From 4fec36c3cc00bd649dfb3c9d7e9079b5c8685a0e Mon Sep 17 00:00:00 2001 From: yum Date: Sun, 9 Jun 2024 16:43:34 -0700 Subject: Bump CUDNN to v8.9.7 Also disable flash-attention when CPU mode is selected --- Scripts/transcribe_v2.py | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'Scripts') diff --git a/Scripts/transcribe_v2.py b/Scripts/transcribe_v2.py index 2f37945..1bdc487 100644 --- a/Scripts/transcribe_v2.py +++ b/Scripts/transcribe_v2.py @@ -423,11 +423,12 @@ class Whisper: model_device = "cuda" if cfg["use_cpu"]: model_device = "cpu" + if cfg["use_flash_attention"]: + print(f"Flash attention disabled on CPU", file=sys.stderr) + cfg["use_flash_attention"] = False already_downloaded = os.path.exists(model_root) - print(f"Use flash attention {cfg['use_flash_attention']}") - self.model = WhisperModel(model_str, device = model_device, device_index = cfg["gpu_idx"], -- cgit v1.2.3