mms-tts-eng模型推理生成不了音频。

[yegoling]

Describe the bug/ 问题描述 (Mandatory / 必填)
mms-tts-eng模型推理时报错。

• Hardware Environment(Ascend/GPU/CPU) / 硬件环境:

CPU

• Software Environment / 软件环境 (Mandatory / 必填):
  -- MindSpore version : 2.4.0
  -- MindNlp version : 0.4.0
  -- Python version : 3.9.20

To Reproduce / 重现步骤 (Mandatory / 必填)

运行推理代码

from mindnlp.transformers import VitsModel, AutoTokenizer
model = VitsModel.from_pretrained("./model/mms-tts-eng",from_pt=True)
tokenizer = AutoTokenizer.from_pretrained("./model/mms-tts-eng",from_pt=True)

text = "some example text in the English language"

inputs = tokenizer(text, return_tensors='ms')
output = model(**inputs).waveform

import scipy
scipy.io.wavfile.write("techno.wav", rate=model.config.sampling_rate, data=output)

Expected behavior / 预期结果 (Mandatory / 必填)
生成音频文件

Screenshots/ 日志 / 截图 (Mandatory / 必填)

Building prefix dict from the default dictionary ...
Loading model from cache C:\Users\23974\AppData\Local\Temp\jieba.cache
Loading model cost 0.572 seconds.
Prefix dict has been built successfully.
Some weights of VitsModel were not initialized from the model checkpoint at ./model/mms-tts-eng and are newly initialized: ['flow.flows.0.wavenet.in_layers.0.weight', 'flow.flows.0.wavenet.in_layers.1.weight', 'flow.flows.0.wavenet.in_layers.2.weight', 'flow.flows.0.wavenet.in_layers.3.weight', 'flow.flows.0.wavenet.res_skip_layers.0.weight', 'flow.flows.0.wavenet.res_skip_layers.1.weight', 'flow.flows.0.wavenet.res_skip_layers.2.weight', 'flow.flows.0.wavenet.res_skip_layers.3.weight', 'flow.flows.1.wavenet.in_layers.0.weight', 'flow.flows.1.wavenet.in_layers.1.weight', 'flow.flows.1.wavenet.in_layers.2.weight', 'flow.flows.1.wavenet.in_layers.3.weight', 'flow.flows.1.wavenet.res_skip_layers.0.weight', 'flow.flows.1.wavenet.res_skip_layers.1.weight', 'flow.flows.1.wavenet.res_skip_layers.2.weight', 'flow.flows.1.wavenet.res_skip_layers.3.weight', 'flow.flows.2.wavenet.in_layers.0.weight', 'flow.flows.2.wavenet.in_layers.1.weight', 'flow.flows.2.wavenet.in_layers.2.weight', 'flow.flows.2.wavenet.in_layers.3.weight', 'flow.flows.2.wavenet.res_skip_layers.0.weight', 'flow.flows.2.wavenet.res_skip_layers.1.weight', 'flow.flows.2.wavenet.res_skip_layers.2.weight', 'flow.flows.2.wavenet.res_skip_layers.3.weight', 'flow.flows.3.wavenet.in_layers.0.weight', 'flow.flows.3.wavenet.in_layers.1.weight', 'flow.flows.3.wavenet.in_layers.2.weight', 'flow.flows.3.wavenet.in_layers.3.weight', 'flow.flows.3.wavenet.res_skip_layers.0.weight', 'flow.flows.3.wavenet.res_skip_layers.1.weight', 'flow.flows.3.wavenet.res_skip_layers.2.weight', 'flow.flows.3.wavenet.res_skip_layers.3.weight', 'posterior_encoder.wavenet.in_layers.0.weight', 'posterior_encoder.wavenet.in_layers.1.weight', 'posterior_encoder.wavenet.in_layers.10.weight', 'posterior_encoder.wavenet.in_layers.11.weight', 'posterior_encoder.wavenet.in_layers.12.weight', 'posterior_encoder.wavenet.in_layers.13.weight', 'posterior_encoder.wavenet.in_layers.14.weight', 'posterior_encoder.wavenet.in_layers.15.weight', 'posterior_encoder.wavenet.in_layers.2.weight', 'posterior_encoder.wavenet.in_layers.3.weight', 'posterior_encoder.wavenet.in_layers.4.weight', 'posterior_encoder.wavenet.in_layers.5.weight', 'posterior_encoder.wavenet.in_layers.6.weight', 'posterior_encoder.wavenet.in_layers.7.weight', 'posterior_encoder.wavenet.in_layers.8.weight', 'posterior_encoder.wavenet.in_layers.9.weight', 'posterior_encoder.wavenet.res_skip_layers.0.weight', 'posterior_encoder.wavenet.res_skip_layers.1.weight', 'posterior_encoder.wavenet.res_skip_layers.10.weight', 'posterior_encoder.wavenet.res_skip_layers.11.weight', 'posterior_encoder.wavenet.res_skip_layers.12.weight', 'posterior_encoder.wavenet.res_skip_layers.13.weight', 'posterior_encoder.wavenet.res_skip_layers.14.weight', 'posterior_encoder.wavenet.res_skip_layers.15.weight', 'posterior_encoder.wavenet.res_skip_layers.2.weight', 'posterior_encoder.wavenet.res_skip_layers.3.weight', 'posterior_encoder.wavenet.res_skip_layers.4.weight', 'posterior_encoder.wavenet.res_skip_layers.5.weight', 'posterior_encoder.wavenet.res_skip_layers.6.weight', 'posterior_encoder.wavenet.res_skip_layers.7.weight', 'posterior_encoder.wavenet.res_skip_layers.8.weight', 'posterior_encoder.wavenet.res_skip_layers.9.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
[WARNING] KERNEL(,4408,?):2024-11-9 15:52:50 [mindspore\ccsrc\kernel/kernel.h:916] mindspore::kernel::CheckShapeNull] For 'ReduceMin', the shape of input cannot contain zero, but got [const vector]{0}
[WARNING] KERNEL(,4408,?):2024-11-9 15:52:50 [mindspore\ccsrc\kernel/kernel.h:916] mindspore::kernel::CheckShapeNull] For 'ReduceMax', the shape of input cannot contain zero, but got [const vector]{0}
[WARNING] KERNEL(,4408,?):2024-11-9 15:52:50 [mindspore\ops\kernel\cpu\arithmetic_cpu_kernel.cc:171] mindspore::kernel::`anonymous-namespace'::ArithmeticCpuTypeFunc<float>::RunFunc] Mul output shape contain 0, output_shape: [const vector]{0, 1}
[WARNING] KERNEL(,4408,?):2024-11-9 15:52:50 [mindspore\ops\kernel\cpu\arithmetic_cpu_kernel.cc:171] mindspore::kernel::`anonymous-namespace'::ArithmeticCpuTypeFunc<float>::RunFunc] Add output shape contain 0, output_shape: [const vector]{0, 1}

Additional context / 备注 (Optional / 选填)
代码.md