yamle.utils.specific.mimo_experiments.plotting_utils module#
- yamle.utils.specific.mimo_experiments.plotting_utils.plot_input_layer_norm_bar(weights, save_path, iteration=None)[source]#
Plots the bar chart of the input layer norm weights for each member.
It is assumed that the length of the weights is the number of members. Calculate the L1 norm of the weight for each member across input features.
- Return type:
Tensor
- yamle.utils.specific.mimo_experiments.plotting_utils.plot_output_layer_norm_bar(weights, save_path, iteration=None)[source]#
Plots the bar chart of the output layer norm weights for each member.
It is assumed that the length of the weights is the number of members. Calculate the L1 norm of the weight for each member across output features.
- Return type:
Tensor
- yamle.utils.specific.mimo_experiments.plotting_utils.calculate_overlap_between_norms(norms)[source]#
Calculates the overlap between the norms of the members.
- Return type:
Tensor
- yamle.utils.specific.mimo_experiments.plotting_utils.plot_overlap_between_members(overlap, save_path, input=True, iteration=None)[source]#
Plots the overlap between the members.
- yamle.utils.specific.mimo_experiments.plotting_utils.plot_weight_trajectories(weight, save_path, iteration=None, predicted=True)[source]#
Given the depth weights plot the weight trajectories over iterations for each member separately.
- yamle.utils.specific.mimo_experiments.plotting_utils.plot_weight_histogram(weight, save_path, iteration=None, predicted=True)[source]#
Given the depth weights plot the weight histogram over iterations for each member separately.
The weight has the shape (N, depth, members). The histogram is plotted for each member. N is the number of samples for the histogram.
For each member create a subplots for each depth. Try to make the number of subplots square in rows and columns.