A BUG in planllm_qformer.py

[ross-Hr]

def get_text_encoder(self):
    # return self.qformer.bert
    return None

I get None in class PlanLLM_qformer when used in retrieval_utils.py

 for j in range(0, len(topk_idx), bs):
                if j + bs > len(topk_idx):
                    output = text_encoder(
                        encoder_embeds=text_feats[topk_idx[j:]],
                        attention_mask=text_atts[topk_idx[j:]],
                        encoder_hidden_states=left_encoder_output,
                        encoder_attention_mask=left_encoder_att,
                        return_dict=True,
                        mode="fusion",
                    )
                else:
                    output = text_encoder(
                        encoder_embeds=text_feats[topk_idx[j : j + bs]],
                        attention_mask=text_atts[topk_idx[j : j + bs]],
                        encoder_hidden_states=encoder_output,
                        encoder_attention_mask=encoder_att,
                        return_dict=True,
                        mode="fusion",
                    )

[idejie]

you can try implement it with the text_encoder from the self.qformer.bert