o  h/@sddlZddlZddlZddlmZmZmZddlZde de dej fddZ   dde d e d ed eed eeef d dZ     dde d e dee deeee e fe fd ed eed eeefddZd de dee e fdefddZde fddZddZdS)!N)OptionalTupleUnionbpayload sampling_ratereturnc Cs|}d}d}dddd|d|d|d d d d g }z!tj|tjtjd  }||}Wdn1s1wYWntyH}ztd|d}~ww|d} t| tj} | j ddkr_td| S)z? Helper function to read an audio file through ffmpeg. 1f32leffmpeg-izpipe:0-ac-ar-f -hide_banner -loglevelquietpipe:1)stdinstdoutNzFffmpeg was not found but is required to load audio files from filenamera Soundfile is either not in the correct format or is malformed. Ensure that the soundfile has a valid audio file extension (e.g. wav, flac or mp3) and is not corrupted. If reading from a remote URL, ensure that the URL is the full address to **download** the audio file.) subprocessPopenPIPE communicateFileNotFoundError ValueErrornp frombufferfloat32shape) rraracformat_for_conversionffmpeg_commandffmpeg_process output_streamerror out_bytesaudior(v/var/www/html/construction_image-detection-poc/venv/lib/python3.10/site-packages/transformers/pipelines/audio_utils.py ffmpeg_read sB  r*r chunk_length_sr!ffmpeg_input_deviceffmpeg_additional_argsccs|}d}|dkr d}n|dkrd}ntd|dt}|dkr+d } |p)d } n|d kr6d } |p4d } n |dkrAd} |p@t} |durGgn|}dd| d| d|d|d|ddddddg} | |tt|||} t| | } | D]}|VqrdS)a Helper function to read audio from a microphone using ffmpeg. The default input device will be used unless another input device is specified using the `ffmpeg_input_device` argument. Uses 'alsa' on Linux, 'avfoundation' on MacOS and 'dshow' on Windows. Arguments: sampling_rate (`int`): The sampling_rate to use when reading the data from the microphone. Try using the model's sampling_rate to avoid resampling later. chunk_length_s (`float` or `int`): The length of the maximum chunk of audio to be sent returned. format_for_conversion (`str`, defaults to `f32le`): The name of the format of the audio samples to be returned by ffmpeg. The standard is `f32le`, `s16le` could also be used. ffmpeg_input_device (`str`, *optional*): The identifier of the input device to be used by ffmpeg (i.e. ffmpeg's '-i' argument). If unset, the default input device will be used. See `https://www.ffmpeg.org/ffmpeg-devices.html#Input-Devices` for how to specify and list input devices. ffmpeg_additional_args (`list[str]`, *optional*): Additional arguments to pass to ffmpeg, can include arguments like -nostdin for running as a background process. For example, to pass -nostdin to the ffmpeg process, pass in ["-nostdin"]. If passing in flags with multiple arguments, use the following convention (eg ["flag", "arg1", "arg2]). Returns: A generator yielding audio chunks of `chunk_length_s` seconds as `bytes` objects of length `int(round(sampling_rate * chunk_length_s)) * size_of_sample`. rs16ler Unhandled format ` `. Please use `s16le` or `f32le`LinuxalsadefaultDarwin avfoundationz:defaultWindowsdshowNr rr r r z-fflagsnobufferrrrr)rplatformsystem_get_microphone_nameextendintround_ffmpeg_stream)rr+r!r,r-rr size_of_sampler<format_input_r" chunk_leniteratoritemr(r(r)ffmpeg_microphone1sV"     rHstream_chunk_sstride_length_sccsp|dur|}n|}t|||||durgn|d}|dkr#tj} d} n|dkr-tj} d} ntd|d|dur=|d }tt||| } t|ttfrR||g}tt||d | } tt||d | } t j }t j |d }t || | | fd dD]6}tj |d| d|d<|dd | |dd | f|d<||d<||7}t j |d|krq|VqdS)a Helper function to read audio from a microphone using ffmpeg. This will output `partial` overlapping chunks starting from `stream_chunk_s` (if it is defined) until `chunk_length_s` is reached. It will make use of striding to avoid errors on the "sides" of the various chunks. The default input device will be used unless another input device is specified using the `ffmpeg_input_device` argument. Uses 'alsa' on Linux, 'avfoundation' on MacOS and 'dshow' on Windows. Arguments: sampling_rate (`int`): The sampling_rate to use when reading the data from the microphone. Try using the model's sampling_rate to avoid resampling later. chunk_length_s (`float` or `int`): The length of the maximum chunk of audio to be sent returned. This includes the eventual striding. stream_chunk_s (`float` or `int`): The length of the minimal temporary audio to be returned. stride_length_s (`float` or `int` or `(float, float)`, *optional*): The length of the striding to be used. Stride is used to provide context to a model on the (left, right) of an audio sample but without using that part to actually make the prediction. Setting this does not change the length of the chunk. format_for_conversion (`str`, *optional*, defaults to `f32le`): The name of the format of the audio samples to be returned by ffmpeg. The standard is `f32le`, `s16le` could also be used. ffmpeg_input_device (`str`, *optional*): The identifier of the input device to be used by ffmpeg (i.e. ffmpeg's '-i' argument). If unset, the default input device will be used. See `https://www.ffmpeg.org/ffmpeg-devices.html#Input-Devices` for how to specify and list input devices. ffmpeg_additional_args (`list[str]`, *optional*): Additional arguments to pass to ffmpeg, can include arguments like -nostdin for running as a background process. For example, to pass -nostdin to the ffmpeg process, pass in ["-nostdin"]. If passing in flags with multiple arguments, use the following convention (eg ["flag", "arg1", "arg2]). Return: A generator yielding dictionaries of the following form `{"sampling_rate": int, "raw": np.array(), "partial" bool}` With optionally a `"stride" (int, int)` key if `stride_length_s` is defined. `stride` and `raw` are all expressed in `samples`, and `partial` is a boolean saying if the current yield item is a whole chunk, or a partial temporary result to be later replaced by another larger chunk. N)r!r,r-r.r/r r0r1r2r)secondsT)stridestreamraw)dtyperNr )rHrint16rrr?r@ isinstancefloatdatetimenow timedeltachunk_bytes_iterr)rr+rIrJr!r,r-chunk_s microphonerQrBrE stride_left stride_right audio_timedeltarGr(r(r)ffmpeg_microphone_livesL0  r`FrErNrOc cs d}|\}}|||krtd|d|d|d}|D]L}||7}|r sz(_get_microphone_name..r"rLzUsing microphone: zaudio=zOffmpeg was not found. Please install it or make sure it is in your system PATH.r5)rrunrrl splitlinessplitprintr)commandffmpeg_devicesmicrophone_linesmicrophone_namer(r(r)r=s  r=)r NN)NNr NN)F)rVr;rtypingrrrnumpyrbytesr?arrayr*rUstrlistrHr`boolrYrAr=r(r(r(r)sV*  X  _#