wangtianrui / dccrn Goto Github PK

Can you tell where I can find the dataset, I find the dataset from "https://github.com/microsoft/DNS-Challenge/tree/master/datasets", but the file of dataset is too big. Are you simplifying the dataset? Thank you very much if you canm answer

请问数据集在哪里下载呀，需要怎么设置？

Traceback (most recent call last):
File "train.py", line 19, in
my_system = MyBaseSystem("./conf.yml")
File "/root/DCCRN-master/BaseModel.py", line 38, in init
data_home=conf["preset"]["data_home"]
File "/root/DCCRN-master/BaseModel.py", line 106, in get_dns_data_loader
val_set = DNSDataset(val_json_file, data_home=data_home)
File "/root/DCCRN-master/dns_loader.py", line 21, in init
self.mix_infos = json.load(f)
File "/root/anaconda3/envs/DCCRN/lib/python3.7/json/init.py", line 296, in load
parse_constant=parse_constant, object_pairs_hook=object_pairs_hook, **kw)
File "/root/anaconda3/envs/DCCRN/lib/python3.7/json/init.py", line 348, in loads
return _default_decoder.decode(s)
File "/root/anaconda3/envs/DCCRN/lib/python3.7/json/decoder.py", line 337, in decode
obj, end = self.raw_decode(s, idx=_w(s, 0).end())
File "/root/anaconda3/envs/DCCRN/lib/python3.7/json/decoder.py", line 355, in raw_decode
raise JSONDecodeError("Expecting value", s, err.value) from None
json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)

"{"mix": "F:\Traindata\DNS-Challenge\make_data\noisy\book_09395_chp_0023_reader_07848_7_door_Freesound_validated_112347_5-door_Freesound_validated_458454_2-P2-3vmp1KWA_snr20_fileid_3154.wav", "clean": "F:\Traindata\DNS-Challenge\make_data\clean\clean_fileid_3154.wav", "noise": "F:\Traindata\DNS-Challenge\make_data\noise\noise_fileid_3154.wav", "snr": 20, "file_len": 480000"
你好请问下大概需要多少的数据可以训练一个通用的去混响模型，我准备训练24k的模型

您好：

当我在模型中使用ComplexBatchNorm时，训练一段时间内存就会爆炸。但是我把ComplexBatchNorm换成torch自带的BatchNorm就不会出现该问题。请问您在训练DCCRN的时候没有出现这个情况吗？

谢谢；
祝好！

Hello, Thank you for sharing your code. Can you please the inference script as well.

can you give me the loss log? thx

below is my loss

Epoch 0: 100%|██████████| 21209/21209 [15:24:41<00:00, 2.62s/it, loss=13.395, v_num=0]ts/DCCRN# tail -f nohup.out
Epoch 00000: val_loss reached 13.25967 (best 13.25967), saving model to /data/juicefs_speech_tts_v3/11139034/work/ptts/DCCRN/logs/_ckpt_epoch_0.ckpt as top 5
Epoch 1: 29%|██▉ | 6162/21209 [4:45:09<11:36:18, 2.78s/it, loss=14.446, v_num=0, val_loss=13.3]

is it right?

你好，我安装了torch==1.6.0和asteroid==0.3.4，我的cuda版本为10.2，但是在运行train.py文件的时候报错：AttributeError: can't set attribute，具体报错如下：
Traceback (most recent call last):
File "/media/li/TOSHIBA EXT/WDCCRN /DCCRN-master/train.py", line 21, in
my_system.init_system_and_trainer(SystemClass=MySystem, model=model_, gpus=[0])
File "/media/li/TOSHIBA EXT/WDCCRN /DCCRN-master/BaseModel.py", line 85, in init_system_and_trainer
config=self.conf
File "/home/li/anaconda3/envs/W-DCCRN/lib/python3.6/site-packages/asteroid/engine/system.py", line 58, in init
self.hparams = Namespace(**self.config_to_hparams(config))
File "/home/li/anaconda3/envs/W-DCCRN/lib/python3.6/site-packages/torch/nn/modules/module.py", line 1178, in setattr
object.setattr(self, name, value)
AttributeError: can't set attribute
请问我该如何解决这个问题呢？

hello， 我有个疑问，训练模型的时候是直接输入原始30s的音频吗，还是要提前切片5s或者更短的音频作为模型输入？谢谢解答

哈喽，请问有生成json文件的代码吗？

maybe the batch size 2 is too small, but I cannot use two RTX 2080Ti to train this model with batch size 4

Hello, I tried to download the data set using the official method, and I also tried to download the data set through "wget", but the data was not available, code show as below:
for root,dirs,files in os.walk(args.datasetpath):
wav_files = []
path = "https://github.com/microsoft/DNS-Challenge/tree/master/datasets"
for line in files:
if line.endswith(".wav"):
wav_files.append(os.path.join(root,line))
else:
pass
for file in wav_files:
URL = path+file[37:]
cmd = "wget -p " +file+" "+URL
os.system(cmd)

https://github.com/maggie0830/DCCRN，这个代码仓库是你这边的第一版代码？复现时发现内存一直增加

请问您能否将您数据集中的的部分clean语音和对应noisy语音发送到我的邮箱，我将不胜感激！
e-mail：[email protected]

您好，我跟着 https://github.com/ShenYi666666/DNS-Challenge 把数据下载完了。也生成了一批数据（也是这个git的代码）。但是我没有找到 json文件。请问是为啥呢。

如果必须下载2022年的数据（太大了）；请问数据集或者说json文件的规则是怎样的；我自己撸代码搞一批。

找到的资料比较少，看不出个所以然；望不吝赐教。

作者你好，我看到你最终给出了3中还原方法，目前这3种你默认选择了E进行数据还原，那么C以及R效果是不如E的效果吗？有类似的实验结果嘛？

This line should be times rather than plus?

Hello,
Could you please share how many epochs did you train the model for.

I see that the training loss plateaus around -21 and then remains around there for many epochs.

Also, how do i interpret the loss value? and how do you know when the training is done.