Fix or silence warnings emitted during tests

examples/llm/vdb_upload/module/content_extractor_module.py

@@ -29,7 +29,7 @@
 import pandas as pd
 from docx import Document
 from langchain.text_splitter import RecursiveCharacterTextSplitter
-from pydantic import BaseModel
+from pydantic import BaseModel  # pylint: disable=no-name-in-module
 from pydantic import Field
 from pydantic import ValidationError
 from pydantic import validator
@@ -197,7 +197,7 @@ def _csv_to_text_converter(input_info: ConverterInputInfo) -> list[str]:
         raise ValueError("The CSV file must either include a 'content' column or have a "
                          "columns specified in the meta configuration with key 'text_column_names'.")
     df.fillna(value='', inplace=True)
-    text_arr = df[text_column_names].apply(lambda x: ' '.join(map(str, x)), axis=1).tolist()
+    text_arr = df[sorted(text_column_names)].apply(lambda x: ' '.join(map(str, x)), axis=1).tolist()
     return text_arr
 
 

examples/llm/vdb_upload/run.py

@@ -67,7 +67,7 @@ def run():
 )
 @click.option(
     "--model_max_batch_size",
-    default=64,
+    default=256,
     type=click.IntRange(min=1),
     help="Max batch size to use for the model",
 )

morpheus/_lib/src/objects/table_info.cpp

@@ -306,7 +306,8 @@ std::optional<std::string> MutableTableInfo::ensure_sliceable_index()
         auto df_index = py_df.attr("index");
 
         // Check to see if we actually need the change
-        if (df_index.attr("is_unique").cast<bool>() && df_index.attr("is_monotonic").cast<bool>())
+        if (df_index.attr("is_unique").cast<bool>() && (df_index.attr("is_monotonic_increasing").cast<bool>() ||
+                                                        df_index.attr("is_monotonic_decreasing").cast<bool>()))
         {
             // Set the outputname to nullopt
             old_index_col_name = std::nullopt;

morpheus/controllers/rss_controller.py

@@ -156,7 +156,7 @@ def _read_file_content(self, file_path: str) -> str:
 
     def _try_parse_feed_with_beautiful_soup(self, feed_input: str) -> "feedparser.FeedParserDict":
 
-        soup = BeautifulSoup(feed_input, 'lxml')
+        soup = BeautifulSoup(feed_input, features='xml')
 
         # Verify whether the given feed has 'item' or 'entry' tags.
         if soup.find('item'):

morpheus/service/vdb/milvus_vector_db_service.py

@@ -698,7 +698,7 @@ def _build_schema_conf(self, df: typing.Union[cudf.DataFrame, pd.DataFrame]) ->
             df = df.to_pandas()
 
         # Loop over all of the columns of the first row and build the schema
-        for col_name, col_val in df.iloc[0].iteritems():
+        for col_name, col_val in df.iloc[0].items():
 
             field_dict = {
                 "name": col_name,

morpheus/stages/general/monitor_stage.py

@@ -99,7 +99,7 @@ def accepted_types(self) -> typing.Tuple:
     def supports_cpp_node(self):
         return False
 
-    def on_start(self):
+    async def start_async(self):
         """
         Starts the pipeline stage's progress bar.
         """

morpheus/stages/output/http_server_sink_stage.py

@@ -142,7 +142,7 @@ def supports_cpp_node(self):
         """Indicates whether or not this stage supports a C++ node."""
         return False
 
-    def on_start(self):
+    async def start_async(self):
         """Starts the HTTP server."""
         from morpheus.common import HttpServer
         self._server = HttpServer(parse_fn=self._request_handler,

morpheus/stages/preprocess/preprocess_nlp_stage.py

@@ -16,6 +16,7 @@
 import json
 import logging
 import typing
+import warnings
 from functools import partial
 
 import cupy as cp
@@ -227,13 +228,18 @@ def process_multi_message(message: MultiMessage,
         # Existing logic for MultiMessage
         text_ser = cudf.Series(message.get_meta(column))
 
-        tokenized = tokenize_text_series(vocab_hash_file=vocab_hash_file,
-                                         do_lower_case=do_lower_case,
-                                         text_ser=text_ser,
-                                         seq_len=seq_len,
-                                         stride=stride,
-                                         truncation=truncation,
-                                         add_special_tokens=add_special_tokens)
+        with warnings.catch_warnings():
+            warnings.filterwarnings(
+                "ignore",
+                message="When truncation is not True, the behavior currently differs from HuggingFace.*",
+                category=UserWarning)
+            tokenized = tokenize_text_series(vocab_hash_file=vocab_hash_file,
+                                             do_lower_case=do_lower_case,
+                                             text_ser=text_ser,
+                                             seq_len=seq_len,
+                                             stride=stride,
+                                             truncation=truncation,
+                                             add_special_tokens=add_special_tokens)
         del text_ser
 
         seg_ids = tokenized.segment_ids

morpheus/utils/column_info.py

@@ -387,7 +387,14 @@ def _process_column(self, df: pd.DataFrame) -> pd.Series:
             The processed column as a datetime Series.
         """
 
-        return pd.to_datetime(df[self.input_name], infer_datetime_format=True, utc=True).astype(self.get_pandas_dtype())
+        dt_series = pd.to_datetime(df[self.input_name], infer_datetime_format=True, utc=True)
+
+        dtype = self.get_pandas_dtype()
+        if dtype == 'datetime64[ns]':
+            # avoid deprecation warning about using .astype to convert from a tz-aware type to a tz-naive type
+            return dt_series.dt.tz_localize(None)
+
+        return dt_series.astype(dtype)
 
 
 @dataclasses.dataclass

morpheus/utils/downloader.py

@@ -21,6 +21,7 @@
 import os
 import threading
 import typing
+import warnings
 from enum import Enum
 
 import fsspec
@@ -131,7 +132,13 @@ def get_dask_client(self):
         dask.config.set({"distributed.client.heartbeat": self._dask_heartbeat_interval})
 
         if (self._merlin_distributed is None):
-            self._merlin_distributed = Distributed(client=dask.distributed.Client(self.get_dask_cluster()))
+            with warnings.catch_warnings():
+                # Merlin.Distributed will warn if a client already exists, the client in question is the one created
+                # and are explicitly passing to it in the constructor.
+                warnings.filterwarnings("ignore",
+                                        message="Existing Dask-client object detected in the current context.*",
+                                        category=UserWarning)
+                self._merlin_distributed = Distributed(client=dask.distributed.Client(self.get_dask_cluster()))
 
         return self._merlin_distributed
 

pyproject.toml

@@ -25,16 +25,18 @@ markers = [
 ]
 
 filterwarnings = [
-  # Warning coming from mlflow's usage of numpy
-  'ignore:`np.object` is a deprecated alias for the builtin `object`. To silence this warning, use `object` by itself. Doing this will not modify any behavior and is safe',
+  # Ignore our own warning about the df property since we still have to test it
   'ignore:Warning the df property returns a copy, please use the copy_dataframe method or the mutable_dataframe context manager to modify the DataFrame in-place instead.',
-  'ignore:`np.MachAr` is deprecated \(NumPy 1.22\):DeprecationWarning',
-  'ignore:Please use `spmatrix` from the `scipy.sparse` namespace, the `scipy.sparse.base` namespace is deprecated:DeprecationWarning', 
 
   # Deprecation warning from any project using distutils, currently known sources of this are:
   # GPUtils https://github.com/anderskm/gputil/issues/48
-  # PySpark https://issues.apache.org/jira/browse/SPARK-45390
+  # PySpark https://issues.apache.org/jira/browse/SPARK-45390 & https://issues.apache.org/jira/browse/SPARK-38660
   'ignore:The distutils package is deprecated and slated for removal in Python 3.12. Use setuptools or check PEP 632 for potential alternatives',
+  'ignore:distutils Version classes are deprecated. Use packaging.version instead.',
+
+  # Ignore cudf warnings about Pandas being used under the hood for processing json
+  'ignore:Using CPU via Pandas to write JSON dataset',
+  'ignore:Using CPU via Pandas to read JSON dataset',
 ]
 
 testpaths = ["tests"]

tests/_utils/dataset_manager.py

@@ -18,6 +18,7 @@
 import os
 import random
 import typing
+import warnings
 
 import cupy as cp
 import pandas as pd
@@ -235,7 +236,10 @@ def compare_df(cls,
                    dfb: typing.Union[pd.DataFrame, cdf.DataFrame],
                    **compare_args):
         """Wrapper for `morpheus.utils.compare_df.compare_df`."""
-        return compare_df.compare_df(cls._value_as_pandas(dfa), cls._value_as_pandas(dfb), **compare_args)
+        with warnings.catch_warnings():
+            # Ignore performance warnings from pandas triggered by the comparison
+            warnings.filterwarnings("ignore", category=pd.errors.PerformanceWarning)
+            return compare_df.compare_df(cls._value_as_pandas(dfa), cls._value_as_pandas(dfb), **compare_args)
 
     @classmethod
     def assert_compare_df(cls,

tests/_utils/kafka.py

@@ -18,6 +18,7 @@
 import subprocess
 import time
 import typing
+import warnings
 from collections import namedtuple
 from functools import partial
 
@@ -73,10 +74,15 @@ def seek_to_beginning(kafka_consumer: "KafkaConsumer", timeout: int = PARTITION_
 
 @pytest.fixture(name='kafka_consumer', scope='function')
 def kafka_consumer_fixture(kafka_topics: KafkaTopics, _kafka_consumer: "KafkaConsumer"):
-    _kafka_consumer.subscribe([kafka_topics.output_topic])
-    seek_to_beginning(_kafka_consumer)
-
-    yield _kafka_consumer
+    with warnings.catch_warnings():
+        # Ignore warnings specific to the test fixture and not the actual morpheus code
+        warnings.filterwarnings("ignore",
+                                message=r"Exception ignored in:.*ConsumerCoordinator\.__del__",
+                                category=pytest.PytestUnraisableExceptionWarning)
+        _kafka_consumer.subscribe([kafka_topics.output_topic])
+        seek_to_beginning(_kafka_consumer)
+
+        yield _kafka_consumer
 
 
 def _init_pytest_kafka() -> (bool, Exception):

tests/dfencoder/test_autoencoder.py

@@ -16,6 +16,7 @@
 
 import os
 import typing
+import warnings
 from unittest.mock import patch
 
 import numpy as np
@@ -493,7 +494,10 @@ def test_auto_encoder_num_only_convergence(train_ae: autoencoder.AutoEncoder):
         'num_feat_2': [3.5, 3.0, 3.2, 3.1, 3.6, 3.9, 3.4, 3.4, 2.9, 3.1],
     })
 
-    train_ae.fit(num_df, epochs=50)
+    with warnings.catch_warnings():
+        # Ignore warning regarding tensorflow not being installed
+        warnings.filterwarnings("ignore", message="Initializing zero-element tensors is a no-op", category=UserWarning)
+        train_ae.fit(num_df, epochs=50)
 
     avg_loss = np.sum([np.array(loss[1])
                        for loss in train_ae.logger.train_fts.values()], axis=0) / len(train_ae.logger.train_fts)

tests/dfencoder/test_scalers.py

@@ -14,6 +14,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import warnings
+
 import numpy as np
 import pytest
 import torch
@@ -25,35 +27,41 @@
 # pylint: disable=redefined-outer-name
 
 
-@pytest.fixture(scope="function")
-def fit_tensor():
+@pytest.fixture(name="fit_tensor", scope="function")
+def fit_tensor_fixture():
     yield torch.tensor([4.4, 5.3, 6.5], dtype=torch.float32)
 
 
-@pytest.fixture(scope="function")
-def tensor():
+@pytest.fixture(name="tensor", scope="function")
+def tensor_fixture():
     yield torch.tensor([7.4, 8.3, 9.5], dtype=torch.float32)
 
 
-@pytest.fixture(scope="function")
-def standard_scaler(fit_tensor):
+@pytest.fixture(name="standard_scaler", scope="function")
+def standard_scaler_fixture(fit_tensor):
     scaler = scalers.StandardScaler()
     scaler.fit(fit_tensor)
     yield scaler
 
 
-@pytest.fixture(scope="function")
-def modified_scaler(fit_tensor):
+@pytest.fixture(name="modified_scaler", scope="function")
+def modified_scaler_fixture(fit_tensor):
     scaler = scalers.ModifiedScaler()
     scaler.fit(fit_tensor)
     yield scaler
 
 
-@pytest.fixture(scope="function")
-def gauss_rank_scaler(fit_tensor):
+@pytest.fixture(name="gauss_rank_scaler", scope="function")
+def gauss_rank_scaler_fixture(fit_tensor):
     scaler = scalers.GaussRankScaler()
-    scaler.fit(fit_tensor)
-    yield scaler
+
+    with warnings.catch_warnings():
+        # This warning is triggered by the abnormally small tensor size used in this test
+        warnings.filterwarnings("ignore",
+                                message=r"n_quantiles \(1000\) is greater than the total number of samples \(3\).*",
+                                category=UserWarning)
+        scaler.fit(fit_tensor)
+        yield scaler
 
 
 def test_ensure_float_type():
@@ -111,8 +119,7 @@ def test_modified_scaler_transform(modified_scaler, tensor):
     assert torch.equal(torch.round(results, decimals=2), expected), f"{results} != {expected}"
 
     # Test alternate path where median absolute deviation is 1
-    test = torch.tensor([3.0, 4.0, 4.0, 5.0])
-    modified_scaler.fit(test)
+    modified_scaler.fit(torch.tensor([3.0, 4.0, 4.0, 5.0]))
     results = modified_scaler.transform(tensor)
     expected = torch.tensor([5.43, 6.86, 8.78])
     assert torch.equal(torch.round(results, decimals=2), expected), f"{results} != {expected}"
@@ -124,8 +131,7 @@ def test_modified_scaler_inverse_transform(modified_scaler, tensor):
     assert torch.equal(torch.round(results, decimals=2), expected), f"{results} != {expected}"
 
     # Test alternate path where median absolute deviation is 1
-    test = torch.tensor([3.0, 4.0, 4.0, 5.0])
-    modified_scaler.fit(test)
+    modified_scaler.fit(torch.tensor([3.0, 4.0, 4.0, 5.0]))
     results = modified_scaler.inverse_transform(tensor)
     expected = torch.tensor([8.64, 9.2, 9.95])
     assert torch.equal(torch.round(results, decimals=2), expected), f"{results} != {expected}"

tests/examples/digital_fingerprinting/test_dfp_postprocessing_stage.py

@@ -51,7 +51,7 @@ def test_process_events_on_data(mock_datetime: mock.MagicMock,
 
     # post-process should replace nans, lets add a nan to the DF
     with dfp_multi_ae_message.meta.mutable_dataframe() as df:
-        df['v2'][10] = np.nan
+        df.loc[10, 'v2'] = np.nan
         df['event_time'] = ''
 
     set_log_level(morpheus_log_level)

tests/examples/digital_fingerprinting/test_dfp_split_users_stage.py

@@ -16,6 +16,7 @@
 import json
 import os
 import typing
+import warnings
 
 import pytest
 
@@ -99,7 +100,12 @@ def test_extract_users(config: Config,
                                skip_users=skip_users,
                                only_users=only_users)
 
-    results = stage.extract_users(df)
+    with warnings.catch_warnings():
+        # Ignore warning about the log message not being set. This happens whenever there aren't any output_messages
+        warnings.filterwarnings("ignore",
+                                message="Must set log msg before end of context! Skipping log",
+                                category=UserWarning)
+        results = stage.extract_users(df)
 
     if not include_generic and not include_individual:
         # Extra check for weird combination

tests/examples/llm/common/conftest.py

@@ -52,10 +52,10 @@ def import_content_extractor_module(restore_sys_path):  # pylint: disable=unused
     return content_extractor_module
 
 
-@pytest.fixture(name="nemollm", autouse=True, scope='session')
-def nemollm_fixture(fail_missing: bool):
+@pytest.fixture(name="langchain", autouse=True, scope='session')
+def langchain_fixture(fail_missing: bool):
     """
-    All the tests in this subdir require nemollm
+    All the tests in this subdir require langchain
     """
 
     skip_reason = ("Tests for the WebScraperStage require the langchain package to be installed, to install this run:\n"

tests/examples/ransomware_detection/test_preprocessing.py

@@ -139,10 +139,10 @@ def test_merge_curr_and_prev_snapshots(self, config: Config, rwd_conf: dict, dat
         }
 
         expected_df = dataset_pandas['examples/ransomware_detection/dask_results.csv'].fillna('')
-        expected_df['pid_process'][1] = 'test_val1'
-        expected_df['pid_process'][3] = 'test_val2'
+        expected_df.loc[1, 'pid_process'] = 'test_val1'
+        expected_df.loc[3, 'pid_process'] = 'test_val2'
 
-        expected_df['snapshot_id'] = snapshot_ids
+        expected_df.loc[:, 'snapshot_id'] = snapshot_ids
         expected_df.index = expected_df.snapshot_id
 
         stage._merge_curr_and_prev_snapshots(df, source_pid_process)