Caffe2 - Python API
A deep learning, cross platform ML framework
Public Member Functions | Public Attributes | List of all members
caffe2.python.layers.gather_record.GatherRecord Class Reference
Inheritance diagram for caffe2.python.layers.gather_record.GatherRecord:

Public Member Functions

def __init__ (self, model, input_record, name='gather_record', kwargs)
def add_ops (self, net)
- Public Member Functions inherited from caffe2.python.layers.layers.ModelLayer
def __init__ (self, model, prefix, input_record, predict_input_record_fields=None, tags=None, kwargs)
def get_type (self)
def predict_input_record (self)
def input_record (self)
def predict_output_schema (self)
def predict_output_schema (self, output_schema)
def output_schema (self)
def output_schema (self, output_schema)
def get_parameters (self)
def get_fp16_compatible_parameters (self)
def get_memory_usage (self)
def add_init_params (self, init_net)
def create_param (self, param_name, shape, initializer, optimizer, ps_param=None, regularizer=None)
def get_next_blob_reference (self, name)
def add_operators (self, net, init_net=None, context=InstantiationContext.TRAINING)
def add_ops (self, net)
def add_eval_ops (self, net)
def add_train_ops (self, net)
def add_ops_to_accumulate_pred (self, net)
def add_param_copy_operators (self, net)
def export_output_for_metrics (self)
def export_params_for_metrics (self)

Public Attributes

- Public Attributes inherited from caffe2.python.layers.layers.ModelLayer

Detailed Description

Given 1-D `indices` tensor, gather elements at `i` in `indices` from all the
blobs in `record`. If a blob is a values blob of a list, all the elements
included by the list's lengths blob are gathered. For example,

    indices = [0, 2]
    record:a = [[0, 1], [2, 3], [4, 5], [6, 7]]
    record:b:lengths = [0, 1, 2, 3]
    record:b:items = [0, 1, 2, 3, 4, 5]

    a = [[0, 1], [4, 5]]
    b:lengths = [0, 2]
    b:items = [1, 2]

This supports nested list.

Definition at line 12 of file

The documentation for this class was generated from the following file: