AMR View
The GUI allows for simple viewing, conversion and plotting of AMR Graphs.
Operation
For pip installations, the executable amr_view
should be installed in the user's path. This file
is simply a wrapper that calls amrlib/amr_view/cli.py::main()
. When running in-place from a GitHub
download the folder scripts/50_Build_AMR_View
has the file run_amr_view.py
which is setup to run the
version of amrlib two levels above it. This is basically the same as cli.py::main()
with the exception of
some path options.
Note that on startup loading models can take some time, during which the various buttons won't work. The Window title at the top will tell you if the system is ready or loading. In addition, log messages to the console outline the loading progress.
The File
menu can be used to load and save AMR data. Use To AMR
to convert the Input Sentence
to an AMR graph. Use Show Graph
to display a plot of the graph with the default system viewer.
Be sure there is a valid graph in the main window to plot. The Generate
button converts the graph
to N (gtos_num_ret_seq below) different output sentences.
Configuration options
The file amrlib/amr_view/amr_view.json
holds the configuration parameters.
"stog_model_dir": "amrlib/data/model_stog",
"stog_model_fn": "model.pt",
"stog_device": "cpu",
"gtos_model_dir": "amrlib/data/model_gtos",
"gtos_num_ret_seq": 8,
"gtos_num_beams": 8,
"gtos_batch_size": 1,
"gtos_device": "cpu"
stog
options refer to the sequence-to-graph model (aka parsing)
while gtos
options refer to the graph-to-sequence model (aka generation)
gtos_num_ret_seq
sets the number of returned sentences that will be generated and printed
gtos_num_beams
sets the number of beams for the beam search. This needs to be >= gtos_num_ret_seq
The x_device
params allow the model to run on the cpu
or can be changed to cuda:0
to run on the gpu.
Because we're only processing one graph or sentence at a time instead of a batch, there probably won't
be a noticeable difference between the two settings.