error in Model.fit_generator

[kodagsourabh]

getting Model.fit_generator error on google colab but on local pc its working properly

[kodagsourabh]

i can help it..

Thanks please help me

[kodagsourabh]

def data_generator(batch_size = 32):
  partial_caps = []
  next_words = []
  images = []
  df = pd.read_csv('flickr8k_training_dataset.txt', delimiter='\t')
  df = df.sample(frac=1)
  iter = df.iterrows()
  c = []
  imgs = []
  for i in range(df.shape[0]):
    x = next(iter)
    c.append(x[1][1])
    imgs.append(x[1][0])

  count = 0
  while True:
    for j, text in enumerate(c):
      current_image = encoding_train[imgs[j]]
      for i in range(len(text.split())-1):
        count += 1
        partial = [word2idx[txt] for txt in text.split()[:i+1]]
        partial_caps.append(partial)
        n = np.zeros(vocab_size)
        n[word2idx[text.split()[i+1]]] = 1
        next_words.append(n)
        images.append(current_image)

      if count>=batch_size:
        next_words = np.asarray(next_words)
        images = np.asarray(images)
        partial_caps = sequence.pad_sequences(partial_caps, maxlen=max_len, padding=
        'post')
        yield ([images, partial_caps], next_words) # change is here
        partial_caps = []
        next_words = []
        images = []
        count = 0

same error

Changed the below function to yours given function but getting the same error it is in load_data.py which is in utils folder

def data_generator(images, captions, tokenizer, max_length, batch_size, random_seed):
  # Setting random seed for reproducibility of results
  random.seed(random_seed)
  # Image ids
  image_ids = list(captions.keys())
  _count=0
  assert batch_size<= len(image_ids), 'Batch size must be less than or equal to {}'.format(len(image_ids))
  while True:
    if _count >= len(image_ids):
      # Generator exceeded or reached the end so restart it
      _count = 0
    # Batch list to store data
    input_img_batch, input_sequence_batch, output_word_batch = list(), list(), list()
    for i in range(_count, min(len(image_ids), _count+batch_size)):
      # Retrieve the image id
      image_id = image_ids[i]
      # Retrieve the image features
      image = images[image_id][0]
      # Retrieve the captions list
      captions_list = captions[image_id]
      # Shuffle captions list
      random.shuffle(captions_list)
      input_img, input_sequence, output_word = create_sequences(tokenizer, max_length, captions_list, image)
      # Add to batch
      for j in range(len(input_img)):
        input_img_batch.append(input_img[j])
        input_sequence_batch.append(input_sequence[j])
        output_word_batch.append(output_word[j])
    _count = _count + batch_size
    yield [[np.array(input_img_batch), np.array(input_sequence_batch)], np.array(output_word_batch)]

[Manuj229]

any solutions?

[kodagsourabh]

Yes there was version issues

…

On Wed, 7 Jul 2021, 1:26 pm Manuj Vyas, ***@***.***> wrote: any solutions? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#11 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEU4D7WH32CSFQZSIAZS25DTWQCEJANCNFSM4XUIGMYQ> .

[Manuj229]

I am still having same issue. Please provide a solution.

[kodagsourabh]

Try this version or else try stack overflow you will find some solutions cuda 10.0 cudnn 7.4 tensorflow 1.13.1 keras 2.2.4

…

On Wed, 7 Jul 2021, 5:01 pm Manuj Vyas, ***@***.***> wrote: I am still having same issue. Please provide a solution. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#11 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEU4D7SEY6HKP3UZSWBIVYLTWQ3JXANCNFSM4XUIGMYQ> .