feat: enable flash_attention_2 for Qwen3-TTS

2026-04-07 03:22:48 -07:00 · 2026-04-07 03:22:48 -07:00 · b75e813c03
parent 78bc3f71c0
commit b75e813c03
1 changed files with 1 additions and 0 deletions
--- a/modules/tts/qwen3_tts.py
+++ b/modules/tts/qwen3_tts.py
@ -41,6 +41,7 @@ class TTSProvider(TTSProviderBase):
            model_path,
            device_map=device,
            dtype=dtype,
+            attn_implementation="flash_attention_2",
        )
        self.tokenizer = Qwen3TTSTokenizer.from_pretrained(tokenizer_path)