|
187 | 187 | { |
188 | 188 | "data": { |
189 | 189 | "application/vnd.jupyter.widget-view+json": { |
190 | | - "model_id": "2ffb58863ba0436296433ed981dc49db", |
| 190 | + "model_id": "e21c690b98854a098fc213c986cdae58", |
191 | 191 | "version_major": 2, |
192 | 192 | "version_minor": 0 |
193 | 193 | }, |
|
261 | 261 | " translated_texts = list()\n", |
262 | 262 | " for batch in get_batches(texts,batch_size=model.generation_config.batch_size):\n", |
263 | 263 | " translated_tokens = model.generate(\n", |
264 | | - " top_k=10,\n", |
265 | | - " top_p=0.75,\n", |
| 264 | + " top_k=100,\n", |
| 265 | + " top_p=0.8,\n", |
266 | 266 | " do_sample=True,\n", |
267 | 267 | " temperature=0.9,\n", |
268 | 268 | " **tokenizer(batch, return_tensors=\"pt\",padding=True),\n", |
|
279 | 279 | { |
280 | 280 | "data": { |
281 | 281 | "application/vnd.jupyter.widget-view+json": { |
282 | | - "model_id": "876bf053074348b9822ecc814931ad2c", |
| 282 | + "model_id": "a27619cff1ed4b7bb054e45a9df98465", |
283 | 283 | "version_major": 2, |
284 | 284 | "version_minor": 0 |
285 | 285 | }, |
|
315 | 315 | { |
316 | 316 | "data": { |
317 | 317 | "text/plain": [ |
318 | | - "{'score': 18.35116938665859,\n", |
319 | | - " 'counts': [129, 61, 33, 19],\n", |
320 | | - " 'totals': [268, 258, 248, 239],\n", |
321 | | - " 'precisions': [48.134328358208954,\n", |
322 | | - " 23.643410852713178,\n", |
323 | | - " 13.306451612903226,\n", |
324 | | - " 7.949790794979079],\n", |
325 | | - " 'bp': 0.9851854581626466,\n", |
326 | | - " 'sys_len': 268,\n", |
| 318 | + "{'score': 16.432731874855403,\n", |
| 319 | + " 'counts': [132, 60, 31, 14],\n", |
| 320 | + " 'totals': [277, 267, 257, 248],\n", |
| 321 | + " 'precisions': [47.65342960288809,\n", |
| 322 | + " 22.471910112359552,\n", |
| 323 | + " 12.062256809338521,\n", |
| 324 | + " 5.645161290322581],\n", |
| 325 | + " 'bp': 1.0,\n", |
| 326 | + " 'sys_len': 277,\n", |
327 | 327 | " 'ref_len': 272}" |
328 | 328 | ] |
329 | 329 | }, |
|
377 | 377 | { |
378 | 378 | "data": { |
379 | 379 | "application/vnd.jupyter.widget-view+json": { |
380 | | - "model_id": "73dd3bbe56c14f7e9f847e2f5afb780b", |
| 380 | + "model_id": "8d09a5ef6f15448295ee3a91e744e854", |
381 | 381 | "version_major": 2, |
382 | 382 | "version_minor": 0 |
383 | 383 | }, |
|
451 | 451 | " translated_texts = list()\n", |
452 | 452 | " for batch in get_batches(texts,batch_size=model.generation_config.batch_size):\n", |
453 | 453 | " translated_tokens = model.generate(\n", |
454 | | - " top_k=10,\n", |
455 | | - " top_p=0.75,\n", |
| 454 | + " top_k=100,\n", |
| 455 | + " top_p=0.8,\n", |
456 | 456 | " num_beams=4,\n", |
457 | 457 | " do_sample=True,\n", |
458 | 458 | " temperature=0.9,\n", |
|
471 | 471 | { |
472 | 472 | "data": { |
473 | 473 | "application/vnd.jupyter.widget-view+json": { |
474 | | - "model_id": "266b195603364b3a9d407108b786065d", |
| 474 | + "model_id": "3ed5415206ab4e7db92c1f78ac4662a0", |
475 | 475 | "version_major": 2, |
476 | 476 | "version_minor": 0 |
477 | 477 | }, |
|
507 | 507 | { |
508 | 508 | "data": { |
509 | 509 | "text/plain": [ |
510 | | - "{'score': 21.323465966136123,\n", |
511 | | - " 'counts': [134, 68, 41, 24],\n", |
| 510 | + "{'score': 21.792077490044267,\n", |
| 511 | + " 'counts': [135, 69, 42, 25],\n", |
512 | 512 | " 'totals': [262, 252, 242, 233],\n", |
513 | | - " 'precisions': [51.14503816793893,\n", |
514 | | - " 26.984126984126984,\n", |
515 | | - " 16.94214876033058,\n", |
516 | | - " 10.300429184549357],\n", |
| 513 | + " 'precisions': [51.52671755725191,\n", |
| 514 | + " 27.38095238095238,\n", |
| 515 | + " 17.355371900826448,\n", |
| 516 | + " 10.729613733905579],\n", |
517 | 517 | " 'bp': 0.9625512774839297,\n", |
518 | 518 | " 'sys_len': 262,\n", |
519 | 519 | " 'ref_len': 272}" |
|
653 | 653 | { |
654 | 654 | "data": { |
655 | 655 | "application/vnd.jupyter.widget-view+json": { |
656 | | - "model_id": "25ce4ae39bfc43b1a3fbd4c8fd83262e", |
| 656 | + "model_id": "d0101acb14924ebc9b74f899d1a28095", |
657 | 657 | "version_major": 2, |
658 | 658 | "version_minor": 0 |
659 | 659 | }, |
|
719 | 719 | }, |
720 | 720 | { |
721 | 721 | "cell_type": "code", |
722 | | - "execution_count": 24, |
| 722 | + "execution_count": 36, |
723 | 723 | "metadata": {}, |
724 | 724 | "outputs": [], |
725 | 725 | "source": [ |
726 | 726 | "greedy_sequences_generator = GreedyGenerator(\n", |
727 | 727 | " use_tqdm=True,\n", |
728 | 728 | " temperature=0.9,\n", |
729 | | - " top_k_sampling=10,\n", |
730 | 729 | " sort_samples=True,\n", |
731 | | - " top_p_sampling=0.75,\n", |
| 730 | + " top_k_sampling=100,\n", |
| 731 | + " top_p_sampling=0.8,\n", |
732 | 732 | " device=model.device,\n", |
733 | 733 | " multinomial_sampling=True,\n", |
734 | 734 | " generation_forward=generate,\n", |
|
741 | 741 | }, |
742 | 742 | { |
743 | 743 | "cell_type": "code", |
744 | | - "execution_count": 25, |
| 744 | + "execution_count": 37, |
745 | 745 | "metadata": {}, |
746 | 746 | "outputs": [ |
747 | 747 | { |
748 | 748 | "data": { |
749 | 749 | "application/vnd.jupyter.widget-view+json": { |
750 | | - "model_id": "cf2eef68941942f5aade85c0567aa0d2", |
| 750 | + "model_id": "e2c36cc8e29a4cd6bca9f30164f0a071", |
751 | 751 | "version_major": 2, |
752 | 752 | "version_minor": 0 |
753 | 753 | }, |
|
764 | 764 | "(10, 10, 10)" |
765 | 765 | ] |
766 | 766 | }, |
767 | | - "execution_count": 25, |
| 767 | + "execution_count": 37, |
768 | 768 | "metadata": {}, |
769 | 769 | "output_type": "execute_result" |
770 | 770 | } |
|
777 | 777 | }, |
778 | 778 | { |
779 | 779 | "cell_type": "code", |
780 | | - "execution_count": 26, |
| 780 | + "execution_count": 38, |
781 | 781 | "metadata": {}, |
782 | 782 | "outputs": [ |
783 | 783 | { |
784 | 784 | "data": { |
785 | 785 | "text/plain": [ |
786 | | - "{'score': 9.729423488181931,\n", |
787 | | - " 'counts': [109, 47, 19, 4],\n", |
788 | | - " 'totals': [266, 256, 246, 237],\n", |
789 | | - " 'precisions': [40.97744360902256,\n", |
790 | | - " 18.359375,\n", |
791 | | - " 7.723577235772358,\n", |
792 | | - " 1.6877637130801688],\n", |
793 | | - " 'bp': 0.9776961023999414,\n", |
794 | | - " 'sys_len': 266,\n", |
| 786 | + "{'score': 16.963549254009152,\n", |
| 787 | + " 'counts': [133, 61, 31, 16],\n", |
| 788 | + " 'totals': [279, 269, 259, 250],\n", |
| 789 | + " 'precisions': [47.67025089605735, 22.676579925650557, 11.96911196911197, 6.4],\n", |
| 790 | + " 'bp': 1.0,\n", |
| 791 | + " 'sys_len': 279,\n", |
795 | 792 | " 'ref_len': 272}" |
796 | 793 | ] |
797 | 794 | }, |
798 | | - "execution_count": 26, |
| 795 | + "execution_count": 38, |
799 | 796 | "metadata": {}, |
800 | 797 | "output_type": "execute_result" |
801 | 798 | } |
|
839 | 836 | { |
840 | 837 | "data": { |
841 | 838 | "application/vnd.jupyter.widget-view+json": { |
842 | | - "model_id": "5eef67a1e84e4349a17cc22558cced89", |
| 839 | + "model_id": "ce3f41dcd2344d669478305a523a112d", |
843 | 840 | "version_major": 2, |
844 | 841 | "version_minor": 0 |
845 | 842 | }, |
|
905 | 902 | }, |
906 | 903 | { |
907 | 904 | "cell_type": "code", |
908 | | - "execution_count": 30, |
| 905 | + "execution_count": 33, |
909 | 906 | "metadata": {}, |
910 | 907 | "outputs": [], |
911 | 908 | "source": [ |
|
914 | 911 | " use_tqdm=True,\n", |
915 | 912 | " temperature=0.9,\n", |
916 | 913 | " sort_samples=True,\n", |
917 | | - " top_k_sampling=10,\n", |
| 914 | + " top_k_sampling=100,\n", |
918 | 915 | " length_penalty=0.6,\n", |
919 | | - " top_p_sampling=0.75,\n", |
| 916 | + " top_p_sampling=0.8,\n", |
920 | 917 | " device=model.device,\n", |
921 | 918 | " multinomial_sampling=True,\n", |
922 | 919 | " generation_forward=generate,\n", |
|
929 | 926 | }, |
930 | 927 | { |
931 | 928 | "cell_type": "code", |
932 | | - "execution_count": 31, |
| 929 | + "execution_count": 34, |
933 | 930 | "metadata": {}, |
934 | 931 | "outputs": [ |
935 | 932 | { |
936 | 933 | "data": { |
937 | 934 | "application/vnd.jupyter.widget-view+json": { |
938 | | - "model_id": "3605ce4f5720498c911d50be793b674f", |
| 935 | + "model_id": "eb933bc8db5f4efc9491dfecb620c4e0", |
939 | 936 | "version_major": 2, |
940 | 937 | "version_minor": 0 |
941 | 938 | }, |
|
952 | 949 | "(10, 10, 10)" |
953 | 950 | ] |
954 | 951 | }, |
955 | | - "execution_count": 31, |
| 952 | + "execution_count": 34, |
956 | 953 | "metadata": {}, |
957 | 954 | "output_type": "execute_result" |
958 | 955 | } |
|
965 | 962 | }, |
966 | 963 | { |
967 | 964 | "cell_type": "code", |
968 | | - "execution_count": 32, |
| 965 | + "execution_count": 35, |
969 | 966 | "metadata": {}, |
970 | 967 | "outputs": [ |
971 | 968 | { |
972 | 969 | "data": { |
973 | 970 | "text/plain": [ |
974 | | - "{'score': 20.27713611022614,\n", |
975 | | - " 'counts': [135, 68, 38, 21],\n", |
976 | | - " 'totals': [259, 249, 239, 230],\n", |
977 | | - " 'precisions': [52.12355212355212,\n", |
978 | | - " 27.309236947791163,\n", |
979 | | - " 15.899581589958158,\n", |
980 | | - " 9.130434782608695],\n", |
981 | | - " 'bp': 0.951045807200927,\n", |
982 | | - " 'sys_len': 259,\n", |
| 971 | + "{'score': 20.760619867696708,\n", |
| 972 | + " 'counts': [136, 69, 39, 22],\n", |
| 973 | + " 'totals': [260, 250, 240, 231],\n", |
| 974 | + " 'precisions': [52.30769230769231, 27.6, 16.25, 9.523809523809524],\n", |
| 975 | + " 'bp': 0.954895043959762,\n", |
| 976 | + " 'sys_len': 260,\n", |
983 | 977 | " 'ref_len': 272}" |
984 | 978 | ] |
985 | 979 | }, |
986 | | - "execution_count": 32, |
| 980 | + "execution_count": 35, |
987 | 981 | "metadata": {}, |
988 | 982 | "output_type": "execute_result" |
989 | 983 | } |
|
0 commit comments