UzRoBERTa-v2

This model achieves the following results on the evaluation set:

• Loss: 1.9097

How to use

>>> from transformers import pipeline
>>> unmasker = pipeline('fill-mask', model='sinonimayzer/UzRoBERTa-v2')
>>> unmasker("Kuchli yomg‘irlar tufayli bir qator <mask> kuchli sel oqishi kuzatildi.")

[{'score': 0.3318027853965759,
  'token': 4877,
  'token_str': ' hududlarda',
  'sequence': 'Kuchli yomg‘irlar tufayli bir qator hududlarda kuchli sel oqishi kuzatildi.'},
 {'score': 0.13175441324710846,
  'token': 14470,
  'token_str': ' viloyatlarda',
  'sequence': 'Kuchli yomg‘irlar tufayli bir qator viloyatlarda kuchli sel oqishi kuzatildi.'},
 {'score': 0.09735308587551117,
  'token': 13555,
  'token_str': ' tumanlarda',
  'sequence': 'Kuchli yomg‘irlar tufayli bir qator tumanlarda kuchli sel oqishi kuzatildi.'},
 {'score': 0.09112472087144852,
  'token': 12261,
  'token_str': ' shaharlarda',
  'sequence': 'Kuchli yomg‘irlar tufayli bir qator shaharlarda kuchli sel oqishi kuzatildi.'},
 {'score': 0.05940879508852959,
  'token': 2767,
  'token_str': ' joylarda',
  'sequence': 'Kuchli yomg‘irlar tufayli bir qator joylarda kuchli sel oqishi kuzatildi.'}]

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 92
• eval_batch_size: 92
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 500000

Training results

Training Loss	Epoch	Step	Validation Loss
2.3673	0.25	100000	2.4588
2.0797	0.51	200000	2.1653
1.9369	0.76	300000	2.0265
1.8545	1.02	400000	1.9456
1.8133	1.27	500000	1.9101

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.1+cu121
• Datasets 2.15.0
• Tokenizers 0.15.0