qaihm-bot commited on
Commit
62d166b
·
verified ·
1 Parent(s): 8c3535f

See https://github.com/quic/ai-hub-models/releases/v0.38.0 for changelog.

Beit_float.dlc CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:754c983bf6a7e2f15133419ca5b10053d772db5de00adafca19f831aaf81f5a2
3
- size 368436740
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:77f55f8b04c8637a3b3d0d9f4c995257abfc3ac4b5217ceadda3fada7c641ab3
3
+ size 368436828
Beit_float.onnx.zip CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5ab7a8a67da713a8f81f5d338629f98842afb76c8d6fc94b144e84ebeb263eb4
3
- size 218716816
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:624b9be0da9e17eda7141271579301e3510a84e0f1cccbda0648e4658a6d79c8
3
+ size 218716832
Beit_float.tflite CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:cc940c9edc13baf359e303c2550304f2368f9384fcad40b1b6d7f1360ba60dc3
3
  size 368085312
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:74755fd840b99e7f4b6e1b9eba7f4964a5ec816c8db801ed26322f9f7c456915
3
  size 368085312
README.md CHANGED
@@ -36,36 +36,33 @@ More details on model performance across various devices, can be found
36
 
37
  | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
38
  |---|---|---|---|---|---|---|---|---|
39
- | Beit | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 43.044 ms | 0 - 309 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
40
- | Beit | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 44.94 ms | 0 - 329 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
41
- | Beit | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 17.621 ms | 0 - 319 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
42
- | Beit | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 26.766 ms | 1 - 323 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
43
- | Beit | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 11.668 ms | 0 - 16 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
44
- | Beit | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 13.897 ms | 1 - 30 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
45
- | Beit | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 13.732 ms | 1 - 24 MB | NPU | [Beit.onnx.zip](https://huggingface.co/qualcomm/Beit/blob/main/Beit.onnx.zip) |
46
- | Beit | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 15.116 ms | 0 - 309 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
47
- | Beit | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 17.096 ms | 0 - 326 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
48
- | Beit | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 43.044 ms | 0 - 309 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
49
- | Beit | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 44.94 ms | 0 - 329 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
50
- | Beit | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 12.079 ms | 0 - 15 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
51
- | Beit | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 13.833 ms | 0 - 30 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
52
- | Beit | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 19.892 ms | 0 - 304 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
53
- | Beit | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 21.644 ms | 1 - 318 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
54
- | Beit | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 11.731 ms | 0 - 15 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
55
- | Beit | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 13.842 ms | 1 - 28 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
56
- | Beit | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 15.116 ms | 0 - 309 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
57
- | Beit | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 17.096 ms | 0 - 326 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
58
- | Beit | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 12.078 ms | 0 - 15 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
59
- | Beit | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 13.867 ms | 0 - 32 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
60
- | Beit | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 13.652 ms | 1 - 28 MB | NPU | [Beit.onnx.zip](https://huggingface.co/qualcomm/Beit/blob/main/Beit.onnx.zip) |
61
- | Beit | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 8.207 ms | 0 - 325 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
62
- | Beit | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 9.516 ms | 1 - 345 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
63
- | Beit | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 9.423 ms | 0 - 335 MB | NPU | [Beit.onnx.zip](https://huggingface.co/qualcomm/Beit/blob/main/Beit.onnx.zip) |
64
- | Beit | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 6.933 ms | 0 - 314 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
65
- | Beit | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 7.272 ms | 1 - 321 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
66
- | Beit | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 10.293 ms | 0 - 318 MB | NPU | [Beit.onnx.zip](https://huggingface.co/qualcomm/Beit/blob/main/Beit.onnx.zip) |
67
- | Beit | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 14.465 ms | 1077 - 1077 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
68
- | Beit | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 14.046 ms | 186 - 186 MB | NPU | [Beit.onnx.zip](https://huggingface.co/qualcomm/Beit/blob/main/Beit.onnx.zip) |
69
 
70
 
71
 
@@ -147,7 +144,7 @@ from qai_hub_models.models.beit import Model
147
  torch_model = Model.from_pretrained()
148
 
149
  # Device
150
- device = hub.Device("Samsung Galaxy S24")
151
 
152
  # Trace model
153
  input_shape = torch_model.get_input_spec()
 
36
 
37
  | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
38
  |---|---|---|---|---|---|---|---|---|
39
+ | Beit | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 42.98 ms | 0 - 309 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
40
+ | Beit | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 45.095 ms | 1 - 336 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
41
+ | Beit | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 17.704 ms | 0 - 319 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
42
+ | Beit | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 25.631 ms | 1 - 319 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
43
+ | Beit | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 12.072 ms | 0 - 21 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
44
+ | Beit | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 14.055 ms | 0 - 33 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
45
+ | Beit | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 13.571 ms | 1 - 30 MB | NPU | [Beit.onnx.zip](https://huggingface.co/qualcomm/Beit/blob/main/Beit.onnx.zip) |
46
+ | Beit | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 15.052 ms | 0 - 309 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
47
+ | Beit | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 17.585 ms | 0 - 330 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
48
+ | Beit | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 42.98 ms | 0 - 309 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
49
+ | Beit | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 45.095 ms | 1 - 336 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
50
+ | Beit | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 12.137 ms | 0 - 15 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
51
+ | Beit | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 14.019 ms | 0 - 33 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
52
+ | Beit | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 19.933 ms | 0 - 305 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
53
+ | Beit | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 22.12 ms | 1 - 319 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
54
+ | Beit | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 12.175 ms | 0 - 14 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
55
+ | Beit | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 14.046 ms | 0 - 35 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
56
+ | Beit | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 15.052 ms | 0 - 309 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
57
+ | Beit | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 17.585 ms | 0 - 330 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
58
+ | Beit | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 8.169 ms | 0 - 324 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
59
+ | Beit | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 9.5 ms | 1 - 357 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
60
+ | Beit | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 9.856 ms | 0 - 348 MB | NPU | [Beit.onnx.zip](https://huggingface.co/qualcomm/Beit/blob/main/Beit.onnx.zip) |
61
+ | Beit | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 5.892 ms | 0 - 314 MB | NPU | [Beit.tflite](https://huggingface.co/qualcomm/Beit/blob/main/Beit.tflite) |
62
+ | Beit | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 7.244 ms | 0 - 322 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
63
+ | Beit | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 7.164 ms | 0 - 327 MB | NPU | [Beit.onnx.zip](https://huggingface.co/qualcomm/Beit/blob/main/Beit.onnx.zip) |
64
+ | Beit | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 14.667 ms | 1102 - 1102 MB | NPU | [Beit.dlc](https://huggingface.co/qualcomm/Beit/blob/main/Beit.dlc) |
65
+ | Beit | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 14.768 ms | 186 - 186 MB | NPU | [Beit.onnx.zip](https://huggingface.co/qualcomm/Beit/blob/main/Beit.onnx.zip) |
 
 
 
66
 
67
 
68
 
 
144
  torch_model = Model.from_pretrained()
145
 
146
  # Device
147
+ device = hub.Device("Samsung Galaxy S25")
148
 
149
  # Trace model
150
  input_shape = torch_model.get_input_spec()
precompiled/qualcomm-snapdragon-x-elite/Beit_float.bin DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:7c237ee76e58dbef2f7eba16bf23711c85fc1b09ed5a3fecb9eeda2bb06dd329
3
- size 194609152
 
 
 
 
precompiled/qualcomm-snapdragon-x-elite/Beit_float.onnx.zip DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:363f3b4b07ff2e890c82e7f8f2ce19b8f624b46037d639a3ad92486185fb3104
3
- size 164961344
 
 
 
 
precompiled/qualcomm-snapdragon-x-elite/tool-versions.yaml DELETED
@@ -1,3 +0,0 @@
1
- tool_versions:
2
- precompiled_qnn_onnx:
3
- qairt: 2.36.4.250725200057_123280
 
 
 
 
tool-versions.yaml CHANGED
@@ -1,4 +1,4 @@
1
  tool_versions:
2
  onnx:
3
- qairt: 2.36.4.250725200057_123280
4
- onnx_runtime: 1.22.0
 
1
  tool_versions:
2
  onnx:
3
+ qairt: 2.37.1.250807093845_124904
4
+ onnx_runtime: 1.22.2