From c9f1b4e8a2e903f74de20d019e70307c26e93c3e Mon Sep 17 00:00:00 2001
From: shixian.shi <shixian.shi@alibaba-inc.com>
Date: 星期四, 23 十一月 2023 20:39:52 +0800
Subject: [PATCH] update

---
 funasr/bin/asr_inference_launch.py |  488 ++++++++++++++++++++++-------------------------------
 1 files changed, 206 insertions(+), 282 deletions(-)

diff --git a/funasr/bin/asr_inference_launch.py b/funasr/bin/asr_inference_launch.py
index 15dbdd4..e1a32c5 100644
--- a/funasr/bin/asr_inference_launch.py
+++ b/funasr/bin/asr_inference_launch.py
@@ -29,6 +29,7 @@
 from funasr.bin.asr_infer import Speech2TextSAASR
 from funasr.bin.asr_infer import Speech2TextTransducer
 from funasr.bin.asr_infer import Speech2TextUniASR
+from funasr.bin.asr_infer import Speech2TextWhisper
 from funasr.bin.punc_infer import Text2Punc
 from funasr.bin.tp_infer import Speech2Timestamp
 from funasr.bin.vad_infer import Speech2VadSegment
@@ -55,6 +56,7 @@
                                         distribute_spk)
 from funasr.build_utils.build_model_from_file import build_model_from_file
 from funasr.utils.cluster_backend import ClusterBackend
+from funasr.utils.modelscope_utils import get_cache_dir
 from tqdm import tqdm
 
 def inference_asr(
@@ -460,18 +462,18 @@
 
 
 def inference_paraformer_vad_punc(
-        maxlenratio: float,
-        minlenratio: float,
-        batch_size: int,
-        beam_size: int,
-        ngpu: int,
-        ctc_weight: float,
-        lm_weight: float,
-        penalty: float,
-        log_level: Union[int, str],
+        maxlenratio: float=0.0,
+        minlenratio: float=0.0,
+        batch_size: int=1,
+        beam_size: int=1,
+        ngpu: int=1,
+        ctc_weight: float=0.0,
+        lm_weight: float=0.0,
+        penalty: float=0.0,
+        log_level: Union[int, str]=logging.ERROR,
         # data_path_and_name_and_type,
-        asr_train_config: Optional[str],
-        asr_model_file: Optional[str],
+        asr_train_config: Optional[str]=None,
+        asr_model_file: Optional[str]=None,
         cmvn_file: Optional[str] = None,
         lm_train_config: Optional[str] = None,
         lm_file: Optional[str] = None,
@@ -485,7 +487,7 @@
         seed: int = 0,
         ngram_weight: float = 0.9,
         nbest: int = 1,
-        num_workers: int = 1,
+        num_workers: int = 0,
         vad_infer_config: Optional[str] = None,
         vad_model_file: Optional[str] = None,
         vad_cmvn_file: Optional[str] = None,
@@ -498,6 +500,7 @@
 ):
     ncpu = kwargs.get("ncpu", 1)
     torch.set_num_threads(ncpu)
+    language = kwargs.get("model_lang", None)
 
     if word_lm_train_config is not None:
         raise NotImplementedError("Word LM is not implemented")
@@ -672,11 +675,13 @@
                 beg_idx = end_idx
                 batch = {"speech": speech_j, "speech_lengths": speech_lengths_j}
                 batch = to_device(batch, device=device)
-                # print("batch: ", speech_j.shape[0])
+
                 beg_asr = time.time()
                 results = speech2text(**batch)
                 end_asr = time.time()
-                # print("time cost asr: ", end_asr - beg_asr)
+                if speech2text.device != "cpu":
+                    print("batch: ", speech_j.shape[0])
+                    print("time cost asr: ", end_asr - beg_asr)
 
                 if len(results) < 1:
                     results = [["", [], [], [], [], [], []]]
@@ -704,10 +709,13 @@
             text, token, token_int = result[0], result[1], result[2]
             time_stamp = result[4] if len(result[4]) > 0 else None
 
-            if use_timestamp and time_stamp is not None and len(time_stamp):
-                postprocessed_result = postprocess_utils.sentence_postprocess(token, time_stamp)
+            if language == "en-bpe":
+                postprocessed_result = postprocess_utils.sentence_postprocess_sentencepiece(token)
             else:
-                postprocessed_result = postprocess_utils.sentence_postprocess(token)
+                if use_timestamp and time_stamp is not None and len(time_stamp):
+                    postprocessed_result = postprocess_utils.sentence_postprocess(token, time_stamp)
+                else:
+                    postprocessed_result = postprocess_utils.sentence_postprocess(token)
             text_postprocessed = ""
             time_stamp_postprocessed = ""
             text_postprocessed_punc = postprocessed_result
@@ -787,7 +795,7 @@
         time_stamp_writer: bool = True,
         punc_infer_config: Optional[str] = None,
         punc_model_file: Optional[str] = None,
-        sv_model_file: Optional[str] = None,
+        sv_model_file: Optional[str] = None, 
         streaming: bool = False,
         embedding_node: str = "resnet1_dense",
         sv_threshold: float = 0.9465,
@@ -808,6 +816,8 @@
         level=log_level,
         format="%(asctime)s (%(module)s:%(lineno)d) %(levelname)s: %(message)s",
     )
+
+    sv_model_file = asr_model_file.replace("model.pb", "campplus_cn_common.bin")
 
     if param_dict is not None:
         hotword_list_or_file = param_dict.get('hotword')
@@ -1084,25 +1094,24 @@
             logging.info("decoding, utt: {}, predictions: {}".format(key, text_postprocessed_punc))
         torch.cuda.empty_cache()
         distribute_spk(asr_result_list[0]['sentences'], sv_output)
-        import pdb; pdb.set_trace()
         return asr_result_list
 
     return _forward
 
 
 def inference_paraformer_online(
-        maxlenratio: float,
-        minlenratio: float,
-        batch_size: int,
-        beam_size: int,
-        ngpu: int,
-        ctc_weight: float,
-        lm_weight: float,
-        penalty: float,
-        log_level: Union[int, str],
+        maxlenratio: float=0.0,
+        minlenratio: float=0.0,
+        batch_size: int=1,
+        beam_size: int=1,
+        ngpu: int=1,
+        ctc_weight: float=0.0,
+        lm_weight: float=0.0,
+        penalty: float=0.0,
+        log_level: Union[int, str]=logging.ERROR,
         # data_path_and_name_and_type,
-        asr_train_config: Optional[str],
-        asr_model_file: Optional[str],
+        asr_train_config: Optional[str]=None,
+        asr_model_file: Optional[str]=None,
         cmvn_file: Optional[str] = None,
         lm_train_config: Optional[str] = None,
         lm_file: Optional[str] = None,
@@ -2013,6 +2022,169 @@
 
     return _forward
 
+def inference_whisper(
+        maxlenratio: float,
+        minlenratio: float,
+        batch_size: int,
+        beam_size: int,
+        ngpu: int,
+        ctc_weight: float,
+        lm_weight: float,
+        penalty: float,
+        log_level: Union[int, str],
+        # data_path_and_name_and_type,
+        asr_train_config: Optional[str],
+        asr_model_file: Optional[str],
+        cmvn_file: Optional[str] = None,
+        lm_train_config: Optional[str] = None,
+        lm_file: Optional[str] = None,
+        token_type: Optional[str] = None,
+        key_file: Optional[str] = None,
+        word_lm_train_config: Optional[str] = None,
+        bpemodel: Optional[str] = None,
+        allow_variable_data_keys: bool = False,
+        streaming: bool = False,
+        output_dir: Optional[str] = None,
+        dtype: str = "float32",
+        seed: int = 0,
+        ngram_weight: float = 0.9,
+        nbest: int = 1,
+        num_workers: int = 1,
+        mc: bool = False,
+        param_dict: dict = None,
+        **kwargs,
+):
+
+    ncpu = kwargs.get("ncpu", 1)
+    torch.set_num_threads(ncpu)
+    if param_dict:
+        language = param_dict.get("language", None)
+        task = param_dict.get("task", "transcribe")
+    else:
+        language = None
+        task = "transcribe"
+    if batch_size > 1:
+        raise NotImplementedError("batch decoding is not implemented")
+    if word_lm_train_config is not None:
+        raise NotImplementedError("Word LM is not implemented")
+    if ngpu > 1:
+        raise NotImplementedError("only single GPU decoding is supported")
+
+    for handler in logging.root.handlers[:]:
+        logging.root.removeHandler(handler)
+
+    logging.basicConfig(
+        level=log_level,
+        format="%(asctime)s (%(module)s:%(lineno)d) %(levelname)s: %(message)s",
+    )
+
+    if ngpu >= 1 and torch.cuda.is_available():
+        device = "cuda"
+    else:
+        device = "cpu"
+
+    # 1. Set random-seed
+    set_all_random_seed(seed)
+
+    # 2. Build speech2text
+    speech2text_kwargs = dict(
+        asr_train_config=asr_train_config,
+        asr_model_file=asr_model_file,
+        cmvn_file=cmvn_file,
+        lm_train_config=lm_train_config,
+        lm_file=lm_file,
+        token_type=token_type,
+        bpemodel=bpemodel,
+        device=device,
+        maxlenratio=maxlenratio,
+        minlenratio=minlenratio,
+        dtype=dtype,
+        beam_size=beam_size,
+        ctc_weight=ctc_weight,
+        lm_weight=lm_weight,
+        ngram_weight=ngram_weight,
+        penalty=penalty,
+        nbest=nbest,
+        streaming=streaming,
+        language=language,
+        task=task,
+    )
+    logging.info("speech2text_kwargs: {}".format(speech2text_kwargs))
+    speech2text = Speech2TextWhisper(**speech2text_kwargs)
+
+    def _forward(data_path_and_name_and_type,
+                 raw_inputs: Union[np.ndarray, torch.Tensor] = None,
+                 output_dir_v2: Optional[str] = None,
+                 fs: dict = None,
+                 param_dict: dict = None,
+                 **kwargs,
+                 ):
+        # 3. Build data-iterator
+        if data_path_and_name_and_type is None and raw_inputs is not None:
+            if isinstance(raw_inputs, torch.Tensor):
+                raw_inputs = raw_inputs.numpy()
+            data_path_and_name_and_type = [raw_inputs, "speech", "waveform"]
+        loader = build_streaming_iterator(
+            task_name="asr",
+            preprocess_args=speech2text.asr_train_args,
+            data_path_and_name_and_type=data_path_and_name_and_type,
+            dtype=dtype,
+            fs=fs,
+            mc=mc,
+            batch_size=batch_size,
+            key_file=key_file,
+            num_workers=num_workers,
+        )
+
+        finish_count = 0
+        file_count = 1
+        # 7 .Start for-loop
+        # FIXME(kamo): The output format should be discussed about
+        asr_result_list = []
+        output_path = output_dir_v2 if output_dir_v2 is not None else output_dir
+        if output_path is not None:
+            writer = DatadirWriter(output_path)
+        else:
+            writer = None
+
+        for keys, batch in loader:
+            assert isinstance(batch, dict), type(batch)
+            assert all(isinstance(s, str) for s in keys), keys
+            _bs = len(next(iter(batch.values())))
+            assert len(keys) == _bs, f"{len(keys)} != {_bs}"
+            # batch = {k: v[0] for k, v in batch.items() if not k.endswith("_lengths")}
+
+            # N-best list of (text, token, token_int, hyp_object)
+            try:
+                results = speech2text(**batch)
+            except TooShortUttError as e:
+                logging.warning(f"Utterance {keys} {e}")
+                hyp = Hypothesis(score=0.0, scores={}, states={}, yseq=[])
+                results = [[" ", ["sil"], [2], hyp]] * nbest
+
+            # Only supporting batch_size==1
+            key = keys[0]
+
+            for n, (text, language) in zip(range(1, nbest + 1), results):
+                # Create a directory: outdir/{n}best_recog
+                if writer is not None:
+                    ibest_writer = writer[f"{n}best_recog"]
+
+                    # Write the result to each file
+                    ibest_writer["language"][key] = language
+
+                if text is not None:
+                    item = {'key': key, 'value': text}
+                    asr_result_list.append(item)
+                    finish_count += 1
+                    if writer is not None:
+                        ibest_writer["text"][key] = text
+
+                logging.info("uttid: {}".format(key))
+                logging.info("text predictions: {}\n".format(text))
+        return asr_result_list
+
+    return _forward
 
 def inference_launch(**kwargs):
     if 'mode' in kwargs:
@@ -2030,7 +2202,7 @@
         return inference_paraformer(**kwargs)
     elif mode == "paraformer_streaming":
         return inference_paraformer_online(**kwargs)
-    elif mode == "paraformer_vad_speaker":
+    elif mode.startswith("paraformer_vad_speaker"):
         return inference_paraformer_vad_speaker(**kwargs)
     elif mode.startswith("paraformer_vad"):
         return inference_paraformer_vad_punc(**kwargs)
@@ -2042,263 +2214,15 @@
         return inference_transducer(**kwargs)
     elif mode == "sa_asr":
         return inference_sa_asr(**kwargs)
+    elif mode == "whisper":
+        return inference_whisper(**kwargs)
     else:
         logging.info("Unknown decoding mode: {}".format(mode))
         return None
 
-
-def get_parser():
-    parser = config_argparse.ArgumentParser(
-        description="ASR Decoding",
-        formatter_class=argparse.ArgumentDefaultsHelpFormatter,
-    )
-
-    # Note(kamo): Use '_' instead of '-' as separator.
-    # '-' is confusing if written in yaml.
-    parser.add_argument(
-        "--log_level",
-        type=lambda x: x.upper(),
-        default="INFO",
-        choices=("CRITICAL", "ERROR", "WARNING", "INFO", "DEBUG", "NOTSET"),
-        help="The verbose level of logging",
-    )
-
-    parser.add_argument("--output_dir", type=str, required=True)
-    parser.add_argument(
-        "--ngpu",
-        type=int,
-        default=0,
-        help="The number of gpus. 0 indicates CPU mode",
-    )
-    parser.add_argument(
-        "--njob",
-        type=int,
-        default=1,
-        help="The number of jobs for each gpu",
-    )
-    parser.add_argument(
-        "--gpuid_list",
-        type=str,
-        default="",
-        help="The visible gpus",
-    )
-    parser.add_argument("--seed", type=int, default=0, help="Random seed")
-    parser.add_argument(
-        "--dtype",
-        default="float32",
-        choices=["float16", "float32", "float64"],
-        help="Data type",
-    )
-    parser.add_argument(
-        "--num_workers",
-        type=int,
-        default=1,
-        help="The number of workers used for DataLoader",
-    )
-
-    group = parser.add_argument_group("Input data related")
-    group.add_argument(
-        "--data_path_and_name_and_type",
-        type=str2triple_str,
-        required=True,
-        action="append",
-    )
-    group.add_argument("--key_file", type=str_or_none)
-    parser.add_argument(
-        "--hotword",
-        type=str_or_none,
-        default=None,
-        help="hotword file path or hotwords seperated by space"
-    )
-    group.add_argument("--allow_variable_data_keys", type=str2bool, default=False)
-    group.add_argument(
-        "--mc",
-        type=bool,
-        default=False,
-        help="MultiChannel input",
-    )
-
-    group = parser.add_argument_group("The model configuration related")
-    group.add_argument(
-        "--vad_infer_config",
-        type=str,
-        help="VAD infer configuration",
-    )
-    group.add_argument(
-        "--vad_model_file",
-        type=str,
-        help="VAD model parameter file",
-    )
-    group.add_argument(
-        "--punc_infer_config",
-        type=str,
-        help="PUNC infer configuration",
-    )
-    group.add_argument(
-        "--punc_model_file",
-        type=str,
-        help="PUNC model parameter file",
-    )
-    group.add_argument(
-        "--cmvn_file",
-        type=str,
-        help="Global CMVN file",
-    )
-    group.add_argument(
-        "--asr_train_config",
-        type=str,
-        help="ASR training configuration",
-    )
-    group.add_argument(
-        "--asr_model_file",
-        type=str,
-        help="ASR model parameter file",
-    )
-    group.add_argument(
-        "--sv_model_file",
-        type=str,
-        help="SV model parameter file",
-    )
-    group.add_argument(
-        "--lm_train_config",
-        type=str,
-        help="LM training configuration",
-    )
-    group.add_argument(
-        "--lm_file",
-        type=str,
-        help="LM parameter file",
-    )
-    group.add_argument(
-        "--word_lm_train_config",
-        type=str,
-        help="Word LM training configuration",
-    )
-    group.add_argument(
-        "--word_lm_file",
-        type=str,
-        help="Word LM parameter file",
-    )
-    group.add_argument(
-        "--ngram_file",
-        type=str,
-        help="N-gram parameter file",
-    )
-    group.add_argument(
-        "--model_tag",
-        type=str,
-        help="Pretrained model tag. If specify this option, *_train_config and "
-             "*_file will be overwritten",
-    )
-    group.add_argument(
-        "--beam_search_config",
-        default={},
-        help="The keyword arguments for transducer beam search.",
-    )
-
-    group = parser.add_argument_group("Beam-search related")
-    group.add_argument(
-        "--batch_size",
-        type=int,
-        default=1,
-        help="The batch size for inference",
-    )
-    group.add_argument("--nbest", type=int, default=5, help="Output N-best hypotheses")
-    group.add_argument("--beam_size", type=int, default=20, help="Beam size")
-    group.add_argument("--penalty", type=float, default=0.0, help="Insertion penalty")
-    group.add_argument(
-        "--maxlenratio",
-        type=float,
-        default=0.0,
-        help="Input length ratio to obtain max output length. "
-             "If maxlenratio=0.0 (default), it uses a end-detect "
-             "function "
-             "to automatically find maximum hypothesis lengths."
-             "If maxlenratio<0.0, its absolute value is interpreted"
-             "as a constant max output length",
-    )
-    group.add_argument(
-        "--minlenratio",
-        type=float,
-        default=0.0,
-        help="Input length ratio to obtain min output length",
-    )
-    group.add_argument(
-        "--ctc_weight",
-        type=float,
-        default=0.0,
-        help="CTC weight in joint decoding",
-    )
-    group.add_argument("--lm_weight", type=float, default=1.0, help="RNNLM weight")
-    group.add_argument("--ngram_weight", type=float, default=0.9, help="ngram weight")
-    group.add_argument("--streaming", type=str2bool, default=False)
-    group.add_argument("--fake_streaming", type=str2bool, default=False)
-    group.add_argument("--full_utt", type=str2bool, default=False)
-    group.add_argument("--chunk_size", type=int, default=16)
-    group.add_argument("--left_context", type=int, default=16)
-    group.add_argument("--right_context", type=int, default=0)
-    group.add_argument(
-        "--display_partial_hypotheses",
-        type=bool,
-        default=False,
-        help="Whether to display partial hypotheses during chunk-by-chunk inference.",
-    )
-
-    group = parser.add_argument_group("Dynamic quantization related")
-    group.add_argument(
-        "--quantize_asr_model",
-        type=bool,
-        default=False,
-        help="Apply dynamic quantization to ASR model.",
-    )
-    group.add_argument(
-        "--quantize_modules",
-        nargs="*",
-        default=None,
-        help="""Module names to apply dynamic quantization on.
-        The module names are provided as a list, where each name is separated
-        by a comma (e.g.: --quantize-config=[Linear,LSTM,GRU]).
-        Each specified name should be an attribute of 'torch.nn', e.g.:
-        torch.nn.Linear, torch.nn.LSTM, torch.nn.GRU, ...""",
-    )
-    group.add_argument(
-        "--quantize_dtype",
-        type=str,
-        default="qint8",
-        choices=["float16", "qint8"],
-        help="Dtype for dynamic quantization.",
-    )
-
-    group = parser.add_argument_group("Text converter related")
-    group.add_argument(
-        "--token_type",
-        type=str_or_none,
-        default=None,
-        choices=["char", "bpe", None],
-        help="The token type for ASR model. "
-             "If not given, refers from the training args",
-    )
-    group.add_argument(
-        "--bpemodel",
-        type=str_or_none,
-        default=None,
-        help="The model path of sentencepiece. "
-             "If not given, refers from the training args",
-    )
-    group.add_argument("--token_num_relax", type=int, default=1, help="")
-    group.add_argument("--decoding_ind", type=int, default=0, help="")
-    group.add_argument("--decoding_mode", type=str, default="model1", help="")
-    group.add_argument(
-        "--ctc_weight2",
-        type=float,
-        default=0.0,
-        help="CTC weight in joint decoding",
-    )
-    return parser
-
-
 def main(cmd=None):
     print(get_commandline_args(), file=sys.stderr)
+    from funasr.bin.argument import get_parser
     parser = get_parser()
     parser.add_argument(
         "--mode",

--
Gitblit v1.9.1