remove flash_attn from requirements

Dockerfile

@@ -18,7 +18,6 @@ WORKDIR /app
 # Copy backend requirements
 COPY --chown=user backend/requirements.txt /app/requirements.txt
 RUN pip install --no-cache-dir --upgrade -r /app/requirements.txt
-RUN pip install --no-cache-dir einops flash_attn
 
 # Copy backend and frontend
 COPY --chown=user backend/ /app/backend/

README.md

@@ -33,7 +33,12 @@ Check out the configuration reference at https://huggingface.co/docs/hub/spaces-
 
 3. Install dependencies
 
-4. Run the app
+4. If you're running on a GPU device, which enables more models options, install the GPU related packages:
+   ```commandline
+   python install_gpu_packages.py
+   ```
+
+5.Run the app
     ```commandline
     uvicorn app:app --reload --host 0.0.0.0 --port 8000
     ```

install_gpu_packages.py

@@ -0,0 +1,10 @@
+import subprocess
+import sys
+import torch
+
+if torch.cuda.is_available():
+    print("GPU detected. Installing GPU packages...")
+    subprocess.check_call([sys.executable, "-m", "pip", "install", "flash_attn", "einops"])
+else:
+    print("No GPU detected. Installing CPU-only packages if needed...")
+    subprocess.check_call([sys.executable, "-m", "pip", "install", "einops"])