From 94de39dde2e616a01683c518023d0fab72b4e103 Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期一, 19 二月 2024 22:21:50 +0800
Subject: [PATCH] aishell example
---
funasr/utils/prepare_data.py | 57 +++++++++++++++++++++++++++++++++++++--------------------
1 files changed, 37 insertions(+), 20 deletions(-)
diff --git a/funasr/utils/prepare_data.py b/funasr/utils/prepare_data.py
index f61e501..36eebdc 100644
--- a/funasr/utils/prepare_data.py
+++ b/funasr/utils/prepare_data.py
@@ -5,6 +5,7 @@
import kaldiio
import numpy as np
+import librosa
import torch.distributed as dist
import torchaudio
@@ -42,7 +43,11 @@
def wav2num_frame(wav_path, frontend_conf):
- waveform, sampling_rate = torchaudio.load(wav_path)
+ try:
+ waveform, sampling_rate = torchaudio.load(wav_path)
+ except:
+ waveform, sampling_rate = librosa.load(wav_path)
+ waveform = np.expand_dims(waveform, axis=0)
n_frames = (waveform.shape[1] * 1000.0) / (sampling_rate * frontend_conf["frame_shift"] * frontend_conf["lfr_n"])
feature_dim = frontend_conf["n_mels"] * frontend_conf["lfr_m"]
return n_frames, feature_dim
@@ -82,6 +87,7 @@
sample_name, feature_path = line.strip().split()
feature = kaldiio.load_mat(feature_path)
n_frames, feature_dim = feature.shape
+ write_flag = True
if n_frames > 0 and length_min > 0:
write_flag = n_frames >= length_min
if n_frames > 0 and length_max > 0:
@@ -185,18 +191,46 @@
for i in range(nj):
path = ""
for file_name in file_names:
- path = path + os.path.join(split_path, str(i + 1), file_name)
+ path = path + " " + os.path.join(split_path, str(i + 1), file_name)
f_data.write(path + "\n")
def prepare_data(args, distributed_option):
+ data_names = args.dataset_conf.get("data_names", "speech,text").split(",")
+ data_types = args.dataset_conf.get("data_types", "sound,text").split(",")
+ file_names = args.data_file_names.split(",")
+ batch_type = args.dataset_conf["batch_conf"]["batch_type"]
+ print("data_names: {}, data_types: {}, file_names: {}".format(data_names, data_types, file_names))
+ assert len(data_names) == len(data_types) == len(file_names)
+ if args.dataset_type == "small":
+ args.train_shape_file = [os.path.join(args.data_dir, args.train_set, "{}_shape".format(data_names[0]))]
+ args.valid_shape_file = [os.path.join(args.data_dir, args.valid_set, "{}_shape".format(data_names[0]))]
+ args.train_data_path_and_name_and_type, args.valid_data_path_and_name_and_type = [], []
+ for file_name, data_name, data_type in zip(file_names, data_names, data_types):
+ args.train_data_path_and_name_and_type.append(
+ ["{}/{}/{}".format(args.data_dir, args.train_set, file_name), data_name, data_type])
+ args.valid_data_path_and_name_and_type.append(
+ ["{}/{}/{}".format(args.data_dir, args.valid_set, file_name), data_name, data_type])
+ if os.path.exists(args.train_shape_file[0]):
+ assert os.path.exists(args.valid_shape_file[0])
+ print('shape file for small dataset already exists.')
+ return
+ else:
+ concat_data_name = "_".join(data_names)
+ args.train_data_file = os.path.join(args.data_dir, args.train_set, "{}_data.list".format(concat_data_name))
+ args.valid_data_file = os.path.join(args.data_dir, args.valid_set, "{}_data.list".format(concat_data_name))
+ if os.path.exists(args.train_data_file):
+ assert os.path.exists(args.valid_data_file)
+ print('data list for large dataset already exists.')
+ return
+
distributed = distributed_option.distributed
if not distributed or distributed_option.dist_rank == 0:
if hasattr(args, "filter_input") and args.filter_input:
filter_wav_text(args.data_dir, args.train_set)
filter_wav_text(args.data_dir, args.valid_set)
- if args.dataset_type == "small":
+ if args.dataset_type == "small" and batch_type != "unsorted":
calc_shape(args, args.train_set)
calc_shape(args, args.valid_set)
@@ -204,22 +238,5 @@
generate_data_list(args, args.data_dir, args.train_set)
generate_data_list(args, args.data_dir, args.valid_set)
- data_names = args.dataset_conf.get("data_names", "speech,text").split(",")
- data_types = args.dataset_conf.get("data_types", "sound,text").split(",")
- file_names = args.data_file_names.split(",")
- assert len(data_names) == len(data_types) == len(file_names)
- if args.dataset_type == "small":
- args.train_shape_file = [os.path.join(args.data_dir, args.train_set, "{}_shape".format(data_names[0]))]
- args.valid_shape_file = [os.path.join(args.data_dir, args.valid_set, "{}}_shape".format(data_names[0]))]
- args.train_data_path_and_name_and_type, args.valid_data_path_and_name_and_type = [], []
- for file_name, data_name, data_type in zip(file_names, data_names, data_types):
- args.train_data_path_and_name_and_type.append(
- ["{}/{}/{}".format(args.data_dir, args.train_set, file_name), data_name, data_type])
- args.valid_data_path_and_name_and_type.append(
- ["{}/{}/{}".format(args.data_dir, args.valid_set, file_name), data_name, data_type])
- else:
- concat_data_name = "_".join(data_names)
- args.train_data_file = os.path.join(args.data_dir, args.train_set, "{}_data.list".format(concat_data_name))
- args.valid_data_file = os.path.join(args.data_dir, args.valid_set, "{}_data.list".format(concat_data_name))
if distributed:
dist.barrier()
--
Gitblit v1.9.1