diff --git a/backends/gaudi/server/text_generation_server/tgi_service.py b/backends/gaudi/server/text_generation_server/tgi_service.py index 18e88a7e..12317127 100644 --- a/backends/gaudi/server/text_generation_server/tgi_service.py +++ b/backends/gaudi/server/text_generation_server/tgi_service.py @@ -31,6 +31,7 @@ def main(args): trust_remote_code=args.trust_remote_code, uds_path=args.uds_path, max_input_tokens=args.max_input_tokens, + kv_cache_dtype="auto", )