search

LiWentomng / BoxInstSeg

A toolbox for box-supervised instance segmentation.
Apache License 2.0
407 stars 38 forks source link

如果我要训练自己的数据集，除了config文件里面要改动相应的clss，别的地方还有哪里要更改吗 #19

Open wubalahao opened 1 year ago

LiWentomng commented 1 year ago

@wubalahao Sorry to rely late. 是的， 只需要修改配置文件中的 class，就可以训练。 可以先训练看看结果，再考虑进一步调整一些参数。

ZHIZIHUABU commented 11 months ago

代码基于mmdet库运行，mmdet里面的class_names也是要修改的

wubalahao commented 10 months ago

是这样的我根据老哥的建议换成了自己的数据集也跑通了，但是发现了一个奇怪的bug，就是我在不改变这个文件里的类别名字的时候数据读取是正常的如下： image 但是我一旦更改乐这个文件里面的类别名字就会 image 出现读取错误的问题，这个我搞了一会也没搞明白是啥问题

wubalahao commented 10 months ago

还有就算可以正常训练，每到第二个epoch就会训练停止出现以下错误

2024-01-28 10:54:54,173 - mmdet - INFO - Epoch [2][750/2800] lr: 5.000e-03, eta: 9:17:20, time: 0.338, data_time: 0.004, memory: 4121, loss_cls: nan, loss_bbox: nan, loss_centerness: nan, loss_prj: nan, loss_pairwise: nan, loss: nan Traceback (most recent call last): File "tools/train.py", line 257, in main() File "tools/train.py", line 245, in main train_detector( File "/mnt/d/deeplearning/wh/boxinstseg1/boxinstseg/mmdet/apis/train.py", line 244, in train_detector runner.run(data_loaders, cfg.workflow) File "/home/wubala/down/yes/envs/boxseg/lib/python3.8/site-packages/mmcv/runner/epoch_based_runner.py", line 127, in run epoch_runner(data_loaders[i], kwargs) File "/home/wubala/down/yes/envs/boxseg/lib/python3.8/site-packages/mmcv/runner/epoch_based_runner.py", line 50, in train self.run_iter(data_batch, train_mode=True, kwargs) File "/home/wubala/down/yes/envs/boxseg/lib/python3.8/site-packages/mmcv/runner/epoch_based_runner.py", line 29, in run_iter outputs = self.model.train_step(data_batch, self.optimizer, File "/home/wubala/down/yes/envs/boxseg/lib/python3.8/site-packages/mmcv/parallel/data_parallel.py", line 75, in train_step return self.module.train_step(inputs[0], kwargs[0]) File "/mnt/d/deeplearning/wh/boxinstseg1/boxinstseg/mmdet/models/detectors/base.py", line 248, in train_step losses = self(data) File "/home/wubala/down/yes/envs/boxseg/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl return forward_call(input, *kwargs) File "/home/wubala/down/yes/envs/boxseg/lib/python3.8/site-packages/mmcv/runner/fp16_utils.py", line 110, in new_func return old_func(args, kwargs) File "/mnt/d/deeplearning/wh/boxinstseg1/boxinstseg/mmdet/models/detectors/base.py", line 172, in forward return self.forward_train(img, img_metas, kwargs) File "/mnt/d/deeplearning/wh/boxinstseg1/boxinstseg/mmdet/models/detectors/condinst.py", line 69, in forward_train param_pred, coors, level_inds, img_inds, gt_inds = self.mask_head.training_sample( File "/mnt/d/deeplearning/wh/boxinstseg1/boxinstseg/mmdet/models/dense_heads/condinst_head.py", line 1225, in training_sample sampled_inds = torch.cat(sampled_inds, dim=0) NotImplementedError: There were no tensor arguments to this function (e.g., you passed an empty list of Tensors), but no fallback function is registered for schema aten::_cat. This us ually means that this function requires a non-empty list of Tensors, or that you (the operator writer) forgot to register a fallback function. Available functions are [CPU, CUDA, Quan tizedCPU, BackendSelect, Named, ADInplaceOrView, AutogradOther, AutogradCPU, AutogradCUDA, AutogradXLA, UNKNOWN_TENSOR_TYPE_ID, AutogradMLC, AutogradHPU, AutogradNestedTensor, AutogradPrivateUse1, AutogradPrivateUse2, AutogradPrivateUse3, Tracer, Autocast, Batched, VmapMode].