wanghaoshuang
commented
6 months ago 当前给出的示例是以 PPYOLOE 为例，仅支持PPYOLOE的后处理逻辑。 如果切换模型，需要该下边链接的代码，来添加 rtdetr 模型结构的后处理逻辑： https://github.com/PaddlePaddle/PaddleSlim/blob/develop/example/auto_compression/detection/run.py#L105-L107
指令：python run.py --config_path=./configs/rtdetr_hgnetv2_x_qat_dis.yaml --save_dir='./output/' --devices='gpu'
配置nms为False Global: reader_config: configs/rtdetr_reader.yml include_nms: False Evaluation: True model_dir: ./models/rtdetr_hgnetv2_6x_coco_no_post/ model_filename: model.pdmodel params_filename: model.pdiparams
量化报错
File "/home/weconai/sda1/PaddleSlim/example/auto_compression/detection/run.py", line 198, in
main()
File "/home/weconai/sda1/PaddleSlim/example/auto_compression/detection/run.py", line 188, in main
ac.compress()
File "/home/weconai/.conda/envs/ppslim/lib/python3.10/site-packages/paddleslim-2.6.0-py3.10.egg/paddleslim/auto_compression/compressor.py", line 586, in compress
self.single_strategy_compress(strategy, config, strategy_idx,
File "/home/weconai/.conda/envs/ppslim/lib/python3.10/site-packages/paddleslim-2.6.0-py3.10.egg/paddleslim/auto_compression/compressor.py", line 780, in single_strategy_compress
test_program_info = self._start_train(
File "/home/weconai/.conda/envs/ppslim/lib/python3.10/site-packages/paddleslim-2.6.0-py3.10.egg/paddleslim/auto_compression/compressor.py", line 827, in _start_train
metric = self.eval_function(
File "/home/weconai/sda1/PaddleSlim/example/auto_compression/detection/run.py", line 109, in eval_function
assert "Not support arch={} now.".format(global_config['arch'])
KeyError: 'arch'