Upload v14p11/ckpt_022000.json with huggingface_hub

v14p11/ckpt_022000.json

@@ -0,0 +1 @@
+{"sent": 22000, "total_steps": 357656, "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}], "mixing_weights": {"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}