Update app.py

app.py

@@ -13,10 +13,13 @@ import random
 import time
 from huggingface_hub import hf_hub_download
 from diffusers import FluxTransformer2DModel, FluxPipeline
+from diffusers import DiffusionPipeline, AutoencoderTiny, AutoencoderKL
 import safetensors.torch
 from safetensors.torch import load_file
-from transformers import CLIPModel, CLIPProcessor, CLIPConfig
+from transformers import CLIPModel, CLIPProcessor, CLIPTextModel, CLIPTokenizer, CLIPConfig, T5EncoderModel, T5Tokenizer
 import gc
+from gradio_client import Client
+Client = Client("AlekseyCalvin/soonfactory4", hf_token=os.getenv("HF_TOKEN"))
 
 cache_path = path.join(path.dirname(path.abspath(__file__)), "models")
 os.environ["TRANSFORMERS_CACHE"] = cache_path
@@ -28,23 +31,15 @@ torch.backends.cuda.matmul.allow_tf32 = True
 
 pipe = FluxPipeline.from_pretrained("AlekseyCalvin/HistoricColorSoonr_v2_FluxSchnell_Diffusers", ignore_mismatched_sizes=True, torch_dtype=torch.bfloat16)
 pipe.to(device="cuda", dtype=torch.bfloat16)
-clipmodel = 'norm' # 'norm', 'long' (my fine-tunes) - 'oai', 'orgL' (OpenAI / BeichenZhang original)
-
-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)
-config.text_config.max_position_embeddings = 77
 
+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
 
 
@@ -91,7 +86,7 @@ def update_selection(evt: gr.SelectData, width, height):
         height,
     )
 
-@spaces.GPU(duration=70)
+@spaces.GPU()
 
 def generate_image(prompt, trigger_word, steps, seed, cfg_scale, width, height, lora_scale, progress):
     pipe.to("cuda")