Dharma-AI R&D
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 1
- Learned predictors improve LLM fine-tuning duration estimates
Dharma-AI R&D analyzed 272 production supervised fine-tuning jobs totaling about 7,382 GPU-hours on NVIDIA H200 clusters. In five-fold cross-validation, its CatBoost predictor achieved R² of 0.98 and a mean absolute error of 1.71 hours, compared with R² of 0.31 and 12.41 hours for the tested FLOP-based baseline. The study covers only SFT workloads on H200 infrastructure, and the full paper is planned for arXiv.
Hugging Face · 🔥 0
Experience and discussion from the community
Share my Dharma-AI R&D experienceAsk about Dharma-AI R&D
Nobody has shared their experience with Dharma-AI R&D yet.