-
Notifications
You must be signed in to change notification settings - Fork 2
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
AttributeError: module 'gentle' has no attribute 'Resources' #3
Comments
Hi Youseff, I've been working on a port that will make this more usable and to replicate the environment. I have also had this problem moving everything to a new virtual environment. Your message has given me new motivation to figure out what is wrong and why it works in one of my environments and not another. I'll get back to you ASAP. |
Hey Cassandra, Awesome! Thanks so much and no pressure whatsoever. I wonder if gently also allows the production of sentence and paragraph or line by line start and end time stamps. I got the Gentle app to work but I’ve been having quite a difficult time working with the Gentle align csv output to transform it from a timestamp per word to a timestamp per line. I tried parsing it but I really suck at that so I couldn’t figure out a way to do it. No worries if not though, appreciate all of what you’ve done with gently so far! |
Interesting application! I think it could in principle -- I don't know if you've tried using the Gentle standalone app, but I've definitely used it to transcribe multiple sentences before. The challenge is that transcription takes an exponential amount of time (if I recall correctly) with the length of the audio file. But if we assume you have the word-by-word alignments, the harder alignment problem is figuring out whether a sequence of transcriptions is a full sentence or not. You could in principle loop over the words in the transcription and terminate when you get to the end of a sentence i and the start of the next sentence i+1. I wonder if a more efficient mechanism would be some kind of greedy sentence-splitting procedure...or if you know if there's a good chance that an utterance will take about a certain amount of time, you could chop the file into pieces and align in segments maybe. |
By the way, I just realized that while the package requires the
Some searching through the Issues on the gentle repo led me to this: lowerquality/gentle#168 If you could try the following commands, I think we could figure out if it's a dependency problem with
It does take a while to install Kaldi (which is a C program), so I don't know if this will work, but I ran into a problem where I was missing resources as well I don't think it depends on where you clone |
Tiny update: I have installed gentle (using the install.sh script provided by that package), but now I have found a bug in my textgrid code. See this issue for ongoing updates on that. I will update this issue as well once I have confirmed that the install procedure for gentle/Kaldi works. |
Yeah I tried using the standalone app. Oh sorry, I forgot to say I already have the transcript for the audio (manually transcribed) the trouble comes when trying to get the starting timestamp for every paragraph in the transcript text file. What I tried was to create an index for all the words that were isolated in the CSV file output by Gentle. Then, to do a paragraph start and end timestamp, I’d basically loop over the transcript text file line by line, and then check if every word in the align csv file is in that line, and if it is, add the index of that word to a dictionary like so: {line_index:[word_index, word_index, etc.]} and then I'd check if the word index was already in one of the values' arrays in that dictionary and if not, I'd add it. Unfortunately, this just gave me a mess of numbers. The word indexes would start at a certain point then you'd get words that aren't even in that line for some reason (or maybe they are, but I imagine they're probably 'the' or some other common word from another paragraph or something. So I think the problem area is that line 'if word in line' because it searches through the entire list of words as opposed to ignore previously looked at words. Anyway, I'd rather just throw away this code snippet or something cause it's the weirdest problem I've ever had to deal with. This is the awful code:
Anyway sorry to bore you with this. I tried the commands you provided and cloned it to my home user folder, it installed, then I tried running the gently code again and got the same error unfortunately. I've placed the 'gently' folder in my Documents directory (which is different from where I installed Gentle) so I could try placing it in my user home folder and running the code there too if that has an effect.
However, I must emphasize, that when I run, I run with python3.6 because gently failed to install on python 3.7. So I'm wondering if Gentle or aspects of Gentle or Kali may be installed in python3.7 when they're really needed in python3.6. Here is what happens when I try to install gently in python3.7.2:
|
Hey Cassandra! So I have been trying to get gentle to work again with a python script, and unfortunately failing to do so. In the process, I tried installing gentle and had more problems with it. Then I said let me try gently again, and surprisingly got a different error this time:
This is different from the error I got in the past which was:
The error I got last time looks awfully similar to the current error I'm getting from gentle, but not the one from gently. Appreciate all the help so far, no need to reply if you're busy! |
Hi again! Thanks for following up. Did you go through all the installation steps for gentle? I found that I had to download the models and everything. Try these steps:
And then reinstall Unrelatedly I am planning on porting the alignment model over to |
Thank you so much! This seems to include much more than on the gentle github. All steps seemed to work fine until this one:
I got this error:
Should I continue or do something else? |
Hi all - reading through this and have had a few issues described here, including
and
I would appreciate any updates you all have. I tried to solve the OpenFST issue with the solution proposed at lowerquality/gentle#257, but no luck. My main issue has just been installing Kaldi (more specifically, installing the version required by gentle). |
Does anyone has kaldi-models-0.03.zip? In 2024 it is expired |
Hey, thanks so much for making this, can't wait to use it.
I tried running this on python3.6 (MacOS Mojave 10.14) and thankfully it installed (3.7 gave me an installation error), and I got this error when importing:
The text was updated successfully, but these errors were encountered: