Update app.py

app.py

@@ -1,20 +1,21 @@
 import requests
+import asyncio
 
 from PIL import Image
 from transformers import AutoProcessor, AutoModelForVision2Seq
 
 
-model = AutoModelForVision2Seq.from_pretrained("microsoft/kosmos-2-patch14-224")
-processor = AutoProcessor.from_pretrained("microsoft/kosmos-2-patch14-224")
+model = await AutoModelForVision2Seq.from_pretrained("microsoft/kosmos-2-patch14-224")
+processor = await AutoProcessor.from_pretrained("microsoft/kosmos-2-patch14-224")
 
 # The original Kosmos-2 demo saves the image first then reload it. For some images, this will give slightly different image input and change the generation outputs.
 
 #prompt = "{question}"
 
-def describe_image(image_path, question : str):
-  inputs = processor(text=question, images=image_path, return_tensors="pt")
+async def describe_image(image_path, question : str):
+  inputs = await processor(text=question, images=image_path, return_tensors="pt")
 
-  generated_ids = model.generate(
+  await generated_ids = model.generate(
       pixel_values=inputs["pixel_values"],
       input_ids=inputs["input_ids"],
       attention_mask=inputs["attention_mask"],
@@ -23,12 +24,12 @@ def describe_image(image_path, question : str):
       use_cache=True,
       max_new_tokens=128,
   )
-  generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
+  generated_text = await processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
 
   # Specify `cleanup_and_extract=False` in order to see the raw model generation.
-  processed_text = processor.post_process_generation(generated_text, cleanup_and_extract=False)
+  processed_text = await processor.post_process_generation(generated_text, cleanup_and_extract=False)
 
-  processed_text, entities = processor.post_process_generation(generated_text)
+  processed_text, entities = await processor.post_process_generation(generated_text)
 
   return processed_text