Upload README.md with huggingface_hub

README.md

@@ -7,7 +7,7 @@ tags:
 - on-device
 - image-classification
 ---
-# efficientnet_b1 — ExecuTorch XNNPACK
+# efficientnet_b1 — ExecuTorch
 
 - **Source**: torchvision efficientnet_b1 IMAGENET1K_V2
 - **License**: BSD-3-Clause
@@ -18,17 +18,31 @@ tags:
 
 All variants take and return fp32 tensors — swap the `.pte` file, keep your app code.
 
-| precision | file | size (MB) | parity vs fp32 eager (worst corr) | Mac median (ms)* |
+| build | file | size (MB) | parity vs fp32 eager (worst corr) | Mac median (ms)* |
 |-----------|------|-----------|------------------------------------|------------------|
 | fp32 | `efficientnet_b1_xnnpack_fp32.pte` | 31.2 | 1.000000 | 9.3 |
 | fp16 | `efficientnet_b1_xnnpack_fp16.pte` | 28.8 | 0.999816 | 54.9 |
+| Core ML (fp16, iOS) | `efficientnet_b1_coreml_all.pte` | 16.3 | 0.992817 — see below | 0.6 |
+
+
+The Core ML build is the same graph lowered to Apple's Neural Engine instead of
+XNNPACK, which is CPU-only. On an iPhone 17 Pro, Depth-Anything-V2-Small runs
+500.8 ms through XNNPACK and 42.7 ms through Core ML, at half the file size. It
+computes in fp16 and is iOS-only; the XNNPACK files stay the portable option and
+are what runs on Android.
 
 \*Mac arm64, single process, median of 10 — a reference point for relative cost
 only, not a device number (torch eager fp32 on the same machine: 352.7 ms).
 
-### Precisions that did not earn a slot
+### Checked in the task's own units
+
+Correlation is a first filter. These are the numbers that decide:
+
+- **Core ML (fp16, iOS)** — measured in the units that matter for this model — fraction of images keeping the fp32 top-1 label: 9 of 10 images keep the fp32 top-1 label.
+
+### Builds that did not earn a slot
 
-- **int8 is not shipped**: measured in the units that matter for this model — fraction of images keeping the fp32 top-1 label: median 0.0000 over 10 real images, worst 0.0000.
+- **int8 is not shipped**: measured in the units that matter for this model — fraction of images keeping the fp32 top-1 label: 0 of 10 images keep the fp32 top-1 label.
 
 ## Verification (executorch 1.4.0, torch 2.13.0)
 
@@ -43,5 +57,5 @@ XNNPACK delegate coverage (fp32): 100.0% (410/410 ops)
 
 ## Conversion
 
-torch.export -> to_edge_transform_and_lower(XnnpackPartitioner) -> .pte
+torch.export -> to_edge_transform_and_lower(partitioner) -> .pte
 (conversion scripts: [executorch-models](https://github.com/john-rocky/executorch-models))