Diya-Roshan commited on
Commit
6c595e3
·
verified ·
1 Parent(s): cdb4b16

Delete Kanmodel1

Browse files
Kanmodel1/config.json DELETED
@@ -1,1446 +0,0 @@
1
- {
2
- "_name_or_path": "../input/kannada-xlm/model111/checkpoint-118-epoch-2",
3
- "accumulate_gradients": 4,
4
- "ae_steps": [],
5
- "amp": 2,
6
- "architectures": [
7
- "XLMForMultiLabelSequenceClassification"
8
- ],
9
- "asm": false,
10
- "attention_dropout": 0.1,
11
- "batch_size": 16,
12
- "beam_size": 1,
13
- "bos_index": 0,
14
- "bos_token_id": 0,
15
- "bptt": 256,
16
- "bt_src_langs": [],
17
- "bt_steps": [],
18
- "causal": false,
19
- "clip_grad_norm": 1.0,
20
- "clm_steps": [],
21
- "command": "python /private/home/aconneau/workdir/xlm_17_100_big.3/2019_08_10_19_23_42/train.py --n_heads 16 --bt_steps '' --max_vocab 200000 --word_mask_keep_rand '0.8,0.1,0.1' --use_lang_emb false --data_path '/private/home/aconneau/projects/XLM/data/wiki/100/175k' --save_periodic 0 --max_len 200 --bptt 256 --ae_steps '' --fp16 true --share_inout_emb true --sinusoidal_embeddings false --word_shuffle 0 --tokens_per_batch '-1' --accumulate_gradients 4 --validation_metrics '_valid_en_mlm_ppl,_valid_mlm_ppl,_valid_zh_mlm_ppl' --attention_dropout '0.1' --split_data true --max_epoch 100000 --stopping_criterion '_valid_zh_mlm_ppl,25' --dump_path '/checkpoint/aconneau/dumped' --epoch_size 200000 --word_blank 0 --gelu_activation true --n_layers 16 --optimizer 'adam_inverse_sqrt,lr=0.00005,warmup_updates=30000,beta1=0.9,beta2=0.999,weight_decay=0.01,eps=0.000001' --mlm_steps 'en,es,fr,de,zh,ru,pt,it,ar,ja,id,tr,nl,pl,simple,fa,vi,sv,ko,he,ro,no,hi,uk,cs,fi,hu,th,da,ca,el,bg,sr,ms,bn,hr,sl,zh_yue,az,sk,eo,ta,sh,lt,et,ml,la,bs,sq,arz,af,ka,mr,eu,tl,ang,gl,nn,ur,kk,be,hy,te,lv,mk,zh_classical,als,is,wuu,my,sco,mn,ceb,ast,cy,kn,br,an,gu,bar,uz,lb,ne,si,war,jv,ga,zh_min_nan,oc,ku,sw,nds,ckb,ia,yi,fy,scn,gan,tt,am' --eval_bleu false --dropout '0.1' --mt_steps '' --batch_size 16 --word_dropout 0 --reload_model '/checkpoint/aconneau/dumped/xlm_17_100_240_big_model_upper.2/14884511/best-valid_zh_mlm_ppl.pth' --min_count 0 --amp 2 --group_by_size true --asm false --sample_alpha '0.5' --word_pred '0.15' --clip_grad_norm 1 --emb_dim 1280 --encoder_only true --lgs 'en-es-fr-de-zh-ru-pt-it-ar-ja-id-tr-nl-pl-simple-fa-vi-sv-ko-he-ro-no-hi-uk-cs-fi-hu-th-da-ca-el-bg-sr-ms-bn-hr-sl-zh_yue-az-sk-eo-ta-sh-lt-et-ml-la-bs-sq-arz-af-ka-mr-eu-tl-ang-gl-nn-ur-kk-be-hy-te-lv-mk-zh_classical-als-is-wuu-my-sco-mn-ceb-ast-cy-kn-br-an-gu-bar-uz-lb-ne-si-war-jv-ga-zh_min_nan-oc-ku-sw-nds-ckb-ia-yi-fy-scn-gan-tt-am' --clm_steps '' --exp_name 'xlm_17_100_big.3' --lg_sampling_factor '0.7' --eval_only false --exp_id 16656234 --master_port 11363 --exp_id \"16656234\"",
22
- "context_size": 0,
23
- "data_path": "/private/home/aconneau/projects/XLM/data/wiki/100/175k",
24
- "debug": false,
25
- "debug_slurm": false,
26
- "debug_train": false,
27
- "dropout": 0.1,
28
- "dump_path": "/checkpoint/aconneau/dumped/xlm_17_100_big.3/16656234",
29
- "emb_dim": 1280,
30
- "embed_init_std": 0.02209708691207961,
31
- "encoder_only": true,
32
- "end_n_top": 5,
33
- "eos_index": 1,
34
- "epoch_size": 200000,
35
- "eval_bleu": false,
36
- "eval_only": false,
37
- "exp_id": "16656234",
38
- "exp_name": "xlm_17_100_big.3",
39
- "fp16": true,
40
- "gelu_activation": true,
41
- "global_rank": 0,
42
- "group_by_size": true,
43
- "hyp_path": "/checkpoint/aconneau/dumped/xlm_17_100_big.3/16656234/hypotheses",
44
- "id2label": {
45
- "0": "LABEL_0",
46
- "1": "LABEL_1",
47
- "2": "LABEL_2"
48
- },
49
- "id2lang": {
50
- "0": "af",
51
- "1": "als",
52
- "10": "be",
53
- "11": "bg",
54
- "12": "bn",
55
- "13": "br",
56
- "14": "bs",
57
- "15": "ca",
58
- "16": "ceb",
59
- "17": "ckb",
60
- "18": "cs",
61
- "19": "cy",
62
- "2": "am",
63
- "20": "da",
64
- "21": "de",
65
- "22": "el",
66
- "23": "en",
67
- "24": "eo",
68
- "25": "es",
69
- "26": "et",
70
- "27": "eu",
71
- "28": "fa",
72
- "29": "fi",
73
- "3": "an",
74
- "30": "fr",
75
- "31": "fy",
76
- "32": "ga",
77
- "33": "gan",
78
- "34": "gl",
79
- "35": "gu",
80
- "36": "he",
81
- "37": "hi",
82
- "38": "hr",
83
- "39": "hu",
84
- "4": "ang",
85
- "40": "hy",
86
- "41": "ia",
87
- "42": "id",
88
- "43": "is",
89
- "44": "it",
90
- "45": "ja",
91
- "46": "jv",
92
- "47": "ka",
93
- "48": "kk",
94
- "49": "kn",
95
- "5": "ar",
96
- "50": "ko",
97
- "51": "ku",
98
- "52": "la",
99
- "53": "lb",
100
- "54": "lt",
101
- "55": "lv",
102
- "56": "mk",
103
- "57": "ml",
104
- "58": "mn",
105
- "59": "mr",
106
- "6": "arz",
107
- "60": "ms",
108
- "61": "my",
109
- "62": "nds",
110
- "63": "ne",
111
- "64": "nl",
112
- "65": "nn",
113
- "66": "no",
114
- "67": "oc",
115
- "68": "pl",
116
- "69": "pt",
117
- "7": "ast",
118
- "70": "ro",
119
- "71": "ru",
120
- "72": "scn",
121
- "73": "sco",
122
- "74": "sh",
123
- "75": "si",
124
- "76": "simple",
125
- "77": "sk",
126
- "78": "sl",
127
- "79": "sq",
128
- "8": "az",
129
- "80": "sr",
130
- "81": "sv",
131
- "82": "sw",
132
- "83": "ta",
133
- "84": "te",
134
- "85": "th",
135
- "86": "tl",
136
- "87": "tr",
137
- "88": "tt",
138
- "89": "uk",
139
- "9": "bar",
140
- "90": "ur",
141
- "91": "uz",
142
- "92": "vi",
143
- "93": "war",
144
- "94": "wuu",
145
- "95": "yi",
146
- "96": "zh",
147
- "97": "zh_classical",
148
- "98": "zh_min_nan",
149
- "99": "zh_yue"
150
- },
151
- "init_std": 0.02,
152
- "is_encoder": true,
153
- "is_master": true,
154
- "is_slurm_job": true,
155
- "label2id": {
156
- "LABEL_0": 0,
157
- "LABEL_1": 1,
158
- "LABEL_2": 2
159
- },
160
- "lambda_ae": 1.0,
161
- "lambda_ae_config": null,
162
- "lambda_bt": 1.0,
163
- "lambda_bt_config": null,
164
- "lambda_clm": 1.0,
165
- "lambda_clm_config": null,
166
- "lambda_mlm": 1.0,
167
- "lambda_mlm_config": null,
168
- "lambda_mt": 1.0,
169
- "lambda_mt_config": null,
170
- "lambda_pc": 1.0,
171
- "lambda_pc_config": null,
172
- "lang2id": {
173
- "af": 0,
174
- "als": 1,
175
- "am": 2,
176
- "an": 3,
177
- "ang": 4,
178
- "ar": 5,
179
- "arz": 6,
180
- "ast": 7,
181
- "az": 8,
182
- "bar": 9,
183
- "be": 10,
184
- "bg": 11,
185
- "bn": 12,
186
- "br": 13,
187
- "bs": 14,
188
- "ca": 15,
189
- "ceb": 16,
190
- "ckb": 17,
191
- "cs": 18,
192
- "cy": 19,
193
- "da": 20,
194
- "de": 21,
195
- "el": 22,
196
- "en": 23,
197
- "eo": 24,
198
- "es": 25,
199
- "et": 26,
200
- "eu": 27,
201
- "fa": 28,
202
- "fi": 29,
203
- "fr": 30,
204
- "fy": 31,
205
- "ga": 32,
206
- "gan": 33,
207
- "gl": 34,
208
- "gu": 35,
209
- "he": 36,
210
- "hi": 37,
211
- "hr": 38,
212
- "hu": 39,
213
- "hy": 40,
214
- "ia": 41,
215
- "id": 42,
216
- "is": 43,
217
- "it": 44,
218
- "ja": 45,
219
- "jv": 46,
220
- "ka": 47,
221
- "kk": 48,
222
- "kn": 49,
223
- "ko": 50,
224
- "ku": 51,
225
- "la": 52,
226
- "lb": 53,
227
- "lt": 54,
228
- "lv": 55,
229
- "mk": 56,
230
- "ml": 57,
231
- "mn": 58,
232
- "mr": 59,
233
- "ms": 60,
234
- "my": 61,
235
- "nds": 62,
236
- "ne": 63,
237
- "nl": 64,
238
- "nn": 65,
239
- "no": 66,
240
- "oc": 67,
241
- "pl": 68,
242
- "pt": 69,
243
- "ro": 70,
244
- "ru": 71,
245
- "scn": 72,
246
- "sco": 73,
247
- "sh": 74,
248
- "si": 75,
249
- "simple": 76,
250
- "sk": 77,
251
- "sl": 78,
252
- "sq": 79,
253
- "sr": 80,
254
- "sv": 81,
255
- "sw": 82,
256
- "ta": 83,
257
- "te": 84,
258
- "th": 85,
259
- "tl": 86,
260
- "tr": 87,
261
- "tt": 88,
262
- "uk": 89,
263
- "ur": 90,
264
- "uz": 91,
265
- "vi": 92,
266
- "war": 93,
267
- "wuu": 94,
268
- "yi": 95,
269
- "zh": 96,
270
- "zh_classical": 97,
271
- "zh_min_nan": 98,
272
- "zh_yue": 99
273
- },
274
- "lang_id": 0,
275
- "langs": [
276
- "en",
277
- "es",
278
- "fr",
279
- "de",
280
- "zh",
281
- "ru",
282
- "pt",
283
- "it",
284
- "ar",
285
- "ja",
286
- "id",
287
- "tr",
288
- "nl",
289
- "pl",
290
- "simple",
291
- "fa",
292
- "vi",
293
- "sv",
294
- "ko",
295
- "he",
296
- "ro",
297
- "no",
298
- "hi",
299
- "uk",
300
- "cs",
301
- "fi",
302
- "hu",
303
- "th",
304
- "da",
305
- "ca",
306
- "el",
307
- "bg",
308
- "sr",
309
- "ms",
310
- "bn",
311
- "hr",
312
- "sl",
313
- "zh_yue",
314
- "az",
315
- "sk",
316
- "eo",
317
- "ta",
318
- "sh",
319
- "lt",
320
- "et",
321
- "ml",
322
- "la",
323
- "bs",
324
- "sq",
325
- "arz",
326
- "af",
327
- "ka",
328
- "mr",
329
- "eu",
330
- "tl",
331
- "ang",
332
- "gl",
333
- "nn",
334
- "ur",
335
- "kk",
336
- "be",
337
- "hy",
338
- "te",
339
- "lv",
340
- "mk",
341
- "zh_classical",
342
- "als",
343
- "is",
344
- "wuu",
345
- "my",
346
- "sco",
347
- "mn",
348
- "ceb",
349
- "ast",
350
- "cy",
351
- "kn",
352
- "br",
353
- "an",
354
- "gu",
355
- "bar",
356
- "uz",
357
- "lb",
358
- "ne",
359
- "si",
360
- "war",
361
- "jv",
362
- "ga",
363
- "zh_min_nan",
364
- "oc",
365
- "ku",
366
- "sw",
367
- "nds",
368
- "ckb",
369
- "ia",
370
- "yi",
371
- "fy",
372
- "scn",
373
- "gan",
374
- "tt",
375
- "am"
376
- ],
377
- "layer_norm_eps": 1e-12,
378
- "lg_sampling_factor": 0.7,
379
- "lgs": "en-es-fr-de-zh-ru-pt-it-ar-ja-id-tr-nl-pl-simple-fa-vi-sv-ko-he-ro-no-hi-uk-cs-fi-hu-th-da-ca-el-bg-sr-ms-bn-hr-sl-zh_yue-az-sk-eo-ta-sh-lt-et-ml-la-bs-sq-arz-af-ka-mr-eu-tl-ang-gl-nn-ur-kk-be-hy-te-lv-mk-zh_classical-als-is-wuu-my-sco-mn-ceb-ast-cy-kn-br-an-gu-bar-uz-lb-ne-si-war-jv-ga-zh_min_nan-oc-ku-sw-nds-ckb-ia-yi-fy-scn-gan-tt-am",
380
- "local_rank": 0,
381
- "mask_index": 5,
382
- "mask_token_id": 0,
383
- "master_addr": "learnfair0332",
384
- "master_port": 11363,
385
- "max_batch_size": 0,
386
- "max_epoch": 100000,
387
- "max_len": 200,
388
- "max_position_embeddings": 512,
389
- "max_vocab": 200000,
390
- "min_count": 0,
391
- "mlm_steps": [
392
- [
393
- "en",
394
- null
395
- ],
396
- [
397
- "es",
398
- null
399
- ],
400
- [
401
- "fr",
402
- null
403
- ],
404
- [
405
- "de",
406
- null
407
- ],
408
- [
409
- "zh",
410
- null
411
- ],
412
- [
413
- "ru",
414
- null
415
- ],
416
- [
417
- "pt",
418
- null
419
- ],
420
- [
421
- "it",
422
- null
423
- ],
424
- [
425
- "ar",
426
- null
427
- ],
428
- [
429
- "ja",
430
- null
431
- ],
432
- [
433
- "id",
434
- null
435
- ],
436
- [
437
- "tr",
438
- null
439
- ],
440
- [
441
- "nl",
442
- null
443
- ],
444
- [
445
- "pl",
446
- null
447
- ],
448
- [
449
- "simple",
450
- null
451
- ],
452
- [
453
- "fa",
454
- null
455
- ],
456
- [
457
- "vi",
458
- null
459
- ],
460
- [
461
- "sv",
462
- null
463
- ],
464
- [
465
- "ko",
466
- null
467
- ],
468
- [
469
- "he",
470
- null
471
- ],
472
- [
473
- "ro",
474
- null
475
- ],
476
- [
477
- "no",
478
- null
479
- ],
480
- [
481
- "hi",
482
- null
483
- ],
484
- [
485
- "uk",
486
- null
487
- ],
488
- [
489
- "cs",
490
- null
491
- ],
492
- [
493
- "fi",
494
- null
495
- ],
496
- [
497
- "hu",
498
- null
499
- ],
500
- [
501
- "th",
502
- null
503
- ],
504
- [
505
- "da",
506
- null
507
- ],
508
- [
509
- "ca",
510
- null
511
- ],
512
- [
513
- "el",
514
- null
515
- ],
516
- [
517
- "bg",
518
- null
519
- ],
520
- [
521
- "sr",
522
- null
523
- ],
524
- [
525
- "ms",
526
- null
527
- ],
528
- [
529
- "bn",
530
- null
531
- ],
532
- [
533
- "hr",
534
- null
535
- ],
536
- [
537
- "sl",
538
- null
539
- ],
540
- [
541
- "zh_yue",
542
- null
543
- ],
544
- [
545
- "az",
546
- null
547
- ],
548
- [
549
- "sk",
550
- null
551
- ],
552
- [
553
- "eo",
554
- null
555
- ],
556
- [
557
- "ta",
558
- null
559
- ],
560
- [
561
- "sh",
562
- null
563
- ],
564
- [
565
- "lt",
566
- null
567
- ],
568
- [
569
- "et",
570
- null
571
- ],
572
- [
573
- "ml",
574
- null
575
- ],
576
- [
577
- "la",
578
- null
579
- ],
580
- [
581
- "bs",
582
- null
583
- ],
584
- [
585
- "sq",
586
- null
587
- ],
588
- [
589
- "arz",
590
- null
591
- ],
592
- [
593
- "af",
594
- null
595
- ],
596
- [
597
- "ka",
598
- null
599
- ],
600
- [
601
- "mr",
602
- null
603
- ],
604
- [
605
- "eu",
606
- null
607
- ],
608
- [
609
- "tl",
610
- null
611
- ],
612
- [
613
- "ang",
614
- null
615
- ],
616
- [
617
- "gl",
618
- null
619
- ],
620
- [
621
- "nn",
622
- null
623
- ],
624
- [
625
- "ur",
626
- null
627
- ],
628
- [
629
- "kk",
630
- null
631
- ],
632
- [
633
- "be",
634
- null
635
- ],
636
- [
637
- "hy",
638
- null
639
- ],
640
- [
641
- "te",
642
- null
643
- ],
644
- [
645
- "lv",
646
- null
647
- ],
648
- [
649
- "mk",
650
- null
651
- ],
652
- [
653
- "zh_classical",
654
- null
655
- ],
656
- [
657
- "als",
658
- null
659
- ],
660
- [
661
- "is",
662
- null
663
- ],
664
- [
665
- "wuu",
666
- null
667
- ],
668
- [
669
- "my",
670
- null
671
- ],
672
- [
673
- "sco",
674
- null
675
- ],
676
- [
677
- "mn",
678
- null
679
- ],
680
- [
681
- "ceb",
682
- null
683
- ],
684
- [
685
- "ast",
686
- null
687
- ],
688
- [
689
- "cy",
690
- null
691
- ],
692
- [
693
- "kn",
694
- null
695
- ],
696
- [
697
- "br",
698
- null
699
- ],
700
- [
701
- "an",
702
- null
703
- ],
704
- [
705
- "gu",
706
- null
707
- ],
708
- [
709
- "bar",
710
- null
711
- ],
712
- [
713
- "uz",
714
- null
715
- ],
716
- [
717
- "lb",
718
- null
719
- ],
720
- [
721
- "ne",
722
- null
723
- ],
724
- [
725
- "si",
726
- null
727
- ],
728
- [
729
- "war",
730
- null
731
- ],
732
- [
733
- "jv",
734
- null
735
- ],
736
- [
737
- "ga",
738
- null
739
- ],
740
- [
741
- "zh_min_nan",
742
- null
743
- ],
744
- [
745
- "oc",
746
- null
747
- ],
748
- [
749
- "ku",
750
- null
751
- ],
752
- [
753
- "sw",
754
- null
755
- ],
756
- [
757
- "nds",
758
- null
759
- ],
760
- [
761
- "ckb",
762
- null
763
- ],
764
- [
765
- "ia",
766
- null
767
- ],
768
- [
769
- "yi",
770
- null
771
- ],
772
- [
773
- "fy",
774
- null
775
- ],
776
- [
777
- "scn",
778
- null
779
- ],
780
- [
781
- "gan",
782
- null
783
- ],
784
- [
785
- "tt",
786
- null
787
- ],
788
- [
789
- "am",
790
- null
791
- ]
792
- ],
793
- "model_type": "xlm",
794
- "mono_dataset": {
795
- "af": {
796
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.af.pth",
797
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.af.pth",
798
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.af.pth"
799
- },
800
- "als": {
801
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.als.pth",
802
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.als.pth",
803
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.als.pth"
804
- },
805
- "am": {
806
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.am.pth",
807
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.am.pth",
808
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.am.pth"
809
- },
810
- "an": {
811
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.an.pth",
812
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.an.pth",
813
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.an.pth"
814
- },
815
- "ang": {
816
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ang.pth",
817
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ang.pth",
818
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ang.pth"
819
- },
820
- "ar": {
821
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ar.pth",
822
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ar.pth",
823
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ar.pth"
824
- },
825
- "arz": {
826
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.arz.pth",
827
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.arz.pth",
828
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.arz.pth"
829
- },
830
- "ast": {
831
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ast.pth",
832
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ast.pth",
833
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ast.pth"
834
- },
835
- "az": {
836
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.az.pth",
837
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.az.pth",
838
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.az.pth"
839
- },
840
- "bar": {
841
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.bar.pth",
842
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.bar.pth",
843
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.bar.pth"
844
- },
845
- "be": {
846
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.be.pth",
847
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.be.pth",
848
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.be.pth"
849
- },
850
- "bg": {
851
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.bg.pth",
852
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.bg.pth",
853
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.bg.pth"
854
- },
855
- "bn": {
856
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.bn.pth",
857
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.bn.pth",
858
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.bn.pth"
859
- },
860
- "br": {
861
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.br.pth",
862
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.br.pth",
863
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.br.pth"
864
- },
865
- "bs": {
866
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.bs.pth",
867
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.bs.pth",
868
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.bs.pth"
869
- },
870
- "ca": {
871
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ca.pth",
872
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ca.pth",
873
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ca.pth"
874
- },
875
- "ceb": {
876
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ceb.pth",
877
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ceb.pth",
878
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ceb.pth"
879
- },
880
- "ckb": {
881
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ckb.pth",
882
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ckb.pth",
883
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ckb.pth"
884
- },
885
- "cs": {
886
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.cs.pth",
887
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.cs.pth",
888
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.cs.pth"
889
- },
890
- "cy": {
891
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.cy.pth",
892
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.cy.pth",
893
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.cy.pth"
894
- },
895
- "da": {
896
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.da.pth",
897
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.da.pth",
898
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.da.pth"
899
- },
900
- "de": {
901
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.de.pth",
902
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.de.pth",
903
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.de.pth"
904
- },
905
- "el": {
906
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.el.pth",
907
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.el.pth",
908
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.el.pth"
909
- },
910
- "en": {
911
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.en.pth",
912
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.en.pth",
913
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.en.pth"
914
- },
915
- "eo": {
916
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.eo.pth",
917
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.eo.pth",
918
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.eo.pth"
919
- },
920
- "es": {
921
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.es.pth",
922
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.es.pth",
923
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.es.pth"
924
- },
925
- "et": {
926
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.et.pth",
927
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.et.pth",
928
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.et.pth"
929
- },
930
- "eu": {
931
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.eu.pth",
932
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.eu.pth",
933
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.eu.pth"
934
- },
935
- "fa": {
936
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.fa.pth",
937
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.fa.pth",
938
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.fa.pth"
939
- },
940
- "fi": {
941
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.fi.pth",
942
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.fi.pth",
943
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.fi.pth"
944
- },
945
- "fr": {
946
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.fr.pth",
947
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.fr.pth",
948
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.fr.pth"
949
- },
950
- "fy": {
951
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.fy.pth",
952
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.fy.pth",
953
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.fy.pth"
954
- },
955
- "ga": {
956
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ga.pth",
957
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ga.pth",
958
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ga.pth"
959
- },
960
- "gan": {
961
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.gan.pth",
962
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.gan.pth",
963
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.gan.pth"
964
- },
965
- "gl": {
966
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.gl.pth",
967
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.gl.pth",
968
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.gl.pth"
969
- },
970
- "gu": {
971
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.gu.pth",
972
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.gu.pth",
973
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.gu.pth"
974
- },
975
- "he": {
976
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.he.pth",
977
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.he.pth",
978
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.he.pth"
979
- },
980
- "hi": {
981
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.hi.pth",
982
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.hi.pth",
983
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.hi.pth"
984
- },
985
- "hr": {
986
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.hr.pth",
987
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.hr.pth",
988
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.hr.pth"
989
- },
990
- "hu": {
991
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.hu.pth",
992
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.hu.pth",
993
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.hu.pth"
994
- },
995
- "hy": {
996
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.hy.pth",
997
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.hy.pth",
998
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.hy.pth"
999
- },
1000
- "ia": {
1001
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ia.pth",
1002
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ia.pth",
1003
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ia.pth"
1004
- },
1005
- "id": {
1006
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.id.pth",
1007
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.id.pth",
1008
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.id.pth"
1009
- },
1010
- "is": {
1011
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.is.pth",
1012
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.is.pth",
1013
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.is.pth"
1014
- },
1015
- "it": {
1016
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.it.pth",
1017
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.it.pth",
1018
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.it.pth"
1019
- },
1020
- "ja": {
1021
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ja.pth",
1022
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ja.pth",
1023
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ja.pth"
1024
- },
1025
- "jv": {
1026
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.jv.pth",
1027
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.jv.pth",
1028
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.jv.pth"
1029
- },
1030
- "ka": {
1031
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ka.pth",
1032
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ka.pth",
1033
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ka.pth"
1034
- },
1035
- "kk": {
1036
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.kk.pth",
1037
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.kk.pth",
1038
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.kk.pth"
1039
- },
1040
- "kn": {
1041
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.kn.pth",
1042
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.kn.pth",
1043
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.kn.pth"
1044
- },
1045
- "ko": {
1046
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ko.pth",
1047
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ko.pth",
1048
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ko.pth"
1049
- },
1050
- "ku": {
1051
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ku.pth",
1052
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ku.pth",
1053
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ku.pth"
1054
- },
1055
- "la": {
1056
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.la.pth",
1057
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.la.pth",
1058
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.la.pth"
1059
- },
1060
- "lb": {
1061
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.lb.pth",
1062
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.lb.pth",
1063
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.lb.pth"
1064
- },
1065
- "lt": {
1066
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.lt.pth",
1067
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.lt.pth",
1068
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.lt.pth"
1069
- },
1070
- "lv": {
1071
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.lv.pth",
1072
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.lv.pth",
1073
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.lv.pth"
1074
- },
1075
- "mk": {
1076
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.mk.pth",
1077
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.mk.pth",
1078
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.mk.pth"
1079
- },
1080
- "ml": {
1081
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ml.pth",
1082
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ml.pth",
1083
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ml.pth"
1084
- },
1085
- "mn": {
1086
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.mn.pth",
1087
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.mn.pth",
1088
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.mn.pth"
1089
- },
1090
- "mr": {
1091
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.mr.pth",
1092
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.mr.pth",
1093
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.mr.pth"
1094
- },
1095
- "ms": {
1096
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ms.pth",
1097
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ms.pth",
1098
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ms.pth"
1099
- },
1100
- "my": {
1101
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.my.pth",
1102
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.my.pth",
1103
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.my.pth"
1104
- },
1105
- "nds": {
1106
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.nds.pth",
1107
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.nds.pth",
1108
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.nds.pth"
1109
- },
1110
- "ne": {
1111
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ne.pth",
1112
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ne.pth",
1113
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ne.pth"
1114
- },
1115
- "nl": {
1116
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.nl.pth",
1117
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.nl.pth",
1118
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.nl.pth"
1119
- },
1120
- "nn": {
1121
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.nn.pth",
1122
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.nn.pth",
1123
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.nn.pth"
1124
- },
1125
- "no": {
1126
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.no.pth",
1127
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.no.pth",
1128
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.no.pth"
1129
- },
1130
- "oc": {
1131
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.oc.pth",
1132
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.oc.pth",
1133
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.oc.pth"
1134
- },
1135
- "pl": {
1136
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.pl.pth",
1137
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.pl.pth",
1138
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.pl.pth"
1139
- },
1140
- "pt": {
1141
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.pt.pth",
1142
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.pt.pth",
1143
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.pt.pth"
1144
- },
1145
- "ro": {
1146
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ro.pth",
1147
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ro.pth",
1148
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ro.pth"
1149
- },
1150
- "ru": {
1151
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ru.pth",
1152
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ru.pth",
1153
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ru.pth"
1154
- },
1155
- "scn": {
1156
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.scn.pth",
1157
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.scn.pth",
1158
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.scn.pth"
1159
- },
1160
- "sco": {
1161
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.sco.pth",
1162
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.sco.pth",
1163
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.sco.pth"
1164
- },
1165
- "sh": {
1166
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.sh.pth",
1167
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.sh.pth",
1168
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.sh.pth"
1169
- },
1170
- "si": {
1171
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.si.pth",
1172
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.si.pth",
1173
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.si.pth"
1174
- },
1175
- "simple": {
1176
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.simple.pth",
1177
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.simple.pth",
1178
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.simple.pth"
1179
- },
1180
- "sk": {
1181
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.sk.pth",
1182
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.sk.pth",
1183
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.sk.pth"
1184
- },
1185
- "sl": {
1186
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.sl.pth",
1187
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.sl.pth",
1188
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.sl.pth"
1189
- },
1190
- "sq": {
1191
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.sq.pth",
1192
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.sq.pth",
1193
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.sq.pth"
1194
- },
1195
- "sr": {
1196
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.sr.pth",
1197
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.sr.pth",
1198
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.sr.pth"
1199
- },
1200
- "sv": {
1201
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.sv.pth",
1202
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.sv.pth",
1203
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.sv.pth"
1204
- },
1205
- "sw": {
1206
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.sw.pth",
1207
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.sw.pth",
1208
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.sw.pth"
1209
- },
1210
- "ta": {
1211
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ta.pth",
1212
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ta.pth",
1213
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ta.pth"
1214
- },
1215
- "te": {
1216
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.te.pth",
1217
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.te.pth",
1218
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.te.pth"
1219
- },
1220
- "th": {
1221
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.th.pth",
1222
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.th.pth",
1223
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.th.pth"
1224
- },
1225
- "tl": {
1226
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.tl.pth",
1227
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.tl.pth",
1228
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.tl.pth"
1229
- },
1230
- "tr": {
1231
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.tr.pth",
1232
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.tr.pth",
1233
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.tr.pth"
1234
- },
1235
- "tt": {
1236
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.tt.pth",
1237
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.tt.pth",
1238
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.tt.pth"
1239
- },
1240
- "uk": {
1241
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.uk.pth",
1242
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.uk.pth",
1243
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.uk.pth"
1244
- },
1245
- "ur": {
1246
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.ur.pth",
1247
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.ur.pth",
1248
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.ur.pth"
1249
- },
1250
- "uz": {
1251
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.uz.pth",
1252
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.uz.pth",
1253
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.uz.pth"
1254
- },
1255
- "vi": {
1256
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.vi.pth",
1257
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.vi.pth",
1258
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.vi.pth"
1259
- },
1260
- "war": {
1261
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.war.pth",
1262
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.war.pth",
1263
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.war.pth"
1264
- },
1265
- "wuu": {
1266
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.wuu.pth",
1267
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.wuu.pth",
1268
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.wuu.pth"
1269
- },
1270
- "yi": {
1271
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.yi.pth",
1272
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.yi.pth",
1273
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.yi.pth"
1274
- },
1275
- "zh": {
1276
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.zh.pth",
1277
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.zh.pth",
1278
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.zh.pth"
1279
- },
1280
- "zh_classical": {
1281
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.zh_classical.pth",
1282
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.zh_classical.pth",
1283
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.zh_classical.pth"
1284
- },
1285
- "zh_min_nan": {
1286
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.zh_min_nan.pth",
1287
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.zh_min_nan.pth",
1288
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.zh_min_nan.pth"
1289
- },
1290
- "zh_yue": {
1291
- "test": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/test.zh_yue.pth",
1292
- "train": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/train.zh_yue.pth",
1293
- "valid": "/private/home/aconneau/projects/XLM/data/wiki/100/175k/valid.zh_yue.pth"
1294
- }
1295
- },
1296
- "mono_list": [
1297
- "en",
1298
- "es",
1299
- "fr",
1300
- "de",
1301
- "zh",
1302
- "ru",
1303
- "pt",
1304
- "it",
1305
- "ar",
1306
- "ja",
1307
- "id",
1308
- "tr",
1309
- "nl",
1310
- "pl",
1311
- "simple",
1312
- "fa",
1313
- "vi",
1314
- "sv",
1315
- "ko",
1316
- "he",
1317
- "ro",
1318
- "no",
1319
- "hi",
1320
- "uk",
1321
- "cs",
1322
- "fi",
1323
- "hu",
1324
- "th",
1325
- "da",
1326
- "ca",
1327
- "el",
1328
- "bg",
1329
- "sr",
1330
- "ms",
1331
- "bn",
1332
- "hr",
1333
- "sl",
1334
- "zh_yue",
1335
- "az",
1336
- "sk",
1337
- "eo",
1338
- "ta",
1339
- "sh",
1340
- "lt",
1341
- "et",
1342
- "ml",
1343
- "la",
1344
- "bs",
1345
- "sq",
1346
- "arz",
1347
- "af",
1348
- "ka",
1349
- "mr",
1350
- "eu",
1351
- "tl",
1352
- "ang",
1353
- "gl",
1354
- "nn",
1355
- "ur",
1356
- "kk",
1357
- "be",
1358
- "hy",
1359
- "te",
1360
- "lv",
1361
- "mk",
1362
- "zh_classical",
1363
- "als",
1364
- "is",
1365
- "wuu",
1366
- "my",
1367
- "sco",
1368
- "mn",
1369
- "ceb",
1370
- "ast",
1371
- "cy",
1372
- "kn",
1373
- "br",
1374
- "an",
1375
- "gu",
1376
- "bar",
1377
- "uz",
1378
- "lb",
1379
- "ne",
1380
- "si",
1381
- "war",
1382
- "jv",
1383
- "ga",
1384
- "zh_min_nan",
1385
- "oc",
1386
- "ku",
1387
- "sw",
1388
- "nds",
1389
- "ckb",
1390
- "ia",
1391
- "yi",
1392
- "fy",
1393
- "scn",
1394
- "gan",
1395
- "tt",
1396
- "am"
1397
- ],
1398
- "mt_steps": [],
1399
- "multi_gpu": true,
1400
- "multi_node": true,
1401
- "n_gpu_per_node": 8,
1402
- "n_heads": 16,
1403
- "n_langs": 100,
1404
- "n_layers": 16,
1405
- "n_nodes": 4,
1406
- "node_id": 0,
1407
- "optimizer": "adam_inverse_sqrt,lr=0.00005,warmup_updates=30000,beta1=0.9,beta2=0.999,weight_decay=0.01,eps=0.000001",
1408
- "pad_index": 2,
1409
- "pad_token_id": 2,
1410
- "para_dataset": {},
1411
- "para_list": [],
1412
- "pc_steps": [],
1413
- "ref_paths": {},
1414
- "reload_checkpoint": "",
1415
- "reload_emb": "",
1416
- "reload_model": "/checkpoint/aconneau/dumped/xlm_17_100_240_big_model_upper.2/14884511/best-valid_zh_mlm_ppl.pth",
1417
- "sample_alpha": 0.5,
1418
- "save_periodic": 0,
1419
- "share_inout_emb": true,
1420
- "sinusoidal_embeddings": false,
1421
- "split_data": true,
1422
- "start_n_top": 5,
1423
- "stopping_criterion": "_valid_zh_mlm_ppl,25",
1424
- "summary_activation": null,
1425
- "summary_first_dropout": 0.1,
1426
- "summary_proj_to_labels": true,
1427
- "summary_type": "first",
1428
- "summary_use_proj": true,
1429
- "tokens_per_batch": -1,
1430
- "torch_dtype": "float32",
1431
- "transformers_version": "4.18.0",
1432
- "unk_index": 3,
1433
- "use_lang_emb": false,
1434
- "use_memory": false,
1435
- "validation_metrics": "_valid_en_mlm_ppl,_valid_mlm_ppl,_valid_zh_mlm_ppl",
1436
- "vocab_size": 200000,
1437
- "word_blank": 0.0,
1438
- "word_dropout": 0.0,
1439
- "word_keep": 0.1,
1440
- "word_mask": 0.8,
1441
- "word_mask_keep_rand": "0.8,0.1,0.1",
1442
- "word_pred": 0.15,
1443
- "word_rand": 0.1,
1444
- "word_shuffle": 0.0,
1445
- "world_size": 32
1446
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Kanmodel1/merges.txt DELETED
The diff for this file is too large to render. See raw diff
 
Kanmodel1/pytorch_model.bin DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:75176900d9607a24034d64b540f950b0d95e3743b0a38457e2d2239f0aa56aba
3
- size 2286087987
 
 
 
 
Kanmodel1/special_tokens_map.json DELETED
@@ -1 +0,0 @@
1
- {"bos_token": "<s>", "unk_token": "<unk>", "sep_token": "</s>", "pad_token": "<pad>", "cls_token": "</s>", "mask_token": "<special1>", "additional_special_tokens": ["<special0>", "<special1>", "<special2>", "<special3>", "<special4>", "<special5>", "<special6>", "<special7>", "<special8>", "<special9>"]}
 
 
Kanmodel1/tokenizer_config.json DELETED
@@ -1 +0,0 @@
1
- {"unk_token": "<unk>", "bos_token": "<s>", "sep_token": "</s>", "pad_token": "<pad>", "cls_token": "</s>", "mask_token": "<special1>", "additional_special_tokens": ["<special0>", "<special1>", "<special2>", "<special3>", "<special4>", "<special5>", "<special6>", "<special7>", "<special8>", "<special9>"], "lang2id": {"af": 0, "als": 1, "am": 2, "an": 3, "ang": 4, "ar": 5, "arz": 6, "ast": 7, "az": 8, "bar": 9, "be": 10, "bg": 11, "bn": 12, "br": 13, "bs": 14, "ca": 15, "ceb": 16, "ckb": 17, "cs": 18, "cy": 19, "da": 20, "de": 21, "el": 22, "en": 23, "eo": 24, "es": 25, "et": 26, "eu": 27, "fa": 28, "fi": 29, "fr": 30, "fy": 31, "ga": 32, "gan": 33, "gl": 34, "gu": 35, "he": 36, "hi": 37, "hr": 38, "hu": 39, "hy": 40, "ia": 41, "id": 42, "is": 43, "it": 44, "ja": 45, "jv": 46, "ka": 47, "kk": 48, "kn": 49, "ko": 50, "ku": 51, "la": 52, "lb": 53, "lt": 54, "lv": 55, "mk": 56, "ml": 57, "mn": 58, "mr": 59, "ms": 60, "my": 61, "nds": 62, "ne": 63, "nl": 64, "nn": 65, "no": 66, "oc": 67, "pl": 68, "pt": 69, "ro": 70, "ru": 71, "scn": 72, "sco": 73, "sh": 74, "si": 75, "simple": 76, "sk": 77, "sl": 78, "sq": 79, "sr": 80, "sv": 81, "sw": 82, "ta": 83, "te": 84, "th": 85, "tl": 86, "tr": 87, "tt": 88, "uk": 89, "ur": 90, "uz": 91, "vi": 92, "war": 93, "wuu": 94, "yi": 95, "zh": 96, "zh_classical": 97, "zh_min_nan": 98, "zh_yue": 99}, "id2lang": {"0": "af", "1": "als", "2": "am", "3": "an", "4": "ang", "5": "ar", "6": "arz", "7": "ast", "8": "az", "9": "bar", "10": "be", "11": "bg", "12": "bn", "13": "br", "14": "bs", "15": "ca", "16": "ceb", "17": "ckb", "18": "cs", "19": "cy", "20": "da", "21": "de", "22": "el", "23": "en", "24": "eo", "25": "es", "26": "et", "27": "eu", "28": "fa", "29": "fi", "30": "fr", "31": "fy", "32": "ga", "33": "gan", "34": "gl", "35": "gu", "36": "he", "37": "hi", "38": "hr", "39": "hu", "40": "hy", "41": "ia", "42": "id", "43": "is", "44": "it", "45": "ja", "46": "jv", "47": "ka", "48": "kk", "49": "kn", "50": "ko", "51": "ku", "52": "la", "53": "lb", "54": "lt", "55": "lv", "56": "mk", "57": "ml", "58": "mn", "59": "mr", "60": "ms", "61": "my", "62": "nds", "63": "ne", "64": "nl", "65": "nn", "66": "no", "67": "oc", "68": "pl", "69": "pt", "70": "ro", "71": "ru", "72": "scn", "73": "sco", "74": "sh", "75": "si", "76": "simple", "77": "sk", "78": "sl", "79": "sq", "80": "sr", "81": "sv", "82": "sw", "83": "ta", "84": "te", "85": "th", "86": "tl", "87": "tr", "88": "tt", "89": "uk", "90": "ur", "91": "uz", "92": "vi", "93": "war", "94": "wuu", "95": "yi", "96": "zh", "97": "zh_classical", "98": "zh_min_nan", "99": "zh_yue"}, "do_lowercase_and_remove_accent": false, "model_max_length": 512, "do_lower_case": false, "special_tokens_map_file": null, "name_or_path": "../input/kannada-xlm/model111/checkpoint-118-epoch-2", "tokenizer_class": "XLMTokenizer"}
 
 
Kanmodel1/vocab.json DELETED
The diff for this file is too large to render. See raw diff