ValueError: too many values to unpack (expected 3)

[ontheway-arch]

File "/data2/xingyc/LG-VQA/train.py", line 305, in
train_loss, train_acc, train_preds = train_or_eval_model(model, train_loader, optimizer, "Train")
File "/data2/xingyc/LG-VQA/train.py", line 199, in train_or_eval_model
loss, pred, label, qid = model(batch)
File "/data2/xingyc/anaconda3/envs/lgvqa/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1518, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/data2/xingyc/anaconda3/envs/lgvqa/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1527, in _call_impl
return forward_call(*args, **kwargs)
File "/data2/xingyc/LG-VQA/models.py", line 178, in forward
similarity_logits = self.score_input(images, texts, knowledge, qid)
File "/data2/xingyc/LG-VQA/models.py", line 152, in score_input
logits, features, image_embeds = self.model(batch, match_head="itm")
ValueError: too many values to unpack (expected 3)

I build LAVIS from source.
I find in the class Blip2ITM, it only return one value, so what the other two params (features, image_embeds)means?

if match_head == "itm":
        query_tokens = self.query_tokens.expand(image_embeds.shape[0], -1, -1)
        query_atts = torch.ones(query_tokens.size()[:-1], dtype=torch.long).to(
            image.device
        )
        attention_mask = torch.cat([query_atts, text.attention_mask], dim=1)
        output_itm = self.Qformer.bert(
            text.input_ids,
            query_embeds=query_tokens,
            attention_mask=attention_mask,
            encoder_hidden_states=image_embeds,
            encoder_attention_mask=image_atts,
            return_dict=True,
        )
        itm_embeddings = output_itm.last_hidden_state[:, : query_tokens.size(1), :]
        itm_logit = self.itm_head(itm_embeddings)
        itm_logit = itm_logit.mean(dim=1)

        return itm_logit