Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -21,7 +21,6 @@ All variants take and return fp32 tensors — swap the `.pte` file, keep your ap
|
|
| 21 |
| build | file | size (MB) | parity vs fp32 eager (worst corr) | Mac median (ms)* |
|
| 22 |
|-----------|------|-----------|------------------------------------|------------------|
|
| 23 |
| fp32 | `ssdlite320_mobilenetv3_xnnpack_fp32.pte` | 13.8 | 1.000000 | 5.1 |
|
| 24 |
-
| int8 | `ssdlite320_mobilenetv3_xnnpack_int8.pte` | 3.9 | 0.968820 | 7.1 |
|
| 25 |
| Core ML (fp16, iOS) | `ssdlite320_mobilenetv3_coreml_all.pte` | 7.5 | 0.999658 | 0.9 |
|
| 26 |
|
| 27 |
|
|
@@ -37,13 +36,44 @@ only, not a device number (torch eager fp32 on the same machine: 115.3 ms).
|
|
| 37 |
|
| 38 |
### Checked in the task's own units
|
| 39 |
|
| 40 |
-
|
|
|
|
|
|
|
| 41 |
|
| 42 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
|
| 44 |
### Builds that did not earn a slot
|
| 45 |
|
| 46 |
-
- **fp16 is not shipped**: it comes out at 100% of the fp32 file (13.8 MB vs 13.8 MB), so it buys nothing. XNNPACK serializes convolution weights as fp32 no matter what dtype the graph carries, so on a conv-heavy model fp16 saves no disk and only adds cast operations.
|
| 47 |
|
| 48 |
## Verification (executorch 1.4.0, torch 2.13.0)
|
| 49 |
|
|
@@ -71,13 +101,3 @@ XNNPACK delegate coverage (fp32): 94.8% (289/305 ops); ops left on the portable
|
|
| 71 |
|
| 72 |
torch.export -> to_edge_transform_and_lower(partitioner) -> .pte
|
| 73 |
(conversion scripts: [executorch-models](https://github.com/john-rocky/executorch-models))
|
| 74 |
-
|
| 75 |
-
<!-- funnel:v1 -->
|
| 76 |
-
|
| 77 |
-
---
|
| 78 |
-
|
| 79 |
-
**More models in this format:** [ExecuTorch Model Zoo](https://huggingface.co/collections/mlboydaisuke/executorch-model-zoo-6a7ff328390b63075ffeae5e) — 31 models, each with the recipe that produced it.
|
| 80 |
-
|
| 81 |
-
**Want a different model on-device?** [Open a request](https://github.com/john-rocky/on-device-requests) — free, open weights only; the export and its measured numbers get published publicly.
|
| 82 |
-
|
| 83 |
-
<!-- /funnel:v1 -->
|
|
|
|
| 21 |
| build | file | size (MB) | parity vs fp32 eager (worst corr) | Mac median (ms)* |
|
| 22 |
|-----------|------|-----------|------------------------------------|------------------|
|
| 23 |
| fp32 | `ssdlite320_mobilenetv3_xnnpack_fp32.pte` | 13.8 | 1.000000 | 5.1 |
|
|
|
|
| 24 |
| Core ML (fp16, iOS) | `ssdlite320_mobilenetv3_coreml_all.pte` | 7.5 | 0.999658 | 0.9 |
|
| 25 |
|
| 26 |
|
|
|
|
| 36 |
|
| 37 |
### Checked in the task's own units
|
| 38 |
|
| 39 |
+
The graph stops at the raw heads, so the numbers that decide have to be taken *after* the
|
| 40 |
+
anchor decode and NMS this card tells callers to run — using torchvision's own
|
| 41 |
+
`anchor_generator` and `postprocess_detections`, not a hand-rolled substitute.
|
| 42 |
|
| 43 |
+
An earlier version of this section counted "firing detections", 26,988 of 27,004 across
|
| 44 |
+
10 images. Those were raw anchors above a floor, not objects: 27,004 of them for ten
|
| 45 |
+
photographs that contain perhaps twenty things. Counting anchors measures the background
|
| 46 |
+
agreeing with itself.
|
| 47 |
+
|
| 48 |
+
Measured properly — **71 detections over 32 photographs** of people and animals, held out
|
| 49 |
+
from the `street` images the int8 build was calibrated on, at a score threshold of 0.30:
|
| 50 |
+
|
| 51 |
+
| build | added | dropped | relabelled | worst matched IoU | worst score shift |
|
| 52 |
+
|---|---|---|---|---|---|
|
| 53 |
+
| fp32 | 0 | 0 | 0 | 1.000 | 0.0000 |
|
| 54 |
+
| Core ML | 1 | 0 | 0 | 0.984 | 0.0732 |
|
| 55 |
+
|
| 56 |
+
Core ML's single disagreement sits within its own score error of the threshold — a
|
| 57 |
+
detection the cut is ambivalent about, which a photograph shifted by a pixel moves in
|
| 58 |
+
eager too. Every object it keeps is in the same place: worst matched IoU 0.984.
|
| 59 |
+
|
| 60 |
+
## Withdrawn: int8 (2026-08-27)
|
| 61 |
+
|
| 62 |
+
`ssdlite320_mobilenetv3_xnnpack_int8.pte` was published and has been **withdrawn**.
|
| 63 |
+
Measured the same way it **adds 19 detections, drops 5 and relabels 1** — 25 changes to a
|
| 64 |
+
71-detection answer, better than a third of it — and the boxes it does keep fall to
|
| 65 |
+
**IoU 0.673**. Its worst score shift is **0.4132**, larger than the 0.30 threshold itself.
|
| 66 |
+
|
| 67 |
+
Its correlation was already the warning: **0.968820**, the lowest on this shelf's detection
|
| 68 |
+
models, and it shipped anyway because the number that was supposed to catch it counted
|
| 69 |
+
anchors instead of objects.
|
| 70 |
+
|
| 71 |
+
At 3.9 MB against fp32's 13.8 it was much the smallest build. The Core ML build is 7.5 MB
|
| 72 |
+
and, on this host, the fastest of the three.
|
| 73 |
|
| 74 |
### Builds that did not earn a slot
|
| 75 |
|
| 76 |
+
- **fp16 is not shipped**: it comes out at 100% of the fp32 file (13.8 MB vs 13.8 MB), so it buys nothing. XNNPACK serializes convolution weights as fp32 no matter what dtype the graph carries, so on a conv-heavy model fp16 saves no disk and only adds cast operations. Quantization, not fp16, is the reduction that can shrink a graph like this.
|
| 77 |
|
| 78 |
## Verification (executorch 1.4.0, torch 2.13.0)
|
| 79 |
|
|
|
|
| 101 |
|
| 102 |
torch.export -> to_edge_transform_and_lower(partitioner) -> .pte
|
| 103 |
(conversion scripts: [executorch-models](https://github.com/john-rocky/executorch-models))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|