by NYU MLL
GLUE (General Language Understanding Evaluation) is a collection of English datasets for training, evaluating and analyzing natural language understanding systems. It is for NLP researchers benchmarking models on tasks such as entailment, paraphrase, sentiment and acceptability.
DatasetFree
Step 1: Install the Hugging Face datasets library with pip install datasets.
Step 2: Load one task by name, for example load_dataset("nyu-mll/glue", "mrpc").
Step 3: Train on the task's train split and tune on its validation split.
Step 4: Generate predictions for the test split; test labels are hidden, so submit to the GLUE leaderboard to get scores.
Step 5: For ax, use a model trained on MNLI to produce predictions.
CoLA, SST-2, MRPC, QQP, STS-B, MNLI, QNLI, RTE and WNLI, plus the ax diagnostic dataset.
GLUE tasks are derived from existing datasets, and the card refers users to the original license of each dataset.
The GLUE leaderboard is at gluebenchmark.com/leaderboard.
English.
MONETMONET is a dataset of 103.8 million English image-text pairs for training text-to-image models. It was distilled from 2.9 billion raw pairs through safety, domain and duplicate filtering.EgoProEgoPro is Lightwheel's 10,000-hour egocentric dataset pairing synchronized head and wrist video with 3D hand pose. It is for teams training embodied AI on human demonstrations.