Document clean release validation
Browse files- README.md +3 -1
- SHA256SUMS.json +4 -4
- VALIDATION.md +17 -3
README.md
CHANGED
|
@@ -123,7 +123,9 @@ another base. For completely offline use, keep `sapiens2_backbone.py` beside the
|
|
| 123 |
|
| 124 |
See `VALIDATION.md` for the checks performed before publishing this loader.
|
| 125 |
The verified standalone loader is included here. Training, evaluation, TailGuard,
|
| 126 |
-
and dataset tooling are maintained in the accompanying source repository.
|
|
|
|
|
|
|
| 127 |
|
| 128 |
## Attribution
|
| 129 |
|
|
|
|
| 123 |
|
| 124 |
See `VALIDATION.md` for the checks performed before publishing this loader.
|
| 125 |
The verified standalone loader is included here. Training, evaluation, TailGuard,
|
| 126 |
+
and dataset tooling are maintained in the accompanying source repository. The
|
| 127 |
+
private source release candidate `v0.1.0-rc2` includes the complete clean HInt
|
| 128 |
+
validation replay and end-to-end TailGuard command.
|
| 129 |
|
| 130 |
## Attribution
|
| 131 |
|
SHA256SUMS.json
CHANGED
|
@@ -1,7 +1,7 @@
|
|
| 1 |
{
|
| 2 |
"README.md": {
|
| 3 |
-
"bytes":
|
| 4 |
-
"sha256": "
|
| 5 |
},
|
| 6 |
"training/S2HAND-Training-WarmStart-5B/manifest.json": {
|
| 7 |
"bytes": 1092,
|
|
@@ -56,8 +56,8 @@
|
|
| 56 |
"sha256": "1c77a629ee16972ddb9844378038b20321143a6faf0cac26c2c04dc9b61c5288"
|
| 57 |
},
|
| 58 |
"VALIDATION.md": {
|
| 59 |
-
"bytes":
|
| 60 |
-
"sha256": "
|
| 61 |
},
|
| 62 |
"requirements.txt": {
|
| 63 |
"bytes": 51,
|
|
|
|
| 1 |
{
|
| 2 |
"README.md": {
|
| 3 |
+
"bytes": 6112,
|
| 4 |
+
"sha256": "1e215e3e017e631c30fcbc79467d53618a40f021e886fc570c090a45ed068797"
|
| 5 |
},
|
| 6 |
"training/S2HAND-Training-WarmStart-5B/manifest.json": {
|
| 7 |
"bytes": 1092,
|
|
|
|
| 56 |
"sha256": "1c77a629ee16972ddb9844378038b20321143a6faf0cac26c2c04dc9b61c5288"
|
| 57 |
},
|
| 58 |
"VALIDATION.md": {
|
| 59 |
+
"bytes": 2336,
|
| 60 |
+
"sha256": "fd538ec44824b292bede76200b5e12754fef056f3e6dd5ad553c427b5795d2a7"
|
| 61 |
},
|
| 62 |
"requirements.txt": {
|
| 63 |
"bytes": 51,
|
VALIDATION.md
CHANGED
|
@@ -25,6 +25,20 @@ The standalone image pipeline is also covered by unit tests for:
|
|
| 25 |
- RGB/ImageNet input tensor shape and finite normalization; and
|
| 26 |
- 308-channel UDP/DARK decoding back to original-image coordinates.
|
| 27 |
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
- RGB/ImageNet input tensor shape and finite normalization; and
|
| 26 |
- 308-channel UDP/DARK decoding back to original-image coordinates.
|
| 27 |
|
| 28 |
+
On 2026-10-30, the standalone Pose-5B trainer additionally passed a real
|
| 29 |
+
one-step L40S/BF16 optimization and independent checkpoint-reload test. The
|
| 30 |
+
released MT-5B route was checked on a real HInt image: all 268 protected
|
| 31 |
+
heatmaps were bitwise equal to the public route, while all 40 adapted channels
|
| 32 |
+
used the learned route.
|
| 33 |
+
|
| 34 |
+
The clean checkout also replayed all 1,175 hands in the HInt NewDays/EPICK
|
| 35 |
+
validation split. The catastrophic-tail rate matched exactly
|
| 36 |
+
(`0.0144680851`). Selection AUC was `0.7656657143`, versus `0.7656003214` in
|
| 37 |
+
the original research record (absolute difference `0.0000653929`), and
|
| 38 |
+
combined NME differed by `0.0000097603`. This is a near-exact numerical replay,
|
| 39 |
+
not a bitwise metric claim.
|
| 40 |
+
|
| 41 |
+
Finally, TailGuard was exercised with the released MT-5B delta, real NewDays
|
| 42 |
+
images, person detections, and WiLoR-YOLO hand detections. Non-triggered rows
|
| 43 |
+
remained bitwise equal to native predictions, while every triggered row used
|
| 44 |
+
the documented coherent 21-joint conditional crop.
|