# Quick Start(M57) **This section applies to the following platforms:** - M57 This section introduces the basic operations of `ONNX` model conversion, and uses the `pulsar2` tool to compile the `ONNX` model into the `axmodel` model. Please refer to the {ref}`《Development Environment Preparation》 ` section to complete the development environment setup. The example model in this section is the open source model `MobileNetv2`. ## Pulsar2 toolchain command description The function commands in the `Pulsar2` toolchain start with `pulsar2`. The commands that are most relevant to users are `pulsar2 build`, `pulsar2 run` and `pulsar2 version`. - `pulsar2 build` is used to convert the `onnx` model to the `axmodel` format model - `pulsar2 run` is used to run the simulation after the model is converted - `pulsar2 version` can be used to view the version information of the current toolchain, which is usually required when reporting issues ```shell root@xxx:/data# pulsar2 --help usage: pulsar2 [-h] {version,build,run} ... positional arguments: {version,build,run} optional arguments: -h, --help show this help message and exit ``` ## Model compilation configuration file description `mobilenet_v2_build_config.json` in the `/data/config/` path shows: ```shell { "model_type": "ONNX", "npu_mode": "NPU1", "quant": { "input_configs": [ { "tensor_name": "input", "calibration_dataset": "./dataset/imagenet-32-images.tar", "calibration_size": 32, "calibration_mean": [103.939, 116.779, 123.68], "calibration_std": [58.0, 58.0, 58.0] } ], "calibration_method": "MinMax", "precision_analysis": false }, "input_processors": [ { "tensor_name": "input", "tensor_format": "BGR", "src_format": "BGR", "src_dtype": "U8", "src_layout": "NHWC", "csc_mode": "NoCSC" } ], "compiler": { "check": 0 } } ``` ::::{attention} The `tensor_name` field in `input_processors`, `output_processors` and `quant` nodes under `input_configs` needs to be set according to the actual input/output node name of the model, or it can be set to `DEFAULT` to indicate that the current configuration applies to all inputs or outputs. :::{figure} ../media/tensor_name.png :align: center :alt: pipeline ::: :::: For more details, please refer to {ref}`Configuration File Detailed Description `. (model_compile_m57)= ## Compile and execute Take `mobilenetv2-sim.onnx` as an example, execute the following `pulsar2 build` command to compile and generate `compiled.axmodel`: ```shell pulsar2 build --target_hardware M57 --input model/mobilenetv2-sim.onnx --output_dir output --config config/mobilenet_v2_build_config.json ``` :::{warning} Before compiling the model, you need to make sure that the original model has been optimized using the `onnxsim` tool. The main purpose is to convert the model into a static graph that is more conducive to `Pulsar2` compilation and obtain better inference performance. There are two ways: 1. Directly execute the command inside the `Pulsar2` docker: `onnxsim in.onnx out.onnx`. 2. When using `pulsar2 build` to convert the model, add the parameter: `--onnx_opt.enable_onnxsim true` (the default value is false). If you want to learn more about `onnxsim`, you can visit the [official website](https://github.com/daquexian/onnx-simplifier). ::: ### log reference information ``` $ pulsar2 build --target_hardware M57 --input model/mobilenetv2-sim.onnx --output_dir output --config config/mobilenet_v2_build_config.json 2025-03-14 14:45:26.362 | WARNING | yamain.command.build:fill_default:265 - apply default output processor configuration to ['output'] 2025-03-14 14:45:26.362 | WARNING | yamain.command.build:fill_default:340 - ignore input csc config because of src_format is AutoColorSpace or src_format and tensor_format are the same 2025-03-14 14:45:26.363 | INFO | yamain.common.util:extract_archive:217 - extract [dataset/imagenet-32-images.tar] to [output/quant/dataset/input]... 32 File(s) Loaded. Building onnx ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% 0:00:00 2025-03-14 14:45:27.529 | INFO | yamain.command.build:quant:748 - save optimized onnx to [output/frontend/optimized.onnx] Quant Config Table ┏━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━┓ ┃ Input ┃ Shape ┃ Dataset Directory ┃ Data Format ┃ Tensor Format ┃ Mean ┃ Std ┃ ┡━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━┩ │ input │ [1, 3, 224, 224] │ output/quant/dataset/input │ Image │ BGR │ [103.93900299072266, 116.77899932861328, 123.68000030517578] │ [58.0, 58.0, 58.0] │ └───────┴──────────────────┴────────────────────────────┴─────────────┴───────────────┴──────────────────────────────────────────────────────────────┴────────────────────┘ Transformer optimize level: 0 32 File(s) Loaded. Stastic Inf tensor: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 9.41it/s] [14:45:28] AX Set Float Op Table Pass Running ... [14:45:29] AX Set MixPrecision Pass Running ... [14:45:29] AX Set LN Quant dtype Quant Pass Running ... [14:45:29] AX Reset Mul Config Pass Running ... [14:45:29] AX Refine Operation Config Pass Running ... [14:45:29] AX Tanh Operation Format Pass Running ... [14:45:29] AX Confused Op Refine Pass Running ... [14:45:29] AX Quantization Fusion Pass Running ... [14:45:29] AX Quantization Simplify Pass Running ... [14:45:29] AX Parameter Quantization Pass Running ... [14:45:29] AX Runtime Calibration Pass Running ... Calibration Progress(Phase 1): 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 32/32 [00:03<00:00, 8.86it/s] [14:45:33] AX Quantization Alignment Pass Running ... [14:45:33] AX Refine Int Parameter Pass Running ... [14:45:33] AX Refine Scale Pass Running ... [14:45:33] AX Passive Parameter Quantization Running ... [14:45:33] AX Parameter Baking Pass Running ... --------- Network Snapshot --------- Num of Op: [100] Num of Quantized Op: [100] Num of Variable: [278] Num of Quantized Var: [278] ------- Quantization Snapshot ------ Num of Quant Config: [387] BAKED: [53] OVERLAPPED: [145] ACTIVATED: [65] SOI: [1] PASSIVE_BAKED: [53] FP32: [70] Network Quantization Finished. Do quant optimization quant.axmodel export success: /opt/pulsar2/quick_start_example/output/quant/quant_axmodel.onnx /opt/pulsar2/quick_start_example/output/quant/quant_axmodel.data ===>export io data to folder: output/quant/debug/io Building native ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% 0:00:00 2025-03-14 14:45:35.003 | INFO | yamain.command.build:compile_ptq_model:1029 - group 0 compiler transformation 2025-03-14 14:45:35.005 | WARNING | yamain.command.load_model:pre_process:616 - preprocess tensor [input] 2025-03-14 14:45:35.005 | INFO | yamain.command.load_model:pre_process:617 - tensor: input, (1, 224, 224, 3), U8 2025-03-14 14:45:35.005 | INFO | yamain.command.load_model:pre_process:617 - op: op:pre_dequant_1, AxDequantizeLinear, {'const_inputs': {'x_zeropoint': array(0, dtype=int32), 'x_scale': array(1., dtype=float32)}, 'output_dtype': , 'quant_method': 0} 2025-03-14 14:45:35.005 | INFO | yamain.command.load_model:pre_process:617 - tensor: tensor:pre_norm_1, (1, 224, 224, 3), FP32 2025-03-14 14:45:35.005 | INFO | yamain.command.load_model:pre_process:617 - op: op:pre_norm_1, AxNormalize, {'dim': 3, 'mean': [103.93900299072266, 116.77899932861328, 123.68000030517578], 'std': [58.0, 58.0, 58.0], 'output_dtype': FP32} 2025-03-14 14:45:35.006 | INFO | yamain.command.load_model:pre_process:617 - tensor: tensor:pre_transpose_1, (1, 224, 224, 3), FP32 2025-03-14 14:45:35.006 | INFO | yamain.command.load_model:pre_process:617 - op: op:pre_transpose_1, AxTranspose, {'perm': [0, 3, 1, 2]} 2025-03-14 14:45:35.006 | WARNING | yamain.command.load_model:post_process:638 - postprocess tensor [output] 2025-03-14 14:45:35.006 | INFO | yamain.command.load_model:ir_compiler_transformation:824 - use random data as gt input: input, uint8, (1, 224, 224, 3) 2025-03-14 14:45:35.209 | INFO | yamain.command.build:compile_ptq_model:1052 - group 0 QuantAxModel macs: 300,774,272 2025-03-14 14:45:35.220 | INFO | yamain.command.build:compile_ptq_model:1182 - subgraph [0], group: 0, type: GraphType.NPU 2025-03-14 14:45:35.221 | INFO | yamain.command.npu_backend_compiler:compile:174 - compile npu subgraph [0] tiling op... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 68/68 0:00:00 new_ddr_tensor = [] build op serially... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 149/149 0:00:00 build op... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 268/268 0:00:00 add ddr swap... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 344/344 0:00:00 calc input dependencies... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 576/576 0:00:00 calc output dependencies... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 576/576 0:00:00 assign eu heuristic ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 576/576 0:00:00 assign eu onepass ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 576/576 0:00:00 assign eu greedy ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 576/576 0:00:00 2025-03-14 14:45:36.965 | INFO | yasched.test_onepass:results2model:2682 - clear job deps 2025-03-14 14:45:36.966 | INFO | yasched.test_onepass:results2model:2683 - max_cycle = 684,124 build jobs ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 576/576 0:00:00 2025-03-14 14:45:37.179 | INFO | yamain.command.npu_backend_compiler:compile:235 - assemble model [0] [subgraph_npu_0] b1 2025-03-14 14:45:37.931 | INFO | yamain.command.build:compile_ptq_model:1221 - fuse 1 subgraph(s) ``` :::{note} The host configuration for this example is: > - Intel(R) Xeon(R) Gold 6336Y CPU @ 2.40GHz > - Memory 32G The whole process takes about `12s`, and the conversion time of hosts with different configurations may vary slightly. ::: ### 模型编译输出文件说明 ```shell root@xxx:/data# tree output/ output/ |-- build_context.json |-- compiled.axmodel # Model will be run on the board |-- compiler # Compiler backend intermediate results and debug information | `-- debug # Front-end graph optimization intermediate results and debug information | `-- subgraph_npu_0 | `-- b1 |-- frontend | |-- optimized.data | `-- optimized.onnx # Input model: floating point ONNX model after graph optimization `-- quant # Quantization tool output and debug information directory |-- dataset | `-- input | |-- ILSVRC2012_val_00000001.JPEG | |-- ...... | `-- ILSVRC2012_val_00000032.JPEG |-- debug | `-- io | |-- float | | |-- input.npy | | `-- output.npy | `-- quant | |-- input.npy | `-- output.npy |-- quant_axmodel.data |-- quant_axmodel.json # Quantitative configuration information `-- quant_axmodel.onnx # Quantized model, QuantAxModel ``` Among them, `compiled.axmodel` is the `.axmodel` model file that can be run on the board generated by the final compilation ::::{note} Since `.axmodel` is developed based on the **ONNX** model storage format, changing the `.axmodel` file suffix to `.axmodel.onnx` can support being directly opened by the network model graphical tool **Netron**. > :::{figure} ../media/axmodel-netron.png > :align: center > :alt: pipeline > ::: :::: #### Model information query You can use onnx inspect --io \${axmodel/onnx_path} to view the input and output information of the compiled axmodel model. You can also use -m -n -t to view the meta / node / tensor information in the model. ```shell root@xxx:/data# onnx inspect -m -n -t output/compiled.axmodel Failed to check model output/compiled.axmodel, statistic could be inaccurate! Meta information -------------------------------------------------------------------------------- IR Version: 10 Opset Import: [domain: "" version: 18 ] Producer name: Pulsar2 Producer version: Domain: Doc string: Pulsar2 Version: 3.4 Pulsar2 Commit: 3dfd5692 meta.{} = {} extra_data CgsKBWlucHV0EAEYAgoICgZvdXRwdXQSATEaQwoOc3ViZ3JhcGhfbnB1XzBSMQoVc3ViZ3JhcGhfbnB1XzBfYjFfbmV1EAEaFgoGcGFyYW1zGgxucHVfMF9wYXJhbXMiACgE Node information -------------------------------------------------------------------------------- Node type "neu mode" has: 1 -------------------------------------------------------------------------------- Node "subgraph_npu_0": type "neu mode", inputs "['input']", outputs "['output']" Tensor information -------------------------------------------------------------------------------- ValueInfo "input": type UINT8, shape [1, 224, 224, 3], ValueInfo "npu_0_params": type UINT8, shape [4324276], ValueInfo "subgraph_npu_0_b1_neu": type UINT8, shape [122368], ValueInfo "output": type FLOAT, shape [1, 1000], Initializer "npu_0_params": type UINT8, shape [4324276], Initializer "subgraph_npu_0_b1_neu": type UINT8, shape [122368], ``` (model_simulator_m57)= ## Simulation Run This chapter introduces the basic operations of `axmodel` simulation. The `pulsar2 run` command can be used to run the `axmodel` model generated by `pulsar2 build` directly on the `PC`. The running results of the network model can be quickly obtained without running on the board. ### Simulation run preparation Some models can only support specific input data formats, and the output data of the model is also output in a module-specific format. Before the model simulation is run, the input data needs to be converted into a data format supported by the model. This part of the data operation is called `pre-processing`. After the model simulation is run, the output data needs to be converted into a data format that can be analyzed and viewed by the tool. This part of the data operation is called `post-processing`. The `pre-processing` and `post-processing` tools required for the simulation run are already included in the `pulsar2-run-helper` folder. The contents of the `pulsar2-run-helper` folder are as follows: ```shell root@xxx:/data# ll pulsar2-run-helper/ drwxr-xr-x 2 root root 4.0K Dec 2 12:23 models/ drwxr-xr-x 5 root root 4.0K Dec 2 12:23 pulsar2_run_helper/ drwxr-xr-x 2 root root 4.0K Dec 2 12:23 sim_images/ drwxr-xr-x 2 root root 4.0K Dec 2 12:23 sim_inputs/ drwxr-xr-x 2 root root 4.0K Dec 2 12:23 sim_outputs/ -rw-r--r-- 1 root root 3.0K Dec 2 12:23 cli_classification.py -rw-r--r-- 1 root root 4.6K Dec 2 12:23 cli_detection.py -rw-r--r-- 1 root root 2 Dec 2 12:23 list.txt -rw-r--r-- 1 root root 29 Dec 2 12:23 requirements.txt -rw-r--r-- 1 root root 308 Dec 2 12:23 setup.cfg ``` ### Simulation run example `mobilenetv2` Copy the `compiled.axmodel` generated in the {ref}`《Compile and Execute》 section to the ``pulsar2-run-helper/models`` path and rename it to `mobilenetv2.axmodel` ```shell root@xxx:/data# cp output/compiled.axmodel pulsar2-run-helper/models/mobilenetv2.axmodel ``` #### Input data preparation Enter the `pulsar2-run-helper` directory and use the `cli_classification.py` script to process `cat.jpg` into the input data format required by `mobilenetv2.axmodel`. ```shell root@xxx:~/data# cd pulsar2-run-helper root@xxx:~/data/pulsar2-run-helper# python3 cli_classification.py --pre_processing --image_path sim_images/cat.jpg --axmodel_path models/mobilenetv2.axmodel --intermediate_path sim_inputs/0 [I] Write [input] to 'sim_inputs/0/input.bin' successfully. ``` #### Simulation Model Reasoning Run the `pulsar2 run` command, use `input.bin` as the input data of `mobilenetv2.axmodel` and perform inference calculations, and output `output.bin` inference results. ```shell root@xxx:~/data/pulsar2-run-helper# pulsar2 run --model models/mobilenetv2.axmodel --input_dir sim_inputs --output_dir sim_outputs --list list.txt Building native ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% 0:00:00 2025-03-14 15:00:34.804 | INFO | yamain.command.run:run:90 - >>> [0] start 2025-03-14 15:00:34.805 | INFO | frontend.npu_subgraph_op:pyrun:89 - running npu subgraph: subgraph_npu_0, version: 1, target batch: 0 2025-03-14 15:00:43.900 | INFO | yamain.command.run:write_output:55 - write [output] to [sim_outputs/0/output.bin] successfully, size: 4000 ``` #### Output data processing Use the `cli_classification.py` script to post-process the `output.bin` data output by the simulation model inference to obtain the final calculation results. ```shell root@xxx:/data/pulsar2-run-helper# python3 cli_classification.py --post_processing --axmodel_path models/mobilenetv2.axmodel --intermediate_path sim_outputs/0 [I] The following are the predicted score index pair. [I] 8.8490, 283 [I] 8.7169, 285 [I] 8.4528, 282 [I] 8.4528, 281 [I] 7.6603, 463 ``` (onboard_running_m57)=