Skip to content

Commit

Permalink
create all models at the same time
Browse files Browse the repository at this point in the history
  • Loading branch information
@lquenti
lquenti committed Mar 1, 2024
1 parent 2aa9789 commit d9fd9be
Show file tree
Hide file tree
Showing 2 changed files with 53 additions and 130 deletions.
181 changes: 52 additions & 129 deletions blackheap/assets/build_models.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -258,7 +258,12 @@ def find_significant_clusters_derivative(
return significant_clusters, biggest_cluster


def find_clusters_meanshift(b: Benchmark, num_points=1000, cutoff_threshold_ratio=0.05):
def find_clusters_meanshift(
b: Benchmark,
num_points=1000,
significant_percentage=0.1,
cutoff_threshold_ratio=0.05,
):
xs = np.linspace(b.min_val, b.max_val, num_points)
ys = b.kde(xs)
xsys = [list(pair) for pair in zip(xs, ys)]
Expand Down Expand Up @@ -337,73 +342,21 @@ def measurements_to_model(
)


def measurements_to_linear_model_derivative(
measurements: Measurements,
num_points=1000,
significant_percentage=0.1,
cutoff_threshold_ratio=0.05,
use_biggest=False,
):
return measurements_to_model(
measurements,
use_derivative=True,
use_linear=True,
num_points=num_points,
significant_percentage=significant_percentage,
cutoff_threshold_ratio=cutoff_threshold_ratio,
use_biggest=use_biggest,
)


def measurements_to_constlinear_model_derivative(
measurements: Measurements,
num_points=1000,
significant_percentage=0.1,
cutoff_threshold_ratio=0.05,
use_biggest=False,
):
return measurements_to_model(
measurements,
use_derivative=True,
use_linear=False,
num_points=num_points,
significant_percentage=significant_percentage,
cutoff_threshold_ratio=cutoff_threshold_ratio,
use_biggest=use_biggest,
)


def measurements_to_linear_model_meanshift(
measurements: Measurements,
num_points=1000,
cutoff_threshold_ratio=0.05,
use_biggest=False,
):
return measurements_to_model(
measurements,
use_derivative=False,
use_linear=True,
num_points=num_points,
cutoff_threshold_ratio=cutoff_threshold_ratio,
use_biggest=use_biggest,
)


def measurements_to_constlinear_model_meanshift(
measurements: Measurements,
num_points=1000,
cutoff_threshold_ratio=0.05,
use_biggest=False,
):
return measurements_to_model(
measurements,
use_derivative=False,
use_linear=False,
num_points=num_points,
cutoff_threshold_ratio=cutoff_threshold_ratio,
use_biggest=use_biggest,
)
def create_all_models_from_measurements(all_measurements, measurements_to_model_f):
all_models = []
for read, write in all_measurements:
print(f"Processing: {read.name}")
all_models.append(measurements_to_model_f(read))
all_models.append(measurements_to_model_f(write))

# If we had constlinear, it is a list of tuples, so we have to flatten it down
all_models_flattened = []
for item in all_models:
if isinstance(item, tuple):
all_models_flattened.extend(item)
else:
all_models_flattened.append(item)
return all_models_flattened

if __name__ == "__main__":
print("Parsing arguments...")
Expand All @@ -412,21 +365,6 @@ def measurements_to_constlinear_model_meanshift(
parser = argparse.ArgumentParser(
description="Creates a model based on the measurements done by blackheap"
)
parser.add_argument(
"--constlinear",
action="store_true",
help="If set, it will generate a constlinear model instead of a linear one",
)
parser.add_argument(
"--meanshift",
action="store_true",
help="If set, it will use the Meanshift based clustering algorithm instead of the derivative based one. See the jupyter notebook for more explaination.",
)
parser.add_argument(
"--usebiggest",
action="store_true",
help="If set, it will use the right bound of the biggest significant cluster instead of the last one",
)
parser.add_argument(
"--cluster-significance",
type=float,
Expand All @@ -441,57 +379,42 @@ def measurements_to_constlinear_model_meanshift(
)
args = parser.parse_args()

def pick_cluster_function_based_on_cli(args):
if args.constlinear and args.meanshift:
return lambda m: measurements_to_constlinear_model_meanshift(
m,
cutoff_threshold_ratio=args.last_cluster_threshold,
use_biggest=args.usebiggest,
)
if args.constlinear and not args.meanshift:
return lambda m: measurements_to_constlinear_model_derivative(
m,
significant_percentage=args.cluster_significance,
cutoff_threshold_ratio=args.last_cluster_threshold,
use_biggest=args.usebiggest,
)
if not args.constlinear and args.meanshift:
return lambda m: measurements_to_linear_model_meanshift(
m,
cutoff_threshold_ratio=args.last_cluster_threshold,
use_biggest=args.usebiggest,
)
if not args.constlinear and not args.meanshift:
return lambda m: measurements_to_linear_model_derivative(
m,
significant_percentage=args.cluster_significance,
cutoff_threshold_ratio=args.last_cluster_threshold,
use_biggest=args.usebiggest,
)
assert False

measurements_to_model_f = pick_cluster_function_based_on_cli(args)

print("Loading all measurements in")
script_dir = os.path.dirname(os.path.realpath(__file__))
all_measurements = [
load_benchmark_folder(x) for x in get_benchmark_dirs(script_dir)
]

all_models = []
for read, write in all_measurements:
print(f"Processing: {read.name}")
all_models.append(measurements_to_model_f(read))
all_models.append(measurements_to_model_f(write))

# If we had constlinear, it is a list of tuples, so we have to flatten it down
all_models_flattened = []
for item in all_models:
if isinstance(item, tuple):
all_models_flattened.extend(item)
else:
all_models_flattened.append(item)
output_folder = "./models"
if os.path.isdir(output_folder):
os.rmdir(output_folder)
os.makedirs(output_folder, exist_ok=True)

"""
All dimensions to choose of:
- linear/constlinear
- derivative/meanshift
- uselast/usebiggest
"""
for use_linear in [False, True]:
linear_str = "linear" if use_linear else "constlinear"
for use_derivative in [False, True]:
cluster_algo_str = "derivative" if use_derivative else "meanshift"
for use_biggest in [False, True]:
cluster_choose_str = "biggest" if use_biggest else "last"
print("Creating:", linear_str, cluster_algo_str, cluster_choose_str)
m2mf = lambda m: measurements_to_model(
m,
use_linear=use_linear,
use_derivative=use_derivative,
use_biggest=use_biggest,
num_points=1000,
significant_percentage=args.cluster_significance,
cutoff_threshold_ratio=args.last_cluster_threshold
)
all_models_flattened = create_all_models_from_measurements(all_measurements, m2mf)

file_str = f"{output_folder}/model_{linear_str}_type_{cluster_algo_str}_algo_{use_biggest}_selection.csv"
with open(file_str, "w") as fp:
fp.write(Model.all_to_csv(all_models_flattened))

with open(f"model.csv", "w") as fp:
fp.write(Model.all_to_csv(all_models_flattened))
print(f"Model successfully saved as model.csv")
2 changes: 1 addition & 1 deletion blackheap/src/assets/mod.rs
View file
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@ use lazy_static::lazy_static;
pub mod progress;

const JUPYTER_NOTEBOOK: &[u8; 7528] = include_bytes!("../../assets/AnalysisTool.ipynb");
const BUILD_MODELS: &[u8; 16151] = include_bytes!("../../assets/build_models.py");
const BUILD_MODELS: &[u8; 13897] = include_bytes!("../../assets/build_models.py");
const GITIGNORE: &[u8; 3079] = include_bytes!("../../assets/.gitignore");
const README: &[u8; 27] = include_bytes!("../../assets/README.md");
const REQUIREMENTS: &[u8; 47] = include_bytes!("../../assets/requirements.txt");
Expand Down

0 comments on commit d9fd9be

Please sign in to comment.