torchao.float8

Created On: Jan 29, 2026 | Last Updated On: Jan 29, 2026

Main float8 training APIs

convert_to_float8_training

Swaps torch.nn.Linear in module with Float8Linear.

Other float8 training types

Float8LinearConfig	Configuration for converting a torch.nn.Linear module to float8 for training.
Float8LinearRecipeName	Pre-made recipes for common float8 training configurations.
CastConfig	Configuration for casting a single tensor to float8.
ScalingType	Defines the type of scaling to use for casting to float8.
ScalingGranularity	Defines the granularity of scaling strategies for casting to float8.
precompute_float8_dynamic_scale_for_fsdp	Calculate scale dynamically for all float8 parameters.