Spaces:

prithivMLmods
/

Vision-to-VibeVoice-en

Running on Zero

prithivMLmods commited on 8 days ago

Commit

4f6fb8d

verified ·

1 Parent(s): 94df60b

update app

Files changed (1) hide show

app.py CHANGED Viewed

@@ -144,7 +144,7 @@ print(f"Loading OCR Model: {QWEN_VL_MODEL_ID}...")
 qwen_processor = AutoProcessor.from_pretrained(QWEN_VL_MODEL_ID, trust_remote_code=True)
 qwen_model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     QWEN_VL_MODEL_ID,
-    attn_implementation="flash_attention_2",
     trust_remote_code=True,
     torch_dtype=torch.float16
 ).to(device).eval()

 qwen_processor = AutoProcessor.from_pretrained(QWEN_VL_MODEL_ID, trust_remote_code=True)
 qwen_model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     QWEN_VL_MODEL_ID,
+    #attn_implementation="flash_attention_2",
     trust_remote_code=True,
     torch_dtype=torch.float16
 ).to(device).eval()