Update app.py

app.py

@@ -45,22 +45,14 @@ torch.cuda.empty_cache()
 
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
-clipmodel = 'norm'
-if clipmodel == "long":
-    model_id = "zer0int/LongCLIP-GmP-ViT-L-14"
-    config = CLIPConfig.from_pretrained(model_id)
-    maxtokens = 77
-if clipmodel == "norm":
-    model_id = "zer0int/CLIP-GmP-ViT-L-14"
-    config = CLIPConfig.from_pretrained(model_id)
-    maxtokens = 77
-clip_model = CLIPModel.from_pretrained(model_id, torch_dtype=torch.bfloat16, config=config, ignore_mismatched_sizes=True).to("cuda")
-clip_processor = CLIPProcessor.from_pretrained(model_id, padding="max_length", max_length=maxtokens, ignore_mismatched_sizes=True, return_tensors="pt", truncation=True)
-#t5 = HFEmbedder("DeepFloyd/t5-v1_1-xxl", max_length=512, torch_dtype=torch.bfloat16).to(device)
-
+model_id = ("zer0int/LongCLIP-GmP-ViT-L-14")
+config = CLIPConfig.from_pretrained(model_id)
+config.text_config.max_position_embeddings = 248
+clip_model = CLIPModel.from_pretrained(model_id, torch_dtype=torch.bfloat16, config=config, ignore_mismatched_sizes=True)
+clip_processor = CLIPProcessor.from_pretrained(model_id, padding="max_length", max_length=248)
 pipe.tokenizer = clip_processor.tokenizer
 pipe.text_encoder = clip_model.text_model
-pipe.tokenizer_max_length = maxtokens
+pipe.tokenizer_max_length = 248
 pipe.text_encoder.dtype = torch.bfloat16
 #pipe.text_encoder_2 = t5.text_model