Closed cthoyt closed 2 years ago
Alright the issue with the torchdrug widths was solved by adding a constant.
Merging #52 (2292b4a) into main (8fdd199) will decrease coverage by
0.60%
. The diff coverage is88.63%
.
@@ Coverage Diff @@
## main #52 +/- ##
==========================================
- Coverage 98.45% 97.84% -0.61%
==========================================
Files 27 28 +1
Lines 712 744 +32
==========================================
+ Hits 701 728 +27
- Misses 11 16 +5
Impacted Files | Coverage Δ | |
---|---|---|
chemicalx/models/__init__.py | 100.00% <ø> (ø) |
|
chemicalx/models/deepsynergy.py | 100.00% <ø> (ø) |
|
chemicalx/data/labeledtriples.py | 97.87% <50.00%> (-2.13%) |
:arrow_down: |
chemicalx/data/datasetloader.py | 90.78% <81.81%> (-3.95%) |
:arrow_down: |
chemicalx/constants.py | 100.00% <100.00%> (ø) |
|
chemicalx/models/epgcnds.py | 100.00% <100.00%> (ø) |
|
tests/unit/test_models.py | 100.00% <100.00%> (ø) |
Continue to review full report at Codecov.
Legend - Click here to learn more
Δ = absolute <relative> (impact)
,ø = not affected
,? = missing data
Powered by Codecov. Last update 8fdd199...2292b4a. Read the comment docs.
Summary
It wasn't clear where the magic parameter numbers came from in the existing two examples - it turns out that a few of the inputs to a given model need to come from the dataset. This PR standardizes the names of those parameters and makes them easier to get from an instance of a dataset by exposing them as python properties
69
for EPGCN-DS + DrugCombDB comes from (update: a constant from insidetorchdrug
)Changes
lru_cache
to functions that get datasets' partsdrug_channels
andcontext_channels
to correspond respectively to # features for each drug and # features for each context.drug_channels
andcontext_channels
(and update the test that checks for default values in each model'sdef __init__()