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

Jump to bottom

Move rag context correctness metrics tests #1092

Merged
merged 4 commits into from
Aug 4, 2024
Merged

Move rag context correctness metrics tests #1092

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
c8190aa
move rag context correctness metrics tests to test_metrics.py
matanor Jul 31, 2024
38f02e4
Merge branch 'main' into move_context_correctness_tests
yoavkatz Jul 31, 2024
938aba5
Merge branch 'main' into move_context_correctness_tests
matanor Aug 1, 2024
6411343
Merge branch 'main' into move_context_correctness_tests
elronbandel Aug 4, 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
173 changes: 1 addition & 172 deletions prepare/metrics/rag_context_correctness.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
from unitxt import add_to_catalog
from unitxt.collections_operators import Wrap
from unitxt.metrics import MetricPipeline
from unitxt.operators import Copy, RenameFields
from unitxt.operators import Copy

for metric_name, catalog_name in [
("map", "metrics.rag.map"),
Expand All @@ -20,174 +20,3 @@
metric=f"metrics.{metric_name}",
)
add_to_catalog(metric, catalog_name, overwrite=True)


if __name__ == "__main__":
from unitxt.test_utils.metrics import test_evaluate, test_metric

task_data = [
{ # MRR is 1, MAP is (1 + 2/3)/2 = 0.833
"context_ids": ["A", "B", "C"],
"ground_truths_context_ids": ["A", "C"],
},
{ # MRR and MAP are both 0.5
"context_ids": ["A", "B"],
"ground_truths_context_ids": ["B"],
},
]

map_instance_targets = [
{"map": 0.83, "score": 0.83, "score_name": "map"},
{"map": 0.5, "score": 0.5, "score_name": "map"},
]
mrr_instance_targets = [
{"mrr": 1.0, "score": 1.0, "score_name": "mrr"},
{"mrr": 0.5, "score": 0.5, "score_name": "mrr"},
]
retrieval_at_k_instance_targets = [
{
"match_at_1": 1.0,
"match_at_3": 1.0,
"match_at_5": 1.0,
"match_at_10": 1.0,
"match_at_20": 1.0,
"match_at_40": 1.0,
"precision_at_1": 1.0,
"precision_at_3": 0.67,
"precision_at_5": 0.67,
"precision_at_10": 0.67,
"precision_at_20": 0.67,
"precision_at_40": 0.67,
"recall_at_1": 0.5,
"recall_at_3": 1.0,
"recall_at_5": 1.0,
"recall_at_10": 1.0,
"recall_at_20": 1.0,
"recall_at_40": 1.0,
"score": 1.0,
"score_name": "match_at_1",
},
{
"match_at_1": 0.0,
"match_at_10": 1.0,
"match_at_20": 1.0,
"match_at_3": 1.0,
"match_at_40": 1.0,
"match_at_5": 1.0,
"precision_at_1": 0.0,
"precision_at_10": 0.5,
"precision_at_20": 0.5,
"precision_at_3": 0.5,
"precision_at_40": 0.5,
"precision_at_5": 0.5,
"recall_at_1": 0.0,
"recall_at_10": 1.0,
"recall_at_20": 1.0,
"recall_at_3": 1.0,
"recall_at_40": 1.0,
"recall_at_5": 1.0,
"score": 0.0,
"score_name": "match_at_1",
},
]

map_global_target = {
"map": 0.67,
"map_ci_high": 0.83,
"map_ci_low": 0.5,
"score": 0.67,
"score_ci_high": 0.83,
"score_ci_low": 0.5,
"score_name": "map",
}
mrr_global_target = {
"mrr": 0.75,
"mrr_ci_high": 1.0,
"mrr_ci_low": 0.5,
"score": 0.75,
"score_ci_high": 1.0,
"score_ci_low": 0.5,
"score_name": "mrr",
}
retrieval_at_k_global_target = {
"match_at_1": 0.5,
"match_at_1_ci_high": 1.0,
"match_at_1_ci_low": 0.0,
"match_at_3": 1.0,
"match_at_5": 1.0,
"match_at_10": 1.0,
"match_at_20": 1.0,
"match_at_40": 1.0,
"precision_at_1": 0.5,
"precision_at_1_ci_high": 1.0,
"precision_at_1_ci_low": 0.0,
"precision_at_3": 0.58,
"precision_at_3_ci_high": 0.67,
"precision_at_3_ci_low": 0.5,
"precision_at_5": 0.58,
"precision_at_5_ci_high": 0.67,
"precision_at_5_ci_low": 0.5,
"precision_at_10": 0.58,
"precision_at_10_ci_high": 0.67,
"precision_at_10_ci_low": 0.5,
"precision_at_20": 0.58,
"precision_at_20_ci_high": 0.67,
"precision_at_20_ci_low": 0.5,
"precision_at_40": 0.58,
"precision_at_40_ci_high": 0.67,
"precision_at_40_ci_low": 0.5,
"recall_at_1": 0.25,
"recall_at_1_ci_high": 0.5,
"recall_at_1_ci_low": 0.0,
"recall_at_3": 1.0,
"recall_at_5": 1.0,
"recall_at_10": 1.0,
"recall_at_20": 1.0,
"recall_at_40": 1.0,
"score": 0.5,
"score_ci_high": 1.0,
"score_ci_low": 0.0,
"score_name": "match_at_1",
}

for catalog_name, global_target, instance_targets in [
("metrics.rag.map", map_global_target, map_instance_targets),
("metrics.rag.mrr", mrr_global_target, mrr_instance_targets),
("metrics.rag.context_correctness", mrr_global_target, mrr_instance_targets),
(
"metrics.rag.retrieval_at_k",
retrieval_at_k_global_target,
retrieval_at_k_instance_targets,
),
]:
# test the evaluate call
test_evaluate(
global_target,
instance_targets=[
{"score": instance["score"]} for instance in instance_targets
],
task_data=task_data,
metric_name=catalog_name,
)

# test using the usual metric pipeline
test_pipeline = MetricPipeline(
main_score="score",
preprocess_steps=[
RenameFields(field_to_field={"task_data/context_ids": "context_ids"}),
RenameFields(
field_to_field={
"task_data/ground_truths_context_ids": "ground_truths_context_ids"
}
),
],
metric=f"{catalog_name}",
)
test_metric(
metric=test_pipeline,
predictions=[None, None],
references=[[], []],
instance_targets=instance_targets,
global_target=global_target,
task_data=task_data,
)
182 changes: 181 additions & 1 deletion tests/library/test_metrics.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,6 +43,7 @@
KendallTauMetric,
LlamaIndexCorrectness,
MaxAccuracy,
MetricPipeline,
MetricsEnsemble,
NormalizedSacrebleu,
Perplexity,
Expand All @@ -53,7 +54,13 @@
TokenOverlap,
UnsortedListExactMatch,
)
from unitxt.test_utils.metrics import apply_metric, check_scores, test_metric
from unitxt.operators import RenameFields
from unitxt.test_utils.metrics import (
apply_metric,
check_scores,
test_evaluate,
test_metric,
)

from tests.utils import UnitxtTestCase

Expand Down Expand Up @@ -1762,3 +1769,176 @@ def test_metrics_ensemble(self):
instance_targets=instance_targets,
global_target=global_target,
)

def text_context_correctness(self):
task_data = [
{ # MRR is 1, MAP is (1 + 2/3)/2 = 0.833
"context_ids": ["A", "B", "C"],
"ground_truths_context_ids": ["A", "C"],
},
{ # MRR and MAP are both 0.5
"context_ids": ["A", "B"],
"ground_truths_context_ids": ["B"],
},
]

map_instance_targets = [
{"map": 0.83, "score": 0.83, "score_name": "map"},
{"map": 0.5, "score": 0.5, "score_name": "map"},
]
mrr_instance_targets = [
{"mrr": 1.0, "score": 1.0, "score_name": "mrr"},
{"mrr": 0.5, "score": 0.5, "score_name": "mrr"},
]
retrieval_at_k_instance_targets = [
{
"match_at_1": 1.0,
"match_at_3": 1.0,
"match_at_5": 1.0,
"match_at_10": 1.0,
"match_at_20": 1.0,
"match_at_40": 1.0,
"precision_at_1": 1.0,
"precision_at_3": 0.67,
"precision_at_5": 0.67,
"precision_at_10": 0.67,
"precision_at_20": 0.67,
"precision_at_40": 0.67,
"recall_at_1": 0.5,
"recall_at_3": 1.0,
"recall_at_5": 1.0,
"recall_at_10": 1.0,
"recall_at_20": 1.0,
"recall_at_40": 1.0,
"score": 1.0,
"score_name": "match_at_1",
},
{
"match_at_1": 0.0,
"match_at_10": 1.0,
"match_at_20": 1.0,
"match_at_3": 1.0,
"match_at_40": 1.0,
"match_at_5": 1.0,
"precision_at_1": 0.0,
"precision_at_10": 0.5,
"precision_at_20": 0.5,
"precision_at_3": 0.5,
"precision_at_40": 0.5,
"precision_at_5": 0.5,
"recall_at_1": 0.0,
"recall_at_10": 1.0,
"recall_at_20": 1.0,
"recall_at_3": 1.0,
"recall_at_40": 1.0,
"recall_at_5": 1.0,
"score": 0.0,
"score_name": "match_at_1",
},
]
map_global_target = {
"map": 0.67,
"map_ci_high": 0.83,
"map_ci_low": 0.5,
"score": 0.67,
"score_ci_high": 0.83,
"score_ci_low": 0.5,
"score_name": "map",
}
mrr_global_target = {
"mrr": 0.75,
"mrr_ci_high": 1.0,
"mrr_ci_low": 0.5,
"score": 0.75,
"score_ci_high": 1.0,
"score_ci_low": 0.5,
"score_name": "mrr",
}
retrieval_at_k_global_target = {
"match_at_1": 0.5,
"match_at_1_ci_high": 1.0,
"match_at_1_ci_low": 0.0,
"match_at_3": 1.0,
"match_at_5": 1.0,
"match_at_10": 1.0,
"match_at_20": 1.0,
"match_at_40": 1.0,
"precision_at_1": 0.5,
"precision_at_1_ci_high": 1.0,
"precision_at_1_ci_low": 0.0,
"precision_at_3": 0.58,
"precision_at_3_ci_high": 0.67,
"precision_at_3_ci_low": 0.5,
"precision_at_5": 0.58,
"precision_at_5_ci_high": 0.67,
"precision_at_5_ci_low": 0.5,
"precision_at_10": 0.58,
"precision_at_10_ci_high": 0.67,
"precision_at_10_ci_low": 0.5,
"precision_at_20": 0.58,
"precision_at_20_ci_high": 0.67,
"precision_at_20_ci_low": 0.5,
"precision_at_40": 0.58,
"precision_at_40_ci_high": 0.67,
"precision_at_40_ci_low": 0.5,
"recall_at_1": 0.25,
"recall_at_1_ci_high": 0.5,
"recall_at_1_ci_low": 0.0,
"recall_at_3": 1.0,
"recall_at_5": 1.0,
"recall_at_10": 1.0,
"recall_at_20": 1.0,
"recall_at_40": 1.0,
"score": 0.5,
"score_ci_high": 1.0,
"score_ci_low": 0.0,
"score_name": "match_at_1",
}

for catalog_name, global_target, instance_targets in [
("metrics.rag.map", map_global_target, map_instance_targets),
("metrics.rag.mrr", mrr_global_target, mrr_instance_targets),
(
"metrics.rag.context_correctness",
mrr_global_target,
mrr_instance_targets,
),
(
"metrics.rag.retrieval_at_k",
retrieval_at_k_global_target,
retrieval_at_k_instance_targets,
),
]:
# test the evaluate call
test_evaluate(
global_target,
instance_targets=[
{"score": instance["score"]} for instance in instance_targets
],
task_data=task_data,
metric_name=catalog_name,
)

# test using the usual metric pipeline
test_pipeline = MetricPipeline(
main_score="score",
preprocess_steps=[
RenameFields(
field_to_field={"task_data/context_ids": "context_ids"}
),
RenameFields(
field_to_field={
"task_data/ground_truths_context_ids": "ground_truths_context_ids"
}
),
],
metric=f"{catalog_name}",
)
test_metric(
metric=test_pipeline,
predictions=[None, None],
references=[[], []],
instance_targets=instance_targets,
global_target=global_target,
task_data=task_data,
)
Loading