We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
如题,当设置inference_type=ar2,cross_attention_forward()函数会报缺少lr_imgs,点进代码(model.py line 426)可以看到确实没有传递lr_imgs,奇怪的是选择full时没有这个bug,请问是什么原因呢
The text was updated successfully, but these errors were encountered:
我也遇到了相同的问题,图比较大的时候full出现oom,ar1跑的比较慢,ar2存在bug不能使用
Sorry, something went wrong.
可以先把推理脚本里的cross_lr删掉跑下试试,我会尽快修一下这个bug
删掉了cross_lr跑的时候,只能单卡跑不能多卡并行。
(model.py line 718) output, *output_per_layers = self.model_forward(*args, hw=[vit_block_bsize, vit_block_bsize], mems=mems, inference=2, **kwargs)
手动添加了 lr_imgs=lr_imgs 依然跑不了,报错k must have shape (batch_size, seqlen_k, num_heads_k, head_size_og).
No branches or pull requests
如题,当设置inference_type=ar2,cross_attention_forward()函数会报缺少lr_imgs,点进代码(model.py line 426)可以看到确实没有传递lr_imgs,奇怪的是选择full时没有这个bug,请问是什么原因呢
The text was updated successfully, but these errors were encountered: