experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Simplify python loops

Browse Source 
This PR replaces loops over range(len(...)) with enumerate() or zip(). This makes the code (arguably) easier to read and in some cases even slightly faster.
This commit is contained in:
Lukas Geiger 2019-02-28 16:19:23 +00:00
parent f6e917b8c2
commit 47b78e8d6b
15 changed files with 37 additions and 55 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
tensorflow/python/debug/cli/analyzer_cli.py
View File

@ -1438,8 +1438,7 @@ class DebugAnalyzer(object):
hang += DEPTH_TEMPLATE % depth
for i in xrange(len(all_inputs)):
inp = all_inputs[i]
for i, inp in enumerate(all_inputs):
op_type = self._debug_dump.node_op_type(debug_graphs.get_node_name(inp))
if op_type in self._GRAPH_STRUCT_OP_TYPE_BLACKLIST:
continue

8
tensorflow/python/debug/cli/debugger_cli_common.py
View File

@ -25,7 +25,6 @@ import traceback
import numpy as np
import six
from six.moves import xrange # pylint: disable=redefined-builtin
from tensorflow.python import pywrap_tensorflow_internal
from tensorflow.python.platform import gfile
@ -412,8 +411,7 @@ def regex_find(orig_screen_output, regex, font_attr):
raise ValueError("Invalid regular expression: \"%s\"" % regex)
regex_match_lines = []
for i in xrange(len(new_screen_output.lines)):
line = new_screen_output.lines[i]
for i, line in enumerate(new_screen_output.lines):
find_it = re_prog.finditer(line)
match_segs = []
@ -466,11 +464,9 @@ def wrap_rich_text_lines(inp, cols):
out = RichTextLines([])
row_counter = 0 # Counter for new row index
for i in xrange(len(inp.lines)):
for i, line in enumerate(inp.lines):
new_line_indices.append(out.num_lines())
line = inp.lines[i]
if i in inp.annotations:
out.annotations[row_counter] = inp.annotations[i]

4
tensorflow/python/debug/cli/profile_analyzer_cli.py
View File

@ -568,8 +568,8 @@ class ProfileAnalyzer(object):
# Add stat totals.
row_str = ""
for col in range(len(device_total_row)):
row_str += ("{:<%d}" % column_widths[col]).format(device_total_row[col])
for width, row in zip(column_widths, device_total_row):
row_str += ("{:<%d}" % width).format(row)
output.append(RL())
output.append(RL(row_str))
return debugger_cli_common.rich_text_lines_from_rich_line_list(output)

4
tensorflow/python/debug/cli/tensor_format.py
View File

@ -249,8 +249,8 @@ def _annotate_ndarray_lines(
curr_indices = [0] * len(dims)
curr_dim = 0
for i in xrange(len(array_lines)):
line = array_lines[i].strip()
for i, raw_line in enumerate(array_lines):
line = raw_line.strip()
if not line:
# Skip empty lines, which can appear for >= 3D arrays.

3
tensorflow/python/debug/lib/debug_graphs.py
View File

@ -346,8 +346,7 @@ class DebugGraph(object):
for node in self._node_inputs:
inputs = self._node_inputs[node]
for i in xrange(len(inputs)):
inp = inputs[i]
for i, inp in enumerate(inputs):
if is_copy_node(inp):
# Find the input to the Copy node, which should be the original
# input to the node.

4
tensorflow/python/distribute/cross_device_ops.py
View File

@ -933,8 +933,8 @@ class MultiWorkerAllReduce(AllReduceCrossDeviceOps):
aggregated_grads = range_agg_grads
else:
assert len(aggregated_grads) == len(range_agg_grads)
for i in range(len(aggregated_grads)):
aggregated_grads[i] += range_agg_grads[i]
for i, range_agg_grad in enumerate(range_agg_grads):
aggregated_grads[i] += range_agg_grad
assert not remaining_grads
return _ungroup_and_make_mirrored(aggregated_grads, per_replica_values[0],

6
tensorflow/python/eager/backprop.py
View File

@ -514,9 +514,9 @@ def make_vjp(f, params=None, persistent=True):
try:
sources = []
args = [
ops.convert_to_tensor(args[i])
if i in parameter_positions else args[i]
for i in range(len(args))
ops.convert_to_tensor(arg)
if i in parameter_positions else arg
for i, arg in enumerate(args)
]
args = _ensure_unique_tensor_objects(parameter_positions, args)
for i in parameter_positions:

3
tensorflow/python/keras/engine/training_distributed.py
View File

@ -565,8 +565,7 @@ def experimental_tpu_predict_loop(model,
prediction_result = np.concatenate(unconcatenated_outs[0], axis=0)
else:
prediction_result = [
np.concatenate(unconcatenated_outs[i], axis=0)
for i in range(len(unconcatenated_outs))
np.concatenate(out, axis=0) for out in unconcatenated_outs
]
if padding_handler:

9
tensorflow/python/keras/engine/training_utils.py
View File

@ -27,6 +27,7 @@ import time
import numpy as np
import six
from six.moves import zip # pylint: disable=redefined-builtin
from tensorflow.python import tf2
from tensorflow.python.data.experimental.ops import cardinality
@ -1589,9 +1590,7 @@ class ModelInputs(object):
# TODO(karmel): There is a side-effect here where what you get
# with as_list and as_dict depends on whether you have called this
# method first, since it modifies in place.
for i in range(len(self._flattened_inputs)):
k = self._input_names[i]
v = self._flattened_inputs[i]
for i, (k, v) in enumerate(zip(self._input_names, self._flattened_inputs)):
if isinstance(v, (list, float, int)):
v = np.asarray(v)
if v.ndim == 1:
@ -1621,8 +1620,8 @@ class ModelInputs(object):
def as_dict(self):
"""An iterable over a dictionary version of inputs."""
for i in range(len(self._flattened_inputs)):
yield self._input_names[i], self._flattened_inputs[i]
for k, v in zip(self._input_names, self._flattened_inputs):
yield k, v
def as_list(self):
"""Returning the inputs as a list."""

3
tensorflow/python/keras/models.py
View File

@ -116,8 +116,7 @@ def _clone_functional_model(model, input_tensors=None, layer_fn=_clone_layer):
# If tensor comes from an input layer: cache the input layer.
input_tensors = nest.flatten(input_tensors)
input_tensors_ = []
for i in range(len(input_tensors)):
input_tensor = input_tensors[i]
for i, input_tensor in enumerate(input_tensors):
if not K.is_keras_tensor(input_tensor):
original_input_layer = model._input_layers[i]
name = original_input_layer.name

8
tensorflow/python/keras/utils/multi_gpu_utils.py
View File

@ -211,9 +211,7 @@ def multi_gpu_model(model, gpus, cpu_merge=True, cpu_relocation=False):
with ops.device('/cpu:0'):
model = clone_model(model)
all_outputs = []
for i in range(len(model.outputs)):
all_outputs.append([])
all_outputs = [[] for _ in range(len(model.outputs))]
# Place a copy of the model on each GPU,
# each getting a slice of the inputs.
@ -241,8 +239,8 @@ def multi_gpu_model(model, gpus, cpu_merge=True, cpu_relocation=False):
outputs = [outputs]
# Save the outputs for merging back together later.
for o in range(len(outputs)):
all_outputs[o].append(outputs[o])
for o, output in enumerate(outputs):
all_outputs[o].append(output)
# Deduplicate output names to handle Siamese networks.
occurrences = {}

5
tensorflow/python/ops/data_flow_ops.py
View File

@ -2426,8 +2426,7 @@ class RecordInput(object):
with ops.name_scope(self._name):
batch_list = [[] for _ in six.moves.range(self._batches)]
records = array_ops.split(records, self._batch_size, 0)
records = [array_ops.reshape(record, []) for record in records]
for index, protobuf in zip(six.moves.range(len(records)), records):
for index, protobuf in enumerate(records):
batch_index = index % self._batches
batch_list[batch_index].append(protobuf)
batch_list[batch_index].append(array_ops.reshape(protobuf, []))
return batch_list

20
tensorflow/python/ops/gradients_util.py
View File

@ -21,7 +21,7 @@ from __future__ import print_function
import collections
import contextlib
from six.moves import xrange # pylint: disable=redefined-builtin
from six.moves import xrange, zip # pylint: disable=redefined-builtin
from tensorflow.core.framework import attr_value_pb2
from tensorflow.python.eager import backprop
@ -160,9 +160,7 @@ def _DefaultGradYs(grad_ys,
raise ValueError("Passed %d grad_ys for %d ys" % (len(grad_ys), len(ys)))
grad_ys = ops.convert_n_to_tensor_or_indexed_slices(grad_ys, name="grad_y")
new_grad_ys = []
for i in xrange(len(grad_ys)):
grad_y = grad_ys[i]
y = ys[i]
for i, (y, grad_y) in enumerate(zip(ys, grad_ys)):
with _maybe_colocate_with(y.op, gradient_uid, colocate_gradients_with_ops):
if grad_y is None:
if y.dtype.is_complex:
@ -883,23 +881,23 @@ class AggregationMethod(object):
aggregating gradients:
* `ADD_N`: All of the gradient terms are summed as part of one
operation using the "AddN" op (see `tf.add_n`). This
method has the property that all gradients must be ready and
operation using the "AddN" op (see `tf.add_n`). This
method has the property that all gradients must be ready and
buffered separately in memory before any aggregation is performed.
* `DEFAULT`: The system-chosen default aggregation method.
The following aggregation methods are experimental and may not
The following aggregation methods are experimental and may not
be supported in future releases:
* `EXPERIMENTAL_TREE`: Gradient terms are summed in pairs using
using the "AddN" op. This method of summing gradients may reduce
performance, but it can improve memory utilization because the
using the "AddN" op. This method of summing gradients may reduce
performance, but it can improve memory utilization because the
gradients can be released earlier.
* `EXPERIMENTAL_ACCUMULATE_N`: Gradient terms are summed using the
"AccumulateN" op (see `tf.accumulate_n`), which accumulates the
"AccumulateN" op (see `tf.accumulate_n`), which accumulates the
overall sum in a single buffer that is shared across threads.
This method of summing gradients can result in a lower memory footprint
This method of summing gradients can result in a lower memory footprint
and lower latency at the expense of higher CPU/GPU utilization.
For gradients of types that "AccumulateN" does not support, this
summation method falls back on the behavior of `EXPERIMENTAL_TREE`

6
tensorflow/python/ops/variable_scope.py
View File

@ -28,7 +28,7 @@ import traceback
import six
from six import iteritems
from six.moves import xrange # pylint: disable=redefined-builtin
from six.moves import xrange, zip # pylint: disable=redefined-builtin
from tensorflow.python import tf2
from tensorflow.python.eager import context
@ -93,9 +93,7 @@ class _PartitionInfo(object):
"full_shape is of length {}.".format(
len(var_offset), len(full_shape)))
for i in xrange(len(full_shape)):
offset = var_offset[i]
shape = full_shape[i]
for offset, shape in zip(var_offset, full_shape):
if offset < 0 or offset >= shape:
raise ValueError(
"Expected 0 <= offset < shape but found offset={}, shape={} for "

6
tensorflow/python/training/checkpoint_management.py
View File

@ -108,8 +108,7 @@ def generate_checkpoint_state_proto(save_dir,
if not os.path.isabs(save_dir):
if not os.path.isabs(model_checkpoint_path):
model_checkpoint_path = os.path.relpath(model_checkpoint_path, save_dir)
for i in range(len(all_model_checkpoint_paths)):
p = all_model_checkpoint_paths[i]
for i, p in enumerate(all_model_checkpoint_paths):
if not os.path.isabs(p):
all_model_checkpoint_paths[i] = os.path.relpath(p, save_dir)
@ -281,8 +280,7 @@ def get_checkpoint_state(checkpoint_dir, latest_filename=None):
if not os.path.isabs(ckpt.model_checkpoint_path):
ckpt.model_checkpoint_path = os.path.join(checkpoint_dir,
ckpt.model_checkpoint_path)
for i in range(len(ckpt.all_model_checkpoint_paths)):
p = ckpt.all_model_checkpoint_paths[i]
for i, p in enumerate(ckpt.all_model_checkpoint_paths):
if not os.path.isabs(p):
ckpt.all_model_checkpoint_paths[i] = os.path.join(checkpoint_dir, p)
except errors.OpError as e: