From 978cffd6f55fd8574b26b12b7ec23fc63c9a5a14 Mon Sep 17 00:00:00 2001 From: Vitor Araujo Date: Mon, 30 Mar 2026 00:07:40 -0300 Subject: [PATCH] [fix] add --ub 16 to run_server - A ubatch_size greater than 16 causes a segmentation fault. fixes: https://github.com/microsoft/BitNet/issues/517 --- run_inference_server.py | 1 + 1 file changed, 1 insertion(+) diff --git a/run_inference_server.py b/run_inference_server.py index 9b0f10d53..3e6a98e72 100644 --- a/run_inference_server.py +++ b/run_inference_server.py @@ -29,6 +29,7 @@ def run_server(): '-t', str(args.threads), '-n', str(args.n_predict), '-ngl', '0', + '-ub', '16', '--temp', str(args.temperature), '--host', args.host, '--port', str(args.port),