This is a roberta pre-trained version on the CodeSearchNet dataset for javascript Mask Language Model mission.

To load the model: (necessary packages: !pip install transformers sentencepiece)

from transformers import AutoTokenizer, AutoModelWithLMHead, pipeline
tokenizer = AutoTokenizer.from_pretrained("dbernsohn/roberta-javascript")
model = AutoModelWithLMHead.from_pretrained("dbernsohn/roberta-javascript")

fill_mask = pipeline(

You can then use this model to fill masked words in a Java code.

code = """
var i;
for (i = 0; i < cars.<mask>; i++) {
  text += cars[i] + "<br>";

pred = {x["token_str"].replace("Ġ", ""): x["score"] for x in fill_mask(code)}
sorted(pred.items(), key=lambda kv: kv[1], reverse=True)
# [('length', 0.9959614872932434),
#  ('i', 0.00027875584783032537),
#  ('len', 0.0002283261710545048),
#  ('nodeType', 0.00013731322542298585),
#  ('index', 7.5289819505997e-05)]

The whole training process and hyperparameters are in my GitHub repo

Created by Dor Bernsohn

Downloads last month
Hosted inference API

Mask token: <mask>