0.66cpu版本启动报错

[gushuaialan1]

运行环境

• 操作系统 (Linux/macOS/Windows)：windows
• 部署方式 (Docker/windows快速部署包/自己搭的环境)：windows快速部署包
• Python 版本 (如果是部署包可不填)：
• 代码版本/部署包版本:0.66 cpu

问题描述

提示：AttributeError: module 'torch' has no attribute 'compiler'是不是torch版本的问题？ Traceback (most recent call last): File "D:\AIVOICE\vits-simple-api-windows-cpu-v0.6.6\app.py", line 10, in from tts_app.frontend.views import frontend File "D:\AIVOICE\vits-simple-api-windows-cpu-v0.6.6\tts_app\frontend\views.py", line 3, in from tts_app.model_manager import model_manager File "D:\AIVOICE\vits-simple-api-windows-cpu-v0.6.6\tts_app\model_manager.py", line 1, in from manager.ModelManager import ModelManager File "D:\AIVOICE\vits-simple-api-windows-cpu-v0.6.6\manager\ModelManager.py", line 15, in from bert_vits2 import Bert_VITS2 File "D:\AIVOICE\vits-simple-api-windows-cpu-v0.6.6\bert_vits2__init.py", line 1, in from bert_vits2.bert_vits2 import Bert_VITS2 File "D:\AIVOICE\vits-simple-api-windows-cpu-v0.6.6\bert_vits2\bert_vits2.py", line 10, in from bert_vits2.models import SynthesizerTrn File "D:\AIVOICE\vits-simple-api-windows-cpu-v0.6.6\bert_vits2\models.py", line 12, in from vector_quantize_pytorch import VectorQuantize File "D:\AIVOICE\vits-simple-api-windows-cpu-v0.6.6\py310\lib\site-packages\vector_quantize_pytorch__init__.py", line 2, in from vector_quantize_pytorch.residual_vq import ResidualVQ, GroupedResidualVQ File "D:\AIVOICE\vits-simple-api-windows-cpu-v0.6.6\py310\lib\site-packages\vector_quantize_pytorch\residual_vq.py", line 13, in from einx import get_at File "D:\AIVOICE\vits-simple-api-windows-cpu-v0.6.6\py310\lib\site-packages\einx__init.py", line 7, in from . import backend File "D:\AIVOICE\vits-simple-api-windows-cpu-v0.6.6\py310\lib\site-packages\einx\backend\init__.py", line 27, in _update() File "D:\AIVOICE\vits-simple-api-windows-cpu-v0.6.6\py310\lib\site-packages\einx\backend\init__.py", line 25, in _update backends.append(backend_factories[backend_name]()) File "D:\AIVOICE\vits-simple-api-windows-cpu-v0.6.6\py310\lib\site-packages\einx\backend_torch.py", line 127, in make_torch_backend einx.lru_cache.decorate_tracedfunctions(torch.compiler.allow_in_graph) AttributeError: module 'torch' has no attribute 'compiler'

问题复现步骤

[gushuaialan1]

Gpu版本也报这个错

[gushuaialan1]

我看compiler is only available for PyTorch > 2.0

[Artrajz]

感谢反馈！应该是vector-quantizer-pytorch更新了，内部使用了新版本torch的函数。但由于之前测试torch 1.13.1 在显存方面表现良好，所以还是继续保持使用1.13.1版本，将vector-quantizer-pytorch替换为稍旧的版本。

[Artrajz]

已更改vector-quantizer-pytorch版本