Export the MobileNetV3 QNN variants as a8w8 (2.2x faster at the same mIoU)
Browse files- qnn/config.json +10 -10
- qnn/deeplab_v3_mobilenet_v3_large_qnn_a8w8_v69.pte +3 -0
- qnn/deeplab_v3_mobilenet_v3_large_qnn_a8w8_v73.pte +3 -0
- qnn/deeplab_v3_mobilenet_v3_large_qnn_a8w8_v75.pte +3 -0
- qnn/deeplab_v3_mobilenet_v3_large_qnn_a8w8_v79.pte +3 -0
- qnn/deeplab_v3_mobilenet_v3_large_qnn_a8w8_v81.pte +3 -0
qnn/config.json
CHANGED
|
@@ -9,8 +9,8 @@
|
|
| 9 |
"license": "bsd-3-clause",
|
| 10 |
"variants": [
|
| 11 |
{
|
| 12 |
-
"file": "
|
| 13 |
-
"precision": "
|
| 14 |
"quantized": true,
|
| 15 |
"default": false,
|
| 16 |
"size": "mobilenet_v3_large",
|
|
@@ -41,8 +41,8 @@
|
|
| 41 |
}
|
| 42 |
},
|
| 43 |
{
|
| 44 |
-
"file": "
|
| 45 |
-
"precision": "
|
| 46 |
"quantized": true,
|
| 47 |
"default": false,
|
| 48 |
"size": "mobilenet_v3_large",
|
|
@@ -73,8 +73,8 @@
|
|
| 73 |
}
|
| 74 |
},
|
| 75 |
{
|
| 76 |
-
"file": "
|
| 77 |
-
"precision": "
|
| 78 |
"quantized": true,
|
| 79 |
"default": false,
|
| 80 |
"size": "mobilenet_v3_large",
|
|
@@ -105,8 +105,8 @@
|
|
| 105 |
}
|
| 106 |
},
|
| 107 |
{
|
| 108 |
-
"file": "
|
| 109 |
-
"precision": "
|
| 110 |
"quantized": true,
|
| 111 |
"default": false,
|
| 112 |
"size": "mobilenet_v3_large",
|
|
@@ -137,8 +137,8 @@
|
|
| 137 |
}
|
| 138 |
},
|
| 139 |
{
|
| 140 |
-
"file": "
|
| 141 |
-
"precision": "
|
| 142 |
"quantized": true,
|
| 143 |
"default": true,
|
| 144 |
"size": "mobilenet_v3_large",
|
|
|
|
| 9 |
"license": "bsd-3-clause",
|
| 10 |
"variants": [
|
| 11 |
{
|
| 12 |
+
"file": "deeplab_v3_mobilenet_v3_large_qnn_a8w8_v69.pte",
|
| 13 |
+
"precision": "a8w8",
|
| 14 |
"quantized": true,
|
| 15 |
"default": false,
|
| 16 |
"size": "mobilenet_v3_large",
|
|
|
|
| 41 |
}
|
| 42 |
},
|
| 43 |
{
|
| 44 |
+
"file": "deeplab_v3_mobilenet_v3_large_qnn_a8w8_v73.pte",
|
| 45 |
+
"precision": "a8w8",
|
| 46 |
"quantized": true,
|
| 47 |
"default": false,
|
| 48 |
"size": "mobilenet_v3_large",
|
|
|
|
| 73 |
}
|
| 74 |
},
|
| 75 |
{
|
| 76 |
+
"file": "deeplab_v3_mobilenet_v3_large_qnn_a8w8_v75.pte",
|
| 77 |
+
"precision": "a8w8",
|
| 78 |
"quantized": true,
|
| 79 |
"default": false,
|
| 80 |
"size": "mobilenet_v3_large",
|
|
|
|
| 105 |
}
|
| 106 |
},
|
| 107 |
{
|
| 108 |
+
"file": "deeplab_v3_mobilenet_v3_large_qnn_a8w8_v79.pte",
|
| 109 |
+
"precision": "a8w8",
|
| 110 |
"quantized": true,
|
| 111 |
"default": false,
|
| 112 |
"size": "mobilenet_v3_large",
|
|
|
|
| 137 |
}
|
| 138 |
},
|
| 139 |
{
|
| 140 |
+
"file": "deeplab_v3_mobilenet_v3_large_qnn_a8w8_v81.pte",
|
| 141 |
+
"precision": "a8w8",
|
| 142 |
"quantized": true,
|
| 143 |
"default": true,
|
| 144 |
"size": "mobilenet_v3_large",
|
qnn/deeplab_v3_mobilenet_v3_large_qnn_a8w8_v69.pte
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ff8ac4a8092cac2bfec0f964c4efd93303fbc6299dac12dc66b0c52c7f46a014
|
| 3 |
+
size 14837248
|
qnn/deeplab_v3_mobilenet_v3_large_qnn_a8w8_v73.pte
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f2e9c4ea35699566077c5114483d559b386c0289faabeeca35d817ef0f925c42
|
| 3 |
+
size 12359168
|
qnn/deeplab_v3_mobilenet_v3_large_qnn_a8w8_v75.pte
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:06382cc2d7d587dc147ee0c65f6ddc040b37abf7a960920f016733fdf534bbac
|
| 3 |
+
size 12359168
|
qnn/deeplab_v3_mobilenet_v3_large_qnn_a8w8_v79.pte
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e4452866ec9e7d3d5eb1fe84f6f1d58bf0ec6fac7169e374cec41612ebded49b
|
| 3 |
+
size 12359168
|
qnn/deeplab_v3_mobilenet_v3_large_qnn_a8w8_v81.pte
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:393dae082d5674634dd118b67738770a5f84015e51b2547b435d010d0f147f58
|
| 3 |
+
size 12404224
|