Artem Plastinkin commited on
Commit ·
9af571b
1
Parent(s): 569ac43
Move network_config in dedicated file
Browse files
compile_config/network_config.yaml
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
optimization_options:
|
| 2 |
+
- -segmentation_json ./DeepLabV3Plus-R50-ONNX-fp32/compile_config/l2_seg_48k_add_split_4_split_2.json
|
| 3 |
+
- -enable_dma_fusion
|
| 4 |
+
- -enable_stu_multi_channel
|
| 5 |
+
- -stu_multi_channel_config 0x15212420
|
| 6 |
+
- -stu_weight
|
| 7 |
+
- -distributed_coef_stu
|
int8/benchmarks/x5h_mwmx_npu_apm80_12core.yaml
CHANGED
|
@@ -43,24 +43,14 @@ memory:
|
|
| 43 |
power:
|
| 44 |
avg_w: null
|
| 45 |
|
| 46 |
-
network_configuration:
|
| 47 |
-
key: deeplabv3plus_r50_oss_sim_inf
|
| 48 |
-
entry_yaml: |
|
| 49 |
-
optimization_options:
|
| 50 |
-
- -segmentation_json ./DeepLabV3Plus-R50-ONNX-fp32/compile_config/l2_seg_48k_add_split_4_split_2.json
|
| 51 |
-
- -enable_dma_fusion
|
| 52 |
-
- -enable_stu_multi_channel
|
| 53 |
-
- -stu_multi_channel_config 0x15212420
|
| 54 |
-
- -stu_weight
|
| 55 |
-
- -distributed_coef_stu
|
| 56 |
-
|
| 57 |
# Exact commands verified against the NNAC "Getting Started" chapter. Rendered
|
| 58 |
# by the AI-Dashboard in place of the generic placeholder flow — see
|
| 59 |
# downloadRunHTML() / parse_reproduce() in AI-Dashboard/app.js.
|
| 60 |
# CAVEAT: this artifact is segment "split_2" of a 4-way split network (see
|
| 61 |
# README) — these steps reproduce only this segment's latency, not an
|
| 62 |
# end-to-end DeepLabV3+ result. The 1-AI-core config also fails to compile
|
| 63 |
-
# for this artifact, so no 1-core reproduce block exists.
|
|
|
|
| 64 |
reproduce:
|
| 65 |
steps:
|
| 66 |
- title: Activate the Python environment
|
|
@@ -71,20 +61,9 @@ reproduce:
|
|
| 71 |
kind: note
|
| 72 |
- title: Download the ONNX model and compile config
|
| 73 |
command: hf download Renesas/DeepLabV3Plus-R50-ONNX --repo-type model --include "fp32/*" "compile_config/*" --local-dir ./DeepLabV3Plus-R50-ONNX-fp32
|
| 74 |
-
- title: "Insert this model's entry into network_configuration: in nnac_config/config.yaml (required — this model has no working 1-core config)"
|
| 75 |
-
command: |
|
| 76 |
-
deeplabv3plus_r50_oss_sim_inf:
|
| 77 |
-
optimization_options:
|
| 78 |
-
- -segmentation_json ./DeepLabV3Plus-R50-ONNX-fp32/compile_config/l2_seg_48k_add_split_4_split_2.json
|
| 79 |
-
- -enable_dma_fusion
|
| 80 |
-
- -enable_stu_multi_channel
|
| 81 |
-
- -stu_multi_channel_config 0x15212420
|
| 82 |
-
- -stu_weight
|
| 83 |
-
- -distributed_coef_stu
|
| 84 |
-
kind: snippet
|
| 85 |
- title: Compile with the NNAC toolchain (INT8 auto-cast from the FP32 graph)
|
| 86 |
command: |
|
| 87 |
-
python3 nnac_frontend/legalize.py -d binary/nnx ./DeepLabV3Plus-R50-ONNX-fp32/fp32/deeplabv3plus_r50_oss_sim_inf.onnx --num-core 12
|
| 88 |
- title: Set up the R-Car X5H board
|
| 89 |
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.
|
| 90 |
kind: note
|
|
|
|
| 43 |
power:
|
| 44 |
avg_w: null
|
| 45 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
# Exact commands verified against the NNAC "Getting Started" chapter. Rendered
|
| 47 |
# by the AI-Dashboard in place of the generic placeholder flow — see
|
| 48 |
# downloadRunHTML() / parse_reproduce() in AI-Dashboard/app.js.
|
| 49 |
# CAVEAT: this artifact is segment "split_2" of a 4-way split network (see
|
| 50 |
# README) — these steps reproduce only this segment's latency, not an
|
| 51 |
# end-to-end DeepLabV3+ result. The 1-AI-core config also fails to compile
|
| 52 |
+
# for this artifact, so no 1-core reproduce block exists. The network config
|
| 53 |
+
# below is required for this model — there is no working default.
|
| 54 |
reproduce:
|
| 55 |
steps:
|
| 56 |
- title: Activate the Python environment
|
|
|
|
| 61 |
kind: note
|
| 62 |
- title: Download the ONNX model and compile config
|
| 63 |
command: hf download Renesas/DeepLabV3Plus-R50-ONNX --repo-type model --include "fp32/*" "compile_config/*" --local-dir ./DeepLabV3Plus-R50-ONNX-fp32
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 64 |
- title: Compile with the NNAC toolchain (INT8 auto-cast from the FP32 graph)
|
| 65 |
command: |
|
| 66 |
+
python3 nnac_frontend/legalize.py -d binary/nnx ./DeepLabV3Plus-R50-ONNX-fp32/fp32/deeplabv3plus_r50_oss_sim_inf.onnx --num-core 12 --network-config ./DeepLabV3Plus-R50-ONNX-fp32/compile_config/network_config.yaml
|
| 67 |
- title: Set up the R-Car X5H board
|
| 68 |
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.
|
| 69 |
kind: note
|