onnx转换出错

[q2547978337]

error log | 日志或报错信息 | ログ

model | 模型 | モデル

1. original model
   pnnxparam = 2.pnnx.param
   pnnxbin = 2.pnnx.bin
   pnnxpy = 2_pnnx.py
   pnnxonnx = 2.pnnx.onnx
   ncnnparam = 2.ncnn.param
   ncnnbin = 2.ncnn.bin
   ncnnpy = 2_ncnn.py
   fp16 = 1
   optlevel = 2
   device = cpu
   inputshape = [1,3,320,320]f32
   inputshape2 =
   customop =
   moduleop =
   ############# pass_level0 onnx
   inline_containers ... 0.00ms
   eliminate_noop ... 0.35ms
   fold_constants ... 0.19ms
   canonicalize ... 0.36ms
   shape_inference ... 106.13ms
   fold_constants_dynamic_shape ... 0.18ms
   inline_if_graph ... 0.03ms
   fuse_constant_as_attribute ... 0.24ms
   eliminate_noop_with_shape ... 0.14ms
   ┌──────────────────┬──────────┬──────────┐
   │ │ orig │ opt │
   ├──────────────────┼──────────┼──────────┤
   │ node │ 277 │ 277 │
   │ initializer │ 173 │ 166 │
   │ functions │ 0 │ 0 │
   ├──────────────────┼──────────┼──────────┤
   │ nn module op │ 0 │ 0 │
   │ custom module op │ 0 │ 0 │
   │ aten op │ 0 │ 0 │
   │ prims op │ 0 │ 0 │
   │ onnx native op │ 277 │ 277 │
   ├──────────────────┼──────────┼──────────┤
   │ Add │ 7 │ 7 │
   │ Concat │ 18 │ 18 │
   │ Conv │ 83 │ 83 │
   │ MaxPool │ 3 │ 3 │
   │ Mul │ 74 │ 74 │
   │ Reshape │ 3 │ 3 │
   │ Resize │ 2 │ 2 │
   │ Sigmoid │ 80 │ 80 │
   │ Slice │ 6 │ 6 │
   │ Transpose │ 1 │ 1 │
   └──────────────────┴──────────┴──────────┘
   ############# pass_level1 onnx
   ############# pass_level2
   ############# pass_level3
   open failed
   ############# pass_level4
   ############# pass_level5
   ############# pass_ncnn
   slice with step 2 is not supported
   slice with step 2 is not supported
   slice with step 2 is not supported
   slice with step 2 is not supported
   slice with step 2 is not supported
   slice with step 2 is not supported

D:\ncnn\zhuanmoxing\pnnx-20240819-windows>pnnx 2.onnx inputshape=[1,3,320,320]
pnnxparam = 2.pnnx.param
pnnxbin = 2.pnnx.bin
pnnxpy = 2_pnnx.py
pnnxonnx = 2.pnnx.onnx
ncnnparam = 2.ncnn.param
ncnnbin = 2.ncnn.bin
ncnnpy = 2_ncnn.py
fp16 = 1
optlevel = 2
device = cpu
inputshape = [1,3,320,320]f32
inputshape2 =
customop =
moduleop =
############# pass_level0 onnx
inline_containers ... 0.01ms
eliminate_noop ... 0.36ms
fold_constants ... 0.18ms
canonicalize ... 0.37ms
shape_inference ... 96.65ms
fold_constants_dynamic_shape ... 0.17ms
inline_if_graph ... 0.04ms
fuse_constant_as_attribute ... 0.23ms
eliminate_noop_with_shape ... 0.13ms
┌──────────────────┬──────────┬──────────┐
│ │ orig │ opt │
├──────────────────┼──────────┼──────────┤
│ node │ 277 │ 277 │
│ initializer │ 173 │ 166 │
│ functions │ 0 │ 0 │
├──────────────────┼──────────┼──────────┤
│ nn module op │ 0 │ 0 │
│ custom module op │ 0 │ 0 │
│ aten op │ 0 │ 0 │
│ prims op │ 0 │ 0 │
│ onnx native op │ 277 │ 277 │
├──────────────────┼──────────┼──────────┤
│ Add │ 7 │ 7 │
│ Concat │ 18 │ 18 │
│ Conv │ 83 │ 83 │
│ MaxPool │ 3 │ 3 │
│ Mul │ 74 │ 74 │
│ Reshape │ 3 │ 3 │
│ Resize │ 2 │ 2 │
│ Sigmoid │ 80 │ 80 │
│ Slice │ 6 │ 6 │
│ Transpose │ 1 │ 1 │
└──────────────────┴──────────┴──────────┘
############# pass_level1 onnx
############# pass_level2
############# pass_level3
open failed
############# pass_level4
############# pass_level5
############# pass_ncnn
slice with step 2 is not supported
slice with step 2 is not supported
slice with step 2 is not supported
slice with step 2 is not supported
slice with step 2 is not supported

how to reproduce | 复现步骤 | 再現方法