Error: Alphabet size does not match loaded model: alphabet has size 1958, but model has 28 classes in its output. Make sure you’re passing an alphabet file with the same size as the one used for training.
there were few suggestions to download files from here.
Which file to use. there are many links and some of them are not active. I have messed it myself. It will be great to hear some suggestions to solve my confusion.
Thanks.
lissyx
((slow to reply) [NOT PROVIDING SUPPORT])
2
for my Q.1) I meant my tensorflow version and deepspeech- tenforflow versions are different. Kindly see the below output from ./deepspeech help.
> (deepspeech-venv) megha@megha-medion:~/Alu_Meg/DeepSpeech_Alug_Meg/DeepSpeech$ ./deepspeech -h
> Usage: deepspeech MODEL_PATH ALPHABET_PATH [LM_PATH] [TRIE_PATH] AUDIO_PATH [-t]
> MODEL_PATH Path to the model (protocol buffer binary file)
> ALPHABET_PATH Path to the configuration file specifying the alphabet used by the network.
> LM_PATH Optional: Path to the language model binary file.
> TRIE_PATH Optional: Path to the language model trie file created with native_client/generate_trie.
> AUDIO_PATH Path to the audio file (or directory of files) to run (any file format supported by libsox).
> -t Run in benchmark mode, output mfcc & inference time
> TensorFlow: v1.6.0-9-g236f83e
> DeepSpeech: v0.1.1-44-gd68fde8
How can I make both the versions same?
I wrote " I also tried downgrading Tensorflow version to 1.4, but still, it does not work. ", meaning I manually changed my tensorflow version to 1.4 using
pip install tensorflow-gpu==1.4
But I am still getting
Error: Alphabet size does not match loaded model: alphabet has size 1948, but model has 28 classes in its output. Make sure you’re passing an alphabet file with the same size as the one used for training.
how can I solve this?
thanks:)
lissyx
((slow to reply) [NOT PROVIDING SUPPORT])
4