Organize RKNN models by Toolkit2 version

.gitattributes

@@ -37,3 +37,7 @@ models/detect/yolox_nano/rk3576/fp16/model.rknn filter=lfs diff=lfs merge=lfs -t
 models/detect/yolox_nano/rk3576/int8/model.rknn filter=lfs diff=lfs merge=lfs -text
 models/detect/yolox_nano/rk3588/fp16/model.rknn filter=lfs diff=lfs merge=lfs -text
 models/detect/yolox_nano/rk3588/int8/model.rknn filter=lfs diff=lfs merge=lfs -text
+models/v2.5.0/detect/yolox_nano/rk3576/fp16/model.rknn filter=lfs diff=lfs merge=lfs -text
+models/v2.5.0/detect/yolox_nano/rk3576/int8/model.rknn filter=lfs diff=lfs merge=lfs -text
+models/v2.5.0/detect/yolox_nano/rk3588/fp16/model.rknn filter=lfs diff=lfs merge=lfs -text
+models/v2.5.0/detect/yolox_nano/rk3588/int8/model.rknn filter=lfs diff=lfs merge=lfs -text

README.md

@@ -12,8 +12,10 @@ tags:
 
 # YOLOX-Nano-RKNN2
 
-Rockchip Model Zoo conversion of YOLOX Nano for RK3576 and RK3588. YOLOX Nano FP16 and INT8 models for RK3576 and RK3588. Download `model.rknn` and `metadata.yaml` from the same directory under `models/detect/yolox_nano/<chip>/<precision>/`. The [manifest](manifest.json) records each model hash.
+Rockchip Model Zoo conversion of YOLOX Nano for RK3576 and RK3588. YOLOX Nano FP16 and INT8 models for RK3576 and RK3588. Download `model.rknn` and `metadata.yaml` from the same directory under `models/v2.5.0/detect/yolox_nano/<chip>/<precision>/`. The [manifest](manifest.json) records each model hash.
 
 The [original YOLOX Nano checkpoint](https://github.com/Megvii-BaseDetection/YOLOX/releases/download/0.1.1rc0/yolox_nano.pth) is included under [source](tree/main/source). It was exported to ONNX with the [Rockchip YOLOX exporter](https://github.com/airockchip/YOLOX/tree/fc69c1e2ccb30fce7e18250bddc13a68090dca09) and compiled with the [Model Zoo YOLOX converter](https://github.com/airockchip/rknn_model_zoo/tree/837db8e992815ff8647eadb709f07067f4bc3a71/examples/yolox) using RKNN Toolkit2 2.5.0. The conversion changed the model format and quantized the INT8 variant.
 
 YOLOX is from [Megvii-BaseDetection](https://github.com/Megvii-BaseDetection/YOLOX). The [Apache-2.0 license](LICENSE) applies to this distribution.
+
+Models are organized by RKNN Toolkit2 version under `models/v<version>/`. Each model and its metadata share a folder.

manifest.json

@@ -1,6 +1,6 @@
 [
   {
-    "path": "models/detect/yolox_nano/rk3576/fp16",
+    "path": "models/v2.5.0/detect/yolox_nano/rk3576/fp16",
     "chip": "rk3576",
     "precision": "fp16",
     "onnx_sha256": "8f2c3337598150e82b4e4b4d3d1a20c1c4a56e9e0df0623ce95e330683aac309",
@@ -9,7 +9,7 @@
     "zoo_commit": "837db8e992815ff8647eadb709f07067f4bc3a71"
   },
   {
-    "path": "models/detect/yolox_nano/rk3576/int8",
+    "path": "models/v2.5.0/detect/yolox_nano/rk3576/int8",
     "chip": "rk3576",
     "precision": "int8",
     "onnx_sha256": "8f2c3337598150e82b4e4b4d3d1a20c1c4a56e9e0df0623ce95e330683aac309",
@@ -18,7 +18,7 @@
     "zoo_commit": "837db8e992815ff8647eadb709f07067f4bc3a71"
   },
   {
-    "path": "models/detect/yolox_nano/rk3588/fp16",
+    "path": "models/v2.5.0/detect/yolox_nano/rk3588/fp16",
     "chip": "rk3588",
     "precision": "fp16",
     "onnx_sha256": "8f2c3337598150e82b4e4b4d3d1a20c1c4a56e9e0df0623ce95e330683aac309",
@@ -27,7 +27,7 @@
     "zoo_commit": "837db8e992815ff8647eadb709f07067f4bc3a71"
   },
   {
-    "path": "models/detect/yolox_nano/rk3588/int8",
+    "path": "models/v2.5.0/detect/yolox_nano/rk3588/int8",
     "chip": "rk3588",
     "precision": "int8",
     "onnx_sha256": "8f2c3337598150e82b4e4b4d3d1a20c1c4a56e9e0df0623ce95e330683aac309",