I-KBA-231226181840
1. Setha Imvelo
1.1. Faka i-Nvidia Driver ne-CUDA
1.2. Faka Ilabhulali Ye-Python Ehlobene
I-python3 -m ukufaka ipayipi -thuthukisa -i-ignore-installed pip
I-python3 -m ukufaka i-pip -i-ignore-efakwe i-gdown
I-python3 -m ukufaka i-pip -i-ignore-efakwe i-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
I-python3 -m ukufaka ipayipi -i-ignore-efakwe i-ftfy
I-python3 -m ukufaka i-pip -i-ignore-installed 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-installed psutil
1.3. Clone aimet-model-zoo
git clone https://github.com/quic/aimet-model-zoo.git
cd aimet-model-zoo
git checkout d09d2b0404d10f71a7640a87e9d5e5257b028802
thekelisa i-PYTHONPATH=${PYTHONPATH}:${PWD}
1.4. Landa i-Set14
wget https://uofi.box.com/shared/static/igsnfieh4lz68l926l8xbklwsnnk8we9.zip
unzip igsnfieh4lz68l926l8xbklwsnnk8we9.zip
1.5. Guqula umugqa 39 aimet-model-zoo/aimet_zoo_torch/quicksrnet/dataloader/utils.py
shintsha
okwe-img_path ku-glob.glob(os.path.join(test_images_dir, “*”)):
ku
okwe-img_path ku-glob.glob(os.path.join(test_images_dir, “*_HR.*”)):
1.6. Yenza ukuhlola.
# sebenzisa ngaphansi kwe-YOURPATH/aimet-model-run
# Okwe-quicksrnet_small_2x_w8a8
python3 aimet_zoo_torch/quicksrnet/evaluators/quicksrnet_quanteval.py \
-model-config quicksrnet_small_2x_w8a8 \
-dataset-path ../Set14/image_SRF_4
# Okwe-quicksrnet_small_4x_w8a8
python3 aimet_zoo_torch/quicksrnet/evaluators/quicksrnet_quanteval.py \
-model-config quicksrnet_small_4x_w8a8 \
-dataset-path ../Set14/image_SRF_4
# Okwe-quicksrnet_medium_2x_w8a8
python3 aimet_zoo_torch/quicksrnet/evaluators/quicksrnet_quanteval.py \
-model-config quicksrnet_medium_2x_w8a8 \
-dataset-path ../Set14/image_SRF_4
# Okwe-quicksrnet_medium_4x_w8a8
python3 aimet_zoo_torch/quicksrnet/evaluators/quicksrnet_quanteval.py \
-model-config quicksrnet_medium_4x_w8a8 \
-dataset-path ../Set14/image_SRF_4
ake sithi uzothola i-PSNRvaluefor theaimet simulated model. Ungashintsha imodeli-config yosayizi ohlukile we-QuickSRNet, inketho ithi underaimet-modelzoo/aimet_zoo_torch/quicksrnet/model/model_cards/.
2 Engeza ipheshana
2.1. Vula okuthi “Thekelisa ku-ONNX Izinyathelo REVISED.docx”
2.2. Yeqa i-id yokuzibophezela ye-git
2.3. Ikhodi yesigaba 1
Engeza yonke ikhodi engu-1 ngaphansi komugqa wokugcina (ngemuva komugqa 366) aimet-model-zoo/aimet_zoo_torch/quicksrnet/model/models.py
2.4. Ikhodi yeSigaba 2 no-3
Engeza amakhodi angu-2, 3 aphelele ngaphansi komugqa 93 aimet-model-zoo/aimet_zoo_torch/quicksrnet/evaluators/quicksrnet_quanteval.py
2.5. Amapharamitha angukhiye ku-Function load_model
imodeli = load_model(MODEL_PATH_INT8,
MODEL_NAME,
MODEL_ARGS.get(MODEL_NAME).get(MODEL_CONFIG),
use_quant_sim_model=Iqiniso,
encoding_path=ENCODING_PATH,
quantsim_config_path=CONFIG_PATH,
calibration_data=IMAGES_LR,
use_cuda=Iqiniso,
before_quantization=Iqiniso,
convert_to_dcr=Iqiniso)
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
Sicela umiselele okuguquguqukayo kosayizi ohlukile we-QuickSRNet
2.6 Ukuguqulwa Kosayizi Wemodeli
- “input_shape” ku-aimet-model-zoo/aimet_zoo_torch/quicksrnet/model/model_cards/*.json
- Ngaphakathi komsebenzi load_model(...) ku-aimet-model-zoo/aimet_zoo_torch/quicksrnet/model/inference.py
- Ipharamitha engaphakathi komsebenzi export_to_onnx(…, input_height, input_width) isuka kokuthi “Thekelisa ku-ONNX Izinyathelo REVISED.docx”
2.7 Qalisa kabusha i-1.6 futhi ukuze uthumele imodeli ye-ONNX
3. Guqula ku-SNPE
3.1. Guqula
${SNPE_ROOT}/bin/x86_64-linux-clang/snpe-onnx-to-dlc \
-input_network model.onnx \
-quantization_overrides ./model.encodings
3.2. (Ongakukhetha) Khipha kuphela i-DLC enamanani amaningi
(ongakukhetha) i-snpe-dlc-quant -input_dlc model.dlc -float_fallback -override_params
3.3. (KUBALULEKILE) I-ONNX I/O ilandelana nge-NCHW; I-DLC eguquliwe ikuhlelo lwe-NHWC
Amadokhumenti / Izinsiza
 |
I-Qualcomm Aimet Efficiency Toolkit Documentation [pdf] Iziyalezo quicksrnet_small_2x_w8a8, quicksrnet_small_4x_w8a8, quicksrnet_medium_2x_w8a8, quicksrnet_medium_4x_w8a8, Aimet Efficiency Toolkit Documentation, Ukusebenza kahle Kombhalo Umbhalo, Ithuluzi Lombhalo, Idokhumenti Yedokhumenti |
