I used 8 3090GPUs to reproduce the result, which was 2 lower than the official result.

[xu19971109]

I trained 20 epochs with the default configuration of the cmt_voxel0100_r50_800x320_cbgs.py.

My result is 65.0% | 69.4% and the official is 67.9% | 70.8%, which is about 2% lower than you.

Is there anything I'm missing?

[junjie18]

@xu19971109 15e + 5e means the last 5 epochs should be trained without GTsample.

[xu19971109]

@junjie18 Thank you, I reproduced your results!

[rubbish001]

能不能把训练日志发我一下，liwenx@whu.edu.cn 训练太慢了，我第十轮是aAP 0.62，nds：0.685

[rubbish001]

@xu19971109 15e + 5e means the last 5 epochs should be trained without GTsample.

能不能把训练日志发我一下，v299+1600*640,的，我只有4张3090，训练太慢了，哎

[xu19971109]

我用的r50 800×320

---- 回复的原邮件 ---- | 发件人 | @.> | | 日期 | 2023年10月23日 16:30 | | 收件人 | @.> | | 抄送至 | @.>@.> | | 主题 | Re: [junjie18/CMT] I used 8 3090GPUs to reproduce the result, which was 2 lower than the official result. (Issue #79) |

@xu19971109 15e + 5e means the last 5 epochs should be trained without GTsample.

能不能把训练日志发我一下，v299+1600*640,的，我只有4张3090，训练太慢了，哎

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

[rubbish001]

@xu19971109 15e + 5e means the last 5 epochs should be trained without GTsample. 你第十五轮的val的精度多少，我的是NDS是70.8，mAP是0.65

[CBY-9527]

@xu19971109 15e + 5e means the last 5 epochs should be trained without GTsample.

能不能把训练日志发我一下，v299+1600*640,的，我只有4张3090，训练太慢了，哎

[CBY-9527]

@xu19971109 15e + 5e means the last 5 epochs should be trained without GTsample.

能不能把训练日志发我一下，v299+1600*640,的，我只有4张3090，训练太慢了，哎

您好，想了解一下4卡3090训练的话大概需要多久呢。