You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently, dglpart.c assumes the following about the inputs:
The nodes, edges, stats files are named "{STEM}_nodes.txt", "{STEM}_edges.txt", "{STEM}_stats.txt", where {STEM} is a common prefix
These nodes/edges/stats files all reside at the working dir where the cmd is invoked
This is a little inflexible. It'd be nice if we instead could invoke pm_dglpart by passing in these paths explicitly, eg something like:
# "some_graph_name" is also explicitly passed in as the graph name
pm_dglpart \
path/to/my_graph/nodes/my_nodes.txt \
path/to/my_graph/edges/edges.txt \
path/to/stats/stats.txt \
some_graph_name \
2
Perhaps this would break downstream consumers that assume that the graph name is encoded in the output files, eg p0-xxx-nodes.txt. So, maybe leaving this as-is is a reasonable compromise to maintain current behavior...
The text was updated successfully, but these errors were encountered:
Currently, dglpart.c assumes the following about the inputs:
This is a little inflexible. It'd be nice if we instead could invoke pm_dglpart by passing in these paths explicitly, eg something like:
Perhaps this would break downstream consumers that assume that the graph name is encoded in the output files, eg
p0-xxx-nodes.txt
. So, maybe leaving this as-is is a reasonable compromise to maintain current behavior...The text was updated successfully, but these errors were encountered: