v0.33.0
Browse filesSee https://github.com/quic/ai-hub-models/releases/v0.33.0 for changelog.
- LeViT.onnx → LeViT.onnx.zip +2 -2
- LeViT.tflite +2 -2
- LeViT_w8a16.dlc +2 -2
- LeViT_w8a16.onnx → LeViT_w8a16.onnx.zip +2 -2
- README.md +27 -27
- precompiled/qualcomm-snapdragon-x-elite/LeViT_w8a16.bin +3 -0
- precompiled/qualcomm-snapdragon-x-elite/LeViT_w8a16.onnx.zip +3 -0
- precompiled/qualcomm-snapdragon-x-elite/sdk_versions.yml +5 -0
LeViT.onnx → LeViT.onnx.zip
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:666ef7315e5bf4ebb4c86de1bcdb532e25c54dc02fdc39889045c39f8edd9630
|
3 |
+
size 27628284
|
LeViT.tflite
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:12cbfdc1c1ac15394c3ddc7ba12034d1ac0e47a724d8ab5515bd345ebeee5c5c
|
3 |
+
size 31342312
|
LeViT_w8a16.dlc
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:712ab65f2b25c83ab27e2ffef6e04120aea8e54da25d5bf80585801a071d914f
|
3 |
+
size 8617341
|
LeViT_w8a16.onnx → LeViT_w8a16.onnx.zip
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3c6f54db63ffb55509e09e0a2cb165f46643b572cc411b0e06487e2ef9711dc2
|
3 |
+
size 10508593
|
README.md
CHANGED
@@ -35,29 +35,29 @@ More details on model performance across various devices, can be found
|
|
35 |
|
36 |
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
|
37 |
|---|---|---|---|---|---|---|---|---|
|
38 |
-
| LeViT | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE |
|
39 |
-
| LeViT | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 1.
|
40 |
-
| LeViT | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 1.
|
41 |
-
| LeViT | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE |
|
42 |
-
| LeViT | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 1.
|
43 |
-
| LeViT | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 1.
|
44 |
-
| LeViT | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE |
|
45 |
-
| LeViT | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 1.
|
46 |
-
| LeViT | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE |
|
47 |
-
| LeViT | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 1.
|
48 |
-
| LeViT | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 1.
|
49 |
-
| LeViT | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC |
|
50 |
-
| LeViT | w8a16 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC |
|
51 |
-
| LeViT | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC |
|
52 |
-
| LeViT | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC |
|
53 |
-
| LeViT | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC |
|
54 |
-
| LeViT | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX |
|
55 |
-
| LeViT | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC |
|
56 |
-
| LeViT | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX |
|
57 |
-
| LeViT | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC |
|
58 |
-
| LeViT | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX |
|
59 |
-
| LeViT | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC |
|
60 |
-
| LeViT | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX |
|
61 |
|
62 |
|
63 |
|
@@ -121,10 +121,10 @@ Profiling Results
|
|
121 |
LeViT
|
122 |
Device : cs_8275 (ANDROID 14)
|
123 |
Runtime : TFLITE
|
124 |
-
Estimated inference time (ms) :
|
125 |
-
Estimated peak memory usage (MB): [0,
|
126 |
-
Total # Ops :
|
127 |
-
Compute Unit(s) : npu (
|
128 |
```
|
129 |
|
130 |
|
|
|
35 |
|
36 |
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
|
37 |
|---|---|---|---|---|---|---|---|---|
|
38 |
+
| LeViT | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 4.151 ms | 0 - 41 MB | NPU | [LeViT.tflite](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT.tflite) |
|
39 |
+
| LeViT | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 1.81 ms | 0 - 47 MB | NPU | [LeViT.tflite](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT.tflite) |
|
40 |
+
| LeViT | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 1.569 ms | 0 - 91 MB | NPU | [LeViT.tflite](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT.tflite) |
|
41 |
+
| LeViT | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 2.094 ms | 0 - 43 MB | NPU | [LeViT.tflite](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT.tflite) |
|
42 |
+
| LeViT | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 1.581 ms | 0 - 84 MB | NPU | [LeViT.tflite](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT.tflite) |
|
43 |
+
| LeViT | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 1.523 ms | 0 - 54 MB | NPU | [LeViT.onnx](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT.onnx) |
|
44 |
+
| LeViT | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 1.074 ms | 0 - 56 MB | NPU | [LeViT.tflite](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT.tflite) |
|
45 |
+
| LeViT | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 1.052 ms | 0 - 49 MB | NPU | [LeViT.onnx](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT.onnx) |
|
46 |
+
| LeViT | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 1.062 ms | 0 - 48 MB | NPU | [LeViT.tflite](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT.tflite) |
|
47 |
+
| LeViT | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 1.135 ms | 1 - 43 MB | NPU | [LeViT.onnx](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT.onnx) |
|
48 |
+
| LeViT | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 1.654 ms | 16 - 16 MB | NPU | [LeViT.onnx](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT.onnx) |
|
49 |
+
| LeViT | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 2.818 ms | 0 - 23 MB | NPU | [LeViT.dlc](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT_w8a16.dlc) |
|
50 |
+
| LeViT | w8a16 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 1.647 ms | 0 - 30 MB | NPU | [LeViT.dlc](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT_w8a16.dlc) |
|
51 |
+
| LeViT | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 1.403 ms | 0 - 10 MB | NPU | [LeViT.dlc](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT_w8a16.dlc) |
|
52 |
+
| LeViT | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 1.728 ms | 0 - 24 MB | NPU | [LeViT.dlc](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT_w8a16.dlc) |
|
53 |
+
| LeViT | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 1.396 ms | 0 - 10 MB | NPU | [LeViT.dlc](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT_w8a16.dlc) |
|
54 |
+
| LeViT | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 3.425 ms | 0 - 53 MB | NPU | [LeViT.onnx](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT_w8a16.onnx) |
|
55 |
+
| LeViT | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 0.958 ms | 0 - 35 MB | NPU | [LeViT.dlc](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT_w8a16.dlc) |
|
56 |
+
| LeViT | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 2.459 ms | 0 - 55 MB | NPU | [LeViT.onnx](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT_w8a16.onnx) |
|
57 |
+
| LeViT | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 0.805 ms | 0 - 24 MB | NPU | [LeViT.dlc](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT_w8a16.dlc) |
|
58 |
+
| LeViT | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 3.071 ms | 0 - 49 MB | NPU | [LeViT.onnx](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT_w8a16.onnx) |
|
59 |
+
| LeViT | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 1.727 ms | 6 - 6 MB | NPU | [LeViT.dlc](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT_w8a16.dlc) |
|
60 |
+
| LeViT | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 3.671 ms | 14 - 14 MB | NPU | [LeViT.onnx](https://huggingface.co/qualcomm/LeViT/blob/main/LeViT_w8a16.onnx) |
|
61 |
|
62 |
|
63 |
|
|
|
121 |
LeViT
|
122 |
Device : cs_8275 (ANDROID 14)
|
123 |
Runtime : TFLITE
|
124 |
+
Estimated inference time (ms) : 4.2
|
125 |
+
Estimated peak memory usage (MB): [0, 41]
|
126 |
+
Total # Ops : 280
|
127 |
+
Compute Unit(s) : npu (280 ops) gpu (0 ops) cpu (0 ops)
|
128 |
```
|
129 |
|
130 |
|
precompiled/qualcomm-snapdragon-x-elite/LeViT_w8a16.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:26c087ad4e8315dcd608aafb76e7c65c0c9b88ce45e7eb47ebb00589c6e9cbf1
|
3 |
+
size 9054312
|
precompiled/qualcomm-snapdragon-x-elite/LeViT_w8a16.onnx.zip
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:99b0850b4a6e681899d630c286856dc08feb1ccf431f78428bf7ac1adf74a699
|
3 |
+
size 5569350
|
precompiled/qualcomm-snapdragon-x-elite/sdk_versions.yml
ADDED
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
sdk_versions:
|
2 |
+
qnn_context_binary:
|
3 |
+
qairt: 2.34.2.250528164111_119506
|
4 |
+
precompiled_qnn_onnx:
|
5 |
+
qairt: 2.33.2.250410134701_117956
|