Hugging Face's logo

Datasets:
arm-bench
/
arm-bench-trace

Languages:
English
Size:
n<1K
Tags:
code
kernels
benchmark
arm
aarch64
sve
License:
Dataset card Files
xet
 Community
2
ArtificialRay7579 commited on
Commit
06f6695
·
verified ·
1 Parent(s): 20669e5

reset convolution kernel definition, workload, and solution; cin now is an axe of input shape

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. definitions/conv/conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c1024_c1024.json +0 -116
  2. definitions/conv/conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c128_c512.json +0 -116
  3. definitions/conv/conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c256_c1024.json +0 -116
  4. definitions/conv/conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c512_c2048.json +0 -116
  5. definitions/conv/conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c64_c256.json +0 -116
  6. definitions/conv/conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c64_c64.json +0 -116
  7. definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh1_dw1_c128_c256.json +0 -116
  8. definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh1_dw1_c256_c512.json +0 -116
  9. definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh1_dw1_c3_c64.json +0 -116
  10. definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh1_dw1_c512_c1024.json +0 -116
  11. definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh1_dw1_c64_c128.json +0 -116
  12. definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh2_dw2_c128_c256.json +0 -116
  13. definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh2_dw2_c256_c256.json +0 -116
  14. definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh2_dw2_c512_c512.json +0 -116
  15. definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh2_dw2_c64_c128.json +0 -116
  16. definitions/conv/conv2d_kh3_kw3_sh2_sw2_dh1_dw1_c128_c256.json +0 -116
  17. definitions/conv/conv2d_kh3_kw3_sh2_sw2_dh1_dw1_c256_c512.json +0 -116
  18. definitions/conv/conv2d_kh3_kw3_sh2_sw2_dh1_dw1_c512_c1024.json +0 -116
  19. definitions/conv/conv2d_kh3_kw3_sh2_sw2_dh1_dw1_c64_c128.json +0 -116
  20. definitions/conv/conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c128_c128.json +0 -116
  21. definitions/conv/conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c256_c256.json +0 -116
  22. definitions/conv/conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c32_c64.json +0 -116
  23. definitions/conv/conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c3_c32.json +0 -116
  24. definitions/conv/conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c512_c512.json +0 -116
  25. definitions/conv/conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c64_c128.json +0 -116
  26. definitions/conv/conv2d_kh7_kw7_sh2_sw2_dh1_dw1_c3_c64.json +0 -116
  27. definitions/conv1d/conv1d_kw1_sw1_dw1_p0_cin1024_cout1024.json +0 -78
  28. definitions/conv1d/conv1d_kw1_sw1_dw1_p0_cin128_cout512.json +0 -78
  29. definitions/conv1d/conv1d_kw1_sw1_dw1_p0_cin256_cout1024.json +0 -78
  30. definitions/conv1d/conv1d_kw1_sw1_dw1_p0_cin512_cout2048.json +0 -78
  31. definitions/conv1d/conv1d_kw1_sw1_dw1_p0_cin64_cout256.json +0 -78
  32. definitions/conv1d/conv1d_kw1_sw1_dw1_p0_cin64_cout64.json +0 -78
  33. definitions/conv1d/conv1d_kw3_sw1_dw1_p1_cin1024_cout1024.json +0 -78
  34. definitions/conv1d/conv1d_kw3_sw1_dw1_p1_cin128_cout256.json +0 -78
  35. definitions/conv1d/conv1d_kw3_sw1_dw1_p1_cin256_cout512.json +0 -78
  36. definitions/conv1d/conv1d_kw3_sw1_dw1_p1_cin3_cout64.json +0 -78
  37. definitions/conv1d/conv1d_kw3_sw1_dw1_p1_cin512_cout1024.json +0 -78
  38. definitions/conv1d/conv1d_kw3_sw1_dw1_p1_cin64_cout128.json +0 -78
  39. definitions/conv2d_depthwise/conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c1024.json +0 -100
  40. definitions/conv2d_depthwise/conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c128.json +0 -100
  41. definitions/conv2d_depthwise/conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c2048.json +0 -100
  42. definitions/conv2d_depthwise/conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c256.json +0 -100
  43. definitions/conv2d_depthwise/conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c32.json +0 -100
  44. definitions/conv2d_depthwise/conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c512.json +0 -100
  45. definitions/conv2d_depthwise/conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c64.json +0 -100
  46. definitions/conv2d_depthwise/conv2d_depthwise_kh5_kw5_sh1_sw1_dh1_dw1_p2_c1024.json +0 -100
  47. definitions/conv2d_depthwise/conv2d_depthwise_kh5_kw5_sh1_sw1_dh1_dw1_p2_c128.json +0 -100
  48. definitions/conv2d_depthwise/conv2d_depthwise_kh5_kw5_sh1_sw1_dh1_dw1_p2_c256.json +0 -100
  49. definitions/conv2d_depthwise/conv2d_depthwise_kh5_kw5_sh1_sw1_dh1_dw1_p2_c32.json +0 -100
  50. definitions/conv2d_depthwise/conv2d_depthwise_kh5_kw5_sh1_sw1_dh1_dw1_p2_c512.json +0 -100
definitions/conv/conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c1024_c1024.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c1024_c1024",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 1x1 kernel, stride (1,1), dilation (1,1), 1024->1024 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 1024
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 1024
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 1
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 1
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c128_c512.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c128_c512",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 1x1 kernel, stride (1,1), dilation (1,1), 128->512 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 128
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 512
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 1
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 1
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c256_c1024.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c256_c1024",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 1x1 kernel, stride (1,1), dilation (1,1), 256->1024 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 256
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 1024
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 1
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 1
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c512_c2048.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c512_c2048",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 1x1 kernel, stride (1,1), dilation (1,1), 512->2048 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 512
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 2048
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 1
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 1
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c64_c256.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c64_c256",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 1x1 kernel, stride (1,1), dilation (1,1), 64->256 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 64
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 256
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 1
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 1
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c64_c64.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh1_kw1_sh1_sw1_dh1_dw1_c64_c64",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 1x1 kernel, stride (1,1), dilation (1,1), 64->64 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 64
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 64
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 1
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 1
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh1_dw1_c128_c256.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh3_kw3_sh1_sw1_dh1_dw1_c128_c256",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 3x3 kernel, stride (1,1), dilation (1,1), 128->256 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 128
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 256
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 3
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 3
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh1_dw1_c256_c512.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh3_kw3_sh1_sw1_dh1_dw1_c256_c512",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 3x3 kernel, stride (1,1), dilation (1,1), 256->512 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 256
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 512
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 3
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 3
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh1_dw1_c3_c64.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh3_kw3_sh1_sw1_dh1_dw1_c3_c64",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 3x3 kernel, stride (1,1), dilation (1,1), 3->64 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 3
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 64
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 3
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 3
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh1_dw1_c512_c1024.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh3_kw3_sh1_sw1_dh1_dw1_c512_c1024",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 3x3 kernel, stride (1,1), dilation (1,1), 512->1024 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 512
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 1024
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 3
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 3
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh1_dw1_c64_c128.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh3_kw3_sh1_sw1_dh1_dw1_c64_c128",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 3x3 kernel, stride (1,1), dilation (1,1), 64->128 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 64
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 128
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 3
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 3
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh2_dw2_c128_c256.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh3_kw3_sh1_sw1_dh2_dw2_c128_c256",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 3x3 kernel, stride (1,1), dilation (2,2), 128->256 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 128
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 256
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 3
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 3
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 2
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 2
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(2, 2))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh2_dw2_c256_c256.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh3_kw3_sh1_sw1_dh2_dw2_c256_c256",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 3x3 kernel, stride (1,1), dilation (2,2), 256->256 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 256
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 256
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 3
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 3
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 2
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 2
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(2, 2))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh2_dw2_c512_c512.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh3_kw3_sh1_sw1_dh2_dw2_c512_c512",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 3x3 kernel, stride (1,1), dilation (2,2), 512->512 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 512
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 512
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 3
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 3
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 2
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 2
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(2, 2))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh3_kw3_sh1_sw1_dh2_dw2_c64_c128.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh3_kw3_sh1_sw1_dh2_dw2_c64_c128",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 3x3 kernel, stride (1,1), dilation (2,2), 64->128 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 64
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 128
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 3
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 3
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 2
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 2
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(2, 2))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh3_kw3_sh2_sw2_dh1_dw1_c128_c256.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh3_kw3_sh2_sw2_dh1_dw1_c128_c256",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 3x3 kernel, stride (2,2), dilation (1,1), 128->256 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 128
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 256
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 3
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 3
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 2
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 2
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(2, 2),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh3_kw3_sh2_sw2_dh1_dw1_c256_c512.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh3_kw3_sh2_sw2_dh1_dw1_c256_c512",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 3x3 kernel, stride (2,2), dilation (1,1), 256->512 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 256
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 512
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 3
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 3
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 2
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 2
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(2, 2),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh3_kw3_sh2_sw2_dh1_dw1_c512_c1024.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh3_kw3_sh2_sw2_dh1_dw1_c512_c1024",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 3x3 kernel, stride (2,2), dilation (1,1), 512->1024 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 512
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 1024
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 3
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 3
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 2
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 2
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(2, 2),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh3_kw3_sh2_sw2_dh1_dw1_c64_c128.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh3_kw3_sh2_sw2_dh1_dw1_c64_c128",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 3x3 kernel, stride (2,2), dilation (1,1), 64->128 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 64
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 128
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 3
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 3
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 2
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 2
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(2, 2),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c128_c128.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c128_c128",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 5x5 kernel, stride (1,1), dilation (1,1), 128->128 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 128
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 128
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 5
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 5
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c256_c256.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c256_c256",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 5x5 kernel, stride (1,1), dilation (1,1), 256->256 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 256
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 256
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 5
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 5
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c32_c64.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c32_c64",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 5x5 kernel, stride (1,1), dilation (1,1), 32->64 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 32
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 64
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 5
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 5
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c3_c32.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c3_c32",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 5x5 kernel, stride (1,1), dilation (1,1), 3->32 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 3
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 32
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 5
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 5
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c512_c512.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c512_c512",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 5x5 kernel, stride (1,1), dilation (1,1), 512->512 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 512
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 512
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 5
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 5
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c64_c128.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh5_kw5_sh1_sw1_dh1_dw1_c64_c128",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 5x5 kernel, stride (1,1), dilation (1,1), 64->128 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 64
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 128
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 5
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 5
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 1
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 1
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(1, 1),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv/conv2d_kh7_kw7_sh2_sw2_dh1_dw1_c3_c64.json DELETED
@@ -1,116 +0,0 @@
1
- {
2
- "name": "conv2d_kh7_kw7_sh2_sw2_dh1_dw1_c3_c64",
3
- "op_type": "conv2d",
4
- "description": "2D convolution: 7x7 kernel, stride (2,2), dilation (1,1), 3->64 channels. Extracted from tests/ncnn/candidate/convolution.cpp.",
5
- "tags": [
6
- "status:phase2",
7
- "isa:sve"
8
- ],
9
- "axes": {
10
- "N": {
11
- "type": "var"
12
- },
13
- "H": {
14
- "type": "var",
15
- "parent": "N"
16
- },
17
- "W": {
18
- "type": "var",
19
- "parent": "N"
20
- },
21
- "H_out": {
22
- "type": "var",
23
- "parent": "N",
24
- "description": "Derived: (H + 2*pad_h - (Kh-1)*Dh - 1) / Sh + 1"
25
- },
26
- "W_out": {
27
- "type": "var",
28
- "parent": "N",
29
- "description": "Derived: (W + 2*pad_w - (Kw-1)*Dw - 1) / Sw + 1"
30
- },
31
- "C_in": {
32
- "type": "const",
33
- "value": 3
34
- },
35
- "C_out": {
36
- "type": "const",
37
- "value": 64
38
- },
39
- "Kh": {
40
- "type": "const",
41
- "value": 7
42
- },
43
- "Kw": {
44
- "type": "const",
45
- "value": 7
46
- },
47
- "Sh": {
48
- "type": "const",
49
- "value": 2
50
- },
51
- "Sw": {
52
- "type": "const",
53
- "value": 2
54
- },
55
- "Dh": {
56
- "type": "const",
57
- "value": 1
58
- },
59
- "Dw": {
60
- "type": "const",
61
- "value": 1
62
- }
63
- },
64
- "inputs": {
65
- "input": {
66
- "shape": [
67
- "N",
68
- "C_in",
69
- "H",
70
- "W"
71
- ],
72
- "dtype": "float32"
73
- },
74
- "weight": {
75
- "shape": [
76
- "C_out",
77
- "C_in",
78
- "Kh",
79
- "Kw"
80
- ],
81
- "dtype": "float32"
82
- },
83
- "pad_top": {
84
- "shape": null,
85
- "dtype": "int32"
86
- },
87
- "pad_left": {
88
- "shape": null,
89
- "dtype": "int32"
90
- },
91
- "activation_type": {
92
- "shape": null,
93
- "dtype": "int32"
94
- },
95
- "with_bias": {
96
- "shape": null,
97
- "dtype": "int32"
98
- }
99
- },
100
- "outputs": {
101
- "output": {
102
- "shape": [
103
- "N",
104
- "C_out",
105
- "H_out",
106
- "W_out"
107
- ],
108
- "dtype": "float32"
109
- }
110
- },
111
- "constraints": [
112
- "H_out == (H + 2*pad_top - (Kh-1)*Dh - 1) / Sh + 1",
113
- "W_out == (W + 2*pad_left - (Kw-1)*Dw - 1) / Sw + 1"
114
- ],
115
- "reference": "import torch\nimport torch.nn.functional as F\n\n\ndef run(input, weight, pad_top, pad_left, activation_type, with_bias):\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight)\n y = F.conv2d(x, w, None,\n stride=(2, 2),\n padding=(int(pad_top), int(pad_left)),\n dilation=(1, 1))\n return y\n"
116
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv1d/conv1d_kw1_sw1_dw1_p0_cin1024_cout1024.json DELETED
@@ -1,78 +0,0 @@
1
- {
2
- "name": "conv1d_kw1_sw1_dw1_p0_cin1024_cout1024",
3
- "op_type": "conv1d",
4
- "description": "1D convolution: kw=1 stride=1 dilation=1 pad=0 C_in=1024 C_out=1024",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "W": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W_out": {
17
- "type": "var",
18
- "parent": "N",
19
- "description": "Derived: (W + 2*pad - dw*(kw-1) - 1) / sw + 1"
20
- },
21
- "C_in": {
22
- "type": "const",
23
- "value": 1024
24
- },
25
- "C_out": {
26
- "type": "const",
27
- "value": 1024
28
- },
29
- "Kw": {
30
- "type": "const",
31
- "value": 1
32
- },
33
- "Sw": {
34
- "type": "const",
35
- "value": 1
36
- },
37
- "Dw": {
38
- "type": "const",
39
- "value": 1
40
- }
41
- },
42
- "inputs": {
43
- "input": {
44
- "shape": [
45
- "C_in",
46
- "W"
47
- ],
48
- "dtype": "float32"
49
- },
50
- "weight": {
51
- "shape": [
52
- "C_out",
53
- "C_in",
54
- "Kw"
55
- ],
56
- "dtype": "float32"
57
- },
58
- "bias": {
59
- "shape": [
60
- "C_out"
61
- ],
62
- "dtype": "float32"
63
- }
64
- },
65
- "outputs": {
66
- "output": {
67
- "shape": [
68
- "C_out",
69
- "W_out"
70
- ],
71
- "dtype": "float32"
72
- }
73
- },
74
- "constraints": [
75
- "W_out == (W + 2*0 - 1*(Kw-1) - 1) // Sw + 1"
76
- ],
77
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n x = torch.from_numpy(input).unsqueeze(0)\n w = torch.from_numpy(weight)\n b = torch.from_numpy(bias)\n return F.conv1d(x, w, b, stride=1, padding=0, dilation=1).squeeze(0).numpy()\n"
78
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv1d/conv1d_kw1_sw1_dw1_p0_cin128_cout512.json DELETED
@@ -1,78 +0,0 @@
1
- {
2
- "name": "conv1d_kw1_sw1_dw1_p0_cin128_cout512",
3
- "op_type": "conv1d",
4
- "description": "1D convolution: kw=1 stride=1 dilation=1 pad=0 C_in=128 C_out=512",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "W": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W_out": {
17
- "type": "var",
18
- "parent": "N",
19
- "description": "Derived: (W + 2*pad - dw*(kw-1) - 1) / sw + 1"
20
- },
21
- "C_in": {
22
- "type": "const",
23
- "value": 128
24
- },
25
- "C_out": {
26
- "type": "const",
27
- "value": 512
28
- },
29
- "Kw": {
30
- "type": "const",
31
- "value": 1
32
- },
33
- "Sw": {
34
- "type": "const",
35
- "value": 1
36
- },
37
- "Dw": {
38
- "type": "const",
39
- "value": 1
40
- }
41
- },
42
- "inputs": {
43
- "input": {
44
- "shape": [
45
- "C_in",
46
- "W"
47
- ],
48
- "dtype": "float32"
49
- },
50
- "weight": {
51
- "shape": [
52
- "C_out",
53
- "C_in",
54
- "Kw"
55
- ],
56
- "dtype": "float32"
57
- },
58
- "bias": {
59
- "shape": [
60
- "C_out"
61
- ],
62
- "dtype": "float32"
63
- }
64
- },
65
- "outputs": {
66
- "output": {
67
- "shape": [
68
- "C_out",
69
- "W_out"
70
- ],
71
- "dtype": "float32"
72
- }
73
- },
74
- "constraints": [
75
- "W_out == (W + 2*0 - 1*(Kw-1) - 1) // Sw + 1"
76
- ],
77
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n x = torch.from_numpy(input).unsqueeze(0)\n w = torch.from_numpy(weight)\n b = torch.from_numpy(bias)\n return F.conv1d(x, w, b, stride=1, padding=0, dilation=1).squeeze(0).numpy()\n"
78
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv1d/conv1d_kw1_sw1_dw1_p0_cin256_cout1024.json DELETED
@@ -1,78 +0,0 @@
1
- {
2
- "name": "conv1d_kw1_sw1_dw1_p0_cin256_cout1024",
3
- "op_type": "conv1d",
4
- "description": "1D convolution: kw=1 stride=1 dilation=1 pad=0 C_in=256 C_out=1024",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "W": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W_out": {
17
- "type": "var",
18
- "parent": "N",
19
- "description": "Derived: (W + 2*pad - dw*(kw-1) - 1) / sw + 1"
20
- },
21
- "C_in": {
22
- "type": "const",
23
- "value": 256
24
- },
25
- "C_out": {
26
- "type": "const",
27
- "value": 1024
28
- },
29
- "Kw": {
30
- "type": "const",
31
- "value": 1
32
- },
33
- "Sw": {
34
- "type": "const",
35
- "value": 1
36
- },
37
- "Dw": {
38
- "type": "const",
39
- "value": 1
40
- }
41
- },
42
- "inputs": {
43
- "input": {
44
- "shape": [
45
- "C_in",
46
- "W"
47
- ],
48
- "dtype": "float32"
49
- },
50
- "weight": {
51
- "shape": [
52
- "C_out",
53
- "C_in",
54
- "Kw"
55
- ],
56
- "dtype": "float32"
57
- },
58
- "bias": {
59
- "shape": [
60
- "C_out"
61
- ],
62
- "dtype": "float32"
63
- }
64
- },
65
- "outputs": {
66
- "output": {
67
- "shape": [
68
- "C_out",
69
- "W_out"
70
- ],
71
- "dtype": "float32"
72
- }
73
- },
74
- "constraints": [
75
- "W_out == (W + 2*0 - 1*(Kw-1) - 1) // Sw + 1"
76
- ],
77
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n x = torch.from_numpy(input).unsqueeze(0)\n w = torch.from_numpy(weight)\n b = torch.from_numpy(bias)\n return F.conv1d(x, w, b, stride=1, padding=0, dilation=1).squeeze(0).numpy()\n"
78
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv1d/conv1d_kw1_sw1_dw1_p0_cin512_cout2048.json DELETED
@@ -1,78 +0,0 @@
1
- {
2
- "name": "conv1d_kw1_sw1_dw1_p0_cin512_cout2048",
3
- "op_type": "conv1d",
4
- "description": "1D convolution: kw=1 stride=1 dilation=1 pad=0 C_in=512 C_out=2048",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "W": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W_out": {
17
- "type": "var",
18
- "parent": "N",
19
- "description": "Derived: (W + 2*pad - dw*(kw-1) - 1) / sw + 1"
20
- },
21
- "C_in": {
22
- "type": "const",
23
- "value": 512
24
- },
25
- "C_out": {
26
- "type": "const",
27
- "value": 2048
28
- },
29
- "Kw": {
30
- "type": "const",
31
- "value": 1
32
- },
33
- "Sw": {
34
- "type": "const",
35
- "value": 1
36
- },
37
- "Dw": {
38
- "type": "const",
39
- "value": 1
40
- }
41
- },
42
- "inputs": {
43
- "input": {
44
- "shape": [
45
- "C_in",
46
- "W"
47
- ],
48
- "dtype": "float32"
49
- },
50
- "weight": {
51
- "shape": [
52
- "C_out",
53
- "C_in",
54
- "Kw"
55
- ],
56
- "dtype": "float32"
57
- },
58
- "bias": {
59
- "shape": [
60
- "C_out"
61
- ],
62
- "dtype": "float32"
63
- }
64
- },
65
- "outputs": {
66
- "output": {
67
- "shape": [
68
- "C_out",
69
- "W_out"
70
- ],
71
- "dtype": "float32"
72
- }
73
- },
74
- "constraints": [
75
- "W_out == (W + 2*0 - 1*(Kw-1) - 1) // Sw + 1"
76
- ],
77
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n x = torch.from_numpy(input).unsqueeze(0)\n w = torch.from_numpy(weight)\n b = torch.from_numpy(bias)\n return F.conv1d(x, w, b, stride=1, padding=0, dilation=1).squeeze(0).numpy()\n"
78
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv1d/conv1d_kw1_sw1_dw1_p0_cin64_cout256.json DELETED
@@ -1,78 +0,0 @@
1
- {
2
- "name": "conv1d_kw1_sw1_dw1_p0_cin64_cout256",
3
- "op_type": "conv1d",
4
- "description": "1D convolution: kw=1 stride=1 dilation=1 pad=0 C_in=64 C_out=256",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "W": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W_out": {
17
- "type": "var",
18
- "parent": "N",
19
- "description": "Derived: (W + 2*pad - dw*(kw-1) - 1) / sw + 1"
20
- },
21
- "C_in": {
22
- "type": "const",
23
- "value": 64
24
- },
25
- "C_out": {
26
- "type": "const",
27
- "value": 256
28
- },
29
- "Kw": {
30
- "type": "const",
31
- "value": 1
32
- },
33
- "Sw": {
34
- "type": "const",
35
- "value": 1
36
- },
37
- "Dw": {
38
- "type": "const",
39
- "value": 1
40
- }
41
- },
42
- "inputs": {
43
- "input": {
44
- "shape": [
45
- "C_in",
46
- "W"
47
- ],
48
- "dtype": "float32"
49
- },
50
- "weight": {
51
- "shape": [
52
- "C_out",
53
- "C_in",
54
- "Kw"
55
- ],
56
- "dtype": "float32"
57
- },
58
- "bias": {
59
- "shape": [
60
- "C_out"
61
- ],
62
- "dtype": "float32"
63
- }
64
- },
65
- "outputs": {
66
- "output": {
67
- "shape": [
68
- "C_out",
69
- "W_out"
70
- ],
71
- "dtype": "float32"
72
- }
73
- },
74
- "constraints": [
75
- "W_out == (W + 2*0 - 1*(Kw-1) - 1) // Sw + 1"
76
- ],
77
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n x = torch.from_numpy(input).unsqueeze(0)\n w = torch.from_numpy(weight)\n b = torch.from_numpy(bias)\n return F.conv1d(x, w, b, stride=1, padding=0, dilation=1).squeeze(0).numpy()\n"
78
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv1d/conv1d_kw1_sw1_dw1_p0_cin64_cout64.json DELETED
@@ -1,78 +0,0 @@
1
- {
2
- "name": "conv1d_kw1_sw1_dw1_p0_cin64_cout64",
3
- "op_type": "conv1d",
4
- "description": "1D convolution: kw=1 stride=1 dilation=1 pad=0 C_in=64 C_out=64",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "W": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W_out": {
17
- "type": "var",
18
- "parent": "N",
19
- "description": "Derived: (W + 2*pad - dw*(kw-1) - 1) / sw + 1"
20
- },
21
- "C_in": {
22
- "type": "const",
23
- "value": 64
24
- },
25
- "C_out": {
26
- "type": "const",
27
- "value": 64
28
- },
29
- "Kw": {
30
- "type": "const",
31
- "value": 1
32
- },
33
- "Sw": {
34
- "type": "const",
35
- "value": 1
36
- },
37
- "Dw": {
38
- "type": "const",
39
- "value": 1
40
- }
41
- },
42
- "inputs": {
43
- "input": {
44
- "shape": [
45
- "C_in",
46
- "W"
47
- ],
48
- "dtype": "float32"
49
- },
50
- "weight": {
51
- "shape": [
52
- "C_out",
53
- "C_in",
54
- "Kw"
55
- ],
56
- "dtype": "float32"
57
- },
58
- "bias": {
59
- "shape": [
60
- "C_out"
61
- ],
62
- "dtype": "float32"
63
- }
64
- },
65
- "outputs": {
66
- "output": {
67
- "shape": [
68
- "C_out",
69
- "W_out"
70
- ],
71
- "dtype": "float32"
72
- }
73
- },
74
- "constraints": [
75
- "W_out == (W + 2*0 - 1*(Kw-1) - 1) // Sw + 1"
76
- ],
77
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n x = torch.from_numpy(input).unsqueeze(0)\n w = torch.from_numpy(weight)\n b = torch.from_numpy(bias)\n return F.conv1d(x, w, b, stride=1, padding=0, dilation=1).squeeze(0).numpy()\n"
78
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv1d/conv1d_kw3_sw1_dw1_p1_cin1024_cout1024.json DELETED
@@ -1,78 +0,0 @@
1
- {
2
- "name": "conv1d_kw3_sw1_dw1_p1_cin1024_cout1024",
3
- "op_type": "conv1d",
4
- "description": "1D convolution: kw=3 stride=1 dilation=1 pad=1 C_in=1024 C_out=1024",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "W": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W_out": {
17
- "type": "var",
18
- "parent": "N",
19
- "description": "Derived: (W + 2*pad - dw*(kw-1) - 1) / sw + 1"
20
- },
21
- "C_in": {
22
- "type": "const",
23
- "value": 1024
24
- },
25
- "C_out": {
26
- "type": "const",
27
- "value": 1024
28
- },
29
- "Kw": {
30
- "type": "const",
31
- "value": 3
32
- },
33
- "Sw": {
34
- "type": "const",
35
- "value": 1
36
- },
37
- "Dw": {
38
- "type": "const",
39
- "value": 1
40
- }
41
- },
42
- "inputs": {
43
- "input": {
44
- "shape": [
45
- "C_in",
46
- "W"
47
- ],
48
- "dtype": "float32"
49
- },
50
- "weight": {
51
- "shape": [
52
- "C_out",
53
- "C_in",
54
- "Kw"
55
- ],
56
- "dtype": "float32"
57
- },
58
- "bias": {
59
- "shape": [
60
- "C_out"
61
- ],
62
- "dtype": "float32"
63
- }
64
- },
65
- "outputs": {
66
- "output": {
67
- "shape": [
68
- "C_out",
69
- "W_out"
70
- ],
71
- "dtype": "float32"
72
- }
73
- },
74
- "constraints": [
75
- "W_out == (W + 2*1 - 1*(Kw-1) - 1) // Sw + 1"
76
- ],
77
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n x = torch.from_numpy(input).unsqueeze(0)\n w = torch.from_numpy(weight)\n b = torch.from_numpy(bias)\n return F.conv1d(x, w, b, stride=1, padding=1, dilation=1).squeeze(0).numpy()\n"
78
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv1d/conv1d_kw3_sw1_dw1_p1_cin128_cout256.json DELETED
@@ -1,78 +0,0 @@
1
- {
2
- "name": "conv1d_kw3_sw1_dw1_p1_cin128_cout256",
3
- "op_type": "conv1d",
4
- "description": "1D convolution: kw=3 stride=1 dilation=1 pad=1 C_in=128 C_out=256",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "W": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W_out": {
17
- "type": "var",
18
- "parent": "N",
19
- "description": "Derived: (W + 2*pad - dw*(kw-1) - 1) / sw + 1"
20
- },
21
- "C_in": {
22
- "type": "const",
23
- "value": 128
24
- },
25
- "C_out": {
26
- "type": "const",
27
- "value": 256
28
- },
29
- "Kw": {
30
- "type": "const",
31
- "value": 3
32
- },
33
- "Sw": {
34
- "type": "const",
35
- "value": 1
36
- },
37
- "Dw": {
38
- "type": "const",
39
- "value": 1
40
- }
41
- },
42
- "inputs": {
43
- "input": {
44
- "shape": [
45
- "C_in",
46
- "W"
47
- ],
48
- "dtype": "float32"
49
- },
50
- "weight": {
51
- "shape": [
52
- "C_out",
53
- "C_in",
54
- "Kw"
55
- ],
56
- "dtype": "float32"
57
- },
58
- "bias": {
59
- "shape": [
60
- "C_out"
61
- ],
62
- "dtype": "float32"
63
- }
64
- },
65
- "outputs": {
66
- "output": {
67
- "shape": [
68
- "C_out",
69
- "W_out"
70
- ],
71
- "dtype": "float32"
72
- }
73
- },
74
- "constraints": [
75
- "W_out == (W + 2*1 - 1*(Kw-1) - 1) // Sw + 1"
76
- ],
77
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n x = torch.from_numpy(input).unsqueeze(0)\n w = torch.from_numpy(weight)\n b = torch.from_numpy(bias)\n return F.conv1d(x, w, b, stride=1, padding=1, dilation=1).squeeze(0).numpy()\n"
78
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv1d/conv1d_kw3_sw1_dw1_p1_cin256_cout512.json DELETED
@@ -1,78 +0,0 @@
1
- {
2
- "name": "conv1d_kw3_sw1_dw1_p1_cin256_cout512",
3
- "op_type": "conv1d",
4
- "description": "1D convolution: kw=3 stride=1 dilation=1 pad=1 C_in=256 C_out=512",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "W": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W_out": {
17
- "type": "var",
18
- "parent": "N",
19
- "description": "Derived: (W + 2*pad - dw*(kw-1) - 1) / sw + 1"
20
- },
21
- "C_in": {
22
- "type": "const",
23
- "value": 256
24
- },
25
- "C_out": {
26
- "type": "const",
27
- "value": 512
28
- },
29
- "Kw": {
30
- "type": "const",
31
- "value": 3
32
- },
33
- "Sw": {
34
- "type": "const",
35
- "value": 1
36
- },
37
- "Dw": {
38
- "type": "const",
39
- "value": 1
40
- }
41
- },
42
- "inputs": {
43
- "input": {
44
- "shape": [
45
- "C_in",
46
- "W"
47
- ],
48
- "dtype": "float32"
49
- },
50
- "weight": {
51
- "shape": [
52
- "C_out",
53
- "C_in",
54
- "Kw"
55
- ],
56
- "dtype": "float32"
57
- },
58
- "bias": {
59
- "shape": [
60
- "C_out"
61
- ],
62
- "dtype": "float32"
63
- }
64
- },
65
- "outputs": {
66
- "output": {
67
- "shape": [
68
- "C_out",
69
- "W_out"
70
- ],
71
- "dtype": "float32"
72
- }
73
- },
74
- "constraints": [
75
- "W_out == (W + 2*1 - 1*(Kw-1) - 1) // Sw + 1"
76
- ],
77
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n x = torch.from_numpy(input).unsqueeze(0)\n w = torch.from_numpy(weight)\n b = torch.from_numpy(bias)\n return F.conv1d(x, w, b, stride=1, padding=1, dilation=1).squeeze(0).numpy()\n"
78
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv1d/conv1d_kw3_sw1_dw1_p1_cin3_cout64.json DELETED
@@ -1,78 +0,0 @@
1
- {
2
- "name": "conv1d_kw3_sw1_dw1_p1_cin3_cout64",
3
- "op_type": "conv1d",
4
- "description": "1D convolution: kw=3 stride=1 dilation=1 pad=1 C_in=3 C_out=64",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "W": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W_out": {
17
- "type": "var",
18
- "parent": "N",
19
- "description": "Derived: (W + 2*pad - dw*(kw-1) - 1) / sw + 1"
20
- },
21
- "C_in": {
22
- "type": "const",
23
- "value": 3
24
- },
25
- "C_out": {
26
- "type": "const",
27
- "value": 64
28
- },
29
- "Kw": {
30
- "type": "const",
31
- "value": 3
32
- },
33
- "Sw": {
34
- "type": "const",
35
- "value": 1
36
- },
37
- "Dw": {
38
- "type": "const",
39
- "value": 1
40
- }
41
- },
42
- "inputs": {
43
- "input": {
44
- "shape": [
45
- "C_in",
46
- "W"
47
- ],
48
- "dtype": "float32"
49
- },
50
- "weight": {
51
- "shape": [
52
- "C_out",
53
- "C_in",
54
- "Kw"
55
- ],
56
- "dtype": "float32"
57
- },
58
- "bias": {
59
- "shape": [
60
- "C_out"
61
- ],
62
- "dtype": "float32"
63
- }
64
- },
65
- "outputs": {
66
- "output": {
67
- "shape": [
68
- "C_out",
69
- "W_out"
70
- ],
71
- "dtype": "float32"
72
- }
73
- },
74
- "constraints": [
75
- "W_out == (W + 2*1 - 1*(Kw-1) - 1) // Sw + 1"
76
- ],
77
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n x = torch.from_numpy(input).unsqueeze(0)\n w = torch.from_numpy(weight)\n b = torch.from_numpy(bias)\n return F.conv1d(x, w, b, stride=1, padding=1, dilation=1).squeeze(0).numpy()\n"
78
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv1d/conv1d_kw3_sw1_dw1_p1_cin512_cout1024.json DELETED
@@ -1,78 +0,0 @@
1
- {
2
- "name": "conv1d_kw3_sw1_dw1_p1_cin512_cout1024",
3
- "op_type": "conv1d",
4
- "description": "1D convolution: kw=3 stride=1 dilation=1 pad=1 C_in=512 C_out=1024",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "W": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W_out": {
17
- "type": "var",
18
- "parent": "N",
19
- "description": "Derived: (W + 2*pad - dw*(kw-1) - 1) / sw + 1"
20
- },
21
- "C_in": {
22
- "type": "const",
23
- "value": 512
24
- },
25
- "C_out": {
26
- "type": "const",
27
- "value": 1024
28
- },
29
- "Kw": {
30
- "type": "const",
31
- "value": 3
32
- },
33
- "Sw": {
34
- "type": "const",
35
- "value": 1
36
- },
37
- "Dw": {
38
- "type": "const",
39
- "value": 1
40
- }
41
- },
42
- "inputs": {
43
- "input": {
44
- "shape": [
45
- "C_in",
46
- "W"
47
- ],
48
- "dtype": "float32"
49
- },
50
- "weight": {
51
- "shape": [
52
- "C_out",
53
- "C_in",
54
- "Kw"
55
- ],
56
- "dtype": "float32"
57
- },
58
- "bias": {
59
- "shape": [
60
- "C_out"
61
- ],
62
- "dtype": "float32"
63
- }
64
- },
65
- "outputs": {
66
- "output": {
67
- "shape": [
68
- "C_out",
69
- "W_out"
70
- ],
71
- "dtype": "float32"
72
- }
73
- },
74
- "constraints": [
75
- "W_out == (W + 2*1 - 1*(Kw-1) - 1) // Sw + 1"
76
- ],
77
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n x = torch.from_numpy(input).unsqueeze(0)\n w = torch.from_numpy(weight)\n b = torch.from_numpy(bias)\n return F.conv1d(x, w, b, stride=1, padding=1, dilation=1).squeeze(0).numpy()\n"
78
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv1d/conv1d_kw3_sw1_dw1_p1_cin64_cout128.json DELETED
@@ -1,78 +0,0 @@
1
- {
2
- "name": "conv1d_kw3_sw1_dw1_p1_cin64_cout128",
3
- "op_type": "conv1d",
4
- "description": "1D convolution: kw=3 stride=1 dilation=1 pad=1 C_in=64 C_out=128",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "W": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W_out": {
17
- "type": "var",
18
- "parent": "N",
19
- "description": "Derived: (W + 2*pad - dw*(kw-1) - 1) / sw + 1"
20
- },
21
- "C_in": {
22
- "type": "const",
23
- "value": 64
24
- },
25
- "C_out": {
26
- "type": "const",
27
- "value": 128
28
- },
29
- "Kw": {
30
- "type": "const",
31
- "value": 3
32
- },
33
- "Sw": {
34
- "type": "const",
35
- "value": 1
36
- },
37
- "Dw": {
38
- "type": "const",
39
- "value": 1
40
- }
41
- },
42
- "inputs": {
43
- "input": {
44
- "shape": [
45
- "C_in",
46
- "W"
47
- ],
48
- "dtype": "float32"
49
- },
50
- "weight": {
51
- "shape": [
52
- "C_out",
53
- "C_in",
54
- "Kw"
55
- ],
56
- "dtype": "float32"
57
- },
58
- "bias": {
59
- "shape": [
60
- "C_out"
61
- ],
62
- "dtype": "float32"
63
- }
64
- },
65
- "outputs": {
66
- "output": {
67
- "shape": [
68
- "C_out",
69
- "W_out"
70
- ],
71
- "dtype": "float32"
72
- }
73
- },
74
- "constraints": [
75
- "W_out == (W + 2*1 - 1*(Kw-1) - 1) // Sw + 1"
76
- ],
77
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n x = torch.from_numpy(input).unsqueeze(0)\n w = torch.from_numpy(weight)\n b = torch.from_numpy(bias)\n return F.conv1d(x, w, b, stride=1, padding=1, dilation=1).squeeze(0).numpy()\n"
78
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv2d_depthwise/conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c1024.json DELETED
@@ -1,100 +0,0 @@
1
- {
2
- "name": "conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c1024",
3
- "op_type": "conv2d_depthwise",
4
- "description": "Depthwise 2D conv: kh=3 kw=3 stride=(1,1) dilation=(1,1) pad=(1,1) C=1024",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "H": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W": {
17
- "type": "var",
18
- "parent": "N"
19
- },
20
- "H_out": {
21
- "type": "var",
22
- "parent": "N",
23
- "description": "Derived: (H + 2*pad_top - dh*(kh-1) - 1) / sh + 1"
24
- },
25
- "W_out": {
26
- "type": "var",
27
- "parent": "N",
28
- "description": "Derived: (W + 2*pad_left - dw*(kw-1) - 1) / sw + 1"
29
- },
30
- "C": {
31
- "type": "const",
32
- "value": 1024
33
- },
34
- "Kh": {
35
- "type": "const",
36
- "value": 3
37
- },
38
- "Kw": {
39
- "type": "const",
40
- "value": 3
41
- },
42
- "Sh": {
43
- "type": "const",
44
- "value": 1
45
- },
46
- "Sw": {
47
- "type": "const",
48
- "value": 1
49
- },
50
- "Dh": {
51
- "type": "const",
52
- "value": 1
53
- },
54
- "Dw": {
55
- "type": "const",
56
- "value": 1
57
- }
58
- },
59
- "inputs": {
60
- "input": {
61
- "shape": [
62
- "N",
63
- "C",
64
- "H",
65
- "W"
66
- ],
67
- "dtype": "float32"
68
- },
69
- "weight": {
70
- "shape": [
71
- "C",
72
- "Kh",
73
- "Kw"
74
- ],
75
- "dtype": "float32"
76
- },
77
- "bias": {
78
- "shape": [
79
- "C"
80
- ],
81
- "dtype": "float32"
82
- }
83
- },
84
- "outputs": {
85
- "output": {
86
- "shape": [
87
- "N",
88
- "C",
89
- "H_out",
90
- "W_out"
91
- ],
92
- "dtype": "float32"
93
- }
94
- },
95
- "constraints": [
96
- "H_out == (H + 2*1 - 1*(Kh-1) - 1) // Sh + 1",
97
- "W_out == (W + 2*1 - 1*(Kw-1) - 1) // Sw + 1"
98
- ],
99
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n n, c = input.shape[0], input.shape[1]\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight).unsqueeze(1)\n b = torch.from_numpy(bias)\n return F.conv2d(x, w, b, stride=(1, 1), padding=(1, 1), dilation=(1, 1), groups=c).numpy()\n"
100
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv2d_depthwise/conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c128.json DELETED
@@ -1,100 +0,0 @@
1
- {
2
- "name": "conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c128",
3
- "op_type": "conv2d_depthwise",
4
- "description": "Depthwise 2D conv: kh=3 kw=3 stride=(1,1) dilation=(1,1) pad=(1,1) C=128",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "H": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W": {
17
- "type": "var",
18
- "parent": "N"
19
- },
20
- "H_out": {
21
- "type": "var",
22
- "parent": "N",
23
- "description": "Derived: (H + 2*pad_top - dh*(kh-1) - 1) / sh + 1"
24
- },
25
- "W_out": {
26
- "type": "var",
27
- "parent": "N",
28
- "description": "Derived: (W + 2*pad_left - dw*(kw-1) - 1) / sw + 1"
29
- },
30
- "C": {
31
- "type": "const",
32
- "value": 128
33
- },
34
- "Kh": {
35
- "type": "const",
36
- "value": 3
37
- },
38
- "Kw": {
39
- "type": "const",
40
- "value": 3
41
- },
42
- "Sh": {
43
- "type": "const",
44
- "value": 1
45
- },
46
- "Sw": {
47
- "type": "const",
48
- "value": 1
49
- },
50
- "Dh": {
51
- "type": "const",
52
- "value": 1
53
- },
54
- "Dw": {
55
- "type": "const",
56
- "value": 1
57
- }
58
- },
59
- "inputs": {
60
- "input": {
61
- "shape": [
62
- "N",
63
- "C",
64
- "H",
65
- "W"
66
- ],
67
- "dtype": "float32"
68
- },
69
- "weight": {
70
- "shape": [
71
- "C",
72
- "Kh",
73
- "Kw"
74
- ],
75
- "dtype": "float32"
76
- },
77
- "bias": {
78
- "shape": [
79
- "C"
80
- ],
81
- "dtype": "float32"
82
- }
83
- },
84
- "outputs": {
85
- "output": {
86
- "shape": [
87
- "N",
88
- "C",
89
- "H_out",
90
- "W_out"
91
- ],
92
- "dtype": "float32"
93
- }
94
- },
95
- "constraints": [
96
- "H_out == (H + 2*1 - 1*(Kh-1) - 1) // Sh + 1",
97
- "W_out == (W + 2*1 - 1*(Kw-1) - 1) // Sw + 1"
98
- ],
99
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n n, c = input.shape[0], input.shape[1]\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight).unsqueeze(1)\n b = torch.from_numpy(bias)\n return F.conv2d(x, w, b, stride=(1, 1), padding=(1, 1), dilation=(1, 1), groups=c).numpy()\n"
100
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv2d_depthwise/conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c2048.json DELETED
@@ -1,100 +0,0 @@
1
- {
2
- "name": "conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c2048",
3
- "op_type": "conv2d_depthwise",
4
- "description": "Depthwise 2D conv: kh=3 kw=3 stride=(1,1) dilation=(1,1) pad=(1,1) C=2048",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "H": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W": {
17
- "type": "var",
18
- "parent": "N"
19
- },
20
- "H_out": {
21
- "type": "var",
22
- "parent": "N",
23
- "description": "Derived: (H + 2*pad_top - dh*(kh-1) - 1) / sh + 1"
24
- },
25
- "W_out": {
26
- "type": "var",
27
- "parent": "N",
28
- "description": "Derived: (W + 2*pad_left - dw*(kw-1) - 1) / sw + 1"
29
- },
30
- "C": {
31
- "type": "const",
32
- "value": 2048
33
- },
34
- "Kh": {
35
- "type": "const",
36
- "value": 3
37
- },
38
- "Kw": {
39
- "type": "const",
40
- "value": 3
41
- },
42
- "Sh": {
43
- "type": "const",
44
- "value": 1
45
- },
46
- "Sw": {
47
- "type": "const",
48
- "value": 1
49
- },
50
- "Dh": {
51
- "type": "const",
52
- "value": 1
53
- },
54
- "Dw": {
55
- "type": "const",
56
- "value": 1
57
- }
58
- },
59
- "inputs": {
60
- "input": {
61
- "shape": [
62
- "N",
63
- "C",
64
- "H",
65
- "W"
66
- ],
67
- "dtype": "float32"
68
- },
69
- "weight": {
70
- "shape": [
71
- "C",
72
- "Kh",
73
- "Kw"
74
- ],
75
- "dtype": "float32"
76
- },
77
- "bias": {
78
- "shape": [
79
- "C"
80
- ],
81
- "dtype": "float32"
82
- }
83
- },
84
- "outputs": {
85
- "output": {
86
- "shape": [
87
- "N",
88
- "C",
89
- "H_out",
90
- "W_out"
91
- ],
92
- "dtype": "float32"
93
- }
94
- },
95
- "constraints": [
96
- "H_out == (H + 2*1 - 1*(Kh-1) - 1) // Sh + 1",
97
- "W_out == (W + 2*1 - 1*(Kw-1) - 1) // Sw + 1"
98
- ],
99
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n n, c = input.shape[0], input.shape[1]\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight).unsqueeze(1)\n b = torch.from_numpy(bias)\n return F.conv2d(x, w, b, stride=(1, 1), padding=(1, 1), dilation=(1, 1), groups=c).numpy()\n"
100
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv2d_depthwise/conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c256.json DELETED
@@ -1,100 +0,0 @@
1
- {
2
- "name": "conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c256",
3
- "op_type": "conv2d_depthwise",
4
- "description": "Depthwise 2D conv: kh=3 kw=3 stride=(1,1) dilation=(1,1) pad=(1,1) C=256",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "H": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W": {
17
- "type": "var",
18
- "parent": "N"
19
- },
20
- "H_out": {
21
- "type": "var",
22
- "parent": "N",
23
- "description": "Derived: (H + 2*pad_top - dh*(kh-1) - 1) / sh + 1"
24
- },
25
- "W_out": {
26
- "type": "var",
27
- "parent": "N",
28
- "description": "Derived: (W + 2*pad_left - dw*(kw-1) - 1) / sw + 1"
29
- },
30
- "C": {
31
- "type": "const",
32
- "value": 256
33
- },
34
- "Kh": {
35
- "type": "const",
36
- "value": 3
37
- },
38
- "Kw": {
39
- "type": "const",
40
- "value": 3
41
- },
42
- "Sh": {
43
- "type": "const",
44
- "value": 1
45
- },
46
- "Sw": {
47
- "type": "const",
48
- "value": 1
49
- },
50
- "Dh": {
51
- "type": "const",
52
- "value": 1
53
- },
54
- "Dw": {
55
- "type": "const",
56
- "value": 1
57
- }
58
- },
59
- "inputs": {
60
- "input": {
61
- "shape": [
62
- "N",
63
- "C",
64
- "H",
65
- "W"
66
- ],
67
- "dtype": "float32"
68
- },
69
- "weight": {
70
- "shape": [
71
- "C",
72
- "Kh",
73
- "Kw"
74
- ],
75
- "dtype": "float32"
76
- },
77
- "bias": {
78
- "shape": [
79
- "C"
80
- ],
81
- "dtype": "float32"
82
- }
83
- },
84
- "outputs": {
85
- "output": {
86
- "shape": [
87
- "N",
88
- "C",
89
- "H_out",
90
- "W_out"
91
- ],
92
- "dtype": "float32"
93
- }
94
- },
95
- "constraints": [
96
- "H_out == (H + 2*1 - 1*(Kh-1) - 1) // Sh + 1",
97
- "W_out == (W + 2*1 - 1*(Kw-1) - 1) // Sw + 1"
98
- ],
99
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n n, c = input.shape[0], input.shape[1]\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight).unsqueeze(1)\n b = torch.from_numpy(bias)\n return F.conv2d(x, w, b, stride=(1, 1), padding=(1, 1), dilation=(1, 1), groups=c).numpy()\n"
100
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv2d_depthwise/conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c32.json DELETED
@@ -1,100 +0,0 @@
1
- {
2
- "name": "conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c32",
3
- "op_type": "conv2d_depthwise",
4
- "description": "Depthwise 2D conv: kh=3 kw=3 stride=(1,1) dilation=(1,1) pad=(1,1) C=32",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "H": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W": {
17
- "type": "var",
18
- "parent": "N"
19
- },
20
- "H_out": {
21
- "type": "var",
22
- "parent": "N",
23
- "description": "Derived: (H + 2*pad_top - dh*(kh-1) - 1) / sh + 1"
24
- },
25
- "W_out": {
26
- "type": "var",
27
- "parent": "N",
28
- "description": "Derived: (W + 2*pad_left - dw*(kw-1) - 1) / sw + 1"
29
- },
30
- "C": {
31
- "type": "const",
32
- "value": 32
33
- },
34
- "Kh": {
35
- "type": "const",
36
- "value": 3
37
- },
38
- "Kw": {
39
- "type": "const",
40
- "value": 3
41
- },
42
- "Sh": {
43
- "type": "const",
44
- "value": 1
45
- },
46
- "Sw": {
47
- "type": "const",
48
- "value": 1
49
- },
50
- "Dh": {
51
- "type": "const",
52
- "value": 1
53
- },
54
- "Dw": {
55
- "type": "const",
56
- "value": 1
57
- }
58
- },
59
- "inputs": {
60
- "input": {
61
- "shape": [
62
- "N",
63
- "C",
64
- "H",
65
- "W"
66
- ],
67
- "dtype": "float32"
68
- },
69
- "weight": {
70
- "shape": [
71
- "C",
72
- "Kh",
73
- "Kw"
74
- ],
75
- "dtype": "float32"
76
- },
77
- "bias": {
78
- "shape": [
79
- "C"
80
- ],
81
- "dtype": "float32"
82
- }
83
- },
84
- "outputs": {
85
- "output": {
86
- "shape": [
87
- "N",
88
- "C",
89
- "H_out",
90
- "W_out"
91
- ],
92
- "dtype": "float32"
93
- }
94
- },
95
- "constraints": [
96
- "H_out == (H + 2*1 - 1*(Kh-1) - 1) // Sh + 1",
97
- "W_out == (W + 2*1 - 1*(Kw-1) - 1) // Sw + 1"
98
- ],
99
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n n, c = input.shape[0], input.shape[1]\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight).unsqueeze(1)\n b = torch.from_numpy(bias)\n return F.conv2d(x, w, b, stride=(1, 1), padding=(1, 1), dilation=(1, 1), groups=c).numpy()\n"
100
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv2d_depthwise/conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c512.json DELETED
@@ -1,100 +0,0 @@
1
- {
2
- "name": "conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c512",
3
- "op_type": "conv2d_depthwise",
4
- "description": "Depthwise 2D conv: kh=3 kw=3 stride=(1,1) dilation=(1,1) pad=(1,1) C=512",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "H": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W": {
17
- "type": "var",
18
- "parent": "N"
19
- },
20
- "H_out": {
21
- "type": "var",
22
- "parent": "N",
23
- "description": "Derived: (H + 2*pad_top - dh*(kh-1) - 1) / sh + 1"
24
- },
25
- "W_out": {
26
- "type": "var",
27
- "parent": "N",
28
- "description": "Derived: (W + 2*pad_left - dw*(kw-1) - 1) / sw + 1"
29
- },
30
- "C": {
31
- "type": "const",
32
- "value": 512
33
- },
34
- "Kh": {
35
- "type": "const",
36
- "value": 3
37
- },
38
- "Kw": {
39
- "type": "const",
40
- "value": 3
41
- },
42
- "Sh": {
43
- "type": "const",
44
- "value": 1
45
- },
46
- "Sw": {
47
- "type": "const",
48
- "value": 1
49
- },
50
- "Dh": {
51
- "type": "const",
52
- "value": 1
53
- },
54
- "Dw": {
55
- "type": "const",
56
- "value": 1
57
- }
58
- },
59
- "inputs": {
60
- "input": {
61
- "shape": [
62
- "N",
63
- "C",
64
- "H",
65
- "W"
66
- ],
67
- "dtype": "float32"
68
- },
69
- "weight": {
70
- "shape": [
71
- "C",
72
- "Kh",
73
- "Kw"
74
- ],
75
- "dtype": "float32"
76
- },
77
- "bias": {
78
- "shape": [
79
- "C"
80
- ],
81
- "dtype": "float32"
82
- }
83
- },
84
- "outputs": {
85
- "output": {
86
- "shape": [
87
- "N",
88
- "C",
89
- "H_out",
90
- "W_out"
91
- ],
92
- "dtype": "float32"
93
- }
94
- },
95
- "constraints": [
96
- "H_out == (H + 2*1 - 1*(Kh-1) - 1) // Sh + 1",
97
- "W_out == (W + 2*1 - 1*(Kw-1) - 1) // Sw + 1"
98
- ],
99
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n n, c = input.shape[0], input.shape[1]\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight).unsqueeze(1)\n b = torch.from_numpy(bias)\n return F.conv2d(x, w, b, stride=(1, 1), padding=(1, 1), dilation=(1, 1), groups=c).numpy()\n"
100
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv2d_depthwise/conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c64.json DELETED
@@ -1,100 +0,0 @@
1
- {
2
- "name": "conv2d_depthwise_kh3_kw3_sh1_sw1_dh1_dw1_p1_c64",
3
- "op_type": "conv2d_depthwise",
4
- "description": "Depthwise 2D conv: kh=3 kw=3 stride=(1,1) dilation=(1,1) pad=(1,1) C=64",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "H": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W": {
17
- "type": "var",
18
- "parent": "N"
19
- },
20
- "H_out": {
21
- "type": "var",
22
- "parent": "N",
23
- "description": "Derived: (H + 2*pad_top - dh*(kh-1) - 1) / sh + 1"
24
- },
25
- "W_out": {
26
- "type": "var",
27
- "parent": "N",
28
- "description": "Derived: (W + 2*pad_left - dw*(kw-1) - 1) / sw + 1"
29
- },
30
- "C": {
31
- "type": "const",
32
- "value": 64
33
- },
34
- "Kh": {
35
- "type": "const",
36
- "value": 3
37
- },
38
- "Kw": {
39
- "type": "const",
40
- "value": 3
41
- },
42
- "Sh": {
43
- "type": "const",
44
- "value": 1
45
- },
46
- "Sw": {
47
- "type": "const",
48
- "value": 1
49
- },
50
- "Dh": {
51
- "type": "const",
52
- "value": 1
53
- },
54
- "Dw": {
55
- "type": "const",
56
- "value": 1
57
- }
58
- },
59
- "inputs": {
60
- "input": {
61
- "shape": [
62
- "N",
63
- "C",
64
- "H",
65
- "W"
66
- ],
67
- "dtype": "float32"
68
- },
69
- "weight": {
70
- "shape": [
71
- "C",
72
- "Kh",
73
- "Kw"
74
- ],
75
- "dtype": "float32"
76
- },
77
- "bias": {
78
- "shape": [
79
- "C"
80
- ],
81
- "dtype": "float32"
82
- }
83
- },
84
- "outputs": {
85
- "output": {
86
- "shape": [
87
- "N",
88
- "C",
89
- "H_out",
90
- "W_out"
91
- ],
92
- "dtype": "float32"
93
- }
94
- },
95
- "constraints": [
96
- "H_out == (H + 2*1 - 1*(Kh-1) - 1) // Sh + 1",
97
- "W_out == (W + 2*1 - 1*(Kw-1) - 1) // Sw + 1"
98
- ],
99
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n n, c = input.shape[0], input.shape[1]\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight).unsqueeze(1)\n b = torch.from_numpy(bias)\n return F.conv2d(x, w, b, stride=(1, 1), padding=(1, 1), dilation=(1, 1), groups=c).numpy()\n"
100
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv2d_depthwise/conv2d_depthwise_kh5_kw5_sh1_sw1_dh1_dw1_p2_c1024.json DELETED
@@ -1,100 +0,0 @@
1
- {
2
- "name": "conv2d_depthwise_kh5_kw5_sh1_sw1_dh1_dw1_p2_c1024",
3
- "op_type": "conv2d_depthwise",
4
- "description": "Depthwise 2D conv: kh=5 kw=5 stride=(1,1) dilation=(1,1) pad=(2,2) C=1024",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "H": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W": {
17
- "type": "var",
18
- "parent": "N"
19
- },
20
- "H_out": {
21
- "type": "var",
22
- "parent": "N",
23
- "description": "Derived: (H + 2*pad_top - dh*(kh-1) - 1) / sh + 1"
24
- },
25
- "W_out": {
26
- "type": "var",
27
- "parent": "N",
28
- "description": "Derived: (W + 2*pad_left - dw*(kw-1) - 1) / sw + 1"
29
- },
30
- "C": {
31
- "type": "const",
32
- "value": 1024
33
- },
34
- "Kh": {
35
- "type": "const",
36
- "value": 5
37
- },
38
- "Kw": {
39
- "type": "const",
40
- "value": 5
41
- },
42
- "Sh": {
43
- "type": "const",
44
- "value": 1
45
- },
46
- "Sw": {
47
- "type": "const",
48
- "value": 1
49
- },
50
- "Dh": {
51
- "type": "const",
52
- "value": 1
53
- },
54
- "Dw": {
55
- "type": "const",
56
- "value": 1
57
- }
58
- },
59
- "inputs": {
60
- "input": {
61
- "shape": [
62
- "N",
63
- "C",
64
- "H",
65
- "W"
66
- ],
67
- "dtype": "float32"
68
- },
69
- "weight": {
70
- "shape": [
71
- "C",
72
- "Kh",
73
- "Kw"
74
- ],
75
- "dtype": "float32"
76
- },
77
- "bias": {
78
- "shape": [
79
- "C"
80
- ],
81
- "dtype": "float32"
82
- }
83
- },
84
- "outputs": {
85
- "output": {
86
- "shape": [
87
- "N",
88
- "C",
89
- "H_out",
90
- "W_out"
91
- ],
92
- "dtype": "float32"
93
- }
94
- },
95
- "constraints": [
96
- "H_out == (H + 2*2 - 1*(Kh-1) - 1) // Sh + 1",
97
- "W_out == (W + 2*2 - 1*(Kw-1) - 1) // Sw + 1"
98
- ],
99
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n n, c = input.shape[0], input.shape[1]\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight).unsqueeze(1)\n b = torch.from_numpy(bias)\n return F.conv2d(x, w, b, stride=(1, 1), padding=(2, 2), dilation=(1, 1), groups=c).numpy()\n"
100
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv2d_depthwise/conv2d_depthwise_kh5_kw5_sh1_sw1_dh1_dw1_p2_c128.json DELETED
@@ -1,100 +0,0 @@
1
- {
2
- "name": "conv2d_depthwise_kh5_kw5_sh1_sw1_dh1_dw1_p2_c128",
3
- "op_type": "conv2d_depthwise",
4
- "description": "Depthwise 2D conv: kh=5 kw=5 stride=(1,1) dilation=(1,1) pad=(2,2) C=128",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "H": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W": {
17
- "type": "var",
18
- "parent": "N"
19
- },
20
- "H_out": {
21
- "type": "var",
22
- "parent": "N",
23
- "description": "Derived: (H + 2*pad_top - dh*(kh-1) - 1) / sh + 1"
24
- },
25
- "W_out": {
26
- "type": "var",
27
- "parent": "N",
28
- "description": "Derived: (W + 2*pad_left - dw*(kw-1) - 1) / sw + 1"
29
- },
30
- "C": {
31
- "type": "const",
32
- "value": 128
33
- },
34
- "Kh": {
35
- "type": "const",
36
- "value": 5
37
- },
38
- "Kw": {
39
- "type": "const",
40
- "value": 5
41
- },
42
- "Sh": {
43
- "type": "const",
44
- "value": 1
45
- },
46
- "Sw": {
47
- "type": "const",
48
- "value": 1
49
- },
50
- "Dh": {
51
- "type": "const",
52
- "value": 1
53
- },
54
- "Dw": {
55
- "type": "const",
56
- "value": 1
57
- }
58
- },
59
- "inputs": {
60
- "input": {
61
- "shape": [
62
- "N",
63
- "C",
64
- "H",
65
- "W"
66
- ],
67
- "dtype": "float32"
68
- },
69
- "weight": {
70
- "shape": [
71
- "C",
72
- "Kh",
73
- "Kw"
74
- ],
75
- "dtype": "float32"
76
- },
77
- "bias": {
78
- "shape": [
79
- "C"
80
- ],
81
- "dtype": "float32"
82
- }
83
- },
84
- "outputs": {
85
- "output": {
86
- "shape": [
87
- "N",
88
- "C",
89
- "H_out",
90
- "W_out"
91
- ],
92
- "dtype": "float32"
93
- }
94
- },
95
- "constraints": [
96
- "H_out == (H + 2*2 - 1*(Kh-1) - 1) // Sh + 1",
97
- "W_out == (W + 2*2 - 1*(Kw-1) - 1) // Sw + 1"
98
- ],
99
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n n, c = input.shape[0], input.shape[1]\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight).unsqueeze(1)\n b = torch.from_numpy(bias)\n return F.conv2d(x, w, b, stride=(1, 1), padding=(2, 2), dilation=(1, 1), groups=c).numpy()\n"
100
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv2d_depthwise/conv2d_depthwise_kh5_kw5_sh1_sw1_dh1_dw1_p2_c256.json DELETED
@@ -1,100 +0,0 @@
1
- {
2
- "name": "conv2d_depthwise_kh5_kw5_sh1_sw1_dh1_dw1_p2_c256",
3
- "op_type": "conv2d_depthwise",
4
- "description": "Depthwise 2D conv: kh=5 kw=5 stride=(1,1) dilation=(1,1) pad=(2,2) C=256",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "H": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W": {
17
- "type": "var",
18
- "parent": "N"
19
- },
20
- "H_out": {
21
- "type": "var",
22
- "parent": "N",
23
- "description": "Derived: (H + 2*pad_top - dh*(kh-1) - 1) / sh + 1"
24
- },
25
- "W_out": {
26
- "type": "var",
27
- "parent": "N",
28
- "description": "Derived: (W + 2*pad_left - dw*(kw-1) - 1) / sw + 1"
29
- },
30
- "C": {
31
- "type": "const",
32
- "value": 256
33
- },
34
- "Kh": {
35
- "type": "const",
36
- "value": 5
37
- },
38
- "Kw": {
39
- "type": "const",
40
- "value": 5
41
- },
42
- "Sh": {
43
- "type": "const",
44
- "value": 1
45
- },
46
- "Sw": {
47
- "type": "const",
48
- "value": 1
49
- },
50
- "Dh": {
51
- "type": "const",
52
- "value": 1
53
- },
54
- "Dw": {
55
- "type": "const",
56
- "value": 1
57
- }
58
- },
59
- "inputs": {
60
- "input": {
61
- "shape": [
62
- "N",
63
- "C",
64
- "H",
65
- "W"
66
- ],
67
- "dtype": "float32"
68
- },
69
- "weight": {
70
- "shape": [
71
- "C",
72
- "Kh",
73
- "Kw"
74
- ],
75
- "dtype": "float32"
76
- },
77
- "bias": {
78
- "shape": [
79
- "C"
80
- ],
81
- "dtype": "float32"
82
- }
83
- },
84
- "outputs": {
85
- "output": {
86
- "shape": [
87
- "N",
88
- "C",
89
- "H_out",
90
- "W_out"
91
- ],
92
- "dtype": "float32"
93
- }
94
- },
95
- "constraints": [
96
- "H_out == (H + 2*2 - 1*(Kh-1) - 1) // Sh + 1",
97
- "W_out == (W + 2*2 - 1*(Kw-1) - 1) // Sw + 1"
98
- ],
99
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n n, c = input.shape[0], input.shape[1]\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight).unsqueeze(1)\n b = torch.from_numpy(bias)\n return F.conv2d(x, w, b, stride=(1, 1), padding=(2, 2), dilation=(1, 1), groups=c).numpy()\n"
100
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv2d_depthwise/conv2d_depthwise_kh5_kw5_sh1_sw1_dh1_dw1_p2_c32.json DELETED
@@ -1,100 +0,0 @@
1
- {
2
- "name": "conv2d_depthwise_kh5_kw5_sh1_sw1_dh1_dw1_p2_c32",
3
- "op_type": "conv2d_depthwise",
4
- "description": "Depthwise 2D conv: kh=5 kw=5 stride=(1,1) dilation=(1,1) pad=(2,2) C=32",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "H": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W": {
17
- "type": "var",
18
- "parent": "N"
19
- },
20
- "H_out": {
21
- "type": "var",
22
- "parent": "N",
23
- "description": "Derived: (H + 2*pad_top - dh*(kh-1) - 1) / sh + 1"
24
- },
25
- "W_out": {
26
- "type": "var",
27
- "parent": "N",
28
- "description": "Derived: (W + 2*pad_left - dw*(kw-1) - 1) / sw + 1"
29
- },
30
- "C": {
31
- "type": "const",
32
- "value": 32
33
- },
34
- "Kh": {
35
- "type": "const",
36
- "value": 5
37
- },
38
- "Kw": {
39
- "type": "const",
40
- "value": 5
41
- },
42
- "Sh": {
43
- "type": "const",
44
- "value": 1
45
- },
46
- "Sw": {
47
- "type": "const",
48
- "value": 1
49
- },
50
- "Dh": {
51
- "type": "const",
52
- "value": 1
53
- },
54
- "Dw": {
55
- "type": "const",
56
- "value": 1
57
- }
58
- },
59
- "inputs": {
60
- "input": {
61
- "shape": [
62
- "N",
63
- "C",
64
- "H",
65
- "W"
66
- ],
67
- "dtype": "float32"
68
- },
69
- "weight": {
70
- "shape": [
71
- "C",
72
- "Kh",
73
- "Kw"
74
- ],
75
- "dtype": "float32"
76
- },
77
- "bias": {
78
- "shape": [
79
- "C"
80
- ],
81
- "dtype": "float32"
82
- }
83
- },
84
- "outputs": {
85
- "output": {
86
- "shape": [
87
- "N",
88
- "C",
89
- "H_out",
90
- "W_out"
91
- ],
92
- "dtype": "float32"
93
- }
94
- },
95
- "constraints": [
96
- "H_out == (H + 2*2 - 1*(Kh-1) - 1) // Sh + 1",
97
- "W_out == (W + 2*2 - 1*(Kw-1) - 1) // Sw + 1"
98
- ],
99
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n n, c = input.shape[0], input.shape[1]\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight).unsqueeze(1)\n b = torch.from_numpy(bias)\n return F.conv2d(x, w, b, stride=(1, 1), padding=(2, 2), dilation=(1, 1), groups=c).numpy()\n"
100
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
definitions/conv2d_depthwise/conv2d_depthwise_kh5_kw5_sh1_sw1_dh1_dw1_p2_c512.json DELETED
@@ -1,100 +0,0 @@
1
- {
2
- "name": "conv2d_depthwise_kh5_kw5_sh1_sw1_dh1_dw1_p2_c512",
3
- "op_type": "conv2d_depthwise",
4
- "description": "Depthwise 2D conv: kh=5 kw=5 stride=(1,1) dilation=(1,1) pad=(2,2) C=512",
5
- "tags": [
6
- "status:draft"
7
- ],
8
- "axes": {
9
- "N": {
10
- "type": "var"
11
- },
12
- "H": {
13
- "type": "var",
14
- "parent": "N"
15
- },
16
- "W": {
17
- "type": "var",
18
- "parent": "N"
19
- },
20
- "H_out": {
21
- "type": "var",
22
- "parent": "N",
23
- "description": "Derived: (H + 2*pad_top - dh*(kh-1) - 1) / sh + 1"
24
- },
25
- "W_out": {
26
- "type": "var",
27
- "parent": "N",
28
- "description": "Derived: (W + 2*pad_left - dw*(kw-1) - 1) / sw + 1"
29
- },
30
- "C": {
31
- "type": "const",
32
- "value": 512
33
- },
34
- "Kh": {
35
- "type": "const",
36
- "value": 5
37
- },
38
- "Kw": {
39
- "type": "const",
40
- "value": 5
41
- },
42
- "Sh": {
43
- "type": "const",
44
- "value": 1
45
- },
46
- "Sw": {
47
- "type": "const",
48
- "value": 1
49
- },
50
- "Dh": {
51
- "type": "const",
52
- "value": 1
53
- },
54
- "Dw": {
55
- "type": "const",
56
- "value": 1
57
- }
58
- },
59
- "inputs": {
60
- "input": {
61
- "shape": [
62
- "N",
63
- "C",
64
- "H",
65
- "W"
66
- ],
67
- "dtype": "float32"
68
- },
69
- "weight": {
70
- "shape": [
71
- "C",
72
- "Kh",
73
- "Kw"
74
- ],
75
- "dtype": "float32"
76
- },
77
- "bias": {
78
- "shape": [
79
- "C"
80
- ],
81
- "dtype": "float32"
82
- }
83
- },
84
- "outputs": {
85
- "output": {
86
- "shape": [
87
- "N",
88
- "C",
89
- "H_out",
90
- "W_out"
91
- ],
92
- "dtype": "float32"
93
- }
94
- },
95
- "constraints": [
96
- "H_out == (H + 2*2 - 1*(Kh-1) - 1) // Sh + 1",
97
- "W_out == (W + 2*2 - 1*(Kw-1) - 1) // Sw + 1"
98
- ],
99
- "reference": "import torch\nimport torch.nn.functional as F\n\ndef run(input, weight, bias):\n n, c = input.shape[0], input.shape[1]\n x = torch.from_numpy(input)\n w = torch.from_numpy(weight).unsqueeze(1)\n b = torch.from_numpy(bias)\n return F.conv2d(x, w, b, stride=(1, 1), padding=(2, 2), dilation=(1, 1), groups=c).numpy()\n"
100
- }