KBA-231226181840
1. Konfiguratu ingurunea
1.1. Instalatu Nvidia Driver eta CUDA
1.2. Instalatu erlazionatutako Python liburutegia
python3 -m pip install –upgrade –ignore-installed pip
python3 -m pip install –ignore-installed gdown
python3 -m pip install –ignore-installed opencv-python
python3 -m pip install –ignore-installed torch==1.9.1+cu111 torchvision==0.10.1+cu111 torchaudio==0.9.1 -f https://download.pytorch.org/whl/torch_stable.html
python3 -m pip install –ignore-installed jax
python3 -m pip install –ignore-installed ftfy
python3 -m pip install –ignore-instaled torchinfo
python3 -m pip install –ignore-installed https://github.com/quic/aimet/releases/download/1.25.0/AimetCommon-torch_gpu_1.25.0-cp38-cp38-linux_x86_64.whl
python3 -m pip install –ignore-installed https://github.com/quic/aimet/releases/download/1.25.0/AimetTorch-torch_gpu_1.25.0-cp38-cp38-linux_x86_64.whl
python3 -m pip install –ignore-installed numpy==1.21.6
python3 -m pip install –ignore-instalatutako psutil
1.3. Klonatu aime-model-zoo
git clone https://github.com/quic/aimet-model-zoo.git
cd aimet-model-zoo
git checkout d09d2b0404d10f71a7640a87e9d5e5257b028802
esportatu PYTHONPATH=${PYTHONPATH}:${PWD}
1.4. Deskargatu Set14
wget https://uofi.box.com/shared/static/igsnfieh4lz68l926l8xbklwsnnk8we9.zip
unzip igsnfieh4lz68l926l8xbklwsnnk8we9.zip
1.5. Aldatu 39. lerroa aimet-model-zoo/aimet_zoo_torch/quicksrnet/dataloader/utils.py
aldatu
img_path-en glob.glob(os.path.join(test_images_dir, “*”)):
to
img_path-en glob.glob(os.path.join(test_images_dir, “*_HR.*”)):
1.6. Egin ebaluazioa.
# exekutatu YOURPATH/aimet-model-run azpian
# Quicksrnet_small_2x_w8a8-rako
python3 aimet_zoo_torch/quicksrnet/evaluators/quicksrnet_quanteval.py \
–model-config quicksrnet_small_2x_w8a8 \
–dataset-path ../Set14/image_SRF_4
# Quicksrnet_small_4x_w8a8-rako
python3 aimet_zoo_torch/quicksrnet/evaluators/quicksrnet_quanteval.py \
–model-config quicksrnet_small_4x_w8a8 \
–dataset-path ../Set14/image_SRF_4
# Quicksrnet_medium_2x_w8a8-rako
python3 aimet_zoo_torch/quicksrnet/evaluators/quicksrnet_quanteval.py \
–model-config quicksrnet_medium_2x_w8a8 \
–dataset-path ../Set14/image_SRF_4
# Quicksrnet_medium_4x_w8a8-rako
python3 aimet_zoo_torch/quicksrnet/evaluators/quicksrnet_quanteval.py \
–model-config quicksrnet_medium_4x_w8a8 \
–dataset-path ../Set14/image_SRF_4
demagun PSNR balioa lortuko duzula simulatutako eredurako. QuickSRNet-en tamaina desberdinetarako eredu-konfigurazioa alda dezakezu, aukera underaimet-modelzoo/aimet_zoo_torch/quicksrnet/model/model_cards/ da.
2 Gehitu adabakia
2.1. Ireki "Esportatu ONNX Steps REVISED.docx-era"
2.2. Saltatu git commit ID
2.3. 1. atala Kodea
Gehitu 1. kode osoa azken lerroan (366. lerroaren ondoren) aimet-model-zoo/aimet_zoo_torch/quicksrnet/model/models.py
2.4. 2. eta 3. atala Kodea
Gehitu 2, 3 kodea osoa 93. lerroan aimet-model-zoo/aimet_zoo_torch/quicksrnet/evaluators/quicksrnet_quanteval.py
2.5. Funtzioaren gako-parametroak load_model
eredua = load_model (MODEL_PATH_INT8,
MODEL_NAME,
MODEL_ARGS.get(MODEL_NAME).get(MODEL_CONFIG),
use_quant_sim_model=Egia,
encoding_path=ENCODING_PATH,
quantsim_config_path=CONFIG_PATH,
calibration_data=IMAGES_LR,
use_cuda=Egia,
before_quantization=Egia,
convert_to_dcr=Egia)
MODEL_PATH_INT8 = aimet_zoo_torch/quicksrnet/model/weights/quicksrnet_small_2x_w8a8/pre_opt_weights
MODEL_NAME = QuickSRNetSmall
MODEL_ARGS.get(MODEL_NAME).get(MODEL_CONFIG) = {'scaling_factor': 2}
ENCODING_PATH = aimet_zoo_torch/quicksrnet/model/weights/quicksrnet_small_2x_w8a8/adaround_encodings
CONFIG_PATH = aimet_zoo_torch/quicksrnet/model/weights/quicksrnet_small_2x_w8a8/aimet_config
Mesedez, ordezkatu QuickSRNet-en tamaina desberdinetarako aldagaiak
2.6 Ereduaren Tamainaren Aldaketa
- "sarrera_forma" aimet-model-zoo/aimet_zoo_torch/quicksrnet/model/model_cards/*.json-en
- load_model(…) funtzioaren barnean aim-model-zoo/aimet_zoo_torch/quicksrnet/model/inference.py
- Export_to_onnx(…, input_height, input_width) funtzioaren barneko parametroa "Esportatu ONNX-ra BERRIKUSI.docx urratsak"-tik
2.7 Berriz exekutatu 1.6 berriro ONNX eredua esportatzeko
3. Bihurtu SNPEn
3.1. Bihurtu
${SNPE_ROOT}/bin/x86_64-linux-clang/snpe-onnx-to-dlc \
–sarrera_sarearen eredua.onnx \
–quantization_overrides ./model.encodings
3.2. (Aukerakoa) Atera ezazu DLC kuantifikatua soilik
(aukerakoa) snpe-dlc-quant –input_dlc model.dlc –float_fallback –override_params
3.3. (GARRANTZITSUA) ONNX I/O NCHW ordenan dago; Bihurtutako DLC ordena NHWC da
Dokumentuak / Baliabideak
 |
Qualcomm Aimet Efficiency Toolkit dokumentazioa [pdfArgibideak quicksrnet_small_2x_w8a8, quicksrnet_small_4x_w8a8, quicksrnet_medium_2x_w8a8, quicksrnet_medium_4x_w8a8, Aimet Efficiency Toolkit Dokumentazioa, Efficiency Toolkit Dokumentazioa, Toolkit Dokumentazioa, Dokumentazioa |
