训练时出现这个错误

(mockingbird) K:\mb>python synthesizer_train.py 75k k:/mb/datame/SV2TTS/synthesizer -b 1000
Arguments:
run_id: 75k
syn_dir: k:/mb/datame/SV2TTS/synthesizer
models_dir: synthesizer/saved_models/
save_every: 1000
backup_every: 1000
log_every: 200
force_restart: False
hparams:

Checkpoint path: synthesizer\saved_models\75k\75k.pt
Loading training data from: k:\mb\datame\SV2TTS\synthesizer\train.txt
Using model: Tacotron
Using device: cpu

Initialising Tacotron Model...

Trainable Parameters: 32.869M

Loading weights at synthesizer\saved_models\75k\75k.pt
Tacotron weights loaded from step 93000
Using inputs from:
k:\mb\datame\SV2TTS\synthesizer\train.txt
k:\mb\datame\SV2TTS\synthesizer\mels
k:\mb\datame\SV2TTS\synthesizer\embeds
Found 12 samples
+----------------+------------+---------------+------------------+
| Steps with r=2 | Batch Size | Learning Rate | Outputs/Step (r) |
+----------------+------------+---------------+------------------+
| 67k Steps | 12 | 5e-06 | 2 |
+----------------+------------+---------------+------------------+

K:\mb\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)
f:\anaconda3\envs\mockingbird\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 "synthesizer_train.py", line 37, in
train(**vars(args))
File "K:\mb\synthesizer\train.py", line 208, in train
optimizer.step()
File "f:\anaconda3\envs\mockingbird\lib\site-packages\torch\optim\optimizer.py", line 88, in wrapper
return func(*args, **kwargs)
File "f:\anaconda3\envs\mockingbird\lib\site-packages\torch\autograd\grad_mode.py", line 28, in decorate_context
return func(*args, **kwargs)
File "f:\anaconda3\envs\mockingbird\lib\site-packages\torch\optim\adam.py", line 133, in step
F.adam(params_with_grad,
File "f:\anaconda3\envs\mockingbird\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

这个模型用的是主页提供的75k steps模型

Start with synthesizer_train.py and follow the training flow into synthesizer/train.py, especially the optimizer.step() call at line 208. Compare the 75k checkpoint loading path and the dataset tensor construction in synthesizer_dataset.py with the reported Adam tensor-shape mismatch. Done means identifying the incompatible tensors or setup and confirming that training can proceed with the supplied 75k model.

Written by the indexing model from the issue text.