Add files using upload-large-folder tool

checkpoints/best.pt/meta.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:23edb7413547ac3374e364fab05a6336e6abc0155221eb070d0078b6bef23a5c
-size 13156046113
+oid sha256:880376b9ccc11cc82242b0ad291716175e8bee150a42280242bab4971e47a056
+size 13156046241

checkpoints/best.pt/model-00000.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:607f33fee8faa1bc61af3845f55942f837dfb24e31a5f417972d160e088c0843
+oid sha256:13daab6cdd98264855f5f992a2595f45d6cb65ff0b95904003bb198ff60a5b16
 size 1070291501

checkpoints/best.pt/model-00001.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b63b499c259e52e1188290d01414107db4cb5981dc34af6d9a20bce75b94651a
+oid sha256:9c6353f1db62922a47d1c3eeaff31983e74c0f5f5dfcc9a5d22daf299cdef8e6
 size 1021705255

checkpoints/best.pt/model-00002.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b83ae49b0dcafff75888e47585cd1e7b045465a0df50135951054ecc335dd457
+oid sha256:f27199ee07a99c48834eeaa6277508e2de424b11012954987ec715a2c999099f
 size 100752777

checkpoints/latest.pt/meta.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:39349834790de3db43d3d6fc9bed484ee65e035c8e694315d23e91bbdf69ba48
+oid sha256:ba34f1c13c03edd084294b0e7fdbb12581d7b25c96d58fc91740baa23b2050b8
 size 13156046241

checkpoints/latest.pt/model-00000.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:14e09140f479f0a9378518443021c3b2931742eb8be4eb4ae132f31b2475b4e2
+oid sha256:ae034082d7b98513188e8a2f3dd3b73c521bc6d830abdc8bd00b8b94ea906366
 size 1070291501

checkpoints/latest.pt/model-00001.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:495cb4f0704cd1ced4b4cbcd35c6bcf3a5b7e011626b985dccac92c6863ef94d
+oid sha256:16b1a1d41bed1723f814e3b9b3e8277edcb31cafe8daf86355ad9e622b32dc32
 size 1021705255

checkpoints/latest.pt/model-00002.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f94d7cc447cf75cf6deda342323d7b0aeb59a8c130cec1d617ee5eb28b5f7c6e
+oid sha256:a78617c40cca3c6258a0e9fb1bff0b773171c05beff04ca6cb6ccdda071a87b2
 size 100752777

config.yaml

@@ -89,6 +89,7 @@ model:
     hf_repo: frankie137/sd_asr_synthesis_data
     split: train
     data_files:
+    - stage3/part-*.parquet
     - stage2/part-*.parquet
     - stage1/part-*.parquet
   validation_ds:
@@ -163,7 +164,7 @@ trainer:
   limit_val_batches: 1
   prestart_validation_loader: true
   prestart_validation_loader_lead_steps: 100
-  max_steps: 1000
+  max_steps: 1500
   limit_realdata_val_batches: 90
 optimizer:
   name: adamw

metrics.jsonl

@@ -127,3 +127,73 @@
 {"val_der_realdata": 0.6255126058549703, "val_der_fa_realdata": 0.25233707539366734, "val_der_miss_realdata": 0.3174154698714251, "val_der_confusion_realdata": 0.055760060589877765, "step": 900}
 {"step": 910, "loss": 0.15686197578907013, "vad_loss": 0.15686197578907013, "vad_sort_loss": 0.1686837375164032, "vad_pil_loss": 0.14504024386405945, "learning_rate": 2.029994902841281e-05}
 {"step": 920, "loss": 0.1857648640871048, "vad_loss": 0.1857648640871048, "vad_sort_loss": 0.1999492645263672, "vad_pil_loss": 0.1715804785490036, "learning_rate": 2.018932173086796e-05}
+{"step": 910, "loss": 0.1768532693386078, "vad_loss": 0.1768532693386078, "vad_sort_loss": 0.18804527819156647, "vad_pil_loss": 0.16566124558448792, "learning_rate": 2.029994902841281e-05}
+{"step": 920, "loss": 0.15799546241760254, "vad_loss": 0.15799546241760254, "vad_sort_loss": 0.16256166994571686, "vad_pil_loss": 0.15342926979064941, "learning_rate": 2.018932173086796e-05}
+{"step": 930, "loss": 0.14951203763484955, "vad_loss": 0.14951203763484955, "vad_sort_loss": 0.15647055208683014, "vad_pil_loss": 0.14255350828170776, "learning_rate": 2.0080482499906793e-05}
+{"step": 940, "loss": 0.20427043735980988, "vad_loss": 0.20427043735980988, "vad_sort_loss": 0.21588988602161407, "vad_pil_loss": 0.19265098869800568, "learning_rate": 1.9973385860794224e-05}
+{"step": 950, "loss": 0.18792028725147247, "vad_loss": 0.18792028725147247, "vad_sort_loss": 0.19753582775592804, "vad_pil_loss": 0.17830471694469452, "learning_rate": 1.986798451980576e-05}
+{"val_loss": 0.16475747525691986, "val_vad_loss": 0.16475747525691986, "val_vad_sort_loss": 0.1737939566373825, "val_vad_pil_loss": 0.15572094917297363, "val_der": 0.2952670038836817, "val_der_fa": 0.11918404038420462, "val_der_miss": 0.13843685111131143, "val_der_confusion": 0.037646112388165676, "step": 950}
+{"val_der_realdata": 0.6258351577746232, "val_der_fa_realdata": 0.29394361980972966, "val_der_miss_realdata": 0.2580736611869108, "val_der_confusion_realdata": 0.0738178767779828, "step": 950}
+{"step": 960, "loss": 0.1656496524810791, "vad_loss": 0.1656496524810791, "vad_sort_loss": 0.1716810017824173, "vad_pil_loss": 0.1596183329820633, "learning_rate": 1.9764234821195714e-05}
+{"step": 970, "loss": 0.17191994190216064, "vad_loss": 0.17191994190216064, "vad_sort_loss": 0.1819942444562912, "vad_pil_loss": 0.1618456393480301, "learning_rate": 1.9662093109218404e-05}
+{"step": 980, "loss": 0.1683615744113922, "vad_loss": 0.1683615744113922, "vad_sort_loss": 0.1861524134874344, "vad_pil_loss": 0.15057073533535004, "learning_rate": 1.956151936610695e-05}
+{"step": 990, "loss": 0.1464417278766632, "vad_loss": 0.1464417278766632, "vad_sort_loss": 0.15215322375297546, "vad_pil_loss": 0.14073023200035095, "learning_rate": 1.9462473574094474e-05}
+{"step": 1000, "loss": 0.1480046510696411, "vad_loss": 0.1480046510696411, "vad_sort_loss": 0.15744248032569885, "vad_pil_loss": 0.13856680691242218, "learning_rate": 1.9364917534403503e-05}
+{"val_loss": 0.14490357041358948, "val_vad_loss": 0.14490357041358948, "val_vad_sort_loss": 0.14490357041358948, "val_vad_pil_loss": 0.14490358531475067, "val_der": 0.2789508304616728, "val_der_fa": 0.12850962194628596, "val_der_miss": 0.13701867863891115, "val_der_confusion": 0.013422529876475681, "step": 1000}
+{"val_der_realdata": 0.6417258545321416, "val_der_fa_realdata": 0.2941713924824327, "val_der_miss_realdata": 0.26817062689506344, "val_der_confusion_realdata": 0.07938383515464552, "step": 1000}
+{"step": 1010, "loss": 0.159490704536438, "vad_loss": 0.159490704536438, "vad_sort_loss": 0.1738385111093521, "vad_pil_loss": 0.14514292776584625, "learning_rate": 1.926881304825656e-05}
+{"step": 1020, "loss": 0.16767293214797974, "vad_loss": 0.16767293214797974, "vad_sort_loss": 0.17592796683311462, "vad_pil_loss": 0.15941788256168365, "learning_rate": 1.917412555485498e-05}
+{"step": 1030, "loss": 0.16871993243694305, "vad_loss": 0.16871993243694305, "vad_sort_loss": 0.17756718397140503, "vad_pil_loss": 0.15987271070480347, "learning_rate": 1.9080818674410693e-05}
+{"step": 1040, "loss": 0.1879580020904541, "vad_loss": 0.1879580020904541, "vad_sort_loss": 0.20310527086257935, "vad_pil_loss": 0.17281073331832886, "learning_rate": 1.898886330309324e-05}
+{"step": 1050, "loss": 0.1899852752685547, "vad_loss": 0.1899852752685547, "vad_sort_loss": 0.2028932124376297, "vad_pil_loss": 0.1770773082971573, "learning_rate": 1.889822306111455e-05}
+{"val_loss": 0.17757514119148254, "val_vad_loss": 0.17757514119148254, "val_vad_sort_loss": 0.194960355758667, "val_vad_pil_loss": 0.1601899266242981, "val_der": 0.31515012380934976, "val_der_fa": 0.13024294628568892, "val_der_miss": 0.1418032361187384, "val_der_confusion": 0.043103941404922456, "step": 1050}
+{"val_der_realdata": 0.6265213636946787, "val_der_fa_realdata": 0.30428277814341936, "val_der_miss_realdata": 0.2489149572199374, "val_der_confusion_realdata": 0.07332362833132201, "step": 1050}
+{"step": 1060, "loss": 0.17263934016227722, "vad_loss": 0.17263934016227722, "vad_sort_loss": 0.18624550104141235, "vad_pil_loss": 0.1590331792831421, "learning_rate": 1.880887066363357e-05}
+{"step": 1070, "loss": 0.16003261506557465, "vad_loss": 0.16003261506557465, "vad_sort_loss": 0.16789495944976807, "vad_pil_loss": 0.15217027068138123, "learning_rate": 1.8720771549851634e-05}
+{"step": 1080, "loss": 0.15407781302928925, "vad_loss": 0.15407781302928925, "vad_sort_loss": 0.1577123999595642, "vad_pil_loss": 0.1504432111978531, "learning_rate": 1.863390025391709e-05}
+{"step": 1090, "loss": 0.14221931993961334, "vad_loss": 0.14221931993961334, "vad_sort_loss": 0.1511334329843521, "vad_pil_loss": 0.13330519199371338, "learning_rate": 1.854822585301008e-05}
+{"step": 1100, "loss": 0.1562451422214508, "vad_loss": 0.1562451422214508, "vad_sort_loss": 0.16454774141311646, "vad_pil_loss": 0.14794255793094635, "learning_rate": 1.8463722881278954e-05}
+{"val_loss": 0.1264890730381012, "val_vad_loss": 0.1264890730381012, "val_vad_sort_loss": 0.1264890730381012, "val_vad_pil_loss": 0.1264890730381012, "val_der": 0.2220519114448172, "val_der_fa": 0.09511803790443288, "val_der_miss": 0.1159322689719107, "val_der_confusion": 0.01100160456847364, "step": 1100}
+{"val_der_realdata": 0.6068951103582719, "val_der_fa_realdata": 0.2949649592538548, "val_der_miss_realdata": 0.24321550345713805, "val_der_confusion_realdata": 0.06871464764727908, "step": 1100}
+{"step": 1110, "loss": 0.18354517221450806, "vad_loss": 0.18354517221450806, "vad_sort_loss": 0.20158322155475616, "vad_pil_loss": 0.16550712287425995, "learning_rate": 1.8380365872872062e-05}
+{"step": 1120, "loss": 0.15723368525505066, "vad_loss": 0.15723368525505066, "vad_sort_loss": 0.1677861511707306, "vad_pil_loss": 0.14668121933937073, "learning_rate": 1.8298125723958947e-05}
+{"step": 1130, "loss": 0.1577104777097702, "vad_loss": 0.1577104777097702, "vad_sort_loss": 0.16748164594173431, "vad_pil_loss": 0.14793933928012848, "learning_rate": 1.8216980606666766e-05}
+{"step": 1140, "loss": 0.18687672913074493, "vad_loss": 0.18687672913074493, "vad_sort_loss": 0.20086392760276794, "vad_pil_loss": 0.17288954555988312, "learning_rate": 1.8136906874133274e-05}
+{"step": 1150, "loss": 0.191878542304039, "vad_loss": 0.191878542304039, "vad_sort_loss": 0.20883500576019287, "vad_pil_loss": 0.17492206394672394, "learning_rate": 1.8057877241517417e-05}
+{"val_loss": 0.15062913298606873, "val_vad_loss": 0.15062913298606873, "val_vad_sort_loss": 0.17134162783622742, "val_vad_pil_loss": 0.12991666793823242, "val_der": 0.19911757613853148, "val_der_fa": 0.0832855804009725, "val_der_miss": 0.09742436464483054, "val_der_confusion": 0.018407631092728444, "step": 1150}
+{"val_der_realdata": 0.5809036929681739, "val_der_fa_realdata": 0.2816984983431396, "val_der_miss_realdata": 0.2373024101399648, "val_der_confusion_realdata": 0.06190278448506948, "step": 1150}
+{"step": 1160, "loss": 0.1571442037820816, "vad_loss": 0.1571442037820816, "vad_sort_loss": 0.166422501206398, "vad_pil_loss": 0.147865891456604, "learning_rate": 1.797987351892516e-05}
+{"step": 1170, "loss": 0.15289995074272156, "vad_loss": 0.15289995074272156, "vad_sort_loss": 0.15760695934295654, "vad_pil_loss": 0.14819291234016418, "learning_rate": 1.7902872059494257e-05}
+{"step": 1180, "loss": 0.17021045088768005, "vad_loss": 0.17021045088768005, "vad_sort_loss": 0.17693915963172913, "vad_pil_loss": 0.16348178684711456, "learning_rate": 1.7826851035351865e-05}
+{"step": 1190, "loss": 0.16504070162773132, "vad_loss": 0.16504070162773132, "vad_sort_loss": 0.17688311636447906, "vad_pil_loss": 0.15319828689098358, "learning_rate": 1.7751790437614545e-05}
+{"step": 1200, "loss": 0.15345904231071472, "vad_loss": 0.15345904231071472, "vad_sort_loss": 0.16109059751033783, "vad_pil_loss": 0.14582745730876923, "learning_rate": 1.767767025739886e-05}
+{"val_loss": 0.15524864196777344, "val_vad_loss": 0.15524864196777344, "val_vad_sort_loss": 0.17386949062347412, "val_vad_pil_loss": 0.13662780821323395, "val_der": 0.23001660274156333, "val_der_fa": 0.10011746172166035, "val_der_miss": 0.09889982752112969, "val_der_confusion": 0.030999313498773273, "step": 1200}
+{"val_der_realdata": 0.6025379638505081, "val_der_fa_realdata": 0.2838383880526649, "val_der_miss_realdata": 0.24973259201390158, "val_der_confusion_realdata": 0.06896698378394164, "step": 1200}
+{"step": 1210, "loss": 0.15837037563323975, "vad_loss": 0.15837037563323975, "vad_sort_loss": 0.16546565294265747, "vad_pil_loss": 0.15127508342266083, "learning_rate": 1.7604470485821366e-05}
+{"step": 1220, "loss": 0.15449555218219757, "vad_loss": 0.15449555218219757, "vad_sort_loss": 0.16281113028526306, "vad_pil_loss": 0.14618001878261566, "learning_rate": 1.753217111399863e-05}
+{"step": 1230, "loss": 0.18104544281959534, "vad_loss": 0.18104544281959534, "vad_sort_loss": 0.1897016167640686, "vad_pil_loss": 0.17238925397396088, "learning_rate": 1.746075759001542e-05}
+{"step": 1240, "loss": 0.1767193228006363, "vad_loss": 0.1767193228006363, "vad_sort_loss": 0.18933866918087006, "vad_pil_loss": 0.16409999132156372, "learning_rate": 1.7390208085998893e-05}
+{"step": 1250, "loss": 0.17970842123031616, "vad_loss": 0.17970842123031616, "vad_sort_loss": 0.19745320081710815, "vad_pil_loss": 0.16196362674236298, "learning_rate": 1.7320508050033823e-05}
+{"val_loss": 0.12551969289779663, "val_vad_loss": 0.12551969289779663, "val_vad_sort_loss": 0.14298784732818604, "val_vad_pil_loss": 0.10805155336856842, "val_der": 0.14763351013867468, "val_der_fa": 0.06599529934531123, "val_der_miss": 0.07029280642365976, "val_der_confusion": 0.011345404369703698, "step": 1250}
+{"val_der_realdata": 0.5905341679869746, "val_der_fa_realdata": 0.2689866189095908, "val_der_miss_realdata": 0.24827946233654344, "val_der_confusion_realdata": 0.07326808674084033, "step": 1250}
+{"step": 1260, "loss": 0.1531403362751007, "vad_loss": 0.1531403362751007, "vad_sort_loss": 0.1632206290960312, "vad_pil_loss": 0.14306005835533142, "learning_rate": 1.7251639292226173e-05}
+{"step": 1270, "loss": 0.19279009103775024, "vad_loss": 0.19279009103775024, "vad_sort_loss": 0.20218788087368011, "vad_pil_loss": 0.18339231610298157, "learning_rate": 1.7183585441671312e-05}
+{"step": 1280, "loss": 0.16555745899677277, "vad_loss": 0.16555745899677277, "vad_sort_loss": 0.17718982696533203, "vad_pil_loss": 0.1539251059293747, "learning_rate": 1.7116330127464607e-05}
+{"step": 1290, "loss": 0.18122223019599915, "vad_loss": 0.18122223019599915, "vad_sort_loss": 0.1955524981021881, "vad_pil_loss": 0.1668919324874878, "learning_rate": 1.704985879769083e-05}
+{"step": 1300, "loss": 0.17036831378936768, "vad_loss": 0.17036831378936768, "vad_sort_loss": 0.18482670187950134, "vad_pil_loss": 0.15590991079807281, "learning_rate": 1.698415508144535e-05}
+{"val_loss": 0.12693104147911072, "val_vad_loss": 0.12693104147911072, "val_vad_sort_loss": 0.14517895877361298, "val_vad_pil_loss": 0.10868310183286667, "val_der": 0.15993868246431886, "val_der_fa": 0.06798647421870621, "val_der_miss": 0.08302772986698047, "val_der_confusion": 0.00892447837863217, "step": 1300}
+{"val_der_realdata": 0.5914508584695971, "val_der_fa_realdata": 0.2732741167844079, "val_der_miss_realdata": 0.25687234612893367, "val_der_confusion_realdata": 0.061304395556255546, "step": 1300}
+{"step": 1310, "loss": 0.1725062131881714, "vad_loss": 0.1725062131881714, "vad_sort_loss": 0.1805669665336609, "vad_pil_loss": 0.16444547474384308, "learning_rate": 1.6919206245802343e-05}
+{"step": 1320, "loss": 0.20368723571300507, "vad_loss": 0.20368723571300507, "vad_sort_loss": 0.2164265513420105, "vad_pil_loss": 0.19094789028167725, "learning_rate": 1.6854995919857174e-05}
+{"step": 1330, "loss": 0.17482586205005646, "vad_loss": 0.17482586205005646, "vad_sort_loss": 0.1908734291791916, "vad_pil_loss": 0.15877828001976013, "learning_rate": 1.6791513189673424e-05}
+{"step": 1340, "loss": 0.15764759480953217, "vad_loss": 0.15764759480953217, "vad_sort_loss": 0.16924405097961426, "vad_pil_loss": 0.14605115354061127, "learning_rate": 1.6728739865357056e-05}
+{"step": 1350, "loss": 0.17926053702831268, "vad_loss": 0.17926053702831268, "vad_sort_loss": 0.1922888308763504, "vad_pil_loss": 0.16623225808143616, "learning_rate": 1.6666666851961054e-05}
+{"val_loss": 0.16108788549900055, "val_vad_loss": 0.16108788549900055, "val_vad_sort_loss": 0.16450363397598267, "val_vad_pil_loss": 0.15767215192317963, "val_der": 0.27419492501448983, "val_der_fa": 0.10210863659505534, "val_der_miss": 0.14727537680735814, "val_der_confusion": 0.024810911612076336, "step": 1350}
+{"val_der_realdata": 0.6293936983502322, "val_der_fa_realdata": 0.2760847298696937, "val_der_miss_realdata": 0.2818200118697333, "val_der_confusion_realdata": 0.07148895661080522, "step": 1350}
+{"step": 1360, "loss": 0.1676727682352066, "vad_loss": 0.1676727682352066, "vad_sort_loss": 0.18301410973072052, "vad_pil_loss": 0.1523314267396927, "learning_rate": 1.660527959757019e-05}
+{"step": 1370, "loss": 0.14743298292160034, "vad_loss": 0.14743298292160034, "vad_sort_loss": 0.15492013096809387, "vad_pil_loss": 0.13994581997394562, "learning_rate": 1.6544565369258635e-05}
+{"step": 1380, "loss": 0.16652429103851318, "vad_loss": 0.16652429103851318, "vad_sort_loss": 0.1759374439716339, "vad_pil_loss": 0.15711112320423126, "learning_rate": 1.6484511434100568e-05}
+{"step": 1390, "loss": 0.17420969903469086, "vad_loss": 0.17420969903469086, "vad_sort_loss": 0.1781802624464035, "vad_pil_loss": 0.17023912072181702, "learning_rate": 1.642510869714897e-05}
+{"step": 1400, "loss": 0.16158883273601532, "vad_loss": 0.16158883273601532, "vad_sort_loss": 0.17560699582099915, "vad_pil_loss": 0.1475706696510315, "learning_rate": 1.6366342606488615e-05}
+{"val_loss": 0.1594265252351761, "val_vad_loss": 0.1594265252351761, "val_vad_sort_loss": 0.17758885025978088, "val_vad_pil_loss": 0.1412642002105713, "val_der": 0.24236476131300774, "val_der_fa": 0.11170638582754754, "val_der_miss": 0.1153735946364466, "val_der_confusion": 0.015284780849013621, "step": 1400}
+{"val_der_realdata": 0.6115668621661116, "val_der_fa_realdata": 0.28530189592392524, "val_der_miss_realdata": 0.25814469184065353, "val_der_confusion_realdata": 0.06812027440153284, "step": 1400}

tensorboard/events.out.tfevents.1782224397.a22d6b5e36ba.34464.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fed3dc672a4dff5838b78b7ef27f3690e76275ce97462e5bac0ad8e174601589
+size 88

tensorboard/events.out.tfevents.1782224552.a22d6b5e36ba.36383.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc3603611d545b174a7ccd1a4750dcaa559651c18cfb6aeb2e1be2ff09ef857a
+size 20328