upload

.gitignore

@@ -0,0 +1,2 @@
+data/
+compose.yaml

Dockerfile

@@ -0,0 +1,28 @@
+FROM debian:12
+
+ARG WYOMING_FASTER_WHIPSER_VERSION=2.2.0
+ARG NVIDIA_DRIVER_VERSION=550.127.05
+
+# install requirements
+RUN apt-get update \
+  && apt-get install -y --no-install-recommends python3-pip wget kmod \
+  && apt-get clean \
+  && rm -rf /var/lib/apt/lists/*
+
+# install nvidia linux drivers
+RUN cd /tmp \
+    && wget "https://de.download.nvidia.com/XFree86/Linux-x86_64/${NVIDIA_DRIVER_VERSION}/NVIDIA-Linux-x86_64-${NVIDIA_DRIVER_VERSION}.run" \
+    && chmod +x NVIDIA-Linux-x86_64-${NVIDIA_DRIVER_VERSION}.run \
+    && bash -c "./NVIDIA-Linux-x86_64-${NVIDIA_DRIVER_VERSION}.run -s --no-kernel-module" \
+    $$ rm NVIDIA-Linux-x86_64-${NVIDIA_DRIVER_VERSION}.run
+
+# install wyoming-faster-whisper and nvidia cuda libs
+RUN python3 -m pip install --no-cache-dir --break-system-packages "wyoming-faster-whisper @ https://github.com/rhasspy/wyoming-faster-whisper/archive/refs/tags/v${WYOMING_FASTER_WHIPSER_VERSION}.tar.gz" \
+    && python3 -m pip install --no-cache-dir --break-system-packages nvidia-cublas-cu12 nvidia-cudnn-cu12==9.*
+
+EXPOSE 10300/tcp
+VOLUME [ "/data" ]
+
+COPY run.sh /
+
+ENTRYPOINT [ "/run.sh" ]

README.md

@@ -0,0 +1,70 @@
+# wyoming-faster-whisper-cuda
+
+This takes the [wyoming-faster-whisper](https://github.com/rhasspy/wyoming-faster-whisper) and wraps it into an nvidia cuda supported container.
+
+**Note** This is only supported on x86_64 systems, yet.
+
+## Usage
+
+### Prerequisits
+
+1. nvidia cuda compatible gpu
+2. [nvidia linux drivers](https://www.nvidia.com/en-us/drivers/unix) installed on the host
+3. up and running [docker](https://docs.docker.com/engine/install) installation on the host
+
+### Installation
+
+1. download this repo
+
+    ```shell
+    $ git clone https://github.com/mib1185/wyoming-faster-whisper-cuda.git
+    ```
+
+2. `compose.yaml` file
+
+    create a `compose.yaml` file, which:
+
+    - builds from the local `Dockerfile`
+    - adds the needed parameters for `model` and `language` as command line parameter
+    - (_optional_) enables `debug` logging via command line parameter
+    - provides a `data` volume or directory
+    - exposes the port `10300/tcp`
+    - maps your nvidia gpu related devices into the container (_obtain with `ls -la /dev/nvidia*`_)
+    - (_optional_) set `restart: always`
+
+    **example `compose.yaml` file**
+
+    ```yaml
+    name: wyoming
+    services:
+    faster-whisper-cuda:
+      container_name: faster-whisper-cuda
+      build: .
+      command: "--model large --language de --debug"
+      volumes:
+        - ./data:/data
+      ports:
+        - 10300:10300/tcp
+      devices:
+        - /dev/nvidia-uvm:/dev/nvidia-uvm
+        - /dev/nvidia-uvm-tools:/dev/nvidia-uvm-tools
+        - /dev/nvidia0:/dev/nvidia0
+        - /dev/nvidiactl:/dev/nvidiactl
+      restart: always
+    ```
+
+3. start service
+
+    on first start, the docker image is build, which needs some time
+
+    ```shell
+    $ docker compose up -d
+    ```
+
+4. check if service is running
+
+    ```shell
+    $ docker ps
+    CONTAINER ID   IMAGE                         COMMAND                  CREATED         STATUS         PORTS                                           NAMES
+    474e37a84326   wyoming-faster-whisper-cuda   "/run.sh --model lar…"   3 minutes ago   Up 3 minutes   0.0.0.0:10300->10300/tcp, :::10300->10300/tcp   faster-whisper-cuda
+    ```

run.sh

@@ -0,0 +1,5 @@
+#!/bin/bash
+
+export LD_LIBRARY_PATH=`python3 -c 'import os; import nvidia.cublas.lib; import nvidia.cudnn.lib; print(os.path.dirname(nvidia.cublas.lib.__file__) + ":" + os.path.dirname(nvidia.cudnn.lib.__file__))'`
+
+python3 -m wyoming_faster_whisper --uri tcp://0.0.0.0:10300 --data-dir /data --download-dir /data --device cuda "$@"