Summer
Summer
> I met this problem under pytorch 0.4.0, python 3.5.6, cuda 11.1, gcc 7.6 and solved it by changing environment to **pytorch 0.4.0, python 3.5.6, cuda 9.0, gcc 5.5.0**. After...
Sure. You can locate the library file "from transformers import CLIPModel". Then you may find the following code， ``` text_outputs = self.text_model( input_ids=input_ids, attention_mask=attention_mask, position_ids=position_ids, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) pooled_output...
> > 对不起，谢谢你的优秀项目。我已经成功地在两个数据集上重现了它，MSRVTT 和 MSVD。 > > 但是当我在 LSMDC 数据集上重现它时，性能比你的差得多。经过分析，我认为可能是因为我压缩了 LSMDC 数据集，降低了帧率 fps=3，并将分辨率调整为 224*298。 > > 因此，我想咨询您是否在实验中对 LSMDC 数据集进行了类似的处理。非常感谢！ > > 请问您能分享一下LSMDC数据集吗，我没有权限下载，希望您能将文件分享到[[email protected]](mailto:[email protected])。非常感谢您 I'm very sorry that I can't share the...
> @overwhelmedxh ， have you reproduced the results on LSMDC dataset ? I doubt the compression may be the key factor which influence the retrieval results Not yet. Our findings...
> it seems that when setting “merge=True”,the results are better than the paper presented? Yes. It seems that the results reported in the paper are obtained by setting “merge=True” without...