数据集过大时训练报错：RuntimeError: The expanded size of the tensor (32) must match the existing size (0) at non-singleton dimension 1. Target sizes: [192, 32]. Tensor sizes: [192, 0]

[Fantendo2001]

RuntimeError: The expanded size of the tensor (32) must match the existing size (0) at non-singleton dimension 1. Target sizes: [192, 32]. Tensor sizes: [192, 0]

当数据集（音频数量）较大时（如500），训练时会出现该报错。当数据集较小时（如200），可以正常训练。

config以lex_base为基础改的，语音数据和标注是自己的。

{
"train": {
"log_interval": 200,
"eval_interval": 1000,
"seed": 1234,
"epochs": 2000,
"learning_rate": 2e-4,
"betas": [0.8, 0.99],
"eps": 1e-9,
"batch_size": 16,
"fp16_run": true,
"lr_decay": 0.999875,
"segment_size": 8192,
"init_lr_ratio": 1,
"warmup_epochs": 0,
"c_mel": 45,
"c_kl": 1.0
},
"data": {
"training_files":"filelists/66_audio_text_train_filelist.txt.cleaned",
"validation_files":"filelists/66_audio_text_train_filelist.txt.cleaned",
"language": "cmn",
"text_cleaners":["chinese_cleaners"],
"max_wav_value": 32768.0,
"sampling_rate": 22050,
"filter_length": 1024,
"hop_length": 256,
"win_length": 1024,
"n_mel_channels": 80,
"mel_fmin": 0.0,
"mel_fmax": null,
"add_blank": true,
"n_speakers": 0,
"cleaned_text": true
},
"model": {
"inter_channels": 192,
"hidden_channels": 192,
"filter_channels": 768,
"n_heads": 2,
"n_layers": 6,
"kernel_size": 3,
"p_dropout": 0.1,
"resblock": "1",
"resblock_kernel_sizes": [3,7,11],
"resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
"upsample_rates": [8,8,2,2],
"upsample_initial_channel": 512,
"upsample_kernel_sizes": [16,16,4,4],
"n_layers_q": 3,
"use_spectral_norm": false
},
"data_loader": {
"use_train_sampler": true,
"num_workers": 2
},
"checkpoints": {
"g_pretrained_path": "pretrained_weights/G_bb_v100_820000.pth",
"d_pretrained_path": "pretrained_weights/D_bb_v100_820000.pth",
"auto_delete_old_checkpoints": true,
"replace_old_checkpoints_mode": false,
"num_checkpoints_to_keep": 2
}
}

[eve2ptp]

batch_size根据自己显存调整

[Fantendo2001]

batch_size根据自己显存调整

谢谢回答！但用colab40GB的A100之后batch size降到了2了还是会出问题:(，有可能是其他方面的问题吗？

[Fantendo2001]

降低batch size无果，最后通过减少train和val的数据量解决了

[zhazl]

同样遇到这个问题

[liziru]

降低batch size无果，最后通过减少train和val的数据量解决了

请问还有更好的解决办法吗，数据集也比较大。@ @Fantendo2001