I want to use the yolov5l-xs-1.pt model to perform inference and optimize it using TensorRT. I understand you are not using TensorRT, but I thought you might understand the issue

I have exported the .pt file to an onnx file using the export.py program (without --dynamic flag). It gave this warning, but I don't understand what it means:

WARNING: The shape inference of prim::Constant type is missing, so it may result in wrong shape inference for the exported graph. Please consider adding it in symbolic function.

Later, when loading the ONNX file with TensorRT, I get this error:

[TensorRT] ERROR: [graphShapeAnalyzer.cpp::throwIfError::1306] Error Code 9: Internal Error (Reshape_218: reshape changes volume)

Apparently there is a node which reshapes to a different volume, which is not allowed according to TensorRT. Do you know what I can do about this issue? Please let me know if you could use more information!

EDIT: I have also tried it with the --dynamic flag. It looks like it is going through more of the ONNX model when loading, but eventually it gives this error:

[TRT]    4: [network.cpp::validate::2713] Error Code 4: Internal Error (images: dynamic input is missing dimensions in profile 0.)

EDIT: This is what the node (from the non-dynamic model) looks like:

Hey,

When I am trying to run the inference command I am getting the next error: ret = input.softmax(dim) RuntimeError: CUDA out of memory. Tried to allocate 962.00 MiB (GPU 0; 3.81 GiB total capacity; 1.79 GiB already allocated; 725.00 MiB free; 1.93 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

I tried to change the batch to 4 and even to 2 but it doesn't solve the problem.

What can I do to solve this?

self.m = nn.ModuleList([nn.Conv2d(c_, c_, kernel_size=3, stride=1, padding= x //2, dilation= x //2, bias=False) for x in k]) TypeError: 'int' object is not iterable

May I ask the author why this is? Without modifying any code, use the training code: python train.py --img 1536 --adam --batch 4 --epochs 80 --data ./data/VisDrone.yaml --weights yolov5l.pt --hy data/hyps /hyp.VisDrone.yaml --cfg models/yolov5l-xs-tph.yaml --name v5l-xs-tph Error message：

Traceback (most recent call last): File "train.py", line 630, in main(opt) File "train.py", line 527, in main train(opt.hyp, opt, device, callbacks) File "train.py", line 119, in train model = Model(cfg or ckpt['model'].yaml, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device) # create File "/root/autodl-tmp/tph-yolov5-main/models/yolo.py", line 104, in init self.model, self.save = parse_model(deepcopy(self.yaml), ch=[ch]) # model, savelist File "/root/autodl-tmp/tph-yolov5-main/models/yolo.py", line 291, in parse_model m_ = nn.Sequential(*(m(args) for _ in range(n))) if n > 1 else m(args) # module File "/root/autodl-tmp/tph-yolov5-main/models/common.py", line 493, in init self.m = SwinTransformerBlock(c_, c_, c_//32, n) File "/root/autodl-tmp/tph-yolov5-main/models/common.py", line 426, in init self.tr = nn.Sequential((SwinTransformerLayer(c2, num_heads=num_heads, window_size=window_size, shift_size=0 if (i % 2 == 0) else self.shift_size ) for i in range(num_layers))) File "/root/autodl-tmp/tph-yolov5-main/models/common.py", line 426, in self.tr = nn.Sequential((SwinTransformerLayer(c2, num_heads=num_heads, window_size=window_size, shift_size=0 if (i % 2 == 0) else self.shift_size ) for i in range(num_layers))) File "/root/autodl-tmp/tph-yolov5-main/models/common.py", line 338, in init self.attn = WindowAttention( File "/root/autodl-tmp/tph-yolov5-main/models/common.py", line 259, in init coords = torch.stack(torch.meshgrid([coords_h, coords_w], indexing="ij")) # [2, Mh, Mw] TypeError: meshgrid() got an unexpected keyword argument 'indexing'

I want to know where the TPH module is？ And I want to know whether the network structure described in the file "yolov5l-xs-tr-cbam-spp-bifpn.yaml" is consistent with the network structure described in the paper.

Hello author, the following error occurred in the reasoning process, please advise, thank you。 AttributeError: Can't get attribute 'NonDynamicallyQuantizableLinear' on <module 'torch.nn.modules.linear' from '/root/miniconda3/lib/python3.8/site-packages/torch/nn/modules/linear.py'>

Traceback (most recent call last): File "train.py", line 48, in from utils.loggers import Loggers ImportError: cannot import name 'Loggers' from 'utils.loggers' (unknown location)

File "/home/test/anaconda3/envs/whn_PT/lib/python3.6/site-packages/torch/nn/modules/module.py", line 727, in _call_impl result = self.forward(input, **kwargs) File "/home/test/anaconda3/envs/whn_PT/lib/python3.6/site-packages/torch/nn/modules/container.py", line 117, in forward input = module(input) File "/home/test/anaconda3/envs/whn_PT/lib/python3.6/site-packages/torch/nn/modules/module.py", line 727, in _call_impl result = self.forward(input, **kwargs) File "/mnt/4T/whn/thesis/7-add swintransformer/models/common.py", line 318, in forward attn_windows = self.attn(x_windows, mask=attn_mask) # [nWB, MhMw, C] File "/home/test/anaconda3/envs/whn_PT/lib/python3.6/site-packages/torch/nn/modules/module.py", line 727, in call_impl result = self.forward(*input, **kwargs) File "/mnt/4T/whn/thesis/7-add swintransformer/models/common.py", line 239, in forward x = (attn @ v).transpose(1, 2).reshape(B, N, C) RuntimeError: Expected object of scalar type Float but got scalar type Half for argument #2 'mat2' in

Thanks for your work. I use this respository to train on the VisDrone dataset just using 10 images to have a fast try, but I find this training is not exactly performed on VisDrone with 0 MAP. However, this code works normally on coco128 dataset. It is very strange and I need your help.

Hi, we ran into an AttributeError running inference with detect.py due to a change of the PyTorch API in version 1.11 (no more "recompute_scale_factor"). So we added this fix according to the official yolov5 "PyTorch 1.11.0 compatibility update".

File "/content/drive/MyDrive/Yolo v5/utils/loss.py", line 240, in build_targets indices.append((b, a, gj.clamp_(0, gain[3] - 1), gi.clamp_(0, gain[2] - 1))) # image, anchor, grid indices RuntimeError: result type Float can't be cast to the desired output type long int

Dear author, I'm confusing the usage of WBF. Could you please give me a guide?

Q: When I get the wbf_labels based on the ensemble for val results of two models, how to calculate the new mAP based on these wbf_labels? Is there a command or code for such calculation??

python train.py --img 1536 --adam --batch 4 --epochs 80 --data ./data/VisDrone.yaml --weights yolov5l.pt --hy data/hyps/hyp.VisDrone.yaml --cfg models/yolov5l-xs-tph.yaml --name v5l-xs-tph

have a error problem File "/data/zhangshilin/wangjun/529_zsl_6.0/tph-yolov5/tph-yolov5/models/common.py", line 493, in init self.m = SwinTransformerBlock(c_, c_, c_//32, n) File "/data/zhangshilin/wangjun/529_zsl_6.0/tph-yolov5/tph-yolov5/models/common.py", line 426, in init self.tr = nn.Sequential((SwinTransformerLayer(c2, num_heads=num_heads, window_size=window_size, shift_size=0 if (i % 2 == 0) else self.shift_size ) for i in range(num_layers))) File "/data/zhangshilin/wangjun/529_zsl_6.0/tph-yolov5/tph-yolov5/models/common.py", line 426, in self.tr = nn.Sequential((SwinTransformerLayer(c2, num_heads=num_heads, window_size=window_size, shift_size=0 if (i % 2 == 0) else self.shift_size ) for i in range(num_layers))) File "/data/zhangshilin/wangjun/529_zsl_6.0/tph-yolov5/tph-yolov5/models/common.py", line 340, in init attn_drop=attn_drop, proj_drop=drop) File "/data/zhangshilin/wangjun/529_zsl_6.0/tph-yolov5/tph-yolov5/models/common.py", line 259, in init coords = torch.stack(torch.meshgrid([coords_h, coords_w], indexing="ij")) # [2, Mh, Mw] TypeError: meshgrid() got an unexpected keyword argument 'indexing'