🧠 KoarAI LFM2.5-350M Training Studio

Interactive Dataset Mixer, Reasoning (CoT) Balancer & Colab Pipeline Generator for LiquidAI/LFM2.5-350M.

🎛️ 1. Dataset Mix Configuration

🇷🇺 Russian Deep CoT (ZeroAgency/ru-thinking-reasoning-r1) 700

💬 Russian Natural Chat (Vikhrmodels/GrandMaster-PRO-MAX) 700

🌐 English General Chat (LDJnr/Capybara) 800

📐 Math & Logic CoT (r0b0tlab/qwen3.8-max-distillation-50k) 600

💻 Clean Python & Code (iamtarun/python_code_instructions) 350

Learning Rate

Epochs