FM / modeling

/usr/local/lib64/python3.6/site-packages/caffe2/python/modeling/

/usr/local/lib64/python3.6/site-packages/caffe2/python/modeling

Name	Size	Mode	Actions
__pycache__/	-	0755	rm
compute_histogram_for_blobs.py	3508	0644	edit dl rm
compute_histogram_for_blobs_test.py	4947	0644	edit dl rm
compute_norm_for_blobs.py	3642	0644	edit dl rm
compute_norm_for_blobs_test.py	7942	0644	edit dl rm
compute_statistics_for_blobs.py	1909	0644	edit dl rm
compute_statistics_for_blobs_test.py	2870	0644	edit dl rm
get_entry_from_blobs.py	3130	0644	edit dl rm
get_entry_from_blobs_test.py	3538	0644	edit dl rm
gradient_clipping.py	5766	0644	edit dl rm
gradient_clipping_test.py	10016	0644	edit dl rm
initializers.py	5378	0644	edit dl rm
initializers_test.py	2106	0644	edit dl rm
net_modifier.py	823	0644	edit dl rm
parameter_info.py	1438	0644	edit dl rm
parameter_sharing.py	3848	0644	edit dl rm
parameter_sharing_test.py	7223	0644	edit dl rm
__init__.py	0	0644	edit dl rm

Edit: /usr/local/lib64/python3.6/site-packages/caffe2/python/modeling/compute_statistics_for_blobs.py (1909B)

from caffe2.python import core, schema
from caffe2.python.modeling.net_modifier import NetModifier

import numpy as np

class ComputeStatisticsForBlobs(NetModifier):
    """
    This class modifies the net passed in by adding ops to compute statistics
    for certain blobs. For each blob in the list, its min, max, mean and standard
    deviation will be computed.

Args:
        blobs: list of blobs to compute norm for
        logging_frequency: frequency for printing norms to logs
    """

def __init__(self, blobs, logging_frequency):
        self._blobs = blobs
        self._logging_frequency = logging_frequency
        self._field_name_suffix = '_summary'

def modify_net(self, net, init_net=None, grad_map=None, blob_to_device=None,
                   modify_output_record=False):

for blob_name in self._blobs:
            blob = core.BlobReference(blob_name)
            assert net.BlobIsDefined(blob), 'blob {} is not defined in net {} whose proto is {}'.format(blob, net.Name(), net.Proto())

cast_blob = net.Cast(blob, to=core.DataType.FLOAT)
            stats_name = net.NextScopedBlob(prefix=blob + self._field_name_suffix)
            stats = net.Summarize(cast_blob, stats_name, to_file=0)
            net.Print(stats, [], every_n=self._logging_frequency)

if modify_output_record:
                output_field_name = str(blob) + self._field_name_suffix
                output_scalar = schema.Scalar((np.float, (1,)), stats)

if net.output_record() is None:
                    net.set_output_record(
                        schema.Struct((output_field_name, output_scalar))
                    )
                else:
                    net.AppendOutputRecordField(
                        output_field_name,
                        output_scalar)

def field_name_suffix(self):
        return self._field_name_suffix

from caffe2.python import core, schema from caffe2.python.modeling.net_modifier import NetModifier import numpy as np class ComputeStatisticsForBlobs(NetModifier): """ This class modifies the net passed in by adding ops to compute statistics for certain blobs. For each blob in the list, its min, max, mean and standard deviation will be computed. Args: blobs: list of blobs to compute norm for logging_frequency: frequency for printing norms to logs """ def __init__(self, blobs, logging_frequency): self._blobs = blobs self._logging_frequency = logging_frequency self._field_name_suffix = '_summary' def modify_net(self, net, init_net=None, grad_map=None, blob_to_device=None, modify_output_record=False): for blob_name in self._blobs: blob = core.BlobReference(blob_name) assert net.BlobIsDefined(blob), 'blob {} is not defined in net {} whose proto is {}'.format(blob, net.Name(), net.Proto()) cast_blob = net.Cast(blob, to=core.DataType.FLOAT) stats_name = net.NextScopedBlob(prefix=blob + self._field_name_suffix) stats = net.Summarize(cast_blob, stats_name, to_file=0) net.Print(stats, [], every_n=self._logging_frequency) if modify_output_record: output_field_name = str(blob) + self._field_name_suffix output_scalar = schema.Scalar((np.float, (1,)), stats) if net.output_record() is None: net.set_output_record( schema.Struct((output_field_name, output_scalar)) ) else: net.AppendOutputRecordField( output_field_name, output_scalar) def field_name_suffix(self): return self._field_name_suffix