gmm-torch
gmm-torch copied to clipboard
ldeecke
Gaussian mixture models in PyTorch.
When I perform the resampling operation, I get the following error: `File "/usr/local/bin/nnUNet_train", line 33, in sys.exit(load_entry_point('nnunet', 'console_scripts', 'nnUNet_train')()) File "/content/drive/MyDrive/nnUNetframe/nnUNet/nnunet/run/run_training.py", line 179, in main trainer.run_training() File "/content/drive/MyDrive/nnUNetframe/nnUNet/nnunet/training/network_training/nnUNetTrainerV2.py", line 440,...
Hello, I have this problem, how can I solve it? File "/usr/local/bin/nnUNet_train", line 33, in sys.exit(load_entry_point('nnunet', 'console_scripts', 'nnUNet_train')()) File "/content/drive/MyDrive/nnUNetframe/nnUNet/nnunet/run/run_training.py", line 179, in main trainer.run_training() File "/content/drive/MyDrive/nnUNetframe/nnUNet/nnunet/training/network_training/nnUNetTrainerV2.py", line 440, in...
Added FisherVectorLayer class, that calculates a fisher vector representation of a batched input x as an nn.Module
Hi, When using "full" covariance to fit the gmm model, the following error occurred: ``` cholesky_cuda: U(31,31) is zero, singular U. File "/Cylinder3D/network/gmm.py", line 320, in _calculate_log_det log_det[k] = 2...
This enables you to install it with: ``` pip install git+https://github.com/ldeecke/gmm-torch ``` rather than cloning and copying the code into a codebase
Improved speed and memory usage with following optimizations: 1. the `(N, K, 1, D) * (1, K, D, D)` matmul at [line275 ](https://github.com/ldeecke/gmm-torch/blob/master/gmm.py#L275) is replaced with an equivalent matmul `(K,...
''' exponent = torch.exp(-.5 * torch.sum((x - mu) * (x - mu) / var, 2, keepdim=True)) RuntimeError: expected device cpu and dtype Float but got device cuda:0 and dtype Float...
When the n_center variable has a high value, some patterns may not belonging to any of the n_center cluser,. In this case the code ` for c in range(n_centers): cost...
there seems to be some limit between the n_components and n_features. If I try and create a model with ``` n_components=1 n_features=99 ``` it will fail with `_LinAlgError: linalg.cholesky: The...
This allows the GMM model to handle data with 1-dimensional features. Specifically, the shape of ```mat_b``` with 1-dimensional features is ```(1,k,1,1)```, and ```mat_b[0, i, :, :].squeeze()``` returns a numeric value...
