Hugging Face's logo

Xenova
/
xlm-mlm-17-1280

Fill-Mask
Transformers.js
ONNX
xlm
Model card Files Community
Xenova HF staff commited on
Commit
50d74ba
·
1 Parent(s): aafafca

Upload folder using huggingface_hub

Browse files
Files changed (9) hide show
  1. .gitattributes +1 -0
  2. config.json +356 -0
  3. generation_config.json +6 -0
  4. merges.txt +0 -0
  5. model.onnx +3 -0
  6. model.onnx_data +3 -0
  7. special_tokens_map.json +20 -0
  8. tokenizer_config.json +62 -0
  9. vocab.json +0 -0
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ model.onnx_data filter=lfs diff=lfs merge=lfs -text
config.json ADDED
@@ -0,0 +1,356 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "xlm-mlm-17-1280",
3
+ "accumulate_gradients": 4,
4
+ "ae_steps": [],
5
+ "amp": 2,
6
+ "architectures": [
7
+ "XLMWithLMHeadModel"
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/17/175k' --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,fr,es,de,it,pt,nl,sv,pl,ru,ar,tr,zh,ja,ko,hi,vi' --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.2/14884510/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-fr-es-de-it-pt-nl-sv-pl-ru-ar-tr-zh-ja-ko-hi-vi' --clm_steps '' --exp_name 'xlm_17_100_big.3' --lg_sampling_factor '0.7' --eval_only false --exp_id 16656237 --master_port 14148 --exp_id \"16656237\"",
22
+ "context_size": 0,
23
+ "data_path": "/private/home/aconneau/projects/XLM/data/wiki/17/175k",
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.3/16656237",
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": "16656237",
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.3/16656237/hypotheses",
44
+ "id2lang": {
45
+ "0": "ar",
46
+ "1": "de",
47
+ "10": "pl",
48
+ "11": "pt",
49
+ "12": "ru",
50
+ "13": "sv",
51
+ "14": "tr",
52
+ "15": "vi",
53
+ "16": "zh",
54
+ "2": "en",
55
+ "3": "es",
56
+ "4": "fr",
57
+ "5": "hi",
58
+ "6": "it",
59
+ "7": "ja",
60
+ "8": "ko",
61
+ "9": "nl"
62
+ },
63
+ "init_std": 0.02,
64
+ "is_encoder": true,
65
+ "is_master": true,
66
+ "is_slurm_job": true,
67
+ "lambda_ae": 1.0,
68
+ "lambda_ae_config": null,
69
+ "lambda_bt": 1.0,
70
+ "lambda_bt_config": null,
71
+ "lambda_clm": 1.0,
72
+ "lambda_clm_config": null,
73
+ "lambda_mlm": 1.0,
74
+ "lambda_mlm_config": null,
75
+ "lambda_mt": 1.0,
76
+ "lambda_mt_config": null,
77
+ "lambda_pc": 1.0,
78
+ "lambda_pc_config": null,
79
+ "lang2id": {
80
+ "ar": 0,
81
+ "de": 1,
82
+ "en": 2,
83
+ "es": 3,
84
+ "fr": 4,
85
+ "hi": 5,
86
+ "it": 6,
87
+ "ja": 7,
88
+ "ko": 8,
89
+ "nl": 9,
90
+ "pl": 10,
91
+ "pt": 11,
92
+ "ru": 12,
93
+ "sv": 13,
94
+ "tr": 14,
95
+ "vi": 15,
96
+ "zh": 16
97
+ },
98
+ "lang_id": 0,
99
+ "langs": [
100
+ "en",
101
+ "fr",
102
+ "es",
103
+ "de",
104
+ "it",
105
+ "pt",
106
+ "nl",
107
+ "sv",
108
+ "pl",
109
+ "ru",
110
+ "ar",
111
+ "tr",
112
+ "zh",
113
+ "ja",
114
+ "ko",
115
+ "hi",
116
+ "vi"
117
+ ],
118
+ "layer_norm_eps": 1e-12,
119
+ "lg_sampling_factor": 0.7,
120
+ "lgs": "en-fr-es-de-it-pt-nl-sv-pl-ru-ar-tr-zh-ja-ko-hi-vi",
121
+ "local_rank": 0,
122
+ "mask_index": 5,
123
+ "mask_token_id": 0,
124
+ "master_addr": "learnfair1605",
125
+ "master_port": 14148,
126
+ "max_batch_size": 0,
127
+ "max_epoch": 100000,
128
+ "max_len": 200,
129
+ "max_position_embeddings": 512,
130
+ "max_vocab": 200000,
131
+ "min_count": 0,
132
+ "mlm_steps": [
133
+ [
134
+ "en",
135
+ null
136
+ ],
137
+ [
138
+ "fr",
139
+ null
140
+ ],
141
+ [
142
+ "es",
143
+ null
144
+ ],
145
+ [
146
+ "de",
147
+ null
148
+ ],
149
+ [
150
+ "it",
151
+ null
152
+ ],
153
+ [
154
+ "pt",
155
+ null
156
+ ],
157
+ [
158
+ "nl",
159
+ null
160
+ ],
161
+ [
162
+ "sv",
163
+ null
164
+ ],
165
+ [
166
+ "pl",
167
+ null
168
+ ],
169
+ [
170
+ "ru",
171
+ null
172
+ ],
173
+ [
174
+ "ar",
175
+ null
176
+ ],
177
+ [
178
+ "tr",
179
+ null
180
+ ],
181
+ [
182
+ "zh",
183
+ null
184
+ ],
185
+ [
186
+ "ja",
187
+ null
188
+ ],
189
+ [
190
+ "ko",
191
+ null
192
+ ],
193
+ [
194
+ "hi",
195
+ null
196
+ ],
197
+ [
198
+ "vi",
199
+ null
200
+ ]
201
+ ],
202
+ "model_type": "xlm",
203
+ "mono_dataset": {
204
+ "ar": {
205
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.ar.pth",
206
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.ar.pth",
207
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.ar.pth"
208
+ },
209
+ "de": {
210
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.de.pth",
211
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.de.pth",
212
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.de.pth"
213
+ },
214
+ "en": {
215
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.en.pth",
216
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.en.pth",
217
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.en.pth"
218
+ },
219
+ "es": {
220
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.es.pth",
221
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.es.pth",
222
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.es.pth"
223
+ },
224
+ "fr": {
225
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.fr.pth",
226
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.fr.pth",
227
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.fr.pth"
228
+ },
229
+ "hi": {
230
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.hi.pth",
231
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.hi.pth",
232
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.hi.pth"
233
+ },
234
+ "it": {
235
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.it.pth",
236
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.it.pth",
237
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.it.pth"
238
+ },
239
+ "ja": {
240
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.ja.pth",
241
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.ja.pth",
242
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.ja.pth"
243
+ },
244
+ "ko": {
245
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.ko.pth",
246
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.ko.pth",
247
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.ko.pth"
248
+ },
249
+ "nl": {
250
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.nl.pth",
251
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.nl.pth",
252
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.nl.pth"
253
+ },
254
+ "pl": {
255
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.pl.pth",
256
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.pl.pth",
257
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.pl.pth"
258
+ },
259
+ "pt": {
260
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.pt.pth",
261
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.pt.pth",
262
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.pt.pth"
263
+ },
264
+ "ru": {
265
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.ru.pth",
266
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.ru.pth",
267
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.ru.pth"
268
+ },
269
+ "sv": {
270
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.sv.pth",
271
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.sv.pth",
272
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.sv.pth"
273
+ },
274
+ "tr": {
275
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.tr.pth",
276
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.tr.pth",
277
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.tr.pth"
278
+ },
279
+ "vi": {
280
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.vi.pth",
281
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.vi.pth",
282
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.vi.pth"
283
+ },
284
+ "zh": {
285
+ "test": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/test.zh.pth",
286
+ "train": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/train.zh.pth",
287
+ "valid": "/private/home/aconneau/projects/XLM/data/wiki/17/175k/valid.zh.pth"
288
+ }
289
+ },
290
+ "mono_list": [
291
+ "en",
292
+ "fr",
293
+ "es",
294
+ "de",
295
+ "it",
296
+ "pt",
297
+ "nl",
298
+ "sv",
299
+ "pl",
300
+ "ru",
301
+ "ar",
302
+ "tr",
303
+ "zh",
304
+ "ja",
305
+ "ko",
306
+ "hi",
307
+ "vi"
308
+ ],
309
+ "mt_steps": [],
310
+ "multi_gpu": true,
311
+ "multi_node": true,
312
+ "n_gpu_per_node": 8,
313
+ "n_heads": 16,
314
+ "n_langs": 17,
315
+ "n_layers": 16,
316
+ "n_nodes": 4,
317
+ "node_id": 0,
318
+ "optimizer": "adam_inverse_sqrt,lr=0.00005,warmup_updates=30000,beta1=0.9,beta2=0.999,weight_decay=0.01,eps=0.000001",
319
+ "pad_index": 2,
320
+ "pad_token_id": 2,
321
+ "para_dataset": {},
322
+ "para_list": [],
323
+ "pc_steps": [],
324
+ "ref_paths": {},
325
+ "reload_checkpoint": "",
326
+ "reload_emb": "",
327
+ "reload_model": "/checkpoint/aconneau/dumped/xlm_17_100_240_big_model_upper.2/14884510/best-valid_zh_mlm_ppl.pth",
328
+ "sample_alpha": 0.5,
329
+ "save_periodic": 0,
330
+ "share_inout_emb": true,
331
+ "sinusoidal_embeddings": false,
332
+ "split_data": true,
333
+ "start_n_top": 5,
334
+ "stopping_criterion": "_valid_zh_mlm_ppl,25",
335
+ "summary_activation": null,
336
+ "summary_first_dropout": 0.1,
337
+ "summary_proj_to_labels": true,
338
+ "summary_type": "first",
339
+ "summary_use_proj": true,
340
+ "tokens_per_batch": -1,
341
+ "transformers_version": "4.34.0.dev0",
342
+ "unk_index": 3,
343
+ "use_lang_emb": false,
344
+ "use_memory": false,
345
+ "validation_metrics": "_valid_en_mlm_ppl,_valid_mlm_ppl,_valid_zh_mlm_ppl",
346
+ "vocab_size": 200000,
347
+ "word_blank": 0.0,
348
+ "word_dropout": 0.0,
349
+ "word_keep": 0.1,
350
+ "word_mask": 0.8,
351
+ "word_mask_keep_rand": "0.8,0.1,0.1",
352
+ "word_pred": 0.15,
353
+ "word_rand": 0.1,
354
+ "word_shuffle": 0.0,
355
+ "world_size": 32
356
+ }
generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 0,
4
+ "pad_token_id": 2,
5
+ "transformers_version": "4.34.0.dev0"
6
+ }
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e9a140da6b8c24d5e43287bd6c17a174f7ea0b4ebf4f4158f698eb3235dfa22
3
+ size 297116
model.onnx_data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5951c7e8e73f5d247113842ff966017e0ade61c5f284dcd4b3be843d63cfe372
3
+ size 2286787840
special_tokens_map.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<special0>",
4
+ "<special1>",
5
+ "<special2>",
6
+ "<special3>",
7
+ "<special4>",
8
+ "<special5>",
9
+ "<special6>",
10
+ "<special7>",
11
+ "<special8>",
12
+ "<special9>"
13
+ ],
14
+ "bos_token": "<s>",
15
+ "cls_token": "</s>",
16
+ "mask_token": "<special1>",
17
+ "pad_token": "<pad>",
18
+ "sep_token": "</s>",
19
+ "unk_token": "<unk>"
20
+ }
tokenizer_config.json ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<special0>",
4
+ "<special1>",
5
+ "<special2>",
6
+ "<special3>",
7
+ "<special4>",
8
+ "<special5>",
9
+ "<special6>",
10
+ "<special7>",
11
+ "<special8>",
12
+ "<special9>"
13
+ ],
14
+ "bos_token": "<s>",
15
+ "clean_up_tokenization_spaces": true,
16
+ "cls_token": "</s>",
17
+ "do_lowercase_and_remove_accent": false,
18
+ "id2lang": {
19
+ "0": "ar",
20
+ "1": "de",
21
+ "10": "pl",
22
+ "11": "pt",
23
+ "12": "ru",
24
+ "13": "sv",
25
+ "14": "tr",
26
+ "15": "vi",
27
+ "16": "zh",
28
+ "2": "en",
29
+ "3": "es",
30
+ "4": "fr",
31
+ "5": "hi",
32
+ "6": "it",
33
+ "7": "ja",
34
+ "8": "ko",
35
+ "9": "nl"
36
+ },
37
+ "lang2id": {
38
+ "ar": 0,
39
+ "de": 1,
40
+ "en": 2,
41
+ "es": 3,
42
+ "fr": 4,
43
+ "hi": 5,
44
+ "it": 6,
45
+ "ja": 7,
46
+ "ko": 8,
47
+ "nl": 9,
48
+ "pl": 10,
49
+ "pt": 11,
50
+ "ru": 12,
51
+ "sv": 13,
52
+ "tr": 14,
53
+ "vi": 15,
54
+ "zh": 16
55
+ },
56
+ "mask_token": "<special1>",
57
+ "model_max_length": 512,
58
+ "pad_token": "<pad>",
59
+ "sep_token": "</s>",
60
+ "tokenizer_class": "XLMTokenizer",
61
+ "unk_token": "<unk>"
62
+ }
vocab.json ADDED
The diff for this file is too large to render. See raw diff