system HF staff commited on
Commit
d8fa21f
1 Parent(s): bd6a443

Update config.json

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