Update Dockerfile

Dockerfile

@@ -25,10 +25,24 @@ COPY --chown=user . /app
 RUN pip install --no-cache-dir \
     torch==2.3.1+cu121 torchaudio==2.3.1 \
     --index-url https://download.pytorch.org/whl/cu121
-    
-RUN pip install text-generation==0.6.1
-ENV ORPHEUS_API_URL=http://127.0.0.1:1234/v1/completions
+
 EXPOSE 7860
-CMD bash -c "text-generation-launcher --model-id SebastianBodza/Kartoffel_Orpheus-3B_german_natural-v0.1 --port 1234 & \
-             uvicorn app:app --host 0.0.0.0 --port 7860"
+
+RUN pip install --no-cache-dir text-generation==0.6.1
+
+ENV HF_HOME=/app/.cache
+ENV MODEL_ID="SebastianBodza/Kartoffel_Orpheus-3B_german_natural-v0.1"
+ENV ORPHEUS_API_URL=http://127.0.0.1:1234/v1/completions
+ENV ORPHEUS_MODEL=$MODEL_ID         # landet im JSON-Payload von OrpheusEngine
+
+CMD bash -c "\
+   text-generation-launcher \
+        --model-id $MODEL_ID \
+        --port 1234 \
+        --max-input-length 4096 \
+        --dtype auto \
+        --quantization fp8 \
+        --enable-chunked-prefill \
+        --gpu-memory-utilization 0.7 & \
+   uvicorn app:app --host 0.0.0.0 --port 7860"