24][UTT][ MSE = 0.1735 | LCC = 0.9711 | SRCC = 0.9698 ] +INFO:root:Average loss=0.03704575976729393 +INFO:root: +[Valid][25][UTT][ MSE = 0.1694 | LCC = 0.9706 | SRCC = 0.9700 ] +INFO:root:Average loss=0.03711046877503395 +INFO:root: +[Valid][26][UTT][ MSE = 0.1843 | LCC = 0.9712 | SRCC = 0.9697 ] +INFO:root:Average loss=0.03382663530111313 +INFO:root: +[Valid][27][UTT][ MSE = 0.1676 | LCC = 0.9713 | SRCC = 0.9700 ] +INFO:root:Average loss=0.03398748384416103 +INFO:root: +[Valid][28][UTT][ MSE = 0.1699 | LCC = 0.9702 | SRCC = 0.9701 ] +INFO:root:Average loss=0.034507860958576206 +INFO:root: +[Valid][29][UTT][ MSE = 0.1808 | LCC = 0.9712 | SRCC = 0.9697 ] +INFO:root:Average loss=0.033300335124135015 +INFO:root: +[Valid][30][UTT][ MSE = 0.1991 | LCC = 0.9694 | SRCC = 0.9695 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1917 | LCC = 0.9709 | SRCC = 0.9711 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run4/config.gin b/results_tencent/w2v2_xlsr_1b/layer28/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run4/model.pt b/results_tencent/w2v2_xlsr_1b/layer28/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..230a3bdbc31513d6a9e27e00f6abc8b3987d7464 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:629b1a8cede28878b913746205f92260fa5a7e5e84f24dcc0219b783d04a5750 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run4/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer28/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..67c0494682c45c58a13d30840e77b7ceae4f5c06 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:66a25b2d2281eab2a369c06f2937cdcfd2f1e823044ac58d5b210a2b1bdbb248 +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer28/run4/train.log b/results_tencent/w2v2_xlsr_1b/layer28/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d0a09b13aea2da78521b103e5f02567a8003b3fb --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer28/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.45104071247577665 +INFO:root: +[Valid][0][UTT][ MSE = 0.2644 | LCC = 0.9444 | SRCC = 0.9415 ] +INFO:root:Average loss=0.15564540696144105 +INFO:root: +[Valid][1][UTT][ MSE = 0.2445 | LCC = 0.9577 | SRCC = 0.9566 ] +INFO:root:Average loss=0.1294523845911026 +INFO:root: +[Valid][2][UTT][ MSE = 0.1808 | LCC = 0.9625 | SRCC = 0.9618 ] +INFO:root:Average loss=0.11473770451545716 +INFO:root: +[Valid][3][UTT][ MSE = 0.1561 | LCC = 0.9640 | SRCC = 0.9631 ] +INFO:root:Average loss=0.10114477506279945 +INFO:root: +[Valid][4][UTT][ MSE = 0.1766 | LCC = 0.9668 | SRCC = 0.9661 ] +INFO:root:Average loss=0.09800687661767006 +INFO:root: +[Valid][5][UTT][ MSE = 0.1804 | LCC = 0.9671 | SRCC = 0.9664 ] +INFO:root:Average loss=0.09040862596035004 +INFO:root: +[Valid][6][UTT][ MSE = 0.2087 | LCC = 0.9679 | SRCC = 0.9678 ] +INFO:root:Average loss=0.0820669005215168 +INFO:root: +[Valid][7][UTT][ MSE = 0.1924 | LCC = 0.9688 | SRCC = 0.9688 ] +INFO:root:Average loss=0.07674883726239204 +INFO:root: +[Valid][8][UTT][ MSE = 0.2165 | LCC = 0.9683 | SRCC = 0.9695 ] +INFO:root:Average loss=0.07455705419182777 +INFO:root: +[Valid][9][UTT][ MSE = 0.1579 | LCC = 0.9709 | SRCC = 0.9700 ] +INFO:root:Average loss=0.06893536764383316 +INFO:root: +[Valid][10][UTT][ MSE = 0.1995 | LCC = 0.9687 | SRCC = 0.9702 ] +INFO:root:Average loss=0.0656939802467823 +INFO:root: +[Valid][11][UTT][ MSE = 0.1622 | LCC = 0.9706 | SRCC = 0.9699 ] +INFO:root:Average loss=0.06102023346722126 +INFO:root: +[Valid][12][UTT][ MSE = 0.1996 | LCC = 0.9682 | SRCC = 0.9698 ] +INFO:root:Average loss=0.060656870007514956 +INFO:root: +[Valid][13][UTT][ MSE = 0.1647 | LCC = 0.9705 | SRCC = 0.9699 ] +INFO:root:Average loss=0.05668517148494721 +INFO:root: +[Valid][14][UTT][ MSE = 0.1930 | LCC = 0.9712 | SRCC = 0.9707 ] +INFO:root:Average loss=0.052656004518270495 +INFO:root: +[Valid][15][UTT][ MSE = 0.1544 | LCC = 0.9704 | SRCC = 0.9702 ] +INFO:root:Average loss=0.05462116216123104 +INFO:root: +[Valid][16][UTT][ MSE = 0.1432 | LCC = 0.9693 | SRCC = 0.9698 ] +INFO:root:Average loss=0.05053271833062172 +INFO:root: +[Valid][17][UTT][ MSE = 0.1677 | LCC = 0.9717 | SRCC = 0.9712 ] +INFO:root:Average loss=0.046558944806456566 +INFO:root: +[Valid][18][UTT][ MSE = 0.1558 | LCC = 0.9718 | SRCC = 0.9709 ] +INFO:root:Average loss=0.04431748333573341 +INFO:root: +[Valid][19][UTT][ MSE = 0.1476 | LCC = 0.9721 | SRCC = 0.9711 ] +INFO:root:Average loss=0.04077638557553291 +INFO:root: +[Valid][20][UTT][ MSE = 0.1632 | LCC = 0.9702 | SRCC = 0.9704 ] +INFO:root:Average loss=0.04296636222302914 +INFO:root: +[Valid][21][UTT][ MSE = 0.1672 | LCC = 0.9724 | SRCC = 0.9714 ] +INFO:root:Average loss=0.040729926034808156 +INFO:root: +[Valid][22][UTT][ MSE = 0.1817 | LCC = 0.9723 | SRCC = 0.9714 ] +INFO:root:Average loss=0.038487766161561014 +INFO:root: +[Valid][23][UTT][ MSE = 0.1636 | LCC = 0.9710 | SRCC = 0.9702 ] +INFO:root:Average loss=0.03926387348771095 +INFO:root: +[Valid][24][UTT][ MSE = 0.1766 | LCC = 0.9709 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03760325735807419 +INFO:root: +[Valid][25][UTT][ MSE = 0.1583 | LCC = 0.9722 | SRCC = 0.9710 ] +INFO:root:Average loss=0.03489084441959858 +INFO:root: +[Valid][26][UTT][ MSE = 0.1890 | LCC = 0.9724 | SRCC = 0.9718 ] +INFO:root:Average loss=0.03126743613183498 +INFO:root: +[Valid][27][UTT][ MSE = 0.1684 | LCC = 0.9721 | SRCC = 0.9710 ] +INFO:root:Average loss=0.03129567071795464 +INFO:root: +[Valid][28][UTT][ MSE = 0.1552 | LCC = 0.9719 | SRCC = 0.9711 ] +INFO:root:Average loss=0.030789371639490127 +INFO:root: +[Valid][29][UTT][ MSE = 0.1541 | LCC = 0.9727 | SRCC = 0.9714 ] +INFO:root:Average loss=0.028977390833199026 +INFO:root: +[Valid][30][UTT][ MSE = 0.1686 | LCC = 0.9726 | SRCC = 0.9715 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1536 | LCC = 0.9722 | SRCC = 0.9717 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run0/config.gin b/results_tencent/w2v2_xlsr_1b/layer30/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run0/model.pt b/results_tencent/w2v2_xlsr_1b/layer30/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..418bf8c6a98d5aee3fd98de514d7233f0ac576d8 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24ecea75d80e33082c7c0bfdc2102e2271850574ac0c067eefe7378754477e09 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run0/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer30/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..74e3d0b85d8a3a88bbb4a1e3dd46bae6d0d72907 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a1d8ec4b28a769da9374763c4d9f5ca22fed4b42dcc1619ce7ce4f178f7f1400 +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run0/train.log b/results_tencent/w2v2_xlsr_1b/layer30/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..62fc54d7690216bc547420233c70f66aec3e69f7 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.45316492462158203 +INFO:root: +[Valid][0][UTT][ MSE = 0.2672 | LCC = 0.9425 | SRCC = 0.9399 ] +INFO:root:Average loss=0.1624820802807808 +INFO:root: +[Valid][1][UTT][ MSE = 0.2652 | LCC = 0.9557 | SRCC = 0.9548 ] +INFO:root:Average loss=0.1350029287338257 +INFO:root: +[Valid][2][UTT][ MSE = 0.2095 | LCC = 0.9613 | SRCC = 0.9609 ] +INFO:root:Average loss=0.11808636611700057 +INFO:root: +[Valid][3][UTT][ MSE = 0.2310 | LCC = 0.9642 | SRCC = 0.9632 ] +INFO:root:Average loss=0.11512292218208313 +INFO:root: +[Valid][4][UTT][ MSE = 0.2522 | LCC = 0.9651 | SRCC = 0.9645 ] +INFO:root:Average loss=0.09936621215939521 +INFO:root: +[Valid][5][UTT][ MSE = 0.1790 | LCC = 0.9674 | SRCC = 0.9666 ] +INFO:root:Average loss=0.09014165899157524 +INFO:root: +[Valid][6][UTT][ MSE = 0.1877 | LCC = 0.9681 | SRCC = 0.9669 ] +INFO:root:Average loss=0.08815925443172455 +INFO:root: +[Valid][7][UTT][ MSE = 0.2044 | LCC = 0.9686 | SRCC = 0.9674 ] +INFO:root:Average loss=0.08451049888134003 +INFO:root: +[Valid][8][UTT][ MSE = 0.1841 | LCC = 0.9688 | SRCC = 0.9681 ] +INFO:root:Average loss=0.07726837801933288 +INFO:root: +[Valid][9][UTT][ MSE = 0.1972 | LCC = 0.9690 | SRCC = 0.9684 ] +INFO:root:Average loss=0.07517916271090508 +INFO:root: +[Valid][10][UTT][ MSE = 0.1791 | LCC = 0.9708 | SRCC = 0.9696 ] +INFO:root:Average loss=0.06823520722985267 +INFO:root: +[Valid][11][UTT][ MSE = 0.1699 | LCC = 0.9696 | SRCC = 0.9693 ] +INFO:root:Average loss=0.06599734270572663 +INFO:root: +[Valid][12][UTT][ MSE = 0.1945 | LCC = 0.9707 | SRCC = 0.9697 ] +INFO:root:Average loss=0.06199913188815117 +INFO:root: +[Valid][13][UTT][ MSE = 0.1759 | LCC = 0.9717 | SRCC = 0.9706 ] +INFO:root:Average loss=0.05940119987726212 +INFO:root: +[Valid][14][UTT][ MSE = 0.1731 | LCC = 0.9712 | SRCC = 0.9703 ] +INFO:root:Average loss=0.05506279167532921 +INFO:root: +[Valid][15][UTT][ MSE = 0.1550 | LCC = 0.9689 | SRCC = 0.9691 ] +INFO:root:Average loss=0.052658567190170286 +INFO:root: +[Valid][16][UTT][ MSE = 0.1663 | LCC = 0.9713 | SRCC = 0.9703 ] +INFO:root:Average loss=0.04984529814124107 +INFO:root: +[Valid][17][UTT][ MSE = 0.1856 | LCC = 0.9717 | SRCC = 0.9707 ] +INFO:root:Average loss=0.051171438083052634 +INFO:root: +[Valid][18][UTT][ MSE = 0.1884 | LCC = 0.9705 | SRCC = 0.9693 ] +INFO:root:Average loss=0.050857501164078715 +INFO:root: +[Valid][19][UTT][ MSE = 0.2093 | LCC = 0.9698 | SRCC = 0.9699 ] +INFO:root:Average loss=0.045717981025576594 +INFO:root: +[Valid][20][UTT][ MSE = 0.1813 | LCC = 0.9713 | SRCC = 0.9705 ] +INFO:root:Average loss=0.04459835375845432 +INFO:root: +[Valid][21][UTT][ MSE = 0.1760 | LCC = 0.9719 | SRCC = 0.9712 ] +INFO:root:Average loss=0.04579315210878849 +INFO:root: +[Valid][22][UTT][ MSE = 0.1638 | LCC = 0.9720 | SRCC = 0.9709 ] +INFO:root:Average loss=0.04119412423670292 +INFO:root: +[Valid][23][UTT][ MSE = 0.1523 | LCC = 0.9716 | SRCC = 0.9704 ] +INFO:root:Average loss=0.03982239583134651 +INFO:root: +[Valid][24][UTT][ MSE = 0.2024 | LCC = 0.9698 | SRCC = 0.9707 ] +INFO:root:Average loss=0.03630590111017227 +INFO:root: +[Valid][25][UTT][ MSE = 0.1730 | LCC = 0.9718 | SRCC = 0.9711 ] +INFO:root:Average loss=0.03552814650535584 +INFO:root: +[Valid][26][UTT][ MSE = 0.1785 | LCC = 0.9719 | SRCC = 0.9716 ] +INFO:root:Average loss=0.035174749866127966 +INFO:root: +[Valid][27][UTT][ MSE = 0.1874 | LCC = 0.9717 | SRCC = 0.9716 ] +INFO:root:Average loss=0.03277854353189468 +INFO:root: +[Valid][28][UTT][ MSE = 0.1827 | LCC = 0.9717 | SRCC = 0.9706 ] +INFO:root:Average loss=0.033440541341900824 +INFO:root: +[Valid][29][UTT][ MSE = 0.1643 | LCC = 0.9726 | SRCC = 0.9718 ] +INFO:root:Average loss=0.032130675315856935 +INFO:root: +[Valid][30][UTT][ MSE = 0.1512 | LCC = 0.9722 | SRCC = 0.9709 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1668 | LCC = 0.9715 | SRCC = 0.9706 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run1/config.gin b/results_tencent/w2v2_xlsr_1b/layer30/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run1/model.pt b/results_tencent/w2v2_xlsr_1b/layer30/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..af5cad377f737a1eb69719548c991780a3e73a4d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5efd30be877525e02fb9605910bcae85faa2f2fcdb2889183352eb9da3b98b97 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run1/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer30/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..225a636143362f31d36370389d90697f055b68db --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ce0a9cb6998d606c66f730f799437f92db5a4d2057eb2b84ddd39abf2c6cf05 +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run1/train.log b/results_tencent/w2v2_xlsr_1b/layer30/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..9ee4d54b283d2f4a97aef5e6ff5e4b7f38c460b7 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.40843772423267366 +INFO:root: +[Valid][0][UTT][ MSE = 0.2443 | LCC = 0.9423 | SRCC = 0.9426 ] +INFO:root:Average loss=0.16387355762720107 +INFO:root: +[Valid][1][UTT][ MSE = 0.2274 | LCC = 0.9568 | SRCC = 0.9575 ] +INFO:root:Average loss=0.1300381636917591 +INFO:root: +[Valid][2][UTT][ MSE = 0.2282 | LCC = 0.9595 | SRCC = 0.9605 ] +INFO:root:Average loss=0.12219942814111709 +INFO:root: +[Valid][3][UTT][ MSE = 0.1983 | LCC = 0.9630 | SRCC = 0.9633 ] +INFO:root:Average loss=0.10121498170495033 +INFO:root: +[Valid][4][UTT][ MSE = 0.1865 | LCC = 0.9653 | SRCC = 0.9645 ] +INFO:root:Average loss=0.09622880652546882 +INFO:root: +[Valid][5][UTT][ MSE = 0.2215 | LCC = 0.9655 | SRCC = 0.9660 ] +INFO:root:Average loss=0.09477669695019722 +INFO:root: +[Valid][6][UTT][ MSE = 0.1866 | LCC = 0.9658 | SRCC = 0.9662 ] +INFO:root:Average loss=0.08931307318806649 +INFO:root: +[Valid][7][UTT][ MSE = 0.1724 | LCC = 0.9684 | SRCC = 0.9683 ] +INFO:root:Average loss=0.07643592369556428 +INFO:root: +[Valid][8][UTT][ MSE = 0.1852 | LCC = 0.9685 | SRCC = 0.9691 ] +INFO:root:Average loss=0.0753068496286869 +INFO:root: +[Valid][9][UTT][ MSE = 0.1570 | LCC = 0.9672 | SRCC = 0.9674 ] +INFO:root:Average loss=0.07496659237146377 +INFO:root: +[Valid][10][UTT][ MSE = 0.1717 | LCC = 0.9693 | SRCC = 0.9689 ] +INFO:root:Average loss=0.06640326967835426 +INFO:root: +[Valid][11][UTT][ MSE = 0.1734 | LCC = 0.9700 | SRCC = 0.9696 ] +INFO:root:Average loss=0.0673659481704235 +INFO:root: +[Valid][12][UTT][ MSE = 0.1882 | LCC = 0.9694 | SRCC = 0.9699 ] +INFO:root:Average loss=0.060909967422485355 +INFO:root: +[Valid][13][UTT][ MSE = 0.1957 | LCC = 0.9697 | SRCC = 0.9696 ] +INFO:root:Average loss=0.05715302738547325 +INFO:root: +[Valid][14][UTT][ MSE = 0.1826 | LCC = 0.9716 | SRCC = 0.9708 ] +INFO:root:Average loss=0.05522725921869278 +INFO:root: +[Valid][15][UTT][ MSE = 0.1885 | LCC = 0.9717 | SRCC = 0.9714 ] +INFO:root:Average loss=0.051086230978369715 +INFO:root: +[Valid][16][UTT][ MSE = 0.1699 | LCC = 0.9718 | SRCC = 0.9709 ] +INFO:root:Average loss=0.050900917604565624 +INFO:root: +[Valid][17][UTT][ MSE = 0.1626 | LCC = 0.9718 | SRCC = 0.9714 ] +INFO:root:Average loss=0.04754983891546726 +INFO:root: +[Valid][18][UTT][ MSE = 0.1518 | LCC = 0.9711 | SRCC = 0.9705 ] +INFO:root:Average loss=0.04838381187617779 +INFO:root: +[Valid][19][UTT][ MSE = 0.1764 | LCC = 0.9711 | SRCC = 0.9704 ] +INFO:root:Average loss=0.045694956660270694 +INFO:root: +[Valid][20][UTT][ MSE = 0.1486 | LCC = 0.9713 | SRCC = 0.9707 ] +INFO:root:Average loss=0.04216049019992352 +INFO:root: +[Valid][21][UTT][ MSE = 0.1552 | LCC = 0.9708 | SRCC = 0.9706 ] +INFO:root:Average loss=0.042669099912047384 +INFO:root: +[Valid][22][UTT][ MSE = 0.1412 | LCC = 0.9704 | SRCC = 0.9704 ] +INFO:root:Average loss=0.038995586141943935 +INFO:root: +[Valid][23][UTT][ MSE = 0.1592 | LCC = 0.9720 | SRCC = 0.9712 ] +INFO:root:Average loss=0.037749811396002766 +INFO:root: +[Valid][24][UTT][ MSE = 0.1630 | LCC = 0.9721 | SRCC = 0.9714 ] +INFO:root:Average loss=0.03578999204933643 +INFO:root: +[Valid][25][UTT][ MSE = 0.1586 | LCC = 0.9724 | SRCC = 0.9716 ] +INFO:root:Average loss=0.03557871122658253 +INFO:root: +[Valid][26][UTT][ MSE = 0.1427 | LCC = 0.9718 | SRCC = 0.9716 ] +INFO:root:Average loss=0.03273314566910267 +INFO:root: +[Valid][27][UTT][ MSE = 0.1567 | LCC = 0.9727 | SRCC = 0.9722 ] +INFO:root:Average loss=0.035910183399915696 +INFO:root: +[Valid][28][UTT][ MSE = 0.1509 | LCC = 0.9717 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03414994797110558 +INFO:root: +[Valid][29][UTT][ MSE = 0.1579 | LCC = 0.9727 | SRCC = 0.9719 ] +INFO:root:Average loss=0.0332096079736948 +INFO:root: +[Valid][30][UTT][ MSE = 0.1442 | LCC = 0.9725 | SRCC = 0.9715 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1569 | LCC = 0.9719 | SRCC = 0.9713 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run2/config.gin b/results_tencent/w2v2_xlsr_1b/layer30/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run2/model.pt b/results_tencent/w2v2_xlsr_1b/layer30/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..e284fd70ed99c750ca282cac35cdc4dcccef22d4 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0626143c1582d76e531511720ee7255f0fb38762f5656c6651ce4a685b629fd +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run2/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer30/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..55827a799b8b010bffdf4f4a6a4c2d721e2184de --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f9fe6297a87c505015aafaf21e89618bb8fd850133d1b7288776f619278aea75 +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run2/train.log b/results_tencent/w2v2_xlsr_1b/layer30/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..65db39ef271bafda4a7ec3d72ccd4f34739d1ffd --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.5147365785837174 +INFO:root: +[Valid][0][UTT][ MSE = 0.3045 | LCC = 0.9394 | SRCC = 0.9388 ] +INFO:root:Average loss=0.17231032329797744 +INFO:root: +[Valid][1][UTT][ MSE = 0.2808 | LCC = 0.9542 | SRCC = 0.9540 ] +INFO:root:Average loss=0.13128359258174896 +INFO:root: +[Valid][2][UTT][ MSE = 0.1686 | LCC = 0.9608 | SRCC = 0.9597 ] +INFO:root:Average loss=0.11753569120168686 +INFO:root: +[Valid][3][UTT][ MSE = 0.1839 | LCC = 0.9636 | SRCC = 0.9628 ] +INFO:root:Average loss=0.1046034106016159 +INFO:root: +[Valid][4][UTT][ MSE = 0.1887 | LCC = 0.9646 | SRCC = 0.9646 ] +INFO:root:Average loss=0.09974093872308731 +INFO:root: +[Valid][5][UTT][ MSE = 0.2340 | LCC = 0.9663 | SRCC = 0.9666 ] +INFO:root:Average loss=0.09217378297448159 +INFO:root: +[Valid][6][UTT][ MSE = 0.2018 | LCC = 0.9680 | SRCC = 0.9678 ] +INFO:root:Average loss=0.08536110657453537 +INFO:root: +[Valid][7][UTT][ MSE = 0.2045 | LCC = 0.9688 | SRCC = 0.9678 ] +INFO:root:Average loss=0.08660328814387322 +INFO:root: +[Valid][8][UTT][ MSE = 0.2011 | LCC = 0.9693 | SRCC = 0.9686 ] +INFO:root:Average loss=0.07892864936590195 +INFO:root: +[Valid][9][UTT][ MSE = 0.2552 | LCC = 0.9684 | SRCC = 0.9696 ] +INFO:root:Average loss=0.0743848489522934 +INFO:root: +[Valid][10][UTT][ MSE = 0.2083 | LCC = 0.9700 | SRCC = 0.9692 ] +INFO:root:Average loss=0.0677552500963211 +INFO:root: +[Valid][11][UTT][ MSE = 0.1861 | LCC = 0.9699 | SRCC = 0.9692 ] +INFO:root:Average loss=0.06810321009159089 +INFO:root: +[Valid][12][UTT][ MSE = 0.1830 | LCC = 0.9705 | SRCC = 0.9697 ] +INFO:root:Average loss=0.06099610039591789 +INFO:root: +[Valid][13][UTT][ MSE = 0.1826 | LCC = 0.9704 | SRCC = 0.9692 ] +INFO:root:Average loss=0.06005183780193329 +INFO:root: +[Valid][14][UTT][ MSE = 0.1668 | LCC = 0.9711 | SRCC = 0.9699 ] +INFO:root:Average loss=0.05585414209961891 +INFO:root: +[Valid][15][UTT][ MSE = 0.1703 | LCC = 0.9711 | SRCC = 0.9702 ] +INFO:root:Average loss=0.056953209355473516 +INFO:root: +[Valid][16][UTT][ MSE = 0.2025 | LCC = 0.9690 | SRCC = 0.9695 ] +INFO:root:Average loss=0.05033880516886711 +INFO:root: +[Valid][17][UTT][ MSE = 0.1804 | LCC = 0.9711 | SRCC = 0.9701 ] +INFO:root:Average loss=0.051505419433116914 +INFO:root: +[Valid][18][UTT][ MSE = 0.1985 | LCC = 0.9703 | SRCC = 0.9704 ] +INFO:root:Average loss=0.04612691509723663 +INFO:root: +[Valid][19][UTT][ MSE = 0.1925 | LCC = 0.9712 | SRCC = 0.9707 ] +INFO:root:Average loss=0.04616462118923664 +INFO:root: +[Valid][20][UTT][ MSE = 0.2092 | LCC = 0.9706 | SRCC = 0.9701 ] +INFO:root:Average loss=0.04419497817754745 +INFO:root: +[Valid][21][UTT][ MSE = 0.1612 | LCC = 0.9695 | SRCC = 0.9698 ] +INFO:root:Average loss=0.0438974684625864 +INFO:root: +[Valid][22][UTT][ MSE = 0.2008 | LCC = 0.9707 | SRCC = 0.9707 ] +INFO:root:Average loss=0.04131982217729092 +INFO:root: +[Valid][23][UTT][ MSE = 0.1765 | LCC = 0.9709 | SRCC = 0.9702 ] +INFO:root:Average loss=0.0402647145986557 +INFO:root: +[Valid][24][UTT][ MSE = 0.1887 | LCC = 0.9709 | SRCC = 0.9707 ] +INFO:root:Average loss=0.040552014455199244 +INFO:root: +[Valid][25][UTT][ MSE = 0.1892 | LCC = 0.9717 | SRCC = 0.9709 ] +INFO:root:Average loss=0.0368333085924387 +INFO:root: +[Valid][26][UTT][ MSE = 0.1779 | LCC = 0.9716 | SRCC = 0.9705 ] +INFO:root:Average loss=0.03468458342552185 +INFO:root: +[Valid][27][UTT][ MSE = 0.1844 | LCC = 0.9721 | SRCC = 0.9713 ] +INFO:root:Average loss=0.03371872448921204 +INFO:root: +[Valid][28][UTT][ MSE = 0.1786 | LCC = 0.9721 | SRCC = 0.9716 ] +INFO:root:Average loss=0.03225068610906601 +INFO:root: +[Valid][29][UTT][ MSE = 0.1916 | LCC = 0.9725 | SRCC = 0.9715 ] +INFO:root:Average loss=0.03146370074152947 +INFO:root: +[Valid][30][UTT][ MSE = 0.1723 | LCC = 0.9716 | SRCC = 0.9711 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1915 | LCC = 0.9718 | SRCC = 0.9711 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run3/config.gin b/results_tencent/w2v2_xlsr_1b/layer30/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run3/model.pt b/results_tencent/w2v2_xlsr_1b/layer30/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..51824ad792f1c1ff3c4f95deff89fd35f8414748 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fb508b2a4fe36bd1a980a0075e31664e19ec6a0bf0651933652e27b2a0ee18b6 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run3/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer30/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..4ba532129aa1233bebbe71c670a3465a96f39d3f --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24ebb4c30385c4305d25031c9aaefebead3f3e19e1cefa4edec46ab86f63f71b +size 4060422 diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run3/train.log b/results_tencent/w2v2_xlsr_1b/layer30/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..8209a24dfee2e9fa081fe83232e29d6e9de98747 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.42114655900001524 +INFO:root: +[Valid][0][UTT][ MSE = 0.2697 | LCC = 0.9403 | SRCC = 0.9393 ] +INFO:root:Average loss=0.16098882400989534 +INFO:root: +[Valid][1][UTT][ MSE = 0.2077 | LCC = 0.9560 | SRCC = 0.9563 ] +INFO:root:Average loss=0.13812776744365693 +INFO:root: +[Valid][2][UTT][ MSE = 0.2164 | LCC = 0.9602 | SRCC = 0.9595 ] +INFO:root:Average loss=0.11964855107665062 +INFO:root: +[Valid][3][UTT][ MSE = 0.2101 | LCC = 0.9634 | SRCC = 0.9629 ] +INFO:root:Average loss=0.10531798213720321 +INFO:root: +[Valid][4][UTT][ MSE = 0.1852 | LCC = 0.9667 | SRCC = 0.9662 ] +INFO:root:Average loss=0.09773841142654419 +INFO:root: +[Valid][5][UTT][ MSE = 0.1963 | LCC = 0.9686 | SRCC = 0.9678 ] +INFO:root:Average loss=0.08897233870625496 +INFO:root: +[Valid][6][UTT][ MSE = 0.2020 | LCC = 0.9682 | SRCC = 0.9679 ] +INFO:root:Average loss=0.08488813725113868 +INFO:root: +[Valid][7][UTT][ MSE = 0.2220 | LCC = 0.9694 | SRCC = 0.9691 ] +INFO:root:Average loss=0.07915021753311158 +INFO:root: +[Valid][8][UTT][ MSE = 0.2026 | LCC = 0.9703 | SRCC = 0.9701 ] +INFO:root:Average loss=0.07714604300260544 +INFO:root: +[Valid][9][UTT][ MSE = 0.1672 | LCC = 0.9709 | SRCC = 0.9705 ] +INFO:root:Average loss=0.06884646371006965 +INFO:root: +[Valid][10][UTT][ MSE = 0.2296 | LCC = 0.9695 | SRCC = 0.9692 ] +INFO:root:Average loss=0.06614958555996418 +INFO:root: +[Valid][11][UTT][ MSE = 0.1500 | LCC = 0.9712 | SRCC = 0.9706 ] +INFO:root:Average loss=0.0637829641699791 +INFO:root: +[Valid][12][UTT][ MSE = 0.2046 | LCC = 0.9713 | SRCC = 0.9711 ] +INFO:root:Average loss=0.058953686863183974 +INFO:root: +[Valid][13][UTT][ MSE = 0.1420 | LCC = 0.9711 | SRCC = 0.9700 ] +INFO:root:Average loss=0.060542968630790714 +INFO:root: +[Valid][14][UTT][ MSE = 0.1748 | LCC = 0.9712 | SRCC = 0.9708 ] +INFO:root:Average loss=0.05629207158088684 +INFO:root: +[Valid][15][UTT][ MSE = 0.1623 | LCC = 0.9716 | SRCC = 0.9710 ] +INFO:root:Average loss=0.057011110931634905 +INFO:root: +[Valid][16][UTT][ MSE = 0.1717 | LCC = 0.9710 | SRCC = 0.9705 ] +INFO:root:Average loss=0.05143259525299072 +INFO:root: +[Valid][17][UTT][ MSE = 0.1750 | LCC = 0.9706 | SRCC = 0.9700 ] +INFO:root:Average loss=0.047790775820612905 +INFO:root: +[Valid][18][UTT][ MSE = 0.1408 | LCC = 0.9715 | SRCC = 0.9708 ] +INFO:root:Average loss=0.04666347596049309 +INFO:root: +[Valid][19][UTT][ MSE = 0.1590 | LCC = 0.9714 | SRCC = 0.9708 ] +INFO:root:Average loss=0.04368059121072292 +INFO:root: +[Valid][20][UTT][ MSE = 0.1542 | LCC = 0.9714 | SRCC = 0.9709 ] +INFO:root:Average loss=0.04247399741411209 +INFO:root: +[Valid][21][UTT][ MSE = 0.1640 | LCC = 0.9720 | SRCC = 0.9713 ] +INFO:root:Average loss=0.04219787603616715 +INFO:root: +[Valid][22][UTT][ MSE = 0.1656 | LCC = 0.9717 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03869632934033871 +INFO:root: +[Valid][23][UTT][ MSE = 0.1677 | LCC = 0.9717 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03811897929012775 +INFO:root: +[Valid][24][UTT][ MSE = 0.1484 | LCC = 0.9711 | SRCC = 0.9712 ] +INFO:root:Average loss=0.036496924832463265 +INFO:root: +[Valid][25][UTT][ MSE = 0.1684 | LCC = 0.9717 | SRCC = 0.9709 ] +INFO:root:Average loss=0.03734706000983715 +INFO:root: +[Valid][26][UTT][ MSE = 0.1635 | LCC = 0.9716 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03438530930876732 +INFO:root: +[Valid][27][UTT][ MSE = 0.1676 | LCC = 0.9717 | SRCC = 0.9711 ] +INFO:root:Average loss=0.03717813888192177 +INFO:root: +[Valid][28][UTT][ MSE = 0.1529 | LCC = 0.9719 | SRCC = 0.9710 ] +INFO:root:Average loss=0.032898233801126484 +INFO:root: +[Valid][29][UTT][ MSE = 0.1560 | LCC = 0.9719 | SRCC = 0.9710 ] +INFO:root:Average loss=0.03096433287858963 +INFO:root: +[Valid][30][UTT][ MSE = 0.1341 | LCC = 0.9702 | SRCC = 0.9702 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1642 | LCC = 0.9715 | SRCC = 0.9713 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run4/config.gin b/results_tencent/w2v2_xlsr_1b/layer30/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run4/model.pt b/results_tencent/w2v2_xlsr_1b/layer30/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..bd19aa8f8ef748cde5704c05a979e725c708746b --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:59263568dde360d06dbeafefee39e359943b8288cb2ef36790a35c3f89837baf +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run4/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer30/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..046328f93ffa5719cce73b1961624b76f3a8ae4c --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3fed7584adfcf7f5f93f4483f8a9b81fb6e902878b88cd4686d6cf7ccf23960c +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer30/run4/train.log b/results_tencent/w2v2_xlsr_1b/layer30/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..fe8d4676d9cae478d73ce18e4f55af807c626f34 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer30/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4132209242582321 +INFO:root: +[Valid][0][UTT][ MSE = 0.2625 | LCC = 0.9425 | SRCC = 0.9394 ] +INFO:root:Average loss=0.16573581475019455 +INFO:root: +[Valid][1][UTT][ MSE = 0.2210 | LCC = 0.9568 | SRCC = 0.9563 ] +INFO:root:Average loss=0.12490762001276016 +INFO:root: +[Valid][2][UTT][ MSE = 0.2346 | LCC = 0.9614 | SRCC = 0.9616 ] +INFO:root:Average loss=0.11714120236039162 +INFO:root: +[Valid][3][UTT][ MSE = 0.1909 | LCC = 0.9634 | SRCC = 0.9634 ] +INFO:root:Average loss=0.0988269785642624 +INFO:root: +[Valid][4][UTT][ MSE = 0.2101 | LCC = 0.9652 | SRCC = 0.9666 ] +INFO:root:Average loss=0.09297812676429748 +INFO:root: +[Valid][5][UTT][ MSE = 0.1651 | LCC = 0.9656 | SRCC = 0.9669 ] +INFO:root:Average loss=0.08815433722734452 +INFO:root: +[Valid][6][UTT][ MSE = 0.1824 | LCC = 0.9679 | SRCC = 0.9678 ] +INFO:root:Average loss=0.08868917629122734 +INFO:root: +[Valid][7][UTT][ MSE = 0.1853 | LCC = 0.9666 | SRCC = 0.9672 ] +INFO:root:Average loss=0.07642530050873757 +INFO:root: +[Valid][8][UTT][ MSE = 0.1771 | LCC = 0.9691 | SRCC = 0.9691 ] +INFO:root:Average loss=0.07091952127218247 +INFO:root: +[Valid][9][UTT][ MSE = 0.2075 | LCC = 0.9681 | SRCC = 0.9689 ] +INFO:root:Average loss=0.06980417910218238 +INFO:root: +[Valid][10][UTT][ MSE = 0.1917 | LCC = 0.9690 | SRCC = 0.9694 ] +INFO:root:Average loss=0.0662857876420021 +INFO:root: +[Valid][11][UTT][ MSE = 0.2336 | LCC = 0.9685 | SRCC = 0.9684 ] +INFO:root:Average loss=0.06567746332287788 +INFO:root: +[Valid][12][UTT][ MSE = 0.1849 | LCC = 0.9698 | SRCC = 0.9697 ] +INFO:root:Average loss=0.059321426570415496 +INFO:root: +[Valid][13][UTT][ MSE = 0.1529 | LCC = 0.9705 | SRCC = 0.9700 ] +INFO:root:Average loss=0.058852547764778135 +INFO:root: +[Valid][14][UTT][ MSE = 0.1400 | LCC = 0.9688 | SRCC = 0.9683 ] +INFO:root:Average loss=0.05331182639300823 +INFO:root: +[Valid][15][UTT][ MSE = 0.1890 | LCC = 0.9702 | SRCC = 0.9697 ] +INFO:root:Average loss=0.05133016519248486 +INFO:root: +[Valid][16][UTT][ MSE = 0.1500 | LCC = 0.9702 | SRCC = 0.9700 ] +INFO:root:Average loss=0.050823039770126345 +INFO:root: +[Valid][17][UTT][ MSE = 0.2125 | LCC = 0.9687 | SRCC = 0.9701 ] +INFO:root:Average loss=0.04614640282094479 +INFO:root: +[Valid][18][UTT][ MSE = 0.1555 | LCC = 0.9708 | SRCC = 0.9700 ] +INFO:root:Average loss=0.04432921105623245 +INFO:root: +[Valid][19][UTT][ MSE = 0.1666 | LCC = 0.9708 | SRCC = 0.9702 ] +INFO:root:Average loss=0.04164204166829586 +INFO:root: +[Valid][20][UTT][ MSE = 0.1649 | LCC = 0.9709 | SRCC = 0.9705 ] +INFO:root:Average loss=0.040818579971790316 +INFO:root: +[Valid][21][UTT][ MSE = 0.1664 | LCC = 0.9707 | SRCC = 0.9703 ] +INFO:root:Average loss=0.04010241144895554 +INFO:root: +[Valid][22][UTT][ MSE = 0.1708 | LCC = 0.9710 | SRCC = 0.9702 ] +INFO:root:Average loss=0.038252696320414545 +INFO:root: +[Valid][23][UTT][ MSE = 0.1523 | LCC = 0.9698 | SRCC = 0.9693 ] +INFO:root:Average loss=0.03743449538946152 +INFO:root: +[Valid][24][UTT][ MSE = 0.1836 | LCC = 0.9709 | SRCC = 0.9704 ] +INFO:root:Average loss=0.038069233641028405 +INFO:root: +[Valid][25][UTT][ MSE = 0.1877 | LCC = 0.9699 | SRCC = 0.9699 ] +INFO:root:Average loss=0.03639935654401779 +INFO:root: +[Valid][26][UTT][ MSE = 0.1641 | LCC = 0.9713 | SRCC = 0.9702 ] +INFO:root:Average loss=0.034121145844459536 +INFO:root: +[Valid][27][UTT][ MSE = 0.1645 | LCC = 0.9710 | SRCC = 0.9705 ] +INFO:root:Average loss=0.03289041595160961 +INFO:root: +[Valid][28][UTT][ MSE = 0.1546 | LCC = 0.9709 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03224160423874855 +INFO:root: +[Valid][29][UTT][ MSE = 0.1571 | LCC = 0.9716 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03060628394782543 +INFO:root: +[Valid][30][UTT][ MSE = 0.1845 | LCC = 0.9715 | SRCC = 0.9707 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1546 | LCC = 0.9722 | SRCC = 0.9717 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run0/config.gin b/results_tencent/w2v2_xlsr_1b/layer32/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run0/model.pt b/results_tencent/w2v2_xlsr_1b/layer32/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..674e98e7e8fcce0e781c6ec7cb148964fc1781ad --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0582810d4e2ff87e1504c063fe88e56e69102cb63688b279519a9dacff7db6fa +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run0/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer32/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..0985ab7934d54ebfad485d4611868e8d5e11161e --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4f9b732bb6a5ab610259b5b4a5b1a663fa2dfbc0570c82b5de988c03c332aa56 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run0/train.log b/results_tencent/w2v2_xlsr_1b/layer32/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..e0e7383ba5fdfbf16e183b72c343927fe9b7c581 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.46447015142440795 +INFO:root: +[Valid][0][UTT][ MSE = 0.2686 | LCC = 0.9332 | SRCC = 0.9305 ] +INFO:root:Average loss=0.16985192304849625 +INFO:root: +[Valid][1][UTT][ MSE = 0.2333 | LCC = 0.9527 | SRCC = 0.9512 ] +INFO:root:Average loss=0.13130974519252778 +INFO:root: +[Valid][2][UTT][ MSE = 0.2132 | LCC = 0.9597 | SRCC = 0.9585 ] +INFO:root:Average loss=0.12209726196527482 +INFO:root: +[Valid][3][UTT][ MSE = 0.2343 | LCC = 0.9626 | SRCC = 0.9621 ] +INFO:root:Average loss=0.10174457955360412 +INFO:root: +[Valid][4][UTT][ MSE = 0.1956 | LCC = 0.9648 | SRCC = 0.9641 ] +INFO:root:Average loss=0.09586338809132576 +INFO:root: +[Valid][5][UTT][ MSE = 0.2054 | LCC = 0.9660 | SRCC = 0.9654 ] +INFO:root:Average loss=0.08620637246966362 +INFO:root: +[Valid][6][UTT][ MSE = 0.1688 | LCC = 0.9670 | SRCC = 0.9667 ] +INFO:root:Average loss=0.08859682378172874 +INFO:root: +[Valid][7][UTT][ MSE = 0.1618 | LCC = 0.9664 | SRCC = 0.9662 ] +INFO:root:Average loss=0.07916825577616692 +INFO:root: +[Valid][8][UTT][ MSE = 0.1542 | LCC = 0.9671 | SRCC = 0.9675 ] +INFO:root:Average loss=0.07498814633488655 +INFO:root: +[Valid][9][UTT][ MSE = 0.1854 | LCC = 0.9686 | SRCC = 0.9684 ] +INFO:root:Average loss=0.06697692891955376 +INFO:root: +[Valid][10][UTT][ MSE = 0.1925 | LCC = 0.9687 | SRCC = 0.9685 ] +INFO:root:Average loss=0.06370404425263405 +INFO:root: +[Valid][11][UTT][ MSE = 0.1742 | LCC = 0.9694 | SRCC = 0.9687 ] +INFO:root:Average loss=0.05918876087665558 +INFO:root: +[Valid][12][UTT][ MSE = 0.1951 | LCC = 0.9697 | SRCC = 0.9688 ] +INFO:root:Average loss=0.05945541340112686 +INFO:root: +[Valid][13][UTT][ MSE = 0.1910 | LCC = 0.9693 | SRCC = 0.9684 ] +INFO:root:Average loss=0.05626878309249878 +INFO:root: +[Valid][14][UTT][ MSE = 0.1710 | LCC = 0.9695 | SRCC = 0.9688 ] +INFO:root:Average loss=0.0542686902731657 +INFO:root: +[Valid][15][UTT][ MSE = 0.1703 | LCC = 0.9696 | SRCC = 0.9690 ] +INFO:root:Average loss=0.05080519795417786 +INFO:root: +[Valid][16][UTT][ MSE = 0.1561 | LCC = 0.9698 | SRCC = 0.9690 ] +INFO:root:Average loss=0.05014731362462044 +INFO:root: +[Valid][17][UTT][ MSE = 0.1793 | LCC = 0.9709 | SRCC = 0.9700 ] +INFO:root:Average loss=0.052243023946881295 +INFO:root: +[Valid][18][UTT][ MSE = 0.1632 | LCC = 0.9704 | SRCC = 0.9699 ] +INFO:root:Average loss=0.04469613282382488 +INFO:root: +[Valid][19][UTT][ MSE = 0.1605 | LCC = 0.9705 | SRCC = 0.9698 ] +INFO:root:Average loss=0.046004252329468726 +INFO:root: +[Valid][20][UTT][ MSE = 0.1677 | LCC = 0.9711 | SRCC = 0.9701 ] +INFO:root:Average loss=0.042027970299124716 +INFO:root: +[Valid][21][UTT][ MSE = 0.1533 | LCC = 0.9696 | SRCC = 0.9698 ] +INFO:root:Average loss=0.04018431308865547 +INFO:root: +[Valid][22][UTT][ MSE = 0.1861 | LCC = 0.9708 | SRCC = 0.9704 ] +INFO:root:Average loss=0.03950634928047657 +INFO:root: +[Valid][23][UTT][ MSE = 0.1852 | LCC = 0.9706 | SRCC = 0.9699 ] +INFO:root:Average loss=0.0397805889248848 +INFO:root: +[Valid][24][UTT][ MSE = 0.1805 | LCC = 0.9698 | SRCC = 0.9692 ] +INFO:root:Average loss=0.034630466982722284 +INFO:root: +[Valid][25][UTT][ MSE = 0.1392 | LCC = 0.9697 | SRCC = 0.9695 ] +INFO:root:Average loss=0.03501557832956314 +INFO:root: +[Valid][26][UTT][ MSE = 0.1667 | LCC = 0.9709 | SRCC = 0.9702 ] +INFO:root:Average loss=0.03272310933470726 +INFO:root: +[Valid][27][UTT][ MSE = 0.1608 | LCC = 0.9708 | SRCC = 0.9696 ] +INFO:root:Average loss=0.03139225581288338 +INFO:root: +[Valid][28][UTT][ MSE = 0.1513 | LCC = 0.9712 | SRCC = 0.9703 ] +INFO:root:Average loss=0.029679319255053998 +INFO:root: +[Valid][29][UTT][ MSE = 0.1618 | LCC = 0.9708 | SRCC = 0.9698 ] +INFO:root:Average loss=0.031831628903746606 +INFO:root: +[Valid][30][UTT][ MSE = 0.1548 | LCC = 0.9696 | SRCC = 0.9695 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1480 | LCC = 0.9716 | SRCC = 0.9714 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run1/config.gin b/results_tencent/w2v2_xlsr_1b/layer32/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run1/model.pt b/results_tencent/w2v2_xlsr_1b/layer32/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..170afb7526864840c6eb453123270ebfd08d7596 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d6f8cee794e028cceaa4395396eba7a9a1e7dae54546c2804535feb7ca15c52f +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run1/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer32/run1/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..49869f0f4e17aa33b8a686929f6f1bf947c5c06b --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run1/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8f20d302fd7cc32eda623a52b9a6d8497abd84891f3022df745ac80378307f9e +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run1/train.log b/results_tencent/w2v2_xlsr_1b/layer32/run1/train.log new file mode 100644 index 0000000000000000000000000000000000000000..bdc3f90d672669fe57ad3bb89738d8cd6a2f9001 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run1/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4370132430791855 +INFO:root: +[Valid][0][UTT][ MSE = 0.2812 | LCC = 0.9367 | SRCC = 0.9323 ] +INFO:root:Average loss=0.16108622789382934 +INFO:root: +[Valid][1][UTT][ MSE = 0.2345 | LCC = 0.9556 | SRCC = 0.9562 ] +INFO:root:Average loss=0.12668674051761628 +INFO:root: +[Valid][2][UTT][ MSE = 0.2048 | LCC = 0.9598 | SRCC = 0.9593 ] +INFO:root:Average loss=0.11061775133013725 +INFO:root: +[Valid][3][UTT][ MSE = 0.2105 | LCC = 0.9622 | SRCC = 0.9623 ] +INFO:root:Average loss=0.09958636555075645 +INFO:root: +[Valid][4][UTT][ MSE = 0.1952 | LCC = 0.9642 | SRCC = 0.9634 ] +INFO:root:Average loss=0.09555594629049301 +INFO:root: +[Valid][5][UTT][ MSE = 0.1956 | LCC = 0.9650 | SRCC = 0.9649 ] +INFO:root:Average loss=0.08744319227337838 +INFO:root: +[Valid][6][UTT][ MSE = 0.2205 | LCC = 0.9645 | SRCC = 0.9656 ] +INFO:root:Average loss=0.08104054376482964 +INFO:root: +[Valid][7][UTT][ MSE = 0.2184 | LCC = 0.9654 | SRCC = 0.9669 ] +INFO:root:Average loss=0.0780131612420082 +INFO:root: +[Valid][8][UTT][ MSE = 0.1819 | LCC = 0.9677 | SRCC = 0.9671 ] +INFO:root:Average loss=0.07073585525155067 +INFO:root: +[Valid][9][UTT][ MSE = 0.2081 | LCC = 0.9679 | SRCC = 0.9682 ] +INFO:root:Average loss=0.0705174660384655 +INFO:root: +[Valid][10][UTT][ MSE = 0.2116 | LCC = 0.9682 | SRCC = 0.9683 ] +INFO:root:Average loss=0.06603280752897263 +INFO:root: +[Valid][11][UTT][ MSE = 0.1663 | LCC = 0.9687 | SRCC = 0.9680 ] +INFO:root:Average loss=0.061892851740121844 +INFO:root: +[Valid][12][UTT][ MSE = 0.1714 | LCC = 0.9695 | SRCC = 0.9691 ] +INFO:root:Average loss=0.058534281462430955 +INFO:root: +[Valid][13][UTT][ MSE = 0.1615 | LCC = 0.9692 | SRCC = 0.9690 ] +INFO:root:Average loss=0.05633677351474762 +INFO:root: +[Valid][14][UTT][ MSE = 0.1484 | LCC = 0.9682 | SRCC = 0.9682 ] +INFO:root:Average loss=0.05302706292271614 +INFO:root: +[Valid][15][UTT][ MSE = 0.1751 | LCC = 0.9693 | SRCC = 0.9689 ] +INFO:root:Average loss=0.04987254509329796 +INFO:root: +[Valid][16][UTT][ MSE = 0.1575 | LCC = 0.9679 | SRCC = 0.9681 ] +INFO:root:Average loss=0.04680662177503109 +INFO:root: +[Valid][17][UTT][ MSE = 0.1639 | LCC = 0.9686 | SRCC = 0.9685 ] +INFO:root:Average loss=0.049415506675839424 +INFO:root: +[Valid][18][UTT][ MSE = 0.1990 | LCC = 0.9694 | SRCC = 0.9691 ] +INFO:root:Average loss=0.04571088869869709 +INFO:root: +[Valid][19][UTT][ MSE = 0.1564 | LCC = 0.9696 | SRCC = 0.9694 ] +INFO:root:Average loss=0.04371769207715988 +INFO:root: +[Valid][20][UTT][ MSE = 0.1715 | LCC = 0.9689 | SRCC = 0.9694 ] +INFO:root:Average loss=0.04400060349702835 +INFO:root: +[Valid][21][UTT][ MSE = 0.1538 | LCC = 0.9695 | SRCC = 0.9691 ] +INFO:root:Average loss=0.0401464142203331 +INFO:root: +[Valid][22][UTT][ MSE = 0.1847 | LCC = 0.9690 | SRCC = 0.9689 ] +INFO:root:Average loss=0.039343042597174645 +INFO:root: +[Valid][23][UTT][ MSE = 0.1566 | LCC = 0.9700 | SRCC = 0.9694 ] +INFO:root:Average loss=0.03953065429627895 +INFO:root: +[Valid][24][UTT][ MSE = 0.1970 | LCC = 0.9693 | SRCC = 0.9694 ] +INFO:root:Average loss=0.0366700514703989 +INFO:root: +[Valid][25][UTT][ MSE = 0.1862 | LCC = 0.9698 | SRCC = 0.9693 ] +INFO:root:Average loss=0.03562344501912594 +INFO:root: +[Valid][26][UTT][ MSE = 0.1878 | LCC = 0.9690 | SRCC = 0.9691 ] +INFO:root:Average loss=0.035513670310378075 +INFO:root: +[Valid][27][UTT][ MSE = 0.1545 | LCC = 0.9695 | SRCC = 0.9692 ] +INFO:root:Average loss=0.03367989644408226 +INFO:root: +[Valid][28][UTT][ MSE = 0.1479 | LCC = 0.9685 | SRCC = 0.9685 ] +INFO:root:Average loss=0.03286314396560192 +INFO:root: +[Valid][29][UTT][ MSE = 0.1652 | LCC = 0.9702 | SRCC = 0.9692 ] +INFO:root:Average loss=0.03171059745550155 +INFO:root: +[Valid][30][UTT][ MSE = 0.1891 | LCC = 0.9699 | SRCC = 0.9685 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1574 | LCC = 0.9718 | SRCC = 0.9713 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run2/config.gin b/results_tencent/w2v2_xlsr_1b/layer32/run2/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run2/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run2/model.pt b/results_tencent/w2v2_xlsr_1b/layer32/run2/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..67e2ce550051910274f69ca66246a3b97dcddc39 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run2/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cb1dea36cddb047b028d0ad49b6c69562cf79e45acbe7f98652ccf72ba938087 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run2/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer32/run2/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..d43025e994b3a896a324690812b34545d0cc80f2 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run2/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3527cb8488a4fa10b72f398c914bff5f154d6289f3c869fbf859886cdae5d239 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run2/train.log b/results_tencent/w2v2_xlsr_1b/layer32/run2/train.log new file mode 100644 index 0000000000000000000000000000000000000000..4bf4fefb1fce24a2861f02d78abbe1ba78b595bb --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run2/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4618257687687874 +INFO:root: +[Valid][0][UTT][ MSE = 0.2596 | LCC = 0.9442 | SRCC = 0.9416 ] +INFO:root:Average loss=0.16335248255729676 +INFO:root: +[Valid][1][UTT][ MSE = 0.2133 | LCC = 0.9542 | SRCC = 0.9527 ] +INFO:root:Average loss=0.1381802773475647 +INFO:root: +[Valid][2][UTT][ MSE = 0.2145 | LCC = 0.9595 | SRCC = 0.9582 ] +INFO:root:Average loss=0.11485369285941124 +INFO:root: +[Valid][3][UTT][ MSE = 0.2261 | LCC = 0.9619 | SRCC = 0.9631 ] +INFO:root:Average loss=0.10144479367136955 +INFO:root: +[Valid][4][UTT][ MSE = 0.1726 | LCC = 0.9661 | SRCC = 0.9653 ] +INFO:root:Average loss=0.09631563094258308 +INFO:root: +[Valid][5][UTT][ MSE = 0.1761 | LCC = 0.9662 | SRCC = 0.9658 ] +INFO:root:Average loss=0.08608627229928971 +INFO:root: +[Valid][6][UTT][ MSE = 0.1766 | LCC = 0.9665 | SRCC = 0.9655 ] +INFO:root:Average loss=0.0791087712943554 +INFO:root: +[Valid][7][UTT][ MSE = 0.1602 | LCC = 0.9664 | SRCC = 0.9653 ] +INFO:root:Average loss=0.0802452906370163 +INFO:root: +[Valid][8][UTT][ MSE = 0.1662 | LCC = 0.9682 | SRCC = 0.9670 ] +INFO:root:Average loss=0.07457903903722764 +INFO:root: +[Valid][9][UTT][ MSE = 0.1767 | LCC = 0.9681 | SRCC = 0.9668 ] +INFO:root:Average loss=0.06676888734102249 +INFO:root: +[Valid][10][UTT][ MSE = 0.1802 | LCC = 0.9679 | SRCC = 0.9678 ] +INFO:root:Average loss=0.06387111237645149 +INFO:root: +[Valid][11][UTT][ MSE = 0.1995 | LCC = 0.9692 | SRCC = 0.9683 ] +INFO:root:Average loss=0.06329297226667405 +INFO:root: +[Valid][12][UTT][ MSE = 0.1570 | LCC = 0.9690 | SRCC = 0.9684 ] +INFO:root:Average loss=0.061888046324253085 +INFO:root: +[Valid][13][UTT][ MSE = 0.1558 | LCC = 0.9681 | SRCC = 0.9679 ] +INFO:root:Average loss=0.05980470445752144 +INFO:root: +[Valid][14][UTT][ MSE = 0.1489 | LCC = 0.9695 | SRCC = 0.9684 ] +INFO:root:Average loss=0.0533517190515995 +INFO:root: +[Valid][15][UTT][ MSE = 0.2054 | LCC = 0.9689 | SRCC = 0.9683 ] +INFO:root:Average loss=0.05192219178378582 +INFO:root: +[Valid][16][UTT][ MSE = 0.1659 | LCC = 0.9704 | SRCC = 0.9697 ] +INFO:root:Average loss=0.050158100843429566 +INFO:root: +[Valid][17][UTT][ MSE = 0.2081 | LCC = 0.9687 | SRCC = 0.9684 ] +INFO:root:Average loss=0.04633115413784981 +INFO:root: +[Valid][18][UTT][ MSE = 0.1629 | LCC = 0.9699 | SRCC = 0.9687 ] +INFO:root:Average loss=0.045051252126693725 +INFO:root: +[Valid][19][UTT][ MSE = 0.1400 | LCC = 0.9701 | SRCC = 0.9696 ] +INFO:root:Average loss=0.0450109489262104 +INFO:root: +[Valid][20][UTT][ MSE = 0.1553 | LCC = 0.9703 | SRCC = 0.9686 ] +INFO:root:Average loss=0.04236969918012619 +INFO:root: +[Valid][21][UTT][ MSE = 0.1593 | LCC = 0.9704 | SRCC = 0.9693 ] +INFO:root:Average loss=0.04114753901958466 +INFO:root: +[Valid][22][UTT][ MSE = 0.1578 | LCC = 0.9702 | SRCC = 0.9689 ] +INFO:root:Average loss=0.03778087808191776 +INFO:root: +[Valid][23][UTT][ MSE = 0.1977 | LCC = 0.9698 | SRCC = 0.9689 ] +INFO:root:Average loss=0.03672968442738056 +INFO:root: +[Valid][24][UTT][ MSE = 0.1544 | LCC = 0.9693 | SRCC = 0.9687 ] +INFO:root:Average loss=0.03627919360995293 +INFO:root: +[Valid][25][UTT][ MSE = 0.1754 | LCC = 0.9708 | SRCC = 0.9694 ] +INFO:root:Average loss=0.03519171524047852 +INFO:root: +[Valid][26][UTT][ MSE = 0.1368 | LCC = 0.9693 | SRCC = 0.9685 ] +INFO:root:Average loss=0.033829599022865295 +INFO:root: +[Valid][27][UTT][ MSE = 0.1616 | LCC = 0.9704 | SRCC = 0.9690 ] +INFO:root:Average loss=0.0340585844963789 +INFO:root: +[Valid][28][UTT][ MSE = 0.1637 | LCC = 0.9691 | SRCC = 0.9680 ] +INFO:root:Average loss=0.03180890321731567 +INFO:root: +[Valid][29][UTT][ MSE = 0.1673 | LCC = 0.9703 | SRCC = 0.9689 ] +INFO:root:Average loss=0.03125601264834404 +INFO:root: +[Valid][30][UTT][ MSE = 0.1628 | LCC = 0.9700 | SRCC = 0.9689 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1704 | LCC = 0.9720 | SRCC = 0.9713 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run3/config.gin b/results_tencent/w2v2_xlsr_1b/layer32/run3/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run3/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run3/model.pt b/results_tencent/w2v2_xlsr_1b/layer32/run3/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..41f950ff727681214503c1c72b88f588230012b7 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run3/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b02a1e7ff8f326eb23fef5ca8c9dd4054cb07b739698a75092af83fd47bd6c1d +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run3/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer32/run3/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..f6279f8dbd62793c252fd39bc4d32e73d2e1ecf3 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run3/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ea8107155b092101b05072cfe67169683af69551975f21915146dd4f66d09244 +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run3/train.log b/results_tencent/w2v2_xlsr_1b/layer32/run3/train.log new file mode 100644 index 0000000000000000000000000000000000000000..03795948a6da4b6fba588ac9166bd9fe94493f94 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run3/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4822829564809799 +INFO:root: +[Valid][0][UTT][ MSE = 0.3573 | LCC = 0.9327 | SRCC = 0.9329 ] +INFO:root:Average loss=0.16125817447900773 +INFO:root: +[Valid][1][UTT][ MSE = 0.2407 | LCC = 0.9562 | SRCC = 0.9549 ] +INFO:root:Average loss=0.13148978912830353 +INFO:root: +[Valid][2][UTT][ MSE = 0.1913 | LCC = 0.9610 | SRCC = 0.9608 ] +INFO:root:Average loss=0.11388853770494461 +INFO:root: +[Valid][3][UTT][ MSE = 0.1560 | LCC = 0.9637 | SRCC = 0.9634 ] +INFO:root:Average loss=0.10407764282822608 +INFO:root: +[Valid][4][UTT][ MSE = 0.1827 | LCC = 0.9645 | SRCC = 0.9637 ] +INFO:root:Average loss=0.09650350213050843 +INFO:root: +[Valid][5][UTT][ MSE = 0.1990 | LCC = 0.9667 | SRCC = 0.9663 ] +INFO:root:Average loss=0.08965162843465806 +INFO:root: +[Valid][6][UTT][ MSE = 0.1910 | LCC = 0.9674 | SRCC = 0.9675 ] +INFO:root:Average loss=0.08131792277097702 +INFO:root: +[Valid][7][UTT][ MSE = 0.1884 | LCC = 0.9686 | SRCC = 0.9680 ] +INFO:root:Average loss=0.07712030732631683 +INFO:root: +[Valid][8][UTT][ MSE = 0.1811 | LCC = 0.9692 | SRCC = 0.9685 ] +INFO:root:Average loss=0.07263929945230484 +INFO:root: +[Valid][9][UTT][ MSE = 0.1999 | LCC = 0.9690 | SRCC = 0.9683 ] +INFO:root:Average loss=0.0686115083694458 +INFO:root: +[Valid][10][UTT][ MSE = 0.1786 | LCC = 0.9700 | SRCC = 0.9695 ] +INFO:root:Average loss=0.0641595167517662 +INFO:root: +[Valid][11][UTT][ MSE = 0.1758 | LCC = 0.9695 | SRCC = 0.9692 ] +INFO:root:Average loss=0.060556914687156675 +INFO:root: +[Valid][12][UTT][ MSE = 0.1742 | LCC = 0.9710 | SRCC = 0.9707 ] +INFO:root:Average loss=0.05865385575592518 +INFO:root: +[Valid][13][UTT][ MSE = 0.1624 | LCC = 0.9707 | SRCC = 0.9701 ] +INFO:root:Average loss=0.05962443363666534 +INFO:root: +[Valid][14][UTT][ MSE = 0.1834 | LCC = 0.9691 | SRCC = 0.9688 ] +INFO:root:Average loss=0.05431556542217732 +INFO:root: +[Valid][15][UTT][ MSE = 0.1878 | LCC = 0.9706 | SRCC = 0.9695 ] +INFO:root:Average loss=0.049739729344844816 +INFO:root: +[Valid][16][UTT][ MSE = 0.1590 | LCC = 0.9708 | SRCC = 0.9699 ] +INFO:root:Average loss=0.048920387491583825 +INFO:root: +[Valid][17][UTT][ MSE = 0.1427 | LCC = 0.9703 | SRCC = 0.9693 ] +INFO:root:Average loss=0.047523624420166014 +INFO:root: +[Valid][18][UTT][ MSE = 0.1660 | LCC = 0.9687 | SRCC = 0.9680 ] +INFO:root:Average loss=0.04385874357819557 +INFO:root: +[Valid][19][UTT][ MSE = 0.1872 | LCC = 0.9658 | SRCC = 0.9688 ] +INFO:root:Average loss=0.04563216626644134 +INFO:root: +[Valid][20][UTT][ MSE = 0.1682 | LCC = 0.9708 | SRCC = 0.9695 ] +INFO:root:Average loss=0.04002584792673588 +INFO:root: +[Valid][21][UTT][ MSE = 0.1769 | LCC = 0.9707 | SRCC = 0.9698 ] +INFO:root:Average loss=0.03975432524085045 +INFO:root: +[Valid][22][UTT][ MSE = 0.1593 | LCC = 0.9697 | SRCC = 0.9684 ] +INFO:root:Average loss=0.03810755568742752 +INFO:root: +[Valid][23][UTT][ MSE = 0.1903 | LCC = 0.9702 | SRCC = 0.9691 ] +INFO:root:Average loss=0.03606504026055336 +INFO:root: +[Valid][24][UTT][ MSE = 0.1622 | LCC = 0.9702 | SRCC = 0.9689 ] +INFO:root:Average loss=0.03578412413597107 +INFO:root: +[Valid][25][UTT][ MSE = 0.1580 | LCC = 0.9709 | SRCC = 0.9696 ] +INFO:root:Average loss=0.034860264003276825 +INFO:root: +[Valid][26][UTT][ MSE = 0.1601 | LCC = 0.9706 | SRCC = 0.9691 ] +INFO:root:Average loss=0.03153924612700939 +INFO:root: +[Valid][27][UTT][ MSE = 0.1983 | LCC = 0.9703 | SRCC = 0.9689 ] +INFO:root:Average loss=0.0332702267318964 +INFO:root: +[Valid][28][UTT][ MSE = 0.1926 | LCC = 0.9686 | SRCC = 0.9682 ] +INFO:root:Average loss=0.032725758790969846 +INFO:root: +[Valid][29][UTT][ MSE = 0.1625 | LCC = 0.9712 | SRCC = 0.9698 ] +INFO:root:Average loss=0.03092787103354931 +INFO:root: +[Valid][30][UTT][ MSE = 0.1723 | LCC = 0.9709 | SRCC = 0.9693 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1565 | LCC = 0.9722 | SRCC = 0.9718 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run4/config.gin b/results_tencent/w2v2_xlsr_1b/layer32/run4/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run4/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run4/model.pt b/results_tencent/w2v2_xlsr_1b/layer32/run4/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..c49f363ec36e6c9cc63abc3fac0eed934380a59e --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run4/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3cffae91a84d62462359c9b8a43fb15c036583058470c2b594b6046d65a6516b +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run4/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer32/run4/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..88412c62fdf10d4d5fc572ef2d362b52992f4904 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run4/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e865511e92bf1dc849cfc56f66d4d02987e3071ad0833ab2953b7cf76714bde +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer32/run4/train.log b/results_tencent/w2v2_xlsr_1b/layer32/run4/train.log new file mode 100644 index 0000000000000000000000000000000000000000..bc0fa7dad2c4f3e52d1ab05ce947a1edda659b40 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer32/run4/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4529640629291534 +INFO:root: +[Valid][0][UTT][ MSE = 0.2385 | LCC = 0.9393 | SRCC = 0.9387 ] +INFO:root:Average loss=0.15188526582717896 +INFO:root: +[Valid][1][UTT][ MSE = 0.1889 | LCC = 0.9536 | SRCC = 0.9527 ] +INFO:root:Average loss=0.11965578007698059 +INFO:root: +[Valid][2][UTT][ MSE = 0.2441 | LCC = 0.9590 | SRCC = 0.9593 ] +INFO:root:Average loss=0.10743962612748147 +INFO:root: +[Valid][3][UTT][ MSE = 0.2282 | LCC = 0.9613 | SRCC = 0.9610 ] +INFO:root:Average loss=0.09735984125733375 +INFO:root: +[Valid][4][UTT][ MSE = 0.1812 | LCC = 0.9637 | SRCC = 0.9632 ] +INFO:root:Average loss=0.08988163110613823 +INFO:root: +[Valid][5][UTT][ MSE = 0.1889 | LCC = 0.9637 | SRCC = 0.9635 ] +INFO:root:Average loss=0.08668176844716072 +INFO:root: +[Valid][6][UTT][ MSE = 0.2017 | LCC = 0.9653 | SRCC = 0.9650 ] +INFO:root:Average loss=0.08116007816791534 +INFO:root: +[Valid][7][UTT][ MSE = 0.2417 | LCC = 0.9641 | SRCC = 0.9634 ] +INFO:root:Average loss=0.08222381076216698 +INFO:root: +[Valid][8][UTT][ MSE = 0.2116 | LCC = 0.9658 | SRCC = 0.9653 ] +INFO:root:Average loss=0.07263830298185349 +INFO:root: +[Valid][9][UTT][ MSE = 0.1788 | LCC = 0.9667 | SRCC = 0.9660 ] +INFO:root:Average loss=0.0685463216304779 +INFO:root: +[Valid][10][UTT][ MSE = 0.1636 | LCC = 0.9674 | SRCC = 0.9663 ] +INFO:root:Average loss=0.06775912380218506 +INFO:root: +[Valid][11][UTT][ MSE = 0.1485 | LCC = 0.9672 | SRCC = 0.9667 ] +INFO:root:Average loss=0.06079507675766945 +INFO:root: +[Valid][12][UTT][ MSE = 0.2230 | LCC = 0.9661 | SRCC = 0.9661 ] +INFO:root:Average loss=0.059743433117866514 +INFO:root: +[Valid][13][UTT][ MSE = 0.1620 | LCC = 0.9661 | SRCC = 0.9652 ] +INFO:root:Average loss=0.05751631912589073 +INFO:root: +[Valid][14][UTT][ MSE = 0.1783 | LCC = 0.9680 | SRCC = 0.9670 ] +INFO:root:Average loss=0.052481439128518104 +INFO:root: +[Valid][15][UTT][ MSE = 0.1613 | LCC = 0.9681 | SRCC = 0.9669 ] +INFO:root:Average loss=0.053654935508966446 +INFO:root: +[Valid][16][UTT][ MSE = 0.1728 | LCC = 0.9678 | SRCC = 0.9674 ] +INFO:root:Average loss=0.048168851017951965 +INFO:root: +[Valid][17][UTT][ MSE = 0.1609 | LCC = 0.9685 | SRCC = 0.9676 ] +INFO:root:Average loss=0.04805943994224072 +INFO:root: +[Valid][18][UTT][ MSE = 0.2092 | LCC = 0.9677 | SRCC = 0.9675 ] +INFO:root:Average loss=0.04709889779984951 +INFO:root: +[Valid][19][UTT][ MSE = 0.1534 | LCC = 0.9678 | SRCC = 0.9671 ] +INFO:root:Average loss=0.04305770762264729 +INFO:root: +[Valid][20][UTT][ MSE = 0.1674 | LCC = 0.9680 | SRCC = 0.9669 ] +INFO:root:Average loss=0.04359705555438995 +INFO:root: +[Valid][21][UTT][ MSE = 0.1513 | LCC = 0.9683 | SRCC = 0.9674 ] +INFO:root:Average loss=0.040460486188530924 +INFO:root: +[Valid][22][UTT][ MSE = 0.1709 | LCC = 0.9676 | SRCC = 0.9673 ] +INFO:root:Average loss=0.03906215062737465 +INFO:root: +[Valid][23][UTT][ MSE = 0.1534 | LCC = 0.9681 | SRCC = 0.9671 ] +INFO:root:Average loss=0.038221860349178315 +INFO:root: +[Valid][24][UTT][ MSE = 0.1799 | LCC = 0.9678 | SRCC = 0.9671 ] +INFO:root:Average loss=0.03992365957796574 +INFO:root: +[Valid][25][UTT][ MSE = 0.1659 | LCC = 0.9684 | SRCC = 0.9672 ] +INFO:root:Average loss=0.03424187138676643 +INFO:root: +[Valid][26][UTT][ MSE = 0.1568 | LCC = 0.9694 | SRCC = 0.9681 ] +INFO:root:Average loss=0.03257403281331062 +INFO:root: +[Valid][27][UTT][ MSE = 0.1746 | LCC = 0.9680 | SRCC = 0.9672 ] +INFO:root:Average loss=0.032589599311351775 +INFO:root: +[Valid][28][UTT][ MSE = 0.1553 | LCC = 0.9683 | SRCC = 0.9669 ] +INFO:root:Average loss=0.030647655591368676 +INFO:root: +[Valid][29][UTT][ MSE = 0.1770 | LCC = 0.9686 | SRCC = 0.9676 ] +INFO:root:Average loss=0.03276904532313347 +INFO:root: +[Valid][30][UTT][ MSE = 0.1649 | LCC = 0.9685 | SRCC = 0.9668 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1492 | LCC = 0.9715 | SRCC = 0.9709 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer34/run0/config.gin b/results_tencent/w2v2_xlsr_1b/layer34/run0/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer34/run0/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer34/run0/model.pt b/results_tencent/w2v2_xlsr_1b/layer34/run0/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..8bd0962a81135cfebb8ef657b78972a06fb1384a --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer34/run0/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b52e00c3f1044d93169d5e7861f9fb419eddea6affec93ea6bef81e7a2baf289 +size 4058699 diff --git a/results_tencent/w2v2_xlsr_1b/layer34/run0/model_best.pt b/results_tencent/w2v2_xlsr_1b/layer34/run0/model_best.pt new file mode 100644 index 0000000000000000000000000000000000000000..b68653d2608eaac9fbefa8e7a70da0ed078e87c7 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer34/run0/model_best.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7668ca4df265e9865880ca21a808a0d1076d3f97839b053605c689fc38bbacdd +size 4060358 diff --git a/results_tencent/w2v2_xlsr_1b/layer34/run0/train.log b/results_tencent/w2v2_xlsr_1b/layer34/run0/train.log new file mode 100644 index 0000000000000000000000000000000000000000..7a76f1c8e948c151b2a13f048ee9fdbe830f4c16 --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer34/run0/train.log @@ -0,0 +1,99 @@ +INFO:root:Num train speech clips: 8000 +INFO:root:Num val speech clips: 2000 +INFO:root:Num test speech clips: 1563 +INFO:root:Device=cuda +INFO:root:Average loss=0.4065350062251091 +INFO:root: +[Valid][0][UTT][ MSE = 0.2306 | LCC = 0.9459 | SRCC = 0.9430 ] +INFO:root:Average loss=0.15101445960998536 +INFO:root: +[Valid][1][UTT][ MSE = 0.2359 | LCC = 0.9583 | SRCC = 0.9571 ] +INFO:root:Average loss=0.1320447953939438 +INFO:root: +[Valid][2][UTT][ MSE = 0.2230 | LCC = 0.9628 | SRCC = 0.9624 ] +INFO:root:Average loss=0.1083694152534008 +INFO:root: +[Valid][3][UTT][ MSE = 0.2132 | LCC = 0.9653 | SRCC = 0.9650 ] +INFO:root:Average loss=0.09857622465491295 +INFO:root: +[Valid][4][UTT][ MSE = 0.1907 | LCC = 0.9663 | SRCC = 0.9663 ] +INFO:root:Average loss=0.0923819639980793 +INFO:root: +[Valid][5][UTT][ MSE = 0.2237 | LCC = 0.9676 | SRCC = 0.9678 ] +INFO:root:Average loss=0.08952340137958527 +INFO:root: +[Valid][6][UTT][ MSE = 0.1930 | LCC = 0.9681 | SRCC = 0.9677 ] +INFO:root:Average loss=0.07941829389333725 +INFO:root: +[Valid][7][UTT][ MSE = 0.1708 | LCC = 0.9689 | SRCC = 0.9684 ] +INFO:root:Average loss=0.07929832696914672 +INFO:root: +[Valid][8][UTT][ MSE = 0.1989 | LCC = 0.9701 | SRCC = 0.9695 ] +INFO:root:Average loss=0.0708783985376358 +INFO:root: +[Valid][9][UTT][ MSE = 0.2232 | LCC = 0.9688 | SRCC = 0.9696 ] +INFO:root:Average loss=0.06654076138138772 +INFO:root: +[Valid][10][UTT][ MSE = 0.1923 | LCC = 0.9696 | SRCC = 0.9694 ] +INFO:root:Average loss=0.06937596201896667 +INFO:root: +[Valid][11][UTT][ MSE = 0.2134 | LCC = 0.9698 | SRCC = 0.9698 ] +INFO:root:Average loss=0.06538755974173546 +INFO:root: +[Valid][12][UTT][ MSE = 0.1883 | LCC = 0.9702 | SRCC = 0.9701 ] +INFO:root:Average loss=0.058827474474906924 +INFO:root: +[Valid][13][UTT][ MSE = 0.1682 | LCC = 0.9705 | SRCC = 0.9699 ] +INFO:root:Average loss=0.056451218649744984 +INFO:root: +[Valid][14][UTT][ MSE = 0.1778 | LCC = 0.9711 | SRCC = 0.9701 ] +INFO:root:Average loss=0.05417767243087292 +INFO:root: +[Valid][15][UTT][ MSE = 0.1813 | LCC = 0.9708 | SRCC = 0.9702 ] +INFO:root:Average loss=0.05219523486495018 +INFO:root: +[Valid][16][UTT][ MSE = 0.1660 | LCC = 0.9705 | SRCC = 0.9698 ] +INFO:root:Average loss=0.05196155297756195 +INFO:root: +[Valid][17][UTT][ MSE = 0.1514 | LCC = 0.9704 | SRCC = 0.9696 ] +INFO:root:Average loss=0.04732499957084656 +INFO:root: +[Valid][18][UTT][ MSE = 0.1759 | LCC = 0.9712 | SRCC = 0.9705 ] +INFO:root:Average loss=0.048098266646265986 +INFO:root: +[Valid][19][UTT][ MSE = 0.1753 | LCC = 0.9705 | SRCC = 0.9699 ] +INFO:root:Average loss=0.04560001254081726 +INFO:root: +[Valid][20][UTT][ MSE = 0.1656 | LCC = 0.9710 | SRCC = 0.9699 ] +INFO:root:Average loss=0.04276911361515522 +INFO:root: +[Valid][21][UTT][ MSE = 0.2090 | LCC = 0.9699 | SRCC = 0.9701 ] +INFO:root:Average loss=0.041299426391720774 +INFO:root: +[Valid][22][UTT][ MSE = 0.1858 | LCC = 0.9717 | SRCC = 0.9708 ] +INFO:root:Average loss=0.03818281401693821 +INFO:root: +[Valid][23][UTT][ MSE = 0.1673 | LCC = 0.9710 | SRCC = 0.9704 ] +INFO:root:Average loss=0.036192685037851335 +INFO:root: +[Valid][24][UTT][ MSE = 0.1628 | LCC = 0.9713 | SRCC = 0.9706 ] +INFO:root:Average loss=0.03579220198094845 +INFO:root: +[Valid][25][UTT][ MSE = 0.1736 | LCC = 0.9711 | SRCC = 0.9702 ] +INFO:root:Average loss=0.03438975633680821 +INFO:root: +[Valid][26][UTT][ MSE = 0.1810 | LCC = 0.9712 | SRCC = 0.9701 ] +INFO:root:Average loss=0.03414388693869114 +INFO:root: +[Valid][27][UTT][ MSE = 0.1675 | LCC = 0.9709 | SRCC = 0.9697 ] +INFO:root:Average loss=0.034106380954384806 +INFO:root: +[Valid][28][UTT][ MSE = 0.1490 | LCC = 0.9709 | SRCC = 0.9705 ] +INFO:root:Average loss=0.031076549619436265 +INFO:root: +[Valid][29][UTT][ MSE = 0.1644 | LCC = 0.9712 | SRCC = 0.9705 ] +INFO:root:Average loss=0.030553173065185547 +INFO:root: +[Valid][30][UTT][ MSE = 0.1583 | LCC = 0.9696 | SRCC = 0.9695 ] +INFO:root: +[Test][31][UTT][ MSE = 0.1810 | LCC = 0.9727 | SRCC = 0.9720 ] diff --git a/results_tencent/w2v2_xlsr_1b/layer34/run1/config.gin b/results_tencent/w2v2_xlsr_1b/layer34/run1/config.gin new file mode 100644 index 0000000000000000000000000000000000000000..cc46fceb0123d2b1295975663a63b17a6b94080d --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer34/run1/config.gin @@ -0,0 +1,31 @@ +# Training config on Tencent dataset. + +# ------------------------------ +# Architecture +# ------------------------------ +ProjectionHead: + conv_channels = (32, 32) + dense_neurons = (64, 32, 1) + use_poolings = (True, True) + kernel_size = 5 + pool_size = 5 + +# ------------------------------ +# Dataset +# ------------------------------ +Tencent: + debug = False + +get_dataloader: + num_workers = 12 + shuffle = True + +# ------------------------------ +# Training loop +# ------------------------------ +TrainingLoop: + num_epochs = 30 + learning_rate = 1e-4 + dataset_cls = @Tencent + weight_decay = 0.0 + batch_size_train = 64 diff --git a/results_tencent/w2v2_xlsr_1b/layer34/run1/model.pt b/results_tencent/w2v2_xlsr_1b/layer34/run1/model.pt new file mode 100644 index 0000000000000000000000000000000000000000..48819d9eed4d73bedcbc39dab8778b08a8d028bc --- /dev/null +++ b/results_tencent/w2v2_xlsr_1b/layer34/run1/model.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:446e0bfdc57330edcd260d4a01d47f1d78755c4cfd408497278473db96738fed +size 4058699