Weight Issue with Pytorch 2.6 Bug Fx

[Daniel-moen]

Sitting on Pytorch 2.6 getting weight (1) In PyTorch 2.6, we changed the default value of the `weights_only` argument in `torch.load` from `False` to `True`. Re-running `torch.load` with `weights_only` set to `False` will likely succeed, but it can result in arbitrary code execution. Do it only if you got the file from a trusted source. (2) Alternatively, to load with `weights_only=True` please check the recommended steps in the following error message. WeightsUnpickler error: Unsupported global: GLOBAL numpy.core.multiarray.scalar was not an allowed global by default. Please use `torch.serialization.add_safe_globals([scalar])` or the `torch.serialization.safe_globals([scalar])` context manager to allowlist this global if you trust this class/function.
on Line 212
checkpoint = torch.load(ckpt_path, map_location=device)
Change it to:
checkpoint = torch.load(ckpt_path, map_location=device, weights_only=False)
This allows PyTorch to load the full model

[Daniel-moen]

Fixed by #619

Работая на PyTorch 2.6, получаем вес (1) In PyTorch 2.6, we changed the default value of the `weights_only` argument in `torch.load` from `False` to `True`. Re-running `torch.load` with `weights_only` set to `False` will likely succeed, but it can result in arbitrary code execution. Do it only if you got the file from a trusted source. (2) Alternatively, to load with `weights_only=True` please check the recommended steps in the following error message. WeightsUnpickler error: Unsupported global: GLOBAL numpy.core.multiarray.scalar was not an allowed global by default. Please use `torch.serialization.add_safe_globals([scalar])` or the `torch.serialization.safe_globals([scalar])` context manager to allowlist this global if you trust this class/function. в строке 212. checkpoint = torch.load(ckpt_path, map_location=device) Измените ее на: checkpoint = torch.load(ckpt_path, map_location=device, weights_only=False) Это позволит PyTorch загрузить полную модель.

Так какой нужно файл открыть?

[1ERRORX]

Работая на PyTorch 2.6, получаем вес (1) In PyTorch 2.6, we changed the default value of the `weights_only` argument in `torch.load` from `False` to `True`. Re-running `torch.load` with `weights_only` set to `False` will likely succeed, but it can result in arbitrary code execution. Do it only if you got the file from a trusted source. (2) Alternatively, to load with `weights_only=True` please check the recommended steps in the following error message. WeightsUnpickler error: Unsupported global: GLOBAL numpy.core.multiarray.scalar was not an allowed global by default. Please use `torch.serialization.add_safe_globals([scalar])` or the `torch.serialization.safe_globals([scalar])` context manager to allowlist this global if you trust this class/function. в строке 212. checkpoint = torch.load(ckpt_path, map_location=device) Измените ее на: checkpoint = torch.load(ckpt_path, map_location=device, weights_only=False) Это позволит PyTorch загрузить полную модель.

Так какой нужно файл открыть?

Just use this code for patch torch.load:
Просто используй этот код чтобы пропатчить torch.load:

import sys
import torch
from importlib import reload

original_load = torch.load
def patched_load(*args, **kwargs):
    kwargs['weights_only'] = False
    return original_load(*args, **kwargs)
torch.load = patched_load

if 'bark' in sys.modules:
    reload(sys.modules['bark'])
from bark import SAMPLE_RATE, generate_audio, preload_models
from IPython.display import Audio

preload_models()

[bendlas]

I've had luck with

import numpy
import torch.serialization

torch.serialization.add_safe_globals([
    (numpy._core.multiarray.scalar, 'numpy.core.multiarray.scalar'),
    numpy.dtype,
    numpy.dtypes.Float64DType
])

preload_models()

that way, I don't have to re-open the security hole, closed by weights_only=True ..