AnonymousSub commited on
Commit
627fbe3
1 Parent(s): df847c3

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +4 -4
app.py CHANGED
@@ -38,7 +38,7 @@ def generate_answer_git(processor, model, image, question):
38
  input_ids = [processor.tokenizer.cls_token_id] + input_ids
39
  input_ids = torch.tensor(input_ids).unsqueeze(0)
40
 
41
- generated_ids = model.generate(pixel_values=pixel_values, input_ids=input_ids, max_length=256)#50)
42
  generated_answer = processor.batch_decode(generated_ids, skip_special_tokens=True)
43
 
44
  return generated_answer
@@ -48,7 +48,7 @@ def generate_answer_blip(processor, model, image, question):
48
  # prepare image + question
49
  inputs = processor(images=image, text=question, return_tensors="pt")
50
 
51
- generated_ids = model.generate(**inputs, max_length=256)#50)
52
  generated_answer = processor.batch_decode(generated_ids, skip_special_tokens=True)
53
 
54
  return generated_answer
@@ -56,10 +56,10 @@ def generate_answer_blip(processor, model, image, question):
56
 
57
  def generate_answer_vilt(processor, model, image, question):
58
  # prepare image + question
59
- encoding = processor(images=image, text=question, return_tensors="pt")
60
 
61
  with torch.no_grad():
62
- outputs = model(**encoding, max_length=256)
63
 
64
  predicted_class_idx = outputs.logits.argmax(-1).item()
65
 
 
38
  input_ids = [processor.tokenizer.cls_token_id] + input_ids
39
  input_ids = torch.tensor(input_ids).unsqueeze(0)
40
 
41
+ generated_ids = model.generate(pixel_values=pixel_values, input_ids=input_ids, max_length=128)#50)
42
  generated_answer = processor.batch_decode(generated_ids, skip_special_tokens=True)
43
 
44
  return generated_answer
 
48
  # prepare image + question
49
  inputs = processor(images=image, text=question, return_tensors="pt")
50
 
51
+ generated_ids = model.generate(**inputs, max_length=128)#50)
52
  generated_answer = processor.batch_decode(generated_ids, skip_special_tokens=True)
53
 
54
  return generated_answer
 
56
 
57
  def generate_answer_vilt(processor, model, image, question):
58
  # prepare image + question
59
+ encoding = processor(images=image, text=question, max_length=128, return_tensors="pt")
60
 
61
  with torch.no_grad():
62
+ outputs = model(**encoding)
63
 
64
  predicted_class_idx = outputs.logits.argmax(-1).item()
65