You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 
nihui f3651ce5b8
pnnx convert onnx depth2space, add pixelshuffle pixelunshuffle test (#5967)
1 year ago
..
pass_level0 pnnx do not fold tensor with dynamic shape, use fp32 module by default (#5755) 1 year ago
pass_level1 RMSNorm (#5630) 1 year ago
pass_level2 pnnx convert onnx depth2space, add pixelshuffle pixelunshuffle test (#5967) 1 year ago
pass_level3 tnn2pnnx (#5898) 1 year ago
pass_level4 PNNX is an open standard for PyTorch model interoperability (#3262) 4 years ago
pass_level5 pnnx eval typed expression, optimize type conversion (#5921) 1 year ago
pass_ncnn pnnx convert onnx conv with asymmatric pads (#5965) 1 year ago
pass_onnx pnnx fix onnx clip conversion, add onnx clamp test, always reserve onnx split outputs, convert onnx mod (#5834) 1 year ago
pass_tnn tnn2pnnx (#5898) 1 year ago
CMakeLists.txt pnnx convert onnx conv with asymmatric pads (#5965) 1 year ago
ir.cpp tnn2pnnx (#5898) 1 year ago
ir.h tnn2pnnx (#5898) 1 year ago
load_onnx.cpp pnnx drop onnx weight-like graph input (#5736) 1 year ago
load_onnx.h pnnx onnx with inputshape (#5480) 2 years ago
load_tnn.cpp tnn2pnnx (#5898) 1 year ago
load_tnn.h tnn2pnnx (#5898) 1 year ago
load_torchscript.cpp pnnx auto inputshape from traced inputs (#5825) 1 year ago
load_torchscript.h split load torchscript sources from pnnx, drop cxxabi hack, link torchscript with whole-archive (#5348) 2 years ago
main.cpp tnn2pnnx (#5898) 1 year ago
onnx-data.proto pnnx ci for torch 2.4 (#5618) 1 year ago
onnx-ml.proto pnnx ci for torch 2.4 (#5618) 1 year ago
onnx-operators-ml.proto pnnx ci for torch 2.4 (#5618) 1 year ago
pass_level0.cpp pnnx support dynamic slice indexes (#5299) 2 years ago
pass_level0.h save foldable constants in file for reducing memory usage (#4337) 3 years ago
pass_level1.cpp fix moduleop attribute order, write weight shape as parameter for ncnn module op (#4970) 2 years ago
pass_level1.h pnnx export weight inside moduleop (#3902) 4 years ago
pass_level2.cpp pnnx pass level2 priority (#5791) 1 year ago
pass_level2.h fuse adaptive pool dynamic output size, implement ncnn adaptive pooling dynamic outsize (#5043) 2 years ago
pass_level3.cpp pnnx pass level2 priority (#5791) 1 year ago
pass_level3.h pnnx recognize foldable scalar and fuse it into expression (#4639) 3 years ago
pass_level4.cpp pnnx fold constant (#3521) 4 years ago
pass_level4.h PNNX is an open standard for PyTorch model interoperability (#3262) 4 years ago
pass_level5.cpp pnnx fuse t5-layernorm as rmsnorm (#5675) 1 year ago
pass_level5.h save foldable constants in file for reducing memory usage (#4337) 3 years ago
pass_ncnn.cpp pnnx convert onnx conv with asymmatric pads (#5965) 1 year ago
pass_ncnn.h fix moduleop attribute order, write weight shape as parameter for ncnn module op (#4970) 2 years ago
pass_onnx.cpp pnnx fix onnx clip conversion, add onnx clamp test, always reserve onnx split outputs, convert onnx mod (#5834) 1 year ago
pass_onnx.h pnnx load dynamo onnx (#5363) 2 years ago
save_ncnn.cpp fix pnnx2ncnn i64 to i32 (#5881) 1 year ago
save_ncnn.h pnnx fp16 option for ncnn and onnx weight type (#4350) 3 years ago
save_onnx.cpp pnnx ci for torch 2.4 (#5618) 1 year ago
save_onnx.h pnnx fp16 option for ncnn and onnx weight type (#4350) 3 years ago
storezip.cpp pnnx model binary over 4g (#5274) 2 years ago
storezip.h pnnx model binary over 4g (#5274) 2 years ago
utils.cpp split load torchscript sources from pnnx, drop cxxabi hack, link torchscript with whole-archive (#5348) 2 years ago
utils.h split load torchscript sources from pnnx, drop cxxabi hack, link torchscript with whole-archive (#5348) 2 years ago