hanrui1sensetime comments

Results 37 comments of


                                            hanrui1sensetime

improvement(cmake): simplify build option and doc

May be still need change [jetsons.md](https://github.com/tpoisonooo/mmdeploy/blob/simplify-build-option/docs/en/01-how-to-build/jetsons.md) to remove `-DMMDEPLOY_TARGET_BACKENDS` options explicitly. The same about zh-cn version.

convnext onnx export loss precision!

Your config and checkpoint doesn't match, please try for [this checkpoint](https://download.openmmlab.com/mmclassification/v0/convnext/convnext-base_3rdparty_32xb128_in1k_20220124-d0915162.pth).

Error: export FCOS model!

@FDInSky Hi, did you remember your script and the env with python tools/check_ev.py? Please post on here, thank you.

Error: export FCOS model!

I will close this issue, if you have other questions about this issue, you can reopen it.

AssertionError: `max_shape` should be tensor of (h,w) for onnx for FCOS model

Now you can use `detection_onnxruntime_dynamic.py` instead of `detection_onnxruntime_static.py`. And later we will fix the detection_onnxruntime_dynamic.py`

Confusing Results on ONNX model

Hi, @sarmientoj24 The output shape (1, 9, 5) means you have 9 detected bboxes, and each of them has 5 values, which means the related coordinate and side length, and...

Score/confidence of prediction drop a lot after convert to trt engine

Hi, what about the result under dynamic config?

Score/confidence of prediction drop a lot after convert to trt engine

Could you please provide me with the files you have used? ```/workdir/detection_onnx_static_1024x1024.py``` ```/workdir/atss_swin-l-p4-w12_fpn_dyhead_mstrain_2x.py``` ```/workdir/latest.pth```. Are they different from our official config and checkpoints?

RuntimeError: Failed to parse onnx, In node 28 (parseGraph): INVALID_NODE: Invalid Node - Conv_28; Conv_28: two inputs (data and weights) are allowed only in explicit-quantization mode.

Hi, @Xia-sj , would you please provide the link of the model config and the checkpoint of the model you mentioned, i.e. `mmdeploy/detectors_cascade_rcnn_r50_1x_coco.py` and `mmdeploy/detectors_htc_r50_1x_coco-329b1453.pth`?

RuntimeError: Failed to parse onnx, In node 28 (parseGraph): INVALID_NODE: Invalid Node - Conv_28; Conv_28: two inputs (data and weights) are allowed only in explicit-quantization mode.

> @hanrui1sensetime Tnanks for your reply. The corresponding link is provided in: > > https://pan.baidu.com/s/1_q_5UkkWSfrKuCEv9gGu0w 提取码：on80 > > Note that the layer ConvAWS is not supported. The download on Baidu...