sam-writer

{
    "title": "Schema for validation Qordoba Spacy Match/Replace format",
    "type": "object",
    "definitions": {
        "spacyMatch": {
            "type": "array",
            "items": {
                "$ref": "#/definitions/spacyAttribute"
            },
            "minItems": 1

sam-writer

from fitbert import FitBert


# currently supported models: bert-large-uncased and distilbert-base-uncased
# this takes a while and loads a whole big BERT into memory
fb = FitBert()

masked_string = "Why Bert, you're looking ***mask*** today!"
options = ['buff', 'handsome', 'strong']

sam-writer

unmasked_string = "Why Bert, you're looks handsome today!"
span_to_mask = (17, 22)

filled_in = fb.mask_fitb(unmasked_string, span_to_mask)
# >>> "Why Bert, you're looking handsome today!"

sam-writer

masked_string = "Your 17 ***mask*** burritos are on their way !"
options = ['hot', 'cold', 'sweet', 'delicious', 'artisanal']
fb.fitb(masked_string, options=options)

# >>> 'Your 17 delicious burritos are on their way !'

sam-writer

# example from "Targeted Syntactic Evaluation of Language Models"
# https://arxiv.org/abs/1808.09031

masked_string = "the author that the guard likes ***mask***"
options = ['laugh', 'laughs']
fb.rank_with_prob(masked_string, options)

# >>> (['laughs', 'laugh'], [4.14195717654553e-12, 3.3748110100755013e-13])

sam-writer

mkpoetryproj ()
{
    if [ $# -eq 1 ]; then
        poetry new "$1"
        cd "$1" || exit
        # get gitignore
        curl https://raw.githubusercontent.com/github/gitignore/master/Python.gitignore -o .gitignore
        {
            echo ""
            echo ".vscode/"

sam-writer

import itertools
from typing import Tuple


def ucp_to_utf16_charmap(s: str):
  """
  mostly copied from
  https://stackoverflow.com/questions/56280011/keeping-java-string-offsets-with-unicode-consistent-in-python

  converts from python indices (unicode code points) to indices

sam-writer

import torch
from torch import nn
from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
from transformers.modeling_outputs import SequenceClassifierOutput


class T5EncoderClassificationHead(nn.Module):
    """Head for sentence-level classification tasks."""

    def __init__(self, config):