modified:   bmtrain/optim/adam.py
	modified:   bmtrain/optim/adam_offload.py
	modified:   csrc/cuda/adam_cuda.cu
	modified:   csrc/include/adam_cpu.hpp

	modified:   csrc/bind.cpp
	modified:   csrc/cuda/adam_cuda.cu
	modified:   csrc/include/adam_cpu.hpp
	modified:   csrc/include/bind.hpp

modified: add bf16.h to csrc/cuda/has_inf_nan.cu
modified: add bf16 support to bmtrain/optim/_function.py
modified: add bf16 support to csrc/cuda/adam_cuda.cu(Todo)
modified: add bf16.h to csrc/cuda/reduce.cuh
modified: add bf16 support to csrc/include/adam_cpu.hpp

	modified:   csrc/cuda/adam_cuda.cu
	modified:   csrc/cuda/reduce.cuh
	modified:   csrc/include/adam_cpu.hpp

	modified:   bmtrain/optim/optim_manager.py
	modified:   csrc/bind.cpp
	modified:   csrc/cuda/has_inf_nan.cu
	modified:   csrc/include/adam_cpu.hpp
	new file:   csrc/include/test.cpp

	modified:   bind bf16 functions
	modified:   revise the logic of bf16 accum
	modified:   add bf16 to is_nan_inf() of CUDA version
	modified:   revise the logic of bf16 accum
	modified:   bind bf16 functions
	deleted:    csrc/include/test.cpp

modified： revise logic of has_inf_nan.cu
FIX：csrc/include/adam_cpu.hpp
FIX：csrc/include/bind.hpp

	new file:   tests/test_has_inf_nan_bf16.py

	modified:   csrc/bind.cpp
	new file:   tests/test_optim_bf16.py
	new file:   tests/test_optim_fp16.py