From bc744d5f51dffd0f7e9b8142e088f053e847475e Mon Sep 17 00:00:00 2001 From: TPTBusiness Date: Mon, 4 May 2026 20:31:11 +0200 Subject: [PATCH] fix: skip Kronos factor on GPUs < 20GB to avoid CUDA OOM (shared with llama-server) --- predix.py | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-) diff --git a/predix.py b/predix.py index 328a3399..a4d438dd 100644 --- a/predix.py +++ b/predix.py @@ -329,8 +329,15 @@ def quant( threading.Thread(target=start_cli_dash, daemon=True).start() time.sleep(1) - # ---- Kronos Factor: auto-generate if not in pool ---- - _ensure_kronos_factor_in_pool(console) + # ---- Kronos Factor: skip if GPU unavailable (CUDA OOM with llama-server) ---- + try: + import torch + if torch.cuda.is_available() and torch.cuda.get_device_properties(0).total_memory > 20 * 1024**3: + _ensure_kronos_factor_in_pool(console) + else: + console.print("[dim]Kronos Factor skipped — GPU < 20GB or CUDA unavailable[/dim]") + except Exception: + console.print("[dim]Kronos Factor skipped — torch not available[/dim]") # ---- Start fin_quant ---- from rdagent.app.qlib_rd_loop.quant import main as fin_quant