git: 471d29fc25d0 - main - misc/py-openai-whisper: New port: Robust Speech Recognition via Large-Scale Weak Supervision

The branch main has been updated by yuri:

URL: https://cgit.FreeBSD.org/ports/commit/?id=471d29fc25d0c664931c0dfe005f07769bb48ca5

commit 471d29fc25d0c664931c0dfe005f07769bb48ca5
Author:     Yuri Victorovich <yuri@FreeBSD.org>
AuthorDate: 2024-03-22 04:02:27 +0000
Commit:     Yuri Victorovich <yuri@FreeBSD.org>
CommitDate: 2024-03-22 04:02:27 +0000

    misc/py-openai-whisper: New port: Robust Speech Recognition via Large-Scale Weak Supervision
---
 misc/Makefile                    |  1 +
 misc/py-openai-whisper/Makefile  | 33 +++++++++++++++++++++++++++++++++
 misc/py-openai-whisper/distinfo  |  3 +++
 misc/py-openai-whisper/pkg-descr |  4 ++++
 4 files changed, 41 insertions(+)

diff --git a/misc/Makefile b/misc/Makefile
index 6ef2c99c1ff5..75b1f28a5c61 100644
--- a/misc/Makefile
+++ b/misc/Makefile
@@ -450,6 +450,7 @@
     SUBDIR += py-onnx
     SUBDIR += py-onnx-tf
     SUBDIR += py-openai
+    SUBDIR += py-openai-whisper
     SUBDIR += py-openpaperwork-core
     SUBDIR += py-openpaperwork-gtk
     SUBDIR += py-openvdb
diff --git a/misc/py-openai-whisper/Makefile b/misc/py-openai-whisper/Makefile
new file mode 100644
index 000000000000..a8e1f4dbc882
--- /dev/null
+++ b/misc/py-openai-whisper/Makefile
@@ -0,0 +1,33 @@
+PORTNAME=	openai-whisper
+DISTVERSIONPREFIX=	v
+DISTVERSION=	20231117
+CATEGORIES=	misc python # machine-learning
+PKGNAMEPREFIX=	${PYTHON_PKGNAMEPREFIX}
+
+MAINTAINER=	yuri@FreeBSD.org
+COMMENT=	Robust Speech Recognition via Large-Scale Weak Supervision
+WWW=		https://openai.com/research/whisper
+
+LICENSE=	MIT
+LICENSE_FILE=	${WRKSRC}/LICENSE
+
+BUILD_DEPENDS=	${PY_SETUPTOOLS} \
+		${PYTHON_PKGNAMEPREFIX}wheel>0:devel/py-wheel@${PY_FLAVOR}
+RUN_DEPENDS=	${PYTHON_PKGNAMEPREFIX}more-itertools>0:devel/py-more-itertools@${PY_FLAVOR} \
+		${PYTHON_PKGNAMEPREFIX}numba>0:devel/py-numba@${PY_FLAVOR} \
+		${PYTHON_PKGNAMEPREFIX}numpy>0:math/py-numpy@${PY_FLAVOR} \
+		${PYTHON_PKGNAMEPREFIX}tiktoken>=0:textproc/py-tiktoken@${PY_FLAVOR} \
+		${PYTHON_PKGNAMEPREFIX}pytorch>=0:misc/py-pytorch@${PY_FLAVOR} \
+		${PYTHON_PKGNAMEPREFIX}tqdm>0:misc/py-tqdm@${PY_FLAVOR}
+TEST_DEPENDS=	${PYTHON_PKGNAMEPREFIX}scipy>0:science/py-scipy@${PY_FLAVOR}
+
+USES=		python
+USE_PYTHON=	pep517 autoplist pytest # tests use 4+ GB each, several tests fail
+
+USE_GITHUB=	yes
+GH_ACCOUNT=	openai
+GH_PROJECT=	whisper
+
+NO_ARCH=	yes
+
+.include <bsd.port.mk>
diff --git a/misc/py-openai-whisper/distinfo b/misc/py-openai-whisper/distinfo
new file mode 100644
index 000000000000..9cf8b2cb969b
--- /dev/null
+++ b/misc/py-openai-whisper/distinfo
@@ -0,0 +1,3 @@
+TIMESTAMP = 1711013814
+SHA256 (openai-whisper-v20231117_GH0.tar.gz) = b0f8b8d3b485fad2c423ba7f8b95eded067aad11ed3165828aad819d168cac06
+SIZE (openai-whisper-v20231117_GH0.tar.gz) = 7261054
diff --git a/misc/py-openai-whisper/pkg-descr b/misc/py-openai-whisper/pkg-descr
new file mode 100644
index 000000000000..c448226b1e34
--- /dev/null
+++ b/misc/py-openai-whisper/pkg-descr
@@ -0,0 +1,4 @@
+Whisper is a general-purpose speech recognition model. It is trained on a large
+dataset of diverse audio and is also a multitasking model that can perform
+multilingual speech recognition, speech translation, and language
+identification.