cannot load model anymore as weights_only

[beefsoup18]

error occurs when using inference demo

Traceback (most recent call last): File "/FastSAM/Inference.py", line 122, in main(args) File "/FastSAM/Inference.py", line 76, in main model = FastSAM(args.model_path) ^^^^^^^^^^^^^^^^^^^^^^^^ File "/FastSAM/ultralytics/yolo/engine/model.py", line 107, in init self._load(model, task) File "/FastSAM/ultralytics/yolo/engine/model.py", line 156, in _load self.model, self.ckpt = attempt_load_one_weight(weights) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/FastSAM/ultralytics/nn/tasks.py", line 578, in attempt_load_one_weight ckpt, weight = torch_safe_load(weight) # load ckpt ^^^^^^^^^^^^^^^^^^^^^^^ File "/FastSAM/ultralytics/nn/tasks.py", line 518, in torch_safe_load return torch.load(file, map_location='cpu'), file # load ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "**********************/lib/python3.11/site-packages/torch/serialization.py", line 1524, in load raise pickle.UnpicklingError(_get_wo_message(str(e))) from None _pickle.UnpicklingError: Weights only load failed. This file can still be loaded, to do so you have two options, do those steps only if you trust the source of the checkpoint. (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 ultralytics.nn.tasks.SegmentationModel was not an allowed global by default. Please use torch.serialization.add_safe_globals([ultralytics.nn.tasks.SegmentationModel]) or the torch.serialization.safe_globals([ultralytics.nn.tasks.SegmentationModel]) context manager to allowlist this global if you trust this class/function.

Check the documentation of torch.load to learn more about types accepted by default with weights_only https://pytorch.org/docs/stable/generated/torch.load.html.

Even if using argument as 'weights_only=False' when load checkpoint cannot resolve this problem

[astral-fi]

I am facing the same issue, Did you find any solution??

[nazar1ous]

Hey, the following issue can be easily solved with downgrading the torch to 1 version. For example, I have used: pip install torch==1.13.1+cu116 torchvision==0.14.1+cu116 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu116

As well as, update the numpy to make it work again: pip install numpy==1.24.3

[beefsoup18]

Hey, the following issue can be easily solved with downgrading the torch to 1 version. For example, I have used: pip install torch==1.13.1+cu116 torchvision==0.14.1+cu116 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu116

As well as, update the numpy to make it work again: pip install numpy==1.24.3

And by this solution your python version could only be 3.9 or 3.10

@astral-fi

[devietti]

I was able to get this working with python-3.13, torch-2.7.1 and numpy-2.2.6 by following the advice in the error messages and adding a bunch of torch.serialization.add_safe_globals calls before loading the model.

from fastsam import FastSAM, FastSAMPrompt
import torch, ultralytics

torch.serialization.add_safe_globals([ultralytics.nn.tasks.SegmentationModel])
torch.serialization.add_safe_globals([torch.nn.modules.container.Sequential])
torch.serialization.add_safe_globals([ultralytics.nn.modules.conv.Conv])
torch.serialization.add_safe_globals([torch.nn.modules.conv.Conv2d])
torch.serialization.add_safe_globals([torch.nn.modules.batchnorm.BatchNorm2d])
torch.serialization.add_safe_globals([torch.nn.modules.activation.SiLU])
torch.serialization.add_safe_globals([ultralytics.nn.modules.block.C2f])
torch.serialization.add_safe_globals([torch.nn.modules.container.ModuleList])
torch.serialization.add_safe_globals([ultralytics.nn.modules.block.Bottleneck])
torch.serialization.add_safe_globals([ultralytics.nn.modules.block.SPPF])
torch.serialization.add_safe_globals([torch.nn.modules.pooling.MaxPool2d])
torch.serialization.add_safe_globals([torch.nn.modules.upsampling.Upsample])
torch.serialization.add_safe_globals([ultralytics.nn.modules.conv.Concat])
torch.serialization.add_safe_globals([ultralytics.nn.modules.head.Segment])
torch.serialization.add_safe_globals([ultralytics.nn.modules.block.DFL])
torch.serialization.add_safe_globals([ultralytics.nn.modules.block.Proto])
torch.serialization.add_safe_globals([torch.nn.modules.conv.ConvTranspose2d])
torch.serialization.add_safe_globals([getattr])
torch.serialization.add_safe_globals([ultralytics.nn.modules.head.Detect])
torch.serialization.add_safe_globals([ultralytics.yolo.utils.IterableSimpleNamespace])

IMAGE_PATH = 'input.png'
DEVICE = 'cpu'
model = FastSAM('weights/FastSAM-x.pt')
everything_results = model(IMAGE_PATH, device=DEVICE, retina_masks=True, imgsz=1024, conf=0.4, iou=0.9,)
prompt_process = FastSAMPrompt(IMAGE_PATH, everything_results, device=DEVICE)
ann = prompt_process.everything_prompt()
prompt_process.plot(annotations=ann,output_path='output.jpg',)