Update app.py

app.py

@@ -36,9 +36,9 @@ torch.backends.cuda.preferred_blas_library="cublas"
 
 hftoken = os.getenv("HF_AUTH_TOKEN") 
 
-image_encoder_path = "google/siglip-so400m-patch14-384"
-ipadapter_path = hf_hub_download(repo_id="InstantX/SD3.5-Large-IP-Adapter", filename="ip-adapter.bin")
-model_path = 'ford442/stable-diffusion-3.5-medium-bf16'
+#image_encoder_path = "google/siglip-so400m-patch14-384"
+#ipadapter_path = hf_hub_download(repo_id="InstantX/SD3.5-Large-IP-Adapter", filename="ip-adapter.bin")
+#model_path = 'ford442/stable-diffusion-3.5-medium-bf16'
 
 def upload_to_ftp(filename):
     try:
@@ -56,7 +56,7 @@ def upload_to_ftp(filename):
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 torch_dtype = torch.bfloat16
 
-checkpoint = "microsoft/Phi-3.5-mini-instruct"
+#checkpoint = "microsoft/Phi-3.5-mini-instruct"
 #vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
 #vae = AutoencoderKL.from_pretrained("ford442/sdxl-vae-bf16")
 #vae = AutoencoderKL.from_pretrained("ford442/sdxl-vae-bf16")
@@ -102,9 +102,9 @@ pipe.to(device=device, dtype=torch.bfloat16)
 #refiner.scheduler = EulerAncestralDiscreteScheduler.from_config(refiner.scheduler.config, beta_schedule="scaled_linear")
 #refiner.scheduler = EulerAncestralDiscreteScheduler.from_config(refiner.scheduler.config)
 
-tokenizer = AutoTokenizer.from_pretrained(checkpoint, add_prefix_space=True)
-tokenizer.tokenizer_legacy=False
-model = AutoModelForCausalLM.from_pretrained(checkpoint).to('cuda')
+#tokenizer = AutoTokenizer.from_pretrained(checkpoint, add_prefix_space=True)
+#tokenizer.tokenizer_legacy=False
+#model = AutoModelForCausalLM.from_pretrained(checkpoint).to('cuda')
 #model = torch.compile(model)
 
 upscaler_2 = UpscaleWithModel.from_pretrained("Kim2091/ClearRealityV1").to(torch.device("cuda:0"))
@@ -153,6 +153,7 @@ def infer(
     torch.set_float32_matmul_precision("highest")
     seed = random.randint(0, MAX_SEED)
     generator = torch.Generator(device='cuda').manual_seed(seed)
+    '''
     if expanded:
         system_prompt_rewrite = (
         "You are an AI assistant that rewrites image prompts to be more descriptive and detailed."
@@ -204,9 +205,10 @@ def infer(
         print('-- filtered prompt 2 --')
         print(enhanced_prompt_2)
     else:
-        enhanced_prompt = prompt
-        enhanced_prompt_2 = prompt
-    model.to('cpu')
+    '''
+    enhanced_prompt = prompt
+    enhanced_prompt_2 = prompt
+    #model.to('cpu')
     if latent_file:  # Check if a latent file is provided
       #  initial_latents = pipe.prepare_latents(
       #      batch_size=1,