Optionally include "passage" in BoolQ samples

ultravox/data/datasets.py

@@ -182,13 +182,26 @@ class DatasetSplit(str, enum.Enum):
 @dataclasses.dataclass
 class VoiceDatasetArgs:
     data_dir: Optional[str] = None
+    prompt: Optional[str] = None
+    """A specific prompt to use for the dataset."""
     num_prompts: int = 1
+    """If `prompt` is not set, the number of canned prompts to use."""
+    include_audio: bool = True
+    """Whether to include audio in the samples."""
+    include_context: bool = False
+    """Whether to include additional textual context from the dataset to the prompt."""
     shuffle: bool = False
+    """Whether to shuffle the dataset."""
     shuffle_seed: int = 42
+    """Seed for shuffling the dataset."""
     max_audio_duration_secs: Optional[float] = None
+    """Whether to skip samples with audio longer than this duration."""
     use_mds: bool = False
+    """Whether to load the dataset from GCP (using MDS) or Hugging Face."""
     mds_batch_size: int = 32
+    """Batch size for MDS."""
     split: DatasetSplit = DatasetSplit.TRAIN
+    """Which split of the dataset to use."""
 
     def __post_init__(self):
         if isinstance(self.split, str):
@@ -261,17 +274,25 @@ def _get_sample(self, idx: int, row: transformers.BatchFeature) -> VoiceSample:
         pass
 
     def _get_answer_prompt(self, idx: int) -> str:
+        if self._args.prompt:
+            return self._args.prompt
         prompt_idx = idx % min(self._args.num_prompts, len(ANSWER_PROMPTS))
         return ANSWER_PROMPTS[prompt_idx]
 
     def _get_transcribe_prompt(self, idx: int) -> str:
+        if self._args.prompt:
+            return self._args.prompt
         prompt_idx = idx % min(self._args.num_prompts, len(TRANSCRIBE_PROMPTS))
         return TRANSCRIBE_PROMPTS[prompt_idx]
 
-    def _get_answer_messages(self, idx: int, text: str) -> List[Dict[str, str]]:
+    def _get_answer_messages(
+        self, idx: int, question: str, answer: str, context: Optional[str] = None
+    ) -> List[Dict[str, str]]:
+        prompt = self._get_answer_prompt(idx) if self._args.include_audio else question
+        user_content = f"{context}\n\n{prompt}" if context else prompt
         return [
-            {"role": "user", "content": self._get_answer_prompt(idx)},
-            {"role": "assistant", "content": text},
+            {"role": "user", "content": user_content},
+            {"role": "assistant", "content": answer},
         ]
 
     def _get_transcribe_messages(self, idx: int, text: str) -> List[Dict[str, str]]:
@@ -294,6 +315,19 @@ def _get_audio(self, row: transformers.BatchFeature) -> np.ndarray:
         assert sampling_rate == SAMPLE_RATE
         return audio
 
+    def _load_audio(self, base_url: str, folder: str, filename: str) -> np.ndarray:
+        if self._args.data_dir:
+            audio_path = f"{self._args.data_dir}/{folder}/{filename}"
+            audio = audio_from_file(audio_path)
+        else:
+            url = f"{base_url}/{filename}"  # hack for GCS bucket naming
+            if self._session is None:
+                self._session = requests.Session()
+            response = self._session.get(url)
+            response.raise_for_status()
+            audio = audio_from_buf(response.content)
+        return audio
+
     def _get_transcribe_sample(
         self,
         idx: int,
@@ -302,24 +336,21 @@ def _get_transcribe_sample(
         tproc: Optional[Callable[[str], str]] = None,
     ) -> VoiceSample:
         text = tproc(row[tcol]) if tproc else row[tcol]
-        return VoiceSample(
+        return self._make_sample(
             self._get_transcribe_messages(idx, text),
             self._get_audio(row),
             audio_transcript=text,
         )
 
-    def _load_audio(self, base_url: str, folder: str, filename: str) -> np.ndarray:
-        if self._args.data_dir:
-            audio_path = f"{self._args.data_dir}/{folder}/{filename}"
-            audio = audio_from_file(audio_path)
-        else:
-            url = f"{base_url}/{filename}"  # hack for GCS bucket naming
-            if self._session is None:
-                self._session = requests.Session()
-            response = self._session.get(url)
-            response.raise_for_status()
-            audio = audio_from_buf(response.content)
-        return audio
+    def _make_sample(
+        self,
+        messages: List[Dict[str, str]],
+        audio: np.ndarray,
+        audio_transcript: Optional[str] = None,
+    ) -> VoiceSample:
+        if not self._args.include_audio:
+            return VoiceSample(messages)
+        return VoiceSample(messages, audio, audio_transcript=audio_transcript)
 
 
 class LibriSpeechDummyDataset(VoiceDataset):
@@ -385,10 +416,11 @@ def __init__(self, args: VoiceDatasetArgs) -> None:
         super().__init__(args)
 
     def _get_sample(self, idx: int, row: transformers.BatchFeature) -> VoiceSample:
-        return VoiceSample(
-            self._get_answer_messages(idx, row["chat"][1]["message"]),
-            self._load_anyinstruct_audio(row["chat"][0]["speech"]),
-            audio_transcript=row["chat"][0]["message"],
+        chat = row["chat"]
+        return self._make_sample(
+            self._get_answer_messages(idx, chat[0]["message"], chat[1]["message"]),
+            self._load_anyinstruct_audio(chat[0]["speech"]),
+            audio_transcript=chat[0]["message"],
         )
 
 
@@ -428,8 +460,11 @@ def __init__(self, args: VoiceDatasetArgs) -> None:
         self._init_dataset(dataset)
 
     def _get_sample(self, idx: int, row: transformers.BatchFeature) -> VoiceSample:
-        return VoiceSample(
-            self._get_answer_messages(idx, "True" if row["answer"] else "False"),
+        question = row["question"]
+        answer = "True" if row["answer"] else "False"
+        context = row["passage"] if self._args.include_context else None
+        return self._make_sample(
+            self._get_answer_messages(idx, question, answer, context),
             self._get_audio(row),
             audio_transcript=row["question"],
         )

ultravox/tools/infer_tool.py

@@ -48,9 +48,13 @@ class InferArgs:
     # Data sets to use for inference
     data_sets: Optional[List[str]] = simple_parsing.field(default=None, alias="-d")
     # Which dataset split to use
-    data_split: datasets.DatasetSplit = datasets.DatasetSplit.VALIDATION
+    data_split: datasets.DatasetSplit = simple_parsing.field(
+        default=datasets.DatasetSplit.VALIDATION, alias="-s"
+    )
     # Directory for existing data
     data_dir: Optional[str] = None
+    # Use dataset context
+    context: bool = False
     # Load datasets using MDS
     mds: bool = False
     # Number of dataset samples to process
@@ -128,10 +132,9 @@ def run_tui(
         eval_str = ""
         if scores is not None:
             assert args.data_sets
-            assert sample.audio_transcript is not None, "Query must have transcript"
             ds_name = args.data_sets[0]
             eval_sample = eval_types.Sample(
-                sample.audio_transcript,
+                sample.audio_transcript or sample.messages[0]["content"],
                 expected_answer=expected_response,
                 generated_answer=text,
             )
@@ -156,7 +159,8 @@ def run_tui(
         print(f"X: {expected_response}{eval_str}")
 
 
-def oneshot_infer(inference: base.VoiceInference, prompt: str, args: InferArgs):
+def oneshot_infer(inference: base.VoiceInference, args: InferArgs):
+    prompt = args.prompt or (DEFAULT_ASR_PROMPT if args.asr else DEFAULT_PROMPT)
     if args.audio_file is not None:
         sample = datasets.VoiceSample.from_prompt_and_buf(
             prompt, args.audio_file.read()
@@ -166,10 +170,13 @@ def oneshot_infer(inference: base.VoiceInference, prompt: str, args: InferArgs):
     run_tui(-1, inference, sample, args)
 
 
-def dataset_infer(inference: base.VoiceInference, prompt: str, args: InferArgs):
+def dataset_infer(inference: base.VoiceInference, args: InferArgs):
     assert args.data_sets, "At least one data set must be provided"
     ds_args = datasets.VoiceDatasetArgs(
         data_dir=args.data_dir,
+        prompt=args.prompt,
+        include_audio=not args.text_only,
+        include_context=args.context,
         shuffle=args.shuffle,
         use_mds=args.mds,
         split=args.data_split,
@@ -184,15 +191,6 @@ def dataset_infer(inference: base.VoiceInference, prompt: str, args: InferArgs):
         expected_answer = sample.messages[1]["content"]
         # Drop any assistant response from the sample.
         sample.messages = sample.messages[:1]
-        # Normally, we overwrite the dataset prompt with our prompt, allowing us to customize
-        # the inference prompt in this tool.
-        # If we're using text-only mode though, there's no audio to inference, so we
-        # just paste the text transcript in as the text prompt.
-        if not args.text_only:
-            sample.messages[0]["content"] = prompt
-        else:
-            sample.messages[0]["content"] = sample.audio_transcript
-            sample.audio = sample.audio_transcript = None
         if not args.json:
             run_tui(i, inference, sample, args, expected_answer, scores)
         else:
@@ -222,11 +220,10 @@ def main(args: InferArgs):
             device=args.device,
             data_type=args.data_type,
         )
-    prompt = args.prompt or (DEFAULT_ASR_PROMPT if args.asr else DEFAULT_PROMPT)
     if args.data_sets is None:
-        oneshot_infer(inference, prompt, args)
+        oneshot_infer(inference, args)
     else:
-        dataset_infer(inference, prompt, args)
+        dataset_infer(inference, args)
 
 
 if __name__ == "__main__":