From 97a689d65da434345a641a909f13b78e5690c86b Mon Sep 17 00:00:00 2001
From: zhifu gao <zhifu.gzf@alibaba-inc.com>
Date: 星期四, 18 五月 2023 19:35:08 +0800
Subject: [PATCH] Merge pull request #526 from alibaba-damo-academy/dev_infer
---
funasr/bin/sv_inference_launch.py | 180 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++---
1 files changed, 169 insertions(+), 11 deletions(-)
diff --git a/funasr/bin/sv_inference_launch.py b/funasr/bin/sv_inference_launch.py
index 8806070..dbddd9f 100755
--- a/funasr/bin/sv_inference_launch.py
+++ b/funasr/bin/sv_inference_launch.py
@@ -1,7 +1,7 @@
+# -*- encoding: utf-8 -*-
#!/usr/bin/env python3
# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
# MIT License (https://opensource.org/licenses/MIT)
-
import argparse
import logging
@@ -14,7 +14,173 @@
from funasr.utils.types import str2bool
from funasr.utils.types import str2triple_str
from funasr.utils.types import str_or_none
+import argparse
+import logging
+import os
+import sys
+from pathlib import Path
+from typing import Any
+from typing import List
+from typing import Optional
+from typing import Sequence
+from typing import Tuple
+from typing import Union
+import numpy as np
+import torch
+from kaldiio import WriteHelper
+from typeguard import check_argument_types
+from typeguard import check_return_type
+
+from funasr.utils.cli_utils import get_commandline_args
+from funasr.tasks.sv import SVTask
+from funasr.torch_utils.device_funcs import to_device
+from funasr.torch_utils.set_all_random_seed import set_all_random_seed
+from funasr.utils import config_argparse
+from funasr.utils.types import str2bool
+from funasr.utils.types import str2triple_str
+from funasr.utils.types import str_or_none
+from funasr.utils.misc import statistic_model_parameters
+from funasr.bin.sv_infer import Speech2Xvector
+
+def inference_sv(
+ output_dir: Optional[str] = None,
+ batch_size: int = 1,
+ dtype: str = "float32",
+ ngpu: int = 1,
+ seed: int = 0,
+ num_workers: int = 0,
+ log_level: Union[int, str] = "INFO",
+ key_file: Optional[str] = None,
+ sv_train_config: Optional[str] = "sv.yaml",
+ sv_model_file: Optional[str] = "sv.pb",
+ model_tag: Optional[str] = None,
+ allow_variable_data_keys: bool = True,
+ streaming: bool = False,
+ embedding_node: str = "resnet1_dense",
+ sv_threshold: float = 0.9465,
+ param_dict: Optional[dict] = None,
+ **kwargs,
+):
+ assert check_argument_types()
+ ncpu = kwargs.get("ncpu", 1)
+ torch.set_num_threads(ncpu)
+
+ if batch_size > 1:
+ raise NotImplementedError("batch decoding is not implemented")
+ if ngpu > 1:
+ raise NotImplementedError("only single GPU decoding is supported")
+
+ logging.basicConfig(
+ level=log_level,
+ format="%(asctime)s (%(module)s:%(lineno)d) %(levelname)s: %(message)s",
+ )
+ logging.info("param_dict: {}".format(param_dict))
+
+ if ngpu >= 1 and torch.cuda.is_available():
+ device = "cuda"
+ else:
+ device = "cpu"
+
+ # 1. Set random-seed
+ set_all_random_seed(seed)
+
+ # 2. Build speech2xvector
+ speech2xvector_kwargs = dict(
+ sv_train_config=sv_train_config,
+ sv_model_file=sv_model_file,
+ device=device,
+ dtype=dtype,
+ streaming=streaming,
+ embedding_node=embedding_node
+ )
+ logging.info("speech2xvector_kwargs: {}".format(speech2xvector_kwargs))
+ speech2xvector = Speech2Xvector.from_pretrained(
+ model_tag=model_tag,
+ **speech2xvector_kwargs,
+ )
+ speech2xvector.sv_model.eval()
+
+ def _forward(
+ data_path_and_name_and_type: Sequence[Tuple[str, str, str]] = None,
+ raw_inputs: Union[np.ndarray, torch.Tensor] = None,
+ output_dir_v2: Optional[str] = None,
+ param_dict: Optional[dict] = None,
+ ):
+ logging.info("param_dict: {}".format(param_dict))
+ 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"]
+
+ # 3. Build data-iterator
+ loader = SVTask.build_streaming_iterator(
+ data_path_and_name_and_type,
+ dtype=dtype,
+ batch_size=batch_size,
+ key_file=key_file,
+ num_workers=num_workers,
+ preprocess_fn=None,
+ collate_fn=None,
+ allow_variable_data_keys=allow_variable_data_keys,
+ inference=True,
+ )
+
+ # 7 .Start for-loop
+ output_path = output_dir_v2 if output_dir_v2 is not None else output_dir
+ embd_writer, ref_embd_writer, score_writer = None, None, None
+ if output_path is not None:
+ os.makedirs(output_path, exist_ok=True)
+ embd_writer = WriteHelper("ark,scp:{}/xvector.ark,{}/xvector.scp".format(output_path, output_path))
+ sv_result_list = []
+ 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")}
+
+ embedding, ref_embedding, score = speech2xvector(**batch)
+ # Only supporting batch_size==1
+ key = keys[0]
+ normalized_score = 0.0
+ if score is not None:
+ score = score.item()
+ normalized_score = max(score - sv_threshold, 0.0) / (1.0 - sv_threshold) * 100.0
+ item = {"key": key, "value": normalized_score}
+ else:
+ item = {"key": key, "value": embedding.squeeze(0).cpu().numpy()}
+ sv_result_list.append(item)
+ if output_path is not None:
+ embd_writer(key, embedding[0].cpu().numpy())
+ if ref_embedding is not None:
+ if ref_embd_writer is None:
+ ref_embd_writer = WriteHelper(
+ "ark,scp:{}/ref_xvector.ark,{}/ref_xvector.scp".format(output_path, output_path)
+ )
+ score_writer = open(os.path.join(output_path, "score.txt"), "w")
+ ref_embd_writer(key, ref_embedding[0].cpu().numpy())
+ score_writer.write("{} {:.6f}\n".format(key, normalized_score))
+
+ if output_path is not None:
+ embd_writer.close()
+ if ref_embd_writer is not None:
+ ref_embd_writer.close()
+ score_writer.close()
+
+ return sv_result_list
+
+ return _forward
+
+
+
+
+def inference_launch(mode, **kwargs):
+ if mode == "sv":
+ return inference_sv(**kwargs)
+ else:
+ logging.info("Unknown decoding mode: {}".format(mode))
+ return None
def get_parser():
parser = config_argparse.ArgumentParser(
@@ -131,15 +297,6 @@
return parser
-def inference_launch(mode, **kwargs):
- if mode == "sv":
- from funasr.bin.sv_inference import inference_modelscope
- return inference_modelscope(**kwargs)
- else:
- logging.info("Unknown decoding mode: {}".format(mode))
- return None
-
-
def main(cmd=None):
print(get_commandline_args(), file=sys.stderr)
parser = get_parser()
@@ -167,7 +324,8 @@
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
os.environ["CUDA_VISIBLE_DEVICES"] = gpuid
- inference_launch(**kwargs)
+ inference_pipeline = inference_launch(**kwargs)
+ return inference_pipeline(kwargs["data_path_and_name_and_type"])
if __name__ == "__main__":
--
Gitblit v1.9.1