Fix config schema and document XNNPACK int8 quality

README.md

@@ -24,17 +24,40 @@ revision `e08cab65ca0ec38e7826075418411ab90cab4da3`.
 | `small/xnnpack/depth_anything_v3_small_xnnpack_fp32.pte` | xnnpack | fp32 |
 | `small/xnnpack/depth_anything_v3_small_xnnpack_int8.pte` | xnnpack | int8 |
 
+## XNNPACK int8
+
+The INT8 export uses dynamic INT8 activations and per-channel INT8 weights for
+linear layers; convolutions remain floating point. Public inputs and outputs
+remain float32.
+
+Numerical agreement with the float reference on three photos (cats and the two
+official DA3 SOH examples) and four upstream example illustrations:
+
+| Image | Range-normalized RMSE | Correlation |
+| --- | --- | --- |
+| depth_cats | 0.42% | 0.999919 |
+| SOH/000 | 0.46% | 0.999890 |
+| SOH/010 | 0.91% | 0.999848 |
+| demo17 | 1.44% | 0.998027 |
+| demo18 | 0.51% | 0.999704 |
+| demo19 | 1.15% | 0.999797 |
+| demo20 | 1.03% | 0.999547 |
+
+Range-normalized RMSE is RMSE divided by the reference depth range. The worst
+value was 1.44%, with a minimum correlation of 0.998027. These are numerical
+export checks, not depth-accuracy benchmarks.
+
 ## Repository structure
 
 ```
-config.json                                                   39 B
-small/coreml/config.json                                      928 B
-small/coreml/depth_anything_v3_small_coreml_fp16.pte          70.9 MB
-small/vulkan/config.json                                      928 B
-small/vulkan/depth_anything_v3_small_vulkan_fp16.pte          56.4 MB
-small/xnnpack/config.json                                     1.6 kB
-small/xnnpack/depth_anything_v3_small_xnnpack_fp32.pte        106.5 MB
-small/xnnpack/depth_anything_v3_small_xnnpack_int8.pte        40.7 MB
+config.json                                              39 B
+small/coreml/config.json                                 879 B
+small/coreml/depth_anything_v3_small_coreml_fp16.pte     70.9 MB
+small/vulkan/config.json                                 879 B
+small/vulkan/depth_anything_v3_small_vulkan_fp16.pte     56.4 MB
+small/xnnpack/config.json                                1.5 kB
+small/xnnpack/depth_anything_v3_small_xnnpack_fp32.pte   106.5 MB
+small/xnnpack/depth_anything_v3_small_xnnpack_int8.pte   40.7 MB
 ```
 
 ## Compatibility

small/coreml/config.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
   "model": "depth_anything_v3_small",
-  "family": "depth-anything-v3",
+  "family": "depth_anything_v3",
   "capabilities": [
     "depth-estimation"
   ],
@@ -11,8 +11,6 @@
     {
       "file": "depth_anything_v3_small_coreml_fp16.pte",
       "precision": "fp16",
-      "quantized": false,
-      "default": true,
       "methods": {
         "forward": {
           "inputs": [

small/vulkan/config.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
   "model": "depth_anything_v3_small",
-  "family": "depth-anything-v3",
+  "family": "depth_anything_v3",
   "capabilities": [
     "depth-estimation"
   ],
@@ -11,8 +11,6 @@
     {
       "file": "depth_anything_v3_small_vulkan_fp16.pte",
       "precision": "fp16",
-      "quantized": false,
-      "default": true,
       "methods": {
         "forward": {
           "inputs": [

small/xnnpack/config.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
   "model": "depth_anything_v3_small",
-  "family": "depth-anything-v3",
+  "family": "depth_anything_v3",
   "capabilities": [
     "depth-estimation"
   ],
@@ -11,8 +11,6 @@
     {
       "file": "depth_anything_v3_small_xnnpack_fp32.pte",
       "precision": "fp32",
-      "quantized": false,
-      "default": true,
       "methods": {
         "forward": {
           "inputs": [
@@ -42,8 +40,6 @@
     {
       "file": "depth_anything_v3_small_xnnpack_int8.pte",
       "precision": "int8",
-      "quantized": true,
-      "default": false,
       "methods": {
         "forward": {
           "inputs": [