qaihm-bot commited on
Commit
962dadb
·
verified ·
1 Parent(s): e101aac

See https://github.com/quic/ai-hub-models/releases/v0.31.0 for changelog.

.gitattributes CHANGED
@@ -35,3 +35,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  WhisperEncoder.so filter=lfs diff=lfs merge=lfs -text
37
  WhisperDecoder.so filter=lfs diff=lfs merge=lfs -text
 
 
 
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  WhisperEncoder.so filter=lfs diff=lfs merge=lfs -text
37
  WhisperDecoder.so filter=lfs diff=lfs merge=lfs -text
38
+ Whisper-Small-En_WhisperDecoderInf.dlc filter=lfs diff=lfs merge=lfs -text
39
+ Whisper-Small-En_WhisperEncoderInf.dlc filter=lfs diff=lfs merge=lfs -text
README.md CHANGED
@@ -38,50 +38,32 @@ More details on model performance across various devices, can be found
38
 
39
  | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
40
  |---|---|---|---|---|---|---|---|---|
41
- | WhisperEncoderInf | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 3217.923 ms | 107 - 140 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
42
- | WhisperEncoderInf | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN | 1115.123 ms | 1 - 10 MB | NPU | Use Export Script |
43
- | WhisperEncoderInf | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 938.784 ms | 109 - 208 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
44
- | WhisperEncoderInf | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 683.69 ms | 37 - 139 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
45
- | WhisperEncoderInf | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 759.812 ms | 1 - 3 MB | NPU | Use Export Script |
46
- | WhisperEncoderInf | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 1268.878 ms | 83 - 115 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
47
- | WhisperEncoderInf | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 580.733 ms | 1 - 9 MB | NPU | Use Export Script |
48
- | WhisperEncoderInf | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 3217.923 ms | 107 - 140 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
49
- | WhisperEncoderInf | float | SA7255P ADP | Qualcomm® SA7255P | QNN | 1115.123 ms | 1 - 10 MB | NPU | Use Export Script |
50
- | WhisperEncoderInf | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 675.98 ms | 102 - 164 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
51
- | WhisperEncoderInf | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN | 753.984 ms | 1 - 3 MB | NPU | Use Export Script |
52
- | WhisperEncoderInf | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 594.91 ms | 109 - 140 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
53
- | WhisperEncoderInf | float | SA8295P ADP | Qualcomm® SA8295P | QNN | 687.861 ms | 0 - 14 MB | NPU | Use Export Script |
54
- | WhisperEncoderInf | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 671.599 ms | 83 - 145 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
55
- | WhisperEncoderInf | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN | 759.003 ms | 4 - 6 MB | NPU | Use Export Script |
56
- | WhisperEncoderInf | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 1268.878 ms | 83 - 115 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
57
- | WhisperEncoderInf | float | SA8775P ADP | Qualcomm® SA8775P | QNN | 580.733 ms | 1 - 9 MB | NPU | Use Export Script |
58
- | WhisperEncoderInf | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 706.921 ms | 110 - 213 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
59
- | WhisperEncoderInf | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 528.505 ms | 108 - 199 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
60
- | WhisperEncoderInf | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 461.162 ms | 111 - 141 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
61
- | WhisperEncoderInf | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 537.812 ms | 0 - 0 MB | NPU | Use Export Script |
62
- | WhisperEncoderInf | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 635.133 ms | 296 - 296 MB | NPU | [Whisper-Small-En.onnx](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.onnx) |
63
- | WhisperDecoderInf | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 70.546 ms | 16 - 384 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
64
- | WhisperDecoderInf | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN | 18.697 ms | 54 - 63 MB | NPU | Use Export Script |
65
- | WhisperDecoderInf | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 49.144 ms | 16 - 391 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
66
- | WhisperDecoderInf | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 48.891 ms | 8 - 45 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
67
- | WhisperDecoderInf | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 12.103 ms | 61 - 68 MB | NPU | Use Export Script |
68
- | WhisperDecoderInf | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 49.479 ms | 16 - 384 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
69
- | WhisperDecoderInf | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 13.366 ms | 53 - 63 MB | NPU | Use Export Script |
70
- | WhisperDecoderInf | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 70.546 ms | 16 - 384 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
71
- | WhisperDecoderInf | float | SA7255P ADP | Qualcomm® SA7255P | QNN | 18.697 ms | 54 - 63 MB | NPU | Use Export Script |
72
- | WhisperDecoderInf | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 48.875 ms | 16 - 53 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
73
- | WhisperDecoderInf | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN | 11.952 ms | 54 - 56 MB | NPU | Use Export Script |
74
- | WhisperDecoderInf | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 42.684 ms | 16 - 353 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
75
- | WhisperDecoderInf | float | SA8295P ADP | Qualcomm® SA8295P | QNN | 13.45 ms | 57 - 71 MB | NPU | Use Export Script |
76
- | WhisperDecoderInf | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 48.796 ms | 16 - 49 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
77
- | WhisperDecoderInf | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN | 12.047 ms | 61 - 64 MB | NPU | Use Export Script |
78
- | WhisperDecoderInf | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 49.479 ms | 16 - 384 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
79
- | WhisperDecoderInf | float | SA8775P ADP | Qualcomm® SA8775P | QNN | 13.366 ms | 53 - 63 MB | NPU | Use Export Script |
80
- | WhisperDecoderInf | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 49.263 ms | 16 - 45 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
81
- | WhisperDecoderInf | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 43.809 ms | 5 - 406 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
82
- | WhisperDecoderInf | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 40.185 ms | 15 - 387 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
83
- | WhisperDecoderInf | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 10.5 ms | 61 - 61 MB | NPU | Use Export Script |
84
- | WhisperDecoderInf | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 34.881 ms | 226 - 226 MB | NPU | [Whisper-Small-En.onnx](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.onnx) |
85
 
86
 
87
 
@@ -145,8 +127,8 @@ Profiling Results
145
  WhisperEncoderInf
146
  Device : cs_8275 (ANDROID 14)
147
  Runtime : TFLITE
148
- Estimated inference time (ms) : 3217.9
149
- Estimated peak memory usage (MB): [107, 140]
150
  Total # Ops : 911
151
  Compute Unit(s) : npu (0 ops) gpu (900 ops) cpu (11 ops)
152
 
@@ -154,7 +136,7 @@ Compute Unit(s) : npu (0 ops) gpu (900 ops) cpu (11 ops)
154
  WhisperDecoderInf
155
  Device : cs_8275 (ANDROID 14)
156
  Runtime : TFLITE
157
- Estimated inference time (ms) : 70.5
158
  Estimated peak memory usage (MB): [16, 384]
159
  Total # Ops : 2573
160
  Compute Unit(s) : npu (2573 ops) gpu (0 ops) cpu (0 ops)
 
38
 
39
  | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
40
  |---|---|---|---|---|---|---|---|---|
41
+ | WhisperEncoderInf | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 3230.789 ms | 89 - 122 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
42
+ | WhisperEncoderInf | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 965.99 ms | 39 - 138 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
43
+ | WhisperEncoderInf | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 675.072 ms | 45 - 119 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
44
+ | WhisperEncoderInf | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 1268.791 ms | 108 - 140 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
45
+ | WhisperEncoderInf | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 3230.789 ms | 89 - 122 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
46
+ | WhisperEncoderInf | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 677.316 ms | 103 - 165 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
47
+ | WhisperEncoderInf | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 655.714 ms | 109 - 142 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
48
+ | WhisperEncoderInf | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 856.713 ms | 90 - 254 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
49
+ | WhisperEncoderInf | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 1268.791 ms | 108 - 140 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
50
+ | WhisperEncoderInf | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 686.994 ms | 85 - 167 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
51
+ | WhisperEncoderInf | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 518.045 ms | 109 - 201 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
52
+ | WhisperEncoderInf | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 545.405 ms | 381 - 381 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) |
53
+ | WhisperEncoderInf | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 631.145 ms | 295 - 295 MB | NPU | [Whisper-Small-En.onnx](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.onnx) |
54
+ | WhisperDecoderInf | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 71.015 ms | 16 - 384 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
55
+ | WhisperDecoderInf | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 49.307 ms | 16 - 397 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
56
+ | WhisperDecoderInf | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 49.176 ms | 5 - 43 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
57
+ | WhisperDecoderInf | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 49.431 ms | 16 - 384 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
58
+ | WhisperDecoderInf | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 71.015 ms | 16 - 384 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
59
+ | WhisperDecoderInf | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 51.019 ms | 10 - 42 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
60
+ | WhisperDecoderInf | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 42.717 ms | 11 - 349 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
61
+ | WhisperDecoderInf | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 48.824 ms | 16 - 51 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
62
+ | WhisperDecoderInf | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 49.431 ms | 16 - 384 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
63
+ | WhisperDecoderInf | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 50.949 ms | 16 - 49 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
64
+ | WhisperDecoderInf | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 44.926 ms | 16 - 419 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) |
65
+ | WhisperDecoderInf | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 34.427 ms | 1205 - 1205 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) |
66
+ | WhisperDecoderInf | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 34.753 ms | 227 - 227 MB | NPU | [Whisper-Small-En.onnx](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.onnx) |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
67
 
68
 
69
 
 
127
  WhisperEncoderInf
128
  Device : cs_8275 (ANDROID 14)
129
  Runtime : TFLITE
130
+ Estimated inference time (ms) : 3230.8
131
+ Estimated peak memory usage (MB): [89, 122]
132
  Total # Ops : 911
133
  Compute Unit(s) : npu (0 ops) gpu (900 ops) cpu (11 ops)
134
 
 
136
  WhisperDecoderInf
137
  Device : cs_8275 (ANDROID 14)
138
  Runtime : TFLITE
139
+ Estimated inference time (ms) : 71.0
140
  Estimated peak memory usage (MB): [16, 384]
141
  Total # Ops : 2573
142
  Compute Unit(s) : npu (2573 ops) gpu (0 ops) cpu (0 ops)
Whisper-Small-En_WhisperDecoderInf.dlc ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c7ba84b066799b259ebb914a33bc3494cbe8efbf4e8d81f9bdd08a3d26776bff
3
+ size 717952345
Whisper-Small-En_WhisperEncoderInf.dlc ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f62dd7c62b5c42f96143b2ec574cec176b0f2ddba7a5b262911ac85ec80cc81d
3
+ size 410233065