sciai.common
Common module, involving datasets, initializers, optimizers, and train cells.
API Name |
Description |
Supported Platforms |
Common data generator. |
|
|
L-BFGS training function, which can only run on PYNATIVE mode currently. |
|
|
L-BFGS second-order optimizer, which is currently only supported in PYNATIVE_MODE. |
|
|
Yann LeCun Normal Initialization \({N}(0, \text{sigma}^2)\) in order to initialize a tensor, where |
|
|
Yann LeCun Normal Initialization \({U}(-\text{boundary}, \text{boundary})\) in order to initialize a tensor, where |
|
|
Common data sampler. |
|
|
Generates an array with values sampled from Standard Uniform distribution \({U}(-\text{boundary}, \text{boundary})\) in order to initialize a tensor, where |
|
|
TrainOneStepCell with callbacks, which can handle multi-losses. |
|
|
Cell with gradient descent, similar to nn.TrainOneStepCell, but can accept multi-losses return. |
|
|
Xavier Truncated Normal Initialization with clip of 2 times of stddev from mean of Xavier Normal Initialization. |
|