Open yantiantianlx opened 1 year ago
OK, thanks, I use the wrong commit. And when I convert cascade-rcnn model to tensorrt, it can only be 1.3 times faster in fp32 (inference time is from 84ms in pytorch to 64ms in tensorrt), and the GPU memery usage is not reduced. Is this normal?
看起来没有定义这个RepeatDimsPluginDynamic plugin, 是需要自己定义么,还是需要从哪里下载插件包。
[TensorRT] ERROR: INVALID_ARGUMENT: getPluginCreator could not find plugin RepeatDimsPluginDynamic version 1 Traceback (most recent call last): File "mmdetection-to-tensorrt/demo/inference.py", line 45, in
main()
File "mmdetection-to-tensorrt/demo/inference.py", line 27, in main
cfg_path, args.checkpoint, fp16_mode=args.fp16, device=args.device)
File "/data/sda/ytt/raycloud_server/test_tensorRT/mmdetection/mmdetection-to-tensorrt/mmdet2trt/mmdet2trt.py", line 156, in mmdet2trt
int8_calib_algorithm=int8_calib_algorithm)
File "/data/sda/ytt/raycloud_server/test_tensorRT/mmdetection/torch2trt_dynamic/torch2trt_dynamic/torch2trt_dynamic.py", line 534, in torch2trt_dynamic
outputs = module(inputs)
File "/usr/local/lib/python3.7/site-packages/torch/nn/modules/module.py", line 722, in _call_impl
result = self.forward(input, *kwargs)
File "/data/sda/ytt/raycloud_server/test_tensorRT/mmdetection/mmdetection-to-tensorrt/mmdet2trt/models/detectors/two_stage.py", line 48, in forward
rois = rpn_head(feat, x)
File "/usr/local/lib/python3.7/site-packages/torch/nn/modules/module.py", line 722, in _call_impl
result = self.forward(input, *kwargs)
File "/data/sda/ytt/raycloud_server/test_tensorRT/mmdetection/mmdetection-to-tensorrt/mmdet2trt/models/dense_heads/rpn_head.py", line 55, in forward
proposals = mm2trt_util.gather_topk(proposals, 1, topk_inds)
File "/data/sda/ytt/raycloud_server/test_tensorRT/mmdetection/mmdetection-to-tensorrt/mmdet2trt/ops/util_ops.py", line 59, in gather_topk
index = index.repeat(repeat_size)
File "/data/sda/ytt/raycloud_server/test_tensorRT/mmdetection/torch2trt_dynamic/torch2trt_dynamic/torch2trt_dynamic.py", line 325, in wrapper
converter'converter'
File "/data/sda/ytt/raycloud_server/test_tensorRT/mmdetection/torch2trt_dynamic/torch2trt_dynamic/converters/repeat.py", line 16, in convert_repeat
plugin = create_repeatplugin("repeat" + str(id(input)), repeat_shape=shape)
File "/data/sda/ytt/raycloud_server/test_tensorRT/mmdetection/torch2trt_dynamic/torch2trt_dynamic/plugins/create_repeatdim_plugin.py", line 24, in create_repeat_plugin
res = creator.create_plugin(layer_name, pfc)
AttributeError: 'NoneType' object has no attribute 'create_plugin'