Need help..error running sample prediction on KTH dataset

[AvaniPitre]

Hi @alexlee-gk

I am trying to workout pretrained model on kth dataset, I have successfully downloaded and processed dataset by running following command
bash data/download_and_preprocess_dataset.sh kth

Also downloaded pretrained model
bash pretrained_models/download_model.sh kth ours_savp

and now trying to run sample prediction on kth dataset but getting following batch size error, command is

CUDA_VISIBLE_DEVICES=0 python scripts/generate.py --input_dir data/kth --dataset_hparams sequence_length=30 --checkpoint pretrained_models/kth/ours_savp --mode test --results_dir results_test_samples/kth

Error is
File "scripts/generate.py", line 193, in
main()
File "scripts/generate.py", line 130, in main
raise ValueError('batch_size should evenly divide the dataset size %d' % num_examples_per_epoch)
ValueError: batch_size should evenly divide the dataset size 819

I am confused whats went wrong..?
I am newbie in deep learning and Tensor-flow..
Please help
also want to try your models on my own dataset.. need your guidance
Thanks in advance
Avani

[wangwen39]

You need to change the batch_size=9 ,maybe it will work

[AvaniPitre]

ok thanks will try that

[AvaniPitre]

thanks that batch issue resolved but encountered other error for tensor shape mismatch..

InvalidArgumentError (see above for traceback): Assign requires shapes of both tensors to match. lhs shape= [4,4,2,64] rhs shape= [4,4,6,64]
[[node save/Assign_1 (defined at /home/video_prediction/video_prediction/utils/tf_utils.py:542)

please help
thanks

[NagabhushanSN95]

Check this:
issue 22

Same problem. Uncomment line 118 in video_prediction/datasets/kth_dataset.py and rerun download_and_preprocess_dataset.sh

I used batch-size=8 and num-samples=816. It has worked for me.

[malalejandra]

Hi, I switched to tf 1.10 and at least testing is working

[staniPetrox]

Hi, I switched to tf 1.10 and at least testing is working

Only thing that worked for me. Being in 2021