Benign huggingface Download

humair025/kikiri-tts-urdu-pipeline main

Module built on the numeric Python stackImports the NumPy array library

SHA-2568e7398683bab0c0483d920400d0fbcbd5cc6ab3263c05f09aefdc790a6295654

URLhttps://huggingface.co/humair025/kikiri-tts-urdu-pipeline

Evidence

Python module-level function call lines 1–14

⋯4 lines
5import time
6from munch import Munch
7import numpy as np
8import torch
9
10torch.autograd.set_detect_anomaly(True)
11
12if getattr(torch, "_original_load", None) is None:
13    torch._original_load = torch.load
14    torch.load = lambda *args, **kwar …

English function-word token "that" lines 45–50

45:7… ptimizers import build_optimizer
46
47
48# simple fix for dataparallel that allows access to class attributes
49class MyDataParallel(torch.nn.DataParallel):
50    def __getatt …

English function-word token "and" lines 152–163

⋯4 lines
156    iters = 0
157
158    load_pretrained = config.get("pretrained_model", "") != "" and config.get(
159        "second_stage_load_pretrained", False
160    )
161
162    if not load_pretrained:
163   …

English function-word token "the" lines 165–168

69… .pth")
          )
          print("Loading the first stage model at %s ..." % first_stage_path)
          model, _, start_epoch, iters = loa …

NumPy array operation (mean) lines 920–925

40… "diffusion"]["dist"]["sigma_data"] = float(
                  np.mean(running_std)
              )
923
              with open(osp.join(log_dir, osp.basename(config_path)), "w") as outfile:
     …