Scalene: a high-performance, high-precision CPU, GPU, and memory profiler for Python

This release adds 3.14 compatibility, uses a more robust CPU/GPU sampling approach to avoid pathological behavior when the sampling rate aliases native execution epochs, and incorporates a prototype enhanced support for profiling certain Neuron workloads (mostly due to @nehamur). It also includes some bug fixes.

What's Changed

• Fixes a line-level attribution bug when using @profile by @emeryberger
• Py314 update by @emeryberger in https://github.com/plasma-umass/scalene/pull/934
• Add neuron profiling support to Scalene GUI and profiler by @nehamur in https://github.com/plasma-umass/scalene/pull/935
• Typed dict for arguments, fixing a bug by @emeryberger in https://github.com/plasma-umass/scalene/pull/937
• Refactor scalene_profiler.py to improve modularity and reduce complexity by @Copilot in https://github.com/plasma-umass/scalene/pull/938
• Cpu Timestamp + Visualization Changes by @nehamur in https://github.com/plasma-umass/scalene/pull/940
• Random time samples added to GUI by @nehamur in https://github.com/plasma-umass/scalene/pull/942
• Geometric sampling update by @emeryberger in https://github.com/plasma-umass/scalene/pull/943

New Contributors

• @nehamur made their first contribution in https://github.com/plasma-umass/scalene/pull/935

Full Changelog: https://github.com/plasma-umass/scalene/compare/v1.5.53...v1.5.54