Houjiang Chen

support non contiguous inplace

4

- 系统解决目前不支持非contiguous的算子的非contiguous输入inplace计算的问题 比如add不支持non-contiguous，那么对于non-contiguous的输入tensor是不允许使用inplace操作的，比如`a += 1`。但该PR会自动将`a += 1`重写为`b = a + 1; a = b`。 Fixes OneFlow-Inc/OneTeam#1621 Fixes OneFlow-Inc/OneTeam#1627

unified autograd engine

15

This PR aims to make nn graph use the unified autograd engine like eager mode. Background and discussion：https://github.com/Oneflow-Inc/OneTeam/issues/1504 The following optimizations have been adapted： - [x] Gradient Accumulation - [x]...

This PR implement a model converter that converts oneflow mlir model to oneflow-lite format. Usage: ./oneflow/ir/bin/oneflow-lite-compile /examples/resnet50/1/model/ --targets=cuda -o resnet50-lite.bin