FAI-863: XAI-Bench (#109)

* submodule

* linting round 1

Author: RobGeada

.gitmodules

@@ -0,0 +1,3 @@
+[submodule "tests/benchmarks/trustyai_xai_bench"]
+	path = tests/benchmarks/trustyai_xai_bench
+	url = https://github.com/trustyai-explainability/trustyai_xai_bench

src/trustyai/explainers.py

@@ -1,5 +1,6 @@
 """Explainers module"""
-# pylint: disable = import-error, too-few-public-methods, wrong-import-order, line-too-long
+# pylint: disable = import-error, too-few-public-methods, wrong-import-order, line-too-long,
+# pylint: disable = unused-argument
 from typing import Dict, Optional, List, Union
 import matplotlib.pyplot as plt
 import matplotlib as mpl
@@ -425,6 +426,8 @@ def __init__(
         penalise_sparse_balance=True,
         track_counterfactuals=False,
         normalise_weights=False,
+        use_wlr_model=True,
+        **kwargs
     ):
         """Initialize the :class:`LimeExplainer`.
 
@@ -454,6 +457,7 @@ def __init__(
             .withEncodingParams(EncodingParams(0.07, 0.3))
             .withAdaptiveVariance(True)
             .withPenalizeBalanceSparse(penalise_sparse_balance)
+            .withUseWLRLinearModel(use_wlr_model)
             .withTrackCounterfactuals(track_counterfactuals)
         )
 
@@ -896,6 +900,7 @@ def __init__(
         seed=0,
         link_type: Optional[_ShapConfig.LinkType] = None,
         track_counterfactuals=False,
+        **kwargs,
     ):
         r"""Initialize the :class:`SHAPxplainer`.
 

tests/benchmarks/trustyai_xai_bench

@@ -0,0 +1,23 @@
+import pytest
+from trustyai_xai_bench import run_benchmark_config
+
+
+@pytest.mark.benchmark(group="xai_bench", min_rounds=1, warmup=False)
+def test_level_0(benchmark):
+    # ~4.5 min
+    result = benchmark(run_benchmark_config, 0)
+    benchmark.extra_info['runs'] = result.to_dict('records')
+
+
+@pytest.mark.skip(reason="full diagnostic benchmark, ~2 hour runtime")
+@pytest.mark.benchmark(group="xai_bench", min_rounds=1, warmup=False)
+def test_level_1(benchmark):
+    result = benchmark(run_benchmark_config, 1)
+    benchmark.extra_info['runs'] = result.to_dict('records')
+
+
+@pytest.mark.skip(reason="very thorough benchmark, >>2 hour runtime")
+@pytest.mark.benchmark(group="xai_bench", min_rounds=1, warmup=False)
+def test_level_2(benchmark):
+    result = benchmark(run_benchmark_config, 2)
+    benchmark.extra_info['runs'] = result.to_dict('records')