Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[WIP] Fix Windows CI #165

Closed
wants to merge 5 commits into from
Closed

[WIP] Fix Windows CI #165

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
538c402
Remove potential data races
MilesCranmer Dec 22, 2022
f4933df
Remove other potential data race
MilesCranmer Dec 23, 2022
b9c0702
Don't skip windows errors
MilesCranmer Dec 23, 2022
e0f0a68
Add TODO
MilesCranmer Dec 23, 2022
08463d2
Force all mutable variables to be copied when threading
MilesCranmer Dec 29, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 0 additions & 12 deletions .github/workflows/CI.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,18 +51,6 @@ jobs:
julia --color=yes --threads=auto --check-bounds=yes --depwarn=yes --code-coverage=user --project=. -e 'import Pkg; Pkg.test(coverage=true)'
julia --color=yes --project=. coverage.jl
shell: bash
if: ${{ matrix.os != 'windows-latest' }}
- name: "Run tests, skipping errors."
run: |
julia --color=yes --project=. -e 'import Pkg; Pkg.add("Coverage")'
{
julia --color=yes --threads=auto --check-bounds=yes --depwarn=yes --code-coverage=user --project=. -e 'import Pkg; Pkg.test(coverage=true)'
} || {
echo "Tests failed, but continuing anyway."
}
julia --color=yes --project=. coverage.jl
shell: bash
if: ${{ matrix.os == 'windows-latest' }}
- name: "Coveralls"
uses: coverallsapp/github-action@master
with:
Expand Down
8 changes: 5 additions & 3 deletions src/Configure.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,7 @@ function assert_operators_defined_over_reals(T, options::Options)
end
end
catch error
# TODO: Only raise error if domain error.
throw(
AssertionError(
"Your configuration is invalid - one of your operators ($cur_op) is not well-defined over the real line. You can get around this by returning `NaN` for invalid inputs.",
Expand Down Expand Up @@ -242,18 +243,19 @@ function test_entire_pipeline(procs, dataset::Dataset{T}, options::Options) wher
(options.verbosity > 0 || options.progress), "Testing entire pipeline on workers..."
)
for proc in procs
cdataset = copy_dataset(dataset)
push!(
futures,
@spawnat proc begin
tmp_pop = Population(
dataset;
cdataset;
npop=20,
nlength=3,
options=options,
nfeatures=dataset.nfeatures,
)
tmp_pop = s_r_cycle(
dataset,
cdataset,
tmp_pop,
5,
5,
Expand All @@ -263,7 +265,7 @@ function test_entire_pipeline(procs, dataset::Dataset{T}, options::Options) wher
record=RecordType(),
)[1]
tmp_pop = optimize_and_simplify_population(
dataset, tmp_pop, options, options.maxsize, RecordType()
cdataset, tmp_pop, options, options.maxsize, RecordType()
)
end
)
Expand Down
2 changes: 1 addition & 1 deletion src/Core.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@ include("Operators.jl")
include("Options.jl")

import .ProgramConstantsModule: MAX_DEGREE, BATCH_DIM, FEATURE_DIM, RecordType
import .DatasetModule: Dataset
import .DatasetModule: Dataset, copy_dataset
import .OptionsStructModule: Options, MutationWeights, sample_mutation
import .OptionsModule: Options
import .OperatorsModule:
Expand Down
14 changes: 14 additions & 0 deletions src/Dataset.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,4 +63,18 @@ function Dataset(
return Dataset{T}(X, y, n, nfeatures, weighted, weights, avg_y, baseline, varMap)
end

function copy_dataset(dataset::Dataset{T}) where {T<:Real}
return Dataset{T}(
copy(dataset.X),
copy(dataset.y),
dataset.n,
dataset.nfeatures,
dataset.weighted,
dataset.weighted ? copy(dataset.weights) : nothing,
dataset.avg_y,
dataset.baseline_loss,
copy(dataset.varMap),
)
end

end
3 changes: 1 addition & 2 deletions src/Population.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,7 @@ import Random: randperm
import DynamicExpressions: string_tree
import ..CoreModule: Options, Dataset, RecordType
import ..ComplexityModule: compute_complexity
import ..LossFunctionsModule: score_func, update_baseline_loss!
import ..LossFunctionsModule: score_func
import ..AdaptiveParsimonyModule: RunningSearchStatistics
import ..MutationFunctionsModule: gen_random_tree
import ..PopMemberModule: PopMember, copy_pop_member
Expand Down Expand Up @@ -60,7 +60,6 @@ function Population(
nfeatures::Int,
) where {T<:Real}
dataset = Dataset(X, y)
update_baseline_loss!(dataset, options)
return Population(dataset; npop=npop, options=options, nfeatures=nfeatures)
end

Expand Down
36 changes: 22 additions & 14 deletions src/SymbolicRegression.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -143,6 +143,7 @@ import .CoreModule:
FEATURE_DIM,
RecordType,
Dataset,
copy_dataset,
Options,
MutationWeights,
plus,
Expand Down Expand Up @@ -546,16 +547,18 @@ function _EquationSearch(

if saved_pop !== nothing && length(saved_pop.members) == options.npop
saved_pop::Population{T}
copy_saved_pop = copy_population(saved_pop)
new_pop = @sr_spawner parallelism worker_idx (
saved_pop, HallOfFame(options, T), RecordType(), 0.0
copy_saved_pop, HallOfFame(options, T), RecordType(), 0.0
)
else
if saved_pop !== nothing
@warn "Recreating population (output=$(j), population=$(i)), as the saved one doesn't have the correct number of members."
end
cdataset = copy_dataset(datasets[j])
new_pop = @sr_spawner parallelism worker_idx (
Population(
datasets[j];
cdataset;
npop=options.npop,
nlength=3,
options=options,
Expand Down Expand Up @@ -584,6 +587,8 @@ function _EquationSearch(

# TODO - why is this needed??
# Multi-threaded doesn't like to fetch within a new task:
copy_search_stats = deepcopy(running_search_statistics)
cdataset = copy_dataset(dataset)
updated_pop = @sr_spawner parallelism worker_idx let
in_pop = if parallelism in (:multiprocessing, :multithreading)
fetch(init_pops[j][i])[1]
Expand All @@ -596,26 +601,26 @@ function _EquationSearch(
"iteration0" => record_population(in_pop, options)
)
tmp_num_evals = 0.0
normalize_frequencies!(running_search_statistics)
normalize_frequencies!(copy_search_stats)
tmp_pop, tmp_best_seen, evals_from_cycle = s_r_cycle(
dataset,
cdataset,
in_pop,
options.ncycles_per_iteration,
curmaxsize,
running_search_statistics;
copy_search_stats;
verbosity=options.verbosity,
options=options,
record=cur_record,
)
tmp_num_evals += evals_from_cycle
tmp_pop, evals_from_optimize = optimize_and_simplify_population(
dataset, tmp_pop, options, curmaxsize, cur_record
cdataset, tmp_pop, options, curmaxsize, cur_record
)
tmp_num_evals += evals_from_optimize
if options.batching
for i_member in 1:(options.maxsize + MAX_DEGREE)
score, result_loss = score_func(
dataset, tmp_best_seen.members[i_member].tree, options
cdataset, tmp_best_seen.members[i_member].tree, options
)
tmp_best_seen.members[i_member].score = score
tmp_best_seen.members[i_member].loss = result_loss
Expand Down Expand Up @@ -787,26 +792,29 @@ function _EquationSearch(
iteration = find_iteration_from_record(key, record) + 1
end

copy_search_stats = deepcopy(all_running_search_statistics[j])
copy_cur_pop = copy_population(cur_pop)
cdataset = copy_dataset(dataset)
allPops[j][i] = @sr_spawner parallelism worker_idx let
cur_record = RecordType()
@recorder cur_record[key] = RecordType(
"iteration$(iteration)" => record_population(cur_pop, options)
"iteration$(iteration)" => record_population(copy_cur_pop, options)
)
tmp_num_evals = 0.0
normalize_frequencies!(all_running_search_statistics[j])
normalize_frequencies!(copy_search_stats)
tmp_pop, tmp_best_seen, evals_from_cycle = s_r_cycle(
dataset,
cur_pop,
cdataset,
copy_cur_pop,
options.ncycles_per_iteration,
curmaxsize,
all_running_search_statistics[j];
copy_search_stats;
verbosity=options.verbosity,
options=options,
record=cur_record,
)
tmp_num_evals += evals_from_cycle
tmp_pop, evals_from_optimize = optimize_and_simplify_population(
dataset, tmp_pop, options, curmaxsize, cur_record
cdataset, tmp_pop, options, curmaxsize, cur_record
)
tmp_num_evals += evals_from_optimize

Expand All @@ -815,7 +823,7 @@ function _EquationSearch(
for i_member in 1:(options.maxsize + MAX_DEGREE)
if tmp_best_seen.exists[i_member]
score, result_loss = score_func(
dataset, tmp_best_seen.members[i_member].tree, options
cdataset, tmp_best_seen.members[i_member].tree, options
)
tmp_best_seen.members[i_member].score = score
tmp_best_seen.members[i_member].loss = result_loss
Expand Down