Artem Plastinkin commited on
Commit ·
9b7a66d
1
Parent(s): c59daae
Update commands reproduce
Browse files
int8/benchmarks/x5h_mwmx_npu_apm80_12core.yaml
CHANGED
|
@@ -52,15 +52,21 @@ power:
|
|
| 52 |
# for this artifact, so no 1-core reproduce block exists.
|
| 53 |
reproduce:
|
| 54 |
steps:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 55 |
- title: Download the ONNX model and compile config
|
| 56 |
-
command: hf download Renesas/DeepLabV3Plus-R50-ONNX --repo-type=model --include "fp32/*" --include "compile_config/*" --local-dir .
|
| 57 |
- title: Configure the network for the 12-core build (required — this model has no working 1-core config)
|
| 58 |
command: |
|
| 59 |
cat >> ${WORKDIR}/nnac_config/config.yaml << 'EOF'
|
| 60 |
network_configuration:
|
| 61 |
deeplabv3plus_r50_oss_sim_inf:
|
| 62 |
optimization_options:
|
| 63 |
-
- -segmentation_json ./compile_config/l2_seg_48k_add_split_4_split_2.json
|
| 64 |
- -enable_dma_fusion
|
| 65 |
- -enable_stu_multi_channel
|
| 66 |
- -stu_multi_channel_config 0x15212420
|
|
@@ -71,12 +77,12 @@ reproduce:
|
|
| 71 |
# previous model), merge this block into it instead of appending a duplicate key.
|
| 72 |
- title: Compile with the NNAC toolchain (INT8 auto-cast from the FP32 graph)
|
| 73 |
command: |
|
| 74 |
-
python3 nnac_frontend/legalize.py -d binary/nnx ./fp32/deeplabv3plus_r50_oss_sim_inf.onnx
|
| 75 |
- title: Set up the R-Car X5H board
|
| 76 |
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.
|
| 77 |
kind: note
|
| 78 |
- title: Copy the compiled artifact to the board
|
| 79 |
-
command: Copy ${WORKDIR}/binary/nnx to the board -- method may vary (NFS mount, scp, USB, etc.).
|
| 80 |
kind: note
|
| 81 |
- title: Run on R-Car X5H (single NPU cluster, 12 AI cores)
|
| 82 |
command: |
|
|
|
|
| 52 |
# for this artifact, so no 1-core reproduce block exists.
|
| 53 |
reproduce:
|
| 54 |
steps:
|
| 55 |
+
- title: Activate the Python environment
|
| 56 |
+
command: >-
|
| 57 |
+
Activate the Python virtual environment that has the `hf` CLI
|
| 58 |
+
(huggingface_hub) and the NNAC toolchain installed, e.g.
|
| 59 |
+
`source nnac_venv/bin/activate` -- path depends on your toolchain install.
|
| 60 |
+
kind: note
|
| 61 |
- title: Download the ONNX model and compile config
|
| 62 |
+
command: hf download Renesas/DeepLabV3Plus-R50-ONNX --repo-type=model --include "fp32/*" --include "compile_config/*" --local-dir ./DeepLabV3Plus-R50-ONNX-fp32
|
| 63 |
- title: Configure the network for the 12-core build (required — this model has no working 1-core config)
|
| 64 |
command: |
|
| 65 |
cat >> ${WORKDIR}/nnac_config/config.yaml << 'EOF'
|
| 66 |
network_configuration:
|
| 67 |
deeplabv3plus_r50_oss_sim_inf:
|
| 68 |
optimization_options:
|
| 69 |
+
- -segmentation_json ./DeepLabV3Plus-R50-ONNX-fp32/compile_config/l2_seg_48k_add_split_4_split_2.json
|
| 70 |
- -enable_dma_fusion
|
| 71 |
- -enable_stu_multi_channel
|
| 72 |
- -stu_multi_channel_config 0x15212420
|
|
|
|
| 77 |
# previous model), merge this block into it instead of appending a duplicate key.
|
| 78 |
- title: Compile with the NNAC toolchain (INT8 auto-cast from the FP32 graph)
|
| 79 |
command: |
|
| 80 |
+
python3 nnac_frontend/legalize.py -d binary/nnx ./DeepLabV3Plus-R50-ONNX-fp32/fp32/deeplabv3plus_r50_oss_sim_inf.onnx --num-core 12
|
| 81 |
- title: Set up the R-Car X5H board
|
| 82 |
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.
|
| 83 |
kind: note
|
| 84 |
- title: Copy the compiled artifact to the board
|
| 85 |
+
command: Copy ${WORKDIR}/binary/nnx (the working directory from the download/compile steps above) to the board -- method may vary (NFS mount, scp, USB, etc.).
|
| 86 |
kind: note
|
| 87 |
- title: Run on R-Car X5H (single NPU cluster, 12 AI cores)
|
| 88 |
command: |
|