Add ROCm build for faster-whisper with Wyoming endpoint

Multi-stage Dockerfile builds CTranslate2 v4.0.0 with HIP/ROCm support
targeting gfx1030/gfx1031 (RX 6000 series), then installs faster-whisper
and wyoming-faster-whisper on top for a Wyoming ASR server on port 10300.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

docker-compose.yml

@@ -0,0 +1,29 @@
+services:
+  wyoming-faster-whisper:
+    build:
+      context: .
+      args:
+        ROCM_VERSION: "6.2"
+        GPU_ARCH: "gfx1030;gfx1031"
+    image: rocm-faster-whisper:latest
+    devices:
+      - /dev/kfd
+      - /dev/dri
+    group_add:
+      - video
+      - render
+    ports:
+      - "10300:10300"
+    volumes:
+      - ./data:/data
+    environment:
+      # Override GFX version if ROCm doesn't natively support your exact chip.
+      # gfx1031 (RX 6000-series variants) can be told to behave as gfx1030.
+      - HSA_OVERRIDE_GFX_VERSION=10.3.0
+    command: >
+      --uri tcp://0.0.0.0:10300
+      --data-dir /data
+      --model Systran/faster-distil-whisper-small.en
+      --device cuda
+      --compute-type float16
+    restart: unless-stopped