v0.42.0
Browse filesSee https://github.com/quic/ai-hub-models/releases/v0.42.0 for changelog.
- DeepLabV3-ResNet50_w8a8.dlc +2 -2
- DeepLabV3-ResNet50_w8a8.onnx.zip +2 -2
- DeepLabV3-ResNet50_w8a8.tflite +1 -1
- README.md +27 -25
- tool-versions.yaml +3 -3
DeepLabV3-ResNet50_w8a8.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 sha256:2c4c2da6d61cb431e6e080b5ef61fd1167fe7791b7f8f1322fbccbea693e8ed3
|
| 3 |
+
size 41460228
|
DeepLabV3-ResNet50_w8a8.onnx.zip
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 sha256:3e63fabb498f182b05b73fbf4f761a9ea1ebec7b73bf19b8909413868ea94a7a
|
| 3 |
+
size 68085002
|
DeepLabV3-ResNet50_w8a8.tflite
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 40457472
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2ae545c4bf8ccf70d2c91d763824791da120889fa5967f31a1d9be8393e07730
|
| 3 |
size 40457472
|
README.md
CHANGED
|
@@ -36,27 +36,29 @@ 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 |
|
| 40 |
-
| DeepLabV3-ResNet50 | w8a8 |
|
| 41 |
-
| DeepLabV3-ResNet50 | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) |
|
| 42 |
-
| DeepLabV3-ResNet50 | w8a8 |
|
| 43 |
-
| DeepLabV3-ResNet50 | w8a8 |
|
| 44 |
-
| DeepLabV3-ResNet50 | w8a8 |
|
| 45 |
-
| DeepLabV3-ResNet50 | w8a8 |
|
| 46 |
-
| DeepLabV3-ResNet50 | w8a8 |
|
| 47 |
-
| DeepLabV3-ResNet50 | w8a8 |
|
| 48 |
-
| DeepLabV3-ResNet50 | w8a8 |
|
| 49 |
-
| DeepLabV3-ResNet50 | w8a8 |
|
| 50 |
-
| DeepLabV3-ResNet50 | w8a8 |
|
| 51 |
-
| DeepLabV3-ResNet50 | w8a8 |
|
| 52 |
-
| DeepLabV3-ResNet50 | w8a8 |
|
| 53 |
-
| DeepLabV3-ResNet50 | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile |
|
| 54 |
-
| DeepLabV3-ResNet50 | w8a8 | Samsung Galaxy
|
| 55 |
-
| DeepLabV3-ResNet50 | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile |
|
| 56 |
-
| DeepLabV3-ResNet50 | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile |
|
| 57 |
-
| DeepLabV3-ResNet50 | w8a8 |
|
| 58 |
-
| DeepLabV3-ResNet50 | w8a8 | Snapdragon
|
| 59 |
-
| DeepLabV3-ResNet50 | w8a8 | Snapdragon
|
|
|
|
|
|
|
| 60 |
|
| 61 |
|
| 62 |
|
|
@@ -70,9 +72,9 @@ pip install qai-hub-models
|
|
| 70 |
```
|
| 71 |
|
| 72 |
|
| 73 |
-
## Configure Qualcomm® AI Hub to run this model on a cloud-hosted device
|
| 74 |
|
| 75 |
-
Sign-in to [Qualcomm® AI Hub](https://
|
| 76 |
Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.
|
| 77 |
|
| 78 |
With this API token, you can configure your client to run models on the cloud
|
|
@@ -80,7 +82,7 @@ hosted devices.
|
|
| 80 |
```bash
|
| 81 |
qai-hub configure --api_token API_TOKEN
|
| 82 |
```
|
| 83 |
-
Navigate to [docs](https://
|
| 84 |
|
| 85 |
|
| 86 |
|
|
@@ -191,7 +193,7 @@ With the output of the model, you can compute like PSNR, relative errors or
|
|
| 191 |
spot check the output with expected output.
|
| 192 |
|
| 193 |
**Note**: This on-device profiling and inference requires access to Qualcomm®
|
| 194 |
-
AI Hub. [Sign up for access](https://myaccount.qualcomm.com/signup).
|
| 195 |
|
| 196 |
|
| 197 |
|
|
|
|
| 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 | 438001.182 ms | 1 - 359 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
|
| 40 |
+
| DeepLabV3-ResNet50 | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 260668.949 ms | 1 - 21 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) | ONNX | 260586.071 ms | 0 - 132 MB | NPU | [DeepLabV3-ResNet50.onnx.zip](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.onnx.zip) |
|
| 42 |
+
| DeepLabV3-ResNet50 | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 363685.021 ms | 5 - 367 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
|
| 43 |
+
| DeepLabV3-ResNet50 | w8a8 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | ONNX | 1212.149 ms | 60 - 76 MB | CPU | [DeepLabV3-ResNet50.onnx.zip](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.onnx.zip) |
|
| 44 |
+
| DeepLabV3-ResNet50 | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | TFLITE | 3262.651 ms | 28 - 96 MB | NPU | [DeepLabV3-ResNet50.tflite](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.tflite) |
|
| 45 |
+
| DeepLabV3-ResNet50 | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | ONNX | 959.412 ms | 51 - 99 MB | CPU | [DeepLabV3-ResNet50.onnx.zip](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.onnx.zip) |
|
| 46 |
+
| DeepLabV3-ResNet50 | w8a8 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 438001.182 ms | 1 - 359 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
|
| 47 |
+
| DeepLabV3-ResNet50 | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 263156.527 ms | 1 - 24 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
|
| 48 |
+
| DeepLabV3-ResNet50 | w8a8 | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 381288.119 ms | 1 - 360 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
|
| 49 |
+
| DeepLabV3-ResNet50 | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 260556.156 ms | 3 - 25 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
|
| 50 |
+
| DeepLabV3-ResNet50 | w8a8 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 363685.021 ms | 5 - 367 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
|
| 51 |
+
| DeepLabV3-ResNet50 | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 235766.818 ms | 0 - 415 MB | NPU | [DeepLabV3-ResNet50.tflite](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.tflite) |
|
| 52 |
+
| DeepLabV3-ResNet50 | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 255996.2 ms | 1 - 384 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
|
| 53 |
+
| DeepLabV3-ResNet50 | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 254083.359 ms | 1 - 464 MB | NPU | [DeepLabV3-ResNet50.onnx.zip](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.onnx.zip) |
|
| 54 |
+
| DeepLabV3-ResNet50 | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 264191.164 ms | 8 - 421 MB | NPU | [DeepLabV3-ResNet50.tflite](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.tflite) |
|
| 55 |
+
| DeepLabV3-ResNet50 | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 274193.398 ms | 19 - 324 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
|
| 56 |
+
| DeepLabV3-ResNet50 | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 279312.091 ms | 13 - 317 MB | NPU | [DeepLabV3-ResNet50.onnx.zip](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.onnx.zip) |
|
| 57 |
+
| DeepLabV3-ResNet50 | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | TFLITE | 989094.256 ms | 42 - 196 MB | NPU | [DeepLabV3-ResNet50.tflite](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.tflite) |
|
| 58 |
+
| DeepLabV3-ResNet50 | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | QNN_DLC | 958670.904 ms | 37 - 224 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
|
| 59 |
+
| DeepLabV3-ResNet50 | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | ONNX | 824.286 ms | 66 - 83 MB | CPU | [DeepLabV3-ResNet50.onnx.zip](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.onnx.zip) |
|
| 60 |
+
| DeepLabV3-ResNet50 | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 351933.012 ms | 26 - 26 MB | NPU | [DeepLabV3-ResNet50.dlc](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.dlc) |
|
| 61 |
+
| DeepLabV3-ResNet50 | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 364813.625 ms | 51 - 51 MB | NPU | [DeepLabV3-ResNet50.onnx.zip](https://huggingface.co/qualcomm/DeepLabV3-ResNet50/blob/main/DeepLabV3-ResNet50_w8a8.onnx.zip) |
|
| 62 |
|
| 63 |
|
| 64 |
|
|
|
|
| 72 |
```
|
| 73 |
|
| 74 |
|
| 75 |
+
## Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device
|
| 76 |
|
| 77 |
+
Sign-in to [Qualcomm® AI Hub Workbench](https://workbench.aihub.qualcomm.com/) with your
|
| 78 |
Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.
|
| 79 |
|
| 80 |
With this API token, you can configure your client to run models on the cloud
|
|
|
|
| 82 |
```bash
|
| 83 |
qai-hub configure --api_token API_TOKEN
|
| 84 |
```
|
| 85 |
+
Navigate to [docs](https://workbench.aihub.qualcomm.com/docs/) for more information.
|
| 86 |
|
| 87 |
|
| 88 |
|
|
|
|
| 193 |
spot check the output with expected output.
|
| 194 |
|
| 195 |
**Note**: This on-device profiling and inference requires access to Qualcomm®
|
| 196 |
+
AI Hub Workbench. [Sign up for access](https://myaccount.qualcomm.com/signup).
|
| 197 |
|
| 198 |
|
| 199 |
|
tool-versions.yaml
CHANGED
|
@@ -1,4 +1,4 @@
|
|
| 1 |
tool_versions:
|
| 2 |
-
|
| 3 |
-
qairt: 2.
|
| 4 |
-
|
|
|
|
| 1 |
tool_versions:
|
| 2 |
+
onnx:
|
| 3 |
+
qairt: 2.37.1.250807093845_124904
|
| 4 |
+
onnx_runtime: 1.23.0
|