20200621--learning-to-fool-the-speaker-recognition-master 实验记录

learning-to-fool-the-speaker-recognition-master 实验记录
timit数据集是之前就prepare好的

出错1：
RuntimeError: Detected that PyTorch and torchvision were compiled with different CUDA versions. PyTorch has CUDA Version=10.2 and torchvision has CUDA Version=10.1. Please reinstall the torchvision that matches your PyTorch install.
解决办法：pip install torchvision0.6.1
原因：PyTorch和torchvision依赖的是不同版本的cuda，不用动cuda。之前的torchvision版本是0.6.0.dev20200319，torch版本是1.5.0，现在安装torchvision0.6.1就好了
一开始写的命令是：pip install torchvision10.13
但是有提示，这个提示告诉了你最新版本的torchvision是0.6.1
ERROR: Could not find a version that satisfies the requirement torchvision10.13 (from versions: 0.1.6, 0.1.7, 0.1.8, 0.1.9, 0.2.0, 0.2.1, 0.2.2, 0.2.2.post2, 0.2.2.post3, 0.3.0, 0.4.0, 0.4.1, 0.4.2, 0.5.0, 0.6.0, 0.6.1)
ERROR: No matching distribution found for torchvision==10.13

出错2：
ModuleNotFoundError: No module named ‘numba.decorators’
解决办法：https://blog.csdn.net/July_Wander/article/details/106857289
卸载掉numba先，然后安装 numba-0.48.0即可
pip uninstall numba
pip install numba==0.48.0

出错3：
AssertionError:
The NVIDIA driver on your system is too old (found version 10010).
Please update your GPU driver by downloading and installing a new
version from the URL: http://www.nvidia.com/Download/index.aspx
Alternatively, go to: https://pytorch.org to install
a PyTorch version that has been compiled with your version
of the CUDA driver.
解决办法：降级torch，
pip install torch1.4.0
原来是1.5.0
这个问题是没有了，但又出现了新问题
RuntimeError: Failed to register operator torchvision::_new_empty_tensor_op. The symbolic name must match the format Domain::Name, and sould start with a letter and contain only alphanumerical characters
网上说这个问题的解决办法是：更新torch到最新版即可
晕，还是不要动NVIDIA driver吧
还是应该降pytorch的版本
经过一番折腾之后，把下面4个都降级，就可以了
torch1.2
torchvision0.4
torchaudio0.3
pip3 install ‘pillow<7.0.0’

出错4：
FileNotFoundError: [Errno 2] No such file or directory: ‘/workspace/TIMIT/processed/TIMIT_labels.npy’
这是少了一个文件，在github上找到了这个文件，解决

出错5：
Traceback (most recent call last):
File “train_transformer.py”, line 669, in
main(args)
File “train_transformer.py”, line 593, in main
train_dataset = TIMIT_speaker(args.data_root, train=True, phoneme=True, norm_factor=True)
File “/workspace/learning-to-fool-the-speaker-recognition-master/dataset.py”, line 66, in init
self.timit_labels = np.load(os.path.join(data_root_processed, “TIMIT_labels.npy”)).item()
File “/usr/local/lib/python3.6/dist-packages/numpy/lib/npyio.py”, line 453, in load
pickle_kwargs=pickle_kwargs)
File “/usr/local/lib/python3.6/dist-packages/numpy/lib/format.py”, line 739, in read_array
raise ValueError("Object arrays cannot be loaded when "
ValueError: Object arrays cannot be loaded when allow_pickle=False
解决办法：
在dataset.py中修改了一句话，加上allow_pickle=True，变成下面这样，就可以了
self.timit_labels = np.load(os.path.join(data_root_processed, “TIMIT_labels.npy”),allow_pickle=True).item()

至此，下面命令终于可以跑起来了

python train_transformer.py --output_dir /workspace/learning-to-fool-the-speaker-recognition-master/output --speaker_factor 1 --speech_factor 5 --norm_factor 1000 --speech_kld_factor 1 --data_root /workspace/TIMIT --no_dist --speaker_model /workspace/learning-to-fool-the-speaker-recognition-master/pretrained-speaker-recognition-model/model_raw.pkl --speech_model /workspace/learning-to-fool-the-speaker-recognition-master/pretrained-phoneme-recognition-model/epoch_23.pth --speaker_cfg /workspace/learning-to-fool-the-speaker-recognition-master/config/timit_speaker_transformer.cfg --speech_cfg /workspace/learning-to-fool-the-speaker-recognition-master/config/timit_speech.cfg --norm_clip 0.01 --no_dist