From 79d8d2cd96193617fbb3f15d346227e72f0561db Mon Sep 17 00:00:00 2001
From: mARTin-B78 <mail@martin-bierschenk.de>
Date: Mon, 22 Jun 2026 11:17:39 +0200
Subject: [PATCH] fix(customvoice): override max_seq_len to 1024 to prevent
 CUDA OOM

---
 config/run_customvoice_server.py | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/config/run_customvoice_server.py b/config/run_customvoice_server.py
index ec2469b..b42220e 100644
--- a/config/run_customvoice_server.py
+++ b/config/run_customvoice_server.py
@@ -252,6 +252,8 @@ def main():
     parser.add_argument("--device", default="cuda")
     parser.add_argument("--max-seq-len", type=int, default=2048)
     args = parser.parse_args()
+    # Force a smaller max_seq_len to save VRAM and prevent OOM
+    args.max_seq_len = 1024
     DEFAULT_MAX_NEW_TOKENS = args.max_seq_len
     _load_model_kwargs = args