allow os kill

main.py

@@ -216,7 +216,7 @@ def generate(request: GenerationRequest, authorized: bool = Depends(verify_token
         if "CUDA out of memory" in e_s or \
                 "an illegal memory access" in e_s or "CUDA" in e_s:
             logger.error("GPU error, committing seppuku.")
-            # os.kill(mainpid, signal.SIGTERM)
+            os.kill(mainpid, signal.SIGTERM)
         return {"error": str(e)}
 
 @app.post('/generate', response_model=Union[GenerationOutput, ErrorOutput])
@@ -267,7 +267,7 @@ def generate(request: GenerationRequest, authorized: bool = Depends(verify_token
         if "CUDA out of memory" in e_s or \
                 "an illegal memory access" in e_s or "CUDA" in e_s:
             logger.error("GPU error, committing seppuku.")
-            # os.kill(mainpid, signal.SIGTERM)
+            os.kill(mainpid, signal.SIGTERM)
         return {"error": str(e)}
 
 @app.post('/generate-text', response_model=Union[TextOutput, ErrorOutput])
@@ -294,7 +294,7 @@ def generate_text(request: TextRequest, authorized: bool = Depends(verify_token)
         if "CUDA out of memory" in e_s or \
                 "an illegal memory access" in e_s or "CUDA" in e_s:
             logger.error("GPU error, committing seppuku.")
-            # os.kill(mainpid, signal.SIGTERM)
+            os.kill(mainpid, signal.SIGTERM)
         return ErrorOutput(error=str(e))
 
 @app.get('/predict-tags', response_model=Union[TagOutput, ErrorOutput])
@@ -321,7 +321,7 @@ async def predict_tags(prompt="", authorized: bool = Depends(verify_token)):
         if "CUDA out of memory" in e_s or \
                 "an illegal memory access" in e_s or "CUDA" in e_s:
             logger.error("GPU error, committing seppuku.")
-            # os.kill(mainpid, signal.SIGTERM)
+            os.kill(mainpid, signal.SIGTERM)
         return ErrorOutput(error=str(e))
 
 #@app.get('/')