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

add --count and --batch args for data_export.py #522

Draft
wants to merge 2 commits into
base: main
Choose a base branch
from
Draft

add --count and --batch args for data_export.py #522

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
19d9f8c
fix: convert X to numpy.array in glove dataset function
jiangyinzuo May 17, 2024
3982de0
add `--count` and `--batch` args for data_export.py
jiangyinzuo May 17, 2024
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
2 changes: 1 addition & 1 deletion ann_benchmarks/datasets.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -213,7 +213,7 @@ def glove(out_fn: str, d: int) -> None:
for line in z.open(z_fn):
v = [float(x) for x in line.strip().split()[1:]]
X.append(numpy.array(v))
X_train, X_test = train_test_split(X)
X_train, X_test = train_test_split(numpy.array(X))
write_output(numpy.array(X_train), numpy.array(X_test), out_fn, "angular")


Expand Down
4 changes: 2 additions & 2 deletions ann_benchmarks/results.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -84,7 +84,7 @@ def load_all_results(dataset: Optional[str] = None,
Yields:
tuple: A tuple containing properties as a dictionary and an h5py file object.
"""
for root, _, files in os.walk(build_result_filepath(dataset, count)):
for root, _, files in os.walk(build_result_filepath(dataset, count, batch_mode=batch_mode)):
Copy link
Author

@jiangyinzuo jiangyinzuo May 21, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If data_export.py is expected to export all the results that are found, then here build_result_filepath(...) should return all the sub-directories.

However, in def build_result_filepath(...):

ann-benchmarks/ann_benchmarks/results.py

Lines 11 to 38 in 3982de0

def build_result_filepath(dataset_name: Optional[str] = None,
count: Optional[int] = None,
definition: Optional[Definition] = None,
query_arguments: Optional[Any] = None,
batch_mode: bool = False) -> str:
"""
Constructs the filepath for storing the results.
Args:
dataset_name (str, optional): The name of the dataset.
count (int, optional): The count of records.
definition (Definition, optional): The definition of the algorithm.
query_arguments (Any, optional): Additional arguments for the query.
batch_mode (bool, optional): If True, the batch mode is activated.
Returns:
str: The constructed filepath.
"""
d = ["results"]
if dataset_name:
d.append(dataset_name)
if count:
d.append(str(count))
if definition:
d.append(definition.algorithm + ("-batch" if batch_mode else ""))
data = definition.arguments + query_arguments
d.append(re.sub(r"\W+", "_", json.dumps(data, sort_keys=True)).strip("_") + ".hdf5")
return os.path.join(*d)

If count and batch_mode are None, directories of specific counts and batch mode will be ignored.

So if we expect data_export.py to export all the results, including every count argument, batch mode, and non-batch mode, maybe we should implement a new function for def load_all_results(...)?

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good point. I think the problem is that

ann-benchmarks/ann_benchmarks/results.py

Lines 90 to 97 in 3982de0

continue
try:
with h5py.File(os.path.join(root, filename), "r+") as f:
properties = dict(f.attrs)
if batch_mode != properties["batch_mode"]:
continue
yield properties, f
except Exception:

doesn't differentiate between batch_mode being True or False, or batch_mode being None.

Since definition is None in the part above, it should try to load both the batch/non-batch/different count values.

for filename in files:
if os.path.splitext(filename)[-1] != ".hdf5":
continue
Expand All @@ -110,4 +110,4 @@ def get_unique_algorithms() -> Set[str]:
for batch_mode in [False, True]:
for properties, _ in load_all_results(batch_mode=batch_mode):
algorithms.add(properties["algo"])
return algorithms
return algorithms
11 changes: 9 additions & 2 deletions data_export.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,14 +9,21 @@
parser = argparse.ArgumentParser()
parser.add_argument("--output", help="Path to the output file", required=True)
parser.add_argument("--recompute", action="store_true", help="Recompute metrics")
parser.add_argument(
"-k", "--count", default=10, type=int, help="The number of near neighbours to search for"
)
parser.add_argument("--batch", action="store_true", help="Batch mode")
args = parser.parse_args()

datasets = DATASETS.keys()
dfs = []
for dataset_name in datasets:
print("Looking at dataset", dataset_name)
if len(list(load_all_results(dataset_name))) > 0:
results = load_all_results(dataset_name)
if len(list(load_all_results(dataset_name,
count=args.count,
batch_mode=args.batch
))) > 0:
results = load_all_results(dataset_name, count=args.count, batch_mode=args.batch)
dataset, _ = get_dataset(dataset_name)
results = compute_metrics_all_runs(dataset, results, args.recompute)
for res in results:
Expand Down
Loading