INSTRUCTION
stringlengths 1
46.3k
| RESPONSE
stringlengths 75
80.2k
|
|---|---|
Convert gradients to grayscale. This gives a saliency map.
|
def to_grayscale(cv2im):
"""Convert gradients to grayscale. This gives a saliency map."""
# How strongly does each position activate the output
grayscale_im = np.sum(np.abs(cv2im), axis=0)
# Normalize between min and 99th percentile
im_max = np.percentile(grayscale_im, 99)
im_min = np.min(grayscale_im)
grayscale_im = np.clip((grayscale_im - im_min) / (im_max - im_min), 0, 1)
grayscale_im = np.expand_dims(grayscale_im, axis=0)
return grayscale_im
|
Helper function for checking shape of label and prediction
Parameters
----------
labels : list of `NDArray`
The labels of the data.
preds : list of `NDArray`
Predicted values.
wrap : boolean
If True, wrap labels/preds in a list if they are single NDArray
shape : boolean
If True, check the shape of labels and preds;
Otherwise only check their length.
|
def check_label_shapes(labels, preds, wrap=False, shape=False):
"""Helper function for checking shape of label and prediction
Parameters
----------
labels : list of `NDArray`
The labels of the data.
preds : list of `NDArray`
Predicted values.
wrap : boolean
If True, wrap labels/preds in a list if they are single NDArray
shape : boolean
If True, check the shape of labels and preds;
Otherwise only check their length.
"""
if not shape:
label_shape, pred_shape = len(labels), len(preds)
else:
label_shape, pred_shape = labels.shape, preds.shape
if label_shape != pred_shape:
raise ValueError("Shape of labels {} does not match shape of "
"predictions {}".format(label_shape, pred_shape))
if wrap:
if isinstance(labels, ndarray.ndarray.NDArray):
labels = [labels]
if isinstance(preds, ndarray.ndarray.NDArray):
preds = [preds]
return labels, preds
|
Creates evaluation metric from metric names or instances of EvalMetric
or a custom metric function.
Parameters
----------
metric : str or callable
Specifies the metric to create.
This argument must be one of the below:
- Name of a metric.
- An instance of `EvalMetric`.
- A list, each element of which is a metric or a metric name.
- An evaluation function that computes custom metric for a given batch of
labels and predictions.
*args : list
Additional arguments to metric constructor.
Only used when metric is str.
**kwargs : dict
Additional arguments to metric constructor.
Only used when metric is str
Examples
--------
>>> def custom_metric(label, pred):
... return np.mean(np.abs(label - pred))
...
>>> metric1 = mx.metric.create('acc')
>>> metric2 = mx.metric.create(custom_metric)
>>> metric3 = mx.metric.create([metric1, metric2, 'rmse'])
|
def create(metric, *args, **kwargs):
"""Creates evaluation metric from metric names or instances of EvalMetric
or a custom metric function.
Parameters
----------
metric : str or callable
Specifies the metric to create.
This argument must be one of the below:
- Name of a metric.
- An instance of `EvalMetric`.
- A list, each element of which is a metric or a metric name.
- An evaluation function that computes custom metric for a given batch of
labels and predictions.
*args : list
Additional arguments to metric constructor.
Only used when metric is str.
**kwargs : dict
Additional arguments to metric constructor.
Only used when metric is str
Examples
--------
>>> def custom_metric(label, pred):
... return np.mean(np.abs(label - pred))
...
>>> metric1 = mx.metric.create('acc')
>>> metric2 = mx.metric.create(custom_metric)
>>> metric3 = mx.metric.create([metric1, metric2, 'rmse'])
"""
if callable(metric):
return CustomMetric(metric, *args, **kwargs)
elif isinstance(metric, list):
composite_metric = CompositeEvalMetric()
for child_metric in metric:
composite_metric.add(create(child_metric, *args, **kwargs))
return composite_metric
return _create(metric, *args, **kwargs)
|
Creates a custom evaluation metric that receives its inputs as numpy arrays.
Parameters
----------
numpy_feval : callable(label, pred)
Custom evaluation function that receives labels and predictions for a minibatch
as numpy arrays and returns the corresponding custom metric as a floating point number.
name : str, optional
Name of the custom metric.
allow_extra_outputs : bool, optional
Whether prediction output is allowed to have extra outputs. This is useful in cases
like RNN where states are also part of output which can then be fed back to the RNN
in the next step. By default, extra outputs are not allowed.
Returns
-------
float
Custom metric corresponding to the provided labels and predictions.
Example
-------
>>> def custom_metric(label, pred):
... return np.mean(np.abs(label-pred))
...
>>> metric = mx.metric.np(custom_metric)
|
def np(numpy_feval, name=None, allow_extra_outputs=False):
"""Creates a custom evaluation metric that receives its inputs as numpy arrays.
Parameters
----------
numpy_feval : callable(label, pred)
Custom evaluation function that receives labels and predictions for a minibatch
as numpy arrays and returns the corresponding custom metric as a floating point number.
name : str, optional
Name of the custom metric.
allow_extra_outputs : bool, optional
Whether prediction output is allowed to have extra outputs. This is useful in cases
like RNN where states are also part of output which can then be fed back to the RNN
in the next step. By default, extra outputs are not allowed.
Returns
-------
float
Custom metric corresponding to the provided labels and predictions.
Example
-------
>>> def custom_metric(label, pred):
... return np.mean(np.abs(label-pred))
...
>>> metric = mx.metric.np(custom_metric)
"""
def feval(label, pred):
"""Internal eval function."""
return numpy_feval(label, pred)
feval.__name__ = numpy_feval.__name__
return CustomMetric(feval, name, allow_extra_outputs)
|
Save configurations of metric. Can be recreated
from configs with metric.create(``**config``)
|
def get_config(self):
"""Save configurations of metric. Can be recreated
from configs with metric.create(``**config``)
"""
config = self._kwargs.copy()
config.update({
'metric': self.__class__.__name__,
'name': self.name,
'output_names': self.output_names,
'label_names': self.label_names})
return config
|
Update the internal evaluation with named label and pred
Parameters
----------
labels : OrderedDict of str -> NDArray
name to array mapping for labels.
preds : OrderedDict of str -> NDArray
name to array mapping of predicted outputs.
|
def update_dict(self, label, pred):
"""Update the internal evaluation with named label and pred
Parameters
----------
labels : OrderedDict of str -> NDArray
name to array mapping for labels.
preds : OrderedDict of str -> NDArray
name to array mapping of predicted outputs.
"""
if self.output_names is not None:
pred = [pred[name] for name in self.output_names]
else:
pred = list(pred.values())
if self.label_names is not None:
label = [label[name] for name in self.label_names]
else:
label = list(label.values())
self.update(label, pred)
|
Resets the internal evaluation result to initial state.
|
def reset(self):
"""Resets the internal evaluation result to initial state."""
self.num_inst = 0
self.sum_metric = 0.0
self.global_num_inst = 0
self.global_sum_metric = 0.0
|
Gets the current evaluation result.
Returns
-------
names : list of str
Name of the metrics.
values : list of float
Value of the evaluations.
|
def get(self):
"""Gets the current evaluation result.
Returns
-------
names : list of str
Name of the metrics.
values : list of float
Value of the evaluations.
"""
if self.num_inst == 0:
return (self.name, float('nan'))
else:
return (self.name, self.sum_metric / self.num_inst)
|
Gets the current global evaluation result.
Returns
-------
names : list of str
Name of the metrics.
values : list of float
Value of the evaluations.
|
def get_global(self):
"""Gets the current global evaluation result.
Returns
-------
names : list of str
Name of the metrics.
values : list of float
Value of the evaluations.
"""
if self._has_global_stats:
if self.global_num_inst == 0:
return (self.name, float('nan'))
else:
return (self.name, self.global_sum_metric / self.global_num_inst)
else:
return self.get()
|
Returns zipped name and value pairs.
Returns
-------
list of tuples
A (name, value) tuple list.
|
def get_name_value(self):
"""Returns zipped name and value pairs.
Returns
-------
list of tuples
A (name, value) tuple list.
"""
name, value = self.get()
if not isinstance(name, list):
name = [name]
if not isinstance(value, list):
value = [value]
return list(zip(name, value))
|
Returns zipped name and value pairs for global results.
Returns
-------
list of tuples
A (name, value) tuple list.
|
def get_global_name_value(self):
"""Returns zipped name and value pairs for global results.
Returns
-------
list of tuples
A (name, value) tuple list.
"""
if self._has_global_stats:
name, value = self.get_global()
if not isinstance(name, list):
name = [name]
if not isinstance(value, list):
value = [value]
return list(zip(name, value))
else:
return self.get_name_value()
|
Update various binary classification counts for a single (label, pred)
pair.
Parameters
----------
label : `NDArray`
The labels of the data.
pred : `NDArray`
Predicted values.
|
def update_binary_stats(self, label, pred):
"""
Update various binary classification counts for a single (label, pred)
pair.
Parameters
----------
label : `NDArray`
The labels of the data.
pred : `NDArray`
Predicted values.
"""
pred = pred.asnumpy()
label = label.asnumpy().astype('int32')
pred_label = numpy.argmax(pred, axis=1)
check_label_shapes(label, pred)
if len(numpy.unique(label)) > 2:
raise ValueError("%s currently only supports binary classification."
% self.__class__.__name__)
pred_true = (pred_label == 1)
pred_false = 1 - pred_true
label_true = (label == 1)
label_false = 1 - label_true
true_pos = (pred_true * label_true).sum()
false_pos = (pred_true * label_false).sum()
false_neg = (pred_false * label_true).sum()
true_neg = (pred_false * label_false).sum()
self.true_positives += true_pos
self.global_true_positives += true_pos
self.false_positives += false_pos
self.global_false_positives += false_pos
self.false_negatives += false_neg
self.global_false_negatives += false_neg
self.true_negatives += true_neg
self.global_true_negatives += true_neg
|
Calculate the Matthew's Correlation Coefficent
|
def matthewscc(self, use_global=False):
"""
Calculate the Matthew's Correlation Coefficent
"""
if use_global:
if not self.global_total_examples:
return 0.
true_pos = float(self.global_true_positives)
false_pos = float(self.global_false_positives)
false_neg = float(self.global_false_negatives)
true_neg = float(self.global_true_negatives)
else:
if not self.total_examples:
return 0.
true_pos = float(self.true_positives)
false_pos = float(self.false_positives)
false_neg = float(self.false_negatives)
true_neg = float(self.true_negatives)
terms = [(true_pos + false_pos),
(true_pos + false_neg),
(true_neg + false_pos),
(true_neg + false_neg)]
denom = 1.
for t in filter(lambda t: t != 0., terms):
denom *= t
return ((true_pos * true_neg) - (false_pos * false_neg)) / math.sqrt(denom)
|
Returns a new dataset with each sample transformed by the
transformer function `fn`.
Parameters
----------
fn : callable
A transformer function that takes a sample as input and
returns the transformed sample.
lazy : bool, default True
If False, transforms all samples at once. Otherwise,
transforms each sample on demand. Note that if `fn`
is stochastic, you must set lazy to True or you will
get the same result on all epochs.
Returns
-------
Dataset
The transformed dataset.
|
def transform(self, fn, lazy=True):
"""Returns a new dataset with each sample transformed by the
transformer function `fn`.
Parameters
----------
fn : callable
A transformer function that takes a sample as input and
returns the transformed sample.
lazy : bool, default True
If False, transforms all samples at once. Otherwise,
transforms each sample on demand. Note that if `fn`
is stochastic, you must set lazy to True or you will
get the same result on all epochs.
Returns
-------
Dataset
The transformed dataset.
"""
trans = _LazyTransformDataset(self, fn)
if lazy:
return trans
return SimpleDataset([i for i in trans])
|
Returns a new dataset with the first element of each sample
transformed by the transformer function `fn`.
This is useful, for example, when you only want to transform data
while keeping label as is.
Parameters
----------
fn : callable
A transformer function that takes the first elemtn of a sample
as input and returns the transformed element.
lazy : bool, default True
If False, transforms all samples at once. Otherwise,
transforms each sample on demand. Note that if `fn`
is stochastic, you must set lazy to True or you will
get the same result on all epochs.
Returns
-------
Dataset
The transformed dataset.
|
def transform_first(self, fn, lazy=True):
"""Returns a new dataset with the first element of each sample
transformed by the transformer function `fn`.
This is useful, for example, when you only want to transform data
while keeping label as is.
Parameters
----------
fn : callable
A transformer function that takes the first elemtn of a sample
as input and returns the transformed element.
lazy : bool, default True
If False, transforms all samples at once. Otherwise,
transforms each sample on demand. Note that if `fn`
is stochastic, you must set lazy to True or you will
get the same result on all epochs.
Returns
-------
Dataset
The transformed dataset.
"""
return self.transform(_TransformFirstClosure(fn), lazy)
|
Forward the image through the LSTM network model
Parameters
----------
img_: int of array
Returns
----------
label_list: string of list
|
def forward_ocr(self, img_):
"""Forward the image through the LSTM network model
Parameters
----------
img_: int of array
Returns
----------
label_list: string of list
"""
img_ = cv2.resize(img_, (80, 30))
img_ = img_.transpose(1, 0)
print(img_.shape)
img_ = img_.reshape((1, 80, 30))
print(img_.shape)
# img_ = img_.reshape((80 * 30))
img_ = np.multiply(img_, 1 / 255.0)
self.predictor.forward(data=img_, **self.init_state_dict)
prob = self.predictor.get_output(0)
label_list = []
for p in prob:
print(np.argsort(p))
max_index = np.argsort(p)[::-1][0]
label_list.append(max_index)
return self.__get_string(label_list)
|
Return a caffe_pb2.NetParameter object that defined in a prototxt file
|
def read_prototxt(fname):
"""Return a caffe_pb2.NetParameter object that defined in a prototxt file
"""
proto = caffe_pb2.NetParameter()
with open(fname, 'r') as f:
text_format.Merge(str(f.read()), proto)
return proto
|
Returns layers in a caffe_pb2.NetParameter object
|
def get_layers(proto):
"""Returns layers in a caffe_pb2.NetParameter object
"""
if len(proto.layer):
return proto.layer
elif len(proto.layers):
return proto.layers
else:
raise ValueError('Invalid proto file.')
|
Return a caffe_pb2.NetParameter object that defined in a binary
caffemodel file
|
def read_caffemodel(prototxt_fname, caffemodel_fname):
"""Return a caffe_pb2.NetParameter object that defined in a binary
caffemodel file
"""
if use_caffe:
caffe.set_mode_cpu()
net = caffe.Net(prototxt_fname, caffemodel_fname, caffe.TEST)
layer_names = net._layer_names
layers = net.layers
return (layers, layer_names)
else:
proto = caffe_pb2.NetParameter()
with open(caffemodel_fname, 'rb') as f:
proto.ParseFromString(f.read())
return (get_layers(proto), None)
|
Iterate over all layers
|
def layer_iter(layers, layer_names):
"""Iterate over all layers"""
if use_caffe:
for layer_idx, layer in enumerate(layers):
layer_name = re.sub('[-/]', '_', layer_names[layer_idx])
layer_type = layer.type
layer_blobs = layer.blobs
yield (layer_name, layer_type, layer_blobs)
else:
for layer in layers:
layer_name = re.sub('[-/]', '_', layer.name)
layer_type = layer.type
layer_blobs = layer.blobs
yield (layer_name, layer_type, layer_blobs)
|
Set up the configure of profiler (only accepts keyword arguments).
Parameters
----------
filename : string,
output file for profile data
profile_all : boolean,
all profile types enabled
profile_symbolic : boolean,
whether to profile symbolic operators
profile_imperative : boolean,
whether to profile imperative operators
profile_memory : boolean,
whether to profile memory usage
profile_api : boolean,
whether to profile the C API
contiguous_dump : boolean,
whether to periodically dump profiling data to file
dump_period : float,
seconds between profile data dumps
aggregate_stats : boolean,
whether to maintain aggregate stats in memory for console
dump. Has some negative performance impact.
profile_process : string
whether to profile kvstore `server` or `worker`.
server can only be profiled when kvstore is of type dist.
if this is not passed, defaults to `worker`
|
def set_config(**kwargs):
"""Set up the configure of profiler (only accepts keyword arguments).
Parameters
----------
filename : string,
output file for profile data
profile_all : boolean,
all profile types enabled
profile_symbolic : boolean,
whether to profile symbolic operators
profile_imperative : boolean,
whether to profile imperative operators
profile_memory : boolean,
whether to profile memory usage
profile_api : boolean,
whether to profile the C API
contiguous_dump : boolean,
whether to periodically dump profiling data to file
dump_period : float,
seconds between profile data dumps
aggregate_stats : boolean,
whether to maintain aggregate stats in memory for console
dump. Has some negative performance impact.
profile_process : string
whether to profile kvstore `server` or `worker`.
server can only be profiled when kvstore is of type dist.
if this is not passed, defaults to `worker`
"""
kk = kwargs.keys()
vv = kwargs.values()
check_call(_LIB.MXSetProcessProfilerConfig(len(kwargs),
c_str_array([key for key in kk]),
c_str_array([str(val) for val in vv]),
profiler_kvstore_handle))
|
Set up the configure of profiler (Deprecated).
Parameters
----------
mode : string, optional
Indicates whether to enable the profiler, can
be 'symbolic', or 'all'. Defaults to `symbolic`.
filename : string, optional
The name of output trace file. Defaults to 'profile.json'.
|
def profiler_set_config(mode='symbolic', filename='profile.json'):
"""Set up the configure of profiler (Deprecated).
Parameters
----------
mode : string, optional
Indicates whether to enable the profiler, can
be 'symbolic', or 'all'. Defaults to `symbolic`.
filename : string, optional
The name of output trace file. Defaults to 'profile.json'.
"""
warnings.warn('profiler.profiler_set_config() is deprecated. '
'Please use profiler.set_config() instead')
keys = c_str_array([key for key in ["profile_" + mode, "filename"]])
values = c_str_array([str(val) for val in [True, filename]])
assert len(keys) == len(values)
check_call(_LIB.MXSetProcessProfilerConfig(len(keys), keys, values, profiler_kvstore_handle))
|
Set up the profiler state to 'run' or 'stop'.
Parameters
----------
state : string, optional
Indicates whether to run the profiler, can
be 'stop' or 'run'. Default is `stop`.
profile_process : string
whether to profile kvstore `server` or `worker`.
server can only be profiled when kvstore is of type dist.
if this is not passed, defaults to `worker`
|
def set_state(state='stop', profile_process='worker'):
"""Set up the profiler state to 'run' or 'stop'.
Parameters
----------
state : string, optional
Indicates whether to run the profiler, can
be 'stop' or 'run'. Default is `stop`.
profile_process : string
whether to profile kvstore `server` or `worker`.
server can only be profiled when kvstore is of type dist.
if this is not passed, defaults to `worker`
"""
state2int = {'stop': 0, 'run': 1}
profile_process2int = {'worker': 0, 'server': 1}
check_call(_LIB.MXSetProcessProfilerState(ctypes.c_int(state2int[state]),
profile_process2int[profile_process],
profiler_kvstore_handle))
|
Dump profile and stop profiler. Use this to save profile
in advance in case your program cannot exit normally.
Parameters
----------
finished : boolean
Indicates whether to stop statistic output (dumping) after this dump.
Default is True
profile_process : string
whether to profile kvstore `server` or `worker`.
server can only be profiled when kvstore is of type dist.
if this is not passed, defaults to `worker`
|
def dump(finished=True, profile_process='worker'):
"""Dump profile and stop profiler. Use this to save profile
in advance in case your program cannot exit normally.
Parameters
----------
finished : boolean
Indicates whether to stop statistic output (dumping) after this dump.
Default is True
profile_process : string
whether to profile kvstore `server` or `worker`.
server can only be profiled when kvstore is of type dist.
if this is not passed, defaults to `worker`
"""
fin = 1 if finished is True else 0
profile_process2int = {'worker': 0, 'server': 1}
check_call(_LIB.MXDumpProcessProfile(fin,
profile_process2int[profile_process],
profiler_kvstore_handle))
|
Return a printable string of aggregate profile stats.
Parameters
----------
reset: boolean
Indicates whether to clean aggeregate statistical data collected up to this point
|
def dumps(reset=False):
"""Return a printable string of aggregate profile stats.
Parameters
----------
reset: boolean
Indicates whether to clean aggeregate statistical data collected up to this point
"""
debug_str = ctypes.c_char_p()
do_reset = 1 if reset is True else 0
check_call(_LIB.MXAggregateProfileStatsPrint(ctypes.byref(debug_str), int(do_reset)))
return py_str(debug_str.value)
|
Pause profiling.
Parameters
----------
profile_process : string
whether to profile kvstore `server` or `worker`.
server can only be profiled when kvstore is of type dist.
if this is not passed, defaults to `worker`
|
def pause(profile_process='worker'):
"""Pause profiling.
Parameters
----------
profile_process : string
whether to profile kvstore `server` or `worker`.
server can only be profiled when kvstore is of type dist.
if this is not passed, defaults to `worker`
"""
profile_process2int = {'worker': 0, 'server': 1}
check_call(_LIB.MXProcessProfilePause(int(1),
profile_process2int[profile_process],
profiler_kvstore_handle))
|
Resume paused profiling.
Parameters
----------
profile_process : string
whether to profile kvstore `server` or `worker`.
server can only be profiled when kvstore is of type dist.
if this is not passed, defaults to `worker`
|
def resume(profile_process='worker'):
"""
Resume paused profiling.
Parameters
----------
profile_process : string
whether to profile kvstore `server` or `worker`.
server can only be profiled when kvstore is of type dist.
if this is not passed, defaults to `worker`
"""
profile_process2int = {'worker': 0, 'server': 1}
check_call(_LIB.MXProcessProfilePause(int(0),
profile_process2int[profile_process],
profiler_kvstore_handle))
|
Set counter value.
Parameters
----------
value : int
Value for the counter
|
def set_value(self, value):
"""Set counter value.
Parameters
----------
value : int
Value for the counter
"""
check_call(_LIB.MXProfileSetCounter(self.handle, int(value)))
|
Increment counter value.
Parameters
----------
value_change : int
Amount by which to add to the counter
|
def increment(self, delta=1):
"""Increment counter value.
Parameters
----------
value_change : int
Amount by which to add to the counter
"""
check_call(_LIB.MXProfileAdjustCounter(self.handle, int(delta)))
|
Decrement counter value.
Parameters
----------
value_change : int
Amount by which to subtract from the counter
|
def decrement(self, delta=1):
"""Decrement counter value.
Parameters
----------
value_change : int
Amount by which to subtract from the counter
"""
check_call(_LIB.MXProfileAdjustCounter(self.handle, -int(delta)))
|
Set up the profiler state to record operator.
Parameters
----------
scope : string, optional
Indicates what scope the marker should refer to.
Can be 'global', 'process', thread', task', and 'marker'
Default is `process`.
|
def mark(self, scope='process'):
"""Set up the profiler state to record operator.
Parameters
----------
scope : string, optional
Indicates what scope the marker should refer to.
Can be 'global', 'process', thread', task', and 'marker'
Default is `process`.
"""
check_call(_LIB.MXProfileSetMarker(self.domain.handle, c_str(self.name), c_str(scope)))
|
r"""Get CUDA kernel from compiled module.
Parameters
----------
name : str
String name of the kernel.
signature : str
Function signature for the kernel. For example, if a kernel is
declared as::
extern "C" __global__ void axpy(const float *x, double *y, int alpha)
Then its signature should be::
const float *x, double *y, int alpha
or::
const float *, double *, int
Note that `*` in signature marks an argument as array and
`const` marks an argument as constant (input) array.
Returns
-------
CudaKernel
CUDA kernels that can be launched on GPUs.
|
def get_kernel(self, name, signature):
r"""Get CUDA kernel from compiled module.
Parameters
----------
name : str
String name of the kernel.
signature : str
Function signature for the kernel. For example, if a kernel is
declared as::
extern "C" __global__ void axpy(const float *x, double *y, int alpha)
Then its signature should be::
const float *x, double *y, int alpha
or::
const float *, double *, int
Note that `*` in signature marks an argument as array and
`const` marks an argument as constant (input) array.
Returns
-------
CudaKernel
CUDA kernels that can be launched on GPUs.
"""
hdl = CudaKernelHandle()
is_ndarray = []
is_const = []
dtypes = []
pattern = re.compile(r"""^\s*(const)?\s*([\w_]+)\s*(\*)?\s*([\w_]+)?\s*$""")
args = re.sub(r"\s+", " ", signature).split(",")
for arg in args:
match = pattern.match(arg)
if not match or match.groups()[1] == 'const':
raise ValueError(
'Invalid function prototype "%s". Must be in the '
'form of "(const) type (*) (name)"'%arg)
is_const.append(bool(match.groups()[0]))
dtype = match.groups()[1]
is_ndarray.append(bool(match.groups()[2]))
if dtype not in _DTYPE_CPP_TO_NP:
raise TypeError(
"Unsupported kernel argument type %s. Supported types are: %s."%(
arg, ','.join(_DTYPE_CPP_TO_NP.keys())))
dtypes.append(_DTYPE_NP_TO_MX[_DTYPE_CPP_TO_NP[dtype]])
check_call(_LIB.MXRtcCudaKernelCreate(
self.handle,
c_str(name),
len(dtypes),
c_array_buf(ctypes.c_int, array('i', is_ndarray)),
c_array_buf(ctypes.c_int, array('i', is_const)),
c_array_buf(ctypes.c_int, array('i', dtypes)),
ctypes.byref(hdl)))
return CudaKernel(hdl, name, is_ndarray, dtypes)
|
Launch cuda kernel.
Parameters
----------
args : tuple of NDArray or numbers
List of arguments for kernel. NDArrays are expected for pointer
types (e.g. `float*`, `double*`) while numbers are expected for
non-pointer types (e.g. `int`, `float`).
ctx : Context
The context to launch kernel on. Must be GPU context.
grid_dims : tuple of 3 integers
Grid dimensions for CUDA kernel.
block_dims : tuple of 3 integers
Block dimensions for CUDA kernel.
shared_mem : integer, optional
Size of dynamically allocated shared memory. Defaults to 0.
|
def launch(self, args, ctx, grid_dims, block_dims, shared_mem=0):
"""Launch cuda kernel.
Parameters
----------
args : tuple of NDArray or numbers
List of arguments for kernel. NDArrays are expected for pointer
types (e.g. `float*`, `double*`) while numbers are expected for
non-pointer types (e.g. `int`, `float`).
ctx : Context
The context to launch kernel on. Must be GPU context.
grid_dims : tuple of 3 integers
Grid dimensions for CUDA kernel.
block_dims : tuple of 3 integers
Block dimensions for CUDA kernel.
shared_mem : integer, optional
Size of dynamically allocated shared memory. Defaults to 0.
"""
assert ctx.device_type == 'gpu', "Cuda kernel can only be launched on GPU"
assert len(grid_dims) == 3, "grid_dims must be a tuple of 3 integers"
assert len(block_dims) == 3, "grid_dims must be a tuple of 3 integers"
assert len(args) == len(self._dtypes), \
"CudaKernel(%s) expects %d arguments but got %d"%(
self._name, len(self._dtypes), len(args))
void_args = []
ref_holder = []
for i, (arg, is_nd, dtype) in enumerate(zip(args, self._is_ndarray, self._dtypes)):
if is_nd:
assert isinstance(arg, NDArray), \
"The %d-th argument is expected to be a NDArray but got %s"%(
i, type(arg))
void_args.append(arg.handle)
else:
assert isinstance(arg, numeric_types), \
"The %d-th argument is expected to be a number, but got %s"%(
i, type(arg))
ref_holder.append(np.array(arg, dtype=dtype))
void_args.append(ref_holder[-1].ctypes.data_as(ctypes.c_void_p))
check_call(_LIB.MXRtcCudaKernelCall(
self.handle,
ctx.device_id,
c_array(ctypes.c_void_p, void_args),
mx_uint(grid_dims[0]), mx_uint(grid_dims[1]), mx_uint(grid_dims[2]),
mx_uint(block_dims[0]), mx_uint(block_dims[1]), mx_uint(block_dims[2]),
mx_uint(shared_mem)))
|
Clear the internal statistics to initial state.
|
def reset(self):
"""Clear the internal statistics to initial state."""
if getattr(self, 'num', None) is None:
self.num_inst = 0
self.sum_metric = 0.0
else:
self.num_inst = [0] * self.num
self.sum_metric = [0.0] * self.num
self.records = dict()
self.counts = dict()
|
Update internal records. This function now only update internal buffer,
sum_metric and num_inst are updated in _update() function instead when
get() is called to return results.
Params:
----------
labels: mx.nd.array (n * 6) or (n * 5), difficult column is optional
2-d array of ground-truths, n objects(id-xmin-ymin-xmax-ymax-[difficult])
preds: mx.nd.array (m * 6)
2-d array of detections, m objects(id-score-xmin-ymin-xmax-ymax)
|
def update(self, labels, preds):
"""
Update internal records. This function now only update internal buffer,
sum_metric and num_inst are updated in _update() function instead when
get() is called to return results.
Params:
----------
labels: mx.nd.array (n * 6) or (n * 5), difficult column is optional
2-d array of ground-truths, n objects(id-xmin-ymin-xmax-ymax-[difficult])
preds: mx.nd.array (m * 6)
2-d array of detections, m objects(id-score-xmin-ymin-xmax-ymax)
"""
def iou(x, ys):
"""
Calculate intersection-over-union overlap
Params:
----------
x : numpy.array
single box [xmin, ymin ,xmax, ymax]
ys : numpy.array
multiple box [[xmin, ymin, xmax, ymax], [...], ]
Returns:
-----------
numpy.array
[iou1, iou2, ...], size == ys.shape[0]
"""
ixmin = np.maximum(ys[:, 0], x[0])
iymin = np.maximum(ys[:, 1], x[1])
ixmax = np.minimum(ys[:, 2], x[2])
iymax = np.minimum(ys[:, 3], x[3])
iw = np.maximum(ixmax - ixmin, 0.)
ih = np.maximum(iymax - iymin, 0.)
inters = iw * ih
uni = (x[2] - x[0]) * (x[3] - x[1]) + (ys[:, 2] - ys[:, 0]) * \
(ys[:, 3] - ys[:, 1]) - inters
ious = inters / uni
ious[uni < 1e-12] = 0 # in case bad boxes
return ious
# independant execution for each image
for i in range(labels[0].shape[0]):
# get as numpy arrays
label = labels[0][i].asnumpy()
if np.sum(label[:, 0] >= 0) < 1:
continue
pred = preds[self.pred_idx][i].asnumpy()
# calculate for each class
while (pred.shape[0] > 0):
cid = int(pred[0, 0])
indices = np.where(pred[:, 0].astype(int) == cid)[0]
if cid < 0:
pred = np.delete(pred, indices, axis=0)
continue
dets = pred[indices]
pred = np.delete(pred, indices, axis=0)
# sort by score, desceding
dets = dets[dets[:,1].argsort()[::-1]]
records = np.hstack((dets[:, 1][:, np.newaxis], np.zeros((dets.shape[0], 1))))
# ground-truths
label_indices = np.where(label[:, 0].astype(int) == cid)[0]
gts = label[label_indices, :]
label = np.delete(label, label_indices, axis=0)
if gts.size > 0:
found = [False] * gts.shape[0]
for j in range(dets.shape[0]):
# compute overlaps
ious = iou(dets[j, 2:], gts[:, 1:5])
ovargmax = np.argmax(ious)
ovmax = ious[ovargmax]
if ovmax > self.ovp_thresh:
if (not self.use_difficult and
gts.shape[1] >= 6 and
gts[ovargmax, 5] > 0):
pass
else:
if not found[ovargmax]:
records[j, -1] = 1 # tp
found[ovargmax] = True
else:
# duplicate
records[j, -1] = 2 # fp
else:
records[j, -1] = 2 # fp
else:
# no gt, mark all fp
records[:, -1] = 2
# ground truth count
if (not self.use_difficult and gts.shape[1] >= 6):
gt_count = np.sum(gts[:, 5] < 1)
else:
gt_count = gts.shape[0]
# now we push records to buffer
# first column: score, second column: tp/fp
# 0: not set(matched to difficult or something), 1: tp, 2: fp
records = records[np.where(records[:, -1] > 0)[0], :]
if records.size > 0:
self._insert(cid, records, gt_count)
# add missing class if not present in prediction
while (label.shape[0] > 0):
cid = int(label[0, 0])
label_indices = np.where(label[:, 0].astype(int) == cid)[0]
label = np.delete(label, label_indices, axis=0)
if cid < 0:
continue
gt_count = label_indices.size
self._insert(cid, np.array([[0, 0]]), gt_count)
|
update num_inst and sum_metric
|
def _update(self):
""" update num_inst and sum_metric """
aps = []
for k, v in self.records.items():
recall, prec = self._recall_prec(v, self.counts[k])
ap = self._average_precision(recall, prec)
aps.append(ap)
if self.num is not None and k < (self.num - 1):
self.sum_metric[k] = ap
self.num_inst[k] = 1
if self.num is None:
self.num_inst = 1
self.sum_metric = np.mean(aps)
else:
self.num_inst[-1] = 1
self.sum_metric[-1] = np.mean(aps)
|
get recall and precision from internal records
|
def _recall_prec(self, record, count):
""" get recall and precision from internal records """
record = np.delete(record, np.where(record[:, 1].astype(int) == 0)[0], axis=0)
sorted_records = record[record[:,0].argsort()[::-1]]
tp = np.cumsum(sorted_records[:, 1].astype(int) == 1)
fp = np.cumsum(sorted_records[:, 1].astype(int) == 2)
if count <= 0:
recall = tp * 0.0
else:
recall = tp / float(count)
prec = tp.astype(float) / (tp + fp)
return recall, prec
|
calculate average precision
Params:
----------
rec : numpy.array
cumulated recall
prec : numpy.array
cumulated precision
Returns:
----------
ap as float
|
def _average_precision(self, rec, prec):
"""
calculate average precision
Params:
----------
rec : numpy.array
cumulated recall
prec : numpy.array
cumulated precision
Returns:
----------
ap as float
"""
# append sentinel values at both ends
mrec = np.concatenate(([0.], rec, [1.]))
mpre = np.concatenate(([0.], prec, [0.]))
# compute precision integration ladder
for i in range(mpre.size - 1, 0, -1):
mpre[i - 1] = np.maximum(mpre[i - 1], mpre[i])
# look for recall value changes
i = np.where(mrec[1:] != mrec[:-1])[0]
# sum (\delta recall) * prec
ap = np.sum((mrec[i + 1] - mrec[i]) * mpre[i + 1])
return ap
|
Insert records according to key
|
def _insert(self, key, records, count):
""" Insert records according to key """
if key not in self.records:
assert key not in self.counts
self.records[key] = records
self.counts[key] = count
else:
self.records[key] = np.vstack((self.records[key], records))
assert key in self.counts
self.counts[key] += count
|
calculate average precision, override the default one,
special 11-point metric
Params:
----------
rec : numpy.array
cumulated recall
prec : numpy.array
cumulated precision
Returns:
----------
ap as float
|
def _average_precision(self, rec, prec):
"""
calculate average precision, override the default one,
special 11-point metric
Params:
----------
rec : numpy.array
cumulated recall
prec : numpy.array
cumulated precision
Returns:
----------
ap as float
"""
ap = 0.
for t in np.arange(0., 1.1, 0.1):
if np.sum(rec >= t) == 0:
p = 0
else:
p = np.max(prec[rec >= t])
ap += p / 11.
return ap
|
symbol: the pre-trained network symbol
arg_params: the argument parameters of the pre-trained model
num_classes: the number of classes for the fine-tune datasets
layer_name: the layer name before the last fully-connected layer
|
def get_fine_tune_model(symbol, arg_params, num_classes, layer_name, dtype='float32'):
"""
symbol: the pre-trained network symbol
arg_params: the argument parameters of the pre-trained model
num_classes: the number of classes for the fine-tune datasets
layer_name: the layer name before the last fully-connected layer
"""
all_layers = symbol.get_internals()
net = all_layers[layer_name+'_output']
net = mx.symbol.FullyConnected(data=net, num_hidden=num_classes, name='fc')
if dtype == 'float16':
net = mx.sym.Cast(data=net, dtype=np.float32)
net = mx.symbol.SoftmaxOutput(data=net, name='softmax')
new_args = dict({k:arg_params[k] for k in arg_params if 'fc' not in k})
return (net, new_args)
|
Description : generate list for lip images
|
def _list_images(self, root):
"""
Description : generate list for lip images
"""
self.labels = []
self.items = []
valid_unseen_sub_idx = [1, 2, 20, 22]
skip_sub_idx = [21]
if self._mode == 'train':
sub_idx = ['s' + str(i) for i in range(1, 35) \
if i not in valid_unseen_sub_idx + skip_sub_idx]
elif self._mode == 'valid':
sub_idx = ['s' + str(i) for i in valid_unseen_sub_idx]
folder_path = []
for i in sub_idx:
folder_path.extend(glob.glob(os.path.join(root, i, "*")))
for folder in folder_path:
filename = glob.glob(os.path.join(folder, "*"))
if len(filename) != self._seq_len:
continue
filename.sort()
label = os.path.split(folder)[-1]
self.items.append((filename, label))
|
Description : Align to lip position
|
def align_generation(self, file_nm, padding=75):
"""
Description : Align to lip position
"""
align = Align(self._align_root + '/' + file_nm + '.align')
return nd.array(align.sentence(padding))
|
Switch on/off verbose mode
Parameters
----------
verbose : bool
switch on/off verbose mode
print_func : function
A function that computes statistics of initialized arrays.
Takes an `NDArray` and returns an `str`. Defaults to mean
absolute value str((abs(x)/size(x)).asscalar()).
|
def set_verbosity(self, verbose=False, print_func=None):
"""Switch on/off verbose mode
Parameters
----------
verbose : bool
switch on/off verbose mode
print_func : function
A function that computes statistics of initialized arrays.
Takes an `NDArray` and returns an `str`. Defaults to mean
absolute value str((abs(x)/size(x)).asscalar()).
"""
self._verbose = verbose
if print_func is None:
def asum_stat(x):
"""returns |x|/size(x), async execution."""
return str((ndarray.norm(x)/sqrt(x.size)).asscalar())
print_func = asum_stat
self._print_func = print_func
return self
|
Internal verbose print function
Parameters
----------
desc : InitDesc or str
name of the array
init : str
initializer pattern
arr : NDArray
initialized array
|
def _verbose_print(self, desc, init, arr):
"""Internal verbose print function
Parameters
----------
desc : InitDesc or str
name of the array
init : str
initializer pattern
arr : NDArray
initialized array
"""
if self._verbose and self._print_func:
logging.info('Initialized %s as %s: %s', desc, init, self._print_func(arr))
|
Legacy initialization method.
Parameters
----------
name : str
Name of corresponding NDArray.
arr : NDArray
NDArray to be initialized.
|
def _legacy_init(self, name, arr):
"""Legacy initialization method.
Parameters
----------
name : str
Name of corresponding NDArray.
arr : NDArray
NDArray to be initialized.
"""
warnings.warn(
"\033[91mCalling initializer with init(str, NDArray) has been deprecated." \
"please use init(mx.init.InitDesc(...), NDArray) instead.\033[0m",
DeprecationWarning, stacklevel=3)
if not isinstance(name, string_types):
raise TypeError('name must be string')
if not isinstance(arr, NDArray):
raise TypeError('arr must be NDArray')
if name.startswith('upsampling'):
self._init_bilinear(name, arr)
elif name.startswith('stn_loc') and name.endswith('weight'):
self._init_zero(name, arr)
elif name.startswith('stn_loc') and name.endswith('bias'):
self._init_loc_bias(name, arr)
elif name.endswith('bias'):
self._init_bias(name, arr)
elif name.endswith('gamma'):
self._init_gamma(name, arr)
elif name.endswith('beta'):
self._init_beta(name, arr)
elif name.endswith('weight'):
self._init_weight(name, arr)
elif name.endswith("moving_mean"):
self._init_zero(name, arr)
elif name.endswith("moving_var"):
self._init_one(name, arr)
elif name.endswith("moving_inv_var"):
self._init_zero(name, arr)
elif name.endswith("moving_avg"):
self._init_zero(name, arr)
elif name.endswith('min'):
self._init_zero(name, arr)
elif name.endswith('max'):
self._init_one(name, arr)
else:
self._init_default(name, arr)
|
save imglist to disk
Parameters:
----------
fname : str
saved filename
|
def save_imglist(self, fname=None, root=None, shuffle=False):
"""
save imglist to disk
Parameters:
----------
fname : str
saved filename
"""
def progress_bar(count, total, suffix=''):
import sys
bar_len = 24
filled_len = int(round(bar_len * count / float(total)))
percents = round(100.0 * count / float(total), 1)
bar = '=' * filled_len + '-' * (bar_len - filled_len)
sys.stdout.write('[%s] %s%s ...%s\r' % (bar, percents, '%', suffix))
sys.stdout.flush()
str_list = []
for index in range(self.num_images):
progress_bar(index, self.num_images)
label = self.label_from_index(index)
if label.size < 1:
continue
path = self.image_path_from_index(index)
if root:
path = osp.relpath(path, root)
str_list.append('\t'.join([str(index), str(2), str(label.shape[1])] \
+ ["{0:.4f}".format(x) for x in label.ravel()] + [path,]) + '\n')
if str_list:
if shuffle:
import random
random.shuffle(str_list)
if not fname:
fname = self.name + '.lst'
with open(fname, 'w') as f:
for line in str_list:
f.write(line)
else:
raise RuntimeError("No image in imdb")
|
load class names from text file
Parameters:
----------
filename: str
file stores class names
dirname: str
file directory
|
def _load_class_names(self, filename, dirname):
"""
load class names from text file
Parameters:
----------
filename: str
file stores class names
dirname: str
file directory
"""
full_path = osp.join(dirname, filename)
classes = []
with open(full_path, 'r') as f:
classes = [l.strip() for l in f.readlines()]
return classes
|
download and read data into numpy
|
def read_data(label, image):
"""
download and read data into numpy
"""
base_url = 'http://yann.lecun.com/exdb/mnist/'
with gzip.open(download_file(base_url+label, os.path.join('data',label))) as flbl:
magic, num = struct.unpack(">II", flbl.read(8))
label = np.fromstring(flbl.read(), dtype=np.int8)
with gzip.open(download_file(base_url+image, os.path.join('data',image)), 'rb') as fimg:
magic, num, rows, cols = struct.unpack(">IIII", fimg.read(16))
image = np.fromstring(fimg.read(), dtype=np.uint8).reshape(len(label), rows, cols)
return (label, image)
|
create data iterator with NDArrayIter
|
def get_mnist_iter(args, kv):
"""
create data iterator with NDArrayIter
"""
(train_lbl, train_img) = read_data(
'train-labels-idx1-ubyte.gz', 'train-images-idx3-ubyte.gz')
(val_lbl, val_img) = read_data(
't10k-labels-idx1-ubyte.gz', 't10k-images-idx3-ubyte.gz')
train = mx.io.NDArrayIter(
to4d(train_img), train_lbl, args.batch_size, shuffle=True)
val = mx.io.NDArrayIter(
to4d(val_img), val_lbl, args.batch_size)
return (train, val)
|
Function factory for file extension argparse assertion
Args:
extension (string): the file extension to assert
Returns:
string: the supplied extension, if assertion is successful.
|
def make_file_extension_assertion(extension):
"""Function factory for file extension argparse assertion
Args:
extension (string): the file extension to assert
Returns:
string: the supplied extension, if assertion is successful.
"""
def file_extension_assertion(file_path):
base, ext = os.path.splitext(file_path)
if ext.lower() != extension:
raise argparse.ArgumentTypeError('File must have ' + extension + ' extension')
return file_path
return file_extension_assertion
|
generates the colormap for visualizing the segmentation mask
Args:
num_colors (int): the number of colors to generate in the output palette
Returns:
string: the supplied extension, if assertion is successful.
|
def get_palette(num_colors=256):
"""generates the colormap for visualizing the segmentation mask
Args:
num_colors (int): the number of colors to generate in the output palette
Returns:
string: the supplied extension, if assertion is successful.
"""
pallete = [0]*(num_colors*3)
for j in range(0, num_colors):
lab = j
pallete[j*3+0] = 0
pallete[j*3+1] = 0
pallete[j*3+2] = 0
i = 0
while (lab > 0):
pallete[j*3+0] |= (((lab >> 0) & 1) << (7-i))
pallete[j*3+1] |= (((lab >> 1) & 1) << (7-i))
pallete[j*3+2] |= (((lab >> 2) & 1) << (7-i))
i = i + 1
lab >>= 3
return pallete
|
get the (1, 3, h, w) np.array data for the supplied image
Args:
img_path (string): the input image path
Returns:
np.array: image data in a (1, 3, h, w) shape
|
def get_data(img_path):
"""get the (1, 3, h, w) np.array data for the supplied image
Args:
img_path (string): the input image path
Returns:
np.array: image data in a (1, 3, h, w) shape
"""
mean = np.array([123.68, 116.779, 103.939]) # (R,G,B)
img = Image.open(img_path)
img = np.array(img, dtype=np.float32)
reshaped_mean = mean.reshape(1, 1, 3)
img = img - reshaped_mean
img = np.swapaxes(img, 0, 2)
img = np.swapaxes(img, 1, 2)
img = np.expand_dims(img, axis=0)
return img
|
Module main execution
|
def main():
"""Module main execution"""
# Initialization variables - update to change your model and execution context
model_prefix = "FCN8s_VGG16"
epoch = 19
# By default, MXNet will run on the CPU. Change to ctx = mx.gpu() to run on GPU.
ctx = mx.cpu()
fcnxs, fcnxs_args, fcnxs_auxs = mx.model.load_checkpoint(model_prefix, epoch)
fcnxs_args["data"] = mx.nd.array(get_data(args.input), ctx)
data_shape = fcnxs_args["data"].shape
label_shape = (1, data_shape[2]*data_shape[3])
fcnxs_args["softmax_label"] = mx.nd.empty(label_shape, ctx)
exector = fcnxs.bind(ctx, fcnxs_args, args_grad=None, grad_req="null", aux_states=fcnxs_args)
exector.forward(is_train=False)
output = exector.outputs[0]
out_img = np.uint8(np.squeeze(output.asnumpy().argmax(axis=1)))
out_img = Image.fromarray(out_img)
out_img.putpalette(get_palette())
out_img.save(args.output)
|
check input imdbs, make sure they have same classes
|
def _check_classes(self):
"""
check input imdbs, make sure they have same classes
"""
try:
self.classes = self.imdbs[0].classes
self.num_classes = len(self.classes)
except AttributeError:
# fine, if no classes is provided
pass
if self.num_classes > 0:
for db in self.imdbs:
assert self.classes == db.classes, "Multiple imdb must have same classes"
|
get total number of images, init indices
Parameters
----------
shuffle : bool
whether to shuffle the initial indices
|
def _load_image_set_index(self, shuffle):
"""
get total number of images, init indices
Parameters
----------
shuffle : bool
whether to shuffle the initial indices
"""
self.num_images = 0
for db in self.imdbs:
self.num_images += db.num_images
indices = list(range(self.num_images))
if shuffle:
random.shuffle(indices)
return indices
|
given index, find out sub-db and sub-index
Parameters
----------
index : int
index of a specific image
Returns
----------
a tuple (sub-db, sub-index)
|
def _locate_index(self, index):
"""
given index, find out sub-db and sub-index
Parameters
----------
index : int
index of a specific image
Returns
----------
a tuple (sub-db, sub-index)
"""
assert index >= 0 and index < self.num_images, "index out of range"
pos = self.image_set_index[index]
for k, v in enumerate(self.imdbs):
if pos >= v.num_images:
pos -= v.num_images
else:
return (k, pos)
|
given image index, find out full path
Parameters
----------
index: int
index of a specific image
Returns
----------
full path of this image
|
def image_path_from_index(self, index):
"""
given image index, find out full path
Parameters
----------
index: int
index of a specific image
Returns
----------
full path of this image
"""
assert self.image_set_index is not None, "Dataset not initialized"
pos = self.image_set_index[index]
n_db, n_index = self._locate_index(index)
return self.imdbs[n_db].image_path_from_index(n_index)
|
Callback to checkpoint Module to prefix every epoch.
Parameters
----------
mod : subclass of BaseModule
The module to checkpoint.
prefix : str
The file prefix for this checkpoint.
period : int
How many epochs to wait before checkpointing. Defaults to 1.
save_optimizer_states : bool
Indicates whether or not to save optimizer states for continued training.
Returns
-------
callback : function
The callback function that can be passed as iter_end_callback to fit.
|
def module_checkpoint(mod, prefix, period=1, save_optimizer_states=False):
"""Callback to checkpoint Module to prefix every epoch.
Parameters
----------
mod : subclass of BaseModule
The module to checkpoint.
prefix : str
The file prefix for this checkpoint.
period : int
How many epochs to wait before checkpointing. Defaults to 1.
save_optimizer_states : bool
Indicates whether or not to save optimizer states for continued training.
Returns
-------
callback : function
The callback function that can be passed as iter_end_callback to fit.
"""
period = int(max(1, period))
# pylint: disable=unused-argument
def _callback(iter_no, sym=None, arg=None, aux=None):
"""The checkpoint function."""
if (iter_no + 1) % period == 0:
mod.save_checkpoint(prefix, iter_no + 1, save_optimizer_states)
return _callback
|
A callback that saves a model checkpoint every few epochs.
Each checkpoint is made up of a couple of binary files: a model description file and a
parameters (weights and biases) file. The model description file is named
`prefix`--symbol.json and the parameters file is named `prefix`-`epoch_number`.params
Parameters
----------
prefix : str
Prefix for the checkpoint filenames.
period : int, optional
Interval (number of epochs) between checkpoints. Default `period` is 1.
Returns
-------
callback : function
A callback function that can be passed as `epoch_end_callback` to fit.
Example
-------
>>> module.fit(iterator, num_epoch=n_epoch,
... epoch_end_callback = mx.callback.do_checkpoint("mymodel", 1))
Start training with [cpu(0)]
Epoch[0] Resetting Data Iterator
Epoch[0] Time cost=0.100
Saved checkpoint to "mymodel-0001.params"
Epoch[1] Resetting Data Iterator
Epoch[1] Time cost=0.060
Saved checkpoint to "mymodel-0002.params"
|
def do_checkpoint(prefix, period=1):
"""A callback that saves a model checkpoint every few epochs.
Each checkpoint is made up of a couple of binary files: a model description file and a
parameters (weights and biases) file. The model description file is named
`prefix`--symbol.json and the parameters file is named `prefix`-`epoch_number`.params
Parameters
----------
prefix : str
Prefix for the checkpoint filenames.
period : int, optional
Interval (number of epochs) between checkpoints. Default `period` is 1.
Returns
-------
callback : function
A callback function that can be passed as `epoch_end_callback` to fit.
Example
-------
>>> module.fit(iterator, num_epoch=n_epoch,
... epoch_end_callback = mx.callback.do_checkpoint("mymodel", 1))
Start training with [cpu(0)]
Epoch[0] Resetting Data Iterator
Epoch[0] Time cost=0.100
Saved checkpoint to "mymodel-0001.params"
Epoch[1] Resetting Data Iterator
Epoch[1] Time cost=0.060
Saved checkpoint to "mymodel-0002.params"
"""
period = int(max(1, period))
def _callback(iter_no, sym, arg, aux):
"""The checkpoint function."""
if (iter_no + 1) % period == 0:
save_checkpoint(prefix, iter_no + 1, sym, arg, aux)
return _callback
|
Callback to log the training evaluation result every period.
Parameters
----------
period : int
The number of batch to log the training evaluation metric.
auto_reset : bool
Reset the metric after each log.
Returns
-------
callback : function
The callback function that can be passed as iter_epoch_callback to fit.
|
def log_train_metric(period, auto_reset=False):
"""Callback to log the training evaluation result every period.
Parameters
----------
period : int
The number of batch to log the training evaluation metric.
auto_reset : bool
Reset the metric after each log.
Returns
-------
callback : function
The callback function that can be passed as iter_epoch_callback to fit.
"""
def _callback(param):
"""The checkpoint function."""
if param.nbatch % period == 0 and param.eval_metric is not None:
name_value = param.eval_metric.get_name_value()
for name, value in name_value:
logging.info('Iter[%d] Batch[%d] Train-%s=%f',
param.epoch, param.nbatch, name, value)
if auto_reset:
param.eval_metric.reset_local()
return _callback
|
install callback to executor.
Supports installing to multiple exes.
Parameters
----------
exe : mx.executor.Executor
The Executor (returned by symbol.bind) to install to.
|
def install(self, exe):
"""install callback to executor.
Supports installing to multiple exes.
Parameters
----------
exe : mx.executor.Executor
The Executor (returned by symbol.bind) to install to.
"""
exe.set_monitor_callback(self.stat_helper, self.monitor_all)
self.exes.append(exe)
|
Start collecting stats for current batch.
Call before calling forward.
|
def tic(self):
"""Start collecting stats for current batch.
Call before calling forward."""
if self.step % self.interval == 0:
for exe in self.exes:
for array in exe.arg_arrays:
array.wait_to_read()
for array in exe.aux_arrays:
array.wait_to_read()
self.queue = []
self.activated = True
self.step += 1
|
End collecting for current batch and return results.
Call after computation of current batch.
Returns
-------
res : list of
|
def toc(self):
"""End collecting for current batch and return results.
Call after computation of current batch.
Returns
-------
res : list of """
if not self.activated:
return []
for exe in self.exes:
for array in exe.arg_arrays:
array.wait_to_read()
for array in exe.aux_arrays:
array.wait_to_read()
for exe in self.exes:
for name, array in zip(exe._symbol.list_arguments(), exe.arg_arrays):
if self.re_prog.match(name):
self.queue.append((self.step, name, self.stat_func(array)))
for name, array in zip(exe._symbol.list_auxiliary_states(), exe.aux_arrays):
if self.re_prog.match(name):
self.queue.append((self.step, name, self.stat_func(array)))
self.activated = False
res = []
if self.sort:
self.queue.sort(key=lambda x: x[1])
for n, k, v_list in self.queue:
if isinstance(v_list, NDArray):
v_list = [v_list]
assert isinstance(v_list, list)
s = ''
for v in v_list:
assert isinstance(v, NDArray)
if v.shape == (1,):
s += str(v.asscalar()) + '\t'
else:
s += str(v.asnumpy()) + '\t'
res.append((n, k, s))
self.queue = []
return res
|
End collecting and print results.
|
def toc_print(self):
"""End collecting and print results."""
res = self.toc()
for n, k, v in res:
logging.info('Batch: {:7d} {:30s} {:s}'.format(n, k, v))
|
make a random data iteration plan
|
def make_data_iter_plan(self):
"make a random data iteration plan"
# truncate each bucket into multiple of batch-size
bucket_n_batches = []
for i in range(len(self.data)):
bucket_n_batches.append(np.floor((self.data[i]) / self.batch_size))
self.data[i] = self.data[i][:int(bucket_n_batches[i]*self.batch_size)]
bucket_plan = np.hstack([np.zeros(n, int)+i for i, n in enumerate(bucket_n_batches)])
np.random.shuffle(bucket_plan)
bucket_idx_all = [np.random.permutation(len(x)) for x in self.data]
self.bucket_plan = bucket_plan
self.bucket_idx_all = bucket_idx_all
self.bucket_curr_idx = [0 for x in self.data]
self.data_buffer = []
self.label_buffer = []
for i_bucket in range(len(self.data)):
if not self.model_parallel:
data = np.zeros((self.batch_size, self.buckets[i_bucket]))
label = np.zeros((self.batch_size, self.buckets[i_bucket]))
self.data_buffer.append(data)
self.label_buffer.append(label)
else:
data = np.zeros((self.buckets[i_bucket], self.batch_size))
self.data_buffer.append(data)
if self.model_parallel:
# Transpose data if model parallel
for i in range(len(self.data)):
bucket_data = self.data[i]
self.data[i] = np.transpose(bucket_data)
|
Expand the pending files in the current stage.
Parameters
----------
x: str
The file to expand.
pending : str
The list of pending files to expand.
stage: str
The current stage for file expansion, used for matching the prefix of files.
|
def expand(x, pending, stage):
"""
Expand the pending files in the current stage.
Parameters
----------
x: str
The file to expand.
pending : str
The list of pending files to expand.
stage: str
The current stage for file expansion, used for matching the prefix of files.
"""
if x in history and x not in ['mshadow/mshadow/expr_scalar-inl.h']: # MULTIPLE includes
return
if x in pending:
#print('loop found: {} in {}'.format(x, pending))
return
whtspace = ' ' * expand.treeDepth
expand.fileCount += 1
comment = u"//=====[{:3d}] STAGE:{:>4} {}EXPANDING: {} =====\n\n".format(expand.fileCount, stage, whtspace, x)
out.write(comment.encode('ascii'))
print(comment)
with open(x, 'rb') as x_h:
for line in x_h.readlines():
uline = line.decode('utf-8')
if '#define DMLC_LOG_STACK_TRACE 1' in uline.strip():
# Do not enable stacktrace logging
continue
if uline.find('#include') < 0:
out.write(line)
continue
if uline.strip().find('#include') > 0:
print(uline)
continue
m = re1.search(uline)
if not m:
m = re2.search(uline)
if m:
path = m.groups()[0]
else:
m = re3.search(uline)
if m:
path = 'execinfo.h'
else:
print(uline + ' not found')
continue
h = path.strip('./') if "../3rdparty/" not in path else path
if h.endswith('complex.h') and x.endswith('openblas_config.h'):
source = ''
elif h.startswith('ps/'):
source = '../3rdparty/ps-lite/include/' + h
else:
source = find_source(h, x, stage)
if not source:
if (h not in blacklist and
h not in sysheaders and
'mkl' not in h and
'nnpack' not in h and
'tensorrt' not in h and
not h.endswith('.cuh')): sysheaders.append(h)
else:
expand.treeDepth += 1
expand(source, pending + [x], stage)
expand.treeDepth -= 1
out.write(u"//===== EXPANDED : {} =====\n\n".format(x).encode('ascii'))
history.add(x)
|
Dataset loader with preprocessing.
|
def get_imagenet_iterator(root, batch_size, num_workers, data_shape=224, dtype='float32'):
"""Dataset loader with preprocessing."""
train_dir = os.path.join(root, 'train')
train_transform, val_transform = get_imagenet_transforms(data_shape, dtype)
logging.info("Loading image folder %s, this may take a bit long...", train_dir)
train_dataset = ImageFolderDataset(train_dir, transform=train_transform)
train_data = DataLoader(train_dataset, batch_size, shuffle=True,
last_batch='discard', num_workers=num_workers)
val_dir = os.path.join(root, 'val')
if not os.path.isdir(os.path.expanduser(os.path.join(root, 'val', 'n01440764'))):
user_warning = 'Make sure validation images are stored in one subdir per category, a helper script is available at https://git.io/vNQv1'
raise ValueError(user_warning)
logging.info("Loading image folder %s, this may take a bit long...", val_dir)
val_dataset = ImageFolderDataset(val_dir, transform=val_transform)
val_data = DataLoader(val_dataset, batch_size, last_batch='keep', num_workers=num_workers)
return DataLoaderIter(train_data, dtype), DataLoaderIter(val_data, dtype)
|
Creates an instance of token embedding.
Creates a token embedding instance by loading embedding vectors from an externally hosted
pre-trained token embedding file, such as those of GloVe and FastText. To get all the valid
`embedding_name` and `pretrained_file_name`, use
`mxnet.contrib.text.embedding.get_pretrained_file_names()`.
Parameters
----------
embedding_name : str
The token embedding name (case-insensitive).
Returns
-------
An instance of `mxnet.contrib.text.glossary._TokenEmbedding`:
A token embedding instance that loads embedding vectors from an externally hosted
pre-trained token embedding file.
|
def create(embedding_name, **kwargs):
"""Creates an instance of token embedding.
Creates a token embedding instance by loading embedding vectors from an externally hosted
pre-trained token embedding file, such as those of GloVe and FastText. To get all the valid
`embedding_name` and `pretrained_file_name`, use
`mxnet.contrib.text.embedding.get_pretrained_file_names()`.
Parameters
----------
embedding_name : str
The token embedding name (case-insensitive).
Returns
-------
An instance of `mxnet.contrib.text.glossary._TokenEmbedding`:
A token embedding instance that loads embedding vectors from an externally hosted
pre-trained token embedding file.
"""
create_text_embedding = registry.get_create_func(_TokenEmbedding, 'token embedding')
return create_text_embedding(embedding_name, **kwargs)
|
Get valid token embedding names and their pre-trained file names.
To load token embedding vectors from an externally hosted pre-trained token embedding file,
such as those of GloVe and FastText, one should use
`mxnet.contrib.text.embedding.create(embedding_name, pretrained_file_name)`.
This method returns all the valid names of `pretrained_file_name` for the specified
`embedding_name`. If `embedding_name` is set to None, this method returns all the valid
names of `embedding_name` with their associated `pretrained_file_name`.
Parameters
----------
embedding_name : str or None, default None
The pre-trained token embedding name.
Returns
-------
dict or list:
A list of all the valid pre-trained token embedding file names (`pretrained_file_name`)
for the specified token embedding name (`embedding_name`). If the text embeding name is
set to None, returns a dict mapping each valid token embedding name to a list of valid
pre-trained files (`pretrained_file_name`). They can be plugged into
`mxnet.contrib.text.embedding.create(embedding_name,
pretrained_file_name)`.
|
def get_pretrained_file_names(embedding_name=None):
"""Get valid token embedding names and their pre-trained file names.
To load token embedding vectors from an externally hosted pre-trained token embedding file,
such as those of GloVe and FastText, one should use
`mxnet.contrib.text.embedding.create(embedding_name, pretrained_file_name)`.
This method returns all the valid names of `pretrained_file_name` for the specified
`embedding_name`. If `embedding_name` is set to None, this method returns all the valid
names of `embedding_name` with their associated `pretrained_file_name`.
Parameters
----------
embedding_name : str or None, default None
The pre-trained token embedding name.
Returns
-------
dict or list:
A list of all the valid pre-trained token embedding file names (`pretrained_file_name`)
for the specified token embedding name (`embedding_name`). If the text embeding name is
set to None, returns a dict mapping each valid token embedding name to a list of valid
pre-trained files (`pretrained_file_name`). They can be plugged into
`mxnet.contrib.text.embedding.create(embedding_name,
pretrained_file_name)`.
"""
text_embedding_reg = registry.get_registry(_TokenEmbedding)
if embedding_name is not None:
if embedding_name not in text_embedding_reg:
raise KeyError('Cannot find `embedding_name` %s. Use '
'`get_pretrained_file_names('
'embedding_name=None).keys()` to get all the valid embedding '
'names.' % embedding_name)
return list(text_embedding_reg[embedding_name].pretrained_file_name_sha1.keys())
else:
return {embedding_name: list(embedding_cls.pretrained_file_name_sha1.keys())
for embedding_name, embedding_cls in registry.get_registry(_TokenEmbedding).items()}
|
Load embedding vectors from the pre-trained token embedding file.
For every unknown token, if its representation `self.unknown_token` is encountered in the
pre-trained token embedding file, index 0 of `self.idx_to_vec` maps to the pre-trained token
embedding vector loaded from the file; otherwise, index 0 of `self.idx_to_vec` maps to the
text embedding vector initialized by `init_unknown_vec`.
If a token is encountered multiple times in the pre-trained text embedding file, only the
first-encountered token embedding vector will be loaded and the rest will be skipped.
|
def _load_embedding(self, pretrained_file_path, elem_delim, init_unknown_vec, encoding='utf8'):
"""Load embedding vectors from the pre-trained token embedding file.
For every unknown token, if its representation `self.unknown_token` is encountered in the
pre-trained token embedding file, index 0 of `self.idx_to_vec` maps to the pre-trained token
embedding vector loaded from the file; otherwise, index 0 of `self.idx_to_vec` maps to the
text embedding vector initialized by `init_unknown_vec`.
If a token is encountered multiple times in the pre-trained text embedding file, only the
first-encountered token embedding vector will be loaded and the rest will be skipped.
"""
pretrained_file_path = os.path.expanduser(pretrained_file_path)
if not os.path.isfile(pretrained_file_path):
raise ValueError('`pretrained_file_path` must be a valid path to '
'the pre-trained token embedding file.')
logging.info('Loading pre-trained token embedding vectors from %s', pretrained_file_path)
vec_len = None
all_elems = []
tokens = set()
loaded_unknown_vec = None
line_num = 0
with io.open(pretrained_file_path, 'r', encoding=encoding) as f:
for line in f:
line_num += 1
elems = line.rstrip().split(elem_delim)
assert len(elems) > 1, 'At line %d of the pre-trained text embedding file: the ' \
'data format of the pre-trained token embedding file %s ' \
'is unexpected.' % (line_num, pretrained_file_path)
token, elems = elems[0], [float(i) for i in elems[1:]]
if token == self.unknown_token and loaded_unknown_vec is None:
loaded_unknown_vec = elems
tokens.add(self.unknown_token)
elif token in tokens:
warnings.warn('At line %d of the pre-trained token embedding file: the '
'embedding vector for token %s has been loaded and a duplicate '
'embedding for the same token is seen and skipped.' %
(line_num, token))
elif len(elems) == 1:
warnings.warn('At line %d of the pre-trained text embedding file: token %s '
'with 1-dimensional vector %s is likely a header and is '
'skipped.' % (line_num, token, elems))
else:
if vec_len is None:
vec_len = len(elems)
# Reserve a vector slot for the unknown token at the very beggining because
# the unknown index is 0.
all_elems.extend([0] * vec_len)
else:
assert len(elems) == vec_len, \
'At line %d of the pre-trained token embedding file: the dimension ' \
'of token %s is %d but the dimension of previous tokens is %d. ' \
'Dimensions of all the tokens must be the same.' \
% (line_num, token, len(elems), vec_len)
all_elems.extend(elems)
self._idx_to_token.append(token)
self._token_to_idx[token] = len(self._idx_to_token) - 1
tokens.add(token)
self._vec_len = vec_len
self._idx_to_vec = nd.array(all_elems).reshape((-1, self.vec_len))
if loaded_unknown_vec is None:
self._idx_to_vec[C.UNKNOWN_IDX] = init_unknown_vec(shape=self.vec_len)
else:
self._idx_to_vec[C.UNKNOWN_IDX] = nd.array(loaded_unknown_vec)
|
Sets the mapping between token indices and token embedding vectors.
Parameters
----------
token_embeddings : instance or list `mxnet.contrib.text.embedding._TokenEmbedding`
One or multiple pre-trained token embeddings to load. If it is a list of multiple
embeddings, these embedding vectors will be concatenated for each token.
vocab_len : int
Length of vocabulary whose tokens are indexed in the token embedding.
vocab_idx_to_token: list of str
A list of indexed tokens in the vocabulary. These tokens are indexed in the token
embedding.
|
def _set_idx_to_vec_by_embeddings(self, token_embeddings, vocab_len, vocab_idx_to_token):
"""Sets the mapping between token indices and token embedding vectors.
Parameters
----------
token_embeddings : instance or list `mxnet.contrib.text.embedding._TokenEmbedding`
One or multiple pre-trained token embeddings to load. If it is a list of multiple
embeddings, these embedding vectors will be concatenated for each token.
vocab_len : int
Length of vocabulary whose tokens are indexed in the token embedding.
vocab_idx_to_token: list of str
A list of indexed tokens in the vocabulary. These tokens are indexed in the token
embedding.
"""
new_vec_len = sum(embed.vec_len for embed in token_embeddings)
new_idx_to_vec = nd.zeros(shape=(vocab_len, new_vec_len))
col_start = 0
# Concatenate all the embedding vectors in token_embeddings.
for embed in token_embeddings:
col_end = col_start + embed.vec_len
# Cancatenate vectors of the unknown token.
new_idx_to_vec[0, col_start:col_end] = embed.idx_to_vec[0]
new_idx_to_vec[1:, col_start:col_end] = embed.get_vecs_by_tokens(vocab_idx_to_token[1:])
col_start = col_end
self._vec_len = new_vec_len
self._idx_to_vec = new_idx_to_vec
|
Look up embedding vectors of tokens.
Parameters
----------
tokens : str or list of strs
A token or a list of tokens.
lower_case_backup : bool, default False
If False, each token in the original case will be looked up; if True, each token in the
original case will be looked up first, if not found in the keys of the property
`token_to_idx`, the token in the lower case will be looked up.
Returns
-------
mxnet.ndarray.NDArray:
The embedding vector(s) of the token(s). According to numpy conventions, if `tokens` is
a string, returns a 1-D NDArray of shape `self.vec_len`; if `tokens` is a list of
strings, returns a 2-D NDArray of shape=(len(tokens), self.vec_len).
|
def get_vecs_by_tokens(self, tokens, lower_case_backup=False):
"""Look up embedding vectors of tokens.
Parameters
----------
tokens : str or list of strs
A token or a list of tokens.
lower_case_backup : bool, default False
If False, each token in the original case will be looked up; if True, each token in the
original case will be looked up first, if not found in the keys of the property
`token_to_idx`, the token in the lower case will be looked up.
Returns
-------
mxnet.ndarray.NDArray:
The embedding vector(s) of the token(s). According to numpy conventions, if `tokens` is
a string, returns a 1-D NDArray of shape `self.vec_len`; if `tokens` is a list of
strings, returns a 2-D NDArray of shape=(len(tokens), self.vec_len).
"""
to_reduce = False
if not isinstance(tokens, list):
tokens = [tokens]
to_reduce = True
if not lower_case_backup:
indices = [self.token_to_idx.get(token, C.UNKNOWN_IDX) for token in tokens]
else:
indices = [self.token_to_idx[token] if token in self.token_to_idx
else self.token_to_idx.get(token.lower(), C.UNKNOWN_IDX)
for token in tokens]
vecs = nd.Embedding(nd.array(indices), self.idx_to_vec, self.idx_to_vec.shape[0],
self.idx_to_vec.shape[1])
return vecs[0] if to_reduce else vecs
|
Updates embedding vectors for tokens.
Parameters
----------
tokens : str or a list of strs
A token or a list of tokens whose embedding vector are to be updated.
new_vectors : mxnet.ndarray.NDArray
An NDArray to be assigned to the embedding vectors of `tokens`. Its length must be equal
to the number of `tokens` and its width must be equal to the dimension of embeddings of
the glossary. If `tokens` is a singleton, it must be 1-D or 2-D. If `tokens` is a list
of multiple strings, it must be 2-D.
|
def update_token_vectors(self, tokens, new_vectors):
"""Updates embedding vectors for tokens.
Parameters
----------
tokens : str or a list of strs
A token or a list of tokens whose embedding vector are to be updated.
new_vectors : mxnet.ndarray.NDArray
An NDArray to be assigned to the embedding vectors of `tokens`. Its length must be equal
to the number of `tokens` and its width must be equal to the dimension of embeddings of
the glossary. If `tokens` is a singleton, it must be 1-D or 2-D. If `tokens` is a list
of multiple strings, it must be 2-D.
"""
assert self.idx_to_vec is not None, 'The property `idx_to_vec` has not been properly set.'
if not isinstance(tokens, list) or len(tokens) == 1:
assert isinstance(new_vectors, nd.NDArray) and len(new_vectors.shape) in [1, 2], \
'`new_vectors` must be a 1-D or 2-D NDArray if `tokens` is a singleton.'
if not isinstance(tokens, list):
tokens = [tokens]
if len(new_vectors.shape) == 1:
new_vectors = new_vectors.expand_dims(0)
else:
assert isinstance(new_vectors, nd.NDArray) and len(new_vectors.shape) == 2, \
'`new_vectors` must be a 2-D NDArray if `tokens` is a list of multiple strings.'
assert new_vectors.shape == (len(tokens), self.vec_len), \
'The length of new_vectors must be equal to the number of tokens and the width of' \
'new_vectors must be equal to the dimension of embeddings of the glossary.'
indices = []
for token in tokens:
if token in self.token_to_idx:
indices.append(self.token_to_idx[token])
else:
raise ValueError('Token %s is unknown. To update the embedding vector for an '
'unknown token, please specify it explicitly as the '
'`unknown_token` %s in `tokens`. This is to avoid unintended '
'updates.' % (token, self.idx_to_token[C.UNKNOWN_IDX]))
self._idx_to_vec[nd.array(indices)] = new_vectors
|
Checks if a pre-trained token embedding file name is valid.
Parameters
----------
pretrained_file_name : str
The pre-trained token embedding file.
|
def _check_pretrained_file_names(cls, pretrained_file_name):
"""Checks if a pre-trained token embedding file name is valid.
Parameters
----------
pretrained_file_name : str
The pre-trained token embedding file.
"""
embedding_name = cls.__name__.lower()
if pretrained_file_name not in cls.pretrained_file_name_sha1:
raise KeyError('Cannot find pretrained file %s for token embedding %s. Valid '
'pretrained files for embedding %s: %s' %
(pretrained_file_name, embedding_name, embedding_name,
', '.join(cls.pretrained_file_name_sha1.keys())))
|
Calculate gradient
|
def calc_grad(exe, exe_grads, params, X, Y, label_name=None, outgrad_f=None):
"""Calculate gradient"""
exe.copy_params_from(params)
exe.arg_dict['data'][:] = X
if outgrad_f is None:
exe.arg_dict[label_name][:] = Y
exe.forward(is_train=True)
exe.backward()
else:
exe.forward(is_train=True)
exe.backward(outgrad_f(exe.outpus, Y))
for k, v in exe_grads.items():
v.wait_to_read()
|
Generate the implementation of step HMC
|
def step_HMC(exe, exe_params, exe_grads, label_key, noise_precision, prior_precision, L=10, eps=1E-6):
"""Generate the implementation of step HMC"""
init_params = {k: v.copyto(v.context) for k, v in exe_params.items()}
end_params = {k: v.copyto(v.context) for k, v in exe_params.items()}
init_momentums = {k: mx.random.normal(0, 1, v.shape) for k, v in init_params.items()}
end_momentums = {k: v.copyto(v.context) for k, v in init_momentums.items()}
init_potential = calc_potential(exe, init_params, label_key, noise_precision, prior_precision)
# 0. Calculate Initial Energy and Kinetic
init_kinetic = sum([nd.sum(nd.square(momentum)) / 2.0
for momentum in init_momentums.values()]).asscalar()
# 1. Make a half step for momentum at the beginning
exe.copy_params_from(end_params)
exe.forward(is_train=True)
exe.backward()
for k, v in exe_grads.items():
v.wait_to_read()
for k, momentum in end_momentums.items():
momentum[:] = momentum - (eps / 2) * exe_grads[k]
# 2. Alternate full steps for position and momentum
for i in range(L):
# 2.1 Full step for position
for k, param in exe_params.items():
param[:] = param + eps * end_momentums[k]
# 2.2 Full step for the momentum, except at the end of trajectory we perform a half step
exe.forward(is_train=True)
exe.backward()
for v in exe_grads.values():
v.wait_to_read()
if i != L - 1:
for k, momentum in end_momentums.items():
momentum[:] = momentum - eps * exe_grads[k]
else:
for k, momentum in end_momentums.items():
# We should reverse the sign of the momentum at the end
momentum[:] = -(momentum - eps / 2.0 * exe_grads[k])
copy_param(exe, end_params)
# 3. Calculate acceptance ratio and accept/reject the move
end_potential = calc_potential(exe, end_params, label_key, noise_precision, prior_precision)
end_kinetic = sum([nd.sum(nd.square(momentum)) / 2.0
for momentum in end_momentums.values()]).asscalar()
# print init_potential, init_kinetic, end_potential, end_kinetic
r = numpy.random.rand(1)
if r < numpy.exp(-(end_potential + end_kinetic) + (init_potential + init_kinetic)):
exe.copy_params_from(end_params)
return end_params, 1
else:
exe.copy_params_from(init_params)
return init_params, 0
|
Generate the implementation of HMC
|
def HMC(sym, data_inputs, X, Y, X_test, Y_test, sample_num,
initializer=None, noise_precision=1 / 9.0, prior_precision=0.1,
learning_rate=1E-6, L=10, dev=mx.gpu()):
"""Generate the implementation of HMC"""
label_key = list(set(data_inputs.keys()) - set(['data']))[0]
exe, exe_params, exe_grads, _ = get_executor(sym, dev, data_inputs, initializer)
exe.arg_dict['data'][:] = X
exe.arg_dict[label_key][:] = Y
sample_pool = []
accept_num = 0
start = time.time()
for i in range(sample_num):
sample_params, is_accept = step_HMC(exe, exe_params, exe_grads, label_key, noise_precision,
prior_precision, L, learning_rate)
accept_num += is_accept
if (i + 1) % 10 == 0:
sample_pool.append(sample_params)
if (i + 1) % 100000 == 0:
end = time.time()
print("Current Iter Num: %d" % (i + 1), "Time Spent: %f" % (end - start), "MSE:",
sample_test_regression(exe, X=X_test, Y=Y_test, sample_pool=sample_pool,
minibatch_size=Y.shape[0],
save_path='regression_HMC.txt'))
start = time.time()
exe.copy_params_from(sample_params)
print('accept ratio', accept_num / float(sample_num))
return sample_pool
|
Generate the implementation of SGD
|
def SGD(sym, data_inputs, X, Y, X_test, Y_test, total_iter_num,
lr=None,
lr_scheduler=None, prior_precision=1,
out_grad_f=None,
initializer=None,
minibatch_size=100, dev=mx.gpu()):
"""Generate the implementation of SGD"""
if out_grad_f is None:
label_key = list(set(data_inputs.keys()) - set(['data']))[0]
exe, params, params_grad, _ = get_executor(sym, dev, data_inputs, initializer)
optimizer = mx.optimizer.create('sgd', learning_rate=lr,
rescale_grad=X.shape[0] / minibatch_size,
lr_scheduler=lr_scheduler,
wd=prior_precision)
updater = mx.optimizer.get_updater(optimizer)
start = time.time()
for i in range(total_iter_num):
indices = numpy.random.randint(X.shape[0], size=minibatch_size)
X_batch = X[indices]
Y_batch = Y[indices]
exe.arg_dict['data'][:] = X_batch
if out_grad_f is None:
exe.arg_dict[label_key][:] = Y_batch
exe.forward(is_train=True)
exe.backward()
else:
exe.forward(is_train=True)
exe.backward(out_grad_f(exe.outputs, nd.array(Y_batch, ctx=dev)))
for k in params:
updater(k, params_grad[k], params[k])
if (i + 1) % 500 == 0:
end = time.time()
print("Current Iter Num: %d" % (i + 1), "Time Spent: %f" % (end - start))
sample_test_acc(exe, X=X_test, Y=Y_test, label_num=10, minibatch_size=100)
start = time.time()
return exe, params, params_grad
|
Generate the implementation of SGLD
|
def SGLD(sym, X, Y, X_test, Y_test, total_iter_num,
data_inputs=None,
learning_rate=None,
lr_scheduler=None, prior_precision=1,
out_grad_f=None,
initializer=None,
minibatch_size=100, thin_interval=100, burn_in_iter_num=1000, task='classification',
dev=mx.gpu()):
"""Generate the implementation of SGLD"""
if out_grad_f is None:
label_key = list(set(data_inputs.keys()) - set(['data']))[0]
exe, params, params_grad, _ = get_executor(sym, dev, data_inputs, initializer)
optimizer = mx.optimizer.create('sgld', learning_rate=learning_rate,
rescale_grad=X.shape[0] / minibatch_size,
lr_scheduler=lr_scheduler,
wd=prior_precision)
updater = mx.optimizer.get_updater(optimizer)
sample_pool = []
start = time.time()
for i in range(total_iter_num):
indices = numpy.random.randint(X.shape[0], size=minibatch_size)
X_batch = X[indices]
Y_batch = Y[indices]
exe.arg_dict['data'][:] = X_batch
if out_grad_f is None:
exe.arg_dict[label_key][:] = Y_batch
exe.forward(is_train=True)
exe.backward()
else:
exe.forward(is_train=True)
exe.backward(out_grad_f(exe.outputs, nd.array(Y_batch, ctx=dev)))
for k in params:
updater(k, params_grad[k], params[k])
if i < burn_in_iter_num:
continue
else:
if (i - burn_in_iter_num) % thin_interval == 0:
if optimizer.lr_scheduler is not None:
lr = optimizer.lr_scheduler(optimizer.num_update)
else:
lr = learning_rate
sample_pool.append([lr, copy_param(exe)])
if (i + 1) % 100000 == 0:
end = time.time()
if task == 'classification':
print("Current Iter Num: %d" % (i + 1), "Time Spent: %f" % (end - start))
test_correct, test_total, test_acc = \
sample_test_acc(exe, sample_pool=sample_pool, X=X_test, Y=Y_test, label_num=10,
minibatch_size=minibatch_size)
print("Test %d/%d=%f" % (test_correct, test_total, test_acc))
else:
print("Current Iter Num: %d" % (i + 1), "Time Spent: %f" % (end - start), "MSE:",
sample_test_regression(exe=exe, sample_pool=sample_pool,
X=X_test,
Y=Y_test, minibatch_size=minibatch_size,
save_path='regression_SGLD.txt'))
start = time.time()
return exe, sample_pool
|
Generate the implementation of DistilledSGLD
|
def DistilledSGLD(teacher_sym, student_sym,
teacher_data_inputs, student_data_inputs,
X, Y, X_test, Y_test, total_iter_num,
teacher_learning_rate, student_learning_rate,
teacher_lr_scheduler=None, student_lr_scheduler=None,
student_optimizing_algorithm='sgd',
teacher_grad_f=None, student_grad_f=None,
teacher_prior_precision=1, student_prior_precision=0.001,
perturb_deviation=0.001,
student_initializer=None,
teacher_initializer=None,
minibatch_size=100,
task='classification',
dev=mx.gpu()):
"""Generate the implementation of DistilledSGLD"""
teacher_exe, teacher_params, teacher_params_grad, _ = \
get_executor(teacher_sym, dev, teacher_data_inputs, teacher_initializer)
student_exe, student_params, student_params_grad, _ = \
get_executor(student_sym, dev, student_data_inputs, student_initializer)
if teacher_grad_f is None:
teacher_label_key = list(set(teacher_data_inputs.keys()) - set(['data']))[0]
if student_grad_f is None:
student_label_key = list(set(student_data_inputs.keys()) - set(['data']))[0]
teacher_optimizer = mx.optimizer.create('sgld',
learning_rate=teacher_learning_rate,
rescale_grad=X.shape[0] / float(minibatch_size),
lr_scheduler=teacher_lr_scheduler,
wd=teacher_prior_precision)
student_optimizer = mx.optimizer.create(student_optimizing_algorithm,
learning_rate=student_learning_rate,
rescale_grad=1.0 / float(minibatch_size),
lr_scheduler=student_lr_scheduler,
wd=student_prior_precision)
teacher_updater = mx.optimizer.get_updater(teacher_optimizer)
student_updater = mx.optimizer.get_updater(student_optimizer)
start = time.time()
for i in range(total_iter_num):
# 1.1 Draw random minibatch
indices = numpy.random.randint(X.shape[0], size=minibatch_size)
X_batch = X[indices]
Y_batch = Y[indices]
# 1.2 Update teacher
teacher_exe.arg_dict['data'][:] = X_batch
if teacher_grad_f is None:
teacher_exe.arg_dict[teacher_label_key][:] = Y_batch
teacher_exe.forward(is_train=True)
teacher_exe.backward()
else:
teacher_exe.forward(is_train=True)
teacher_exe.backward(
teacher_grad_f(teacher_exe.outputs, nd.array(Y_batch, ctx=dev)))
for k in teacher_params:
teacher_updater(k, teacher_params_grad[k], teacher_params[k])
# 2.1 Draw random minibatch and do random perturbation
if task == 'classification':
indices = numpy.random.randint(X.shape[0], size=minibatch_size)
X_student_batch = X[indices] + numpy.random.normal(0,
perturb_deviation,
X_batch.shape).astype('float32')
else:
X_student_batch = mx.random.uniform(-6, 6, X_batch.shape, mx.cpu())
# 2.2 Get teacher predictions
teacher_exe.arg_dict['data'][:] = X_student_batch
teacher_exe.forward(is_train=False)
teacher_pred = teacher_exe.outputs[0]
teacher_pred.wait_to_read()
# 2.3 Update student
student_exe.arg_dict['data'][:] = X_student_batch
if student_grad_f is None:
student_exe.arg_dict[student_label_key][:] = teacher_pred
student_exe.forward(is_train=True)
student_exe.backward()
else:
student_exe.forward(is_train=True)
student_exe.backward(student_grad_f(student_exe.outputs, teacher_pred))
for k in student_params:
student_updater(k, student_params_grad[k], student_params[k])
if (i + 1) % 2000 == 0:
end = time.time()
if task == 'classification':
print("Current Iter Num: %d" % (i + 1), "Time Spent: %f" % (end - start))
test_correct, test_total, test_acc = \
sample_test_acc(student_exe, X=X_test, Y=Y_test, label_num=10,
minibatch_size=minibatch_size)
train_correct, train_total, train_acc = \
sample_test_acc(student_exe, X=X, Y=Y, label_num=10,
minibatch_size=minibatch_size)
teacher_test_correct, teacher_test_total, teacher_test_acc = \
sample_test_acc(teacher_exe, X=X_test, Y=Y_test, label_num=10,
minibatch_size=minibatch_size)
teacher_train_correct, teacher_train_total, teacher_train_acc = \
sample_test_acc(teacher_exe, X=X, Y=Y, label_num=10,
minibatch_size=minibatch_size)
print("Student: Test ACC %d/%d=%f, Train ACC %d/%d=%f" % (test_correct, test_total,
test_acc, train_correct,
train_total, train_acc))
print("Teacher: Test ACC %d/%d=%f, Train ACC %d/%d=%f" \
% (teacher_test_correct, teacher_test_total, teacher_test_acc,
teacher_train_correct, teacher_train_total, teacher_train_acc))
else:
print("Current Iter Num: %d" % (i + 1), "Time Spent: %f" % (end - start), "MSE:",
sample_test_regression(exe=student_exe, X=X_test, Y=Y_test,
minibatch_size=minibatch_size,
save_path='regression_DSGLD.txt'))
start = time.time()
return student_exe, student_params, student_params_grad
|
Get a list of architectures given our dockerfiles
|
def get_platforms(path: str = get_dockerfiles_path()) -> List[str]:
"""Get a list of architectures given our dockerfiles"""
dockerfiles = glob.glob(os.path.join(path, "Dockerfile.*"))
dockerfiles = list(filter(lambda x: x[-1] != '~', dockerfiles))
files = list(map(lambda x: re.sub(r"Dockerfile.(.*)", r"\1", x), dockerfiles))
platforms = list(map(lambda x: os.path.split(x)[1], sorted(files)))
return platforms
|
:return: docker tag to be used for the container
|
def get_docker_tag(platform: str, registry: str) -> str:
""":return: docker tag to be used for the container"""
platform = platform if any(x in platform for x in ['build.', 'publish.']) else 'build.{}'.format(platform)
if not registry:
registry = "mxnet_local"
return "{0}/{1}".format(registry, platform)
|
Build a container for the given platform
:param platform: Platform
:param docker_binary: docker binary to use (docker/nvidia-docker)
:param registry: Dockerhub registry name
:param num_retries: Number of retries to build the docker image
:param no_cache: pass no-cache to docker to rebuild the images
:return: Id of the top level image
|
def build_docker(platform: str, docker_binary: str, registry: str, num_retries: int, no_cache: bool) -> str:
"""
Build a container for the given platform
:param platform: Platform
:param docker_binary: docker binary to use (docker/nvidia-docker)
:param registry: Dockerhub registry name
:param num_retries: Number of retries to build the docker image
:param no_cache: pass no-cache to docker to rebuild the images
:return: Id of the top level image
"""
tag = get_docker_tag(platform=platform, registry=registry)
logging.info("Building docker container tagged '%s' with %s", tag, docker_binary)
#
# We add a user with the same group as the executing non-root user so files created in the
# container match permissions of the local user. Same for the group.
#
# These variables are used in the docker files to create user and group with these ids.
# see: docker/install/ubuntu_adduser.sh
#
# cache-from is needed so we use the cached images tagged from the remote via
# docker pull see: docker_cache.load_docker_cache
#
# This also prevents using local layers for caching: https://github.com/moby/moby/issues/33002
# So to use local caching, we should omit the cache-from by using --no-dockerhub-cache argument to this
# script.
#
# This doesn't work with multi head docker files.
#
cmd = [docker_binary, "build",
"-f", get_dockerfile(platform),
"--build-arg", "USER_ID={}".format(os.getuid()),
"--build-arg", "GROUP_ID={}".format(os.getgid())]
if no_cache:
cmd.append("--no-cache")
elif registry:
cmd.extend(["--cache-from", tag])
cmd.extend(["-t", tag, get_dockerfiles_path()])
@retry(subprocess.CalledProcessError, tries=num_retries)
def run_cmd():
logging.info("Running command: '%s'", ' '.join(cmd))
check_call(cmd)
run_cmd()
# Get image id by reading the tag. It's guaranteed (except race condition) that the tag exists. Otherwise, the
# check_call would have failed
image_id = _get_local_image_id(docker_binary=docker_binary, docker_tag=tag)
if not image_id:
raise FileNotFoundError('Unable to find docker image id matching with {}'.format(tag))
return image_id
|
Get the image id of the local docker layer with the passed tag
:param docker_tag: docker tag
:return: Image id as string or None if tag does not exist
|
def _get_local_image_id(docker_binary, docker_tag):
"""
Get the image id of the local docker layer with the passed tag
:param docker_tag: docker tag
:return: Image id as string or None if tag does not exist
"""
cmd = [docker_binary, "images", "-q", docker_tag]
image_id_b = check_output(cmd)
image_id = image_id_b.decode('utf-8').strip()
if not image_id:
raise RuntimeError('Unable to find docker image id matching with tag {}'.format(docker_tag))
return image_id
|
:return: ccache directory for the current platform
|
def default_ccache_dir() -> str:
""":return: ccache directory for the current platform"""
# Share ccache across containers
if 'CCACHE_DIR' in os.environ:
ccache_dir = os.path.realpath(os.environ['CCACHE_DIR'])
try:
os.makedirs(ccache_dir, exist_ok=True)
return ccache_dir
except PermissionError:
logging.info('Unable to make dirs at %s, falling back to local temp dir', ccache_dir)
# In osx tmpdir is not mountable by default
import platform
if platform.system() == 'Darwin':
ccache_dir = "/tmp/_mxnet_ccache"
os.makedirs(ccache_dir, exist_ok=True)
return ccache_dir
return os.path.join(os.path.expanduser("~"), ".ccache")
|
Run command in a container
|
def container_run(platform: str,
nvidia_runtime: bool,
docker_registry: str,
shared_memory_size: str,
local_ccache_dir: str,
command: List[str],
cleanup: Cleanup,
environment: Dict[str, str],
dry_run: bool = False) -> int:
"""Run command in a container"""
container_wait_s = 600
#
# Environment setup
#
environment.update({
'CCACHE_MAXSIZE': '500G',
'CCACHE_TEMPDIR': '/tmp/ccache', # temp dir should be local and not shared
'CCACHE_DIR': '/work/ccache', # this path is inside the container as /work/ccache is
# mounted
'CCACHE_LOGFILE': '/tmp/ccache.log', # a container-scoped log, useful for ccache
# verification.
})
# These variables are passed to the container to the process tree killer can find runaway
# process inside the container
# https://wiki.jenkins.io/display/JENKINS/ProcessTreeKiller
# https://github.com/jenkinsci/jenkins/blob/578d6bacb33a5e99f149de504c80275796f0b231/core/src/main/java/hudson/model/Run.java#L2393
#
jenkins_env_vars = ['BUILD_NUMBER', 'BUILD_ID', 'BUILD_TAG']
environment.update({k: os.environ[k] for k in jenkins_env_vars if k in os.environ})
environment.update({k: os.environ[k] for k in ['CCACHE_MAXSIZE'] if k in os.environ})
tag = get_docker_tag(platform=platform, registry=docker_registry)
mx_root = get_mxnet_root()
local_build_folder = buildir()
# We need to create it first, otherwise it will be created by the docker daemon with root only permissions
os.makedirs(local_build_folder, exist_ok=True)
os.makedirs(local_ccache_dir, exist_ok=True)
logging.info("Using ccache directory: %s", local_ccache_dir)
docker_client = docker.from_env()
# Equivalent command
docker_cmd_list = [
get_docker_binary(nvidia_runtime),
'run',
"--cap-add",
"SYS_PTRACE", # Required by ASAN
'--rm',
'--shm-size={}'.format(shared_memory_size),
# mount mxnet root
'-v', "{}:/work/mxnet".format(mx_root),
# mount mxnet/build for storing build
'-v', "{}:/work/build".format(local_build_folder),
'-v', "{}:/work/ccache".format(local_ccache_dir),
'-u', '{}:{}'.format(os.getuid(), os.getgid()),
'-e', 'CCACHE_MAXSIZE={}'.format(environment['CCACHE_MAXSIZE']),
# temp dir should be local and not shared
'-e', 'CCACHE_TEMPDIR={}'.format(environment['CCACHE_TEMPDIR']),
# this path is inside the container as /work/ccache is mounted
'-e', "CCACHE_DIR={}".format(environment['CCACHE_DIR']),
# a container-scoped log, useful for ccache verification.
'-e', "CCACHE_LOGFILE={}".format(environment['CCACHE_LOGFILE']),
'-ti',
tag]
docker_cmd_list.extend(command)
docker_cmd = ' \\\n\t'.join(docker_cmd_list)
logging.info("Running %s in container %s", command, tag)
logging.info("Executing the equivalent of:\n%s\n", docker_cmd)
# return code of the command inside docker
ret = 0
if not dry_run:
#############################
#
signal.pthread_sigmask(signal.SIG_BLOCK, {signal.SIGINT, signal.SIGTERM})
# noinspection PyShadowingNames
runtime = None
if nvidia_runtime:
# noinspection PyShadowingNames
# runc is default (docker info | grep -i runtime)
runtime = 'nvidia'
container = docker_client.containers.run(
tag,
runtime=runtime,
detach=True,
command=command,
shm_size=shared_memory_size,
user='{}:{}'.format(os.getuid(), os.getgid()),
cap_add='SYS_PTRACE',
volumes={
mx_root:
{'bind': '/work/mxnet', 'mode': 'rw'},
local_build_folder:
{'bind': '/work/build', 'mode': 'rw'},
local_ccache_dir:
{'bind': '/work/ccache', 'mode': 'rw'},
},
environment=environment)
try:
logging.info("Started container: %s", trim_container_id(container.id))
# Race condition:
# If the previous call is interrupted then it's possible that the container is not cleaned up
# We avoid by masking the signals temporarily
cleanup.add_container(container)
signal.pthread_sigmask(signal.SIG_UNBLOCK, {signal.SIGINT, signal.SIGTERM})
#
#############################
stream = container.logs(stream=True, stdout=True, stderr=True)
sys.stdout.flush()
for chunk in stream:
sys.stdout.buffer.write(chunk)
sys.stdout.buffer.flush()
sys.stdout.flush()
stream.close()
try:
logging.info("Waiting for status of container %s for %d s.",
trim_container_id(container.id),
container_wait_s)
wait_result = container.wait(timeout=container_wait_s)
logging.info("Container exit status: %s", wait_result)
ret = wait_result.get('StatusCode', 200)
if ret != 0:
logging.error("Container exited with an error 😞")
else:
logging.info("Container exited with success 👍")
except Exception as e:
logging.exception(e)
ret = 150
# Stop
try:
logging.info("Stopping container: %s", trim_container_id(container.id))
container.stop()
except Exception as e:
logging.exception(e)
ret = 151
# Remove
try:
logging.info("Removing container: %s", trim_container_id(container.id))
container.remove()
except Exception as e:
logging.exception(e)
ret = 152
cleanup.remove_container(container)
containers = docker_client.containers.list()
if containers:
logging.info("Other running containers: %s", [trim_container_id(x.id) for x in containers])
except docker.errors.NotFound as e:
logging.info("Container was stopped before cleanup started: %s", e)
return ret
|
Imports tagged container from the given docker registry
|
def load_docker_cache(tag, docker_registry) -> None:
"""Imports tagged container from the given docker registry"""
if docker_registry:
# noinspection PyBroadException
try:
import docker_cache
logging.info('Docker cache download is enabled from registry %s', docker_registry)
docker_cache.load_docker_cache(registry=docker_registry, docker_tag=tag)
except Exception:
logging.exception('Unable to retrieve Docker cache. Continue without...')
else:
logging.info('Distributed docker cache disabled')
|
Load a list of arrays into a list of arrays specified by slices.
|
def _load_general(data, targets, major_axis):
"""Load a list of arrays into a list of arrays specified by slices."""
for d_src, d_targets, axis in zip(data, targets, major_axis): # pylint: disable=too-many-nested-blocks
if isinstance(d_targets, nd.NDArray):
d_src.copyto(d_targets)
elif isinstance(d_src, (list, tuple)):
for src, dst in zip(d_src, d_targets):
src.copyto(dst)
else:
for slice_idx, d_dst in d_targets:
if axis >= 0:
# copy slice
shape = d_src.shape
do_crop = (slice_idx.start != 0 or shape[axis] != slice_idx.stop)
# pylint: disable=no-member,protected-access
if do_crop:
if axis == 0:
d_src[slice_idx.start:slice_idx.stop].copyto(d_dst)
else:
if d_src.context == d_dst.context:
nd.slice_axis(d_src, axis=axis, begin=slice_idx.start,
end=slice_idx.stop, out=d_dst)
else:
# on different device, crop and then do cross device copy
d_dst_copy = nd.slice_axis(d_src, axis=axis, begin=slice_idx.start,
end=slice_idx.stop)
d_dst_copy.copyto(d_dst)
else:
d_src.copyto(d_dst)
# pylint: enable=no-member,protected-access
else:
d_src.copyto(d_dst)
|
Load data into sliced arrays.
|
def _load_data(batch, targets, major_axis):
"""Load data into sliced arrays."""
if isinstance(batch, list):
new_batch = []
for i in range(len(targets)):
new_batch.append([b.data[i] for b in batch])
new_targets = [[dst for _, dst in d_target] for d_target in targets]
_load_general(new_batch, new_targets, major_axis)
else:
_load_general(batch.data, targets, major_axis)
|
Merge outputs that lives on multiple context into one, so that they look
like living on one context.
|
def _merge_multi_context(outputs, major_axis):
"""Merge outputs that lives on multiple context into one, so that they look
like living on one context.
"""
rets = []
for tensors, axis in zip(outputs, major_axis):
if axis >= 0:
# pylint: disable=no-member,protected-access
if len(tensors) == 1:
rets.append(tensors[0])
else:
# Concatenate if necessary
rets.append(nd.concat(*[tensor.as_in_context(tensors[0].context)
for tensor in tensors],
dim=axis))
# pylint: enable=no-member,protected-access
else:
# negative axis means the there is no batch_size axis, and all the
# results should be the same on each device. We simply take the
# first one, without checking they are actually the same
rets.append(tensors[0])
return rets
|
Prepare the group2contexts, will duplicate the context
if some ctx_group map to only one context.
|
def _prepare_group2ctxs(group2ctxs, ctx_len):
"""Prepare the group2contexts, will duplicate the context
if some ctx_group map to only one context.
"""
if group2ctxs is None:
return [None] * ctx_len
elif isinstance(group2ctxs, list):
assert(len(group2ctxs) == ctx_len), "length of group2ctxs\
should be %d" % ctx_len
return group2ctxs
elif isinstance(group2ctxs, dict):
ret = [{} for i in range(ctx_len)]
for k, v in group2ctxs.items():
ctxs = None
if isinstance(v, ctx.Context):
ctxs = [v] * ctx_len
else:
if len(v) == 1:
ctxs = v * ctx_len
else:
assert(len(v) == ctx_len), "length of group2ctxs[%s]\
should be %d or 1" % (k, ctx_len)
ctxs = v
for i in range(ctx_len):
ret[i][k] = ctxs[i]
return ret
else:
assert(False), "group2ctxs should be list of dict of str to context,\
or dict of str to context or list of context"
return False
|
Decide the slices for each context according to the workload.
Parameters
----------
data_shapes : list
list of (name, shape) specifying the shapes for the input data or label.
|
def decide_slices(self, data_shapes):
"""Decide the slices for each context according to the workload.
Parameters
----------
data_shapes : list
list of (name, shape) specifying the shapes for the input data or label.
"""
assert len(data_shapes) > 0
major_axis = [DataDesc.get_batch_axis(x.layout) for x in data_shapes]
for (name, shape), axis in zip(data_shapes, major_axis):
if axis == -1:
continue
batch_size = shape[axis]
if self.batch_size is not None:
assert batch_size == self.batch_size, ("all data must have the same batch size: "
+ ("batch_size = %d, but " % self.batch_size)
+ ("%s has shape %s" % (name, shape)))
else:
self.batch_size = batch_size
self.slices = _split_input_slice(self.batch_size, self.workload)
return major_axis
|
Collect internal arrays from executors.
|
def _collect_arrays(self):
"""Collect internal arrays from executors."""
# convenient data structures
self.data_arrays = [[(self.slices[i], e.arg_dict[name]) for i, e in enumerate(self.execs)]
for name, _ in self.data_shapes]
self.state_arrays = [[e.arg_dict[name] for e in self.execs]
for name in self.state_names]
if self.label_shapes is not None:
self.label_arrays = [[(self.slices[i], e.arg_dict[name])
for i, e in enumerate(self.execs)]
for name, _ in self.label_shapes]
else:
self.label_arrays = None
self.param_arrays = [[exec_.arg_arrays[i] for exec_ in self.execs]
for i, name in enumerate(self.arg_names)
if name in self.param_names]
if self.for_training:
self.grad_arrays = [[exec_.grad_arrays[i] for exec_ in self.execs]
for i, name in enumerate(self.arg_names)
if name in self.param_names]
else:
self.grad_arrays = None
data_names = [x[0] for x in self.data_shapes]
if self.inputs_need_grad:
self.input_grad_arrays = [[exec_.grad_arrays[self.arg_names.index(name)]
for exec_ in self.execs]
for name in data_names if name in self.arg_names]
else:
self.input_grad_arrays = None
self.aux_arrays = [[exec_.aux_arrays[i] for exec_ in self.execs]
for i in range(len(self.aux_names))]
|
Bind executors on their respective devices.
Parameters
----------
data_shapes : list
label_shapes : list
shared_group : DataParallelExecutorGroup
reshape : bool
|
def bind_exec(self, data_shapes, label_shapes, shared_group=None, reshape=False):
"""Bind executors on their respective devices.
Parameters
----------
data_shapes : list
label_shapes : list
shared_group : DataParallelExecutorGroup
reshape : bool
"""
assert reshape or not self.execs
self.batch_size = None
# calculate workload and bind executors
self.data_layouts = self.decide_slices(data_shapes)
if label_shapes is not None:
# call it to make sure labels has the same batch size as data
self.label_layouts = self.decide_slices(label_shapes)
for i in range(len(self.contexts)):
data_shapes_i = self._sliced_shape(data_shapes, i, self.data_layouts)
if label_shapes is not None:
label_shapes_i = self._sliced_shape(label_shapes, i, self.label_layouts)
else:
label_shapes_i = []
if reshape:
self.execs[i] = self._default_execs[i].reshape(
allow_up_sizing=True, **dict(data_shapes_i + label_shapes_i))
else:
self.execs.append(self._bind_ith_exec(i, data_shapes_i, label_shapes_i,
shared_group))
self.data_shapes = data_shapes
self.label_shapes = label_shapes
self.data_names = [i.name for i in self.data_shapes]
if label_shapes is not None:
self.label_names = [i.name for i in self.label_shapes]
self._collect_arrays()
|
Reshape executors.
Parameters
----------
data_shapes : list
label_shapes : list
|
def reshape(self, data_shapes, label_shapes):
"""Reshape executors.
Parameters
----------
data_shapes : list
label_shapes : list
"""
if data_shapes == self.data_shapes and label_shapes == self.label_shapes:
return
if self._default_execs is None:
self._default_execs = [i for i in self.execs]
self.bind_exec(data_shapes, label_shapes, reshape=True)
|
Assign, i.e. copy parameters to all the executors.
Parameters
----------
arg_params : dict
A dictionary of name to `NDArray` parameter mapping.
aux_params : dict
A dictionary of name to `NDArray` auxiliary variable mapping.
allow_extra : boolean, optional
Whether allow extra parameters that are not needed by symbol.
If this is True, no error will be thrown when arg_params or aux_params
contain extra parameters that is not needed by the executor.
|
def set_params(self, arg_params, aux_params, allow_extra=False):
"""Assign, i.e. copy parameters to all the executors.
Parameters
----------
arg_params : dict
A dictionary of name to `NDArray` parameter mapping.
aux_params : dict
A dictionary of name to `NDArray` auxiliary variable mapping.
allow_extra : boolean, optional
Whether allow extra parameters that are not needed by symbol.
If this is True, no error will be thrown when arg_params or aux_params
contain extra parameters that is not needed by the executor.
"""
for exec_ in self.execs:
exec_.copy_params_from(arg_params, aux_params, allow_extra_params=allow_extra)
|
Copy data from each executor to `arg_params` and `aux_params`.
Parameters
----------
arg_params : list of NDArray
Target parameter arrays.
aux_params : list of NDArray
Target aux arrays.
Notes
-----
- This function will inplace update the NDArrays in arg_params and aux_params.
|
def get_params(self, arg_params, aux_params):
""" Copy data from each executor to `arg_params` and `aux_params`.
Parameters
----------
arg_params : list of NDArray
Target parameter arrays.
aux_params : list of NDArray
Target aux arrays.
Notes
-----
- This function will inplace update the NDArrays in arg_params and aux_params.
"""
for name, block in zip(self.param_names, self.param_arrays):
weight = sum(w.copyto(ctx.cpu()) for w in block) / len(block)
weight.astype(arg_params[name].dtype).copyto(arg_params[name])
for name, block in zip(self.aux_names, self.aux_arrays):
weight = sum(w.copyto(ctx.cpu()) for w in block) / len(block)
weight.astype(aux_params[name].dtype).copyto(aux_params[name])
|
Split `data_batch` according to workload and run forward on each devices.
Parameters
----------
data_batch : DataBatch
Or could be any object implementing similar interface.
is_train : bool
The hint for the backend, indicating whether we are during training phase.
Default is `None`, then the value `self.for_training` will be used.
Returns
-------
|
def forward(self, data_batch, is_train=None):
"""Split `data_batch` according to workload and run forward on each devices.
Parameters
----------
data_batch : DataBatch
Or could be any object implementing similar interface.
is_train : bool
The hint for the backend, indicating whether we are during training phase.
Default is `None`, then the value `self.for_training` will be used.
Returns
-------
"""
_load_data(data_batch, self.data_arrays, self.data_layouts)
if is_train is None:
is_train = self.for_training
if isinstance(data_batch, list):
if self.label_arrays is not None and data_batch is not None and data_batch[0].label:
_load_label(data_batch, self.label_arrays, self.label_layouts)
else:
if self.label_arrays is not None and data_batch.label:
_load_label(data_batch, self.label_arrays, self.label_layouts)
for exec_ in self.execs:
exec_.forward(is_train=is_train)
|
Get the shapes of the outputs.
|
def get_output_shapes(self):
"""Get the shapes of the outputs."""
outputs = self.execs[0].outputs
shapes = [out.shape for out in outputs]
concat_shapes = []
for key, the_shape, axis in zip(self.symbol.list_outputs(), shapes, self.output_layouts):
the_shape = list(the_shape)
if axis >= 0:
the_shape[axis] = self.batch_size
concat_shapes.append((key, tuple(the_shape)))
return concat_shapes
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.