Push to ghcr (#2)

* push to ghcr

* typing-extensions>=4.8.0, use legacy

* pr should try to build it too

* pr should try to build it too

* pr should try to build it too

* only push main builds

* only push main builds

* only push main builds

* only push main builds

* only push main builds

* use workdir

* use workdir

* use workdir

* use workdir

* use workdir

* use workdir

* use workdir

* use workdir

* fixes for comments

* fixes main build cache ref
pr will use cache but will not update

* add some logs

* use numpy 1.24.3 (works locally)

* pr caching

* pr caching

* pr number as cache, remove tags on pr

* multiple cache

* multiple cache

* multiple cache

* NUM_THREADS env var

.github/workflows/build-and-push-to-ghcr.yml

@@ -0,0 +1,46 @@
+name: Build and push to GHCR
+on:
+  push:
+    branches: [main]
+  pull_request:
+env:
+  GHCR_URL: ghcr.io/coqui-ai/xtts-streaming-server
+jobs:
+  build-and-push-to-ghcr:
+    runs-on: ubuntu-22.04
+    steps:
+      -
+        name: Checkout
+        uses: actions/checkout@v3
+
+      -
+        name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: 'Login to GitHub Container Registry'
+        run: |
+          set -xe
+          docker login --username ${{ github.actor }} --password ${{ secrets.GITHUB_TOKEN }} ghcr.io
+
+      - name: Build only for PR
+        if: github.ref != 'refs/heads/main'
+        uses: docker/build-push-action@v5
+        with:
+          context: "{{defaultContext}}:server"
+          file: Dockerfile
+          push: false # Do not push image for PR
+          cache-from: type=registry,ref=ghcr.io/coqui-ai/xtts-streaming-server:cache-latest;type=registry,ref=ghcr.io/coqui-ai/xtts-streaming-server:cache-pr-${{ github.event.number }}
+          cache-to: type=registry,ref=ghcr.io/coqui-ai/xtts-streaming-server:cache-pr-${{ github.event.number }}
+
+      - name: Build and Push image
+        if: github.ref == 'refs/heads/main'
+        uses: docker/build-push-action@v5
+        with:
+          context: "{{defaultContext}}:server"
+          file: Dockerfile
+          push: true # Push if merged
+          cache-from: type=registry,ref=ghcr.io/coqui-ai/xtts-streaming-server:cache-latest
+          cache-to: type=registry,ref=ghcr.io/coqui-ai/xtts-streaming-server:cache-latest
+          tags: coqui-ai/xtts-streaming-server:latest, coqui-ai/xtts-streaming-server:${{ env.RELEASE_VERSION }}-${{ github.sha }}
+          #build-args:
+

server/Dockerfile

@@ -5,5 +5,6 @@ RUN python -m pip install -r requirements.txt --use-deprecated=legacy-resolver
 
 COPY main.py .
 
+ENV NUM_THREADS=8
 EXPOSE 80
 CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "80"]

server/main.py

@@ -18,18 +18,22 @@
 from TTS.utils.generic_utils import get_user_data_dir
 from TTS.utils.manage import ModelManager
 
-torch.set_num_threads(8)
+torch.set_num_threads(int(os.environ.get("NUM_THREADS", "8")))
 device = torch.device("cuda")
 
 model_name = "tts_models/multilingual/multi-dataset/xtts_v1.1"
+print("Downloading XTTS Model:",model_name)
 ModelManager().download_model(model_name)
 model_path = os.path.join(get_user_data_dir("tts"), model_name.replace("/", "--"))
+print("XTTS Model downloaded")
 
+print("Loading XTTS")
 config = XttsConfig()
 config.load_json(os.path.join(model_path, "config.json"))
 model = Xtts.init_from_config(config)
 model.load_checkpoint(config, checkpoint_dir=model_path, eval=True, use_deepspeed=True)
 model.to(device)
+print("XTTS Loaded.")
 
 ##### Run fastapi #####
 app = FastAPI(

server/requirements.txt

@@ -1,6 +1,8 @@
-TTS==0.18.2
+TTS==0.19.1
 uvicorn[standard]==0.23.2
-fastapi==0.104.0
-deepspeed==0.8.3
+fastapi==0.95.2
+deepspeed==0.10.3
 pydantic==1.10.13
 python-multipart==0.0.6
+typing-extensions>=4.8.0
+numpy==1.24.3