🚀 Efficiently (pre)training foundation models with native PyTorch features, including FSDP for training and SDPA implementation of Flash attention v2.

This is the official repository for Peacock: A Family of Arabic Multimodal Large Language Models and Benchmarks.

French instruction-following and chat models

BEAMetrics: Benchmark to Evaluate Automatic Metrics in Natural Language Generation

The French summarization dataset introduced in "BARThez: a Skilled Pretrained French Sequence-to-Sequence Model".

Keras, PyTorch, and NumPy Implementations of Deep Learning Architectures for NLP