Hugging Face study: predicting fine-tuning duration
What happened
Hugging Face published a study on 19 November 2026 examining why fine-tuning jobs take longer than FLOPs estimates suggest. The study draws on 272 production SFT jobs totalling about 7,382 GPU-hours on NVIDIA H200: conventional FLOP estimates reached only R²=0.31 with an MAE of 12.41 hours, whereas a CatBoost regressor using only pre-scheduling information achieved R²=0.98, MAE 1.71 hours and median error 0.51 hours in 5-fold cross-validation. So far only the results have been published, with no reports yet of follow-up applications or reproductions.
Written by AI from the reports · updated 2 h ago
Timeline
Follow the reports to see the event from each side.
- Hugging Face BlogWhy your fine-tuning job takes longer than the FLOPs say
Predicting end-to-end fine-tuning time on shared GPU clusters, from 272 production runs. A preview of our upcoming paper, Predicting End-to-End Fine-Tuning Time for Generative Language Models . The full version will be on arXiv soon. GPUs are among the most expensive and scarcest resources in AI today.
Heat of this event
Not enough continuous observations yet to draw a trend.