[Bug] use new 4bits quantizated models of internlm2, decoded word starts with a blank.

### Checklist

- [ ] 1. I have searched related issues but cannot get the expected help.
- [ ] 2. The bug has not been fixed in the latest version.
- [ ] 3. Please note that if the bug-related issue you submitted lacks corresponding environment info and a minimal reproducible demo, it will be challenging for us to reproduce and resolve the issue, reducing the likelihood of receiving feedback.

### Describe the bug


### Reproduction

1. do quantization on transformers>=4.45.0

lmdeploy lite auto_awq internlm/internlm2_5-7b-chat --work-dir internlm2_5-7b-chat-inner-4bits --batch-size 32

3. run script
```from transformers import AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained('/nvme/qa_test_models/internlm/internlm2_5-7b-chat-inner-4bits', trust_remote_code=True)

decode = tokenizer.decode(2)
print(decode)
print(tokenizer.encode(decode))
```

### Environment

```Shell
transformers>=4.45.0
```

### Error traceback

_No response_