GitHub Repository

Profine automatically profiles and optimizes PyTorch training jobs on real GPUs, delivering measurable speedups and lower GPU costs before teams waste days tuning configs by hand.

17 starsPython

Profine – Profile and rewrite your PyTorch training loop on real GPUs

Name: Profine – Profile and rewrite your PyTorch training loop on real GPUs
Availability: InStock
Author: aisinghal

by aisinghal·May 12, 2026·4 points·0 comments

Visit Project View on HN

AI Analysis

●●●BangerBig BrainSolve My Problem

Automates the painful torch.compile and mixed-precision tuning loop with measured 3x speedups.

Strengths

•Runs on real Modal GPU hardware instead of relying on static analysis or simulators.
•Outputs reproducible artifacts with JSON reports for every pipeline optimization step.
•Supports local LLMs via OpenAI-compatible servers to keep the rewrite logic private.

Weaknesses

•Requires a Modal account and external LLM setup, adding friction to the local dev loop.
•Smaller models under 7B parameters may struggle to generate valid code suggestions.