public_git/wyoming-piper-gpu
1
0
Fork 0
mirror of https://github.com/slackr31337/wyoming-piper-gpu.git synced 2026-10-18 22:48:09 +02:00
Code Issues Packages Projects Releases Wiki Activity

Change patches

Browse Source
This commit is contained in:
Robert Dunmire III
2025-01-04 16:17:03 -05:00
parent deadd0af87
commit cbced1fd5e
6 changed files with 61 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.vscode/settings.json
Vendored
+2
View File
@@ -5,10 +5,12 @@
"cudnn",
"cufft",
"curand",
"espeak",
"manylinux",
"nvrtc",
"onnxruntime",
"phonemize",
"rhasspy",
"setuptools",
"slackr",
"venv",
Dockerfile
+10 -15
View File
@@ -4,8 +4,9 @@ WORKDIR /usr/src
ARG TARGETARCH=amd64
ARG TARGETVARIANT=
ARG WYOMING_PIPER_VERSION='1.4.0'
ARG PIPER_RELEASE='1.2.0'
ARG WYOMING_PIPER_VERSION="1.4.0"
ARG PIPER_RELEASE="1.2.0"
ARG PIPER_URL="https://github.com/rhasspy/piper/releases/download/v${PIPER_RELEASE}/piper_${TARGETARCH}${TARGETVARIANT}.tar.gz"
ENV DEBIAN_FRONTEND=noninteractive
@@ -45,32 +46,26 @@ RUN \
onnxruntime-gpu \
&&\
\
/app/bin/python3 -m pip install --no-cache-dir\
/app/bin/python3 -m pip install --force-reinstall --no-cache-dir\
"wyoming-piper==${WYOMING_PIPER_VERSION}"\
&&\
\
wget \
"https://github.com/rhasspy/piper/releases/download/v${PIPER_RELEASE}/piper_${TARGETARCH}${TARGETVARIANT}.tar.gz" -O -|tar -zxvf - -C /usr/share
wget "${PIPER_URL}" -O -|tar -zxvf - -C /app
# pip3 install --no-cache-dir --force-reinstall --no-deps\
# piper_phonemize-1.1.0-py3-none-any.whl &&\
# \
# "wyoming-piper @ https://github.com/rhasspy/wyoming-piper/archive/refs/tags/v${WYOMING_PIPER_VERSION}.tar.gz"
# wget https://github.com/rhasspy/piper-phonemize/releases/download/v1.1.0/piper_phonemize-1.1.0-cp310-cp310-manylinux_2_28_x86_64.whl &&\
# mv piper_phonemize-1.1.0-cp310-cp310-manylinux_2_28_x86_64.whl piper_phonemize-1.1.0-py3-none-any.whl &&\
# rm -r piper_phonemize-1.1.0-py3-none-any.whl &&\
# Patch to enable CUDA arguments for piper
COPY patch/wyoming-piper_cuda.patch /tmp/
COPY patch/* /tmp/
RUN \
cd /app/lib/python3.10/site-packages/wyoming_piper/;\
patch -p0 --forward < /tmp/wyoming-piper_cuda.patch || true
patch -p0 --forward < /tmp/__main__.py.diff;\
patch -p0 --forward < /tmp/process.py.diff;\
patch -p0 --forward < /tmp/handler.py.diff || true
# Clean up
RUN \
rm -rf /root/.cache/pip /var/lib/apt/lists/* /tmp/*
COPY tests/* /app/tests/
# COPY tests/* /app/tests/
WORKDIR /app
COPY run.sh /app/
patch/__main__.diff
+16
View File
@@ -0,0 +1,16 @@
--- __main__.py 2025-01-04 17:48:27.645862671 +0000
+++ __main__.py.patched 2025-01-04 17:51:50.193357609 +0000
@@ -66,6 +66,15 @@
help="Download latest voices.json during startup",
)
#
+ parser.add_argument(
+ "--cuda",
+ action="store_true",
+ help="Use GPU"
+ )
+ parser.add_argument(
+ "--espeak-data-dir",
+ help="Path to espeak-ng data directory"
+ )
+
patch/handler.diff
+10
View File
@@ -0,0 +1,10 @@
--- handler.py 2025-01-04 19:50:00.000000000 +0000
+++ handler.py.patched 2025-01-04 20:58:52.201781636 +0000
@@ -88,7 +88,7 @@
_LOGGER.debug("input: %s", input_obj)
piper_proc.proc.stdin.write(
- (json.dumps(input_obj, ensure_ascii=False) + "\n").encode()
+ f{text}"\n").encode()
)
await piper_proc.proc.stdin.drain()
patch/process.diff
+23
View File
@@ -0,0 +1,23 @@
--- process.py 2025-01-04 19:50:00.000000000 +0000
+++ process.py.patched 2025-01-04 20:58:10.813174885 +0000
@@ -130,7 +130,6 @@
str(config_path),
"--output_dir",
str(wav_dir.name),
- "--json-input", # piper 1.1+
]
if voice_speaker is not None:
@@ -148,6 +147,12 @@
if self.args.noise_w:
piper_args.extend(["--noise-w", str(self.args.noise_w)])
+ if self.args.espeak_data_dir:
+ piper_args.extend(["--espeak_data", str(self.args.espeak_data_dir)])
+
+ if self.args.cuda:
+ piper_args.extend(["--cuda"])
+
_LOGGER.debug(
"Starting piper process: %s args=%s", self.args.piper, piper_args
)
patch/wyoming-piper_cuda.patch
-30
View File
@@ -1,30 +0,0 @@
--- __main__.py 2025-01-04 17:48:27.645862671 +0000
+++ __main__.py.patched 2025-01-04 17:51:50.193357609 +0000
@@ -66,6 +66,15 @@
help="Download latest voices.json during startup",
)
#
+ parser.add_argument(
+ "--cuda",
+ action="store_true",
+ help="Use GPU"
+ )
+ parser.add_argument(
+ "--espeak-data-dir",
+ help="Path to espeak-ng data directory"
+ )
+
--- process.py 2025-01-04 17:48:27.653862810 +0000
+++ process.py.patched 2025-01-04 17:52:48.494362993 +0000
@@ -148,6 +148,14 @@
if self.args.noise_w:
piper_args.extend(["--noise-w", str(self.args.noise_w)])
+ if self.args.espeak_data_dir:
+ piper_args.extend(["--espeak_data", str(self.args.espeak_data_dir)])
+
+ if self.args.cuda:
+ piper_args.extend(["--debug"])
+ piper_args.extend(["--cuda"])
+