Add support for RoFormer models

[xenova]

Example usage:

import { pipeline } from '@xenova/transformers';

const pipe = await pipeline('fill-mask', 'Xenova/antiberta2');

const output = await pipe('Ḣ Q V Q ... C A [MASK] D ... T V S S');
console.log(output);

See output

[
  {
    score: 0.48774364590644836,
    token: 19,
    token_str: 'R',
    sequence: 'Ḣ Q V Q C A R D T V S S'
  },
  {
    score: 0.2768442928791046,
    token: 18,
    token_str: 'Q',
    sequence: 'Ḣ Q V Q C A Q D T V S S'
  },
  {
    score: 0.0890476182103157,
    token: 13,
    token_str: 'K',
    sequence: 'Ḣ Q V Q C A K D T V S S'
  },
  {
    score: 0.05106702819466591,
    token: 14,
    token_str: 'L',
    sequence: 'Ḣ Q V Q C A L D T V S S'
  },
  {
    score: 0.021606773138046265,
    token: 8,
    token_str: 'E',
    sequence: 'Ḣ Q V Q C A E D T V S S'
  }
]

Also, as expected, the (unquantized) JS version produces the exact same results as the python library. Example code can be found in the alchemab/antiberta2 model card.

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.