revert changes

python/cugraph/cugraph/tests/link_analysis/test_pagerank_mg.py

@@ -22,9 +22,7 @@
 import dask_cudf
 from cugraph.testing import utils
 from cugraph.dask.common.mg_utils import is_single_gpu
-import cugraph.dask.comms.comms as Comms
 from cugraph.testing.utils import RAPIDS_DATASET_ROOT_DIR_PATH
-from cudf.testing.testing import assert_frame_equal
 
 
 # The function selects personalization_perc% of accessible vertices in graph M
@@ -274,108 +272,3 @@ def test_pagerank_non_convergence(dask_client):
     assert type(df) is dask_cudf.DataFrame
     assert type(converged) is bool
     assert converged is True
-
-
-IS_DIRECTED = [True]
-HAS_PRECOMPUTED = [False, True]
-HAS_GUESS = [True]
-
-
-@pytest.mark.skipif(is_single_gpu(), reason="skipping MG testing on Single GPU system")
-@pytest.mark.parametrize("personalization_perc", PERSONALIZATION_PERC[1:])
-@pytest.mark.parametrize("directed", IS_DIRECTED)
-@pytest.mark.parametrize("has_precomputed_vertex_out_weight", HAS_PRECOMPUTED)
-@pytest.mark.parametrize("has_guess", HAS_GUESS)
-def test_dask_batch_pagerank(
-    dask_client,
-    personalization_perc,
-    directed,
-    has_precomputed_vertex_out_weight,
-    has_guess,
-):
-
-    input_data_path = (RAPIDS_DATASET_ROOT_DIR_PATH / "karate.csv").as_posix()
-    print(f"dataset={input_data_path}")
-    # chunksize = dcg.get_chunksize(input_data_path)
-
-    # FIXME. Check with dask_cudf dataframe too
-    """
-    ddf = dask_cudf.read_csv(
-        input_data_path,
-        chunksize=chunksize,
-        delimiter=" ",
-        names=["src", "dst", "value"],
-        dtype=["int32", "int32", "float32"],
-    )
-    """
-
-    df = cudf.read_csv(
-        input_data_path,
-        delimiter=" ",
-        names=["src", "dst", "value"],
-        dtype=["int32", "int32", "float32"],
-    )
-
-    g = cugraph.Graph(directed=directed)
-    # df = ddf.compute().reset_index(drop=True)
-    g.from_cudf_edgelist(df, "src", "dst", "value")
-
-    parallel_g = cugraph.Graph(directed=directed, parallel=True)
-    parallel_g.from_cudf_edgelist(df, "src", "dst", "value")
-
-    personalization = None
-    pre_vtx_o_wgt = None
-    nstart = None
-    max_iter = 100
-    if personalization_perc != 0:
-        personalization, p = personalize(g.nodes(), personalization_perc)
-
-    if has_precomputed_vertex_out_weight == 1:
-        df = df[["src", "value"]]
-        pre_vtx_o_wgt = (
-            df.groupby(["src"], as_index=False)
-            .sum()
-            .rename(columns={"src": "vertex", "value": "sums"})
-        )
-
-    if has_guess == 1:
-        nstart = cugraph.pagerank(g, personalization=personalization, tol=1e-6).rename(
-            columns={"pagerank": "values"}
-        )
-        # FIXME: Why is 'batch' not converging in 20 iterations
-        # max_iter = 20
-
-    personalization_ddf = dask_cudf.from_cudf(
-        personalization, npartitions=len(Comms.get_workers())
-    )
-
-    batch_pr = cugraph.pagerank(
-        parallel_g,
-        personalization=personalization_ddf,
-        precomputed_vertex_out_weight=pre_vtx_o_wgt,
-        max_iter=max_iter,
-        tol=1e-6,
-        nstart=nstart,
-    )
-
-    for i in range(batch_pr.npartitions):
-
-        # print("batch_pr = \n", batch_pr.get_partition(i).compute())
-
-        personalization = (
-            personalization_ddf.get_partition(i).compute().reset_index(drop=True)
-        )
-        expected_pr = cugraph.pagerank(
-            g,
-            personalization=personalization,
-            precomputed_vertex_out_weight=pre_vtx_o_wgt,
-            max_iter=max_iter,
-            tol=1e-6,
-            nstart=nstart,
-        )
-
-        result_pr = batch_pr.get_partition(i).compute().reset_index(drop=True)
-
-        assert_frame_equal(result_pr, expected_pr, check_dtype=False, check_like=True)
-
-        print("pr = \n", expected_pr)