在跑NWPU-VHR-10数据集时出现的问题

777Zhao commented 2 years ago

Traceback (most recent call last): File "E:/Python_3/yolov7-pytorch-master/train.py", line 552, in fit_one_epoch(model_train, model, ema, yolo_loss, loss_history, eval_callback, optimizer, epoch, epoch_step, epoch_step_val, gen, gen_val, UnFreeze_Epoch, Cuda, fp16, scaler, save_period, save_dir, local_rank) File "E:\Python_3\yolov7-pytorch-master\utils\utils_fit.py", line 48, in fit_one_epoch loss_value = yolo_loss(outputs, targets, images) File "E:\Python_3\yolov7-pytorch-master\nets\yolotraining.py", line 104, in call bs, as, gjs, gis, targets, anchors = self.build_targets(predictions, targets, imgs) File "E:\Python_3\yolov7-pytorch-master\nets\yolo_training.py", line 406, in build_targets matching_bs[i] = torch.cat(matching_bs[i], dim=0) RuntimeError: torch.cat(): expected a non-empty list of Tensors 出现这样的问题会和数据集中尺寸大小数量有关系吗？我调了一下输入尺寸也还是会报同样错误。如果不是，请教一下有什么解决办法呢，因为看了一些博客但是里面都没有说明具体原因。谢谢

misck commented 2 years ago

那你在用官方库训练的时候，每个epoch有几个step呀

正常的，batchsize开1的话，有648个step

bubbliiiing commented 2 years ago

嗯……你试出来么

misck commented 2 years ago

batchsize开成1的话，step是648，也就是数据集800*0.81。

------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2022年9月20日(星期二) 晚上9:00 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [bubbliiiing/yolov7-pytorch] 在跑NWPU-VHR-10数据集时出现的问题 (Issue #35)

那你在用官方库训练的时候，每个epoch有几个step呀

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>

bubbliiiing commented 2 years ago

emm……不是很理解。现在我没有很明白你的情况

bubbliiiing / yolov7-pytorch

在跑NWPU-VHR-10数据集时出现的问题 #35