search

MediaBrain-SJTU / MVFA-AD

[CVPR2024 Highlight] Adapting Visual-Language Models for Generalizable Anomaly Detection in Medical Images
MIT License
110 stars 16 forks source link

batch_size problem #10

Closed sydney72380 closed 3 months ago

sydney72380 commented 3 months ago

代码默认的batch_size是1，但是论文里说用的是16，改成batch_size=16，会报错： File "/home/cwj/sata1/work/paper/medical/code/MVFA-AD-main/CLIP/adapter.py", line 81, in forward out_attn = out_attn + attn_out[i][0, 0, 1:].view(H, H)报错IndexError: list index out of range， 有人遇到过吗？

7HHHHH commented 2 months ago

请问您是如何解决的？

sydney72380 commented 2 months ago

好像是哪一步维度不太对，调整一下就好了，具体的有点忘了。你一步步断点跟一下，我印象中好像不难解决