MCPcopy Create free account
hub / github.com/conflow-dev/ConFlow / _e_reduce_max_zero_grad

Function _e_reduce_max_zero_grad

operators.py:144–146  ·  view source on GitHub ↗
(op, grad)

Source from the content-addressed store, hash-verified

142 return trainer_ops.e_reduce_max_zero(inputs,eid_low=(eid&0xFFFFFFFF),eid_high=(eid>>32),times=times)
143@ops.RegisterGradient("EReduceMaxZero")
144def _e_reduce_max_zero_grad(op, grad):
145 with tf.name_scope("EReduceMaxZeroGrad"), tf.xla.experimental.jit_scope(compile_ops=False):
146 return trainer_ops.e_reduce_max_zero_grad(grad,op.inputs[0],eid_low=op.get_attr("eid_low"), eid_high=op.get_attr("eid_high"),times = op.get_attr("times"))
147
148def e_reduce_min_one(inputs):
149 global eid,times

Callers

nothing calls this directly

Calls

no outgoing calls

Tested by

no test coverage detected