ConvertToInt8Pass¶
- class paddle.fluid.contrib.slim.quantization.quantization_pass. ConvertToInt8Pass ( scope, place, quantizable_op_type=None ) [source]
-
-
apply
(
graph
)
apply¶
-
Convert weights’ type of the graph. After that, the data type of the graph weights is int8_t.
- Parameters
-
graph (IrGraph) – the applied graph.
- Returns
-
None
-
apply
(
graph
)