No-Regret Strategy Solving in Imperfect-Information Games via Pre-Trained Embedding