strisunshinewentingwang/pytorch

Pull Requests Commits

try to fix XLA error: torch_xla/csrc/helpers.cpp:85 : Check failed: min_shape_dim <= dim && dim <= max_shape_dim, when running self.slice(dim, start, stop, step)

Will Feng committed 4 years ago

9797b4e5

remove getTensorSize

Will Feng committed 4 years ago

e6a544b1

try to speed up 1-D get: none / ellipsis

Will Feng committed 4 years ago

4232f981

try to optimize dispatch_index / dispatch_index_put_

Will Feng committed 4 years ago

f1f57df0

try to remove fast path for N-D integer indexing

Will Feng committed 4 years ago

363abdd0

Optimize sizes() and dim()

Will Feng committed 4 years ago

85c931bd

avoid repeated calls to .device()

Will Feng committed 4 years ago

751aaf09

simplify recordSelectTrace

Will Feng committed 4 years ago

0629ce42

perf debug: add fast path for N-D integer indexing

Will Feng committed 4 years ago

578d351b

perf debug: use old indexing code for 1-D get

Will Feng committed 4 years ago

7ebfe937

perf debug: remove traceAndConvertPythonIndexToTensorIndex and inline its logic into callsites

Will Feng committed 4 years ago

db654748

add AutoNonVariableTypeMode for C++ set_item path

Will Feng committed 4 years ago

95a4682a

try to improve applySlice

Will Feng committed 4 years ago

567f8d2d

fix at::alias usage and reorder checks in handleSimpleTypesInSingleDimIndexingGet

Will Feng committed 4 years ago

ba4abb44

optimization: at::native::size, device dispatch for boolToIndexingTensor and scalarToTensor, THPVariable_Wrap

Will Feng committed 4 years ago

bc9918b2

try to clean up more header includes

Will Feng committed 4 years ago

df19e6b4

try to use at:: instead of at::native::

Will Feng committed 4 years ago

c8dd0f8b

Will Feng committed 4 years ago

daaa08d2

reorder single-dim fast path condition

Will Feng committed 4 years ago

529ac006

move tracing into indexToTensorIndex

Will Feng committed 4 years ago

afaf602b

Revert "perf debug: tensor_index_tmp"

Will Feng committed 4 years ago

b0788af4

perf debug: tensor_index_tmp

Will Feng committed 4 years ago

c9aa4b78

simplify handle* functions

Will Feng committed 4 years ago

dd48dafe

simplify slice extract tensor

Will Feng committed 4 years ago

fadeeb5b

python single-dim set uses C++ path

Will Feng committed 4 years ago

7a62831f

simplify c++ single dim set path

Will Feng committed 4 years ago

304306a2

merge python/c++ getitem single dim path

Will Feng committed 4 years ago

cf89bb01

simplify single dim get

Will Feng committed 4 years ago

ab9503f1

inline -> static inline

Will Feng committed 4 years ago

ae5a63d7

remove redundant isTracing check

Will Feng committed 4 years ago

577ad57b

Older