Hugging Face's logo

qualcomm
/
HuggingFace-WavLM-Base-Plus

Automatic Speech Recognition
PyTorch
LiteRT
backbone
android
Model card Files
xet
Community
bhushans commited on
Commit
47ebbeb
·
verified ·
1 Parent(s): 629b888

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +29 -12
README.md CHANGED
@@ -35,15 +35,16 @@ More details on model performance across various devices, can be found
35
 
36
  | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
37
  |---|---|---|---|---|---|---|---|---|
38
- | HuggingFace-WavLM-Base-Plus | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 814.795 ms | 63 - 98 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
39
- | HuggingFace-WavLM-Base-Plus | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 739.112 ms | 30 - 50 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
40
- | HuggingFace-WavLM-Base-Plus | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 530.552 ms | 63 - 77 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
41
- | HuggingFace-WavLM-Base-Plus | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 742.901 ms | 63 - 66 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
42
- | HuggingFace-WavLM-Base-Plus | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 927.766 ms | 63 - 610 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
43
- | HuggingFace-WavLM-Base-Plus | SA8775 (Proxy) | SA8775P Proxy | TFLITE | 919.003 ms | 63 - 100 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
44
- | HuggingFace-WavLM-Base-Plus | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 887.749 ms | 60 - 63 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
45
- | HuggingFace-WavLM-Base-Plus | SA8295P ADP | SA8295P | TFLITE | 1027.812 ms | 63 - 82 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
46
- | HuggingFace-WavLM-Base-Plus | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 1159.685 ms | 30 - 57 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
 
47
 
48
 
49
 
@@ -108,8 +109,8 @@ Profiling Results
108
  HuggingFace-WavLM-Base-Plus
109
  Device : Samsung Galaxy S23 (13)
110
  Runtime : TFLITE
111
- Estimated inference time (ms) : 814.8
112
- Estimated peak memory usage (MB): [63, 98]
113
  Total # Ops : 871
114
  Compute Unit(s) : CPU (871 ops)
115
  ```
@@ -130,13 +131,29 @@ in memory using the `jit.trace` and then call the `submit_compile_job` API.
130
  import torch
131
 
132
  import qai_hub as hub
133
- from qai_hub_models.models.huggingface_wavlm_base_plus import
134
 
135
  # Load the model
 
136
 
137
  # Device
138
  device = hub.Device("Samsung Galaxy S23")
139
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
140
 
141
  ```
142
 
 
35
 
36
  | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
37
  |---|---|---|---|---|---|---|---|---|
38
+ | HuggingFace-WavLM-Base-Plus | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 768.657 ms | 63 - 103 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
39
+ | HuggingFace-WavLM-Base-Plus | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 818.517 ms | 32 - 54 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
40
+ | HuggingFace-WavLM-Base-Plus | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 543.868 ms | 61 - 76 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
41
+ | HuggingFace-WavLM-Base-Plus | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 953.526 ms | 63 - 96 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
42
+ | HuggingFace-WavLM-Base-Plus | SA7255P ADP | SA7255P | TFLITE | 1567.307 ms | 63 - 83 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
43
+ | HuggingFace-WavLM-Base-Plus | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 924.443 ms | 44 - 46 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
44
+ | HuggingFace-WavLM-Base-Plus | SA8295P ADP | SA8295P | TFLITE | 985.456 ms | 63 - 81 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
45
+ | HuggingFace-WavLM-Base-Plus | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 726.672 ms | 63 - 66 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
46
+ | HuggingFace-WavLM-Base-Plus | SA8775P ADP | SA8775P | TFLITE | 1317.403 ms | 63 - 83 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
47
+ | HuggingFace-WavLM-Base-Plus | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 1062.167 ms | 64 - 89 MB | FP32 | CPU | [HuggingFace-WavLM-Base-Plus.tflite](https://huggingface.co/qualcomm/HuggingFace-WavLM-Base-Plus/blob/main/HuggingFace-WavLM-Base-Plus.tflite) |
48
 
49
 
50
 
 
109
  HuggingFace-WavLM-Base-Plus
110
  Device : Samsung Galaxy S23 (13)
111
  Runtime : TFLITE
112
+ Estimated inference time (ms) : 768.7
113
+ Estimated peak memory usage (MB): [63, 103]
114
  Total # Ops : 871
115
  Compute Unit(s) : CPU (871 ops)
116
  ```
 
131
  import torch
132
 
133
  import qai_hub as hub
134
+ from qai_hub_models.models.huggingface_wavlm_base_plus import Model
135
 
136
  # Load the model
137
+ torch_model = Model.from_pretrained()
138
 
139
  # Device
140
  device = hub.Device("Samsung Galaxy S23")
141
 
142
+ # Trace model
143
+ input_shape = torch_model.get_input_spec()
144
+ sample_inputs = torch_model.sample_inputs()
145
+
146
+ pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])
147
+
148
+ # Compile model on a specific device
149
+ compile_job = hub.submit_compile_job(
150
+ model=pt_model,
151
+ device=device,
152
+ input_specs=torch_model.get_input_spec(),
153
+ )
154
+
155
+ # Get target model to run on-device
156
+ target_model = compile_job.get_target_model()
157
 
158
  ```
159