Hugging Face's logo

qualcomm
/
DeepLabV3-ResNet50

Image Segmentation
PyTorch
LiteRT
android
Model card Files
xet
Community
qaihm-bot commited on
Commit
57add31
·
verified ·
1 Parent(s): a6b282c

v0.38.0

Browse files

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

Files changed (6) hide show
  1. DeepLabV3-ResNet50_w8a8.dlc +2 -2
  2. precompiled/qualcomm-snapdragon-x-elite/DeepLabV3-ResNet50_w8a8.bin → DeepLabV3-ResNet50_w8a8.tflite +2 -2
  3. README.md +15 -14
  4. precompiled/qualcomm-snapdragon-x-elite/DeepLabV3-ResNet50_w8a8.onnx.zip +0 -3
  5. precompiled/qualcomm-snapdragon-x-elite/tool-versions.yaml +0 -3
  6. tool-versions.yaml +3 -2
DeepLabV3-ResNet50_w8a8.dlc CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:606386276b04c040fdac2f24a986892f7cbfb0d3d8519f07d8208c971a98366e
3
- size 41458716
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5395744697d27cfb50fc992df427707f869f14ed84cadd46749c4f68c17f84ee
3
+ size 41458708
precompiled/qualcomm-snapdragon-x-elite/DeepLabV3-ResNet50_w8a8.bin → DeepLabV3-ResNet50_w8a8.tflite RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b6bd3e177e1e87b6b641054e41aa3a6730e550850b01831fb6437a6b213efb0a
3
- size 53858304
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28c3e66c0d9b950f69c8bfe2131b17cfdbd1300246849f429402027e57b18650
3
+ size 40457472
README.md CHANGED
@@ -36,19 +36,20 @@ 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
- | DeepLabV3-ResNet50 | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 441771.584 ms | 1 - 394 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
40
- | DeepLabV3-ResNet50 | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 430789.715 ms | 11 - 550 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
41
- | DeepLabV3-ResNet50 | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 261816.134 ms | 10 - 32 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
42
- | DeepLabV3-ResNet50 | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 363154.415 ms | 1 - 394 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
43
- | DeepLabV3-ResNet50 | w8a8 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 441771.584 ms | 1 - 394 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
44
- | DeepLabV3-ResNet50 | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 263004.139 ms | 1 - 21 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
45
- | DeepLabV3-ResNet50 | w8a8 | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 381864.98 ms | 1 - 393 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
46
- | DeepLabV3-ResNet50 | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 264272.795 ms | 0 - 24 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
47
- | DeepLabV3-ResNet50 | w8a8 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 363154.415 ms | 1 - 394 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
48
- | DeepLabV3-ResNet50 | w8a8 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 273100.243 ms | 0 - 24 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
49
- | DeepLabV3-ResNet50 | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 257100.118 ms | 1 - 509 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
50
- | DeepLabV3-ResNet50 | w8a8 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 309501.214 ms | 42 - 347 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
51
- | DeepLabV3-ResNet50 | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 352853.208 ms | 121 - 121 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
 
52
 
53
 
54
 
@@ -130,7 +131,7 @@ from qai_hub_models.models.deeplabv3_resnet50 import Model
130
  torch_model = Model.from_pretrained()
131
 
132
  # Device
133
- device = hub.Device("Samsung Galaxy S24")
134
 
135
  # Trace model
136
  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
+ | DeepLabV3-ResNet50 | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 441225.106 ms | 1 - 395 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
40
+ | DeepLabV3-ResNet50 | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 425275.28 ms | 13 - 565 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
41
+ | DeepLabV3-ResNet50 | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 264705.408 ms | 10 - 31 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
42
+ | DeepLabV3-ResNet50 | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 363242.695 ms | 1 - 390 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
43
+ | DeepLabV3-ResNet50 | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | TFLITE | 3276.866 ms | 0 - 106 MB | NPU | [DeepLabV3-ResNet50.tflite](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.tflite) |
44
+ | DeepLabV3-ResNet50 | w8a8 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 441225.106 ms | 1 - 395 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
45
+ | DeepLabV3-ResNet50 | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 260939.61 ms | 12 - 34 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
46
+ | DeepLabV3-ResNet50 | w8a8 | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 381873.4 ms | 1 - 394 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
47
+ | DeepLabV3-ResNet50 | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 261728.526 ms | 1 - 21 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
48
+ | DeepLabV3-ResNet50 | w8a8 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 363242.695 ms | 1 - 390 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
49
+ | DeepLabV3-ResNet50 | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 254186.213 ms | 1 - 515 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
50
+ | DeepLabV3-ResNet50 | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 262956.065 ms | 11 - 422 MB | NPU | [DeepLabV3-ResNet50.tflite](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.tflite) |
51
+ | DeepLabV3-ResNet50 | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 274680.73 ms | 17 - 324 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
52
+ | DeepLabV3-ResNet50 | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 358211.043 ms | 132 - 132 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
53
 
54
 
55
 
 
131
  torch_model = Model.from_pretrained()
132
 
133
  # Device
134
+ device = hub.Device("Samsung Galaxy S25")
135
 
136
  # Trace model
137
  input_shape = torch_model.get_input_spec()
precompiled/qualcomm-snapdragon-x-elite/DeepLabV3-ResNet50_w8a8.onnx.zip DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:4950e16b03785942a37b5032198ce6a3958d39df4f130005f2edd397611f9c37
3
- size 35956025
 
 
 
 
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
@@ -1,3 +1,4 @@
1
  tool_versions:
2
- qnn_dlc:
3
- qairt: 2.37.0.250724175447_124859
 
 
1
  tool_versions:
2
+ tflite:
3
+ qairt: 2.38.0.250901140452_125126
4
+ tflite: 2.17.0