Download whisper model and convert for faster_whisper while building Dockerfile

Dockerfile.gpu

@@ -34,4 +34,9 @@ COPY . .
 RUN poetry install
 RUN $POETRY_VENV/bin/pip install torch==1.13.0+cu117 -f https://download.pytorch.org/whl/torch
 
+ENV ASR_MODEL="base"
+
+# TODO: Skip based on ENV variable
+RUN ./faster_whisper_model_conversion.sh ${ASR_MODEL}
+
 CMD gunicorn --bind 0.0.0.0:9000 --workers 1 --timeout 0 app.webservice:app -k uvicorn.workers.UvicornWorker

README.md

@@ -70,6 +70,11 @@ Starting the Webservice:
 poetry run gunicorn --bind 0.0.0.0:9000 --workers 1 --timeout 0 app.webservice:app -k uvicorn.workers.UvicornWorker
 ```
 
+With docker compose:
+```sh
+docker-compose up --build
+```
+
 ## Quick start
 
 After running the docker image interactive Swagger API documentation is available at [localhost:9000/docs](http://localhost:9000/docs)

app/webservice.py

@@ -137,11 +137,10 @@ def run_asr(
     with model_lock:   
         model = get_modal(faster)
         if faster:
-            # TODO: options_dict
             segments = []
             text = ""
             i = 0
-            segment_generator, info = model.transcribe(audio, beam_size=5)
+            segment_generator, info = model.transcribe(audio, beam_size=5, **options_dict)
             for segment in segment_generator:
                 segments.append(segment)
                 text = text + segment.text

docker-compose.yml

@@ -13,12 +13,18 @@ services:
               count: 1
               capabilities: [gpu]
     environment:
-      - ASR_MODEL=large
+      - ASR_MODEL=base
     ports:
       - 9000:9000
     volumes:
       - ./app:/app/app
-      - ~/.cache/poetry:/root/.cache/poetry
-      - ~/.cache/whisper:/root/.cache/whisper
-      - ~/.cache/faster_whisper:/root/.cache/faster_whisper
+      - cache-pip:/root/.cache/pip
+      - cache-poetry:/root/.cache/poetry
+      - cache-whisper:/root/.cache/whisper
+      - cache-faster-whisper:/root/.cache/faster_whisper
 
+volumes:
+  cache-pip:
+  cache-poetry:
+  cache-whisper:
+  cache-faster-whisper:

faster_whisper_model_conversion.sh

+#!/bin/bash
+
+ct2-transformers-converter --model openai/whisper-"$ASR_MODEL" --output_dir /root/.cache/faster_whisper --quantization float16

model_conversion.sh

-#!/bin/bash
-
-# TODO: Add step to build setup based on ENV variable
-ct2-transformers-converter --model openai/whisper-large-v2 --output_dir /root/.cache/faster_whisper --quantization float16