Why is there an optimizer in the test.py file? about gpt2-chest-x-ray-report-generation HOT 4 CLOSED

Hello,
I send the optimizer because the TensorFlow version that I started with required an optimizer to be passed.
CheckpointManager takes a ckpt_path and the model weights are loaded in this part:

ckpt.restore(ckpt_manager.latest_checkpoint)

I hope this helps.

from gpt2-chest-x-ray-report-generation.

Hi, thanks for your reply!

I still have a few doubts, I am new to tensorflow.

I downloaded the checkpoint from the drive link and placed it in an appropriate folder, gave the path to that folder in configs.py under ckpt_path. (Used absolute paths)

When I run the code, I get the following error:
tensorflow.python.framework.errors_impl.NotFoundError: Unsuccessful TensorSliceReader constructor: Failed to find any matching files for /content/drive/MyDrive/new/GPT2-Chest-X-Ray-Report-Generation/ckpt-27

I used the 'checkpoint_name' function parameter of the CheckpointManager to redirect it to the right filename, but now my latest_checkpoint is None. What could be the reason?

from gpt2-chest-x-ray-report-generation.

Hey,

I just tried it on googleColab using TF 2.1 and it worked normally. I just pointed to the folder containing the checkpoint in config.py under ckpt_path using absolute paths like the following:

/content/drive/My Drive/colab/GPT2-Chest-X-Ray-Report-Generation/checkpoints/paper-ckpt

or a relative path

./checkpoints/paper-ckpt

No need to use 'checkpoint_name' as CheckpointManager automatically fetches the latest checkpoint from the folder. If it's still not working try to check if the folder is readable from your directory using

import os

if os.path.exists('path_to_folder_or_file'):
   print("found")

from gpt2-chest-x-ray-report-generation.

Hey, the issue is resolved. I just picked the 'checkpoint' file from the drive rather than the whole folder.

Thanks a lot.

from gpt2-chest-x-ray-report-generation.