FIx: cannot identify one class to a dict(needed)

To implement recursive construction while retaining the characteristics of the original dictionary, we can slightly modify the DictToAttrRecursive class. This allows each object to retain its characteristics as a dictionary while accessing keys and values as attributes.

GPT_SoVITS/inference_webui.py

@@ -10,25 +10,15 @@ if("_CUDA_VISIBLE_DEVICES"in os.environ):
 is_half=eval(os.environ.get("is_half","True"))
 import gradio as gr
 from transformers import AutoModelForMaskedLM, AutoTokenizer
-import sys,torch,numpy as np
+import numpy as np
-from pathlib import Path
+import librosa,torch
-import os,pdb,utils,librosa,math,traceback,requests,argparse,torch,multiprocessing,pandas as pd,torch.multiprocessing as mp,soundfile
-# torch.backends.cuda.sdp_kernel("flash")
-# torch.backends.cuda.enable_flash_sdp(True)
-# torch.backends.cuda.enable_mem_efficient_sdp(True)  # Not avaliable if torch version is lower than 2.0
-# torch.backends.cuda.enable_math_sdp(True)
-from random import shuffle
-from AR.utils import get_newest_ckpt
-from glob import glob
-from tqdm import tqdm
 from feature_extractor import cnhubert
 cnhubert.cnhubert_base_path=cnhubert_base_path
-from io import BytesIO
+
 from module.models import SynthesizerTrn
 from AR.models.t2s_lightning_module import Text2SemanticLightningModule
-from AR.utils.io import load_yaml_config
 from text import cleaned_text_to_sequence
-from text.cleaner import text_to_sequence, clean_text
+from text.cleaner import clean_text
 from time import time as ttime
 from module.mel_processing import spectrogram_torch
 from my_utils import load_audio
@@ -58,16 +48,36 @@ def get_bert_feature(text, word2ph):
 n_semantic = 1024
 dict_s2=torch.load(sovits_path,map_location="cpu")
 hps=dict_s2["config"]
-class DictToAttrRecursive:
+
+class DictToAttrRecursive(dict):
     def __init__(self, input_dict):
+        super().__init__(input_dict)
         for key, value in input_dict.items():
             if isinstance(value, dict):
-                # 如果值是字典，递归调用构造函数
+                value = DictToAttrRecursive(value)
-                setattr(self, key, DictToAttrRecursive(value))
+            self[key] = value
-            else:
             setattr(self, key, value)
 
+    def __getattr__(self, item):
+        try:
+            return self[item]
+        except KeyError:
+            raise AttributeError(f"Attribute {item} not found")
+
+    def __setattr__(self, key, value):
+        if isinstance(value, dict):
+            value = DictToAttrRecursive(value)
+        super(DictToAttrRecursive, self).__setitem__(key, value)
+        super().__setattr__(key, value)
+
+    def __delattr__(self, item):
+        try:
+            del self[item]
+        except KeyError:
+            raise AttributeError(f"Attribute {item} not found")
+
 hps = DictToAttrRecursive(hps)
+
 hps.model.semantic_frame_rate="25hz"
 dict_s1=torch.load(gpt_path,map_location="cpu")
 config=dict_s1["config"]