digital-umuganda / deepspeech-kinyarwanda Goto Github PK

I'm very interested in your project but I'm surprised how little information is given to use it.

I was planning to have a real-time Kinyarwanda speech to Text with your work and I found the following video
https://www.youtube.com/watch?v=c_0Q3T0XYTA and tested your pbmm and scorer models but I'm told that the Model File 1200h_output_graph.pbmm is a correpted memmaped model file.

(DeepSpeech) F:\DeepSpeech\mic_vad_streaming>python mic_vad_streaming.py -m F:\DeepSpeech\1200h_output_graph.pbmm -s "F:\DeepSpeech\kinyarwanda_deepspeech_model(full_vocabulary).scorer"
Initializing model...
INFO:root:ARGS.model: F:\DeepSpeech\1200h_output_graph.pbmm
TensorFlow: v2.3.0-6-g23ad988fcd
DeepSpeech: v0.9.3-0-gf2e9c858
Data loss: Corrupted memmapped model file: F:\DeepSpeech\1200h_output_graph.pbmm Invalid directory offset
Traceback (most recent call last):
File "mic_vad_streaming.py", line 224, in
main(ARGS)
File "mic_vad_streaming.py", line 163, in main
model = deepspeech.Model(ARGS.model)
File "F:\DeepSpeech\lib\site-packages\deepspeech_init_.py", line 38, in init
raise RuntimeError("CreateModel failed with '{}' (0x{:X})".format(deepspeech.impl.ErrorCodeToErrorMessage(status),status))
RuntimeError: CreateModel failed with 'Failed to initialize memory mapped model.' (0x3000)

It seems your 1200h_output_graph.pbmm is very small and is text based while other pbmm files such as english deepspeech-0.9.3-models.pbmm are binary and big.

Could you please show us how to test your model?

OS: Windows 10
Python: 3.8.10
DeepSpeech version 0.9.3