{ }
View source on GitHub |
A general quantization scheme is being developed in tf.contrib.quantize
. (deprecated)
tf.compat.v1.train.do_quantize_training_on_graphdef(
input_graph, num_bits
)
Consider using that instead, though since it is in the tf.contrib namespace, it is not subject to backward compatibility guarantees.
Args | |
---|---|
input_graph
|
A GraphDef .
|
num_bits
|
The number of bits for quantize training. |
Returns | |
---|---|
The graph with quantize training done. |