Where can I get the final model

[thunder-007]

I don't have enough compute power to train the model by myself. It would be helpful if I can get the huggingfaces link also the script to load and test a test script with nii.gz file and slice as input and predicts the output https://github.com/Project-MONAI/tutorials/blob/main/3d_segmentation/brats_segmentation_3d.ipynb

model.load_state_dict(torch.load(os.path.join(root_dir, "best_metric_model.pth")))

[KumoLiu]

Hi @thunder-007, sorry that I didn't fully understand your question.
In the tutorial link you shared, it saved the best model by using:

torch.save(
                    model.state_dict(),
                    os.path.join(root_dir, "best_metric_model.pth"),
                )

You can also use torch.save to save the final model.
Hope it helps, thanks!

[thunder-007]

HI @KumoLiu I mean I don''t have infrastructure to train where can I directly find the pretrained model also I'm confused with the inference and preprocessing is there some simple method which takes the .nii.gz file and slice number as input and give the mask and the image as ouput.

[KumoLiu]

Hi @thunder-007, I didn't fully get your question, if you are using a dataset.json, you may also update the json file.
And for what I mean, you can only put one file in this files.
https://github.com/Project-MONAI/tutorials/blob/d0de28e3e184847a4adf84bbc8f8f6e8b699c012/3d_segmentation/torch/unet_inference_dict.py#L51C5-L51C10

[thunder-007]

@KumoLiu sorry for the confusion I will explain as simple as possible kindly tell me if something isn't understandable

import monai
model = monai.bundle.load("brats_mri_segmentation")
images , masks = model.predict("path/of/nii.gz file") # I wanted to get the tumor images and segmentation from the model

I'm will get the pretrained model from the bundle.load method how can make inference now use the model to create a segment as numpu array simply with just the single nni.gz

I just wanted to make a simple ui for the model.

[thunder-007]

thanks a lot for the above reply https://github.com/Project-MONAI/tutorials/blob/d0de28e3e184847a4adf84bbc8f8f6e8b699c012/3d_segmentation/torch/unet_inference_dict.py#L51C5-L51C10 can I have same such inferece.py for SegRestNet similarly?

[KumoLiu]

Yes, sure.

[thunder-007]

Thanks I'm finally able to achieve it I will make a notebook and a pull request thank you...