Monday, November 7th, 2022 - Saturday, November 12th, 2022
I supposed him to be a professor of deep learning, but he turned out to be a rookie. He even didn't finish the machine learning course.
Luckily, he is interested in the project and agreed to write reports every week.
As the d2l vision transformer program file uses PyTorch, I had first to achieve all functions using TensorFlow. At this stage, I encountered a problem. The batch_size in TensorFlow is dynamic and when I use the command to get the batch_size it always returns a None, which raises an error. I fixed it by using tf.config.run_functions_eagerly(True)
.
The code worked but the result was as terrible as shit. The loss and accuracy didn't change no matter how many epochs I set.
I had achieved applied the vision transformer using TensorFlow. Seeing hundreds of lines of code I wrote, I felt proud.
When doing a classification task on the Autism Dataset, I got shit!
Eddie He