Datasets:

Modalities:
Text
Libraries:
Datasets
syntaxgym / test.py
Jon Gauthier
Download test suite materials dynamically rather than distributing JSONs
3dad049
raw
history blame
383 Bytes
import itertools
from typing import List
import datasets
import numpy as np
import transformers
import torch
dataset = datasets.load_dataset("syntaxgym", "subordination_src-src")
metric = datasets.load_metric("syntaxgym")
model_ref = "gpt2"
# model_ref = "hf-internal-testing/tiny-random-gpt_neo"
result = metric.compute(suite=dataset["test"], model_id=model_ref)
print(result)