# Multi-task Pre-training

Tarsier-34b is an open-source large-scale video-language model focused on generating high-quality video captions and achieving leading results in multiple public benchmarks.

A lightweight sentence embedding model based on BERT for efficient calculation of sentence similarity

This is a pre-trained language model specifically for Nepali, with research results published in the 2nd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 12th International Joint Conference on Natural Language Processing (2022).

The MTL-data-to-text model is specifically designed for data-to-text generation tasks, utilizing supervised pre-training on mixed-annotated data-to-text datasets.

The MTL-Summarization Model is a text generation model specifically designed for summarization tasks, trained through supervised pre-training on multiple annotated summarization datasets.

This is a T5 model customized for Arabic and English tasks, serving as an alternative to the google/mt5-base model, with a smaller size and a focus on Arabic and English-related tasks.

This is a model based on sentence-transformers that can map sentences and paragraphs into a 768-dimensional dense vector space, suitable for tasks such as sentence similarity calculation and semantic search.

A T5 model customized for Arabic and English tasks, smaller in size and optimized for specific languages.

RuBERT is a Russian BERT model trained on Russian Wikipedia and news data, with 180 million parameters, supporting masked language modeling and next sentence prediction tasks.

A pre-trained model based on the T5 architecture, specifically designed for Python code summarization with multi-task support

Python code documentation generation model based on T5-large architecture, supporting multi-task processing