search

flashlight / wav2letter

Facebook AI Research's Automatic Speech Recognition Toolkit
https://github.com/facebookresearch/wav2letter/wiki
Other
6.37k stars 1.01k forks source link

streaming_tds_model_converter error #646

Closed bjl21012 closed 4 years ago

bjl21012 commented 4 years ago

after Train model with network.arch and test， I try to General feature_extractor.bin by streaming_tds_model_converter , terminate called after throwing an instance of 'std::invalid_argument' command : /root/wav2letter/build/tools/streaming_tds_model_converter -am /home/bjl/data/data_aishell/aishelltrainlogs/001_model_last.bin --outdir ./ output ： terminate called after throwing an instance of 'std::invalid_argument' what(): stoi Aborted at 1588934677 (unix time) try "date -d @1588934677" if you are using GNU date PC: @ 0x7fc2bd67ee97 gsignal SIGABRT (@0x4498) received by PID 17560 (TID 0x7fc302b4e3c0) from PID 17560; stack trace: @ 0x7fc2c537c890 (unknown) @ 0x7fc2bd67ee97 gsignal @ 0x7fc2bd680801 abort @ 0x7fc2be073957 (unknown) @ 0x7fc2be079ab6 (unknown) @ 0x7fc2be079af1 std::terminate() @ 0x7fc2be079d24 cxa_throw @ 0x7fc2be0757e3 (unknown) @ 0x561416a36db3 _ZN9gnu_cxx6__stoaIlicJiEEET0_PFT_PKT1_PPS3_DpT2_EPKcS5PmS9.constprop.24545 @ 0x5614169cc839 main @ 0x7fc2bd661b97 __libc_start_main @ 0x561416a364ba _start Aborted (core dumped) Is there problem with the command ？ Thank you!

vineelpratap commented 4 years ago

Hi, could you provide the arch file you are using. Also, if you can gdb backtrace, it would help us debug easily.

bjl21012 commented 4 years ago

Hi, could you provide the arch file you are using. Also, if you can gdb backtrace, it would help us debug easily.

i use network.arch

8 Temporal Convolutions followed by 2 Linear layers with ReLU activations.

First Conv layer has stride of 2. All Conv layers perform "SAME" padding.

V -1 1 NFEAT 0 C2 NFEAT 256 8 1 2 1 -1 -1 R C2 256 256 8 1 1 1 -1 -1 R C2 256 256 8 1 1 1 -1 -1 R C2 256 256 8 1 1 1 -1 -1 R C2 256 256 8 1 1 1 -1 -1 R C2 256 256 8 1 1 1 -1 -1 R C2 256 256 8 1 1 1 -1 -1 R C2 256 256 8 1 1 1 -1 -1 R RO 2 0 3 1 L 256 512 R L 512 NLABEL

vineelpratap commented 4 years ago

Hi, We support only TDS based modules for streaming inference at the moment. See https://github.com/facebookresearch/wav2letter/tree/master/recipes/models/streaming_convnets for an example.

bjl21012 commented 4 years ago

Hi, We support only TDS based modules for streaming inference at the moment. See https://github.com/facebookresearch/wav2letter/tree/master/recipes/models/streaming_convnets for an example.

Thanks