update for cpu

Dockerfile

@@ -2,7 +2,7 @@ FROM pytorch/pytorch:2.2.2-cuda12.1-cudnn8-devel
 
 WORKDIR /app
 
-# Install system dependencies
+# Common system dependencies
 RUN apt-get update && apt-get install -y \
     libportaudio2 \
     libportaudiocpp0 \
@@ -10,22 +10,41 @@ RUN apt-get update && apt-get install -y \
     libasound-dev \
     libsndfile1-dev \
     kmod \
-    build-essential \
     && rm -rf /var/lib/apt/lists/*
 
+# Conditional CUDA configuration for AMD64 only
+ARG TARGETARCH
+RUN if [ "$TARGETARCH" = "amd64" ]; then \
+    apt-get update && apt-get install -y \
+    ocl-icd-opencl-dev \
+    libopenblas-dev \
+    && rm -rf /var/lib/apt/lists/*; \
+    fi
+
+# Mac-specific audio fixes
+RUN if [ "$TARGETARCH" = "arm64" ]; then \
+    apt-get update && apt-get install -y \
+    libomp5 \
+    libopenblas0-pthread \
+    && rm -rf /var/lib/apt/lists/*; \
+    fi
+
 # Update pip first
 RUN pip install --upgrade pip setuptools wheel
 
 # Install base requirements first
 RUN pip install numpy==1.26.4
 
-# Install vLLM with CUDA 12.1 support
-RUN pip install vllm==0.6.4.post1 \
-    --extra-index-url https://pypi.nvidia.com \
-    --extra-index-url https://download.pytorch.org/whl/cu121
+# Install appropriate PyTorch and TorchAudio versions
+RUN if [ "$TARGETARCH" = "amd64" ]; then \
+    pip install torchaudio==2.2.2 --index-url https://download.pytorch.org/whl/cu121; \
+    else \
+    pip install torchaudio==2.2.2 --index-url https://download.pytorch.org/whl/cpu; \
+    fi
 
+# Install remaining requirements
 COPY . .
-RUN pip install -r requirements.txt
+RUN pip install --no-cache-dir -r requirements.txt
 
 EXPOSE 7860
 CMD ["python", "app.py"]

tts_ui/tts/auralis_tts_engine.py

@@ -14,6 +14,7 @@ import hashlib
 import torchaudio
 import time
 from pathlib import Path
+import os
 
 # Loading the TTS engine first and assign it to the class.
 # This looks ugly, but it works
@@ -24,6 +25,11 @@ model_path = "AstraMindAI/xttsv2"  # change this if you have a different model
 gpt_model = "AstraMindAI/xtts2-gpt"
 
 try:
+    cpu_only = torch.cuda.is_available()
+    if cpu_only:
+        os.environ["VLLM_NO_GPU"] = "1"
+        os.environ["TRITON_CPU_ONLY"] = "1"
+
     tts: TTS = tts.from_pretrained(
         model_name_or_path=model_path,
         gpt_model=gpt_model,
@@ -31,6 +37,7 @@ try:
         max_seq_len_to_capture=4096,  # Match WSL2 page size
         scheduler_max_concurrency=4,
     )
+
     logger.info(f"Successfully loaded model {model_path}")
 except Exception as e:
     error_msg = f"Failed to load model: {e}."