Upload v14p11/ckpt_060000.json with huggingface_hub
Browse files- v14p11/ckpt_060000.json +1 -0
v14p11/ckpt_060000.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"sent": 60000, "total_steps": 1040622, "eval_results": [{"sent": 2000, "train_loss": 2.432787154142005, "train_acc": 0.3502440560541647, "eval_loss": 2.118043747758509, "eval_acc": 0.36486486486486486, "ppl": 8.314855612223376, "w_freq": 0.002980692174428512, "w_bigram": 0.08442807804126903, "w_trigram": 0.5600977178934385, "w_skip_bi": 0.0, "w_skip_tri": 0.30558859322324583, "w_neural_phoneme": 0.11086792535243935, "w_neural_word": 0.10704230938567356, "word_vocab": 1593, "elapsed_s": 22.772148609161377, "steps_per_sec": 1394.466571644661}, {"sent": 4000, "train_loss": 2.2786028354161605, "train_acc": 0.3700189753320683, "eval_loss": 2.0249313112630087, "eval_acc": 0.4268292682926829, "ppl": 7.57559056901693, "w_freq": 0.001900615251889618, "w_bigram": 0.04091509370044348, "w_trigram": 0.6025034681103408, "w_skip_bi": 0.016269361181900557, "w_skip_tri": 0.30044624013305987, "w_neural_phoneme": 0.14248896115220244, "w_neural_word": 0.12885198020550945, "word_vocab": 2198, "elapsed_s": 42.032264709472656, "steps_per_sec": 1382.3904184469293}, {"sent": 6000, "train_loss": 2.2638648611260805, "train_acc": 0.3696285880860112, "eval_loss": 2.2661469111895056, "eval_acc": 0.29591836734693877, "ppl": 9.642176982521246, "w_freq": 0.005071080760866514, "w_bigram": 0.1004709937035607, "w_trigram": 0.7215903272616544, "w_skip_bi": 0.0, "w_skip_tri": 0.15471821963118054, "w_neural_phoneme": 0.12349327549711486, "w_neural_word": 0.12592625245814743, "word_vocab": 2819, "elapsed_s": 61.89241814613342, "steps_per_sec": 1409.9303697257724}, {"sent": 8000, "train_loss": 2.2579223470919176, "train_acc": 0.3729170426517476, "eval_loss": 1.977694124366012, "eval_acc": 0.4067796610169492, "ppl": 7.226061360855239, "w_freq": 0.0032242825955699755, "w_bigram": 0.06519526018729675, "w_trigram": 0.686213882826786, "w_skip_bi": 0.004398696612972909, "w_skip_tri": 0.2961216714064021, "w_neural_phoneme": 0.11345740581842013, "w_neural_word": 0.11019285304894107, "word_vocab": 3459, "elapsed_s": 84.53820824623108, "steps_per_sec": 1425.5092756283088}, {"sent": 10000, "train_loss": 2.2407470616472644, "train_acc": 0.37700868857699127, "eval_loss": 1.8662516498131747, "eval_acc": 0.4264705882352941, "ppl": 6.464021516069884, "w_freq": 0.0016634194364748773, "w_bigram": 0.08551293040231268, "w_trigram": 0.8904272942562684, "w_skip_bi": 0.002753659917702692, "w_skip_tri": 0.10463157830770219, "w_neural_phoneme": 0.14308896595294693, "w_neural_word": 0.1091067059825302, "word_vocab": 4067, "elapsed_s": 110.5045576095581, "steps_per_sec": 1401.9602752257877}, {"sent": 12000, "train_loss": 2.145701733835004, "train_acc": 0.39764410240301556, "eval_loss": 2.0540618566454256, "eval_acc": 0.39473684210526316, "ppl": 7.799517373940953, "w_freq": 0.0025111378447461898, "w_bigram": 0.02319721069751405, "w_trigram": 1.0068745534000523, "w_skip_bi": 0.0, "w_skip_tri": 0.1011492495659415, "w_neural_phoneme": 0.1090752779054336, "w_neural_word": 0.11504807088145023, "word_vocab": 4455, "elapsed_s": 132.12165808677673, "steps_per_sec": 1413.5305498311143}, {"sent": 14000, "train_loss": 2.203491558671496, "train_acc": 0.3805706679013775, "eval_loss": 2.220696383678586, "eval_acc": 0.4225352112676056, "ppl": 9.21374493388752, "w_freq": 0.004144665671882547, "w_bigram": 0.007991062477394301, "w_trigram": 0.8784748119715345, "w_skip_bi": 0.01230585893434202, "w_skip_tri": 0.00534574999733973, "w_neural_phoneme": 0.12642355084730628, "w_neural_word": 0.10973620996738914, "word_vocab": 4946, "elapsed_s": 155.3891613483429, "steps_per_sec": 1424.2563514701674}, {"sent": 16000, "train_loss": 2.168354578516426, "train_acc": 0.3860417666826692, "eval_loss": 2.0685103558852416, "eval_acc": 0.45454545454545453, "ppl": 7.913026740115201, "w_freq": 0.0018057590135707087, "w_bigram": 0.024685564394030707, "w_trigram": 0.8688733919989445, "w_skip_bi": 0.003252207009275675, "w_skip_tri": 0.09758470430097636, "w_neural_phoneme": 0.12342357090711002, "w_neural_word": 0.11577498176896588, "word_vocab": 5338, "elapsed_s": 177.91618943214417, "steps_per_sec": 1431.2469304381007}, {"sent": 18000, "train_loss": 2.189501757602934, "train_acc": 0.3812078309921469, "eval_loss": 2.0181485175927567, "eval_acc": 0.4230769230769231, "ppl": 7.524380770337071, "w_freq": 0.0017931810925152865, "w_bigram": 0.0023788496422169807, "w_trigram": 0.9988364307265897, "w_skip_bi": 0.0025461379046776587, "w_skip_tri": 0.10618368675643003, "w_neural_phoneme": 0.10391532984906614, "w_neural_word": 0.10270729266818661, "word_vocab": 5752, "elapsed_s": 202.25098371505737, "steps_per_sec": 1437.847147431945}, {"sent": 20000, "train_loss": 2.148403896381157, "train_acc": 0.39114577155734787, "eval_loss": 1.8308064300583136, "eval_acc": 0.4594594594594595, "ppl": 6.238915879691277, "w_freq": 0.0008110459876357098, "w_bigram": 0.05521536514067184, "w_trigram": 0.9525261835647582, "w_skip_bi": 0.0034460457712969903, "w_skip_tri": 0.046515023196689945, "w_neural_phoneme": 0.12355741554891311, "w_neural_word": 0.10257844190021748, "word_vocab": 6098, "elapsed_s": 227.18542671203613, "steps_per_sec": 1428.4807115350354}, {"sent": 22000, "train_loss": 2.157125638741271, "train_acc": 0.3881543198695888, "eval_loss": 2.2514547735249275, "eval_acc": 0.34782608695652173, "ppl": 9.501548387895875, "w_freq": 0.0008045319795472094, "w_bigram": 0.017566520834153776, "w_trigram": 1.036018300904672, "w_skip_bi": 0.0013008332882517345, "w_skip_tri": 0.07463912252609632, "w_neural_phoneme": 0.11745308591098921, "w_neural_word": 0.10980822456400445, "word_vocab": 6393, "elapsed_s": 249.61171889305115, "steps_per_sec": 1432.8493933942325}, {"sent": 24000, "train_loss": 2.166642168083094, "train_acc": 0.38901120567842, "eval_loss": 1.981601900975726, "eval_acc": 0.4044943820224719, "ppl": 7.2543544399332385, "w_freq": 0.0005989243644755175, "w_bigram": 0.013516189000682539, "w_trigram": 0.9757727063133742, "w_skip_bi": 0.004295254266240299, "w_skip_tri": 0.017335311235351442, "w_neural_phoneme": 0.10812098611069718, "w_neural_word": 0.10561047785533544, "word_vocab": 6673, "elapsed_s": 272.60766768455505, "steps_per_sec": 1423.6099934249487}, {"sent": 26000, "train_loss": 2.2162730674125717, "train_acc": 0.3731794449024457, "eval_loss": 2.2940609343141585, "eval_acc": 0.3870967741935484, "ppl": 9.915120689509424, "w_freq": 0.0005009553010530371, "w_bigram": 0.004033512775049364, "w_trigram": 0.9949173973202257, "w_skip_bi": 0.0, "w_skip_tri": 0.04671435437813999, "w_neural_phoneme": 0.10156565188303651, "w_neural_word": 0.10122616229844661, "word_vocab": 6984, "elapsed_s": 295.0455267429352, "steps_per_sec": 1426.3493659629833}, {"sent": 28000, "train_loss": 2.3349580686849114, "train_acc": 0.35027531667082096, "eval_loss": 2.1471757754783005, "eval_acc": 0.3803680981595092, "ppl": 8.560647034831863, "w_freq": 0.0007111893844633669, "w_bigram": 0.060228408259701244, "w_trigram": 0.8169010341548681, "w_skip_bi": 0.002297272280085888, "w_skip_tri": 0.08731229999490946, "w_neural_phoneme": 0.11097362932551486, "w_neural_word": 0.10072951741801331, "word_vocab": 7648, "elapsed_s": 326.4926710128784, "steps_per_sec": 1424.133039671385}, {"sent": 30000, "train_loss": 2.3953538370950977, "train_acc": 0.33940063211606564, "eval_loss": 2.174915131392821, "eval_acc": 0.36423841059602646, "ppl": 8.801438124694922, "w_freq": 0.000844758730976373, "w_bigram": 0.03125119597200782, "w_trigram": 0.8183211406515284, "w_skip_bi": 0.0037810012486415513, "w_skip_tri": 0.032706153482974384, "w_neural_phoneme": 0.10665008873567963, "w_neural_word": 0.10356355595208931, "word_vocab": 8515, "elapsed_s": 361.0735294818878, "steps_per_sec": 1433.2011564031266}, {"sent": 32000, "train_loss": 2.370509402421815, "train_acc": 0.3400035646580359, "eval_loss": 1.9827539899773676, "eval_acc": 0.4017094017094017, "ppl": 7.26271691813211, "w_freq": 0.00024439414043867327, "w_bigram": 0.03717681609081632, "w_trigram": 0.928620467869658, "w_skip_bi": 0.0, "w_skip_tri": 0.061650265248159086, "w_neural_phoneme": 0.10037571495906418, "w_neural_word": 0.10026030989027548, "word_vocab": 9486, "elapsed_s": 401.62839460372925, "steps_per_sec": 1442.1490307513777}, {"sent": 34000, "train_loss": 2.3614921359348937, "train_acc": 0.3453666178523993, "eval_loss": 2.098575636099161, "eval_acc": 0.4025974025974026, "ppl": 8.154546594818305, "w_freq": 0.00022036682841590656, "w_bigram": 0.015794162113159014, "w_trigram": 0.9465458403521261, "w_skip_bi": 0.0029195686868849708, "w_skip_tri": 0.01138276633078188, "w_neural_phoneme": 0.10043348293866802, "w_neural_word": 0.10027480221906118, "word_vocab": 10546, "elapsed_s": 440.23353695869446, "steps_per_sec": 1449.0808774067912}, {"sent": 36000, "train_loss": 2.3614486116869124, "train_acc": 0.3519731173530932, "eval_loss": 2.138724256554134, "eval_acc": 0.3867924528301887, "ppl": 8.488601440605288, "w_freq": 0.0003747182513018575, "w_bigram": 0.002789287970172509, "w_trigram": 0.9258438392261324, "w_skip_bi": 0.0011260176104164693, "w_skip_tri": 0.0026962371538783152, "w_neural_phoneme": 0.1005525163899946, "w_neural_word": 0.10055618618495057, "word_vocab": 11088, "elapsed_s": 463.72850251197815, "steps_per_sec": 1450.7454175358193}, {"sent": 38000, "train_loss": 2.34023034975833, "train_acc": 0.35172017524958704, "eval_loss": 2.1535453088385252, "eval_acc": 0.32558139534883723, "ppl": 8.615348387811272, "w_freq": 0.001166950701105949, "w_bigram": 0.004555404293321961, "w_trigram": 0.9726397467382298, "w_skip_bi": 0.00359410552828447, "w_skip_tri": 0.0019243890237378384, "w_neural_phoneme": 0.10092916267009587, "w_neural_word": 0.10116679587155879, "word_vocab": 11395, "elapsed_s": 482.90501618385315, "steps_per_sec": 1450.798762739019}, {"sent": 40000, "train_loss": 2.3545010743446824, "train_acc": 0.3528838198794469, "eval_loss": 2.0920821164200225, "eval_acc": 0.41935483870967744, "ppl": 8.10176643599007, "w_freq": 6.877921298001959e-05, "w_bigram": 0.012395967603322268, "w_trigram": 0.9900739476754092, "w_skip_bi": 0.0, "w_skip_tri": 0.03205652819958863, "w_neural_phoneme": 0.10073410700800188, "w_neural_word": 0.10068276665316908, "word_vocab": 11827, "elapsed_s": 505.8135368824005, "steps_per_sec": 1452.0014717810027}, {"sent": 42000, "train_loss": 2.3371681404678326, "train_acc": 0.3484729095092508, "eval_loss": 2.3049973856160535, "eval_acc": 0.3412322274881517, "ppl": 10.02415204540851, "w_freq": 0.00013757911831188934, "w_bigram": 0.013710322789071822, "w_trigram": 0.9986118108021049, "w_skip_bi": 0.0005575665036098186, "w_skip_tri": 0.01971635819378755, "w_neural_phoneme": 0.10029222149837813, "w_neural_word": 0.10025323221025317, "word_vocab": 12148, "elapsed_s": 528.8179748058319, "steps_per_sec": 1448.8331269021578}, {"sent": 44000, "train_loss": 2.3368983976089885, "train_acc": 0.33988803358992303, "eval_loss": 2.4245802555374403, "eval_acc": 0.3153153153153153, "ppl": 11.297486366570952, "w_freq": 5.2209258270114504e-05, "w_bigram": 0.04928819523979438, "w_trigram": 0.9898752936387495, "w_skip_bi": 0.00018351158720042983, "w_skip_tri": 0.0008697350566886753, "w_neural_phoneme": 0.10012567641275837, "w_neural_word": 0.1001002002950189, "word_vocab": 12373, "elapsed_s": 548.1976408958435, "steps_per_sec": 1449.749033398341}, {"sent": 46000, "train_loss": 2.342050991997343, "train_acc": 0.35287343084187495, "eval_loss": 2.2461093193945234, "eval_acc": 0.3448275862068966, "ppl": 9.45089380331619, "w_freq": 0.0003120583066123222, "w_bigram": 0.049556422184625686, "w_trigram": 0.9899769384437657, "w_skip_bi": 0.000897932675955054, "w_skip_tri": 0.02453802353673541, "w_neural_phoneme": 0.10010545134789105, "w_neural_word": 0.10009934989349702, "word_vocab": 12688, "elapsed_s": 573.2708399295807, "steps_per_sec": 1446.9269012564664}, {"sent": 48000, "train_loss": 2.2744664619421595, "train_acc": 0.3517450245168734, "eval_loss": 2.1525416633790826, "eval_acc": 0.40625, "ppl": 8.606705970207384, "w_freq": 0.00022096073348991347, "w_bigram": 0.06538324030350139, "w_trigram": 0.9939165955072082, "w_skip_bi": 0.0011271896028859846, "w_skip_tri": 0.0025874691652559034, "w_neural_phoneme": 0.10029382918550124, "w_neural_word": 0.10027188686889761, "word_vocab": 12870, "elapsed_s": 592.383679151535, "steps_per_sec": 1447.0638374571408}, {"sent": 50000, "train_loss": 2.2574540297465684, "train_acc": 0.3598308364991757, "eval_loss": 2.134478521220761, "eval_acc": 0.37037037037037035, "ppl": 8.452637486278439, "w_freq": 2.8496173679618738e-05, "w_bigram": 0.014852416017252313, "w_trigram": 1.016712460903424, "w_skip_bi": 8.434557569852601e-05, "w_skip_tri": 0.0013188767600837844, "w_neural_phoneme": 0.10005893420418674, "w_neural_word": 0.10005735364662664, "word_vocab": 12991, "elapsed_s": 611.6337506771088, "steps_per_sec": 1447.1389111868493}, {"sent": 52000, "train_loss": 2.3950161040822078, "train_acc": 0.33680125505738584, "eval_loss": 2.1059195893770775, "eval_acc": 0.37168141592920356, "ppl": 8.214653645550225, "w_freq": 4.540757312370499e-06, "w_bigram": 0.02526359071049371, "w_trigram": 0.9707119674320835, "w_skip_bi": 0.0, "w_skip_tri": 0.014258978681487594, "w_neural_phoneme": 0.10000738103145258, "w_neural_word": 0.10001269113943692, "word_vocab": 13361, "elapsed_s": 636.1322839260101, "steps_per_sec": 1448.5226159456734}, {"sent": 54000, "train_loss": 2.328858525779356, "train_acc": 0.3422616966221414, "eval_loss": 2.128960007239139, "eval_acc": 0.3263157894736842, "ppl": 8.406119959992086, "w_freq": 2.6792867754738535e-05, "w_bigram": 0.020840029076528245, "w_trigram": 0.9863829708761431, "w_skip_bi": 0.0, "w_skip_tri": 0.02320990633742573, "w_neural_phoneme": 0.10002749589528112, "w_neural_word": 0.10001343661949207, "word_vocab": 13527, "elapsed_s": 655.7029104232788, "steps_per_sec": 1448.9031311249023}, {"sent": 56000, "train_loss": 2.2659868448683005, "train_acc": 0.367722918201916, "eval_loss": 2.1963255199229907, "eval_acc": 0.33695652173913043, "ppl": 8.991912119560165, "w_freq": 3.5969161986200666e-06, "w_bigram": 0.045302301300784956, "w_trigram": 1.031739132919099, "w_skip_bi": 0.0, "w_skip_tri": 0.04342132954783599, "w_neural_phoneme": 0.10002166597666615, "w_neural_word": 0.10001300352094959, "word_vocab": 13709, "elapsed_s": 674.3656370639801, "steps_per_sec": 1447.0384408205223}, {"sent": 58000, "train_loss": 2.355835789473543, "train_acc": 0.33464828067725605, "eval_loss": 2.104755077264415, "eval_acc": 0.40860215053763443, "ppl": 8.205093149615719, "w_freq": 7.169439513032591e-05, "w_bigram": 0.03152966930518009, "w_trigram": 1.002858476964674, "w_skip_bi": 0.0006488483900344139, "w_skip_tri": 0.019266814315636382, "w_neural_phoneme": 0.10015921905116675, "w_neural_word": 0.1000373188584095, "word_vocab": 13922, "elapsed_s": 693.854793548584, "steps_per_sec": 1447.6775390752791}, {"sent": 60000, "train_loss": 2.3552519860355616, "train_acc": 0.34625387339530767, "eval_loss": 2.2986235063901375, "eval_acc": 0.29591836734693877, "ppl": 9.960462501282553, "w_freq": 8.640250957099613e-05, "w_bigram": 0.016515337940124657, "w_trigram": 0.9954826845513458, "w_skip_bi": 0.0, "w_skip_tri": 0.016804818988953217, "w_neural_phoneme": 0.10002297566569668, "w_neural_word": 0.1000232084223452, "word_vocab": 14280, "elapsed_s": 717.9318759441376, "steps_per_sec": 1449.4717881574757}], "mixing_weights": {"w_freq": 8.640250957099613e-05, "w_bigram": 0.016515337940124657, "w_trigram": 0.9954826845513458, "w_skip_bi": 0.0, "w_skip_tri": 0.016804818988953217, "w_neural_phoneme": 0.10002297566569668, "w_neural_word": 0.1000232084223452}, "word_vocab": 14280, "elapsed_s": 717.9318759441376, "steps_per_sec": 1449.4717881574757}
|