[Bug]: return_dict=False is not working in minicpm3

[TianmengChen]

Is there an existing issue ? / 是否已有相关的 issue ?

• I have searched, and there is no existing issue. / 我已经搜索过了，没有相关的 issue。

Describe the bug / 描述这个 bug

Add parameter return_dict=False to model.generate or from_pretrained dose not affect the value of return_dict=True in the pipeline. If direct change return_dict=False, error occurs.

To Reproduce / 如何复现

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

path = "MiniCPM3-4B"
device = "cpu"

tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(path, torch_dtype=torch.bfloat16, device_map=device, trust_remote_code=True, return_dict=False)

messages = [
    {"role": "user", "content": "推荐5个北京的景点。"},
]
model_inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True).to(device)
model.config.torchscript = True

model_outputs = model.generate(
    model_inputs,
    max_new_tokens=128,
    return_dict=False
)

output_token_ids = [
    model_outputs[i][len(model_inputs[i]):] for i in range(len(model_inputs))
]

responses = tokenizer.batch_decode(output_token_ids, skip_special_tokens=True)[0]
print(responses)

then print retunr_dict after return_dict = return_dict if return_dict is not None else self.config.use_return_dict

Expected behavior / 期望的结果

No response

Screenshots / 截图

No response

Environment / 环境

- OS: [e.g. Ubuntu 20.04]
- Pytorch: [e.g. torch 2.0.0]
- CUDA: [e.g. CUDA 11.8]
- Device: [e.g. A10, RTX3090]

Additional context / 其他信息

No response