◆ OpPrecision
| Enumerator |
|---|
| FP32 | |
| FP16 | |
| BF16 | |
| FP8 | |
| Auto | |
| Default | |
◆ AMPConfig()
◆ getComputeType()
template<typename T >
| cudaDataType_t Mila::Dnn::Compute::AMPConfig::getComputeType |
( |
| ) |
const |
|
inline |
◆ getCublasComputeType()
template<typename TPrecision , typename TCompute = float>
| cublasComputeType_t Mila::Dnn::Compute::AMPConfig::getCublasComputeType |
( |
| ) |
const |
|
inline |
Get the appropriate cuBLASLt compute type based on storage and precision settings.
- Template Parameters
-
| TPrecision | The storage data type |
| TCompute | The desired computation data type |
- Returns
- cublasComputeType_t The recommended cuBLASLt compute type
◆ useLossScaling()
| bool Mila::Dnn::Compute::AMPConfig::useLossScaling |
( |
| ) |
const |
|
inline |
◆ useReducedPrecision()
| bool Mila::Dnn::Compute::AMPConfig::useReducedPrecision |
( |
| ) |
const |
|
inline |
◆ useTensorCores()
| bool Mila::Dnn::Compute::AMPConfig::useTensorCores |
( |
| ) |
const |
|
inline |
Determine if TensorCore acceleration should be used.
- Returns
- bool True if TensorCore acceleration should be used
◆ precision_
The documentation for this class was generated from the following file: