LoRA Speedrun: A Public Wall-Clock Leaderboard for Fine-Tuning Techniques

GitHub / Hacker News·medium signal

A new GitHub project (Saivineeth147/lora-speedrun) publishes a wall-clock leaderboard for LoRA fine-tuning techniques, ranking approaches by actual elapsed training time rather than the FLOP counts and loss curves that dominate published comparisons. It picked up 114 points and 23 comments on HN, with the discussion focused on whether wall-clock is the honest metric for practitioners who care about iteration speed on a single GPU. Useful as a reproducible baseline for anyone choosing between PEFT variants.