I am trying to compile pretrained yolov5s6_ti_lite model by TI using edge ai benchmark repo, but Compilation stuck inbetween and there is no further progress in jupyter notebook tutorials_detection.
Steps I have done:
1.Downloaded model weight and prototxt file from edgeai-yolov5 for yolov5s6 model.
2. Installed edgeai-benchmark 8.5
3. Created a folder named models in edgeai-benchmark and moved model and prototxt file in it.
4. set tidl_tools path
Fullscreen
1
2
3
4
5
6
7
8
9
10
11
# the cwd must be the root of the respository
if os.path.split(os.getcwd())[-1] in ('scripts', 'tutorials'):
os.chdir('../')
#
os.environ['TIDL_TOOLS_PATH']=f"/home/mugu/edgeai-benchmark/tools/TDA4VM/tidl_tools"
print(os.environ['TIDL_TOOLS_PATH'])
print(os.getcwd())
output:
/home/mugu/edgeai-benchmark/tools/TDA4VM/tidl_tools
/home/mugu/edgeai-benchmark
5. copied pipeline configs for yolov5s6 from benchmark configs :
Fullscreen
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
pipeline_configs = {
'imagedet-7': dict(
task_type='detection',
calibration_dataset=calib_dataset,
input_dataset=val_dataset,
preprocess=preproc_transforms.get_transform_onnx(640, 640, resize_with_pad=True, backend='cv2', pad_color=[114,114,114]),
session=sessions.ONNXRTSession(**utils.dict_update(onnx_session_cfg, input_optimization=False, input_mean=(0.0, 0.0, 0.0), input_scale=(0.003921568627, 0.003921568627, 0.003921568627)),
runtime_options=settings.runtime_options_onnx_np2(
det_options=True, ext_options={'object_detection:meta_arch_type': 6,
'object_detection:meta_layers_names_list':f'/home/mugu/edgeai-benchmark/models/yolov5s6_640_ti_lite_metaarch.prototxt',
'advanced_options:output_feature_16bit_names_list':'370,680,990,1300'
}),
model_path=f'/home/mugu/edgeai-benchmark/models/yolov5s6_640_ti_lite_37p4_56p0.onnx'),
postprocess=postproc_transforms.get_transform_detection_yolov5_onnx(squeeze_axis=None, normalized_detections=False, resize_with_pad=True, formatter=postprocess.DetectionBoxSL2BoxLS()), #TODO: check this
metric=dict(label_offset_pred=datasets.coco_det_label_offset_80to90(label_offset=1)),
model_info=dict(metric_reference={'accuracy_ap[.5:.95]%':37.4})
)
}
print(pipeline_configs)
6.Started compilation, but stucks inbetween and sometimes kernal dead issues arrives
