Hugging Face's logo

qualcomm
/
LaMa-Dilated

Image-to-Image
PyTorch
LiteRT
backbone
android
Model card Files
xet
Community
qaihm-bot commited on
Commit
e5726c6
·
verified ·
1 Parent(s): 3b9ac59

v0.32.0

Browse files

See https://github.com/quic/ai-hub-models/releases/v0.32.0 for changelog.

Files changed (7) hide show
  1. .gitattributes +1 -0
  2. DEPLOYMENT_MODEL_LICENSE.pdf +3 -0
  3. LICENSE +2 -0
  4. LaMa-Dilated.dlc +2 -2
  5. LaMa-Dilated.onnx +2 -2
  6. LaMa-Dilated.tflite +2 -2
  7. README.md +34 -34
.gitattributes CHANGED
@@ -35,3 +35,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  LaMa-Dilated.so filter=lfs diff=lfs merge=lfs -text
37
  LaMa-Dilated.dlc filter=lfs diff=lfs merge=lfs -text
 
 
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  LaMa-Dilated.so filter=lfs diff=lfs merge=lfs -text
37
  LaMa-Dilated.dlc filter=lfs diff=lfs merge=lfs -text
38
+ DEPLOYMENT_MODEL_LICENSE.pdf filter=lfs diff=lfs merge=lfs -text
DEPLOYMENT_MODEL_LICENSE.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4409f93b0e82531303b3e10f52f1fdfb56467a25f05b7441c6bbd8bb8a64b42c
3
+ size 109629
LICENSE ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ The license of the original trained model can be found at https://github.com/advimman/lama/blob/main/LICENSE.
2
+ The license for the deployable model files (.tflite, .onnx, .dlc, .bin, etc.) can be found in DEPLOYMENT_MODEL_LICENSE.pdf.
LaMa-Dilated.dlc CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c557d4596f6df546d01a2bfe4b0305dc75178ced4e9a22251d530e6c5d94095b
3
- size 182934911
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8117616480f6875b80bc569d6a70d223e30b9f522bcd4699c437ee450e55bbab
3
+ size 182934503
LaMa-Dilated.onnx CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e90ae4943ed438642c0d78c6b524813a7ab863e0a21094e904fb7974088e2d61
3
- size 182781009
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:75b3abe38e19d2f8f332a25f4cfaa400dc0dde07dc5558550c61de406ebdc766
3
+ size 182788300
LaMa-Dilated.tflite CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ea9b30519fb9a5814007f74b0b7fc078c997539104fc55f8b8005b7bf66c3eb5
3
- size 182653032
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5de02609582df172a20966d2e0355cece2e3178b7f1147b354a63446b229c130
3
+ size 182648372
README.md CHANGED
@@ -31,39 +31,39 @@ More details on model performance across various devices, can be found
31
  - Model checkpoint: Dilated CelebAHQ
32
  - Input resolution: 512x512
33
  - Number of parameters: 45.6M
34
- - Model size: 174 MB
35
 
36
  | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
37
  |---|---|---|---|---|---|---|---|---|
38
- | LaMa-Dilated | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 434.02 ms | 3 - 140 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
39
- | LaMa-Dilated | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 406.967 ms | 1 - 130 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
40
- | LaMa-Dilated | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 119.97 ms | 3 - 183 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
41
- | LaMa-Dilated | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 124.79 ms | 4 - 129 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
42
- | LaMa-Dilated | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 89.247 ms | 3 - 322 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
43
- | LaMa-Dilated | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 77.431 ms | 6 - 57 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
44
- | LaMa-Dilated | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 123.027 ms | 3 - 140 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
45
- | LaMa-Dilated | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 110.268 ms | 2 - 130 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
46
- | LaMa-Dilated | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 434.02 ms | 3 - 140 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
47
- | LaMa-Dilated | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 406.967 ms | 1 - 130 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
48
- | LaMa-Dilated | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 89.221 ms | 3 - 328 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
49
- | LaMa-Dilated | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 78.183 ms | 3 - 55 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
50
- | LaMa-Dilated | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 128.883 ms | 3 - 109 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
51
- | LaMa-Dilated | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 116.524 ms | 0 - 118 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
52
- | LaMa-Dilated | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 89.161 ms | 3 - 324 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
53
- | LaMa-Dilated | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 77.621 ms | 3 - 54 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
54
- | LaMa-Dilated | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 123.027 ms | 3 - 140 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
55
- | LaMa-Dilated | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 110.268 ms | 2 - 130 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
56
- | LaMa-Dilated | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 89.212 ms | 3 - 315 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
57
- | LaMa-Dilated | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 78.013 ms | 3 - 56 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
58
- | LaMa-Dilated | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 77.578 ms | 7 - 247 MB | NPU | [LaMa-Dilated.onnx](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.onnx) |
59
- | LaMa-Dilated | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 63.084 ms | 3 - 218 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
60
- | LaMa-Dilated | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 55.226 ms | 4 - 135 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
61
- | LaMa-Dilated | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 54.852 ms | 0 - 142 MB | NPU | [LaMa-Dilated.onnx](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.onnx) |
62
- | LaMa-Dilated | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 59.128 ms | 3 - 145 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
63
- | LaMa-Dilated | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 50.298 ms | 4 - 134 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
64
- | LaMa-Dilated | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 49.944 ms | 10 - 146 MB | NPU | [LaMa-Dilated.onnx](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.onnx) |
65
- | LaMa-Dilated | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 89.662 ms | 101 - 101 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
66
- | LaMa-Dilated | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 87.072 ms | 89 - 89 MB | NPU | [LaMa-Dilated.onnx](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.onnx) |
67
 
68
 
69
 
@@ -127,10 +127,10 @@ Profiling Results
127
  LaMa-Dilated
128
  Device : cs_8275 (ANDROID 14)
129
  Runtime : TFLITE
130
- Estimated inference time (ms) : 434.0
131
- Estimated peak memory usage (MB): [3, 140]
132
- Total # Ops : 343
133
- Compute Unit(s) : npu (343 ops) gpu (0 ops) cpu (0 ops)
134
  ```
135
 
136
 
 
31
  - Model checkpoint: Dilated CelebAHQ
32
  - Input resolution: 512x512
33
  - Number of parameters: 45.6M
34
+ - Model size (float): 174 MB
35
 
36
  | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
37
  |---|---|---|---|---|---|---|---|---|
38
+ | LaMa-Dilated | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 429.471 ms | 3 - 138 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
39
+ | LaMa-Dilated | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 406.956 ms | 1 - 130 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
40
+ | LaMa-Dilated | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 115.12 ms | 3 - 180 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
41
+ | LaMa-Dilated | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 124.295 ms | 4 - 132 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
42
+ | LaMa-Dilated | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 86.771 ms | 0 - 315 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
43
+ | LaMa-Dilated | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 77.725 ms | 3 - 54 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
44
+ | LaMa-Dilated | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 120.557 ms | 3 - 139 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
45
+ | LaMa-Dilated | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 110.243 ms | 1 - 129 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
46
+ | LaMa-Dilated | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 429.471 ms | 3 - 138 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
47
+ | LaMa-Dilated | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 406.956 ms | 1 - 130 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
48
+ | LaMa-Dilated | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 86.079 ms | 3 - 323 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
49
+ | LaMa-Dilated | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 77.978 ms | 2 - 53 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
50
+ | LaMa-Dilated | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 126.414 ms | 3 - 107 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
51
+ | LaMa-Dilated | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 116.51 ms | 2 - 120 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
52
+ | LaMa-Dilated | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 86.558 ms | 3 - 320 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
53
+ | LaMa-Dilated | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 78.392 ms | 3 - 54 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
54
+ | LaMa-Dilated | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 120.557 ms | 3 - 139 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
55
+ | LaMa-Dilated | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 110.243 ms | 1 - 129 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
56
+ | LaMa-Dilated | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 86.106 ms | 3 - 330 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
57
+ | LaMa-Dilated | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 77.828 ms | 2 - 54 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
58
+ | LaMa-Dilated | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 76.618 ms | 7 - 222 MB | NPU | [LaMa-Dilated.onnx](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.onnx) |
59
+ | LaMa-Dilated | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 61.154 ms | 2 - 217 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
60
+ | LaMa-Dilated | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 55.296 ms | 4 - 137 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
61
+ | LaMa-Dilated | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 53.893 ms | 6 - 140 MB | NPU | [LaMa-Dilated.onnx](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.onnx) |
62
+ | LaMa-Dilated | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 57.525 ms | 2 - 144 MB | NPU | [LaMa-Dilated.tflite](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.tflite) |
63
+ | LaMa-Dilated | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 50.28 ms | 4 - 134 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
64
+ | LaMa-Dilated | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 48.981 ms | 10 - 144 MB | NPU | [LaMa-Dilated.onnx](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.onnx) |
65
+ | LaMa-Dilated | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 90.085 ms | 104 - 104 MB | NPU | [LaMa-Dilated.dlc](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.dlc) |
66
+ | LaMa-Dilated | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 84.178 ms | 90 - 90 MB | NPU | [LaMa-Dilated.onnx](https://huggingface.co/qualcomm/LaMa-Dilated/blob/main/LaMa-Dilated.onnx) |
67
 
68
 
69
 
 
127
  LaMa-Dilated
128
  Device : cs_8275 (ANDROID 14)
129
  Runtime : TFLITE
130
+ Estimated inference time (ms) : 429.5
131
+ Estimated peak memory usage (MB): [3, 138]
132
+ Total # Ops : 340
133
+ Compute Unit(s) : npu (340 ops) gpu (0 ops) cpu (0 ops)
134
  ```
135
 
136