Add model

README.md

@@ -15,8 +15,6 @@ tags:
 
 # DeepLabV3Plus-R50 (ONNX) – Renesas X5H
 
-> ⏳ **Model file not yet uploaded.** Benchmark results on this page were published ahead of the model weights — see **Provided Artifacts** below. Download/deployment steps will not work until the file is added to this repository.
-
 > **⚠️ Partial-model caveat.** The source checkpoint name ends in
 > `custom_seg_split_4_split_2` — this artifact is **one segment of a 4-way split network**, not
 > the full end-to-end DeepLabV3+ model. The latency below reflects only that segment; do not
@@ -41,7 +39,7 @@ The FP32 ONNX model is auto-cast to **INT8** by the Renesas MWMX toolchain at co
 separate quantization step is required.
 
 ```
-deeplabv3plus_..._split_2.onnx (FP32)
+deeplabv3plus_r50_oss_sim_inf.onnx (FP32, split_2 segment)
         │
         └─▶  MWMX Runtime  ──▶  INT8 auto-cast  ──▶  NPX6 NPU
 ```
@@ -50,7 +48,7 @@ deeplabv3plus_..._split_2.onnx (FP32)
 
 | Artifact | Status | Notes |
 |----------|--------|-------|
-| **FP32 (ONNX)** | ⏳ Not yet uploaded | Benchmark numbers below exist; the model file has not been published to this repo yet |
+| **FP32 (ONNX)** | ✅ Published | `fp32/deeplabv3plus_r50_oss_sim_inf.onnx` — segment `split_2` of 4; auto-cast to INT8 by the MWMX toolchain at compile time (see Deployment Flow above); no separate INT8 file is shipped |
 
 ## Performance
 
@@ -65,6 +63,8 @@ Measured on **Renesas R-Car X5H** via the MWMX runtime (APM80 ship-performance C
 |---------|-----------|--------|---------------|------|
 | MWMX Runtime | INT8 (auto) | X5H · 1× NPU · 12 Cores · 850 MHz | 38.104 | Measured |
 
+> Reconfirmed: 1-core compile still fails as of the 2026-09-16 benchmark run.
+
 ### Accuracy (mIoU)
 
 TBD — not yet measured/published for this repo.
@@ -88,11 +88,13 @@ To run inference on Renesas R-Car X5H, you need:
 
 1. **Renesas R-Car X5H board** with NPX6 NPU
 2. **Renesas MWMX Runtime**
-3. **Hugging Face CLI** to download the model (once the model file is published)
+3. **Hugging Face CLI** to download the model
 
 ## Download
 
-TBD — model file not yet published to this repository.
+```bash
+hf download Renesas/DeepLabV3Plus-R50-ONNX --repo-type=model --include "fp32/*"
+```
 
 ---
 

compile_config/l2_seg_48k_add_split_4_split_2.json

@@ -0,0 +1,233 @@
+{
+  "segments": [
+    {
+      "name": "Segment",
+      "inputs": [
+        "input"
+      ],
+      "outputs": [
+        "/backbone/maxpool/MaxPool_output_0"
+      ],
+      "n_buffer": 2
+    },
+    {
+      "name": "Segment_0",
+      "inputs": [
+        "/backbone/maxpool/MaxPool_output_0"
+      ],
+      "outputs": [
+        "/backbone/layer1/layer1.2/relu_2/Relu_output_0"
+      ],
+      "n_buffer": 2
+    },
+    {
+      "name": "Segment_layer2.{0,1,2,3}",
+      "inputs": [
+        "/backbone/layer1/layer1.2/relu_2/Relu_output_0"
+      ],
+      "outputs": [
+        "/backbone/layer2/layer2.3/relu_2/Relu_output_0"
+      ],
+      "n_buffer": 2
+    },
+    {
+      "name": "Segment_layer3.{0,1,2}",
+      "inputs": [
+        "/backbone/layer2/layer2.3/relu_2/Relu_output_0"
+      ],
+      "outputs": [
+        "/backbone/layer3/layer3.2/relu_2/Relu_output_0"
+      ],
+      "n_buffer": 2
+    },
+    {
+      "name": "Segment_layer3.{3,4,5}",
+      "inputs": [
+        "/backbone/layer3/layer3.2/relu_2/Relu_output_0"
+      ],
+      "outputs": [
+        "/backbone/layer3/layer3.5/relu_2/Relu_output_0"
+      ],
+      "n_buffer": 2
+    },
+    {
+      "name": "Segment_layer4.0",
+      "inputs": [
+        "/backbone/layer3/layer3.5/relu_2/Relu_output_0"
+      ],
+      "outputs": [
+        "/backbone/layer4/layer4.0/relu_2/Relu_output_0"
+      ],
+      "n_buffer": 2
+    },
+    {
+      "name": "Segment_aspp_modules.1",
+      "inputs": [
+        "/backbone/layer4/layer4.2/relu_2/Relu_output_0"
+      ],
+      "outputs": [
+        "/aspp_modules/aspp_modules.1/depthwise_conv/activate/Relu_output_0"
+      ],
+      "n_buffer": 1
+    },
+    {
+      "name": "Segment_aspp_modules.2",
+      "inputs": [
+        "/backbone/layer4/layer4.2/relu_2/Relu_output_0"
+      ],
+      "outputs": [
+        "/aspp_modules/aspp_modules.2/depthwise_conv/activate/Relu_output_0"
+      ],
+      "n_buffer": 1
+    },
+    {
+      "name": "Segment_bottleneck_0",
+      "inputs": [
+        "/Concat_1_output_0_2",
+        "/bottleneck/conv/Conv_output_0_1"
+      ],
+      "outputs": [
+        "/bottleneck/activate/Relu_output_0"
+      ],
+      "n_buffer": 1
+    },
+    {
+      "name": "Segment_bottleneck_1",
+      "inputs": [
+        "/Concat_1_output_0_1"
+      ],
+      "outputs": [
+        "/bottleneck/conv/Conv_output_0_1"
+      ],
+      "n_buffer": 1
+    },
+    {
+      "name": "Segment_output",
+      "inputs": [
+        "/Concat_3_output_0"
+      ],
+      "outputs": [
+        "/conv_seg/Conv_output_0"
+      ],
+      "n_buffer": 2
+    },
+    {
+      "name": "seg_11",
+      "inputs": [
+        "/backbone/layer1/layer1.2/relu_2/Relu_output_0"
+      ],
+      "outputs": [
+        "/c1_bottleneck/activate/Relu_output_0"
+      ],
+      "n_buffer": 1
+    },
+    {
+      "name": "seg_12",
+      "inputs": [
+        "/backbone/layer4/layer4.0/relu_2/Relu_output_0"
+      ],
+      "outputs": [
+        "/backbone/layer4/layer4.2/relu_2/Relu_output_0"
+      ],
+      "n_buffer": 1
+    },
+    {
+      "name": "seg_14",
+      "inputs": [
+        "/image_pool/image_pool.0/GlobalAveragePool_output_0"
+      ],
+      "outputs": [
+        "/image_pool/image_pool.1/activate/Relu_output_0"
+      ],
+      "n_buffer": 1
+    },
+    {
+      "name": "seg_16",
+      "inputs": [
+        "/backbone/layer4/layer4.2/relu_2/Relu_output_0"
+      ],
+      "outputs": [
+        "/aspp_modules/aspp_modules.0/activate/Relu_output_0"
+      ],
+      "n_buffer": 1
+    },
+    {
+      "name": "seg_17",
+      "inputs": [
+        "/aspp_modules/aspp_modules.1/depthwise_conv/activate/Relu_output_0"
+      ],
+      "outputs": [
+        "/aspp_modules/aspp_modules.1/pointwise_conv/activate/Relu_output_0"
+      ],
+      "n_buffer": 1
+    },
+    {
+      "name": "seg_19",
+      "inputs": [
+        "/aspp_modules/aspp_modules.2/depthwise_conv/activate/Relu_output_0"
+      ],
+      "outputs": [
+        "/aspp_modules/aspp_modules.2/pointwise_conv/activate/Relu_output_0"
+      ],
+      "n_buffer": 2
+    },
+    {
+      "name": "seg_18",
+      "inputs": [
+        "/aspp_modules/aspp_modules.3/depthwise_conv/activate/Relu_output_0"
+      ],
+      "outputs": [
+        "/aspp_modules/aspp_modules.3/pointwise_conv/activate/Relu_output_0"
+      ],
+      "n_buffer": 2
+    },
+    {
+      "name": "Segment_27",
+      "inputs": [
+        "/backbone/layer4/layer4.2/relu_2/Relu_output_0_slice_1"
+      ],
+      "outputs": [
+        "/aspp_modules/aspp_modules.3/depthwise_conv/activate/Relu_1_out"
+      ],
+      "n_buffer": 1
+    },
+    {
+      "name": "Segment_29",
+      "inputs": [
+        "/backbone/layer4/layer4.2/relu_2/Relu_output_0_slice_2"
+      ],
+      "outputs": [
+        "/aspp_modules/aspp_modules.3/depthwise_conv/activate/Relu_2_out"
+      ],
+      "n_buffer": 1
+    },
+    {
+      "name": "Segment_31",
+      "inputs": [
+        "/backbone/layer4/layer4.2/relu_2/Relu_output_0_slice_3"
+      ],
+      "outputs": [
+        "/aspp_modules/aspp_modules.3/depthwise_conv/activate/Relu_3_out"
+      ],
+      "n_buffer": 1
+    },
+    {
+      "name": "Segment_33",
+      "inputs": [
+        "/backbone/layer4/layer4.2/relu_2/Relu_output_0_slice_4"
+      ],
+      "outputs": [
+        "/aspp_modules/aspp_modules.3/depthwise_conv/activate/Relu_4_out"
+      ],
+      "n_buffer": 1
+    }
+  ],
+   "tiling_info": {
+      "/Resize_1_output_0": {
+        "l1_tile_shape" : [1,16,64,128]
+      },
+      "/Resize_2_output_0": {
+        "l1_tile_shape" : [1,32,172,19]
+      }
+   }
+}

fp32/.metadata.yaml

@@ -0,0 +1,5 @@
+variant:
+  id: onnx_fp32
+  format: onnx
+  precision: fp32
+  method: onnx_export        # exported from the OpenMMLab checkpoint, no quantization

fp32/deeplabv3plus_r50_oss_sim_inf.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc33270e904d85a8ddcd85d8c4603f577f68255d272a1b8635a0e783c4499f81
+size 165154945

int8/benchmarks/x5h_mwmx_npu_apm80_12core.yaml

@@ -42,3 +42,46 @@ memory:
 
 power:
   avg_w: null
+
+# Exact commands verified against the NNAC "Getting Started" chapter. Rendered
+# by the AI-Dashboard in place of the generic placeholder flow — see
+# downloadRunHTML() / parse_reproduce() in AI-Dashboard/app.js.
+# CAVEAT: this artifact is segment "split_2" of a 4-way split network (see
+# README) — these steps reproduce only this segment's latency, not an
+# end-to-end DeepLabV3+ result. The 1-AI-core config also fails to compile
+# for this artifact, so no 1-core reproduce block exists.
+reproduce:
+  steps:
+    - title: Download the ONNX model and compile config
+      command: hf download Renesas/DeepLabV3Plus-R50-ONNX --repo-type=model --include "fp32/*" --include "compile_config/*"
+    - title: Configure the network for the 12-core build (required — this model has no working 1-core config)
+      command: |
+        cat >> ${WORKDIR}/nnac_config/config.yaml << 'EOF'
+        network_configuration:
+          deeplabv3plus_r50_oss_sim_inf:
+            optimization_options:
+              - -segmentation_json ./compile_config/l2_seg_48k_add_split_4_split_2.json
+              - -enable_dma_fusion
+              - -enable_stu_multi_channel
+              - -stu_multi_channel_config 0x15212420
+              - -stu_weight
+              - -distributed_coef_stu
+        EOF
+        # If config.yaml already has a network_configuration: key (e.g. from a
+        # previous model), merge this block into it instead of appending a duplicate key.
+    - title: Compile with the NNAC toolchain (INT8 auto-cast from the FP32 graph)
+      command: |
+        python3 nnac_frontend/legalize.py -d binary/nnx ./fp32/deeplabv3plus_r50_oss_sim_inf.onnx
+    - title: Set up the R-Car X5H board
+      command: Configure the board per the AI Compiler (NNAC) "Getting Started" guide, section 3.4 (host TFTP/NFS setup, bootloader flashing, U-Boot, Linux boot, login) -- exact steps depend on your board/network setup.
+    - title: Copy the compiled artifact to the board
+      command: Copy ${WORKDIR}/binary/nnx to the board -- method may vary (NFS mount, scp, USB, etc.).
+    - title: Run on R-Car X5H (single NPU cluster, 12 AI cores)
+      command: |
+        cd binary
+        ./host_app ./arc_prog_npus ./nnx/deeplabv3plus_r50_oss_sim_inf
+  notes: >-
+    hash[n] = 0x...(OK) means the run's output matches the reference hash in
+    hash.txt; latency is the NPX execution time reported in cycles and ms.
+    This is one segment (split_2 of 4) of the full segmentation pipeline —
+    not whole-model latency.