# Two-stage Fine-tuning

Slush is a two-stage model trained with high LoRA dropout rate, focusing on enhancing creativity and role-playing capabilities

ALMA is a two-stage trained large language model translation system based on LLaMA-2-13B, significantly improving translation performance through an innovative paradigm of monolingual data fine-tuning + parallel corpus optimization

ALMA is an advanced translator based on large language models, employing a two-stage training paradigm (monolingual fine-tuning + parallel corpus optimization). The 13B-LoRA version achieves optimal performance through LoRA fine-tuning on the LLaMA-2-13B foundation.

ALMA-13B-R is an advanced translation model based on large language models, fine-tuned using Contrastive Preference Optimization (CPO), capable of matching or even surpassing GPT-4 or WMT competition winners.

TrOCR base model for French handwritten text, trained with a two-stage fine-tuning strategy, suitable for single-line text image recognition

A DPR-based question encoder model specifically designed for long-form QA (LFQA) tasks, optimized for retrieval performance through two-stage training