![]() More specifically, we made TVM able to generate OpenGL shading language (GLSL) kernels to perform tensor computation on the GPU. ![]() In our project, we added a new backend platform for TVM: OpenGL. In other words, TVM is considered the LLVM for deep learning. Then, TVM supports further transformations into platform-specific code: CUDA, OpenCL, etc. After the user uses MXNet (or other frameworks that TVM intends to support) to create a machine learning program, the computation graph is transformed into a lower-level but still cross-platform representation in TVM. ![]() MXNet supports running deep learning algorithms in various environments: CPUs, GPUs, or even mobile devices.Īn active project within MXNet is TVM, an intermediate representation for tensor computation. The programmer specifies a high-level computation graph, and MXNet utilizes a data-flow runtime scheduler to execute the graph in a parallel / distributed setting, depending on the available computation resources. MXNet is an open-source deep learning framework, similar to TensorFlow, Caffe, CNTK, etc.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |