Hugging Face's logo

qualcomm
/
OpenAI-Clip

Image Classification
PyTorch
foundation
android
Model card Files
xet
 Community
1
qaihm-bot commited on
Commit
90ee4a0
·
verified ·
1 Parent(s): 94a0719

v0.46.1

Browse files

See https://github.com/quic/ai-hub-models/releases/v0.46.1 for changelog.

Files changed (5) hide show
  1. OpenAI-Clip_float.dlc +0 -3
  2. OpenAI-Clip_float.onnx.zip +0 -3
  3. OpenAI-Clip_float.tflite +0 -3
  4. README.md +74 -213
  5. tool-versions.yaml +0 -4
OpenAI-Clip_float.dlc DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:b9d43080f00ad5214d7268b1a90ea2313a02f33e99082fd5e156b09217fc0bbb
3
- size 599439284
 
 
 
 
OpenAI-Clip_float.onnx.zip DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:bbd9cf0f9a317b3400283b9dab145a843e4a7ba4bf7396c3a273d0073ccf49ef
3
- size 406345322
 
 
 
 
OpenAI-Clip_float.tflite DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:4cc191e297a62574c8f713be4324796c79da5a6ea859c1754e962a84d0e5959a
3
- size 598745484
 
 
 
 
README.md CHANGED
@@ -10,231 +10,92 @@ pipeline_tag: image-classification
10
 
11
  ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/openai_clip/web-assets/model_demo.png)
12
 
13
- # OpenAI-Clip: Optimized for Mobile Deployment
14
- ## Multi-modal foundational model for vision and language tasks like image/text similarity and for zero-shot image classification
15
-
16
 
17
  Contrastive Language-Image Pre-Training (CLIP) uses a ViT like transformer to get visual features and a causal language model to get the text features. Both the text and visual features can then be used for a variety of zero-shot learning tasks.
18
 
19
- This model is an implementation of OpenAI-Clip found [here](https://github.com/openai/CLIP/).
20
-
21
-
22
- This repository provides scripts to run OpenAI-Clip on Qualcomm® devices.
23
- More details on model performance across various devices, can be found
24
- [here](https://aihub.qualcomm.com/models/openai_clip).
25
-
26
-
27
-
28
- ### Model Details
29
-
30
- - **Model Type:** Model_use_case.image_classification
31
- - **Model Stats:**
32
- - Model checkpoint: ViT-B/16
33
- - Image input resolution: 224x224
34
- - Text context length: 77
35
- - Number of parameters: 150M
36
- - Model size (float): 571 MB
37
-
38
- | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
39
- |---|---|---|---|---|---|---|---|---|
40
- | OpenAI-Clip | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 55.909 ms | 0 - 764 MB | NPU | [OpenAI-Clip.tflite](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.tflite) |
41
- | OpenAI-Clip | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 59.966 ms | 1 - 767 MB | NPU | [OpenAI-Clip.dlc](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.dlc) |
42
- | OpenAI-Clip | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 24.443 ms | 0 - 748 MB | NPU | [OpenAI-Clip.tflite](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.tflite) |
43
- | OpenAI-Clip | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 25.861 ms | 0 - 752 MB | NPU | [OpenAI-Clip.dlc](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.dlc) |
44
- | OpenAI-Clip | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 18.45 ms | 0 - 3 MB | NPU | [OpenAI-Clip.tflite](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.tflite) |
45
- | OpenAI-Clip | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 20.95 ms | 1 - 3 MB | NPU | [OpenAI-Clip.dlc](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.dlc) |
46
- | OpenAI-Clip | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 22.179 ms | 0 - 322 MB | NPU | [OpenAI-Clip.onnx.zip](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.onnx.zip) |
47
- | OpenAI-Clip | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 21.813 ms | 0 - 764 MB | NPU | [OpenAI-Clip.tflite](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.tflite) |
48
- | OpenAI-Clip | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 23.858 ms | 1 - 767 MB | NPU | [OpenAI-Clip.dlc](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.dlc) |
49
- | OpenAI-Clip | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 55.909 ms | 0 - 764 MB | NPU | [OpenAI-Clip.tflite](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.tflite) |
50
- | OpenAI-Clip | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 59.966 ms | 1 - 767 MB | NPU | [OpenAI-Clip.dlc](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.dlc) |
51
- | OpenAI-Clip | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 24.923 ms | 0 - 751 MB | NPU | [OpenAI-Clip.tflite](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.tflite) |
52
- | OpenAI-Clip | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 26.378 ms | 1 - 753 MB | NPU | [OpenAI-Clip.dlc](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.dlc) |
53
- | OpenAI-Clip | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 21.813 ms | 0 - 764 MB | NPU | [OpenAI-Clip.tflite](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.tflite) |
54
- | OpenAI-Clip | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 23.858 ms | 1 - 767 MB | NPU | [OpenAI-Clip.dlc](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.dlc) |
55
- | OpenAI-Clip | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 13.364 ms | 0 - 820 MB | NPU | [OpenAI-Clip.tflite](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.tflite) |
56
- | OpenAI-Clip | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 15.095 ms | 0 - 821 MB | NPU | [OpenAI-Clip.dlc](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.dlc) |
57
- | OpenAI-Clip | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 15.584 ms | 1 - 799 MB | NPU | [OpenAI-Clip.onnx.zip](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.onnx.zip) |
58
- | OpenAI-Clip | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 10.192 ms | 0 - 729 MB | NPU | [OpenAI-Clip.tflite](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.tflite) |
59
- | OpenAI-Clip | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 11.404 ms | 0 - 731 MB | NPU | [OpenAI-Clip.dlc](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.dlc) |
60
- | OpenAI-Clip | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 12.271 ms | 1 - 712 MB | NPU | [OpenAI-Clip.onnx.zip](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.onnx.zip) |
61
- | OpenAI-Clip | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | TFLITE | 7.878 ms | 0 - 680 MB | NPU | [OpenAI-Clip.tflite](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.tflite) |
62
- | OpenAI-Clip | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | QNN_DLC | 9.308 ms | 1 - 679 MB | NPU | [OpenAI-Clip.dlc](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.dlc) |
63
- | OpenAI-Clip | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | ONNX | 10.064 ms | 1 - 661 MB | NPU | [OpenAI-Clip.onnx.zip](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.onnx.zip) |
64
- | OpenAI-Clip | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 21.992 ms | 1 - 1 MB | NPU | [OpenAI-Clip.dlc](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.dlc) |
65
- | OpenAI-Clip | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 22.565 ms | 294 - 294 MB | NPU | [OpenAI-Clip.onnx.zip](https://huggingface.co/qualcomm/OpenAI-Clip/blob/main/OpenAI-Clip.onnx.zip) |
66
-
67
-
68
-
69
-
70
- ## Installation
71
-
72
-
73
- Install the package via pip:
74
- ```bash
75
- pip install qai-hub-models
76
- ```
77
-
78
-
79
- ## Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device
80
-
81
- Sign-in to [Qualcomm® AI Hub Workbench](https://workbench.aihub.qualcomm.com/) with your
82
- Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.
83
-
84
- With this API token, you can configure your client to run models on the cloud
85
- hosted devices.
86
- ```bash
87
- qai-hub configure --api_token API_TOKEN
88
- ```
89
- Navigate to [docs](https://workbench.aihub.qualcomm.com/docs/) for more information.
90
-
91
-
92
-
93
- ## Demo off target
94
-
95
- The package contains a simple end-to-end demo that downloads pre-trained
96
- weights and runs this model on a sample input.
97
-
98
- ```bash
99
- python -m qai_hub_models.models.openai_clip.demo
100
- ```
101
-
102
- The above demo runs a reference implementation of pre-processing, model
103
- inference, and post processing.
104
-
105
- **NOTE**: If you want running in a Jupyter Notebook or Google Colab like
106
- environment, please add the following to your cell (instead of the above).
107
- ```
108
- %run -m qai_hub_models.models.openai_clip.demo
109
- ```
110
-
111
-
112
- ### Run model on a cloud-hosted device
113
-
114
- In addition to the demo, you can also run the model on a cloud-hosted Qualcomm®
115
- device. This script does the following:
116
- * Performance check on-device on a cloud-hosted device
117
- * Downloads compiled assets that can be deployed on-device for Android.
118
- * Accuracy check between PyTorch and on-device outputs.
119
-
120
- ```bash
121
- python -m qai_hub_models.models.openai_clip.export
122
- ```
123
-
124
-
125
-
126
- ## How does this work?
127
-
128
- This [export script](https://aihub.qualcomm.com/models/openai_clip/qai_hub_models/models/OpenAI-Clip/export.py)
129
- leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) to optimize, validate, and deploy this model
130
- on-device. Lets go through each step below in detail:
131
-
132
- Step 1: **Compile model for on-device deployment**
133
-
134
- To compile a PyTorch model for on-device deployment, we first trace the model
135
- in memory using the `jit.trace` and then call the `submit_compile_job` API.
136
-
137
- ```python
138
- import torch
139
-
140
- import qai_hub as hub
141
- from qai_hub_models.models.openai_clip import Model
142
-
143
- # Load the model
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()
151
- sample_inputs = torch_model.sample_inputs()
152
-
153
- pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])
154
-
155
- # Compile model on a specific device
156
- compile_job = hub.submit_compile_job(
157
- model=pt_model,
158
- device=device,
159
- input_specs=torch_model.get_input_spec(),
160
- )
161
-
162
- # Get target model to run on-device
163
- target_model = compile_job.get_target_model()
164
-
165
- ```
166
-
167
-
168
- Step 2: **Performance profiling on cloud-hosted device**
169
-
170
- After compiling models from step 1. Models can be profiled model on-device using the
171
- `target_model`. Note that this scripts runs the model on a device automatically
172
- provisioned in the cloud. Once the job is submitted, you can navigate to a
173
- provided job URL to view a variety of on-device performance metrics.
174
- ```python
175
- profile_job = hub.submit_profile_job(
176
- model=target_model,
177
- device=device,
178
- )
179
-
180
- ```
181
-
182
- Step 3: **Verify on-device accuracy**
183
-
184
- To verify the accuracy of the model on-device, you can run on-device inference
185
- on sample input data on the same cloud hosted device.
186
- ```python
187
- input_data = torch_model.sample_inputs()
188
- inference_job = hub.submit_inference_job(
189
- model=target_model,
190
- device=device,
191
- inputs=input_data,
192
- )
193
- on_device_output = inference_job.download_output_data()
194
-
195
- ```
196
- With the output of the model, you can compute like PSNR, relative errors or
197
- spot check the output with expected output.
198
-
199
- **Note**: This on-device profiling and inference requires access to Qualcomm®
200
- AI Hub Workbench. [Sign up for access](https://myaccount.qualcomm.com/signup).
201
-
202
-
203
-
204
-
205
- ## Deploying compiled model to Android
206
-
207
-
208
- The models can be deployed using multiple runtimes:
209
- - TensorFlow Lite (`.tflite` export): [This
210
- tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a
211
- guide to deploy the .tflite model in an Android application.
212
-
213
-
214
- - QNN (`.so` export ): This [sample
215
- app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html)
216
- provides instructions on how to use the `.so` shared library in an Android application.
217
-
218
-
219
- ## View on Qualcomm® AI Hub
220
- Get more details on OpenAI-Clip's performance across various devices [here](https://aihub.qualcomm.com/models/openai_clip).
221
- Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
222
-
223
 
224
  ## License
225
  * The license for the original implementation of OpenAI-Clip can be found
226
  [here](https://github.com/openai/CLIP/blob/main/LICENSE).
227
 
228
-
229
-
230
  ## References
231
  * [Learning Transferable Visual Models From Natural Language Supervision](https://arxiv.org/abs/2103.00020)
232
  * [Source Model Implementation](https://github.com/openai/CLIP/)
233
 
234
-
235
-
236
  ## Community
237
  * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
238
  * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).
239
-
240
-
 
10
 
11
  ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/openai_clip/web-assets/model_demo.png)
12
 
13
+ # OpenAI-Clip: Optimized for Qualcomm Devices
 
 
14
 
15
  Contrastive Language-Image Pre-Training (CLIP) uses a ViT like transformer to get visual features and a causal language model to get the text features. Both the text and visual features can then be used for a variety of zero-shot learning tasks.
16
 
17
+ This is based on the implementation of OpenAI-Clip found [here](https://github.com/openai/CLIP/).
18
+ This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/openai_clip) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).
19
+
20
+ Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.
21
+
22
+ ## Getting Started
23
+ There are two ways to deploy this model on your device:
24
+
25
+ ### Option 1: Download Pre-Exported Models
26
+
27
+ Below are pre-exported model assets ready for deployment.
28
+
29
+ | Runtime | Precision | Chipset | SDK Versions | Download |
30
+ |---|---|---|---|---|
31
+ | ONNX | float | Universal | QAIRT 2.37, ONNX Runtime 1.23.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/openai_clip/releases/v0.46.1/openai_clip-onnx-float.zip)
32
+ | QNN_DLC | float | Universal | QAIRT 2.42 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/openai_clip/releases/v0.46.1/openai_clip-qnn_dlc-float.zip)
33
+ | TFLITE | float | Universal | QAIRT 2.42, TFLite 2.17.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/openai_clip/releases/v0.46.1/openai_clip-tflite-float.zip)
34
+
35
+ For more device-specific assets and performance metrics, visit **[OpenAI-Clip on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/openai_clip)**.
36
+
37
+
38
+ ### Option 2: Export with Custom Configurations
39
+
40
+ Use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/openai_clip) Python library to compile and export the model with your own:
41
+ - Custom weights (e.g., fine-tuned checkpoints)
42
+ - Custom input shapes
43
+ - Target device and runtime configurations
44
+
45
+ This option is ideal if you need to customize the model beyond the default configuration provided here.
46
+
47
+ See our repository for [OpenAI-Clip on GitHub](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/openai_clip) for usage instructions.
48
+
49
+ ## Model Details
50
+
51
+ **Model Type:** Model_use_case.image_classification
52
+
53
+ **Model Stats:**
54
+ - Model checkpoint: ViT-B/16
55
+ - Image input resolution: 224x224
56
+ - Text context length: 77
57
+ - Number of parameters: 150M
58
+ - Model size (float): 571 MB
59
+
60
+ ## Performance Summary
61
+ | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
62
+ |---|---|---|---|---|---|---
63
+ | OpenAI-Clip | ONNX | float | Snapdragon® X Elite | 22.459 ms | 294 - 294 MB | NPU
64
+ | OpenAI-Clip | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 15.269 ms | 0 - 796 MB | NPU
65
+ | OpenAI-Clip | ONNX | float | Qualcomm® QCS8550 (Proxy) | 22.212 ms | 0 - 323 MB | NPU
66
+ | OpenAI-Clip | ONNX | float | Qualcomm® QCS9075 | 25.758 ms | 0 - 4 MB | NPU
67
+ | OpenAI-Clip | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 12.318 ms | 1 - 713 MB | NPU
68
+ | OpenAI-Clip | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 10.095 ms | 1 - 661 MB | NPU
69
+ | OpenAI-Clip | QNN_DLC | float | Snapdragon® X Elite | 18.808 ms | 1 - 1 MB | NPU
70
+ | OpenAI-Clip | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 12.624 ms | 0 - 550 MB | NPU
71
+ | OpenAI-Clip | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 55.883 ms | 1 - 507 MB | NPU
72
+ | OpenAI-Clip | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 17.922 ms | 1 - 593 MB | NPU
73
+ | OpenAI-Clip | QNN_DLC | float | Qualcomm® SA8775P | 20.876 ms | 1 - 504 MB | NPU
74
+ | OpenAI-Clip | QNN_DLC | float | Qualcomm�� QCS9075 | 20.902 ms | 1 - 3 MB | NPU
75
+ | OpenAI-Clip | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 21.094 ms | 0 - 501 MB | NPU
76
+ | OpenAI-Clip | QNN_DLC | float | Qualcomm® SA7255P | 55.883 ms | 1 - 507 MB | NPU
77
+ | OpenAI-Clip | QNN_DLC | float | Qualcomm® SA8295P | 22.195 ms | 0 - 492 MB | NPU
78
+ | OpenAI-Clip | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 10.588 ms | 1 - 515 MB | NPU
79
+ | OpenAI-Clip | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 8.462 ms | 0 - 485 MB | NPU
80
+ | OpenAI-Clip | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 11.008 ms | 0 - 562 MB | NPU
81
+ | OpenAI-Clip | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 52.168 ms | 0 - 512 MB | NPU
82
+ | OpenAI-Clip | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 15.689 ms | 0 - 4 MB | NPU
83
+ | OpenAI-Clip | TFLITE | float | Qualcomm® SA8775P | 18.638 ms | 0 - 509 MB | NPU
84
+ | OpenAI-Clip | TFLITE | float | Qualcomm® QCS9075 | 20.357 ms | 0 - 294 MB | NPU
85
+ | OpenAI-Clip | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 20.361 ms | 0 - 503 MB | NPU
86
+ | OpenAI-Clip | TFLITE | float | Qualcomm® SA7255P | 52.168 ms | 0 - 512 MB | NPU
87
+ | OpenAI-Clip | TFLITE | float | Qualcomm® SA8295P | 21.567 ms | 0 - 495 MB | NPU
88
+ | OpenAI-Clip | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 9.059 ms | 0 - 526 MB | NPU
89
+ | OpenAI-Clip | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 7.023 ms | 0 - 497 MB | NPU
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
90
 
91
  ## License
92
  * The license for the original implementation of OpenAI-Clip can be found
93
  [here](https://github.com/openai/CLIP/blob/main/LICENSE).
94
 
 
 
95
  ## References
96
  * [Learning Transferable Visual Models From Natural Language Supervision](https://arxiv.org/abs/2103.00020)
97
  * [Source Model Implementation](https://github.com/openai/CLIP/)
98
 
 
 
99
  ## Community
100
  * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
101
  * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).
 
 
tool-versions.yaml DELETED
@@ -1,4 +0,0 @@
1
- tool_versions:
2
- onnx:
3
- qairt: 2.37.1.250807093845_124904
4
- onnx_runtime: 1.23.0