fix: empty prompt crashing the server (#912)

Co-authored-by: Max de Bayser <[email protected]>

aphrodite/engine/aphrodite_engine.py

@@ -524,6 +524,8 @@ def _add_processed_request(
         lora_request: Optional[LoRARequest],
         prompt_adapter_request: Optional[PromptAdapterRequest] = None,
     ) -> None:
+        # to prevent empty prompts crashing the engine
+        self._validate_model_inputs(processed_inputs)
         # Create the sequences.
         block_size = self.cache_config.block_size
         seq_id = next(self.seq_counter)
@@ -1467,5 +1469,12 @@ def is_encoder_decoder_model(self):
     def is_embedding_model(self):
         return self.model_config.is_embedding_model
 
+    def _validate_model_inputs(self, inputs: Union[LLMInputs,
+                                                   EncoderDecoderLLMInputs]):
+        prompt_key = "encoder_prompt_token_ids" \
+            if self.is_encoder_decoder_model() else "prompt_token_ids"
+        if not inputs.get(prompt_key):
+            raise ValueError("Prompt cannot be empty")
+
 
 setup_logger()

tests/endpoints/llm/test_prompt_validation.py

@@ -0,0 +1,9 @@
+import pytest
+
+from aphrodite import LLM
+
+
+def test_empty_prompt():
+    llm = LLM(model="gpt2")
+    with pytest.raises(ValueError, match='Prompt cannot be empty'):
+        llm.generate([""])

tests/endpoints/openai/test_prompt_validation.py

@@ -0,0 +1,21 @@
+# imports for guided decoding tests
+import re
+
+import openai
+import pytest
+
+from ...utils import RemoteOpenAIServer
+
+
+@pytest.mark.asyncio
+async def test_empty_prompt():
+    model_name = "gpt2"
+    server_args = ["--enforce-eager"]
+    with RemoteOpenAIServer(model_name, server_args) as remote_server:
+        client = remote_server.get_async_client()
+        with pytest.raises(openai.BadRequestError,
+                           match=re.compile('.+Prompt cannot be empty.+')):
+            await client.completions.create(model=model_name,
+                                            prompt="",
+                                            max_tokens=5,
+                                            temperature=0.0)