stefan-it
Follow
Stars
TransMI: A Framework to Create Strong Baselines from Multilingual Pretrained Language Models for Transliterated Data
ParaNames: A multilingual resource for parallel names
A native PyTorch Library for large model training
XAMPLER: Learning to Retrieve Cross-Lingual In-Context Examples
Raw data, scripts, etc. to produce the tables and figures of our technical report
A byte-level decoder architecture that matches the performance of tokenized Transformers.
Inspection tool for characterizing the semantic compositionality of subword tokenization in English
Evaluation of language models on mono- or multilingual tasks.
Investigating Gender Bias in Turkish Language Models
Experiments for efforts to train a new and improved t5
Official implementation of "A Multi-level Framework for Accelerating Training Transformer Models""
Open weights language model from Google DeepMind, based on Griffin.
Code for 'LLM2Vec: Large Language Models Are Secretly Powerful Text Encoders'
Recognition of news agency mentions in historical news articles (BERT-based token classification).
Temporary remove unused tokens during training to save ram and speed.
Pocket-Sized Multimodal AI for content understanding and generation across multilingual texts, images, and 🔜 video, up to 5x faster than OpenAI CLIP and LLaVA 🖼️ & 🖋️
Evaluation of the Fundus News Scraper https://github.com/flairNLP/fundus
Hetzner Online Community Project
ChroniclingAmericaQA: A Large-scale Question Answering Dataset based on Historical American Newspaper Pages
OCR, layout analysis, reading order, line detection in 90+ languages
The implementation our EMNLP 2021 paper "Enhanced Language Representation with Label Knowledge for Span Extraction".
Code for preprocessing data for UD annotations and for tagging/parsing experiments of MaiBaam
Data and code: "Answering legal questions from laymen in German civil law system", Büttner & Habernal, EACL'24
Language models scale reliably with over-training and on downstream tasks