Update app.py

app.py

@@ -43,7 +43,7 @@ def fixed_get_imports(filename):
 
 
 import subprocess
-#subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
+subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
 
 def download_file(url, folder_path, filename):
     if not os.path.exists(folder_path):
@@ -125,13 +125,13 @@ DEFAULT_NEGATIVE_SUFFIX = "Nsfw oversaturated crappy_art low_quality blurry bad_
 # Initialize Florence model
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
-#def load_models():
-    #with patch("transformers.dynamic_module_utils.get_imports", fixed_get_imports):
-florence_model = AutoModelForCausalLM.from_pretrained('microsoft/Florence-2-base', trust_remote_code=True).to(device).eval()
-florence_processor = AutoProcessor.from_pretrained('microsoft/Florence-2-base', trust_remote_code=True)
-    #return florence_model, florence_processor
+def load_models():
+    with patch("transformers.dynamic_module_utils.get_imports", fixed_get_imports):
+        florence_model = AutoModelForCausalLM.from_pretrained('microsoft/Florence-2-base', trust_remote_code=True).to(device).eval()
+        florence_processor = AutoProcessor.from_pretrained('microsoft/Florence-2-base', trust_remote_code=True)
+    return florence_model, florence_processor
 
-#florence_model, florence_processor = load_models()
+florence_model, florence_processor = load_models()
 
 # Prompt Enhancer
 enhancer_medium = pipeline("summarization", model="gokaygokay/Lamini-Prompt-Enchance", device=device)