MCPcopy Create free account
hub / github.com/DeepRec-AI/DeepRec / call

Method call

tensorflow/python/eager/backprop_test.py:1165–1183  ·  view source on GitHub ↗
(c, x)

Source from the content-addressed store, hash-verified

1163
1164 @def_function.function
1165 def call(c, x):
1166
1167 @custom_gradient.custom_gradient
1168 def _call():
1169 y = c(x)
1170
1171 def grad(dy, variables=None): # pylint: disable=redefined-outer-name
1172 with backprop.GradientTape(persistent=True) as g:
1173 g.watch(variables)
1174 y = c(x)
1175 grad_vars = [
1176 2 * math_ops.reduce_sum(dy * g.jacobian(y, v)) for v in variables
1177 ]
1178 del g
1179 return (), grad_vars
1180
1181 return y, grad
1182
1183 return _call()
1184
1185 c = MyCallable()
1186 x = constant_op.constant([1., 2., 3.])

Callers

nothing calls this directly

Calls 1

_callFunction · 0.85

Tested by

no test coverage detected