Why does the modified code section of the DetectionModel (BaseModel) function need to change the tensor to (2, ch, s, s)?

[18270737873a]

Why does the modified code section of the DetectionModel (BaseModel) function need to change the tensor to (2, ch, s, s)?I tried to modify it to (1,ch,s,s) and there was no problem。

    try: #there add exception
        m.stride = torch.tensor([s / x.shape[-2] for x in forward(torch.zeros(1, ch, s, s))])  # forward
    except Exception as e:
        if 'Not implemented on the CPU' in str(
                e) or 'Input type (torch.FloatTensor) and weight type (torch.cuda.FloatTensor)' in str(
            e) or 'CUDA tensor' in str(e) or 'is_cuda()' in str(e) or "Cannot find backend for cpu" in str(e):
            self.model.to(torch.device('cuda'))
            m.stride = torch.tensor([s / x.shape[-2] for x in
                                     forward(torch.zeros(2, ch, s, s).to(torch.device('cuda')))])  # forward
        else:
            raise e

[zhang123-lf]

你好，请问你用自己的训练成功了吗

[18270737873a]

你好，请问你用自己的训练成功了吗

成功了，就是不理解 forward(torch.zeros(2, ch, s, s).to(torch.device('cuda')))]) # forward这为什么要改成torch.zeros(2, ch, s, s)

[zhang123-lf]

你好，请问你用自己的训练成功了吗

成功了，就是不明白forward(torch.zeros(2, ch, s, s).to(torch.device('cuda')))]) #forward这为什么要改成torch.zeros(2, ch, s, s)

请问你出现过训练时，均值平均精度挺高的，但是在验证集上各项数据为0的情况吗

[18270737873a]

不会

…

------------------ 原始邮件 ------------------ 发件人: ***@***.***>; 发送时间: 2025年2月28日(星期五) 下午4:57 收件人: ***@***.***>; 抄送: ***@***.***>; ***@***.***>; 主题: Re: [HZAI-ZJNU/Mamba-YOLO] Why does the modified code section of the DetectionModel (BaseModel) function need to change the tensor to (2, ch, s, s)? (Issue #72) 你好，请问你用自己的训练成功了吗 成功了，就是不明白forward(torch.zeros(2, ch, s, s).to(torch.device('cuda')))]) #forward这为什么要改成torch.zeros(2, ch, s, s) 请问你出现过训练时，均值平均精度挺高的，但是在验证集上各项数据为0的情况吗 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***> zhang123-lf left a comment (HZAI-ZJNU/Mamba-YOLO#72) 你好，请问你用自己的训练成功了吗 成功了，就是不明白forward(torch.zeros(2, ch, s, s).to(torch.device('cuda')))]) #forward这为什么要改成torch.zeros(2, ch, s, s) 请问你出现过训练时，均值平均精度挺高的，但是在验证集上各项数据为0的情况吗 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[zhang123-lf]

你好，请问你用自己的训练成功了吗

成功了，就是不明白forward(torch.zeros(2, ch, s, s).to(torch.device('cuda')))]) #forward这为什么要改成torch.zeros(2, ch, s, s)

[zhang123-lf]

你好，请问你用自己的训练成功了吗

成功了，就是不明白forward(torch.zeros(2, ch, s, s).to(torch.device('cuda')))]) #forward这为什么要改成torch.zeros(2, ch, s, s)
请问，你是做了什么更改吗，可以分享一下你的操作吗