Commit @af0231d0cf933267013855f0e0b9181e8665e747 - yjyoon/whisper_server

a8efff4

af0231d

README.md

--- README.md

+++ README.md


 Features:
 - GPU and CPU support.
 - Easily deployable using Docker.
-- Configurable through environment variables (see [config.py](./faster_whisper_server/config.py)).
+- **Configurable through environment variables (see [config.py](./faster_whisper_server/config.py))**.
 - OpenAI API compatible.
 
 Please create an issue if you find a bug, have a question, or a feature suggestion.

a8efff4

af0231d

examples/live-audio/script.sh

--- examples/live-audio/script.sh

+++ examples/live-audio/script.sh


 # ffmpeg -y -hide_banner -loglevel quiet -i audio.mp3 -ac 1 -ar 16000 -f s16le -acodec pcm_s16le audio.pcm
 # rm -f audio.mp3
 
-export WHISPER_MODEL=Systran/faster-distil-whisper-large-v3 # or Systran/faster-whisper-tiny.en if you are running on a CPU for a faster inference.
+export WHISPER__MODEL=Systran/faster-distil-whisper-large-v3 # or Systran/faster-whisper-tiny.en if you are running on a CPU for a faster inference.
 
 # Ensure you have `faster-whisper-server` running. If this is your first time running it expect to wait up-to a minute for the model to be downloaded and loaded into memory. You can run `curl localhost:8000/health` to check if the server is ready or watch the logs with `docker logs -f <container_id>`.
-docker run --detach --gpus=all --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER_MODEL=$WHISPER_MODEL fedirz/faster-whisper-server:latest-cuda
+docker run --detach --gpus=all --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER__MODEL=$WHISPER__MODEL fedirz/faster-whisper-server:latest-cuda
 # or you can run it on a CPU
-# docker run --detach --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER_MODEL=$WHISPER_MODEL fedirz/faster-whisper-server:latest-cpu
+# docker run --detach --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER__MODEL=$WHISPER__MODEL fedirz/faster-whisper-server:latest-cpu
 
 # `pv` is used to limit the rate at which the audio is streamed to the server. Audio is being streamed at a rate of 32kb/s(16000 sample rate * 16-bit sample / 8 bits per byte = 32000 bytes per second). This emulutes live audio input from a microphone: `ffmpeg -loglevel quiet -f alsa -i default -ac 1 -ar 16000 -f s16le`
 # shellcheck disable=SC2002

a8efff4

af0231d

examples/youtube/script.sh

--- examples/youtube/script.sh

+++ examples/youtube/script.sh


 set -e
 
 # NOTE: do not use any distil-* model other than the large ones as they don't work on long audio files for some reason.
-export WHISPER_MODEL=Systran/faster-distil-whisper-large-v3 # or Systran/faster-whisper-tiny.en if you are running on a CPU for a faster inference.
+export WHISPER__MODEL=Systran/faster-distil-whisper-large-v3 # or Systran/faster-whisper-tiny.en if you are running on a CPU for a faster inference.
 
 # Ensure you have `faster-whisper-server` running. If this is your first time running it expect to wait up-to a minute for the model to be downloaded and loaded into memory. You can run `curl localhost:8000/health` to check if the server is ready or watch the logs with `docker logs -f <container_id>`.
-docker run --detach --gpus=all --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER_MODEL=$WHISPER_MODEL fedirz/faster-whisper-server:latest-cuda
+docker run --detach --gpus=all --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER__MODEL=$WHISPER__MODEL fedirz/faster-whisper-server:latest-cuda
 # or you can run it on a CPU
-# docker run --detach --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER_MODEL=$WHISPER_MODEL fedirz/faster-whisper-server:latest-cpu
+# docker run --detach --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER__MODEL=$WHISPER__MODEL fedirz/faster-whisper-server:latest-cpu
 
 # Download the audio from a YouTube video. In this example I'm downloading "The Evolution of the Operating System" by Asionometry YouTube channel. I highly checking this channel out, the guy produces very high content. If you don't have `youtube-dl`, you'll have to install it. https://github.com/ytdl-org/youtube-dl
 youtube-dl --extract-audio --audio-format mp3 -o the-evolution-of-the-operating-system.mp3 'https://www.youtube.com/watch?v=1lG7lFLXBIs'

a8efff4

af0231d

faster_whisper_server/config.py

--- faster_whisper_server/config.py

+++ faster_whisper_server/config.py


 
     Pydantic will automatically handle mapping uppercased environment variables to the corresponding fields.
     To populate nested, the environment should be prefixed with the nested field name and an underscore. For example,
-    the environment variable `LOG_LEVEL` will be mapped to `log_level`, `WHISPER_MODEL` to `whisper.model`, etc.
-    """
+    the environment variable `LOG_LEVEL` will be mapped to `log_level`, `WHISPER__MODEL`(note the double underscore) to `whisper.model`, etc.
+    """  # noqa: E501
 
     model_config = SettingsConfigDict(env_nested_delimiter="__")
 

...	...	@@ -162,8 +162,8 @@
162	162
163	163	Pydantic will automatically handle mapping uppercased environment variables to the corresponding fields.
164	164	To populate nested, the environment should be prefixed with the nested field name and an underscore. For example,
165		- the environment variable `LOG_LEVEL` will be mapped to `log_level`, `WHISPER_MODEL` to `whisper.model`, etc.
166		- """
	165	+ the environment variable `LOG_LEVEL` will be mapped to `log_level`, `WHISPER__MODEL`(note the double underscore) to `whisper.model`, etc.
	166	+ """ # noqa: E501
167	167
168	168	model_config = SettingsConfigDict(env_nested_delimiter="__")
169	169

...	...	@@ -3,7 +3,7 @@
3	3	Features:
4	4	- GPU and CPU support.
5	5	- Easily deployable using Docker.
6		-- Configurable through environment variables (see [config.py](./faster_whisper_server/config.py)).
	6	+- Configurable through environment variables (see [config.py](./faster_whisper_server/config.py)).
7	7	- OpenAI API compatible.
8	8
9	9	Please create an issue if you find a bug, have a question, or a feature suggestion.

...	...	@@ -7,12 +7,12 @@
7	7	# ffmpeg -y -hide_banner -loglevel quiet -i audio.mp3 -ac 1 -ar 16000 -f s16le -acodec pcm_s16le audio.pcm
8	8	# rm -f audio.mp3
9	9
10		-export WHISPER_MODEL=Systran/faster-distil-whisper-large-v3 # or Systran/faster-whisper-tiny.en if you are running on a CPU for a faster inference.
	10	+export WHISPER__MODEL=Systran/faster-distil-whisper-large-v3 # or Systran/faster-whisper-tiny.en if you are running on a CPU for a faster inference.
11	11
12	12	# Ensure you have `faster-whisper-server` running. If this is your first time running it expect to wait up-to a minute for the model to be downloaded and loaded into memory. You can run `curl localhost:8000/health` to check if the server is ready or watch the logs with `docker logs -f <container_id>`.
13		-docker run --detach --gpus=all --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER_MODEL=$WHISPER_MODEL fedirz/faster-whisper-server:latest-cuda
	13	+docker run --detach --gpus=all --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER__MODEL=$WHISPER__MODEL fedirz/faster-whisper-server:latest-cuda
14	14	# or you can run it on a CPU
15		-# docker run --detach --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER_MODEL=$WHISPER_MODEL fedirz/faster-whisper-server:latest-cpu
	15	+# docker run --detach --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER__MODEL=$WHISPER__MODEL fedirz/faster-whisper-server:latest-cpu
16	16
17	17	# `pv` is used to limit the rate at which the audio is streamed to the server. Audio is being streamed at a rate of 32kb/s(16000 sample rate * 16-bit sample / 8 bits per byte = 32000 bytes per second). This emulutes live audio input from a microphone: `ffmpeg -loglevel quiet -f alsa -i default -ac 1 -ar 16000 -f s16le`
18	18	# shellcheck disable=SC2002

...	...	@@ -3,12 +3,12 @@
3	3	set -e
4	4
5	5	# NOTE: do not use any distil-* model other than the large ones as they don't work on long audio files for some reason.
6		-export WHISPER_MODEL=Systran/faster-distil-whisper-large-v3 # or Systran/faster-whisper-tiny.en if you are running on a CPU for a faster inference.
	6	+export WHISPER__MODEL=Systran/faster-distil-whisper-large-v3 # or Systran/faster-whisper-tiny.en if you are running on a CPU for a faster inference.
7	7
8	8	# Ensure you have `faster-whisper-server` running. If this is your first time running it expect to wait up-to a minute for the model to be downloaded and loaded into memory. You can run `curl localhost:8000/health` to check if the server is ready or watch the logs with `docker logs -f <container_id>`.
9		-docker run --detach --gpus=all --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER_MODEL=$WHISPER_MODEL fedirz/faster-whisper-server:latest-cuda
	9	+docker run --detach --gpus=all --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER__MODEL=$WHISPER__MODEL fedirz/faster-whisper-server:latest-cuda
10	10	# or you can run it on a CPU
11		-# docker run --detach --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER_MODEL=$WHISPER_MODEL fedirz/faster-whisper-server:latest-cpu
	11	+# docker run --detach --publish 8000:8000 --volume ~/.cache/huggingface:/root/.cache/huggingface --env WHISPER__MODEL=$WHISPER__MODEL fedirz/faster-whisper-server:latest-cpu
12	12
13	13	# Download the audio from a YouTube video. In this example I'm downloading "The Evolution of the Operating System" by Asionometry YouTube channel. I highly checking this channel out, the guy produces very high content. If you don't have `youtube-dl`, you'll have to install it. https://github.com/ytdl-org/youtube-dl
14	14	youtube-dl --extract-audio --audio-format mp3 -o the-evolution-of-the-operating-system.mp3 'https://www.youtube.com/watch?v=1lG7lFLXBIs'

Delete comment