Merge pull request ben0oil1#3 from CNXudiandian/directml

增加对AMD用户的友好功能~

config.json

@@ -0,0 +1,10 @@
+{
+    "cnhubert_path":"./pretrained/chinese-hubert-base",
+    "bert_path":"./pretrained/chinese-roberta-wwm-ext-large",
+    "sovits_path":"./models/sovits.pth",
+    "gpt_path":"./models/gpt.ckpt",
+    "default_refer_path":"./models/referaudio.wav",
+    "default_refer_text":"参考文本",
+    "default_refer_language":"zh",
+    "is_half":false
+}

requirements.txt

@@ -5,95 +5,79 @@ anyio==4.3.0
 async-timeout==4.0.3
 attrs==23.2.0
 audioread==3.0.1
-blinker==1.4
 certifi==2024.2.2
 cffi==1.16.0
 charset-normalizer==3.3.2
 click==8.1.7
 cn2an==0.5.22
-contourpy==1.2.0
-cryptography==3.4.8
+colorama==0.4.6
+contourpy==1.2.1
 cycler==0.12.1
-dbus-python==1.2.18
 decorator==5.1.1
-distro==1.7.0
-distro-info==1.1+ubuntu0.2
 einops==0.7.0
 exceptiongroup==1.2.0
 fastapi==0.110.0
-feature-extractor==0.0.1
-ffmpeg-python==0.1.18
-filelock==3.13.1
-fonttools==4.49.0
+ffmpeg-python==0.2.0
+filelock==3.13.3
+fonttools==4.51.0
 frozenlist==1.4.1
-fsspec==2024.2.0
+fsspec==2024.3.1
 future==1.0.0
 h11==0.14.0
-httplib2==0.20.2
-huggingface-hub==0.21.4
+huggingface-hub==0.22.2
 idna==3.6
-importlib-metadata==4.6.4
-jeepney==0.7.1
+importlib_resources==6.4.0
 jieba==0.42.1
 jieba-fast==0.53
 Jinja2==3.1.3
 joblib==1.3.2
-keyring==23.5.0
 kiwisolver==1.4.5
-launchpadlib==1.10.16
-lazr.restfulclient==0.14.4
-lazr.uri==1.0.6
-librosa==0.9.2
-lightning-utilities==0.10.1
-llvmlite==0.39.1
+lazy_loader==0.4
+librosa==0.10.1
+lightning-utilities==0.11.2
+llvmlite==0.42.0
 MarkupSafe==2.1.5
 matplotlib==3.8.3
-more-itertools==8.10.0
 mpmath==1.3.0
+msgpack==1.0.8
 multidict==6.0.5
 networkx==3.2.1
-numba==0.56.4
-numpy==1.23.5
-oauthlib==3.2.0
-packaging==23.2
-pillow==10.2.0
-platformdirs==2.5.1
+numba==0.59.1
+numpy==1.26.4
+packaging==24.0
+pillow==10.3.0
+platformdirs==4.2.0
 pooch==1.8.1
 proces==0.1.7
-pycparser==2.21
-pydantic==2.6.3
+pycparser==2.22
+pydantic==2.6.4
 pydantic_core==2.16.3
-PyGObject==3.42.1
-PyJWT==2.3.0
 pyparsing==3.1.2
 pypinyin==0.50.0
-python-apt==2.4.0+ubuntu3
 python-dateutil==2.9.0.post0
 pytorch-lightning==2.2.1
 PyYAML==6.0.1
 regex==2023.12.25
 requests==2.31.0
-resampy==0.4.3
 safetensors==0.4.2
 scikit-learn==1.4.1.post1
-scipy==1.12.0
-SecretStorage==3.3.1
+scipy==1.13.0
 six==1.16.0
 sniffio==1.3.1
 soundfile==0.12.1
+soxr==0.3.7
 starlette==0.36.3
 sympy==1.12
-threadpoolctl==3.3.0
+threadpoolctl==3.4.0
 tokenizers==0.15.2
-torch==2.2.1
-torchaudio==2.2.1
-torchmetrics==1.3.1
+torch==2.0.0
+torch-directml==0.2.0.dev230426
+torchmetrics==1.3.2
+torchvision==0.15.1
 tqdm==4.66.2
 transformers==4.38.2
-typing_extensions==4.10.0
-unattended-upgrades==0.1
+typing_extensions==4.11.0
 urllib3==2.2.1
 uvicorn==0.27.1
-wadllib==1.3.6
 yarl==1.9.4
-zipp==1.0.0
+zipp==3.18.1

server.py

@@ -11,6 +11,7 @@
 import numpy as np
 import soundfile as sf
 import torch
+import torch_directml
 import uvicorn
 from AR.models.t2s_lightning_module import Text2SemanticLightningModule
 from fastapi import FastAPI, HTTPException, Request
@@ -22,21 +23,32 @@
 from transformers import AutoModelForMaskedLM, AutoTokenizer 
 from text import cleaned_text_to_sequence
 from text.cleaner import clean_text
-
-# -------pretrained_models和训练好的模型，丢在下面的位置----------------
-cnhubert_path = "./data/pretrained_models/chinese-hubert-base"
-bert_path = "./data/pretrained_models/chinese-roberta-wwm-ext-large" 
-# 从云端下载回来的训练好的内容
-sovits_path = './data/_models/svc/0128-0359_e12_s144.pth'
-gpt_path = './data/_models/gpt/0128-0359-e30.ckpt' 
-# 推理引用的音频文件地址和文本信息 
-default_refer_path =   './data/_models/000.wav'  
-default_refer_text =  "云南凤庆给您发货，一斤装四十九，两斤装九十五，三斤装一百二十九，规格越大价格越划算。" 
-# 语言，我个人把日语韩语乱七八糟的直接删除了，因为我用不上，大家需要的话自己做适配
-default_refer_language = 'zh' 
-# 使用cuda就是用英伟达的gpu，cpu就是用cpu
-device = 'cpu'  # cpu cuda
-is_half = False 
+import json
+
+# 增加一个读取配置文件的功能
+with open("config.json", "r", encoding="utf-8") as f:
+    config = json.load(f)
+    cnhubert_path = config["cnhubert_path"]
+    bert_path = config["bert_path"]
+    sovits_path = config["sovits_path"]
+    gpt_path = config["gpt_path"]
+    default_refer_path = config["default_refer_path"]
+    default_refer_text = config["default_refer_text"]
+    default_refer_language = config["default_refer_language"]
+    is_half = config["is_half"]
+# 自动判断环境是否支持CUDA和DirectML
+if(torch.cuda.is_available()):
+    print("CUDA可用，将使用CUDA进行推理加速。")
+    print("设备名称:",torch.cuda.get_device_name(0))
+    device = "cuda"
+else:
+    if(torch_directml.is_available()==False):
+        device = "cpu"
+        print("在本机没有发现可以用于加速的显卡，使用CPU进行推理运算。")
+    else:
+        device = torch_directml.device(0)
+        print("DirectML可用，将使用DirectML进行推理加速。")
+        print("设备名称:",torch_directml.device_name(0))
 # -----------------------
 
 # 如果要增加更多的参数选项，在这里设定