Comments (3)
coreeey
commented on July 22, 2024
Hi, I was wondering if you were able to reproduce the results and how? I had a similar PESQ of 2.2. I set training for 55 epochs and evaluated the model saved at the last epoch. I didn't modify the code other than changing the return/input of the STFT and iSTFT.
I'm sorry, I haven't reproduced the results of the article here either. I guess the torch version may have caused some differences in feature extraction, but due to time constraints, I'm not prepared to continue trying
from cmgan.
coreeey
commented on July 22, 2024
Hi, I was wondering if you were able to reproduce the results and how? I had a similar PESQ of 2.2. I set training for 55 epochs and evaluated the model saved at the last epoch. I didn't modify the code other than changing the return/input of the STFT and iSTFT.
I'm sorry, I haven't reproduced the results of the article here either. I guess the torch version may have caused some differences in feature extraction, but due to time constraints, I'm not prepared to continue trying
Thanks for the reply. I did get up to ~3.0 PESQ but couldn't get to 3.41. I wonder what others used in the other threads about reproducing the results, they seemed to be able to get ~3.2... Anyway I may try another repo
If it reaches 3.0, the network may converge normally. How many epochs did you get this result? Can you share your loss curve in tensorboard? I'm still quite curious about this
from cmgan.
coreeey
commented on July 22, 2024
So I ended up looking at this again and I got it to 3.28 PESQ, similar to this thread.
I used Sherif's version of this repository, here. I didn't do a deep compare between the repos but the biggest thing I noticed is the training dataloader was shuffled in that version whereas it isn't here. I didn't try this repo with shuffling on but I'm assuming that's the major difference.
Anyways here are my hyperparameters: epochs=55, batch_size=4, log_interval=500, decay_epoch=12, init_lr=0.0005, cut_len=32000 Loss weights: 0.3, 0.7, 1, 0.01
I evaluated at 3 different epochs (31, 52, 54) and the best result came from the very last epoch, even though the generator loss was not the lowest. So it's possible the results could have improved if it was trained longer.
Here is the training log:
INFO:root:Epoch 0, Step 500, loss: 0.312759131193161, disc_loss: 0.014180301688611507 INFO:root:Epoch 0, Step 1000, loss: 0.3000572919845581, disc_loss: 0.03669717535376549 INFO:root:Epoch 0, Step 1500, loss: 0.3671327233314514, disc_loss: 0.02478821389377117 INFO:root:Epoch 0, Step 2000, loss: 0.1893772929906845, disc_loss: 0.015164515003561974 INFO:root:Epoch 0, Step 2500, loss: 0.2994377017021179, disc_loss: 0.017490889877080917 INFO:root:Generator loss: 0.1954199135375833, Discriminator loss: 0.010855219110188717 INFO:root:Epoch 1, Step 500, loss: 0.3214994966983795, disc_loss: 0.0008018735097721219 INFO:root:Epoch 1, Step 1000, loss: 0.25855952501296997, disc_loss: 0.007563697639852762 INFO:root:Epoch 1, Step 1500, loss: 0.3672272861003876, disc_loss: 0.008135653100907803 INFO:root:Epoch 1, Step 2000, loss: 0.2542274296283722, disc_loss: 0.006575008388608694 INFO:root:Epoch 1, Step 2500, loss: 0.3384028375148773, disc_loss: 0.03157956898212433 INFO:root:Generator loss: 0.1993667078249663, Discriminator loss: 0.007349369336878632 INFO:root:Epoch 2, Step 500, loss: 0.2520577609539032, disc_loss: 0.011542283929884434 INFO:root:Epoch 2, Step 1000, loss: 0.283456027507782, disc_loss: 0.005646033212542534 INFO:root:Epoch 2, Step 1500, loss: 0.24863296747207642, disc_loss: 0.006484236102551222 INFO:root:Epoch 2, Step 2000, loss: 0.2130897492170334, disc_loss: 0.0032469583675265312 INFO:root:Epoch 2, Step 2500, loss: 0.24691542983055115, disc_loss: 0.023507077246904373 INFO:root:Generator loss: 0.18021206854327212, Discriminator loss: 0.007001019695839275 INFO:root:Epoch 3, Step 500, loss: 0.36660319566726685, disc_loss: 0.0022028894163668156 INFO:root:Epoch 3, Step 1000, loss: 0.2732304632663727, disc_loss: 0.0022656084038317204 INFO:root:Epoch 3, Step 1500, loss: 0.2375778704881668, disc_loss: 0.0063988384790718555 INFO:root:Epoch 3, Step 2000, loss: 0.2714841067790985, disc_loss: 0.007243407424539328 INFO:root:Epoch 3, Step 2500, loss: 0.20046576857566833, disc_loss: 0.0051153902895748615 INFO:root:Generator loss: 0.17687235407314253, Discriminator loss: 0.005908454136410954 INFO:root:Epoch 4, Step 500, loss: 0.26929739117622375, disc_loss: 0.002962844679132104 INFO:root:Epoch 4, Step 1000, loss: 0.18813900649547577, disc_loss: 0.0035468311980366707 INFO:root:Epoch 4, Step 1500, loss: 0.34768155217170715, disc_loss: 0.002427500206977129 INFO:root:Epoch 4, Step 2000, loss: 0.2553648054599762, disc_loss: 0.010068511590361595 INFO:root:Epoch 4, Step 2500, loss: 0.23005756735801697, disc_loss: 0.0028513793367892504 INFO:root:Generator loss: 0.16993942122580935, Discriminator loss: 0.0087586595099425 INFO:root:Epoch 5, Step 500, loss: 0.30520328879356384, disc_loss: 0.00796982366591692 INFO:root:Epoch 5, Step 1000, loss: 0.26169344782829285, disc_loss: 0.0033064233139157295 INFO:root:Epoch 5, Step 1500, loss: 0.2837878167629242, disc_loss: 0.0033368244767189026 INFO:root:Epoch 5, Step 2000, loss: 0.3022649884223938, disc_loss: 0.0015388906467705965 INFO:root:Epoch 5, Step 2500, loss: 0.19645152986049652, disc_loss: 0.0026794704608619213 INFO:root:Generator loss: 0.17689822230813573, Discriminator loss: 0.01272232923427121 INFO:root:Epoch 6, Step 500, loss: 0.29179835319519043, disc_loss: 0.0010662907734513283 INFO:root:Epoch 6, Step 1000, loss: 0.2581096589565277, disc_loss: 0.004199117422103882 INFO:root:Epoch 6, Step 1500, loss: 0.297472208738327, disc_loss: 0.0029484110418707132 INFO:root:Epoch 6, Step 2000, loss: 0.14254534244537354, disc_loss: 0.0005551559734158218 INFO:root:Epoch 6, Step 2500, loss: 0.23787181079387665, disc_loss: 0.001854499219916761 INFO:root:Generator loss: 0.1655087476071802, Discriminator loss: 0.005884178486614244 INFO:root:Epoch 7, Step 500, loss: 0.2807033061981201, disc_loss: 0.003636919194832444 INFO:root:Epoch 7, Step 1000, loss: 0.2307279258966446, disc_loss: 0.0006609580595977604 INFO:root:Epoch 7, Step 1500, loss: 0.17860105633735657, disc_loss: 0.0018876992398872972 INFO:root:Epoch 7, Step 2000, loss: 0.17049115896224976, disc_loss: 0.003694066544994712 INFO:root:Epoch 7, Step 2500, loss: 0.19690778851509094, disc_loss: 0.004949377384036779 INFO:root:Generator loss: 0.16698780728196636, Discriminator loss: 0.007069916313765452 INFO:root:Epoch 8, Step 500, loss: 0.23593047261238098, disc_loss: 0.007670776918530464 INFO:root:Epoch 8, Step 1000, loss: 0.2838055491447449, disc_loss: 0.002970515750348568 INFO:root:Epoch 8, Step 1500, loss: 0.14858031272888184, disc_loss: 0.00419649388641119 INFO:root:Epoch 8, Step 2000, loss: 0.18173342943191528, disc_loss: 0.006143204402178526 INFO:root:Epoch 8, Step 2500, loss: 0.27629354596138, disc_loss: 0.005271576810628176 INFO:root:Generator loss: 0.16036879802936488, Discriminator loss: 0.0040861366057284715 INFO:root:Epoch 9, Step 500, loss: 0.2189137190580368, disc_loss: 0.0057810521684587 INFO:root:Epoch 9, Step 1000, loss: 0.21471352875232697, disc_loss: 0.004761099815368652 INFO:root:Epoch 9, Step 1500, loss: 0.1557544767856598, disc_loss: 0.00562161672860384 INFO:root:Epoch 9, Step 2000, loss: 0.24498240649700165, disc_loss: 0.0012258212082087994 INFO:root:Epoch 9, Step 2500, loss: 0.23638907074928284, disc_loss: 0.003574251662939787 INFO:root:Generator loss: 0.1707929576050888, Discriminator loss: 0.010396833450887216 INFO:root:Epoch 10, Step 500, loss: 0.24539637565612793, disc_loss: 0.0010290932841598988 INFO:root:Epoch 10, Step 1000, loss: 0.29992005228996277, disc_loss: 0.01089942455291748 INFO:root:Epoch 10, Step 1500, loss: 0.20314528048038483, disc_loss: 0.0011890940368175507 INFO:root:Epoch 10, Step 2000, loss: 0.214199498295784, disc_loss: 0.00865689292550087 INFO:root:Epoch 10, Step 2500, loss: 0.16544727981090546, disc_loss: 0.0019958531484007835 INFO:root:Generator loss: 0.15670098707947916, Discriminator loss: 0.006327984908159689 INFO:root:Epoch 11, Step 500, loss: 0.2719419598579407, disc_loss: 0.002507247729226947 INFO:root:Epoch 11, Step 1000, loss: 0.2289179414510727, disc_loss: 0.0014765533851459622 INFO:root:Epoch 11, Step 1500, loss: 0.15163052082061768, disc_loss: 0.0020043537952005863 INFO:root:Epoch 11, Step 2000, loss: 0.2619889974594116, disc_loss: 0.00040794862434268 INFO:root:Epoch 11, Step 2500, loss: 0.21947583556175232, disc_loss: 0.0033102056477218866 INFO:root:Generator loss: 0.159215243985352, Discriminator loss: 0.008283218649471432 INFO:root:Epoch 12, Step 500, loss: 0.25795814394950867, disc_loss: 0.002124498598277569 INFO:root:Epoch 12, Step 1000, loss: 0.21739135682582855, disc_loss: 0.0023645160254091024 INFO:root:Epoch 12, Step 1500, loss: 0.21675364673137665, disc_loss: 0.0009649371495470405 INFO:root:Epoch 12, Step 2000, loss: 0.25408288836479187, disc_loss: 0.00045945390593260527 INFO:root:Epoch 12, Step 2500, loss: 0.2382950633764267, disc_loss: 0.00102323479950428 INFO:root:Generator loss: 0.15589472546594815, Discriminator loss: 0.004769237699522298 INFO:root:Epoch 13, Step 500, loss: 0.12218505889177322, disc_loss: 0.00337812677025795 INFO:root:Epoch 13, Step 1000, loss: 0.21033677458763123, disc_loss: 0.0019488180987536907 INFO:root:Epoch 13, Step 1500, loss: 0.16327118873596191, disc_loss: 0.0028526477981358767 INFO:root:Epoch 13, Step 2000, loss: 0.16071835160255432, disc_loss: 0.002321660052984953 INFO:root:Epoch 13, Step 2500, loss: 0.16742675006389618, disc_loss: 0.007365376688539982 INFO:root:Generator loss: 0.15445987167722972, Discriminator loss: 0.00741506577935635 INFO:root:Epoch 14, Step 500, loss: 0.23686109483242035, disc_loss: 0.000746804173104465 INFO:root:Epoch 14, Step 1000, loss: 0.22421026229858398, disc_loss: 0.001401600195094943 INFO:root:Epoch 14, Step 1500, loss: 0.24702338874340057, disc_loss: 0.0036804969422519207 INFO:root:Epoch 14, Step 2000, loss: 0.20147472620010376, disc_loss: 0.0014264158671721816 INFO:root:Epoch 14, Step 2500, loss: 0.21461667120456696, disc_loss: 0.0011242892360314727 INFO:root:Generator loss: 0.15769419533390444, Discriminator loss: 0.010084183753913978 INFO:root:Epoch 15, Step 500, loss: 0.18812422454357147, disc_loss: 0.00045774210593663156 INFO:root:Epoch 15, Step 1000, loss: 0.21720658242702484, disc_loss: 0.0015750795137137175 INFO:root:Epoch 15, Step 1500, loss: 0.18461938202381134, disc_loss: 0.00022904400248080492 INFO:root:Epoch 15, Step 2000, loss: 0.2020915150642395, disc_loss: 0.0012964744819328189 INFO:root:Epoch 15, Step 2500, loss: 0.24323663115501404, disc_loss: 0.0013249609619379044 INFO:root:Generator loss: 0.1528906009512619, Discriminator loss: 0.007542880900623503 INFO:root:Epoch 16, Step 500, loss: 0.2666374444961548, disc_loss: 0.0026336447335779667 INFO:root:Epoch 16, Step 1000, loss: 0.1554514616727829, disc_loss: 0.0012972827535122633 INFO:root:Epoch 16, Step 1500, loss: 0.14579275250434875, disc_loss: 0.0010206089355051517 INFO:root:Epoch 16, Step 2000, loss: 0.2060222327709198, disc_loss: 0.0013034377479925752 INFO:root:Epoch 16, Step 2500, loss: 0.1689905822277069, disc_loss: 0.0031409799121320248 INFO:root:Generator loss: 0.15393709026060057, Discriminator loss: 0.007572623159018618 INFO:root:Epoch 17, Step 500, loss: 0.2191568911075592, disc_loss: 0.0015926468186080456 INFO:root:Epoch 17, Step 1000, loss: 0.21130087971687317, disc_loss: 0.0025851901154965162 INFO:root:Epoch 17, Step 1500, loss: 0.13330042362213135, disc_loss: 0.0006709725712426007 INFO:root:Epoch 17, Step 2000, loss: 0.2143322229385376, disc_loss: 0.001584080164320767 INFO:root:Epoch 17, Step 2500, loss: 0.2092394232749939, disc_loss: 0.0036217852029949427 INFO:root:Generator loss: 0.1504001126272007, Discriminator loss: 0.00593943976636386 INFO:root:Epoch 18, Step 500, loss: 0.19369980692863464, disc_loss: 0.001540558529086411 INFO:root:Epoch 18, Step 1000, loss: 0.2042849361896515, disc_loss: 0.0012967906659469008 INFO:root:Epoch 18, Step 1500, loss: 0.296305388212204, disc_loss: 0.00024529595975764096 INFO:root:Epoch 18, Step 2000, loss: 0.19594815373420715, disc_loss: 0.0018008043989539146 INFO:root:Epoch 18, Step 2500, loss: 0.289691299200058, disc_loss: 0.0032066514249891043 INFO:root:Generator loss: 0.1606973894663806, Discriminator loss: 0.009119160932791645 INFO:root:Epoch 19, Step 500, loss: 0.16933570802211761, disc_loss: 0.0008373638265766203 INFO:root:Epoch 19, Step 1000, loss: 0.26960769295692444, disc_loss: 0.0008637731662020087 INFO:root:Epoch 19, Step 1500, loss: 0.1529017984867096, disc_loss: 0.0011711203260347247 INFO:root:Epoch 19, Step 2000, loss: 0.23268896341323853, disc_loss: 0.0012849016347900033 INFO:root:Epoch 19, Step 2500, loss: 0.2509724199771881, disc_loss: 0.0006878843414597213 INFO:root:Generator loss: 0.1548184705807746, Discriminator loss: 0.007865976493280647 INFO:root:Epoch 20, Step 500, loss: 0.23991236090660095, disc_loss: 0.0014005980920046568 INFO:root:Epoch 20, Step 1000, loss: 0.20016886293888092, disc_loss: 0.00015595517470501363 INFO:root:Epoch 20, Step 1500, loss: 0.10266516357660294, disc_loss: 0.006176771596074104 INFO:root:Epoch 20, Step 2000, loss: 0.1987983137369156, disc_loss: 0.003655677428469062 INFO:root:Epoch 20, Step 2500, loss: 0.16489088535308838, disc_loss: 0.01451909076422453 INFO:root:Generator loss: 0.15392981896412025, Discriminator loss: 0.009336405254817664 INFO:root:Epoch 21, Step 500, loss: 0.1588912457227707, disc_loss: 0.003106571501120925 INFO:root:Epoch 21, Step 1000, loss: 0.2122514247894287, disc_loss: 0.001715341117233038 INFO:root:Epoch 21, Step 1500, loss: 0.17192889750003815, disc_loss: 0.0022574549075216055 INFO:root:Epoch 21, Step 2000, loss: 0.20623578131198883, disc_loss: 0.0018650146666914225 INFO:root:Epoch 21, Step 2500, loss: 0.2032129466533661, disc_loss: 0.001579348579980433 INFO:root:Generator loss: 0.15073225383185646, Discriminator loss: 0.00716394909186508 INFO:root:Epoch 22, Step 500, loss: 0.21944525837898254, disc_loss: 0.0006659588543698192 INFO:root:Epoch 22, Step 1000, loss: 0.1950821578502655, disc_loss: 0.0035487571731209755 INFO:root:Epoch 22, Step 1500, loss: 0.13821372389793396, disc_loss: 0.0026758627500385046 INFO:root:Epoch 22, Step 2000, loss: 0.16071826219558716, disc_loss: 0.000786982593126595 INFO:root:Epoch 22, Step 2500, loss: 0.1513395607471466, disc_loss: 0.003028666600584984 INFO:root:Generator loss: 0.1656777167566193, Discriminator loss: 0.00689646471691605 INFO:root:Epoch 23, Step 500, loss: 0.17124043405056, disc_loss: 0.00040915425051935017 INFO:root:Epoch 23, Step 1000, loss: 0.23920418322086334, disc_loss: 0.0017987532773986459 INFO:root:Epoch 23, Step 1500, loss: 0.14390625059604645, disc_loss: 0.001338382251560688 INFO:root:Epoch 23, Step 2000, loss: 0.19782759249210358, disc_loss: 0.001112670754082501 INFO:root:Epoch 23, Step 2500, loss: 0.20089788734912872, disc_loss: 0.0020863921381533146 INFO:root:Generator loss: 0.15288031542474784, Discriminator loss: 0.005006089134554908 INFO:root:Epoch 24, Step 500, loss: 0.21623064577579498, disc_loss: 0.0004943881649523973 INFO:root:Epoch 24, Step 1000, loss: 0.1983931064605713, disc_loss: 0.0024461522698402405 INFO:root:Epoch 24, Step 1500, loss: 0.22622570395469666, disc_loss: 0.054060615599155426 INFO:root:Epoch 24, Step 2000, loss: 0.1344752013683319, disc_loss: 0.00027848611352965236 INFO:root:Epoch 24, Step 2500, loss: 0.22571316361427307, disc_loss: 0.06606058031320572 INFO:root:Generator loss: 0.1486151763464062, Discriminator loss: 0.0073336784491084914 INFO:root:Epoch 25, Step 500, loss: 0.1643536388874054, disc_loss: 0.002177179092541337 INFO:root:Epoch 25, Step 1000, loss: 0.1477476805448532, disc_loss: 0.003139293985441327 INFO:root:Epoch 25, Step 1500, loss: 0.23426830768585205, disc_loss: 0.0012699420331045985 INFO:root:Epoch 25, Step 2000, loss: 0.15443092584609985, disc_loss: 0.0009152680868282914 INFO:root:Epoch 25, Step 2500, loss: 0.14525850117206573, disc_loss: 0.0014707488007843494 INFO:root:Generator loss: 0.151314642063333, Discriminator loss: 0.006845492455804295 INFO:root:Epoch 26, Step 500, loss: 0.17136962711811066, disc_loss: 0.0028529104311019182 INFO:root:Epoch 26, Step 1000, loss: 0.19324925541877747, disc_loss: 0.0018980796448886395 INFO:root:Epoch 26, Step 1500, loss: 0.1829681247472763, disc_loss: 0.027104754000902176 INFO:root:Epoch 26, Step 2000, loss: 0.2756149172782898, disc_loss: 0.0010534025495871902 INFO:root:Epoch 26, Step 2500, loss: 0.22364293038845062, disc_loss: 0.000749379163607955 INFO:root:Generator loss: 0.14807583926950843, Discriminator loss: 0.010684788025703277 INFO:root:Epoch 27, Step 500, loss: 0.20551946759223938, disc_loss: 0.0022747379262000322 INFO:root:Epoch 27, Step 1000, loss: 0.2037193924188614, disc_loss: 0.0014990002382546663 INFO:root:Epoch 27, Step 1500, loss: 0.28234779834747314, disc_loss: 0.001453203265555203 INFO:root:Epoch 27, Step 2000, loss: 0.2405531257390976, disc_loss: 0.002639435464516282 INFO:root:Epoch 27, Step 2500, loss: 0.20126855373382568, disc_loss: 0.0028889309614896774 INFO:root:Generator loss: 0.15275396493453425, Discriminator loss: 0.00758786126028196 INFO:root:Epoch 28, Step 500, loss: 0.17823928594589233, disc_loss: 0.0006160064949654043 INFO:root:Epoch 28, Step 1000, loss: 0.18538382649421692, disc_loss: 0.001021132804453373 INFO:root:Epoch 28, Step 1500, loss: 0.18735086917877197, disc_loss: 0.00253280415199697 INFO:root:Epoch 28, Step 2000, loss: 0.1933680772781372, disc_loss: 0.0010457060998305678 INFO:root:Epoch 28, Step 2500, loss: 0.1937006562948227, disc_loss: 0.0017968968022614717 INFO:root:Generator loss: 0.15479988152685675, Discriminator loss: 0.008467569512857471 INFO:root:Epoch 29, Step 500, loss: 0.17594964802265167, disc_loss: 0.0018970801029354334 INFO:root:Epoch 29, Step 1000, loss: 0.16756503283977509, disc_loss: 0.001571777742356062 INFO:root:Epoch 29, Step 1500, loss: 0.20182061195373535, disc_loss: 0.0032018704805523157 INFO:root:Epoch 29, Step 2000, loss: 0.23015911877155304, disc_loss: 0.0011170469224452972 INFO:root:Epoch 29, Step 2500, loss: 0.21530109643936157, disc_loss: 0.0010174177587032318 INFO:root:Generator loss: 0.15145057978705295, Discriminator loss: 0.0073985461544863445 INFO:root:Epoch 30, Step 500, loss: 0.18501895666122437, disc_loss: 0.0007586692809127271 INFO:root:Epoch 30, Step 1000, loss: 0.16187520325183868, disc_loss: 0.004758385010063648 INFO:root:Epoch 30, Step 1500, loss: 0.17666107416152954, disc_loss: 0.0039044057484716177 INFO:root:Epoch 30, Step 2000, loss: 0.14470462501049042, disc_loss: 0.0010651007760316133 INFO:root:Epoch 30, Step 2500, loss: 0.13727925717830658, disc_loss: 0.0047175087966024876 INFO:root:Generator loss: 0.14848027671280417, Discriminator loss: 0.00486005955364146 INFO:root:Epoch 31, Step 500, loss: 0.1879275143146515, disc_loss: 0.001492742565460503 INFO:root:Epoch 31, Step 1000, loss: 0.13713416457176208, disc_loss: 0.0016783855389803648 INFO:root:Epoch 31, Step 1500, loss: 0.20209118723869324, disc_loss: 0.004142406862229109 INFO:root:Epoch 31, Step 2000, loss: 0.168260857462883, disc_loss: 0.00043510296382009983 INFO:root:Epoch 31, Step 2500, loss: 0.24361303448677063, disc_loss: 0.0011475462233647704 INFO:root:Generator loss: 0.1451130717341761, Discriminator loss: 0.006827290648372794 INFO:root:Epoch 32, Step 500, loss: 0.2178606390953064, disc_loss: 0.0003761161060538143 INFO:root:Epoch 32, Step 1000, loss: 0.1325763314962387, disc_loss: 0.0015204480150714517 INFO:root:Epoch 32, Step 1500, loss: 0.19770699739456177, disc_loss: 0.0013429338578134775 INFO:root:Epoch 32, Step 2000, loss: 0.15416398644447327, disc_loss: 0.001986399292945862 INFO:root:Epoch 32, Step 2500, loss: 0.1255073994398117, disc_loss: 0.04426322504878044 INFO:root:Generator loss: 0.1506036072607758, Discriminator loss: 0.008968745695384112 INFO:root:Epoch 33, Step 500, loss: 0.15152476727962494, disc_loss: 0.001831963425502181 INFO:root:Epoch 33, Step 1000, loss: 0.20520329475402832, disc_loss: 0.0021505120676010847 INFO:root:Epoch 33, Step 1500, loss: 0.15221679210662842, disc_loss: 0.0006668538553640246 INFO:root:Epoch 33, Step 2000, loss: 0.16760966181755066, disc_loss: 0.00034050486283376813 INFO:root:Epoch 33, Step 2500, loss: 0.13106703758239746, disc_loss: 0.0009691817685961723 INFO:root:Generator loss: 0.14916171046715337, Discriminator loss: 0.006695696125920999 INFO:root:Epoch 34, Step 500, loss: 0.17272314429283142, disc_loss: 0.000887170375790447 INFO:root:Epoch 34, Step 1000, loss: 0.2132679969072342, disc_loss: 0.0016812332905828953 INFO:root:Epoch 34, Step 1500, loss: 0.16384224593639374, disc_loss: 0.0019412646070122719 INFO:root:Epoch 34, Step 2000, loss: 0.2238304764032364, disc_loss: 0.0032721676398068666 INFO:root:Epoch 34, Step 2500, loss: 0.22253461182117462, disc_loss: 0.0025769562926143408 INFO:root:Generator loss: 0.151082797862083, Discriminator loss: 0.007573019889481727 INFO:root:Epoch 35, Step 500, loss: 0.22639697790145874, disc_loss: 0.0017826772527769208 INFO:root:Epoch 35, Step 1000, loss: 0.17547865211963654, disc_loss: 0.003049442544579506 INFO:root:Epoch 35, Step 1500, loss: 0.13126714527606964, disc_loss: 0.0017061545513570309 INFO:root:Epoch 35, Step 2000, loss: 0.24197342991828918, disc_loss: 0.0005070678307674825 INFO:root:Epoch 35, Step 2500, loss: 0.19099745154380798, disc_loss: 0.0006508170044980943 INFO:root:Generator loss: 0.15685435558117708, Discriminator loss: 0.008581994546439445 INFO:root:Epoch 36, Step 500, loss: 0.17108827829360962, disc_loss: 0.0014752569841220975 INFO:root:Epoch 36, Step 1000, loss: 0.2389165610074997, disc_loss: 0.0015733256004750729 INFO:root:Epoch 36, Step 1500, loss: 0.19396057724952698, disc_loss: 0.0002904093998949975 INFO:root:Epoch 36, Step 2000, loss: 0.16288608312606812, disc_loss: 0.0021133290138095617 INFO:root:Epoch 36, Step 2500, loss: 0.20927363634109497, disc_loss: 0.000661283906083554 INFO:root:Generator loss: 0.14912311179400647, Discriminator loss: 0.0067926698288432 INFO:root:Epoch 37, Step 500, loss: 0.18722005188465118, disc_loss: 0.002159852534532547 INFO:root:Epoch 37, Step 1000, loss: 0.15937204658985138, disc_loss: 0.0018713129684329033 INFO:root:Epoch 37, Step 1500, loss: 0.15392547845840454, disc_loss: 0.0020612888038158417 INFO:root:Epoch 37, Step 2000, loss: 0.14782842993736267, disc_loss: 0.0010478856274858117 INFO:root:Epoch 37, Step 2500, loss: 0.2138817310333252, disc_loss: 0.00011799987987615168 INFO:root:Generator loss: 0.15218570727167777, Discriminator loss: 0.008030950438694066 INFO:root:Epoch 38, Step 500, loss: 0.12195228040218353, disc_loss: 0.0006151677225716412 INFO:root:Epoch 38, Step 1000, loss: 0.1600419580936432, disc_loss: 0.0010397137375548482 INFO:root:Epoch 38, Step 1500, loss: 0.15388871729373932, disc_loss: 0.0020387968979775906 INFO:root:Epoch 38, Step 2000, loss: 0.14046302437782288, disc_loss: 0.0007483579684048891 INFO:root:Epoch 38, Step 2500, loss: 0.15302549302577972, disc_loss: 0.0016489940462633967 INFO:root:Generator loss: 0.14867327575689382, Discriminator loss: 0.006284861755048748 INFO:root:Epoch 39, Step 500, loss: 0.17320281267166138, disc_loss: 0.0036755544133484364 INFO:root:Epoch 39, Step 1000, loss: 0.15600933134555817, disc_loss: 0.0006042426102794707 INFO:root:Epoch 39, Step 1500, loss: 0.20386555790901184, disc_loss: 0.004750823602080345 INFO:root:Epoch 39, Step 2000, loss: 0.15265820920467377, disc_loss: 0.0014454181073233485 INFO:root:Epoch 39, Step 2500, loss: 0.19368837773799896, disc_loss: 0.002038124715909362 INFO:root:Generator loss: 0.14978806808301545, Discriminator loss: 0.008053443771185132 INFO:root:Epoch 40, Step 500, loss: 0.1634645313024521, disc_loss: 0.0008760954951867461 INFO:root:Epoch 40, Step 1000, loss: 0.24363578855991364, disc_loss: 0.0014732916606590152 INFO:root:Epoch 40, Step 1500, loss: 0.1921803057193756, disc_loss: 0.001860494608990848 INFO:root:Epoch 40, Step 2000, loss: 0.12478583306074142, disc_loss: 0.0015014746459200978 INFO:root:Epoch 40, Step 2500, loss: 0.20051966607570648, disc_loss: 0.0013981866650283337 INFO:root:Generator loss: 0.1495014434879266, Discriminator loss: 0.008858757580755373 INFO:root:Epoch 41, Step 500, loss: 0.1557866483926773, disc_loss: 0.0009908645879477262 INFO:root:Epoch 41, Step 1000, loss: 0.14158287644386292, disc_loss: 0.01940055377781391 INFO:root:Epoch 41, Step 1500, loss: 0.15362367033958435, disc_loss: 0.0011035962961614132 INFO:root:Epoch 41, Step 2000, loss: 0.18089531362056732, disc_loss: 0.0012486111372709274 INFO:root:Epoch 41, Step 2500, loss: 0.1619526594877243, disc_loss: 0.0009440151043236256 INFO:root:Generator loss: 0.1488416255677788, Discriminator loss: 0.0070254697599470865 INFO:root:Epoch 42, Step 500, loss: 0.19800350069999695, disc_loss: 0.001883466960862279 INFO:root:Epoch 42, Step 1000, loss: 0.22153475880622864, disc_loss: 0.0011596799595281482 INFO:root:Epoch 42, Step 1500, loss: 0.1665220558643341, disc_loss: 0.0014407207490876317 INFO:root:Epoch 42, Step 2000, loss: 0.13804322481155396, disc_loss: 0.0003912303363904357 INFO:root:Epoch 42, Step 2500, loss: 0.22652436792850494, disc_loss: 0.0013467944227159023 INFO:root:Generator loss: 0.15468112484199328, Discriminator loss: 0.007592531941854847 INFO:root:Epoch 43, Step 500, loss: 0.1732303947210312, disc_loss: 0.0006291429745033383 INFO:root:Epoch 43, Step 1000, loss: 0.22385995090007782, disc_loss: 0.09368821978569031 INFO:root:Epoch 43, Step 1500, loss: 0.19474774599075317, disc_loss: 0.0015064507024362683 INFO:root:Epoch 43, Step 2000, loss: 0.19128721952438354, disc_loss: 0.004481759387999773 INFO:root:Epoch 43, Step 2500, loss: 0.1631593257188797, disc_loss: 0.0011000009253621101 INFO:root:Generator loss: 0.14729728560569216, Discriminator loss: 0.009075007488435694 INFO:root:Epoch 44, Step 500, loss: 0.17715486884117126, disc_loss: 0.0018717246130108833 INFO:root:Epoch 44, Step 1000, loss: 0.25556033849716187, disc_loss: 0.000396895018639043 INFO:root:Epoch 44, Step 1500, loss: 0.21550242602825165, disc_loss: 0.0014451321912929416 INFO:root:Epoch 44, Step 2000, loss: 0.1984427273273468, disc_loss: 0.0024722693488001823 INFO:root:Epoch 44, Step 2500, loss: 0.21361075341701508, disc_loss: 0.000819200009573251 INFO:root:Generator loss: 0.14986323091445616, Discriminator loss: 0.006630657830224593 INFO:root:Epoch 45, Step 500, loss: 0.20685818791389465, disc_loss: 0.0010365750640630722 INFO:root:Epoch 45, Step 1000, loss: 0.2404719889163971, disc_loss: 0.0006383370491676033 INFO:root:Epoch 45, Step 1500, loss: 0.17369501292705536, disc_loss: 0.0033343026880174875 INFO:root:Epoch 45, Step 2000, loss: 0.17553487420082092, disc_loss: 0.0008124223095364869 INFO:root:Epoch 45, Step 2500, loss: 0.20491598546504974, disc_loss: 9.308904554927722e-05 INFO:root:Generator loss: 0.14902781203244497, Discriminator loss: 0.00557117249770528 INFO:root:Epoch 46, Step 500, loss: 0.18839885294437408, disc_loss: 0.001555745955556631 INFO:root:Epoch 46, Step 1000, loss: 0.2238050103187561, disc_loss: 0.0016249532345682383 INFO:root:Epoch 46, Step 1500, loss: 0.13696785271167755, disc_loss: 0.0016735948156565428 INFO:root:Epoch 46, Step 2000, loss: 0.1770334541797638, disc_loss: 0.0014958776300773025 INFO:root:Epoch 46, Step 2500, loss: 0.2255016267299652, disc_loss: 0.0013764165341854095 INFO:root:Generator loss: 0.15075078678941264, Discriminator loss: 0.007199312354343631 INFO:root:Epoch 47, Step 500, loss: 0.26174628734588623, disc_loss: 0.0010621807305142283 INFO:root:Epoch 47, Step 1000, loss: 0.16118288040161133, disc_loss: 0.002037838101387024 INFO:root:Epoch 47, Step 1500, loss: 0.21030934154987335, disc_loss: 0.00024075487453956157 INFO:root:Epoch 47, Step 2000, loss: 0.2071937471628189, disc_loss: 0.000964963692240417 INFO:root:Epoch 47, Step 2500, loss: 0.1436789333820343, disc_loss: 0.0022595354821532965 INFO:root:Generator loss: 0.1490943245951412, Discriminator loss: 0.00782963457995985 INFO:root:Epoch 48, Step 500, loss: 0.16596905887126923, disc_loss: 0.0016859389143064618 INFO:root:Epoch 48, Step 1000, loss: 0.20538239181041718, disc_loss: 0.000355234689777717 INFO:root:Epoch 48, Step 1500, loss: 0.12697157263755798, disc_loss: 0.00154679617844522 INFO:root:Epoch 48, Step 2000, loss: 0.15941494703292847, disc_loss: 0.00042753099114634097 INFO:root:Epoch 48, Step 2500, loss: 0.12591500580310822, disc_loss: 0.001462430227547884 INFO:root:Generator loss: 0.15104802408554022, Discriminator loss: 0.0062026559350219805 INFO:root:Epoch 49, Step 500, loss: 0.19275020062923431, disc_loss: 0.000377537973690778 INFO:root:Epoch 49, Step 1000, loss: 0.18330758810043335, disc_loss: 8.651253301650286e-05 INFO:root:Epoch 49, Step 1500, loss: 0.21846988797187805, disc_loss: 0.000389557215385139 INFO:root:Epoch 49, Step 2000, loss: 0.2317807674407959, disc_loss: 0.0027107701171189547 INFO:root:Epoch 49, Step 2500, loss: 0.2366754412651062, disc_loss: 0.00023985184088815004 INFO:root:Generator loss: 0.1506285196152127, Discriminator loss: 0.006962537650429775 INFO:root:Epoch 50, Step 500, loss: 0.18307413160800934, disc_loss: 0.0006114450516179204 INFO:root:Epoch 50, Step 1000, loss: 0.18070276081562042, disc_loss: 0.0014288920210674405 INFO:root:Epoch 50, Step 1500, loss: 0.16713868081569672, disc_loss: 0.0002362645900575444 INFO:root:Epoch 50, Step 2000, loss: 0.1450747698545456, disc_loss: 0.00239872420206666 INFO:root:Epoch 50, Step 2500, loss: 0.26365476846694946, disc_loss: 0.001221455866470933 INFO:root:Generator loss: 0.15042886561126384, Discriminator loss: 0.006950676874783189 INFO:root:Epoch 51, Step 500, loss: 0.2152542918920517, disc_loss: 0.0007097358466126025 INFO:root:Epoch 51, Step 1000, loss: 0.12946553528308868, disc_loss: 0.000846027338411659 INFO:root:Epoch 51, Step 1500, loss: 0.16722026467323303, disc_loss: 0.0010668804170563817 INFO:root:Epoch 51, Step 2000, loss: 0.2161879539489746, disc_loss: 0.0006570870755240321 INFO:root:Epoch 51, Step 2500, loss: 0.19870325922966003, disc_loss: 0.0002888255112338811 INFO:root:Generator loss: 0.15483897298719118, Discriminator loss: 0.010537156224512446 INFO:root:Epoch 52, Step 500, loss: 0.1790369302034378, disc_loss: 0.000788058852776885 INFO:root:Epoch 52, Step 1000, loss: 0.16181029379367828, disc_loss: 0.0019506269600242376 INFO:root:Epoch 52, Step 1500, loss: 0.1600004881620407, disc_loss: 0.0030104797333478928 INFO:root:Epoch 52, Step 2000, loss: 0.13153015077114105, disc_loss: 0.0012453808449208736 INFO:root:Epoch 52, Step 2500, loss: 0.17636753618717194, disc_loss: 0.0017272598342970014 INFO:root:Generator loss: 0.15071525269197028, Discriminator loss: 0.007104984784693213 INFO:root:Epoch 53, Step 500, loss: 0.16577422618865967, disc_loss: 0.0005849118460901082 INFO:root:Epoch 53, Step 1000, loss: 0.1668897420167923, disc_loss: 0.0014300784096121788 INFO:root:Epoch 53, Step 1500, loss: 0.1756283938884735, disc_loss: 0.0009977087611332536 INFO:root:Epoch 53, Step 2000, loss: 0.1373414546251297, disc_loss: 0.0019349803915247321 INFO:root:Epoch 53, Step 2500, loss: 0.15059304237365723, disc_loss: 0.00017773735453374684 INFO:root:Generator loss: 0.1531875601526603, Discriminator loss: 0.006457391921305793 INFO:root:Epoch 54, Step 500, loss: 0.16114716231822968, disc_loss: 0.005647638346999884 INFO:root:Epoch 54, Step 1000, loss: 0.13460399210453033, disc_loss: 0.000806011026725173 INFO:root:Epoch 54, Step 1500, loss: 0.19509609043598175, disc_loss: 0.0008489741594530642 INFO:root:Epoch 54, Step 2000, loss: 0.16986267268657684, disc_loss: 0.0007771639502607286 INFO:root:Epoch 54, Step 2500, loss: 0.24058441817760468, disc_loss: 0.002403619699180126 INFO:root:Generator loss: 0.15293034997958582, Discriminator loss: 0.007628766357564159
Although there are still some differences between this result with the paper provided, compared to many other methods, this result is already very good. Thank you for your detailed response. I think I may try this repo again in the next few days :)
from cmgan.
Related Issues (20)
- RuntimeError HOT 3
- the change of gen_loss during training HOT 1
- RuntimeError
- RuntimeError HOT 2
- About the decreasing of loss HOT 1
- Can not reproduce the results HOT 12
- Training can get stuck HOT 6
- Inferior results trained from scratch HOT 7
- RuntimeeError HOT 1
- Training GPU requirements
- File "pesq/cypesq.pyx", line 1, in init cypesq ImportError: numpy.core.multiarray failed to import (auto-generated because you didn't call 'numpy.import_array()' after cimporting numpy; use '<void>numpy._import_array' to disable if you are certain you don't need it)
- File "/anaconda3/envs/cmg/lib/python3.9/site-packages/torch/nn/parallel/distributed.py", line 578, in __init__ dist._verify_model_across_ranks(self.process_group, parameters) RuntimeError: NCCL error in: ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:957, invalid usage, NCCL version 21.0.3 ncclInvalidUsage: This usually reflects invalid usage of NCCL library (such as too many async ops, too many collectives at once, mixing streams in a group, etc). HOT 2
- How do you resample to 16000? HOT 1
- 时域Loss计算疑惑
- the training speed confusion
- My server has a 3090, but reports that I don't have a gpu HOT 1
- Test set requirements when training
- epochs
- 模型训练的采样率以及显卡训练配置咨询
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from cmgan.