ValueError: Cannot feed value of shape (4096,) for Tensor 'cudnn_lstm/rnn/multi_rnn_cell/cell_0/cudnn_compatible_lstm_cell/bias/Initializer/Const:0', which has shape '(8192,)'

I want to evaluate the pretrained model(greedy search without language model). This is the code I execute
./evaluate.py --export_dir ../deepspeech-0.6.1-models --test_files ../../librispeech/librivox-test-clean.csv --beam_width 1 --lm_alpha 0 --lm_beta 0 --n_steps 1

and I got this error
ValueError: Cannot feed value of shape (4096,) for Tensor 'cudnn_lstm/rnn/multi_rnn_cell/cell_0/cudnn_compatible_lstm_cell/bias/Initializer/Const:0', which has shape '(8192,)'

any advice?
Thanks!

  1. Don’t set the LM parameters to zero, pass an empty LM path.

  2. You need a checkpoint and --checkpoint_dir for evaluate.py, not exported models.

  3. You need to specify --n_hidden 2048 for our official models/checkpoints.

  4. Your command is missing some _ characters (looks like they were replaced with spaces).

Hi, @reuben

Thanks for your advice.

Do you mean --scorer_path ?
I’ve set that before but got ValueError: Scorer initialization failed with error code 1

I’ve also tried this before but got this error.

I Could not find best validating checkpoint.
I Could not find most recent checkpoint.

It seems that there is not just one error so I post two version here.

new command_v1
./evaluate.py --n_hidden 2048 --checkpoint_dir ../deepspeech-0.6.1-models --test_files ../../librispeech/librivox-test-clean.csv --scorer --scorer_path --beam_width 1 --n_steps 1

error_v1
ValueError: Scorer initialization failed with error code 1

new command_v2
./evaluate.py --n_hidden 2048 --checkpoint_dir ../deepspeech-0.6.1-models --test_files ../../librispeech/librivox-test-clean.csv --beam_width 1 --n_steps 1

error_v2

I Could not find best validating checkpoint.
I Could not find most recent checkpoint.

Thanks!

See the output of --helpfull or the documentation for the version you’re using, 0.6.1. Pretty sure --scorer_path did not exist back then.

This seems like a completely unrelated problem. Please open a new thread.

Sorry, will delete above comment and create another thread.