Articles by tag: fine-tuning

Qwen: 942 mln downloads of the open-source leader

Qwen from Alibaba surpassed Llama in downloads and fine-tunes. ATOM Report analysis: 942 mln downloads, 69% new adaptations. Trends for developers — read details and choose stack.

Fine-tuning Mistral 7B on Russian for dialogues

Learn why LoRA didn't work and how full fine-tuning Mistral 7B achieved BertScore 0.93. Architecture, optimizations, MERA benchmarks for ML engineers.

Restoration of ruGPT-3 XL for transformers

Learn how to convert legacy ruGPT-3 XL to HF format for fine-tuning and inference. Weight structure, custom classes, MERA benchmarks. Instructions for developers.