Update docker image to CUDA 12.3.2 and CUDNN 9

Dockerfile

@@ -1,5 +1,5 @@
 ##########################################
-FROM nvidia/cuda:11.8.0-cudnn8-runtime-ubuntu22.04
+FROM nvidia/cuda:12.3.2-cudnn9-runtime-ubuntu22.04
 
 ARG WHISPER_VERSION='2.4.0'
 
@@ -42,7 +42,7 @@ RUN \
         build-essential \
         python3-dev &&\
     \
-    rm -rf /var/lib/apt/lists/*
+    rm -rf /root/.cache/* /var/lib/apt/lists/*
 
 COPY run.sh .
 

README.md

@@ -9,9 +9,7 @@ https://github.com/rhasspy/wyoming-faster-whisper
 docker pull ghcr.io/slackr31337/wyoming-whisper-gpu:latest
 
 
-Default model: tiny-int8
-
-Example models:
+Models:
 
 - tiny-int8
 - tiny.en
@@ -34,15 +32,16 @@ Example models:
 - distil-small.en
 
 
-Use environment variable to set model
-
 Environment variables:
 
-> MODEL=base-int8
+> MODEL=base-int8 (Name of faster-whisper model to use)
 >
-> LANGUAGE=en
+> LANGUAGE=en (Default language to set for transcription)
+>
+> COMPUTE_TYPE=int8 (float16, int8, etc.)
+>
+> BEAM_SIZE=5 (Size of beam during decoding. 0 for auto)
 >
-
 
 # Docker compose
 
@@ -52,6 +51,9 @@ Environment variables:
     container_name: wyoming-whisper
     environment:  
       - MODEL=base-int8
+      - LANGUAGE=en
+      - COMPUTE_TYPE=int8
+      - BEAM_SIZE=5
     ports:  
       - 10300:10300
     volumes:  

run.sh

@@ -1,10 +1,16 @@
 #!/usr/bin/env bash
 
+if [ -z "${HF_HUB_CACHE}" ];then
+    HF_HUB_CACHE=/tmp
+fi
+
 /app/bin/python3 -m  wyoming_faster_whisper \
     --uri 'tcp://0.0.0.0:10300' \
     --data-dir /data \
     --download-dir /data \
-    --model "${MODEL:-tiny-int8}" \
+    --model "${MODEL:-auto}" \
     --language "${LANGUAGE:-en}" \
+    --compute-type "${COMPUTE_TYPE:-int8}" \
+    --beam-size "${BEAM_SIZE:-5}" \
     --device cuda \
     "$@"