Finish inference_gendb. #220
Conversation
cxx/gendb.hh
Outdated
| @@ -220,7 +225,8 @@ class GenDB { | |||
| // This data structure contains entity sets and linkages. Semantics are | |||
| // map<tuple<class_name, reference_field_name, class_primary_key> ref_val>>, | |||
| GenDB gendb(&prng, schema); | ||
| setup_gendb(&prng, gendb, 20); | ||
| gendb.transition_reference_class_and_ancestors(&prng, "Record"); | ||
| } |
There was a problem hiding this comment.
Please add one or more BOOST_TEST's as to the state of the gendb at this point.
cxx/inference.cc
Outdated
| inference_hirm(prng, gendb->hirm, iters, timeout, verbose); | ||
| gendb->transition_reference_class_and_ancestors( |
There was a problem hiding this comment.
This will do iters iterations of hirm training, followed by a single pass of entity transitions. I don't think that's what we want.
Rather, we probably want to add a parameter for specifying how many iterations of normal hirm training to run between entity transitions, and then use that to control the relative frequencies.
|
I have another debugging PR that needs to go in by tomorrow so I'm going to merge this and the other PR in the next couple hours (the command-line options could be more fully fleshed out, but I'm fine with leaving that to future maintainers to decide what they want to do there). |
emilyfertig
force-pushed
the
093024-emilyaf-inference_gendb
branch
from
e9e5ce8 to
7aecc4b
Compare
cxx/inference.cc
Outdated
| clock_t t_begin = clock(); | ||
| for (int i = 0; i < iters; ++i) { | ||
| // TRANSITION HIRM | ||
| printf("Starting iteration %d, model score = %f\n", i + 1, |
There was a problem hiding this comment.
I suggest changing this to "Starting outer iteration %d", otherwise it is going to be very confusing distinguishing between these and the messages printed out by inference_hirm.
No description provided.