LogSoftmax¶
Overview¶
A primitive to perform logsoftmax. More…
// typedefs typedef dnnl_softmax_desc_t dnnl_logsoftmax_desc_t; // structs struct dnnl::logsoftmax_backward; struct dnnl::logsoftmax_forward; // global functions dnnl_status_t DNNL_API dnnl_logsoftmax_forward_desc_init( dnnl_logsoftmax_desc_t* logsoftmax_desc, dnnl_prop_kind_t prop_kind, const dnnl_memory_desc_t* data_desc, int logsoftmax_axis ); dnnl_status_t DNNL_API dnnl_logsoftmax_backward_desc_init( dnnl_logsoftmax_desc_t* logsoftmax_desc, const dnnl_memory_desc_t* diff_data_desc, const dnnl_memory_desc_t* data_desc, int logsoftmax_axis );
Detailed Documentation¶
A primitive to perform logsoftmax.
See also:
LogSoftmax in developer guide
Typedefs¶
typedef dnnl_softmax_desc_t dnnl_logsoftmax_desc_t
A descriptor of a LogSoftmax operation.
An alias of Softmax structure, but primitive_kind must be dnnl_logsoftmax.
Global Functions¶
dnnl_status_t DNNL_API dnnl_logsoftmax_forward_desc_init( dnnl_logsoftmax_desc_t* logsoftmax_desc, dnnl_prop_kind_t prop_kind, const dnnl_memory_desc_t* data_desc, int logsoftmax_axis )
Initializes a descriptor for logsoftmax forward propagation primitive.
Parameters:
logsoftmax_desc |
Output descriptor for a logsoftmax primitive. |
prop_kind |
Propagation kind. Possible values are dnnl_forward_training and dnnl_forward_inference. |
data_desc |
Source and destination memory descriptor. |
logsoftmax_axis |
Axis over which logsoftmax is computed. |
Returns:
dnnl_success on success and a status describing the error otherwise.
dnnl_status_t DNNL_API dnnl_logsoftmax_backward_desc_init( dnnl_logsoftmax_desc_t* logsoftmax_desc, const dnnl_memory_desc_t* diff_data_desc, const dnnl_memory_desc_t* data_desc, int logsoftmax_axis )
Initializes a descriptor for logsoftmax backward propagation primitive.
Parameters:
logsoftmax_desc |
Output descriptor for a logsoftmax primitive. |
diff_data_desc |
Diff source and diff destination memory descriptors. |
data_desc |
Destination memory descriptor. |
logsoftmax_axis |
Axis over which softmax is computed. |
Returns:
dnnl_success on success and a status describing the error otherwise.