MusIre commited on
Commit
15785c1
·
verified ·
1 Parent(s): dc1a793

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +6 -1
app.py CHANGED
@@ -88,10 +88,14 @@ scheduler = ReduceLROnPlateau(optimizer, mode='min', factor=0.5, patience=3, ver
88
  # Load SentenceTransformer model
89
  clip_model = SentenceTransformer('sentence-transformers/clip-ViT-B-32-multilingual-v1').to(device)
90
 
 
91
  model_name = "EleutherAI/gpt-neo-1.3B"
92
  tokenizer = AutoTokenizer.from_pretrained(model_name)
 
 
93
  model_gptneo = AutoModelForCausalLM.from_pretrained(model_name).to(device)
94
 
 
95
  def generate_description(image):
96
  image_resnet = data_transforms(image).unsqueeze(0).to(device)
97
 
@@ -123,13 +127,14 @@ def generate_description(image):
123
  top_p=0.9,
124
  repetition_penalty=1.2,
125
  do_sample=True,
126
- pad_token_id=tokenizer.eos_token_id
127
  )
128
 
129
  description_text = tokenizer.decode(output[0], skip_special_tokens=True)
130
 
131
  return predicted_style, predicted_artist, description_text
132
 
 
133
  # Gradio interface
134
  def gradio_interface(image):
135
  if image is None:
 
88
  # Load SentenceTransformer model
89
  clip_model = SentenceTransformer('sentence-transformers/clip-ViT-B-32-multilingual-v1').to(device)
90
 
91
+ # Load GPT-Neo and set padding token
92
  model_name = "EleutherAI/gpt-neo-1.3B"
93
  tokenizer = AutoTokenizer.from_pretrained(model_name)
94
+ if tokenizer.pad_token is None:
95
+ tokenizer.pad_token = tokenizer.eos_token # Set pad_token to eos_token
96
  model_gptneo = AutoModelForCausalLM.from_pretrained(model_name).to(device)
97
 
98
+
99
  def generate_description(image):
100
  image_resnet = data_transforms(image).unsqueeze(0).to(device)
101
 
 
127
  top_p=0.9,
128
  repetition_penalty=1.2,
129
  do_sample=True,
130
+ pad_token_id=tokenizer.pad_token_id
131
  )
132
 
133
  description_text = tokenizer.decode(output[0], skip_special_tokens=True)
134
 
135
  return predicted_style, predicted_artist, description_text
136
 
137
+
138
  # Gradio interface
139
  def gradio_interface(image):
140
  if image is None: