[Bugfix] Always apply MM processor even when no MM items are passed (#26240)

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

tests/test_inputs.py

@@ -3,8 +3,11 @@
 
 import pytest
 
+from vllm.config import ModelConfig
 from vllm.inputs import zip_enc_dec_prompts
 from vllm.inputs.parse import parse_raw_prompts
+from vllm.inputs.preprocess import InputPreprocessor
+from vllm.transformers_utils.tokenizer import init_tokenizer_from_configs
 
 pytestmark = pytest.mark.cpu_test
 
@@ -80,3 +83,50 @@ def test_zip_enc_dec_prompts(mm_processor_kwargs, expected_mm_kwargs):
         assert zipped['encoder_prompt'] == enc
         assert zipped['decoder_prompt'] == dec
         assert zipped['mm_processor_kwargs'] == exp_kwargs
+
+
+@pytest.mark.parametrize("model_id", [
+    "facebook/opt-125m",
+])
+@pytest.mark.parametrize("prompt", [
+    {
+        "prompt": "",
+        "multi_modal_data": {
+            "dummy": []
+        },
+    },
+    {
+        "prompt_token_ids": [],
+        "multi_modal_data": {
+            "dummy": []
+        },
+    },
+])
+def test_preprocessor_text_no_mm_inputs(model_id, prompt):
+    model_config = ModelConfig(model=model_id)
+    tokenizer = init_tokenizer_from_configs(model_config)
+    input_preprocessor = InputPreprocessor(model_config, tokenizer)
+
+    with pytest.raises(ValueError, match="does not support multimodal inputs"):
+        input_preprocessor.preprocess(prompt)
+
+
+@pytest.mark.parametrize("model_id", [
+    "facebook/chameleon-7b",
+])
+@pytest.mark.parametrize("prompt", [
+    "",
+    {
+        "prompt_token_ids": []
+    },
+])
+def test_preprocessor_always_mm_code_path(model_id, prompt):
+    model_config = ModelConfig(model=model_id)
+    tokenizer = init_tokenizer_from_configs(model_config)
+    input_preprocessor = InputPreprocessor(model_config, tokenizer)
+
+    # HF processor adds sep token
+    sep_token_id = tokenizer.vocab[tokenizer.sep_token]
+
+    processed_inputs = input_preprocessor.preprocess(prompt)
+    assert sep_token_id in processed_inputs["prompt_token_ids"]