0xAlcibiades

Use when editing or adding Burn 0.21 (Rust ML framework on cubecl/wgpu/cuda/ndarray/flex backends) code — models, training loops, tensor ops, datasets, save/load, or backend setup. Reach for it on errors like trait-method-not-in-scope (Module/Config/Dataset), recursion_limit on wgpu, bf16-on-CPU, AutodiffBackend / inner-backend mismatches, and when unsure about Param vs constant, Learner vs manual loop, or which recorder/store to use.