I ask because I noticed that PyTorch 1.0 announced its script mode, which, according to the source code, extracts the AST of the forward pass defined as a Python function, like what tangent does.
The script mode generates a graph to be executed by the Caffe2 runtime, thus differs from tangent that generates Python code. Is it a reasonable idea to make tangent the script mode of Eager Execution?
I ask because I noticed that PyTorch 1.0 announced its script mode, which, according to the source code, extracts the AST of the forward pass defined as a Python function, like what tangent does.
The script mode generates a graph to be executed by the Caffe2 runtime, thus differs from tangent that generates Python code. Is it a reasonable idea to make tangent the script mode of Eager Execution?