add video summarization pre-trained model

app.py

@@ -8,14 +8,12 @@ import numpy as np
 from inference import Inference
 import utils
 
-model_checkpoint = 'saved_model'
 encoder_model_name = 'google/vit-large-patch32-224-in21k'
 decoder_model_name = 'gpt2'
 frame_step = 300
 
 inference = Inference(
     decoder_model_name=decoder_model_name,
-    model_checkpoint=model_checkpoint,
 )
 
 model = SentenceTransformer('all-mpnet-base-v2')

inference.py

@@ -1,14 +1,14 @@
 import torch
-from transformers import AutoTokenizer, VisionEncoderDecoderModel
+from transformers import AutoTokenizer, AutoModel
 
 import utils
 
 class Inference:
-  def __init__(self, decoder_model_name, model_checkpoint, max_length=32):
+  def __init__(self, decoder_model_name, max_length=32):
     self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 
     self.tokenizer = AutoTokenizer.from_pretrained(decoder_model_name)
-    self.encoder_decoder_model = VisionEncoderDecoderModel.from_pretrained(model_checkpoint)
+    self.encoder_decoder_model = AutoModel.from_pretrained('armgabrielyan/video-summarization')
     self.encoder_decoder_model.to(self.device)
 
     self.max_length = max_length