can not reproduce the results of VOC

[wuwuwuxxx]

Hello,

Thank you for your valuable contribution to Cascade-CLIP. Your work has been truly inspiring to me. However, I have encountered challenges in reproducing the results of the VOC dataset using the released code and the default settings (The VOC_aug datasets have been used). Despite multiple attempts, I achieved the best result with an 80% mIoU on the unseen classes. Could you kindly provide the training log or offer some suggestions to help me address this issue? By the way, I have successfully reproduced the results of Context.