Edit model card

Model Description

This is a Khmer Language fill masked build on top of pre-trained model of FacebookAI/xlm-roberta-base. This model is fine-tunned with around 26K+ khmer sentences/clauses (80% for training set & 20% for validation set). This model is perform well with Khmer Language ONLY.

Model Usage

>>> from transformers import pipeline
>>> unmasker = pipeline('fill-mask', model='channudam/khmer-xlm-roberta-base')
>>> unmasker("αž’αžΆαž€αžΆαžŸαž’αžΆαžαž»αž€αŸ’αžŠαŸ…αžαŸ’αž›αžΆαŸ†αž„ αž…αžΌαžšαž’αŸ’αž“αž€αž•αžΉαž€<mask>αž²αŸ’αž™αž”αžΆαž“αž…αŸ’αžšαžΎαž“αŸ”")

[
  {
    'score': 0.9788032174110413,
    'token': 41440,
    'token_str': 'αž‘αžΉαž€',
    'sequence': 'αž’αžΆαž€αžΆαžŸαž’αžΆαžαž»αž€αŸ’αžŠαŸ…αžαŸ’αž›αžΆαŸ†αž„ αž…αžΌαžšαž’αŸ’αž“αž€αž•αžΉαž€αž‘αžΉαž€ αž²αŸ’αž™αž”αžΆαž“αž…αŸ’αžšαžΎαž“αŸ”'
  },
  {
    'score': 0.012485685758292675,
    'token': 191670,
    'token_str': 'αžŸαŸ’αžšαžΆ',
    'sequence': 'αž’αžΆαž€αžΆαžŸαž’αžΆαžαž»αž€αŸ’αžŠαŸ…αžαŸ’αž›αžΆαŸ†αž„ αž…αžΌαžšαž’αŸ’αž“αž€αž•αžΉαž€αžŸαŸ’αžšαžΆ αž²αŸ’αž™αž”αžΆαž“αž…αŸ’αžšαžΎαž“αŸ”'
  },
  {
    'score': 0.0014946138253435493,
    'token': 162483,
    'token_str': 'αž”αžΆαž™',
    'sequence': 'αž’αžΆαž€αžΆαžŸαž’αžΆαžαž»αž€αŸ’αžŠαŸ…αžαŸ’αž›αžΆαŸ†αž„ αž…αžΌαžšαž’αŸ’αž“αž€αž•αžΉαž€αž”αžΆαž™ αž²αŸ’αž™αž”αžΆαž“αž…αŸ’αžšαžΎαž“αŸ”'
  },
  {
    'score': 0.001305083278566599,
    'token': 49245,
    'token_str': 'ស៊ី',
    'sequence': 'αž’αžΆαž€αžΆαžŸαž’αžΆαžαž»αž€αŸ’αžŠαŸ…αžαŸ’αž›αžΆαŸ†αž„ αž…αžΌαžšαž’αŸ’αž“αž€αž•αžΉαž€αžŸαŸŠαžΈ αž²αŸ’αž™αž”αžΆαž“αž…αŸ’αžšαžΎαž“αŸ”'
  },
  {
    'score': 0.0007108347490429878,
    'token': 51863,
    'token_str': 'αž‘αžΉαž€',
    'sequence': 'αž’αžΆαž€αžΆαžŸαž’αžΆαžαž»αž€αŸ’αžŠαŸ…αžαŸ’αž›αžΆαŸ†αž„ αž…αžΌαžšαž’αŸ’αž“αž€αž•αžΉαž€ αž‘αžΉαž€ αž²αŸ’αž™αž”αžΆαž“αž…αŸ’αžšαžΎαž“αŸ”'
  }
]
Downloads last month
17
Safetensors
Model size
278M params
Tensor type
F32
Β·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.