One of basic units of computation in Caffe2 are the Operators. Each operator contains the logic necessary to compute the output given the appropriate number and types of inputs and parameters. The overall difference between operators’ functionality in Caffe and Caffe2 is illustrated in the following graphic, respectively:

operators comparison

As a result, for example, in the Fully Connected operator, each of the input X, bias b, and the weight matrix W must be provided, and a single output will be computed.

Operators Catalogue

For a full listing of Caffe2 Operators, refer to the Operators Catalogue:

Writing Your Own Custom Operators

Fantastic idea! Write custom operators and share them with the community! Refer to the guide on writing operators:

Sparse Operations

Caffe2 provides support for representing sparse features and performing corresponding operations on segments of tensors. Refer to the guide on sparse operators:

Edit on GitHub