init

pipeline/kotoba_whisper.py

@@ -1,6 +1,5 @@
 import requests
-from typing import Union, Optional, Dict, List, Any
-from collections import defaultdict
+from typing import Union, Optional, Dict
 
 import torch
 import numpy as np
@@ -24,25 +23,13 @@ class Punctuator:
     def __init__(self, model: str = "pcs_47lang"):
         self.punctuation_model = PunctCapSegModelONNX.from_pretrained(model)
 
-    def punctuate(self, pipeline_chunk: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
-
-        def validate_punctuation(raw: str, punctuated: str):
-            if 'unk' in punctuated.lower() or any(p in raw for p in self.ja_punctuations):
-                return raw
-            if punctuated.count("。") > 1:
-                ind = punctuated.rfind("。")
-                punctuated = punctuated.replace("。", "")
-                punctuated = punctuated[:ind] + "。" + punctuated[ind:]
-            return punctuated
-
-        text_edit = self.punctuation_model.infer([c['text'] for c in pipeline_chunk])
-        return [
-            {
-                'timestamp': c['timestamp'],
-                'speaker': c['speaker'],
-                'text': validate_punctuation(c['text'], "".join(e))
-            } for c, e in zip(pipeline_chunk, text_edit)
-        ]
+    def punctuate(self, text: str) -> str:
+        if any(p in text for p in self.ja_punctuations):
+            return text
+        punctuated = "".join(self.punctuation_model.infer([text])[0])
+        if 'unk' in punctuated.lower():
+            return text
+        return punctuated
 
 
 class SpeakerDiarization:
@@ -114,104 +101,68 @@ class KotobaWhisperPipeline(AutomaticSpeechRecognitionPipeline):
         )
 
     def _sanitize_parameters(self,
-                             chunk_length_s=None,
-                             stride_length_s=None,
-                             ignore_warning=None,
-                             decoder_kwargs=None,
-                             return_timestamps=None,
-                             return_language=None,
-                             generate_kwargs=None,
-                             max_new_tokens=None,
-                             add_punctuation: bool =False,
-                             return_unique_speaker: bool =True,
+                             chunk_length_s: Optional[int] = None,
+                             stride_length_s: Optional[int] = None,
+                             generate_kwargs: Optional[Dict] = None,
+                             max_new_tokens: Optional[int] = None,
+                             add_punctuation: bool = False,
+                             return_unique_speaker: bool = True,
+                             add_silence_end: Optional[float] = None,
+                             add_silence_start: Optional[float] = None,
                              num_speakers: Optional[int] = None,
                              min_speakers: Optional[int] = None,
                              max_speakers: Optional[int] = None):
-        # No parameters on this pipeline right now
-        preprocess_params = {}
-        if chunk_length_s is not None:
-            preprocess_params["chunk_length_s"] = chunk_length_s
-        if stride_length_s is not None:
-            preprocess_params["stride_length_s"] = stride_length_s
-
-        forward_params = defaultdict(dict)
-        if max_new_tokens is not None:
-            forward_params["max_new_tokens"] = max_new_tokens
-        if generate_kwargs is not None:
-            if max_new_tokens is not None and "max_new_tokens" in generate_kwargs:
-                raise ValueError(
-                    "`max_new_tokens` is defined both as an argument and inside `generate_kwargs` argument, please use"
-                    " only 1 version"
-                )
-            forward_params.update(generate_kwargs)
-
-        postprocess_params = {}
-        if decoder_kwargs is not None:
-            postprocess_params["decoder_kwargs"] = decoder_kwargs
-        if return_timestamps is not None:
-            # Check whether we have a valid setting for return_timestamps and throw an error before we perform a forward pass
-            if self.type == "seq2seq" and return_timestamps:
-                raise ValueError("We cannot return_timestamps yet on non-CTC models apart from Whisper!")
-            if self.type == "ctc_with_lm" and return_timestamps != "word":
-                raise ValueError("CTC with LM can only predict word level timestamps, set `return_timestamps='word'`")
-            if self.type == "ctc" and return_timestamps not in ["char", "word"]:
-                raise ValueError(
-                    "CTC can either predict character level timestamps, or word level timestamps. "
-                    "Set `return_timestamps='char'` or `return_timestamps='word'` as required."
-                )
-            if self.type == "seq2seq_whisper" and return_timestamps == "char":
-                raise ValueError(
-                    "Whisper cannot return `char` timestamps, only word level or segment level timestamps. "
-                    "Use `return_timestamps='word'` or `return_timestamps=True` respectively."
-                )
-            forward_params["return_timestamps"] = return_timestamps
-            postprocess_params["return_timestamps"] = return_timestamps
-        if return_language is not None:
-            if self.type != "seq2seq_whisper":
-                raise ValueError("Only Whisper can return language for now.")
-            postprocess_params["return_language"] = return_language
-        postprocess_params["return_language"] = return_language
-        postprocess_params["add_punctuation"] = add_punctuation
-        postprocess_params["return_unique_speaker"] = return_unique_speaker
-        postprocess_params["num_speakers"] = num_speakers
-        postprocess_params["min_speakers"] = min_speakers
-        postprocess_params["max_speakers"] = max_speakers
+        preprocess_params = {
+            "chunk_length_s": chunk_length_s,
+            "stride_length_s": stride_length_s,
+            "add_silence_end": add_silence_end,
+            "add_silence_start": add_silence_start,
+            "num_speakers": num_speakers,
+            "min_speakers": min_speakers,
+            "max_speakers": max_speakers,
+        }
+        postprocess_params = {"add_punctuation": add_punctuation, "return_timestamps": True, "return_language": False}
+        forward_params = {} if generate_kwargs is None else generate_kwargs
+        forward_params.update({"max_new_tokens": max_new_tokens, "return_timestamps": True})
         return preprocess_params, forward_params, postprocess_params
 
-    def preprocess(self, inputs, chunk_length_s=0, stride_length_s=None):
+    def preprocess(self,
+                   inputs,
+                   chunk_length_s: Optional[int] = None,
+                   stride_length_s: Optional[int] = None,
+                   add_silence_end: Optional[float] = None,
+                   add_silence_start: Optional[float] = None,
+                   num_speakers: Optional[int] = None,
+                   min_speakers: Optional[int] = None,
+                   max_speakers: Optional[int] = None):
+
+        def _pad_audio_array(_audio):
+            if add_silence_start:
+                _audio = np.concatenate([np.zeros(int(self.feature_extractor.sampling_rate * add_silence_start)), _audio])
+            if add_silence_end:
+                _audio = np.concatenate([_audio, np.zeros(int(self.feature_extractor.sampling_rate * add_silence_end))])
+            return _audio
+
+        # load file
         if isinstance(inputs, str):
             if inputs.startswith("http://") or inputs.startswith("https://"):
-                # We need to actually check for a real protocol, otherwise it's impossible to use a local file
-                # like http_huggingface_co.png
+                # We need to actually check for a real protocol, otherwise it's impossible to use a local file like http_huggingface_co.png
                 inputs = requests.get(inputs).content
             else:
                 with open(inputs, "rb") as f:
                     inputs = f.read()
-
         if isinstance(inputs, bytes):
             inputs = ffmpeg_read(inputs, self.feature_extractor.sampling_rate)
-
-        stride = None
-        extra = {}
         if isinstance(inputs, dict):
-            stride = inputs.pop("stride", None)
-            # Accepting `"array"` which is the key defined in `datasets` for
-            # better integration
-            if not ("sampling_rate" in inputs and ("raw" in inputs or "array" in inputs)):
+            # Accepting `"array"` which is the key defined in `datasets` for better integration
+            if not ("sampling_rate" in inputs and "array" in inputs):
                 raise ValueError(
                     "When passing a dictionary to AutomaticSpeechRecognitionPipeline, the dict needs to contain a "
-                    '"raw" key containing the numpy array representing the audio and a "sampling_rate" key, '
+                    '"array" key containing the numpy array representing the audio and a "sampling_rate" key, '
                     "containing the sampling_rate associated with that array"
                 )
-
-            _inputs = inputs.pop("raw", None)
-            if _inputs is None:
-                # Remove path which will not be used from `datasets`.
-                inputs.pop("path", None)
-                _inputs = inputs.pop("array", None)
             in_sampling_rate = inputs.pop("sampling_rate")
-            extra = inputs
-            inputs = _inputs
+            inputs = inputs.pop("array", None)
             if in_sampling_rate != self.feature_extractor.sampling_rate:
                 if is_torchaudio_available():
                     from torchaudio import functional as F
@@ -220,190 +171,100 @@ class KotobaWhisperPipeline(AutomaticSpeechRecognitionPipeline):
                         "torchaudio is required to resample audio samples in AutomaticSpeechRecognitionPipeline. "
                         "The torchaudio package can be installed through: `pip install torchaudio`."
                     )
-
                 inputs = F.resample(
                     torch.from_numpy(inputs), in_sampling_rate, self.feature_extractor.sampling_rate
                 ).numpy()
-                ratio = self.feature_extractor.sampling_rate / in_sampling_rate
-            else:
-                ratio = 1
-            if stride is not None:
-                if stride[0] + stride[1] > inputs.shape[0]:
-                    raise ValueError("Stride is too large for input")
 
-                # Stride needs to get the chunk length here, it's going to get
-                # swallowed by the `feature_extractor` later, and then batching
-                # can add extra data in the inputs, so we need to keep track
-                # of the original length in the stride so we can cut properly.
-                stride = (inputs.shape[0], int(round(stride[0] * ratio)), int(round(stride[1] * ratio)))
+        # validate audio array
         if not isinstance(inputs, np.ndarray):
             raise ValueError(f"We expect a numpy ndarray as input, got `{type(inputs)}`")
         if len(inputs.shape) != 1:
             raise ValueError("We expect a single channel audio input for AutomaticSpeechRecognitionPipeline")
 
-        if chunk_length_s:
-            if stride_length_s is None:
-                stride_length_s = chunk_length_s / 6
-
-            if isinstance(stride_length_s, (int, float)):
-                stride_length_s = [stride_length_s, stride_length_s]
-
-            # XXX: Carefuly, this variable will not exist in `seq2seq` setting.
-            # Currently chunking is not possible at this level for `seq2seq` so
-            # it's ok.
-            align_to = getattr(self.model.config, "inputs_to_logits_ratio", 1)
-            chunk_len = int(round(chunk_length_s * self.feature_extractor.sampling_rate / align_to) * align_to)
-            stride_left = int(round(stride_length_s[0] * self.feature_extractor.sampling_rate / align_to) * align_to)
-            stride_right = int(round(stride_length_s[1] * self.feature_extractor.sampling_rate / align_to) * align_to)
-
-            if chunk_len < stride_left + stride_right:
-                raise ValueError("Chunk length must be superior to stride length")
-
-            for item in chunk_iter(
-                    inputs, self.feature_extractor, chunk_len, stride_left, stride_right, self.torch_dtype
-            ):
-                item["audio_array"] = inputs
-                yield item
-        else:
-            if inputs.shape[0] > self.feature_extractor.n_samples:
-                processed = self.feature_extractor(
-                    inputs,
-                    sampling_rate=self.feature_extractor.sampling_rate,
-                    truncation=False,
-                    padding="longest",
-                    return_tensors="pt",
-                )
-            else:
-                processed = self.feature_extractor(
-                    inputs, sampling_rate=self.feature_extractor.sampling_rate, return_tensors="pt"
-                )
-
-            if self.torch_dtype is not None:
-                processed = processed.to(dtype=self.torch_dtype)
-            if stride is not None:
-                processed["stride"] = stride
-            yield {"is_last": True, "audio_array": inputs, **processed, **extra}
-
-    def _forward(self, model_inputs, **generate_kwargs):
-        attention_mask = model_inputs.pop("attention_mask", None)
-        stride = model_inputs.pop("stride", None)
-        is_last = model_inputs.pop("is_last")
-        audio_array = model_inputs.pop("audio_array")
-        encoder = self.model.get_encoder()
-        # Consume values so we can let extra information flow freely through
-        # the pipeline (important for `partial` in microphone)
-        if "input_features" in model_inputs:
-            inputs = model_inputs.pop("input_features")
-        elif "input_values" in model_inputs:
-            inputs = model_inputs.pop("input_values")
-        else:
-            raise ValueError(
-                "Seq2Seq speech recognition model requires either a "
-                f"`input_features` or `input_values` key, but only has {model_inputs.keys()}"
-            )
-
-        # custom processing for Whisper timestamps and word-level timestamps
-        generate_kwargs["return_timestamps"] = True
-        if inputs.shape[-1] > self.feature_extractor.nb_max_frames:
-            generate_kwargs["input_features"] = inputs
-        else:
-            generate_kwargs["encoder_outputs"] = encoder(inputs, attention_mask=attention_mask)
-
-        tokens = self.model.generate(attention_mask=attention_mask, **generate_kwargs)
-        # whisper longform generation stores timestamps in "segments"
-        out = {"tokens": tokens}
-        if self.type == "seq2seq_whisper":
-            if stride is not None:
-                out["stride"] = stride
-
-        # Leftover
-        extra = model_inputs
-        return {"is_last": is_last, "audio_array": audio_array, **out, **extra}
-
-    def postprocess(self,
-                    model_outputs,
-                    decoder_kwargs: Optional[Dict] = None,
-                    return_language=None,
-                    add_punctuation: bool = False,
-                    return_unique_speaker: bool = True,
-                    num_speakers: Optional[int] = None,
-                    min_speakers: Optional[int] = None,
-                    max_speakers: Optional[int] = None,
-                    *args,
-                    **kwargs):
-        assert len(model_outputs) > 0
-        outputs = super().postprocess(
-            model_outputs=model_outputs,
-            decoder_kwargs=decoder_kwargs,
-            return_timestamps=True,
-            return_language=return_language
-        )
-        audio_array = outputs.pop("audio_array")[0]
+        # diarization
         sd = self.model_speaker_diarization(
-            audio_array,
+            inputs,
             num_speakers=num_speakers,
             min_speakers=min_speakers,
             max_speakers=max_speakers,
             sampling_rate=self.feature_extractor.sampling_rate
         )
-        diarization_result = {s: [[i.start, i.end] for i in sd.label_timeline(s)] for s in sd.labels()}
-        timelines = sd.get_timeline()
-
-        pointer_ts = 0
-        pointer_chunk = 0
-        new_chunks = []
-        while True:
-            if pointer_ts == len(timelines):
-                ts = timelines[-1]
-                for chunk in outputs["chunks"][pointer_chunk:]:
-                    chunk["speaker"] = sd.get_labels(ts)
-                    new_chunks.append(chunk)
-                break
-            if pointer_chunk == len(outputs["chunks"]):
-                break
-            ts = timelines[pointer_ts]
-
-            chunk = outputs["chunks"][pointer_chunk]
-            if "speaker" not in chunk:
-                chunk["speaker"] = []
 
-            start, end = chunk["timestamp"]
-            if ts.end <= start:
-                pointer_ts += 1
-            elif end <= ts.start:
-                if len(chunk["speaker"]) == 0:
-                    chunk["speaker"] += list(sd.get_labels(ts))
-                new_chunks.append(chunk)
-                pointer_chunk += 1
-            else:
-                chunk["speaker"] += list(sd.get_labels(ts))
-                if ts.end >= end:
-                    new_chunks.append(chunk)
-                    pointer_chunk += 1
-                else:
-                    pointer_ts += 1
-        for i in new_chunks:
-            if "speaker" in i:
-                if return_unique_speaker:
-                    i["speaker"] = [i["speaker"][0]]
+        # loop over audio chunks and speakers
+        labels = list(sd.labels())
+        for n, s in enumerate(labels):
+            timelines = list(sd.label_timeline(s))
+            for m, i in enumerate(timelines):
+                start = int(i.start * self.feature_extractor.sampling_rate)
+                end = int(i.end * self.feature_extractor.sampling_rate)
+                audio_array = _pad_audio_array(inputs[start: end])
+
+                if chunk_length_s is not None:
+                    stride_length_s = chunk_length_s / 6 if stride_length_s is None else stride_length_s
+                    stride_length_s = [stride_length_s, stride_length_s] if isinstance(stride_length_s, (int, float)) else stride_length_s
+                    align_to = getattr(self.model.config, "inputs_to_logits_ratio", 1)
+                    chunk_len = int(round(chunk_length_s * self.feature_extractor.sampling_rate / align_to) * align_to)
+                    stride_left = int(round(stride_length_s[0] * self.feature_extractor.sampling_rate / align_to) * align_to)
+                    stride_right = int(round(stride_length_s[1] * self.feature_extractor.sampling_rate / align_to) * align_to)
+                    if chunk_len < stride_left + stride_right:
+                        raise ValueError("Chunk length must be superior to stride length")
+                    for item in chunk_iter(
+                            audio_array, self.feature_extractor, chunk_len, stride_left, stride_right, self.torch_dtype
+                    ):
+                        item["speaker_id"] = s
+                        item["speaker_span"] = [i.start, i.end]
+                        item["is_last"] = m == len(timelines) - 1 and n == len(labels) - 1 and item["is_last"]
+                        yield item
                 else:
-                    i["speaker"] = list(set(i["speaker"]))
-            else:
-                i["speaker"] = []
-        outputs["chunks"] = new_chunks
-        if add_punctuation:
-            if self.punctuator is None:
-                self.punctuator = Punctuator()
-            outputs["chunks"] = self.punctuator.punctuate(outputs["chunks"])
-        outputs["text"] = "".join([c["text"] for c in outputs["chunks"]])
-        outputs["speakers"] = sd.labels()
-        speakers = []
-        for s in outputs["speakers"]:
-            chunk_s = [c for c in outputs["chunks"] if s in c["speaker"]]
-            if len(chunk_s) != 0:
-                outputs[f"chunks/{s}"] = chunk_s
-                outputs[f"text/{s}"] = "".join([c["text"] for c in outputs["chunks"] if s in c["speaker"]])
-                speakers.append(s)
-        outputs["speakers"] = speakers
-        outputs["diarization_result"] = diarization_result
+                    if inputs.shape[0] > self.feature_extractor.n_samples:
+                        processed = self.feature_extractor(
+                            audio_array,
+                            sampling_rate=self.feature_extractor.sampling_rate,
+                            truncation=False,
+                            padding="longest",
+                            return_tensors="pt",
+                        )
+                    else:
+                        processed = self.feature_extractor(
+                            audio_array,
+                            sampling_rate=self.feature_extractor.sampling_rate,
+                            return_tensors="pt"
+                        )
+                    if self.torch_dtype is not None:
+                        processed = processed.to(dtype=self.torch_dtype)
+                    processed["speaker_id"] = s
+                    processed["speaker_span"] = [i.start, i.end]
+                    processed["is_last"] = m == len(timelines) - 1 and n == len(labels) - 1
+                    yield processed
+
+    def _forward(self, model_inputs, **generate_kwargs):
+        generate_kwargs["attention_mask"] = model_inputs.pop("attention_mask", None)
+        generate_kwargs["input_features"] = model_inputs.pop("input_features")
+        tokens = self.model.generate(**generate_kwargs)
+        return {"tokens": tokens, **model_inputs}
+
+    def postprocess(self, model_outputs, **postprocess_parameters):
+        if postprocess_parameters["add_punctuation"] and self.punctuator is None:
+            self.punctuator = Punctuator()
+        outputs = {"chunks": []}
+        for o in model_outputs:
+            text, chunks = self.tokenizer._decode_asr(
+                [o],
+                return_language=postprocess_parameters["return_language"],
+                return_timestamps=postprocess_parameters["return_timestamps"],
+                time_precision=self.feature_extractor.chunk_length / self.model.config.max_source_positions,
+            )
+            start, end = o["speaker_span"]
+            new_chunk = []
+            for c in chunks["chunks"]:
+                c["timestamp"] = [round(c["timestamp"][0] + start, 2), round(c["timestamp"][0] + end, 2)]
+                c["speaker_id"] = o["speaker_id"]
+                new_chunk.append(c)
+            outputs["chunks"] += new_chunk
+        outputs["speaker_ids"] = sorted(set([o["speaker_id"] for o in outputs["chunks"]]))
+        for s in outputs["speaker_ids"]:
+            outputs[f"chunk/{s}"] = sorted([o for o in outputs["chunks"] if o["speaker_id"] == s], key=lambda x: x["timestamp"][0])
+            outputs[f"text/{s}"] = "".join([i["text"] for i in outputs[f"chunk/{s}"]])
+            if postprocess_parameters["add_punctuation"]:
+                outputs[f"text/{s}"] = self.punctuator.punctuate(outputs[f"text/{s}"])
         return outputs

sample_audio/sample_diarization_japanese.mp3

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e7252359b53264c767da33a48e39ff57a8f31641c4a80a1702c6940f8914697b
-size 780064
+oid sha256:75afe01062ebd4b4e3555e8ef41becd56274e2d9b4f8c26919f870636a1c55f6
+size 857719