torchref.base.targets.xray_bhattacharyya module

Bhattacharyya overlap X-ray loss math.

Mirrors the loss computation in BhattacharyyaXrayTarget.forward after sigma_m has been computed in a no-grad block (so sigma_m is treated as a constant input here).

torchref.base.targets.xray_bhattacharyya.bhattacharyya_xray_loss_math(F_obs, F_calc, sigma_d, sigma_m, mask)

Bhattacharyya overlap loss between data and model Gaussians.

L_h = (F_obs - |F_calc|)^2 / (4 * (sigma_d^2 + sigma_m^2))

• 0.5 * log((sigma_d^2 + sigma_m^2) / (2 * sigma_d * sigma_m))

Dispatches to torchref.base.targets.triton.xray_bhattacharyya.bhattacharyya_xray_loss_math_triton() on CUDA float32; falls back to the eager implementation otherwise.