README.md

---
library_name: transformers
license: mit
base_model: xlm-roberta-base
tags:
- generated_from_trainer
- language-identification
- codeswitching
metrics:
- precision
- recall
- f1
- accuracy
language:
- multilingual
- af
- am
- ar
- as
- ba
- be
- bg
- bn
- bo
- br
- bs
- ca
- ce
- ckb
- cs
- cy
- da
- de
- dv
- el
- en
- eo
- es
- et
- eu
- fa
- fi
- fr
- ga
- gd
- gl
- gu
- he
- hi
- hr
- hu
- hy
- id
- is
- it
- ja
- jv
- ka
- kk
- km
- kn
- ko
- ku
- ky
- la
- lb
- lo
- lt
- lv
- mg
- mk
- ml
- mn
- mr
- ms
- mt
- my
- ne
- nl
- 'no'
- ny
- oc
- om
- or
- pa
- pl
- ps
- pt
- rm
- ro
- ru
- sd
- si
- sk
- sl
- so
- sq
- sr
- su
- sv
- sw
- ta
- te
- tg
- th
- ti
- tl
- tr
- tt
- ug
- uk
- ur
- uz
- vi
- yo
- yi
- zh
- zu
model-index:
- name: polyglot-tagger
  results: []
datasets:
- wikimedia/wikipedia
- HuggingFaceFW/finetranslations
- google/smol
- polyglot-tagger/nlp-noise-snippets
- polyglot-tagger/wikipedia-language-snippets-filtered
- polyglot-tagger/finetranslations-filtered
- polyglot-tagger/tatoeba-filtered
pipeline_tag: token-classification
---


![image](https://cdn-uploads.huggingface.co/production/uploads/67ee3f0a66388136438834cc/OnfV_fN2br5c4cPnOn6O0.png)


Fine-tuned `xlm-roberta-base` for sentence-level language tagging across 100 languages.
The model predicts BIO-style language tags over tokens, which makes it useful for
language identification, code-switch detection, and multilingual document analysis.

> Compared to version 2.2, this version had training data that attempted to fix the model scoring common grade-school words from major langauges as low confidence or in a minor language bucket.

## Model description

Introducing Polyglot Tagger, a new way to classify multi-lingual documents. By training specifically on token classification on individual sentences, the model 
generalizes well on a variety of languages, while also behaves as a multi-label classifier, and extracts sentences based on its language. 

## Intended uses & limitations
This model can be treated as a base model for further fine-tuning on specific language identification extraction tasks. 
Note that as a general language tagging model, it can potentially get confused from shared language families or from short texts. For example, Danish and Norwegian, Spanish and Portuguese, and Russian and Ukrainian.

The model is trained on a sentence with a minimum of four tokens, so it may not accurately classify very short and ambigous statements. Note that this model is experimental
and may produce unexpected results compared to generic text classifiers. It is trained on cleaned text, therefore, "messy" text may unexpectedly produce different results.

> Note that Romanized versions of any language may have no representation in the training set, such as Romanized Russian, and Hindi.

### Training and Evaluation Data

A synthetic training row consists of 1-6 individual and mostly independent sentences extracted from various sources. To generalize well against multiple languages, several
factors were used to simulate messy text, and to reduce single character bias on certain languages:
- Low chance of deliberate accent stripping for languages such as Spanish and Portugeuse
- Random chance to add in, replace or delete punctuation, numeric, and delimiter artifiacts
- Insert same-script alphabets to family language. For example, randomly injecting Arabic characters in Arabic languages
- Random chance to change the casing of compatible language scripts, such as Latin and Cyrllic.
- Low chance of simulating OCR and messy text with character mutation.

To generalize well on both the target language and code switching a curriculum is provided:
- Pure documents 55%: Single language to learn its vocabulary, simulating a short paragraph of a single language.
- Homogenous 25%: Single language + one foreign sentence to learn simple code switching.
- Spliced 10%: A foreign sentence is centered between two same-language sentence, with the first sentence's punctuation stripped, and second sentence's forced to be lowercased.
- Mixed 10%: Generic mix of any languages.




| lang | train sentences | train tokens | eval sentences | eval tokens | all sentences | all tokens |
| :--- | ---: | ---: | ---: | ---: | ---: | ---: |
| en | 423264 (2.41%) | 9841704 (1.71%) | 3157 (3.84%) | 35025 (1.79%) | 426421 (2.41%) | 9876729 (1.71%) |
| es | 359106 (2.04%) | 9729675 (1.69%) | 2201 (2.68%) | 22340 (1.14%) | 361307 (2.04%) | 9752015 (1.69%) |
| ru | 356083 (2.02%) | 8945224 (1.56%) | 2226 (2.71%) | 21578 (1.10%) | 358309 (2.03%) | 8966802 (1.55%) |
| fr | 354645 (2.02%) | 10591338 (1.84%) | 2213 (2.69%) | 26148 (1.34%) | 356858 (2.02%) | 10617486 (1.84%) |
| ja | 352243 (2.00%) | 7945312 (1.38%) | 2219 (2.70%) | 25849 (1.32%) | 354462 (2.01%) | 7971161 (1.38%) |
| pt | 346042 (1.97%) | 8793674 (1.53%) | 2059 (2.50%) | 20881 (1.07%) | 348101 (1.97%) | 8814555 (1.53%) |
| de | 344644 (1.96%) | 8847457 (1.54%) | 2151 (2.61%) | 24958 (1.27%) | 346795 (1.96%) | 8872415 (1.54%) |
| it | 343887 (1.95%) | 8790806 (1.53%) | 2000 (2.43%) | 17342 (0.89%) | 345887 (1.96%) | 8808148 (1.53%) |
| fi | 299568 (1.70%) | 6905536 (1.20%) | 1576 (1.92%) | 14458 (0.74%) | 301144 (1.70%) | 6919994 (1.20%) |
| uk | 297565 (1.69%) | 7228987 (1.26%) | 1398 (1.70%) | 11391 (0.58%) | 298963 (1.69%) | 7240378 (1.26%) |
| zh | 294064 (1.67%) | 7329413 (1.28%) | 1717 (2.09%) | 33433 (1.71%) | 295781 (1.67%) | 7362846 (1.28%) |
| tr | 289328 (1.64%) | 6606625 (1.15%) | 1384 (1.68%) | 11089 (0.57%) | 290712 (1.64%) | 6617714 (1.15%) |
| he | 289239 (1.64%) | 7792338 (1.36%) | 1100 (1.34%) | 10342 (0.53%) | 290339 (1.64%) | 7802680 (1.35%) |
| pl | 288423 (1.64%) | 7707293 (1.34%) | 1305 (1.59%) | 11306 (0.58%) | 289728 (1.64%) | 7718599 (1.34%) |
| hu | 286880 (1.63%) | 7547282 (1.31%) | 1232 (1.50%) | 11115 (0.57%) | 288112 (1.63%) | 7558397 (1.31%) |
| nl | 280682 (1.60%) | 6057971 (1.05%) | 1296 (1.58%) | 10453 (0.53%) | 281978 (1.60%) | 6068424 (1.05%) |
| lt | 273386 (1.55%) | 6507776 (1.13%) | 1165 (1.42%) | 10491 (0.54%) | 274551 (1.55%) | 6518267 (1.13%) |
| eo | 267885 (1.52%) | 6554654 (1.14%) | 1055 (1.28%) | 13271 (0.68%) | 268940 (1.52%) | 6567925 (1.14%) |
| ar | 257437 (1.46%) | 5125573 (0.89%) | 1327 (1.61%) | 15180 (0.78%) | 258764 (1.46%) | 5140753 (0.89%) |
| cs | 240507 (1.37%) | 6406086 (1.11%) | 1082 (1.32%) | 9473 (0.48%) | 241589 (1.37%) | 6415559 (1.11%) |
| mk | 231103 (1.31%) | 6478376 (1.13%) | 953 (1.16%) | 7713 (0.39%) | 232056 (1.31%) | 6486089 (1.12%) |
| mr | 228596 (1.30%) | 5886608 (1.02%) | 776 (0.94%) | 6332 (0.32%) | 229372 (1.30%) | 5892940 (1.02%) |
| no | 223605 (1.27%) | 6137131 (1.07%) | 1396 (1.70%) | 40226 (2.05%) | 225001 (1.27%) | 6177357 (1.07%) |
| da | 222243 (1.26%) | 5375746 (0.94%) | 1201 (1.46%) | 10373 (0.53%) | 223444 (1.26%) | 5386119 (0.93%) |
| hy | 207937 (1.18%) | 6345675 (1.10%) | 791 (0.96%) | 9276 (0.47%) | 208728 (1.18%) | 6354951 (1.10%) |
| tl | 207674 (1.18%) | 5561702 (0.97%) | 1017 (1.24%) | 10926 (0.56%) | 208691 (1.18%) | 5572628 (0.97%) |
| hi | 206552 (1.17%) | 7796062 (1.36%) | 1079 (1.31%) | 47351 (2.42%) | 207631 (1.17%) | 7843413 (1.36%) |
| ko | 205625 (1.17%) | 6481034 (1.13%) | 1156 (1.41%) | 32355 (1.65%) | 206781 (1.17%) | 6513389 (1.13%) |
| el | 202334 (1.15%) | 7105554 (1.24%) | 826 (1.00%) | 13412 (0.68%) | 203160 (1.15%) | 7118966 (1.23%) |
| ro | 194999 (1.11%) | 6206913 (1.08%) | 820 (1.00%) | 14575 (0.74%) | 195819 (1.11%) | 6221488 (1.08%) |
| fa | 192050 (1.09%) | 5728246 (1.00%) | 696 (0.85%) | 14765 (0.75%) | 192746 (1.09%) | 5743011 (1.00%) |
| sk | 189330 (1.08%) | 5318617 (0.93%) | 873 (1.06%) | 20779 (1.06%) | 190203 (1.08%) | 5339396 (0.93%) |
| la | 188201 (1.07%) | 4591159 (0.80%) | 824 (1.00%) | 8557 (0.44%) | 189025 (1.07%) | 4599716 (0.80%) |
| bg | 187685 (1.07%) | 5860353 (1.02%) | 762 (0.93%) | 16804 (0.86%) | 188447 (1.07%) | 5877157 (1.02%) |
| be | 181543 (1.03%) | 6528657 (1.14%) | 869 (1.06%) | 25944 (1.32%) | 182412 (1.03%) | 6554601 (1.14%) |
| is | 180452 (1.03%) | 6146455 (1.07%) | 959 (1.17%) | 39591 (2.02%) | 181411 (1.03%) | 6186046 (1.07%) |
| lv | 179142 (1.02%) | 5867897 (1.02%) | 762 (0.93%) | 33481 (1.71%) | 179904 (1.02%) | 5901378 (1.02%) |
| ckb | 174282 (0.99%) | 7825141 (1.36%) | 667 (0.81%) | 28756 (1.47%) | 174949 (0.99%) | 7853897 (1.36%) |
| ms | 172573 (0.98%) | 4614764 (0.80%) | 815 (0.99%) | 24769 (1.26%) | 173388 (0.98%) | 4639533 (0.80%) |
| ka | 170876 (0.97%) | 5505127 (0.96%) | 673 (0.82%) | 20651 (1.05%) | 171549 (0.97%) | 5525778 (0.96%) |
| kk | 170695 (0.97%) | 5132560 (0.89%) | 676 (0.82%) | 18695 (0.95%) | 171371 (0.97%) | 5151255 (0.89%) |
| bn | 170721 (0.97%) | 6393448 (1.11%) | 441 (0.54%) | 14727 (0.75%) | 171162 (0.97%) | 6408175 (1.11%) |
| eu | 168462 (0.96%) | 5737310 (1.00%) | 746 (0.91%) | 37196 (1.90%) | 169208 (0.96%) | 5774506 (1.00%) |
| as | 168746 (0.96%) | 8564682 (1.49%) | 445 (0.54%) | 24444 (1.25%) | 169191 (0.96%) | 8589126 (1.49%) |
| mn | 167543 (0.95%) | 5678049 (0.99%) | 703 (0.85%) | 20347 (1.04%) | 168246 (0.95%) | 5698396 (0.99%) |
| ur | 165992 (0.94%) | 5361179 (0.93%) | 684 (0.83%) | 22622 (1.16%) | 166676 (0.94%) | 5383801 (0.93%) |
| oc | 165863 (0.94%) | 5735536 (1.00%) | 730 (0.89%) | 18599 (0.95%) | 166593 (0.94%) | 5754135 (1.00%) |
| ba | 164919 (0.94%) | 8387828 (1.46%) | 699 (0.85%) | 35927 (1.83%) | 165618 (0.94%) | 8423755 (1.46%) |
| th | 164429 (0.93%) | 5495248 (0.96%) | 649 (0.79%) | 22113 (1.13%) | 165078 (0.93%) | 5517361 (0.96%) |
| ky | 164374 (0.93%) | 5199548 (0.90%) | 683 (0.83%) | 18956 (0.97%) | 165057 (0.93%) | 5218504 (0.90%) |
| hr | 163828 (0.93%) | 5183677 (0.90%) | 711 (0.86%) | 33845 (1.73%) | 164539 (0.93%) | 5217522 (0.90%) |
| ps | 163238 (0.93%) | 4735113 (0.82%) | 674 (0.82%) | 18515 (0.95%) | 163912 (0.93%) | 4753628 (0.82%) |
| id | 163187 (0.93%) | 4025079 (0.70%) | 723 (0.88%) | 13371 (0.68%) | 163910 (0.93%) | 4038450 (0.70%) |
| pa | 162180 (0.92%) | 7621059 (1.33%) | 581 (0.71%) | 29036 (1.48%) | 162761 (0.92%) | 7650095 (1.33%) |
| sw | 161777 (0.92%) | 5013161 (0.87%) | 653 (0.79%) | 26493 (1.35%) | 162430 (0.92%) | 5039654 (0.87%) |
| af | 160455 (0.91%) | 4676798 (0.81%) | 932 (1.13%) | 27369 (1.40%) | 161387 (0.91%) | 4704167 (0.82%) |
| jv | 156292 (0.89%) | 4752381 (0.83%) | 576 (0.70%) | 22573 (1.15%) | 156868 (0.89%) | 4774954 (0.83%) |
| tt | 154833 (0.88%) | 5165763 (0.90%) | 578 (0.70%) | 7298 (0.37%) | 155411 (0.88%) | 5173061 (0.90%) |
| cy | 153551 (0.87%) | 5656404 (0.98%) | 653 (0.79%) | 29503 (1.51%) | 154204 (0.87%) | 5685907 (0.99%) |
| ga | 150458 (0.86%) | 5488243 (0.95%) | 680 (0.83%) | 33471 (1.71%) | 151138 (0.86%) | 5521714 (0.96%) |
| kn | 150184 (0.85%) | 14992479 (2.61%) | 697 (0.85%) | 49288 (2.52%) | 150881 (0.85%) | 15041767 (2.61%) |
| bs | 150037 (0.85%) | 4582900 (0.80%) | 649 (0.79%) | 25588 (1.31%) | 150686 (0.85%) | 4608488 (0.80%) |
| ca | 149401 (0.85%) | 5477662 (0.95%) | 629 (0.76%) | 21391 (1.09%) | 150030 (0.85%) | 5499053 (0.95%) |
| ne | 148716 (0.85%) | 4855198 (0.84%) | 535 (0.65%) | 16246 (0.83%) | 149251 (0.84%) | 4871444 (0.84%) |
| ku | 147702 (0.84%) | 4973601 (0.87%) | 574 (0.70%) | 28196 (1.44%) | 148276 (0.84%) | 5001797 (0.87%) |
| gl | 147011 (0.84%) | 4554907 (0.79%) | 658 (0.80%) | 20127 (1.03%) | 147669 (0.84%) | 4575034 (0.79%) |
| uz | 145433 (0.83%) | 4704898 (0.82%) | 573 (0.70%) | 21862 (1.12%) | 146006 (0.83%) | 4726760 (0.82%) |
| sl | 144084 (0.82%) | 3851696 (0.67%) | 651 (0.79%) | 18164 (0.93%) | 144735 (0.82%) | 3869860 (0.67%) |
| sv | 143041 (0.81%) | 4006332 (0.70%) | 905 (1.10%) | 7012 (0.36%) | 143946 (0.81%) | 4013344 (0.70%) |
| tg | 136703 (0.78%) | 7664329 (1.33%) | 572 (0.70%) | 34220 (1.75%) | 137275 (0.78%) | 7698549 (1.33%) |
| et | 131007 (0.74%) | 3280590 (0.57%) | 549 (0.67%) | 14021 (0.72%) | 131556 (0.74%) | 3294611 (0.57%) |
| br | 130223 (0.74%) | 4495403 (0.78%) | 546 (0.66%) | 17304 (0.88%) | 130769 (0.74%) | 4512707 (0.78%) |
| lb | 129528 (0.74%) | 4421411 (0.77%) | 495 (0.60%) | 17761 (0.91%) | 130023 (0.74%) | 4439172 (0.77%) |
| su | 129144 (0.73%) | 4215719 (0.73%) | 535 (0.65%) | 21391 (1.09%) | 129679 (0.73%) | 4237110 (0.73%) |
| mt | 128626 (0.73%) | 6671441 (1.16%) | 508 (0.62%) | 26729 (1.36%) | 129134 (0.73%) | 6698170 (1.16%) |
| sq | 119431 (0.68%) | 4107917 (0.71%) | 561 (0.68%) | 18633 (0.95%) | 119992 (0.68%) | 4126550 (0.72%) |
| sr | 117855 (0.67%) | 3160599 (0.55%) | 427 (0.52%) | 3505 (0.18%) | 118282 (0.67%) | 3164104 (0.55%) |
| or | 110709 (0.63%) | 3922431 (0.68%) | 410 (0.50%) | 13094 (0.67%) | 111119 (0.63%) | 3935525 (0.68%) |
| ml | 110085 (0.63%) | 10929013 (1.90%) | 464 (0.56%) | 36922 (1.89%) | 110549 (0.63%) | 10965935 (1.90%) |
| yi | 104494 (0.59%) | 4085563 (0.71%) | 400 (0.49%) | 6005 (0.31%) | 104894 (0.59%) | 4091568 (0.71%) |
| te | 101076 (0.57%) | 9757033 (1.70%) | 430 (0.52%) | 37897 (1.94%) | 101506 (0.57%) | 9794930 (1.70%) |
| ta | 94122 (0.53%) | 7917169 (1.38%) | 386 (0.47%) | 26610 (1.36%) | 94508 (0.53%) | 7943779 (1.38%) |
| mg | 93939 (0.53%) | 3291017 (0.57%) | 391 (0.48%) | 11698 (0.60%) | 94330 (0.53%) | 3302715 (0.57%) |
| si | 92723 (0.53%) | 5275463 (0.92%) | 364 (0.44%) | 18426 (0.94%) | 93087 (0.53%) | 5293889 (0.92%) |
| vi | 74916 (0.43%) | 2535825 (0.44%) | 335 (0.41%) | 3396 (0.17%) | 75251 (0.43%) | 2539221 (0.44%) |
| rm | 74806 (0.43%) | 2826708 (0.49%) | 318 (0.39%) | 12654 (0.65%) | 75124 (0.43%) | 2839362 (0.49%) |
| gu | 70961 (0.40%) | 7859622 (1.37%) | 335 (0.41%) | 28389 (1.45%) | 71296 (0.40%) | 7888011 (1.37%) |
| bo | 69565 (0.40%) | 1378245 (0.24%) | 263 (0.32%) | 5407 (0.28%) | 69828 (0.40%) | 1383652 (0.24%) |
| ug | 64297 (0.37%) | 1427585 (0.25%) | 260 (0.32%) | 4769 (0.24%) | 64557 (0.37%) | 1432354 (0.25%) |
| dv | 60328 (0.34%) | 1557497 (0.27%) | 215 (0.26%) | 5844 (0.30%) | 60543 (0.34%) | 1563341 (0.27%) |
| am | 59339 (0.34%) | 2705311 (0.47%) | 235 (0.29%) | 10768 (0.55%) | 59574 (0.34%) | 2716079 (0.47%) |
| yo | 59246 (0.34%) | 3649130 (0.63%) | 260 (0.32%) | 21157 (1.08%) | 59506 (0.34%) | 3670287 (0.64%) |
| my | 58575 (0.33%) | 2165089 (0.38%) | 214 (0.26%) | 8142 (0.42%) | 58789 (0.33%) | 2173231 (0.38%) |
| km | 57081 (0.32%) | 3056236 (0.53%) | 193 (0.23%) | 10606 (0.54%) | 57274 (0.32%) | 3066842 (0.53%) |
| so | 56160 (0.32%) | 2044409 (0.36%) | 212 (0.26%) | 8847 (0.45%) | 56372 (0.32%) | 2053256 (0.36%) |
| sd | 55359 (0.31%) | 3226018 (0.56%) | 217 (0.26%) | 10847 (0.55%) | 55576 (0.31%) | 3236865 (0.56%) |
| zu | 52465 (0.30%) | 2406841 (0.42%) | 203 (0.25%) | 9751 (0.50%) | 52668 (0.30%) | 2416592 (0.42%) |
| lo | 50641 (0.29%) | 1747495 (0.30%) | 189 (0.23%) | 6221 (0.32%) | 50830 (0.29%) | 1753716 (0.30%) |
| ti | 47785 (0.27%) | 2895617 (0.50%) | 195 (0.24%) | 12699 (0.65%) | 47980 (0.27%) | 2908316 (0.50%) |
| ce | 45014 (0.26%) | 2425219 (0.42%) | 188 (0.23%) | 9950 (0.51%) | 45202 (0.26%) | 2435169 (0.42%) |
| ny | 43552 (0.25%) | 2051132 (0.36%) | 171 (0.21%) | 8286 (0.42%) | 43723 (0.25%) | 2059418 (0.36%) |
| gd | 36623 (0.21%) | 1273243 (0.22%) | 156 (0.19%) | 3615 (0.18%) | 36779 (0.21%) | 1276858 (0.22%) |
| xh | 24432 (0.14%) | 911850 (0.16%) | 93 (0.11%) | 3528 (0.18%) | 24525 (0.14%) | 915378 (0.16%) |
| om | 15372 (0.09%) | 545603 (0.09%) | 77 (0.09%) | 2564 (0.13%) | 15449 (0.09%) | 548167 (0.10%) |
| sco | 8772 (0.05%) | 233030 (0.04%) | 37 (0.04%) | 828 (0.04%) | 8809 (0.05%) | 233858 (0.04%) |
| **total** | 17593786 (100.00%) | 574735483 (100.00%) | 82270 (100.00%) | 1958217 (100.00%) | 17676056 (100.00%) | 576693700 (100.00%) |


This model is a fine-tuned version of [xlm-roberta-base](https://huggingface.co/xlm-roberta-base) on an unknown dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0306
- Precision: 0.9507
- Recall: 0.9644
- F1: 0.9575
- Accuracy: 0.9917

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 72
- eval_batch_size: 36
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 144
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- num_epochs: 2
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch  | Step  | Validation Loss | Precision | Recall | F1     | Accuracy |
|:-------------:|:------:|:-----:|:---------------:|:---------:|:------:|:------:|:--------:|
| 0.0918        | 0.0731 | 2500  | 0.1050          | 0.7984    | 0.8818 | 0.8381 | 0.9735   |
| 0.0717        | 0.1463 | 5000  | 0.0797          | 0.8393    | 0.9041 | 0.8705 | 0.9782   |
| 0.0624        | 0.2194 | 7500  | 0.0762          | 0.8664    | 0.9166 | 0.8908 | 0.9804   |
| 0.0562        | 0.2925 | 10000 | 0.0620          | 0.8758    | 0.9247 | 0.8995 | 0.9830   |
| 0.0516        | 0.3657 | 12500 | 0.0576          | 0.8844    | 0.9298 | 0.9065 | 0.9845   |
| 0.0465        | 0.4388 | 15000 | 0.0543          | 0.8993    | 0.9357 | 0.9172 | 0.9857   |
| 0.0433        | 0.5119 | 17500 | 0.0558          | 0.9005    | 0.9356 | 0.9177 | 0.9856   |
| 0.0411        | 0.5851 | 20000 | 0.0499          | 0.9012    | 0.9385 | 0.9195 | 0.9867   |
| 0.0420        | 0.6582 | 22500 | 0.0460          | 0.9167    | 0.9438 | 0.9300 | 0.9873   |
| 0.0392        | 0.7313 | 25000 | 0.0441          | 0.9149    | 0.9448 | 0.9296 | 0.9878   |
| 0.0386        | 0.8045 | 27500 | 0.0434          | 0.9200    | 0.9476 | 0.9336 | 0.9885   |
| 0.0357        | 0.8776 | 30000 | 0.0422          | 0.9235    | 0.9503 | 0.9367 | 0.9886   |
| 0.0356        | 0.9507 | 32500 | 0.0404          | 0.9272    | 0.9520 | 0.9395 | 0.9890   |
| 0.0261        | 1.0238 | 35000 | 0.0381          | 0.9293    | 0.9529 | 0.9409 | 0.9898   |
| 0.0322        | 1.0970 | 37500 | 0.0371          | 0.9346    | 0.9558 | 0.9451 | 0.9899   |
| 0.0303        | 1.1701 | 40000 | 0.0374          | 0.9375    | 0.9580 | 0.9476 | 0.9903   |
| 0.0276        | 1.2432 | 42500 | 0.0378          | 0.9355    | 0.9566 | 0.9460 | 0.9901   |
| 0.0264        | 1.3164 | 45000 | 0.0353          | 0.9373    | 0.9574 | 0.9472 | 0.9904   |
| 0.0228        | 1.3895 | 47500 | 0.0366          | 0.9398    | 0.9589 | 0.9493 | 0.9903   |
| 0.0234        | 1.4626 | 50000 | 0.0343          | 0.9430    | 0.9602 | 0.9516 | 0.9907   |
| 0.0274        | 1.5358 | 52500 | 0.0339          | 0.9396    | 0.9591 | 0.9492 | 0.9906   |
| 0.0236        | 1.6089 | 55000 | 0.0324          | 0.9438    | 0.9613 | 0.9525 | 0.9913   |
| 0.0244        | 1.6820 | 57500 | 0.0322          | 0.9478    | 0.9624 | 0.9551 | 0.9914   |
| 0.0222        | 1.7552 | 60000 | 0.0323          | 0.9483    | 0.9628 | 0.9555 | 0.9914   |
| 0.0238        | 1.8283 | 62500 | 0.0320          | 0.9480    | 0.9630 | 0.9554 | 0.9913   |
| 0.0223        | 1.9014 | 65000 | 0.0320          | 0.9485    | 0.9637 | 0.9560 | 0.9913   |
| 0.0208        | 1.9746 | 67500 | 0.0306          | 0.9507    | 0.9644 | 0.9575 | 0.9917   |


### Framework versions

- Transformers 5.0.0
- Pytorch 2.10.0+cu128
- Datasets 4.0.0
- Tokenizers 0.22.2