From 14a1b5eb20c951b1fe23ca7ea389778a6899332a Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期六, 25 二月 2023 17:45:41 +0800
Subject: [PATCH] onnx
---
funasr/bin/vad_inference.py | 49 ++++++++++++++++++++++++++++++++-----------------
1 files changed, 32 insertions(+), 17 deletions(-)
diff --git a/funasr/bin/vad_inference.py b/funasr/bin/vad_inference.py
index 9f1d0f3..607f131 100644
--- a/funasr/bin/vad_inference.py
+++ b/funasr/bin/vad_inference.py
@@ -1,6 +1,7 @@
import argparse
import logging
import sys
+import json
from pathlib import Path
from typing import Any
from typing import List
@@ -80,6 +81,7 @@
self.device = device
self.dtype = dtype
self.frontend = frontend
+ self.batch_size = batch_size
@torch.no_grad()
def __call__(
@@ -105,17 +107,29 @@
feats_len = feats_len.int()
else:
raise Exception("Need to extract feats first, please configure frontend configuration")
- batch = {"feats": feats, "feats_lengths": feats_len, "waveform": speech}
- # a. To device
- batch = to_device(batch, device=self.device)
-
- # b. Forward Encoder
- segments = self.vad_model(**batch)
-
+ # b. Forward Encoder streaming
+ t_offset = 0
+ step = min(feats_len, 6000)
+ segments = [[]] * self.batch_size
+ for t_offset in range(0, feats_len, min(step, feats_len - t_offset)):
+ if t_offset + step >= feats_len - 1:
+ step = feats_len - t_offset
+ is_final_send = True
+ else:
+ is_final_send = False
+ batch = {
+ "feats": feats[:, t_offset:t_offset + step, :],
+ "waveform": speech[:, t_offset * 160:min(speech.shape[-1], (t_offset + step - 1) * 160 + 400)],
+ "is_final_send": is_final_send
+ }
+ # a. To device
+ batch = to_device(batch, device=self.device)
+ segments_part = self.vad_model(**batch)
+ if segments_part:
+ for batch_num in range(0, self.batch_size):
+ segments[batch_num] += segments_part[batch_num]
return segments
-
-
def inference(
@@ -152,11 +166,12 @@
)
return inference_pipeline(data_path_and_name_and_type, raw_inputs)
+
def inference_modelscope(
batch_size: int,
ngpu: int,
log_level: Union[int, str],
- #data_path_and_name_and_type,
+ # data_path_and_name_and_type,
vad_infer_config: Optional[str],
vad_model_file: Optional[str],
vad_cmvn_file: Optional[str] = None,
@@ -167,7 +182,6 @@
dtype: str = "float32",
seed: int = 0,
num_workers: int = 1,
- param_dict: dict = None,
**kwargs,
):
assert check_argument_types()
@@ -201,11 +215,11 @@
speech2vadsegment = Speech2VadSegment(**speech2vadsegment_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,
+ 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,
):
# 3. Build data-iterator
loader = VADTask.build_streaming_iterator(
@@ -238,14 +252,15 @@
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")}
# do vad segment
results = speech2vadsegment(**batch)
for i, _ in enumerate(keys):
+ results[i] = json.dumps(results[i])
item = {'key': keys[i], 'value': results[i]}
vad_results.append(item)
if writer is not None:
+ results[i] = json.loads(results[i])
ibest_writer["text"][keys[i]] = "{}".format(results[i])
return vad_results
--
Gitblit v1.9.1