| /external/tensorflow/tensorflow/python/keras/optimizer_v2/ |
| D | adagrad.py | 86 def _prepare_local(self, var_device, var_dtype, apply_state): argument
87 super(Adagrad, self)._prepare_local(var_device, var_dtype, apply_state)
88 apply_state[(var_device, var_dtype)].update(
92 neg_lr_t=-apply_state[(var_device, var_dtype)]['lr_t'],
127 def _resource_apply_dense(self, grad, var, apply_state=None): argument
129 coefficients = ((apply_state or {}).get((var_device, var_dtype))
141 def _resource_apply_sparse(self, grad, var, indices, apply_state=None): argument
143 coefficients = ((apply_state or {}).get((var_device, var_dtype))
|
| D | adam.py | 134 def _prepare_local(self, var_device, var_dtype, apply_state): argument
135 super(Adam, self)._prepare_local(var_device, var_dtype, apply_state)
142 lr = (apply_state[(var_device, var_dtype)]['lr_t'] *
144 apply_state[(var_device, var_dtype)].update(
166 def _resource_apply_dense(self, grad, var, apply_state=None): argument
168 coefficients = ((apply_state or {}).get((var_device, var_dtype))
203 def _resource_apply_sparse(self, grad, var, indices, apply_state=None): argument
205 coefficients = ((apply_state or {}).get((var_device, var_dtype))
386 def _prepare_local(self, var_device, var_dtype, apply_state): argument
387 super(NonFusedAdam, self)._prepare_local(var_device, var_dtype, apply_state)
[all …]
|
| D | adamax.py | 113 def _prepare_local(self, var_device, var_dtype, apply_state): argument
114 super(Adamax, self)._prepare_local(var_device, var_dtype, apply_state)
120 lr_t = apply_state[(var_device, var_dtype)]['lr_t']
122 apply_state[(var_device, var_dtype)].update(
133 def _resource_apply_dense(self, grad, var, apply_state=None): argument
135 coefficients = ((apply_state or {}).get((var_device, var_dtype))
152 def _resource_apply_sparse(self, grad, var, indices, apply_state=None): argument
154 coefficients = ((apply_state or {}).get((var_device, var_dtype))
|
| D | adadelta.py | 100 def _prepare_local(self, var_device, var_dtype, apply_state): argument
101 super(Adadelta, self)._prepare_local(var_device, var_dtype, apply_state)
102 apply_state[(var_device, var_dtype)].update(
117 def _resource_apply_dense(self, grad, var, apply_state=None): argument
119 coefficients = ((apply_state or {}).get((var_device, var_dtype))
134 def _resource_apply_sparse(self, grad, var, indices, apply_state=None): argument
136 coefficients = ((apply_state or {}).get((var_device, var_dtype))
|
| D | gradient_descent.py | 127 def _prepare_local(self, var_device, var_dtype, apply_state): argument
128 super(SGD, self)._prepare_local(var_device, var_dtype, apply_state)
129 apply_state[(var_device, var_dtype)]["momentum"] = array_ops.identity(
132 def _resource_apply_dense(self, grad, var, apply_state=None): argument
134 coefficients = ((apply_state or {}).get((var_device, var_dtype))
169 def _resource_apply_sparse(self, grad, var, indices, apply_state=None): argument
172 coefficients = ((apply_state or {}).get((var_device, var_dtype))
|
| D | ftrl.py | 139 def _prepare_local(self, var_device, var_dtype, apply_state): argument
140 super(Ftrl, self)._prepare_local(var_device, var_dtype, apply_state)
141 apply_state[(var_device, var_dtype)].update(
153 def _resource_apply_dense(self, grad, var, apply_state=None): argument
155 coefficients = ((apply_state or {}).get((var_device, var_dtype))
191 def _resource_apply_sparse(self, grad, var, indices, apply_state=None): argument
193 coefficients = ((apply_state or {}).get((var_device, var_dtype))
|
| D | rmsprop.py | 163 def _prepare_local(self, var_device, var_dtype, apply_state): argument
164 super(RMSprop, self)._prepare_local(var_device, var_dtype, apply_state)
167 apply_state[(var_device, var_dtype)].update(
169 neg_lr_t=-apply_state[(var_device, var_dtype)]["lr_t"],
176 def _resource_apply_dense(self, grad, var, apply_state=None): argument
178 coefficients = ((apply_state or {}).get((var_device, var_dtype))
222 def _resource_apply_sparse(self, grad, var, indices, apply_state=None): argument
224 coefficients = ((apply_state or {}).get((var_device, var_dtype))
|
| D | optimizer_v2.py | 663 apply_state = self._prepare(var_list)
670 functools.partial(self._distributed_apply, apply_state=apply_state),
676 def _distributed_apply(self, distribution, grads_and_vars, name, apply_state): argument
690 apply_kwargs["apply_state"] = apply_state
695 apply_kwargs["apply_state"] = apply_state
937 apply_state = {}
939 apply_state[(var_device, var_dtype)] = {}
941 self._prepare_local(var_device, var_dtype, apply_state)
943 return apply_state
945 def _prepare_local(self, var_device, var_dtype, apply_state): argument
[all …]
|
| D | nadam.py | 109 def _prepare_local(self, var_device, var_dtype, apply_state): argument
129 apply_state[(var_device, var_dtype)] = dict(
150 def _resource_apply_dense(self, grad, var, apply_state=None): argument
152 coefficients = ((apply_state or {}).get((var_device, var_dtype))
173 def _resource_apply_sparse(self, grad, var, indices, apply_state=None): argument
175 coefficients = ((apply_state or {}).get((var_device, var_dtype))
|