Adds regression tests for #2007 (#2018)

* first commit * added regression test * cleanup * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add ray tests * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Co-authored-by: Geoffrey Angus <geoffrey@predibase.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
ludwig-ai · May 13, 2022 · 7e2b464 · 7e2b464
1 parent 2af21e4
commit 7e2b464
Show file tree

Hide file tree

Showing 2 changed files with 71 additions and 13 deletions.
diff --git a/tests/integration_tests/test_ray.py b/tests/integration_tests/test_ray.py
@@ -26,7 +26,7 @@
 from ludwig.api import LudwigModel
 from ludwig.backend import create_ray_backend, LOCAL_BACKEND
 from ludwig.backend.ray import get_trainer_kwargs, RayBackend
-from ludwig.constants import BALANCE_PERCENTAGE_TOLERANCE, NAME, TRAINER
+from ludwig.constants import BACKFILL, BALANCE_PERCENTAGE_TOLERANCE, NAME, TRAINER
 from ludwig.data.dataframe.dask import DaskEngine
 from ludwig.data.preprocessing import balance_data
 from ludwig.utils.data_utils import read_parquet
@@ -260,11 +260,25 @@ def test_ray_sequence():
     run_test_with_features(input_features, output_features)
 
 
+@pytest.mark.parametrize("feature_type", ["raw", "stft", "stft_phase", "group_delay", "fbank"])
 @pytest.mark.distributed
-def test_ray_audio():
+def test_ray_audio(feature_type):
     with tempfile.TemporaryDirectory() as tmpdir:
+        preprocessing_params = {
+            "audio_file_length_limit_in_s": 3.0,
+            "missing_value_strategy": BACKFILL,
+            "in_memory": True,
+            "padding_value": 0,
+            "norm": "per_file",
+            "audio_feature": {
+                "type": feature_type,
+                "window_length_in_s": 0.04,
+                "window_shift_in_s": 0.02,
+                "num_filter_bands": 80,
+            },
+        }
         audio_dest_folder = os.path.join(tmpdir, "generated_audio")
-        input_features = [audio_feature(folder=audio_dest_folder)]
+        input_features = [audio_feature(folder=audio_dest_folder, preprocessing=preprocessing_params)]
         output_features = [binary_feature()]
         run_test_with_features(input_features, output_features)
 

diff --git a/tests/ludwig/features/test_audio_feature.py b/tests/ludwig/features/test_audio_feature.py
@@ -1,11 +1,15 @@
+import os
 from random import choice
 from string import ascii_lowercase, ascii_uppercase, digits
-from typing import Dict
 
+import pandas as pd
 import pytest
 import torch
 
-from ludwig.features.audio_feature import AudioInputFeature
+from ludwig.backend import LOCAL_BACKEND
+from ludwig.constants import BACKFILL, PROC_COLUMN
+from ludwig.features.audio_feature import AudioFeatureMixin, AudioInputFeature
+from tests.integration_tests.utils import audio_feature, category_feature, generate_data
 
 BATCH_SIZE = 2
 SEQ_SIZE = 20
@@ -16,9 +20,9 @@
 DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
 
 
-@pytest.fixture(scope="module")
-def audio_config():
-    return {
+@pytest.mark.parametrize("encoder", ["rnn", "stacked_cnn", "parallel_cnn", "stacked_parallel_cnn", "rnn", "cnnrnn"])
+def test_audio_input_feature(encoder: str) -> None:
+    audio_config = {
         "name": "audio_feature",
         "type": "audio",
         "preprocessing": {
@@ -34,13 +38,53 @@ def audio_config():
         "vocab": VOCAB,
         "max_sequence_length": SEQ_SIZE,
         "embedding_size": AUDIO_W_SIZE,
+        "encoder": encoder,
     }
-
-
-@pytest.mark.parametrize("encoder", ["rnn", "stacked_cnn", "parallel_cnn", "stacked_parallel_cnn", "rnn", "cnnrnn"])
-def test_audio_input_feature(audio_config: Dict, encoder: str) -> None:
-    audio_config.update({"encoder": encoder})
     audio_input_feature = AudioInputFeature(audio_config).to(DEVICE)
     audio_tensor = torch.randn([BATCH_SIZE, SEQ_SIZE, AUDIO_W_SIZE], dtype=torch.float32).to(DEVICE)
     encoder_output = audio_input_feature(audio_tensor)
     assert encoder_output["encoder_output"].shape[1:] == audio_input_feature.output_shape
+
+
+@pytest.mark.parametrize("feature_type", ["raw", "stft", "stft_phase", "group_delay", "fbank"])
+def test_add_feature_data(feature_type, tmpdir):
+    preprocessing_params = {
+        "audio_file_length_limit_in_s": 3.0,
+        "missing_value_strategy": BACKFILL,
+        "in_memory": True,
+        "padding_value": 0,
+        "norm": "per_file",
+        "audio_feature": {
+            "type": feature_type,
+            "window_length_in_s": 0.04,
+            "window_shift_in_s": 0.02,
+            "num_filter_bands": 80,
+        },
+    }
+    audio_dest_folder = os.path.join(tmpdir, "generated_audio")
+    audio_feature_config = audio_feature(audio_dest_folder, preprocessing=preprocessing_params)
+    data_df_path = generate_data(
+        [audio_feature_config],
+        [category_feature(vocab_size=5, reduce_input="sum")],
+        os.path.join(tmpdir, "data.csv"),
+        num_examples=10,
+    )
+    data_df = pd.read_csv(data_df_path)
+    metadata = {
+        audio_feature_config["name"]: AudioFeatureMixin.get_feature_meta(
+            data_df[audio_feature_config["name"]], preprocessing_params, LOCAL_BACKEND
+        )
+    }
+
+    proc_df = {}
+    AudioFeatureMixin.add_feature_data(
+        feature_config=audio_feature_config,
+        input_df=data_df,
+        proc_df=proc_df,
+        metadata=metadata,
+        preprocessing_parameters=preprocessing_params,
+        backend=LOCAL_BACKEND,
+        skip_save_processed_input=False,
+    )
+
+    assert len(proc_df[audio_feature_config[PROC_COLUMN]]) == 10