• Chris Austen's avatar
    Final performance improvements for release (#1369) · a85b183b
    Chris Austen authored
    * Improvements to handling and add constant passed to dot operator (#1280)
    * Improve horizontal fusion of contiguous (#1292)
    * Add pass to rewrite gelu as fast gelu (#1299)
    * Add jit layernorm fusion (#1301)
    a85b183b
compile_gen.cpp 6.26 KB