[Core] Make encoder-decoder inputs a nested structure to be more composable (#9604)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

tests/tokenization/test_detokenize.py

@@ -3,6 +3,7 @@ from typing import Any, Dict, Generator, List, Optional
 import pytest
 from transformers import AutoTokenizer
 
+from vllm.inputs import token_inputs
 from vllm.sequence import Logprob, SamplingParams, Sequence, SequenceGroup
 from vllm.transformers_utils.detokenizer import (Detokenizer,
                                                  detokenize_incrementally)
@@ -169,10 +170,7 @@ def create_sequence(prompt_token_ids=None):
     prompt_token_ids = prompt_token_ids or [1]
     return Sequence(
         seq_id=0,
-        inputs={
-            "prompt": "<s>",
-            "prompt_token_ids": prompt_token_ids,
-        },
+        inputs=token_inputs(prompt_token_ids, prompt="<s>"),
         block_size=16,
     )