def set_param_values(self, flattened_params, **tags):
#import ipdb
#ipdb.set_trace()
debug = tags.pop("debug", False)
# import ipdb
# ipdb.set_trace()
param_values = unflatten_tensors(flattened_params,
self.get_param_shapes(**tags))
ops = []
feed_dict = dict()
for param, dtype, value in zip(self.get_params(**tags),
self.get_param_dtypes(**tags),
param_values):
if param not in self._cached_assign_ops:
assign_placeholder = tf.placeholder(
dtype=param.dtype.base_dtype)
assign_op = tf.assign(param, assign_placeholder)
self._cached_assign_ops[param] = assign_op
self._cached_assign_placeholders[param] = assign_placeholder
ops.append(self._cached_assign_ops[param])
feed_dict[self._cached_assign_placeholders[param]] = value.astype(
dtype)
if debug:
print("setting value of %s" % param.name)
tf.get_default_session().run(ops, feed_dict=feed_dict)
def set_param_values(self, flattened_params, all_params=False, **tags):
debug = tags.pop("debug", False)
# print("debug, all params", all_params) True
# print("debug, param shapes", self.get_param_shapes(all_params, **tags))
#TODO: remove this hacky code
param_values = unflatten_tensors(
flattened_params, self.get_param_shapes(all_params, **tags))
ops = []
feed_dict = dict()
for param, dtype, value in zip(
self.get_params(all_params, **tags),
self.get_param_dtypes(all_params, **tags), param_values):
if param not in self._cached_assign_ops:
assign_placeholder = tf.placeholder(
dtype=param.dtype.base_dtype)
assign_op = tf.assign(param, assign_placeholder)
self._cached_assign_ops[param] = assign_op
self._cached_assign_placeholders[param] = assign_placeholder
ops.append(self._cached_assign_ops[param])
feed_dict[self._cached_assign_placeholders[param]] = value.astype(
dtype)
if debug:
print("setting value of %s" % param.name)
tf.get_default_session().run(ops, feed_dict=feed_dict)
def set_param_values_transfer(self, flattened_params, **tags):
debug = tags.pop("debug", False)
param_values = unflatten_tensors(flattened_params,
self.get_param_shapes(**tags))
for param, dtype, value in zip(self.get_params(**tags),
self.get_param_dtypes(**tags),
param_values):
if param.name != "output_log_std.param":
param.set_value(value.astype(dtype))
def set_param_values(self, flattened_params, **tags):
debug = tags.pop("debug", False)
param_values = unflatten_tensors(flattened_params,
self.get_param_shapes(**tags))
for param, dtype, value in zip(self.get_params(**tags),
self.get_param_dtypes(**tags),
param_values):
param.set_value(value.astype(dtype))
if debug:
print("setting value of %s" % param.name)
def set_param_values(self, flattened_params, **tags):
debug = tags.pop("debug", False)
param_values = unflatten_tensors(
flattened_params, self.get_param_shapes(**tags))
for param, dtype, value in zip(
self.get_params(**tags),
self.get_param_dtypes(**tags),
param_values):
param.set_value(value.astype(dtype))
if debug:
print("setting value of %s" % param.name)
def set_param_values(self, flattened_params, **tags):
debug = tags.pop("debug", False)
param_values = unflatten_tensors(flattened_params,
self.get_param_shapes(**tags))
for param, dtype, value in zip(self.get_params(**tags),
self.get_param_dtypes(**tags),
param_values):
if param.name == "leak_rate":
value = np.minimum(1.0, np.maximum(0.0, value))
param.set_value(value.astype(dtype))
if debug:
print("setting value of %s" % param.name)
def set_param_values(self, flattened_params, **tags):
debug = tags.pop("debug", False)
param_values = unflatten_tensors(flattened_params,
self.get_param_shapes(**tags))
for param, dtype, value in zip(self.get_params(**tags),
self.get_param_dtypes(**tags),
param_values):
if not self.reparam and param.name == "tc":
if np.any(value < self.dt):
print("Tc constraint violated:", self.dt, value)
value = np.maximum(self.dt, value)
param.set_value(value.astype(dtype))
if debug:
print("setting value of %s" % param.name)
def set_param_values(self, flattened_params, **tags):
debug = tags.pop("debug", False)
param_values = unflatten_tensors(
flattened_params, self.get_param_shapes(**tags))
ops = []
feed_dict = dict()
for param, dtype, value in zip(
self.get_params(**tags),
self.get_param_dtypes(**tags),
param_values):
if param not in self._cached_assign_ops:
assign_placeholder = tf.placeholder(dtype=param.dtype.base_dtype)
assign_op = tf.assign(param, assign_placeholder)
self._cached_assign_ops[param] = assign_op
self._cached_assign_placeholders[param] = assign_placeholder
ops.append(self._cached_assign_ops[param])
feed_dict[self._cached_assign_placeholders[param]] = value.astype(dtype)
if debug:
print("setting value of %s" % param.name)
tf.get_default_session().run(ops, feed_dict=feed_dict)
def flat_to_params(self, flattened_params, all_params=False, **tags):
return unflatten_tensors(flattened_params,
self.get_param_shapes(all_params, **tags))
def flat_to_params(self, flattened_params, **tags):
# Not used.
import pdb
pdb.set_trace()
return unflatten_tensors(flattened_params,
self.get_param_shapes(**tags))
def flat_to_params(self, flattened_params, **tags):
return unflatten_tensors(flattened_params, self.get_param_shapes(**tags))
def flat_to_params(self, flattened_params, **tags):
import numpy as np
print(np.shape(flattened_params))
return unflatten_tensors(flattened_params,
self.get_param_shapes(**tags))
def flat_to_params(self, flattened_params, **tags):
if config.TF_NN_SETTRACE:
ipdb.set_trace()
return unflatten_tensors(flattened_params,
self.get_param_shapes(**tags))
