quark.torch.export.nn.modules.qparamslinear

Module Contents

Functions

normalize_e4m3fn_to_e4m3fnuz(→ Tuple[torch.Tensor, ...)

normalize_e4m3fn_to_e4m3fnuz for amd gpu

quark.torch.export.nn.modules.qparamslinear.normalize_e4m3fn_to_e4m3fnuz(weight: torch.Tensor, qinput: torch.Tensor, weight_scale: torch.Tensor, input_scale: torch.Tensor | None = None) → Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor | None]

normalize_e4m3fn_to_e4m3fnuz for amd gpu