something wrong about [pip3 install $(python3) util/taskcluster.py --decoder]

[ruby11dog]

[tom-doerr]

There's another issue for this (https://github.com/carlini/audio_adversarial_examples/issues/33). The script your executing tries to download the ds-ctcdecoder, which is not available any more on the server. Rogthlu from issue https://github.com/carlini/audio_adversarial_examples/issues/33 seems to have found the whl file that the script is trying to download and installed it himself. Alternatively you could use the fixed code I provide in this repo: https://github.com/tom-doerr/audio_adversarial_examples

[ruby11dog]

There's another issue for this (#33). The script your executing tries to download the ds-ctcdecoder, which is not available any more on the server. Rogthlu from issue #33 seems to have found the whl file that the script is trying to download and installed it himself. Alternatively you could use the fixed code I provide in this repo: https://github.com/tom-doerr/audio_adversarial_examples

thanks a lot! Do you post your docker image in the docker hub? If you do, i'd like to pull the images directly.

[dijksterhuis]

The .whl can be built from the local DeepSpeech 0.4.1 repository:

make -C native_client/ctcdecode

https://discourse.mozilla.org/t/could-not-install-requirement-ds-ctcdecoder-0-4-1/47013/5

[tom-doerr]

Just pushed them. For the GPU version run:

docker run --gpus all -it --mount src=$(pwd),target=/audio_adversarial_examples,type=bind -w /audio_adversarial_examples tomdoerr/aae_deepspeech_041_gpu

For the CPU-only version:

docker run -it --mount src=$(pwd),target=/audio_adversarial_examples,type=bind -w /audio_adversarial_examples tomdoerr/aae_deepspeech_041_cpu

[JeetShah10]

The .whl can be built from the local DeepSpeech 0.4.1 repository:

make -C native_client/ctcdecode

https://discourse.mozilla.org/t/could-not-install-requirement-ds-ctcdecoder-0-4-1/47013/5

@dijksterhuis by running this command, I am getting this error:

Traceback (most recent call last): File "./setup.py", line 53, in num_parallel=args[0].num_processes) File "C:\Users\Jeet\Anaconda3\Lib\site-packages\DeepSpeech\native_client\ctcdecode\build_common.py", line 67, in build_common obj_files = list(pool.imap_unordered(build_one, COMMON_FILES)) File "C:\Users\Jeet\Anaconda3\lib\multiprocessing\pool.py", line 748, in next raise value File "C:\Users\Jeet\Anaconda3\lib\multiprocessing\pool.py", line 121, in worker result = (True, func(*args, **kwds)) File "C:\Users\Jeet\Anaconda3\Lib\site-packages\DeepSpeech\native_client\ctcdecode\build_common.py", line 63, in build_one subprocess.check_call(shlex.split(cmd)) File "C:\Users\Jeet\Anaconda3\lib\subprocess.py", line 342, in check_call retcode = call(*popenargs, **kwargs) File "C:\Users\Jeet\Anaconda3\lib\subprocess.py", line 323, in call with Popen(*popenargs, **kwargs) as p: File "C:\Users\Jeet\Anaconda3\lib\subprocess.py", line 775, in init restore_signals, start_new_session) File "C:\Users\Jeet\Anaconda3\lib\subprocess.py", line 1178, in _execute_child startupinfo) FileNotFoundError: [WinError 2] The system cannot find the file specified make: *** [bindings] Error 1 make: Leaving directory `C:/Users/Jeet/Anaconda3/Lib/site-packages/DeepSpeech/native_client/ctcdecode'

can you please tell me how can I resolve this?

[tom-doerr]

You could just extract the .whl file from one of the docker images in case you don't want to use them.

[dijksterhuis]

@JeetShah10 You're probably better off asking in the DeepSpeech Discourse forum tbh.

It's either one of two things:

• the fact you're trying to use Windows
• problems with make finding files

The first is probably going to be most likely.

windows

This thread seems to suggest windows support is patchy for v0.4.1.

IIRC I don't think there were any decoder wheels produced for windows? native_client/definitions.mk only contains build definitions for OS X and Linux.

If this is the issue then you're probably better off using the docker images provided by @tom-doerr

make finding files

make calls setup.py on two different occasions, each time calling build_common.py:

\build_common.py", line 67, in build_common

build_common.py looks for a bunch of common files, including *.cpp extensions, to compile the package.

        cmd = '{cc} -fPIC -c {cflags} {args} {includes} {infile} -o {outfile}'.format(
            cc=compiler,
            cflags=cflags,
            args=' '.join(ARGS),
            includes=' '.join('-I' + i for i in INCLUDES),
            infile=file,
            outfile=outfile,
        )
        print(cmd)
        subprocess.check_call(shlex.split(cmd))

That last line is the one that is failing. I think it's because it can't find some file it needs to run the string cmd on your system. make should print the command cmd it was going to run just before the traceback. Check the files that are listed and see if they exist on your system somewhere?

[tom-doerr]

@JeetShah10 I just uploaded the ctcdecoder .whl file: https://github.com/tom-doerr/audio_adversarial_examples/blob/master/ds_ctcdecoder-0.4.1-cp35-cp35m-linux_x86_64.whl

[kkfkk]

Hi Tom, I am trying to reproduce code with your GPU images. but cannot figure out the right path DeepSpeech in classify.py. Many thanks

File "classify.py", line 20, in import DeepSpeech ImportError: No module named 'DeepSpeech'

[tom-doerr]

@kkfkk Sorry, didn't see your comment. In case you still want to get it running: There were some issues with the docker images due to a numpy update. I fixed the Dockerfiles, you shouldn't have an issue with the newest versions.