Diffusion Autoencoder tutorial #361
Conversation
| # ## Setup imports | ||
|
|
||
| # %% jupyter={"outputs_hidden": false} | ||
| import os |
| from torch.cuda.amp import GradScaler, autocast | ||
| from tqdm import tqdm |
| # %% jupyter={"outputs_hidden": false} | ||
| directory = os.environ.get("MONAI_DATA_DIRECTORY") | ||
| root_dir = tempfile.mkdtemp() if directory is None else directory | ||
| root_dir = '/home/s2086085/pedro_idcom/experiment_data' |
There was a problem hiding this comment.
please remove this line specific to your system
| transforms.Lambdad(keys=["image"], func=lambda x: x[channel, :, :, :]), | ||
| transforms.AddChanneld(keys=["image"]), |
There was a problem hiding this comment.
Gives the warning <class 'monai.transforms.utility.array.AddChannel'>: Class AddChannel has been deprecated since version 0.8. please use MetaTensor data type and monai.transforms.EnsureChannelFirst instead.
Could get rid of this by deleted AddChannel and adding. channel during the Lambda like:
transforms.Lambdad(keys=["image"], func=lambda x: x[channel, None, :, :, :]),
| # 1. `LoadImaged` loads the brain images from files. | ||
| # 2. `EnsureChannelFirstd` ensures the original data to construct "channel first" shape. | ||
| # 3. The first `Lambdad` transform chooses the first channel of the image, which is the T1-weighted image. | ||
| # 4. `Spacingd` resamples the image to the specified voxel spacing, we use 3,3,2 mm to match the original paper. |
There was a problem hiding this comment.
hm what do you mean by the original paper here? The CVPR paper only uses computer vision data as far as I can see
There was a problem hiding this comment.
Ohh, sorry, this was copied from another tutorial
| # 6. The last `Lambdad` transform obtains `slice_label` by summing up the label to have a single scalar value (healthy `=1` or not `=2` ). | ||
|
|
||
| # %% | ||
| channel = 0 # 0 = Flair |
There was a problem hiding this comment.
Says T1-weighted in step 3. of the transforms list but FLAIR here
| section="training", | ||
| cache_rate=1.0, # you may need a few Gb of RAM... Set to 0 otherwise | ||
| num_workers=4, | ||
| download=False, # Set download to True if the dataset hasnt been downloaded yet |
There was a problem hiding this comment.
we could set this to True so it just runs out the box for users without having to change anything - I think it doesn't actually do the download if it detects the file is already there
| noise_pred = self.unet(x=xt, timesteps=t, context=latent.unsqueeze(2)) | ||
| return noise_pred, latent | ||
|
|
||
| device = torch.device("cuda:2") |
There was a problem hiding this comment.
i think it might be best for users if we default to device 0
| # ## Training a diffusion model and semantic encoder | ||
|
|
||
| # %% | ||
| n_iterations = 1e4 # training for longer helps a lot with reconstruction quality, even if the loss is already low |
There was a problem hiding this comment.
could you just give an estimate of training time in the comments? useful so users know how to change this variable if they just want 'the best results they can get in 10 mins of training' or something similar
There was a problem hiding this comment.
just noticed it says at the end of training, ~3 hours
There was a problem hiding this comment.
Yes, it's not super quick...
| # get latent space of training set | ||
| latents_train = [] | ||
| classes_train = [] | ||
| for i in range(15): # 15 slices from each volume |
There was a problem hiding this comment.
i stared at this a while working out why i was not being used! could we replace i with _ to make clear we don't actually use the variable, its just a way of running the transform 15 times per batch?
Create a Diffusion Autoencoder tutorial using the components that we already have for image manipulation.