Card: correct the GPU node counts — the previous figures added the XNNPACK fallback partition to the GPU delegate partition

README.md

@@ -116,10 +116,10 @@ Measured on a **Pixel 8a** (Tensor G3, Android 16) with the standard TFLite [`be
 
 | Runtime | Backend | Graph on GPU | Latency |
 |---|---|---|---|
-| TFLite `benchmark_model` (`TfLiteGpuDelegateV2`) — `ram_taghead_fp16.tflite` | GPU (OpenCL) | 139 / 249 | 3499.7 ms |
-| TFLite `benchmark_model` (`TfLiteGpuDelegateV2`) — `ram_reweight_fp16.tflite` | GPU (OpenCL) | 17 / 28 | 612.0 ms |
+| TFLite `benchmark_model` (`TfLiteGpuDelegateV2`) — `ram_taghead_fp16.tflite` | GPU (OpenCL) | 30 / 139 | 3499.7 ms |
+| TFLite `benchmark_model` (`TfLiteGpuDelegateV2`) — `ram_reweight_fp16.tflite` | GPU (OpenCL) | 9 / 18 | 612.0 ms |
 | TFLite `benchmark_model` (`TfLiteGpuDelegateV2`) — `ram_swin_s012_fp16.tflite` | GPU (OpenCL) | 105 / 2129 | did not run |
-| TFLite `benchmark_model` (`TfLiteGpuDelegateV2`) — `ram_stage3_tail_fp16.tflite` | GPU (OpenCL) | 174 / 316 | 523.5 ms |
+| TFLite `benchmark_model` (`TfLiteGpuDelegateV2`) — `ram_stage3_tail_fp16.tflite` | GPU (OpenCL) | 41 / 178 | 523.5 ms |
 | TFLite `benchmark_model` — `ram_taghead_fp16.tflite` | CPU (XNNPACK, 4 threads) | — | 1550.1 ms |
 | TFLite `benchmark_model` — `ram_reweight_fp16.tflite` | CPU (XNNPACK, 4 threads) | — | 312.8 ms |
 | TFLite `benchmark_model` — `ram_swin_s012_fp16.tflite` | CPU (XNNPACK, 4 threads) | — | 2786.9 ms |
@@ -129,5 +129,5 @@ Measured on a **Pixel 8a** (Tensor G3, Android 16) with the standard TFLite [`be
 
 On this delegate the CPU is the faster choice for `ram_taghead_fp16.tflite` (1550.1 ms on CPU against 3499.7 ms on GPU), `ram_reweight_fp16.tflite` (312.8 ms on CPU against 612.0 ms on GPU), `ram_stage3_tail_fp16.tflite` (217.0 ms on CPU against 523.5 ms on GPU) — worth knowing before you reach for the GPU on a mid-range phone.
 
-Note that the GPU does not take the whole graph here (139 of 249 in `ram_taghead_fp16.tflite`, 17 of 28 in `ram_reweight_fp16.tflite`, 105 of 2129 in `ram_swin_s012_fp16.tflite`, 174 of 316 in `ram_stage3_tail_fp16.tflite`); the remainder runs on the CPU and the split costs a per-partition round trip.
+Note that the GPU does not take the whole graph here (30 / 139 in `ram_taghead_fp16.tflite`, 9 / 18 in `ram_reweight_fp16.tflite`, 105 / 2129 in `ram_swin_s012_fp16.tflite`, 41 / 178 in `ram_stage3_tail_fp16.tflite`); the remainder runs on the CPU and the split costs a per-partition round trip.