|
{ |
|
"batch_size": 2, |
|
"best_eval_metric_checkpoint_number": 7, |
|
"best_eval_metric_epoch": 7, |
|
"best_eval_metric_steps": 19390, |
|
"best_eval_metric_value": 0.028326265513896942, |
|
"best_eval_test_metrics": { |
|
"combined": { |
|
"loss": 0.02837251126766205 |
|
}, |
|
"output": { |
|
"bleu": 0.2571483254432678, |
|
"char_error_rate": 2.8023552894592285, |
|
"loss": 0.02837251126766205, |
|
"next_token_perplexity": 11895.607421875, |
|
"perplexity": 31894.921875, |
|
"rouge1_fmeasure": 0.5043888688087463, |
|
"rouge1_precision": 0.35199737548828125, |
|
"rouge1_recall": 0.9673399329185486, |
|
"rouge2_fmeasure": 0.48816248774528503, |
|
"rouge2_precision": 0.3395858705043793, |
|
"rouge2_recall": 0.9494638442993164, |
|
"rougeL_fmeasure": 0.5036171078681946, |
|
"rougeL_precision": 0.35143783688545227, |
|
"rougeL_recall": 0.9660221934318542, |
|
"rougeLsum_fmeasure": 0.5034310817718506, |
|
"rougeLsum_precision": 0.35130569338798523, |
|
"rougeLsum_recall": 0.9656680822372437, |
|
"sequence_accuracy": 0.0, |
|
"token_accuracy": 0.0004706201143562794, |
|
"word_error_rate": 2.537461757659912 |
|
} |
|
}, |
|
"best_eval_train_metrics": { |
|
"combined": { |
|
"loss": 0.010627036914229393 |
|
}, |
|
"output": { |
|
"bleu": 0.19371718168258667, |
|
"char_error_rate": 3.023742914199829, |
|
"loss": 0.010627036914229393, |
|
"next_token_perplexity": 11850.654296875, |
|
"perplexity": 31999.515625, |
|
"rouge1_fmeasure": 0.5099087953567505, |
|
"rouge1_precision": 0.34651726484298706, |
|
"rouge1_recall": 1.0, |
|
"rouge2_fmeasure": 0.5034307837486267, |
|
"rouge2_precision": 0.3408345580101013, |
|
"rouge2_recall": 1.0, |
|
"rougeL_fmeasure": 0.5099087953567505, |
|
"rougeL_precision": 0.34651726484298706, |
|
"rougeL_recall": 1.0, |
|
"rougeLsum_fmeasure": 0.5099087953567505, |
|
"rougeLsum_precision": 0.34651726484298706, |
|
"rougeLsum_recall": 1.0, |
|
"sequence_accuracy": 0.0, |
|
"token_accuracy": 0.0, |
|
"word_error_rate": 2.925373077392578 |
|
} |
|
}, |
|
"best_eval_validation_metrics": { |
|
"combined": { |
|
"loss": 0.028326265513896942 |
|
}, |
|
"output": { |
|
"bleu": 0.2575768232345581, |
|
"char_error_rate": 2.8845763206481934, |
|
"loss": 0.028326265513896942, |
|
"next_token_perplexity": 11894.1640625, |
|
"perplexity": 31888.74609375, |
|
"rouge1_fmeasure": 0.5003785490989685, |
|
"rouge1_precision": 0.3479803204536438, |
|
"rouge1_recall": 0.9642844796180725, |
|
"rouge2_fmeasure": 0.4844965636730194, |
|
"rouge2_precision": 0.3358590602874756, |
|
"rouge2_recall": 0.946923017501831, |
|
"rougeL_fmeasure": 0.49971580505371094, |
|
"rougeL_precision": 0.34750500321388245, |
|
"rougeL_recall": 0.9631230235099792, |
|
"rougeLsum_fmeasure": 0.499582976102829, |
|
"rougeLsum_precision": 0.34741097688674927, |
|
"rougeLsum_recall": 0.9628673195838928, |
|
"sequence_accuracy": 0.0, |
|
"token_accuracy": 0.00046722154365852475, |
|
"word_error_rate": 2.64115047454834 |
|
} |
|
}, |
|
"best_increase_batch_size_eval_metric": Infinity, |
|
"checkpoint_number": 7, |
|
"epoch": 8, |
|
"last_improvement_steps": 0, |
|
"last_increase_batch_size": 0, |
|
"last_increase_batch_size_eval_metric_improvement": 0, |
|
"last_increase_batch_size_steps": 0, |
|
"last_learning_rate_reduction": 0, |
|
"last_learning_rate_reduction_steps": 0, |
|
"learning_rate": 0.0001, |
|
"num_increases_batch_size": 0, |
|
"num_reductions_learning_rate": 0, |
|
"steps": 19390, |
|
"test_metrics": { |
|
"combined": { |
|
"loss": [ |
|
[ |
|
1, |
|
2770, |
|
0.056570280343294144 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.03862074762582779 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.0322396382689476 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.029598653316497803 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.029046066105365753 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.029180288314819336 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.02837251126766205 |
|
] |
|
] |
|
}, |
|
"output": { |
|
"bleu": [ |
|
[ |
|
1, |
|
2770, |
|
0.2881736159324646 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.30035901069641113 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.29325926303863525 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.2912749648094177 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.29473328590393066 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.2825150489807129 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.2571483254432678 |
|
] |
|
], |
|
"char_error_rate": [ |
|
[ |
|
1, |
|
2770, |
|
3.7703747749328613 |
|
], |
|
[ |
|
2, |
|
5540, |
|
2.7834644317626953 |
|
], |
|
[ |
|
3, |
|
8310, |
|
2.8107423782348633 |
|
], |
|
[ |
|
4, |
|
11080, |
|
2.7512333393096924 |
|
], |
|
[ |
|
5, |
|
13850, |
|
2.7991256713867188 |
|
], |
|
[ |
|
6, |
|
16620, |
|
2.74367618560791 |
|
], |
|
[ |
|
7, |
|
19390, |
|
2.8023552894592285 |
|
] |
|
], |
|
"loss": [ |
|
[ |
|
1, |
|
2770, |
|
0.056570280343294144 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.03862074762582779 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.0322396382689476 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.029598653316497803 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.029046066105365753 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.029180288314819336 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.02837251126766205 |
|
] |
|
], |
|
"next_token_perplexity": [ |
|
[ |
|
1, |
|
2770, |
|
12036.4189453125 |
|
], |
|
[ |
|
2, |
|
5540, |
|
11973.7529296875 |
|
], |
|
[ |
|
3, |
|
8310, |
|
11931.3291015625 |
|
], |
|
[ |
|
4, |
|
11080, |
|
11918.92578125 |
|
], |
|
[ |
|
5, |
|
13850, |
|
11906.0029296875 |
|
], |
|
[ |
|
6, |
|
16620, |
|
11907.7744140625 |
|
], |
|
[ |
|
7, |
|
19390, |
|
11895.607421875 |
|
] |
|
], |
|
"perplexity": [ |
|
[ |
|
1, |
|
2770, |
|
31891.08984375 |
|
], |
|
[ |
|
2, |
|
5540, |
|
31882.876953125 |
|
], |
|
[ |
|
3, |
|
8310, |
|
31891.08984375 |
|
], |
|
[ |
|
4, |
|
11080, |
|
31888.59375 |
|
], |
|
[ |
|
5, |
|
13850, |
|
31894.5859375 |
|
], |
|
[ |
|
6, |
|
16620, |
|
31886.49609375 |
|
], |
|
[ |
|
7, |
|
19390, |
|
31894.921875 |
|
] |
|
], |
|
"rouge1_fmeasure": [ |
|
[ |
|
1, |
|
2770, |
|
0.45583832263946533 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.46054157614707947 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.46443498134613037 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.4832415282726288 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.4780387878417969 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.49345338344573975 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.5043888688087463 |
|
] |
|
], |
|
"rouge1_precision": [ |
|
[ |
|
1, |
|
2770, |
|
0.31005945801734924 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.31317880749702454 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.3162654638290405 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.33289361000061035 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.3283507823944092 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.3420336842536926 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.35199737548828125 |
|
] |
|
], |
|
"rouge1_recall": [ |
|
[ |
|
1, |
|
2770, |
|
0.9494807720184326 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.9610881805419922 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.9647314548492432 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.9663017988204956 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.9644144177436829 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.9666685461997986 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.9673399329185486 |
|
] |
|
], |
|
"rouge2_fmeasure": [ |
|
[ |
|
1, |
|
2770, |
|
0.4300532042980194 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.44140955805778503 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.4470669627189636 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.46725624799728394 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.4617525637149811 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.47793200612068176 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.48816248774528503 |
|
] |
|
], |
|
"rouge2_precision": [ |
|
[ |
|
1, |
|
2770, |
|
0.29176947474479675 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.29934147000312805 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.3036028742790222 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.32093188166618347 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.31624865531921387 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.33026161789894104 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.3395858705043793 |
|
] |
|
], |
|
"rouge2_recall": [ |
|
[ |
|
1, |
|
2770, |
|
0.9086052179336548 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.9347748756408691 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.9421030879020691 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.9477611780166626 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.9448394179344177 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.9496001601219177 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.9494638442993164 |
|
] |
|
], |
|
"rougeL_fmeasure": [ |
|
[ |
|
1, |
|
2770, |
|
0.45452964305877686 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.45963969826698303 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.4637087881565094 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.48262497782707214 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.4772895872592926 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.4928056597709656 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.5036171078681946 |
|
] |
|
], |
|
"rougeL_precision": [ |
|
[ |
|
1, |
|
2770, |
|
0.30914732813835144 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.3125472366809845 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.3157593905925751 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.33245277404785156 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.3278239071369171 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.3415752649307251 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.35143783688545227 |
|
] |
|
], |
|
"rougeL_recall": [ |
|
[ |
|
1, |
|
2770, |
|
0.9469588994979858 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.9593668580055237 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.963313639163971 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.9652055501937866 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.9630056023597717 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.9654869437217712 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.9660221934318542 |
|
] |
|
], |
|
"rougeLsum_fmeasure": [ |
|
[ |
|
1, |
|
2770, |
|
0.4544520974159241 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.4593540132045746 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.463616281747818 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.4826125502586365 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.47729507088661194 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.4928112030029297 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.5034310817718506 |
|
] |
|
], |
|
"rougeLsum_precision": [ |
|
[ |
|
1, |
|
2770, |
|
0.30910027027130127 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.31235170364379883 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.31568989157676697 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.33244553208351135 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.3278290927410126 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.3415812849998474 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.35130569338798523 |
|
] |
|
], |
|
"rougeLsum_recall": [ |
|
[ |
|
1, |
|
2770, |
|
0.9467605948448181 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.9587725400924683 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.9631702303886414 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.9651756882667542 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.9630022644996643 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.9654796719551086 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.9656680822372437 |
|
] |
|
], |
|
"sequence_accuracy": [ |
|
[ |
|
1, |
|
2770, |
|
0.0 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.0 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.0 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.0 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.0 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.0 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.0 |
|
] |
|
], |
|
"token_accuracy": [ |
|
[ |
|
1, |
|
2770, |
|
0.00046411342918872833 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.00046866878983564675 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.00046727192238904536 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.0004687990585807711 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.0004691674548666924 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.0004687117470894009 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.0004706201143562794 |
|
] |
|
], |
|
"word_error_rate": [ |
|
[ |
|
1, |
|
2770, |
|
1.9269943237304688 |
|
], |
|
[ |
|
2, |
|
5540, |
|
1.9473299980163574 |
|
], |
|
[ |
|
3, |
|
8310, |
|
2.0793614387512207 |
|
], |
|
[ |
|
4, |
|
11080, |
|
2.1502275466918945 |
|
], |
|
[ |
|
5, |
|
13850, |
|
2.092580556869507 |
|
], |
|
[ |
|
6, |
|
16620, |
|
2.2052907943725586 |
|
], |
|
[ |
|
7, |
|
19390, |
|
2.537461757659912 |
|
] |
|
] |
|
} |
|
}, |
|
"train_metrics": { |
|
"combined": { |
|
"loss": [ |
|
[ |
|
1, |
|
2770, |
|
0.16293130815029144 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.012954902835190296 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.03785233944654465 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.020756859332323074 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.011303349398076534 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.012777533382177353 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.010627036914229393 |
|
] |
|
] |
|
}, |
|
"output": { |
|
"bleu": [ |
|
[ |
|
1, |
|
2770, |
|
0.20123623311519623 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.3794978857040405 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.2412903904914856 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.36655688285827637 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.1585085093975067 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.2643570303916931 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.19371718168258667 |
|
] |
|
], |
|
"char_error_rate": [ |
|
[ |
|
1, |
|
2770, |
|
7.55555534362793 |
|
], |
|
[ |
|
2, |
|
5540, |
|
2.9573256969451904 |
|
], |
|
[ |
|
3, |
|
8310, |
|
4.503154754638672 |
|
], |
|
[ |
|
4, |
|
11080, |
|
3.002485513687134 |
|
], |
|
[ |
|
5, |
|
13850, |
|
2.344075918197632 |
|
], |
|
[ |
|
6, |
|
16620, |
|
4.294294357299805 |
|
], |
|
[ |
|
7, |
|
19390, |
|
3.023742914199829 |
|
] |
|
], |
|
"loss": [ |
|
[ |
|
1, |
|
2770, |
|
0.16293130815029144 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.012954902835190296 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.03785233944654465 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.020756859332323074 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.011303349398076534 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.012777533382177353 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.010627036914229393 |
|
] |
|
], |
|
"next_token_perplexity": [ |
|
[ |
|
1, |
|
2770, |
|
12304.353515625 |
|
], |
|
[ |
|
2, |
|
5540, |
|
11869.2724609375 |
|
], |
|
[ |
|
3, |
|
8310, |
|
11982.583984375 |
|
], |
|
[ |
|
4, |
|
11080, |
|
11896.8330078125 |
|
], |
|
[ |
|
5, |
|
13850, |
|
11849.265625 |
|
], |
|
[ |
|
6, |
|
16620, |
|
11852.1123046875 |
|
], |
|
[ |
|
7, |
|
19390, |
|
11850.654296875 |
|
] |
|
], |
|
"perplexity": [ |
|
[ |
|
1, |
|
2770, |
|
31993.228515625 |
|
], |
|
[ |
|
2, |
|
5540, |
|
31764.84765625 |
|
], |
|
[ |
|
3, |
|
8310, |
|
31999.63671875 |
|
], |
|
[ |
|
4, |
|
11080, |
|
31889.169921875 |
|
], |
|
[ |
|
5, |
|
13850, |
|
31996.4921875 |
|
], |
|
[ |
|
6, |
|
16620, |
|
31996.15625 |
|
], |
|
[ |
|
7, |
|
19390, |
|
31999.515625 |
|
] |
|
], |
|
"rouge1_fmeasure": [ |
|
[ |
|
1, |
|
2770, |
|
0.2195121943950653 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.4458082616329193 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.4645876884460449 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.5988943576812744 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.5516705513000488 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.45763128995895386 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.5099087953567505 |
|
] |
|
], |
|
"rouge1_precision": [ |
|
[ |
|
1, |
|
2770, |
|
0.140625 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.2905769348144531 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.3042147159576416 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.43105676770210266 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.40321463346481323 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.2989655137062073 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.34651726484298706 |
|
] |
|
], |
|
"rouge1_recall": [ |
|
[ |
|
1, |
|
2770, |
|
0.5 |
|
], |
|
[ |
|
2, |
|
5540, |
|
1.0 |
|
], |
|
[ |
|
3, |
|
8310, |
|
1.0 |
|
], |
|
[ |
|
4, |
|
11080, |
|
1.0 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.9948453903198242 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.9900000095367432 |
|
], |
|
[ |
|
7, |
|
19390, |
|
1.0 |
|
] |
|
], |
|
"rouge2_fmeasure": [ |
|
[ |
|
1, |
|
2770, |
|
0.20370370149612427 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.438376247882843 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.451366126537323 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.5951134562492371 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.5413873195648193 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.4451362192630768 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.5034307837486267 |
|
] |
|
], |
|
"rouge2_precision": [ |
|
[ |
|
1, |
|
2770, |
|
0.12992125749588013 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.28464847803115845 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.29411762952804565 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.4273049831390381 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.39483463764190674 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.2894570827484131 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.3408345580101013 |
|
] |
|
], |
|
"rouge2_recall": [ |
|
[ |
|
1, |
|
2770, |
|
0.4714285731315613 |
|
], |
|
[ |
|
2, |
|
5540, |
|
1.0 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.9893617033958435 |
|
], |
|
[ |
|
4, |
|
11080, |
|
1.0 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.9895833730697632 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.9795918464660645 |
|
], |
|
[ |
|
7, |
|
19390, |
|
1.0 |
|
] |
|
], |
|
"rougeL_fmeasure": [ |
|
[ |
|
1, |
|
2770, |
|
0.2195121943950653 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.4458082616329193 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.4645876884460449 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.5949573516845703 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.5516705513000488 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.45763128995895386 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.5099087953567505 |
|
] |
|
], |
|
"rougeL_precision": [ |
|
[ |
|
1, |
|
2770, |
|
0.140625 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.2905769348144531 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.3042147159576416 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.428098201751709 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.40321463346481323 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.2989655137062073 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.34651726484298706 |
|
] |
|
], |
|
"rougeL_recall": [ |
|
[ |
|
1, |
|
2770, |
|
0.5 |
|
], |
|
[ |
|
2, |
|
5540, |
|
1.0 |
|
], |
|
[ |
|
3, |
|
8310, |
|
1.0 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.9941176176071167 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.9948453903198242 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.9900000095367432 |
|
], |
|
[ |
|
7, |
|
19390, |
|
1.0 |
|
] |
|
], |
|
"rougeLsum_fmeasure": [ |
|
[ |
|
1, |
|
2770, |
|
0.2195121943950653 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.4458082616329193 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.4645876884460449 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.5949573516845703 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.5516705513000488 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.45763128995895386 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.5099087953567505 |
|
] |
|
], |
|
"rougeLsum_precision": [ |
|
[ |
|
1, |
|
2770, |
|
0.140625 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.2905769348144531 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.3042147159576416 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.428098201751709 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.40321463346481323 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.2989655137062073 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.34651726484298706 |
|
] |
|
], |
|
"rougeLsum_recall": [ |
|
[ |
|
1, |
|
2770, |
|
0.5 |
|
], |
|
[ |
|
2, |
|
5540, |
|
1.0 |
|
], |
|
[ |
|
3, |
|
8310, |
|
1.0 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.9941176176071167 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.9948453903198242 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.9900000095367432 |
|
], |
|
[ |
|
7, |
|
19390, |
|
1.0 |
|
] |
|
], |
|
"sequence_accuracy": [ |
|
[ |
|
1, |
|
2770, |
|
0.0 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.0 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.0 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.0 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.0 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.0 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.0 |
|
] |
|
], |
|
"token_accuracy": [ |
|
[ |
|
1, |
|
2770, |
|
0.0 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.0011185682378709316 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.0 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.0004681647988036275 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.0 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.0 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.0 |
|
] |
|
], |
|
"word_error_rate": [ |
|
[ |
|
1, |
|
2770, |
|
4.363636493682861 |
|
], |
|
[ |
|
2, |
|
5540, |
|
2.318840503692627 |
|
], |
|
[ |
|
3, |
|
8310, |
|
2.492063522338867 |
|
], |
|
[ |
|
4, |
|
11080, |
|
1.4741379022598267 |
|
], |
|
[ |
|
5, |
|
13850, |
|
1.7572815418243408 |
|
], |
|
[ |
|
6, |
|
16620, |
|
2.890625 |
|
], |
|
[ |
|
7, |
|
19390, |
|
2.925373077392578 |
|
] |
|
] |
|
} |
|
}, |
|
"tune_checkpoint_num": 0, |
|
"validation_metrics": { |
|
"combined": { |
|
"loss": [ |
|
[ |
|
1, |
|
2770, |
|
0.056642960757017136 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.03854465112090111 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.03133770078420639 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.029445933178067207 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.0286291241645813 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.028967903926968575 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.028326265513896942 |
|
] |
|
] |
|
}, |
|
"output": { |
|
"bleu": [ |
|
[ |
|
1, |
|
2770, |
|
0.294668972492218 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.30392342805862427 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.2956589162349701 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.2938247621059418 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.2953544855117798 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.2841779291629791 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.2575768232345581 |
|
] |
|
], |
|
"char_error_rate": [ |
|
[ |
|
1, |
|
2770, |
|
3.8609488010406494 |
|
], |
|
[ |
|
2, |
|
5540, |
|
2.8585927486419678 |
|
], |
|
[ |
|
3, |
|
8310, |
|
2.888094663619995 |
|
], |
|
[ |
|
4, |
|
11080, |
|
2.827108383178711 |
|
], |
|
[ |
|
5, |
|
13850, |
|
2.8774003982543945 |
|
], |
|
[ |
|
6, |
|
16620, |
|
2.8187196254730225 |
|
], |
|
[ |
|
7, |
|
19390, |
|
2.8845763206481934 |
|
] |
|
], |
|
"loss": [ |
|
[ |
|
1, |
|
2770, |
|
0.056642960757017136 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.03854465112090111 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.03133770078420639 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.029445933178067207 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.0286291241645813 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.028967903926968575 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.028326265513896942 |
|
] |
|
], |
|
"next_token_perplexity": [ |
|
[ |
|
1, |
|
2770, |
|
12037.177734375 |
|
], |
|
[ |
|
2, |
|
5540, |
|
11975.810546875 |
|
], |
|
[ |
|
3, |
|
8310, |
|
11929.7431640625 |
|
], |
|
[ |
|
4, |
|
11080, |
|
11916.7890625 |
|
], |
|
[ |
|
5, |
|
13850, |
|
11905.4638671875 |
|
], |
|
[ |
|
6, |
|
16620, |
|
11907.5517578125 |
|
], |
|
[ |
|
7, |
|
19390, |
|
11894.1640625 |
|
] |
|
], |
|
"perplexity": [ |
|
[ |
|
1, |
|
2770, |
|
31882.267578125 |
|
], |
|
[ |
|
2, |
|
5540, |
|
31876.583984375 |
|
], |
|
[ |
|
3, |
|
8310, |
|
31884.275390625 |
|
], |
|
[ |
|
4, |
|
11080, |
|
31881.873046875 |
|
], |
|
[ |
|
5, |
|
13850, |
|
31888.412109375 |
|
], |
|
[ |
|
6, |
|
16620, |
|
31880.4453125 |
|
], |
|
[ |
|
7, |
|
19390, |
|
31888.74609375 |
|
] |
|
], |
|
"rouge1_fmeasure": [ |
|
[ |
|
1, |
|
2770, |
|
0.4522210359573364 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.4561161994934082 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.46019619703292847 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.47886744141578674 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.47372379899024963 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.48917827010154724 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.5003785490989685 |
|
] |
|
], |
|
"rouge1_precision": [ |
|
[ |
|
1, |
|
2770, |
|
0.30646950006484985 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.309090793132782 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.31218209862709045 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.3286076486110687 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.32413795590400696 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.3378482758998871 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.3479803204536438 |
|
] |
|
], |
|
"rouge1_recall": [ |
|
[ |
|
1, |
|
2770, |
|
0.947241485118866 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.9578168988227844 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.9620707035064697 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.963258683681488 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.9619985818862915 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.9631357192993164 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.9642844796180725 |
|
] |
|
], |
|
"rouge2_fmeasure": [ |
|
[ |
|
1, |
|
2770, |
|
0.42693406343460083 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.4367271065711975 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.4432367980480194 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.46300628781318665 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.45774954557418823 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.4741453528404236 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.4844965636730194 |
|
] |
|
], |
|
"rouge2_precision": [ |
|
[ |
|
1, |
|
2770, |
|
0.28853029012680054 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.2951086163520813 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.29979175329208374 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.3167739510536194 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.31229183077812195 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.32645225524902344 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.3358590602874756 |
|
] |
|
], |
|
"rouge2_recall": [ |
|
[ |
|
1, |
|
2770, |
|
0.9076516628265381 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.9309775829315186 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.9407065510749817 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.9448210000991821 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.9431052803993225 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.9469495415687561 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.946923017501831 |
|
] |
|
], |
|
"rougeL_fmeasure": [ |
|
[ |
|
1, |
|
2770, |
|
0.4507341980934143 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.4551938772201538 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.4593889117240906 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.47822490334510803 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.47310617566108704 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.4886751174926758 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.49971580505371094 |
|
] |
|
], |
|
"rougeL_precision": [ |
|
[ |
|
1, |
|
2770, |
|
0.3054444193840027 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.30845171213150024 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.3116239011287689 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.3281573951244354 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.32370176911354065 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.3374961018562317 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.34750500321388245 |
|
] |
|
], |
|
"rougeL_recall": [ |
|
[ |
|
1, |
|
2770, |
|
0.9443382620811462 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.95603346824646 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.9604740738868713 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.962070882320404 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.9608785510063171 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.9621948599815369 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.9631230235099792 |
|
] |
|
], |
|
"rougeLsum_fmeasure": [ |
|
[ |
|
1, |
|
2770, |
|
0.4507194757461548 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.4550528824329376 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.45933279395103455 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.47812482714653015 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.4730019271373749 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.48857295513153076 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.499582976102829 |
|
] |
|
], |
|
"rougeLsum_precision": [ |
|
[ |
|
1, |
|
2770, |
|
0.30543628334999084 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.30835309624671936 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.31157732009887695 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.32808494567871094 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.3236253559589386 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.33741870522499084 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.34741097688674927 |
|
] |
|
], |
|
"rougeLsum_recall": [ |
|
[ |
|
1, |
|
2770, |
|
0.9442805647850037 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.9557678699493408 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.9604008197784424 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.9618760943412781 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.9606844186782837 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.9620285034179688 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.9628673195838928 |
|
] |
|
], |
|
"sequence_accuracy": [ |
|
[ |
|
1, |
|
2770, |
|
0.0 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.0 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.0 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.0 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.0 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.0 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.0 |
|
] |
|
], |
|
"token_accuracy": [ |
|
[ |
|
1, |
|
2770, |
|
0.00046871474478393793 |
|
], |
|
[ |
|
2, |
|
5540, |
|
0.00047089491272345185 |
|
], |
|
[ |
|
3, |
|
8310, |
|
0.0004673982912208885 |
|
], |
|
[ |
|
4, |
|
11080, |
|
0.0004686332249548286 |
|
], |
|
[ |
|
5, |
|
13850, |
|
0.0004653023788705468 |
|
], |
|
[ |
|
6, |
|
16620, |
|
0.0004659830592572689 |
|
], |
|
[ |
|
7, |
|
19390, |
|
0.00046722154365852475 |
|
] |
|
], |
|
"word_error_rate": [ |
|
[ |
|
1, |
|
2770, |
|
1.997908115386963 |
|
], |
|
[ |
|
2, |
|
5540, |
|
2.019251585006714 |
|
], |
|
[ |
|
3, |
|
8310, |
|
2.154338836669922 |
|
], |
|
[ |
|
4, |
|
11080, |
|
2.2291877269744873 |
|
], |
|
[ |
|
5, |
|
13850, |
|
2.173459768295288 |
|
], |
|
[ |
|
6, |
|
16620, |
|
2.2866809368133545 |
|
], |
|
[ |
|
7, |
|
19390, |
|
2.64115047454834 |
|
] |
|
] |
|
} |
|
} |
|
} |