Weights Metrics

[ElliotStein]

Implemented:

• Framework to compute metrics based on layer weights using existing mergekit infrastructure (run_measure is based on run_merge, metric_methods based on merge_methods etc).
• plot_tools.MetricsHandler to load metrics output, process and interact with statistics.
• plot_tools.ModelGraph to generate a graph to represent the model structure, with node level statistics visible by hovering over a node, and more detailed stats (histograms, rather than means) available by clicking on a node.
• run_metrics.py ties everything together and generates an interactive dashboard displaying the ModelGraph graph.

Not Implemented:

• Split layers into individual heads
• Activation based metrics
• Unit tests