A lot of Ops with their implementations.

[ZhangZhiPku]

Hi, developer team of onnx2pytorch. I am currently developing an neural network quantization framework: https://github.com/openppl-public/ppq/tree/master/ppq. The really interesting part is that we both need to run an onnx model with pytorch : ) I am glad to share our operator implementations with you: https://github.com/openppl-public/ppq/blob/master/ppq/executor/op/torch/default.py

We support following onnx operators by now(Still work in progress):

1. 'Abs': Abs_forward,
2. 'AdaptiveAvgPool2d': AdaptiveAvgPool2d_forward,
3. 'And':And_forward,
4. 'Add': Add_forward,
5. 'ArgMax': ArgMax_forward,
6. 'AveragePool': AveragePool_forward,
7. 'BatchNormalization': BatchNormalization_forward,
8. 'Cast': Cast_forward,
9. 'Clip': Clip_forward,
10. 'Concat': Concat_forward,
11. 'Constant': Constant_forward,
12. 'ConstantOfShape': ConstantOfShape_forward,
13. 'Conv': Conv_forward,
14. 'ConvTranspose': ConvTranspose_forward,
15. 'Cos': Cos_forward,
16. 'Div': Eltwise_forward,
17. 'Equal': Equal_forward,
18. 'Exp': UnaryEltwise_forward,
19. 'Expand': Expand_forward,
20. 'Flatten': Flatten_forward,
21. 'Gather': Gather_forward,
22. 'GatherElements': Gather_forward,
23. 'GatherND': GatherND_forward,
24. 'Gelu': Gelu_forward,
25. 'Gemm': Gemm_forward,
26. 'grid_sampler': Grid_sampler_forward,
27. 'GlobalAveragePool': AveragePool_forward,
28. 'GlobalMaxPool': MaxPool2d_forward,
29. 'Greater': Greater_forward,
30. 'LayerNorm': LayerNorm_forward,
31. 'LeakyRelu': LeakyRelu_forward,
32. 'Less': Less_forward,
33. 'LogSoftmax': LogSoftmax_forward,
34. 'MatMul': MatMul_forward,
35. 'Max': Eltwise_forward,
36. 'MaxPool': MaxPool2d_forward,
37. 'Min': Eltwise_forward,
38. 'Mul': Mul_forward,
39. 'MultiHeadAttention': MultiHeadAttention_forward,
40. 'NonMaxSuppression': _NMS_forward,
41. 'NonZero': NonZero_forward,
42. 'Not': Not_forward,
43. 'Pad': Pad_forward,
44. 'PRelu': PRelu_forward,
45. 'Range': Range_forward,
46. 'ReduceL2': ReduceL2_forward,
47. 'ReduceMax': ReduceMax_forward,
48. 'ReduceMean': ReduceMean_forward,
49. 'ReduceSum': ReduceSum_forward,
50. 'Relu': UnaryEltwise_forward,
51. 'Reshape': Reshape_forward,
52. 'Resize': Resize_forward,
53. 'ScatterElements': ScatterElements_forward,
54. 'ScatterND': ScatterND_forward,
55. 'Shape': Shape_forward,
56. 'Sigmoid': UnaryEltwise_forward,
57. 'Sin': Sin_forward,
58. 'Slice': Slice_forward,
59. 'Softmax': Softmax_forward,
60. 'Softplus': Softplus_forward,
61. 'Split': Split_forward,
62. 'Squeeze': Squeeze_forward,
63. 'Sub': Eltwise_forward,
64. 'Tile': Tile_forward,
65. 'TopK': TopK_forward,
66. 'Transpose': Transpose_forward,
67. 'Unsqueeze': Unsqueeze_forward,
68. 'Where': Where_forward,
69. 'Sqrt': Sqrt_forward,
70. 'Log': Log_forward,
71. 'Floor': Floor_forward,
72. 'RoiAlign': RoiAlign_forward,
73. 'MMCVRoiAlign': MMCVRoiAlign_forward,
74. 'SpaceToDepth': SpaceToDepth_forward,
75. 'DepthToSpace': DepthToSpace_forward,
76. 'Scale': Scale_forward, # caffe op
77. 'Tanh': Tanh_forward,
78. 'Pow': Pow_forward,
79. 'Crop': Crop_forward, # caffe op
80. 'ChannelShuffle': ChannelShuffle_forward, # caffe op
81. 'InstanceNormalization': InstanceNormalization_forward,
82. 'Parameter': Parameter_forward, # caffe op
83. 'Interp': Interp_forward, # caffe op
84. 'CaffeArgMax': CaffeArgMax_forward, # caffe op
85. 'HardSigmoid': HardSigmoid_forward,
86. 'HardSwish': HardSwish_forward,
87. 'Neg': Neg_forward,
88. 'GRU': GRU_forward,
89. 'PPQDeviceSwitch': PPQDeviceSwitch_forward,
90. 'Identity': Identity_forward,
91. 'OneHot': Onehot_forward,
92. 'Reciprocal': Reciprocal_forward,
93. 'LSTM': LSTM_forward,