ludysama
ludysama
I have experienced this problem, and finally I found that I changed too much codes about importing packages. so I downloaded a new one, just change the path, and it...
不好意思 刚没看到#6已经提过这个问题了 有个老哥给的答案很棒 --gpu_ids 0,0 这个参数把同一张卡写两遍就可以了
I faced with the same bug when I use Ubuntu20.04 as developing platform. The solution is type a '#' before 40 line of train.py (torch.set_default_tensor_type('torch.cuda.FloatTensor)) There may remain some bugs...
最好仿着app里的例子来写,目前我用着只有直接优化可学习的坐标是能gpu反传梯度的,如果这个坐标由网络得到,就会报错,只能用.cpu()操作先迁到cpu上才行。 你可以试试
I try to sovle this problem by initing Path with the value that net predicted, then let gpu just optimize the path's point tensor, then let it backward first, after...
我遇到了同样的问题,经过检查,是我的vicuna_delta_weights文件夹中缺少了tokenzier.model这个文件