UnboundLocalError: local variable 'word_start' referenced before assignment and installing whisperx v3 in google colaboratory

[akashmamun610]

When i run whisperx in google colab it always says , UnboundLocalError: local variable 'word_start' referenced before assignment.

My command was..

!whisperx "/content/drive/MyDrive/Dvd.mp4" --model large-v2 --language "en"

Performing transcription... Performing alignment... Failed to align segment (" Thank you."): backtrack failed, resorting to original... Traceback (most recent call last): File "/usr/local/bin/whisperx", line 8, in sys.exit(cli()) File "/usr/local/lib/python3.10/site-packages/whisperx/transcribe.py", line 183, in cli result = align(result["segments"], align_model, align_metadata, input_audio, device, interpolate_method=interpolate_method) File "/usr/local/lib/python3.10/site-packages/whisperx/alignment.py", line 442, in align "start": word_start,

UnboundLocalError: local variable 'word_start' referenced before assignment

[arnavmehta7]

Here's the fix: https://github.com/m-bain/whisperX/pull/208

[arnavmehta7]

It's merged. Re-install library and it won't occur again :D

[akashmamun610]

Thanks a lot @arnavmehta7 .. it has worked.. 😊😊😊....

I have also some issues on running whisperx in google colaboratory.. can you help me to fix it ???

Example..

1. First i have to input this command in google colab.. !pip install -q condacolab import condacolab condacolab.install()

When i input this command then it is working otherwise not... but it restarts the whole running cell..

2. When i input this command below in google colab it works.. !conda create --name whisperx python=3.8 !conda activate whisperx

But it doesn't work in python=3.10 I think whisperx v3 should be work in latest version of python..

3. when i install torch, torchvision and torchaudio latest version it doesn't work .. it only works when i input below command...

!pip install torch==1.11.0+cu113 torchvision==0.12.0+cu113 torchaudio==0.11.0 --extra-index-url https://download.pytorch.org/whl/cu113

4. Now when i input the most important command below this..

!pip install git+https://github.com/m-bain/whisperx.git@v3 --upgrade

It install and unstall many things..

But at the end it always said..

WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv WARNING: The following packages were previously imported in this runtime: [cycler,dateutil,google,kiwisolver,matplotlib,mpl_toolkits,six] You must restart the runtime in order to use newly installed versions.

When i restart the runtime then all my data's have lost and i have to install all these things i said above..

5. My last input is. !whisperx "/content/drive/MyDrive/dvd" --language "en" --model large-v2 --max_line_count 2 --max_line_width 42 --device "cuda" --batch_size 32 --beam_size 5 --compute_type "float16"

And it works super super fast.. movie length is about 1 hour 43 minutes and 53 seconds and it transcribes within about 4 minutes.. if batch_size is 32 then it works faster than batch_size 16..

But something i missing is that, VAD, diarize, word _timestamps, highlights word , hf_token is not working in this version.. Also some align segment is missing such as

(Failed to align segment (" Thank you."): backtrack failed, resorting to original)

Can you help me how i can solve all these problems..???? 🤔🤔🤔🤔

Thanks a lot ...

[arnavmehta7]

You can firstly uninstall torch, torchvision, torchaudio, and also no need of install "condalab". After that directly install whisperx from whisperX repo by running commands in readme.

[akashmamun610]

Thanks for reply @arnavmehta7

Whisperx v3 does not work on Google colab if you not install condalab.. It says SyntaxError: invalid syntax

In readme i follow every steps of it.. it does not work in google colaboratory in their method.. in only works for me that i said earlier.. In whisperx setup method..

1. pip install git+https://github.com/m-bain/whisperx.git@v3 If i input this command then following error occurs....

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts. ipython 7.34.0 requires jedi>=0.16, which is not installed. tensorflow 2.12.0 requires protobuf!=4.21.0,!=4.21.1,!=4.21.2,!=4.21.3,!=4.21.4,!=4.21.5,<5.0.0dev,>=3.20.3, but you have protobuf 3.20.1 which is incompatible. Successfully installed av-10.0.0 coloredlogs-15.0.1 ctranslate2-3.13.0

2.. if i install torch in whisperx method following error also occur

!pip install torch==1.11.0+cu113 torchvision==0.12.0+cu113 torchaudio==0.11.0 --extra-index-url https://download.pytorch.org/whl/cu113

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts. torchdata 0.6.0 requires torch==2.0.0, but you have torch 1.11.0+cu113 which is incompatible. torchtext 0.15.1 requires torch==2.0.0, but you have torch 1.11.0+cu113 which is incompatible.

3. Last input..

!Whisperx "/content/drive/MyDrive/Dvd" --language "en" --model large-v2

It says.. /bin/bash: Whisperx: command not found