Card: add a measured Performance table (Pixel 8a, benchmark_model GPU+CPU), naming the runtime for each row
Browse files
README.md
CHANGED
|
@@ -123,3 +123,23 @@ repository under **`compiled_model_api/audio_codec`** (google-ai-edge/litert-sam
|
|
| 123 |
to the app's `filesDir` with that sample's `install_to_device.sh`.
|
| 124 |
|
| 125 |
License follows upstream Mimi (CC-BY-4.0).
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 123 |
to the app's `filesDir` with that sample's `install_to_device.sh`.
|
| 124 |
|
| 125 |
License follows upstream Mimi (CC-BY-4.0).
|
| 126 |
+
|
| 127 |
+
## Performance
|
| 128 |
+
|
| 129 |
+
Measured on a **Pixel 8a** (Tensor G3, Android 16) with the standard TFLite [`benchmark_model`](https://ai.google.dev/edge/litert/models/measurement) tool — 10 warm-up runs then 50 timed runs, reported as the tool's mean.
|
| 130 |
+
|
| 131 |
+
| Runtime | Backend | Graph on GPU | Latency |
|
| 132 |
+
|---|---|---|---|
|
| 133 |
+
| TFLite `benchmark_model` (`TfLiteGpuDelegateV2`) — `mimi_dec_tx_fp16.tflite` | GPU (OpenCL) | 689 / 689 | 21.7 ms |
|
| 134 |
+
| TFLite `benchmark_model` (`TfLiteGpuDelegateV2`) — `mimi_deconly_fp16.tflite` | GPU (OpenCL) | 220 / 220 | 114.3 ms |
|
| 135 |
+
| TFLite `benchmark_model` (`TfLiteGpuDelegateV2`) — `mimi_enc_conv_fp16.tflite` | GPU (OpenCL) | 189 / 189 | 93.4 ms |
|
| 136 |
+
| TFLite `benchmark_model` (`TfLiteGpuDelegateV2`) — `mimi_enc_tx_fp16.tflite` | GPU (OpenCL) | 681 / 681 | 28.0 ms |
|
| 137 |
+
| TFLite `benchmark_model` — `mimi_dec_tx_fp16.tflite` | CPU (XNNPACK, 4 threads) | — | XNNPACK declined the graph |
|
| 138 |
+
| TFLite `benchmark_model` — `mimi_deconly_fp16.tflite` | CPU (XNNPACK, 4 threads) | — | 936.2 ms |
|
| 139 |
+
| TFLite `benchmark_model` — `mimi_enc_conv_fp16.tflite` | CPU (XNNPACK, 4 threads) | — | XNNPACK declined the graph |
|
| 140 |
+
| TFLite `benchmark_model` — `mimi_enc_tx_fp16.tflite` | CPU (XNNPACK, 4 threads) | — | 53.1 ms |
|
| 141 |
+
|
| 142 |
+
**Any on-device figure recorded when this model shipped came from a different runtime.** It was taken through LiteRT's own `CompiledModel` accelerator (logcat reports it as `LITERT_CL`), which is the path the Kotlin sample app and the LiteRT API use, and it appears elsewhere on this card. The rows above are the classic TFLite OpenCL delegate, measured with a tool anyone can download and re-run. The two are not comparable, so read the rows above as a reproducible floor rather than as this model's speed on LiteRT.
|
| 143 |
+
|
| 144 |
+
XNNPACK declines these fp16 graphs — it reports `failed to delegate DEPTHWISE_CONV_2D` and then fails to allocate tensors — so there is no usable CPU number. Disabling XNNPACK falls back to reference kernels, which measured about 20× slower than the GPU on models of this size and would not represent CPU inference anyone would ship.
|
| 145 |
+
|