Trouble when used on a fine-tuned on my own dataset: TypeError: argument after ** must be a mapping, not Tensor

[Novampe]

I fine-tuned a Huggingface model (hfl/chinese-roberta-wwm-ext) to do sequence classification task on my own dataset. The fine-tuning process followed the official getting-started guide (https://transformers.run/) and the fine-tuned model ran successfully completed the sequence classification task. I am now seeking to do interpretable analysis with this package but there are some problems. I would greatly appreciate any help or insights that anyone could provide regarding this issue.

Here is part of my code. I firstly instantiated a model and loaded the weights from the saved fine-tuned model. I also loaded the pretrained tokenizer:

class BertForClas(BertPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.bert = BertModel(config, add_pooling_layer=False)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)
        self.classifier = nn.Linear(768, 2)
        self.post_init()

    def forward(self, x):
        bert_output = self.bert(**x)
        cls_vectors = bert_output.last_hidden_state[:, 0, :]
        cls_vectors = self.dropout(cls_vectors)
        logits = self.classifier(cls_vectors)
        return logits

model_0 = BertForClas(config).to(device)
model_0.load_state_dict(torch.load("/path_of_my_finetuned_model.bin"))

tokenizer = BertTokenizerFast.from_pretrained("/path_of_the_downloaded_original_model")

Then I made an explainer and tried:

cls_explainer = SequenceClassificationExplainer(model_0, tokenizer)
text = 'I love you, I like you'
word_attributions = cls_explainer(text)

Here I got the error:

  File /this_code.py:186 in forward
    bert_output = self.bert(**x)

TypeError: BertModel(
  (embeddings): BertEmbeddings(
    (word_embeddings): Embedding(21128, 768, padding_idx=0)
    (position_embeddings): Embedding(512, 768)
    (token_type_embeddings): Embedding(2, 768)
    (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
    (dropout): Dropout(p=0.1, inplace=False)
  )
  (encoder): BertEncoder(
    (layer): ModuleList(
      (0-11): 12 x BertLayer(
        (attention): BertAttention(
          (self): BertSdpaSelfAttention(
            (query): Linear(in_features=768, out_features=768, bias=True)
            (key): Linear(in_features=768, out_features=768, bias=True)
            (value): Linear(in_features=768, out_features=768, bias=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
          (output): BertSelfOutput(
            (dense): Linear(in_features=768, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (intermediate): BertIntermediate(
          (dense): Linear(in_features=768, out_features=3072, bias=True)
          (intermediate_act_fn): GELUActivation()
        )
        (output): BertOutput(
          (dense): Linear(in_features=3072, out_features=768, bias=True)
          (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
          (dropout): Dropout(p=0.1, inplace=False)
        )
      )
    )
  )
) argument after ** must be a mapping, not Tensor

I guess the inputed text was not converted by the tokenizer into the mapping format ({input_ids, attention_mask, token_type_ids}), but I have no idea how to fix this. Any assistance or guidance on this matter would be greatly appreciated!