sci-libs/datasets: new package, add 2.11.0

Signed-off-by: Alfredo Tupone <tupone@gentoo.org>

4 files changed, 234 insertions, 0 deletions

diff --git a/sci-libs/datasets/Manifest b/sci-libs/datasets/Manifest
new file mode 100644
index 000000000000..37284ce3b3e7
--- /dev/null
+++ b/sci-libs/datasets/Manifest
@@ -0,0 +1 @@
+DIST datasets-2.11.0.gh.tar.gz 2141289 BLAKE2B 0fb471dd6ee5de3831eb6586c4a15e67381262470b72d5ab02ee87dfc7977cb4d40e04da6507049d1e47cb8948cad11988bb7627293b48231e1cd413d2cfb885 SHA512 9ec2274d7978e3dde1b2f8ce78dd65bdf66742bbfee7b8672af46216aeaae3ef5c4604a8a5ea0bdee808f1c362cca9a122c16d2e9a161678148e581e4cd5c863
diff --git a/sci-libs/datasets/datasets-2.11.0.ebuild b/sci-libs/datasets/datasets-2.11.0.ebuild
new file mode 100644
index 000000000000..0ff1a4f0b386
--- /dev/null
+++ b/sci-libs/datasets/datasets-2.11.0.ebuild
@@ -0,0 +1,53 @@
+# Copyright 2023 Gentoo Authors
+# Distributed under the terms of the GNU General Public License v2
+
+EAPI=8
+
+DISTUTILS_USE_PEP517=setuptools
+PYTHON_COMPAT=( python3_{9..11} )
+DISTUTILS_SINGLE_IMPL=1
+inherit distutils-r1
+
+DESCRIPTION="access and share datasets for Audio, Computer Vision, and NLP tasks"
+HOMEPAGE="
+	https://pypi.org/project/datasets/
+"
+SRC_URI="https://github.com/huggingface/${PN}/archive/refs/tags/${PV}.tar.gz
+	-> ${P}.gh.tar.gz"
+IUSE="test"
+
+LICENSE="Apache-2.0"
+SLOT="0"
+KEYWORDS="~amd64"
+
+RDEPEND="
+	${PYTHON_DEPS}
+	sci-libs/pytorch[${PYTHON_SINGLE_USEDEP}]
+	$(python_gen_cond_dep '
+		dev-python/pyarrow[${PYTHON_USEDEP},parquet,snappy]
+		dev-python/fsspec[${PYTHON_USEDEP}]
+		dev-python/multiprocess[${PYTHON_USEDEP}]
+		dev-python/aiohttp[${PYTHON_USEDEP}]
+		dev-python/xxhash[${PYTHON_USEDEP}]
+		dev-python/zstandard[${PYTHON_USEDEP}]
+		dev-python/absl-py[${PYTHON_USEDEP}]
+		sci-libs/scikit-learn[${PYTHON_USEDEP}]
+	')
+"
+DEPEND="${RDEPEND}"
+BDEPEND="test? (
+	$(python_gen_cond_dep '
+		dev-python/pytest-datadir[${PYTHON_USEDEP}]
+	')
+)"
+
+PATCHES=( "${FILESDIR}"/${P}-tests.patch )
+
+distutils_enable_tests pytest
+
+src_prepare() {
+	# Require jiwer
+	rm metrics/cer/test_cer.py || die
+
+	distutils-r1_src_prepare
+}
diff --git a/sci-libs/datasets/files/datasets-2.11.0-tests.patch b/sci-libs/datasets/files/datasets-2.11.0-tests.patch
new file mode 100644
index 000000000000..01e5d9c70e7b
--- /dev/null
+++ b/sci-libs/datasets/files/datasets-2.11.0-tests.patch
@@ -0,0 +1,168 @@
+--- a/tests/test_metric_common.py	2023-05-04 18:48:48.550861318 +0200
++++ b/tests/test_metric_common.py	2023-05-04 18:50:25.787364577 +0200
+@@ -93,6 +93,7 @@
+     INTENSIVE_CALLS_PATCHER = {}
+     metric_name = None
+ 
++    @pytest.mark.skip(reason="disabling, depends on bert_score, bleurt, math_equivalence, coval, nltk, faiss, mauve, rouge_score, sacrebleu, sacremoses ...")
+     def test_load_metric(self, metric_name):
+         doctest.ELLIPSIS_MARKER = "[...]"
+         metric_module = importlib.import_module(
+--- a/tests/test_hf_gcp.py	2023-05-04 19:33:31.150825303 +0200
++++ b/tests/test_hf_gcp.py	2023-05-04 19:40:08.401759538 +0200
+@@ -69,6 +69,7 @@
+             self.assertTrue(os.path.exists(datset_info_path))
+ 
+ 
++@pytest.mark.skip(reason="require apache_beam")
+ @pytest.mark.integration
+ def test_wikipedia_frr(tmp_path_factory):
+     tmp_dir = tmp_path_factory.mktemp("test_hf_gcp") / "test_wikipedia_simple"
+--- a/tests/test_distributed.py	2023-05-04 19:43:09.861275030 +0200
++++ b/tests/test_distributed.py	2023-05-04 19:44:17.608326722 +0200
+@@ -55,6 +55,7 @@
+     assert len({tuple(x.values()) for ds in datasets_per_rank for x in ds}) == full_size
+ 
+ 
++@pytest.mark.skip(reason="require distributed torch")
+ @pytest.mark.parametrize("streaming", [False, True])
+ @require_torch
+ @pytest.mark.skipif(os.name == "nt", reason="execute_subprocess_async doesn't support windows")
+@@ -76,6 +77,7 @@
+     execute_subprocess_async(cmd, env=os.environ.copy())
+ 
+ 
++@pytest.mark.skip(reason="require distributed torch")
+ @pytest.mark.parametrize(
+     "nproc_per_node, num_workers",
+     [
+--- a/tests/utils.py	2023-05-06 08:43:16.251987543 +0200
++++ b/tests/utils.py	2023-05-06 08:44:24.467952870 +0200
+@@ -54,8 +54,8 @@
+ # Audio
+ require_sndfile = pytest.mark.skipif(
+     # On Windows and OS X, soundfile installs sndfile
+-    find_spec("soundfile") is None or version.parse(importlib_metadata.version("soundfile")) < version.parse("0.12.0"),
+-    reason="test requires sndfile>=0.12.1: 'pip install \"soundfile>=0.12.1\"'; ",
++    True,
++    reason="test requires librosa",
+ )
+ 
+ # Beam
+--- a/tests/features/test_audio.py	2023-05-06 09:03:58.680108142 +0200
++++ a/tests/features/test_audio.py	2023-05-06 09:05:50.463407967 +0200
+@@ -57,6 +57,7 @@
+     assert features.arrow_schema == pa.schema({"sequence_of_audios": pa.list_(Audio().pa_type)})
+ 
+ 
++@pytest.mark.skip(reason="require librosa")
+ @pytest.mark.parametrize(
+     "build_example",
+     [
+@@ -82,6 +82,7 @@
+     assert decoded_example.keys() == {"path", "array", "sampling_rate"}
+ 
+ 
++@pytest.mark.skip(reason="require librosa")
+ @pytest.mark.parametrize(
+     "build_example",
+     [
+@@ -148,6 +149,7 @@
+     assert decoded_example["sampling_rate"] == 48000
+ 
+ 
++@pytest.mark.skip(reason="require librosa")
+ @pytest.mark.parametrize("sampling_rate", [16_000, 48_000])
+ def test_audio_decode_example_pcm(shared_datadir, sampling_rate):
+     audio_path = str(shared_datadir / "test_audio_16000.pcm")
+@@ -416,6 +417,7 @@
+     assert column[0]["sampling_rate"] == 16000
+ 
+ 
++@pytest.mark.skip(reason="require librosa")
+ @pytest.mark.parametrize(
+     "build_data",
+     [
+@@ -440,6 +442,7 @@
+     assert item["audio"].keys() == {"path", "array", "sampling_rate"}
+ 
+ 
++@pytest.mark.skip(reason="require librosa")
+ def test_dataset_concatenate_audio_features(shared_datadir):
+     # we use a different data structure between 1 and 2 to make sure they are compatible with each other
+     audio_path = str(shared_datadir / "test_audio_44100.wav")
+@@ -453,6 +456,7 @@
+     assert concatenated_dataset[1]["audio"]["array"].shape == dset2[0]["audio"]["array"].shape
+ 
+ 
++@pytest.mark.skip(reason="require librosa")
+ def test_dataset_concatenate_nested_audio_features(shared_datadir):
+     # we use a different data structure between 1 and 2 to make sure they are compatible with each other
+     audio_path = str(shared_datadir / "test_audio_44100.wav")
+@@ -616,6 +616,7 @@
+     assert isinstance(ds, Dataset)
+ 
+ 
++@require_sndfile
+ def test_dataset_with_audio_feature_undecoded(shared_datadir):
+     audio_path = str(shared_datadir / "test_audio_44100.wav")
+     data = {"audio": [audio_path]}
+@@ -633,6 +634,7 @@
+     assert column[0] == {"path": audio_path, "bytes": None}
+ 
+ 
++@require_sndfile
+ def test_formatted_dataset_with_audio_feature_undecoded(shared_datadir):
+     audio_path = str(shared_datadir / "test_audio_44100.wav")
+     data = {"audio": [audio_path]}
+@@ -664,6 +666,7 @@
+         assert column[0] == {"path": audio_path, "bytes": None}
+ 
+ 
++@require_sndfile
+ def test_dataset_with_audio_feature_map_undecoded(shared_datadir):
+     audio_path = str(shared_datadir / "test_audio_44100.wav")
+     data = {"audio": [audio_path]}
+--- a/tests/test_metric_common.py	2023-05-06 13:20:24.496197629 +0200
++++ b/tests/test_metric_common.py	2023-05-06 13:21:09.916732417 +0200
+@@ -210,6 +210,7 @@
+             yield
+ 
+ 
++@pytest.mark.skip(reason="require seqeval")
+ def test_seqeval_raises_when_incorrect_scheme():
+     metric = load_metric(os.path.join("metrics", "seqeval"))
+     wrong_scheme = "ERROR"
+--- a/tests/packaged_modules/test_audiofolder.py	2023-05-06 14:00:39.560876163 +0200
++++ b/tests/packaged_modules/test_audiofolder.py	2023-05-06 14:01:26.005212423 +0200
+@@ -4,7 +4,6 @@
+ import librosa
+ import numpy as np
+ import pytest
+-import soundfile as sf
+ 
+ from datasets import Audio, ClassLabel, Features, Value
+ from datasets.data_files import DataFilesDict, get_data_patterns_locally
+@@ -191,9 +190,11 @@
+     assert len(data_files_with_two_splits_and_metadata["test"]) == 2
+     return data_files_with_two_splits_and_metadata
+ 
+-
++@pytest.mark.skip(reason="require soundfile")
+ @pytest.fixture
+ def data_files_with_zip_archives(tmp_path, audio_file):
++    import soundfile as sf
++
+     data_dir = tmp_path / "audiofolder_data_dir_with_zip_archives"
+     data_dir.mkdir(parents=True, exist_ok=True)
+     archive_dir = data_dir / "archive"
+--- a/tests/test_arrow_dataset.py	2023-05-06 15:36:11.080459079 +0200
++++ b/tests/test_arrow_dataset.py	2023-05-06 15:38:07.452828528 +0200
+@@ -3928,6 +3928,7 @@
+                 )
+                 self.assertDictEqual(features_after_cast, dset.features)
+ 
++    @pytest.mark.skip(reason="require soundfile")
+     def test_task_automatic_speech_recognition(self):
+         # Include a dummy extra column `dummy` to test we drop it correctly
+         features_before_cast = Features(
diff --git a/sci-libs/datasets/metadata.xml b/sci-libs/datasets/metadata.xml
new file mode 100644
index 000000000000..94c112402049
--- /dev/null
+++ b/sci-libs/datasets/metadata.xml
@@ -0,0 +1,12 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE pkgmetadata SYSTEM "https://www.gentoo.org/dtd/metadata.dtd">
+<pkgmetadata>
+	<maintainer type="person">
+		<email>tupone@gentoo.org</email>
+		<name>Tupone Alfredo</name>
+	</maintainer>
+	<upstream>
+		<remote-id type="github">huggingface/datasets</remote-id>
+		<remote-id type="pypi">datasets</remote-id>
+	</upstream>
+</pkgmetadata>