README.md

metadata

license: apache-2.0
language:
  - en
base_model:
  - answerdotai/ModernBERT-base
  - answerdotai/ModernBERT-large
base_model_relation: quantized
tags:
  - fill-mask
  - masked-lm
  - long-context
  - modernbert

ModernBERT-CoreML

This repo contains ModernBERT-base and ModernBERT-large converted to CoreML.

Example Usage

import CoreML
import Tokenizers

let text = "The capital of Ireland is [MASK]."

print("Loading…")
let model = try await ModernBERT_base.load()
let tokenizer = try await AutoTokenizer.from(pretrained: "answerdotai/ModernBERT-base")

let tokens = tokenizer(text)
let inputIDs = MLShapedArray(scalars: tokens.map(Int32.init), shape: [1, tokens.count])
let attentionMask = MLShapedArray<Int32>(scalars: tokens.map { _ in 1 }, shape: [1, tokens.count])
let input = ModernBERT_baseInput(
    input_ids: inputIDs,
    attention_mask: attentionMask
)

print("Predicting…")
let output = try await model.prediction(input: input)
let logits = output.logitsShapedArray
let maskPosition = tokens.firstIndex(
    of: tokenizer.encode(text: "[MASK]", addSpecialTokens: false)[0]
)!
let maskLogits = logits[0, maskPosition]
let maskToken = Int(await MLTensor(maskLogits).argmax().shapedArray(of: Int32.self).scalar!)
let maskText = tokenizer.decode(tokens: [maskToken])

// The capital of Ireland is Dublin.
print(text.replacingOccurrences(of: "[MASK]", with: maskText.trimmingCharacters(in: .whitespaces)))