Datasets:

---

jhu-clsp

/

mmBERT-pretrain-p3-others

like 0

Follow

Center for Language and Speech Processing @ JHU 183

This PR updates the dataset card for mmbert-pretrain-p1-fineweb2-langs to:

• Change the language metadata tag from en to mul to accurately reflect its multilingual nature (covering 60+ languages in this phase).
• Update the task_categories from fill-mask to feature-extraction to better describe the primary use of models trained with this data (encoders for various downstream tasks like classification and retrieval).
• Add a "Sample Usage of mmBERT Models" section, including code snippets from the GitHub README, to demonstrate how models trained using this pre-training data can be installed and used for tasks such as feature extraction and masked language modeling.

@nielsr is "mul" an actual label? I'm beginning to wonder if these are AI-generated or perhaps just typos?

Hi @orionweller , yes those PRs are automated, "mul" is indeed wrong, it should be "multilingual". Thanks for the feedback