替换模型训练遇到的问题

就是接着这个大佬的模型继续训练，因为我自己的模型太过于小了，无法收敛。但是一更换就报错
截图在下面 然后我再复制粘贴一下防止我没传上来图片

E:\数据集制作\MockingBird-main\synthesizer\synthesizer_dataset.py:84: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ..\torch\csrc\utils\tensor_new.cpp:201.)
embeds = torch.tensor(embeds)
C:\Users\11351\AppData\Local\Programs\Python\Python39\lib\site-packages\torch\nn\functional.py:1795: UserWarning: nn.functional.tanh is deprecated. Use torch.tanh instead.
warnings.warn("nn.functional.tanh is deprecated. Use torch.tanh instead.")
Traceback (most recent call last):
File "E:\数据集制作\MockingBird-main\synthesizer_train.py", line 37, in
train(**vars(args))
File "E:\数据集制作\MockingBird-main\synthesizer\train.py", line 208, in train
optimizer.step()
File "C:\Users\11351\AppData\Local\Programs\Python\Python39\lib\site-packages\torch\optim\optimizer.py", line 88, in wrapper
return func(*args, **kwargs)
File "C:\Users\11351\AppData\Local\Programs\Python\Python39\lib\site-packages\torch\autograd\grad_mode.py", line 28, in decorate_context
return func(*args, **kwargs)
File "C:\Users\11351\AppData\Local\Programs\Python\Python39\lib\site-packages\torch\optim\adam.py", line 133, in step
F.adam(params_with_grad,
File "C:\Users\11351\AppData\Local\Programs\Python\Python39\lib\site-packages\torch\optim\_functional.py", line 86, in adam
exp_avg.mul_(beta1).add_(grad, alpha=1 - beta1)
RuntimeError: The size of tensor a (1024) must match the size of tensor b (3) at non-singleton dimension 3

Start with synthesizer_train.py and follow train() into synthesizer/train.py, especially the optimizer.step() failure at line 208. Inspect synthesizer/synthesizer_dataset.py at line 84 and compare the replacement model, checkpoint, and batch tensor dimensions involved in training. Done means reproducing the error, identifying the incompatible dimensions, and confirming that training proceeds with the replacement model.

Written by the indexing model from the issue text.