Hugging Face's logo

qualcomm
/
Video-MAE

Video Classification
PyTorch
LiteRT
backbone
android
Model card Files Community
qaihm-bot commited on
Commit
e6c75a4
·
verified ·
1 Parent(s): 5e8f469

v0.34.0

Browse files

See https://github.com/quic/ai-hub-models/releases/v0.34.0 for changelog.

Files changed (3) hide show
  1. README.md +14 -23
  2. precompiled/qualcomm-snapdragon-x-elite/Video-MAE.onnx.zip +2 -2
  3. precompiled/qualcomm-snapdragon-x-elite/sdk_versions.yml +5 -0
README.md CHANGED
@@ -24,6 +24,7 @@ More details on model performance across various devices, can be found
24
  [here](https://aihub.qualcomm.com/models/video_mae).
25
 
26
 
 
27
  ### Model Details
28
 
29
  - **Model Type:** Model_use_case.video_classification
@@ -35,31 +36,31 @@ 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
- | Video-MAE | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 525.792 ms | 0 - 630 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
39
  | Video-MAE | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 1208.954 ms | 9 - 450 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
40
- | Video-MAE | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 281.207 ms | 0 - 596 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
41
  | Video-MAE | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 591.579 ms | 9 - 412 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
42
- | Video-MAE | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 199.253 ms | 0 - 38 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
43
  | Video-MAE | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 478.111 ms | 9 - 89 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
44
- | Video-MAE | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 223.631 ms | 0 - 629 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
45
  | Video-MAE | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 534.195 ms | 1 - 547 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
46
- | Video-MAE | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 525.792 ms | 0 - 630 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
47
  | Video-MAE | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 1208.954 ms | 9 - 450 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
48
- | Video-MAE | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 197.317 ms | 0 - 42 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
49
  | Video-MAE | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 478.825 ms | 9 - 50 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
50
- | Video-MAE | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 310.367 ms | 0 - 600 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
51
  | Video-MAE | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 630.153 ms | 5 - 464 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
52
- | Video-MAE | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 200.04 ms | 0 - 40 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
53
  | Video-MAE | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 479.622 ms | 10 - 88 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
54
- | Video-MAE | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 223.631 ms | 0 - 629 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
55
  | Video-MAE | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 534.195 ms | 1 - 547 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
56
- | Video-MAE | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 199.162 ms | 0 - 44 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
57
  | Video-MAE | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 479.976 ms | 9 - 84 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
58
  | Video-MAE | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 573.448 ms | 9 - 47 MB | NPU | [Video-MAE.onnx](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.onnx) |
59
- | Video-MAE | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 143.541 ms | 0 - 629 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
60
  | Video-MAE | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 365.895 ms | 9 - 450 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
61
  | Video-MAE | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 390.293 ms | 9 - 554 MB | NPU | [Video-MAE.onnx](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.onnx) |
62
- | Video-MAE | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 134.448 ms | 0 - 616 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
63
  | Video-MAE | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 335.714 ms | 9 - 497 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
64
  | Video-MAE | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 588.855 ms | 9 - 564 MB | NPU | [Video-MAE.onnx](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.onnx) |
65
  | Video-MAE | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 507.184 ms | 571 - 571 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
@@ -121,17 +122,7 @@ device. This script does the following:
121
  ```bash
122
  python -m qai_hub_models.models.video_mae.export
123
  ```
124
- ```
125
- Profiling Results
126
- ------------------------------------------------------------
127
- Video-MAE
128
- Device : cs_8275 (ANDROID 14)
129
- Runtime : TFLITE
130
- Estimated inference time (ms) : 525.8
131
- Estimated peak memory usage (MB): [0, 630]
132
- Total # Ops : 558
133
- Compute Unit(s) : npu (558 ops) gpu (0 ops) cpu (0 ops)
134
- ```
135
 
136
 
137
  ## How does this work?
 
24
  [here](https://aihub.qualcomm.com/models/video_mae).
25
 
26
 
27
+
28
  ### Model Details
29
 
30
  - **Model Type:** Model_use_case.video_classification
 
36
 
37
  | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
38
  |---|---|---|---|---|---|---|---|---|
39
+ | Video-MAE | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 525.556 ms | 0 - 629 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
40
  | Video-MAE | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 1208.954 ms | 9 - 450 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
41
+ | Video-MAE | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 277.573 ms | 0 - 593 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
42
  | Video-MAE | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 591.579 ms | 9 - 412 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
43
+ | Video-MAE | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 200.201 ms | 0 - 35 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
44
  | Video-MAE | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 478.111 ms | 9 - 89 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
45
+ | Video-MAE | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 223.838 ms | 0 - 630 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
46
  | Video-MAE | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 534.195 ms | 1 - 547 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
47
+ | Video-MAE | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 525.556 ms | 0 - 629 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
48
  | Video-MAE | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 1208.954 ms | 9 - 450 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
49
+ | Video-MAE | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 200.747 ms | 0 - 49 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
50
  | Video-MAE | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 478.825 ms | 9 - 50 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
51
+ | Video-MAE | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 310.362 ms | 0 - 600 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
52
  | Video-MAE | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 630.153 ms | 5 - 464 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
53
+ | Video-MAE | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 199.732 ms | 0 - 46 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
54
  | Video-MAE | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 479.622 ms | 10 - 88 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
55
+ | Video-MAE | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 223.838 ms | 0 - 630 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
56
  | Video-MAE | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 534.195 ms | 1 - 547 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
57
+ | Video-MAE | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 200.097 ms | 0 - 37 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
58
  | Video-MAE | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 479.976 ms | 9 - 84 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
59
  | Video-MAE | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 573.448 ms | 9 - 47 MB | NPU | [Video-MAE.onnx](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.onnx) |
60
+ | Video-MAE | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 144.599 ms | 0 - 630 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
61
  | Video-MAE | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 365.895 ms | 9 - 450 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
62
  | Video-MAE | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 390.293 ms | 9 - 554 MB | NPU | [Video-MAE.onnx](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.onnx) |
63
+ | Video-MAE | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 112.718 ms | 0 - 616 MB | NPU | [Video-MAE.tflite](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.tflite) |
64
  | Video-MAE | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 335.714 ms | 9 - 497 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
65
  | Video-MAE | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 588.855 ms | 9 - 564 MB | NPU | [Video-MAE.onnx](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.onnx) |
66
  | Video-MAE | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 507.184 ms | 571 - 571 MB | NPU | [Video-MAE.dlc](https://huggingface.co/qualcomm/Video-MAE/blob/main/Video-MAE.dlc) |
 
122
  ```bash
123
  python -m qai_hub_models.models.video_mae.export
124
  ```
125
+
 
 
 
 
 
 
 
 
 
 
126
 
127
 
128
  ## How does this work?
precompiled/qualcomm-snapdragon-x-elite/Video-MAE.onnx.zip CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:042090a463ef235ceda02d488d97eaa711c6afbf53c88d034fce259b0e2a37cd
3
- size 165419363
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:384e320ad2d176b813807b9b148b3a675b9bbdec6d0d7383cfe9f6806b06b7ea
3
+ size 165419380
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