Hugging Face's logo

qualcomm
/
EfficientFormer

Image Classification
PyTorch
bu_auto
android
Model card Files
xet
Community
qaihm-bot commited on
Commit
7c86b86
·
verified ·
1 Parent(s): 11223ba

v0.42.0

Browse files

See https://github.com/quic/ai-hub-models/releases/v0.42.0 for changelog.

Files changed (7) hide show
  1. .gitattributes +1 -0
  2. EfficientFormer_float.dlc +2 -2
  3. EfficientFormer_float.onnx.zip +2 -2
  4. EfficientFormer_w8a8.dlc → EfficientFormer_w8a16.dlc +2 -2
  5. EfficientFormer_w8a8.onnx.zip → EfficientFormer_w8a16.onnx.zip +2 -2
  6. EfficientFormer_w8a8.tflite +0 -3
  7. README.md +40 -49
.gitattributes CHANGED
@@ -37,3 +37,4 @@ DEPLOYMENT_MODEL_LICENSE.pdf filter=lfs diff=lfs merge=lfs -text
37
  EfficientFormer.dlc filter=lfs diff=lfs merge=lfs -text
38
  EfficientFormer_float.dlc filter=lfs diff=lfs merge=lfs -text
39
  EfficientFormer_w8a8.dlc filter=lfs diff=lfs merge=lfs -text
 
 
37
  EfficientFormer.dlc filter=lfs diff=lfs merge=lfs -text
38
  EfficientFormer_float.dlc filter=lfs diff=lfs merge=lfs -text
39
  EfficientFormer_w8a8.dlc filter=lfs diff=lfs merge=lfs -text
40
+ EfficientFormer_w8a16.dlc filter=lfs diff=lfs merge=lfs -text
EfficientFormer_float.dlc CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:dbb16f5b9478dc6ae9db33abb6d9fcd2c470da33c33171685c6acbdcc560437a
3
- size 49274068
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d7015091e23aa284fefbe0fe478960320f7803681b6f966f5322a92b2f49190
3
+ size 49274172
EfficientFormer_float.onnx.zip CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b868cffe01de9a95ff384109ee251eba9c2e52549ebb90b64bdfb8b997b31560
3
- size 45831790
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9360d6bd0a858068e0010e5d9597354eb0ab6412d15413e36ebb61a98b90337e
3
+ size 45831792
EfficientFormer_w8a8.dlc → EfficientFormer_w8a16.dlc RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f80a6bfafcbc055ba71735cde354669cec8a9aeda4a3632746fe447b6e46d719
3
- size 13539044
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e655d6818515851b32e46d8dd0190f1f951461f254d8f10cc808983e53b49ca3
3
+ size 13565308
EfficientFormer_w8a8.onnx.zip → EfficientFormer_w8a16.onnx.zip RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2db5a5e87b6fb7d36d4fb75f25b8c95f31e9ccb982e7dcfcb2ad676eeee65a6c
3
- size 27903208
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc7b1793642a8a78a37d4a53ad233de132e5820254cc4bcd1280bd1894de0a30
3
+ size 27608368
EfficientFormer_w8a8.tflite DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:62b8461ed44bb9535c537fa47db0145a56118ec0aca477942fcdf09079fa0b81
3
- size 12793736
 
 
 
 
README.md CHANGED
@@ -36,51 +36,42 @@ 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
- | EfficientFormer | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 4.92 ms | 0 - 48 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.tflite) |
40
- | EfficientFormer | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 5.007 ms | 1 - 33 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
41
- | EfficientFormer | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 3.87 ms | 0 - 57 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.tflite) |
42
- | EfficientFormer | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 5.712 ms | 0 - 45 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
43
- | EfficientFormer | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 1.519 ms | 0 - 163 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.tflite) |
44
- | EfficientFormer | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 1.573 ms | 0 - 48 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
45
- | EfficientFormer | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 1.669 ms | 0 - 66 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.onnx.zip) |
46
- | EfficientFormer | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 2.135 ms | 0 - 49 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.tflite) |
47
- | EfficientFormer | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 8.603 ms | 1 - 33 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
48
- | EfficientFormer | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 1.072 ms | 0 - 60 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.tflite) |
49
- | EfficientFormer | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 1.109 ms | 1 - 43 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
50
- | EfficientFormer | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 1.113 ms | 0 - 44 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.onnx.zip) |
51
- | EfficientFormer | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 0.817 ms | 0 - 53 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.tflite) |
52
- | EfficientFormer | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 0.818 ms | 1 - 40 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
53
- | EfficientFormer | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 0.901 ms | 0 - 37 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.onnx.zip) |
54
- | EfficientFormer | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 0.718 ms | 0 - 52 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.tflite) |
55
- | EfficientFormer | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 0.697 ms | 1 - 40 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
56
- | EfficientFormer | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 0.846 ms | 1 - 38 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.onnx.zip) |
57
- | EfficientFormer | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 1.815 ms | 94 - 94 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
58
- | EfficientFormer | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 1.626 ms | 25 - 25 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.onnx.zip) |
59
- | EfficientFormer | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 1.699 ms | 0 - 27 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.tflite) |
60
- | EfficientFormer | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 1.726 ms | 0 - 31 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.dlc) |
61
- | EfficientFormer | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 0.864 ms | 0 - 52 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.tflite) |
62
- | EfficientFormer | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 1.097 ms | 0 - 41 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.dlc) |
63
- | EfficientFormer | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 0.744 ms | 0 - 59 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.tflite) |
64
- | EfficientFormer | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 0.811 ms | 0 - 50 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.dlc) |
65
- | EfficientFormer | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 6.826 ms | 0 - 36 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.onnx.zip) |
66
- | EfficientFormer | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 0.998 ms | 0 - 27 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.tflite) |
67
- | EfficientFormer | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 1.055 ms | 0 - 32 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.dlc) |
68
- | EfficientFormer | w8a8 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | TFLITE | 4.496 ms | 0 - 37 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.tflite) |
69
- | EfficientFormer | w8a8 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | QNN_DLC | 2.307 ms | 0 - 37 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.dlc) |
70
- | EfficientFormer | w8a8 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | ONNX | 30.405 ms | 13 - 27 MB | CPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.onnx.zip) |
71
- | EfficientFormer | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | TFLITE | 42.778 ms | 1 - 79 MB | GPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.tflite) |
72
- | EfficientFormer | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | ONNX | 26.715 ms | 12 - 20 MB | CPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.onnx.zip) |
73
- | EfficientFormer | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 0.516 ms | 0 - 40 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.tflite) |
74
- | EfficientFormer | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 0.562 ms | 0 - 45 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.dlc) |
75
- | EfficientFormer | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 5.645 ms | 5 - 48 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.onnx.zip) |
76
- | EfficientFormer | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 0.41 ms | 0 - 30 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.tflite) |
77
- | EfficientFormer | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 0.409 ms | 0 - 33 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.dlc) |
78
- | EfficientFormer | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 4.657 ms | 5 - 37 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.onnx.zip) |
79
- | EfficientFormer | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 0.381 ms | 0 - 34 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.tflite) |
80
- | EfficientFormer | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 0.355 ms | 0 - 34 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.dlc) |
81
- | EfficientFormer | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 4.229 ms | 5 - 39 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.onnx.zip) |
82
- | EfficientFormer | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 0.963 ms | 62 - 62 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.dlc) |
83
- | EfficientFormer | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 7.035 ms | 10 - 10 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a8.onnx.zip) |
84
 
85
 
86
 
@@ -95,9 +86,9 @@ pip install "qai-hub-models[efficientformer]"
95
  ```
96
 
97
 
98
- ## Configure Qualcomm® AI Hub to run this model on a cloud-hosted device
99
 
100
- Sign-in to [Qualcomm® AI Hub](https://app.aihub.qualcomm.com/) with your
101
  Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.
102
 
103
  With this API token, you can configure your client to run models on the cloud
@@ -105,7 +96,7 @@ hosted devices.
105
  ```bash
106
  qai-hub configure --api_token API_TOKEN
107
  ```
108
- Navigate to [docs](https://app.aihub.qualcomm.com/docs/) for more information.
109
 
110
 
111
 
@@ -216,7 +207,7 @@ With the output of the model, you can compute like PSNR, relative errors or
216
  spot check the output with expected output.
217
 
218
  **Note**: This on-device profiling and inference requires access to Qualcomm®
219
- AI Hub. [Sign up for access](https://myaccount.qualcomm.com/signup).
220
 
221
 
222
 
 
36
 
37
  | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
38
  |---|---|---|---|---|---|---|---|---|
39
+ | EfficientFormer | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 4.931 ms | 0 - 49 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.tflite) |
40
+ | EfficientFormer | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 4.999 ms | 1 - 35 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
41
+ | EfficientFormer | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 3.847 ms | 0 - 59 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.tflite) |
42
+ | EfficientFormer | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 5.613 ms | 0 - 42 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
43
+ | EfficientFormer | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 1.514 ms | 0 - 165 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.tflite) |
44
+ | EfficientFormer | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 1.586 ms | 0 - 45 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
45
+ | EfficientFormer | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 1.7 ms | 0 - 82 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.onnx.zip) |
46
+ | EfficientFormer | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 2.121 ms | 0 - 50 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.tflite) |
47
+ | EfficientFormer | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 2.243 ms | 1 - 35 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
48
+ | EfficientFormer | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 1.073 ms | 0 - 59 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.tflite) |
49
+ | EfficientFormer | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 1.104 ms | 0 - 45 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
50
+ | EfficientFormer | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 1.119 ms | 0 - 43 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.onnx.zip) |
51
+ | EfficientFormer | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 0.819 ms | 0 - 54 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.tflite) |
52
+ | EfficientFormer | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 0.819 ms | 1 - 40 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
53
+ | EfficientFormer | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 0.896 ms | 0 - 38 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.onnx.zip) |
54
+ | EfficientFormer | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 0.72 ms | 0 - 54 MB | NPU | [EfficientFormer.tflite](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.tflite) |
55
+ | EfficientFormer | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 0.693 ms | 1 - 43 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
56
+ | EfficientFormer | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 0.813 ms | 0 - 36 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.onnx.zip) |
57
+ | EfficientFormer | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 1.819 ms | 109 - 109 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.dlc) |
58
+ | EfficientFormer | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 1.623 ms | 25 - 25 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer.onnx.zip) |
59
+ | EfficientFormer | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 3.218 ms | 0 - 35 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.dlc) |
60
+ | EfficientFormer | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 1.566 ms | 0 - 10 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.dlc) |
61
+ | EfficientFormer | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 7.56 ms | 0 - 28 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.onnx.zip) |
62
+ | EfficientFormer | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 7.877 ms | 0 - 35 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.dlc) |
63
+ | EfficientFormer | w8a16 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | ONNX | 59.93 ms | 20 - 33 MB | CPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.onnx.zip) |
64
+ | EfficientFormer | w8a16 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | ONNX | 52.736 ms | 18 - 25 MB | CPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.onnx.zip) |
65
+ | EfficientFormer | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 1.075 ms | 0 - 45 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.dlc) |
66
+ | EfficientFormer | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 6.673 ms | 5 - 47 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.onnx.zip) |
67
+ | EfficientFormer | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 0.704 ms | 0 - 38 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.dlc) |
68
+ | EfficientFormer | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 4.937 ms | 0 - 29 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.onnx.zip) |
69
+ | EfficientFormer | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | QNN_DLC | 1.702 ms | 0 - 43 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.dlc) |
70
+ | EfficientFormer | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | ONNX | 60.676 ms | 22 - 38 MB | CPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.onnx.zip) |
71
+ | EfficientFormer | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 0.598 ms | 0 - 42 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.dlc) |
72
+ | EfficientFormer | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 4.309 ms | 0 - 35 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.onnx.zip) |
73
+ | EfficientFormer | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 1.798 ms | 38 - 38 MB | NPU | [EfficientFormer.dlc](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.dlc) |
74
+ | EfficientFormer | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 6.469 ms | 10 - 10 MB | NPU | [EfficientFormer.onnx.zip](https://huggingface.co/qualcomm/EfficientFormer/blob/main/EfficientFormer_w8a16.onnx.zip) |
 
 
 
 
 
 
 
 
 
75
 
76
 
77
 
 
86
  ```
87
 
88
 
89
+ ## Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device
90
 
91
+ Sign-in to [Qualcomm® AI Hub Workbench](https://workbench.aihub.qualcomm.com/) with your
92
  Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.
93
 
94
  With this API token, you can configure your client to run models on the cloud
 
96
  ```bash
97
  qai-hub configure --api_token API_TOKEN
98
  ```
99
+ Navigate to [docs](https://workbench.aihub.qualcomm.com/docs/) for more information.
100
 
101
 
102
 
 
207
  spot check the output with expected output.
208
 
209
  **Note**: This on-device profiling and inference requires access to Qualcomm®
210
+ AI Hub Workbench. [Sign up for access](https://myaccount.qualcomm.com/signup).
211
 
212
 
213