ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behavior chex 0.1.86 requires numpy>=1.24.1, but you have numpy 1.22.0 which is incompatible plotly 5.15.0 requires tenacity>=6.2.0, but you have tenacity 5.1.5 which is incompatible plotnine 0.12.4 requires numpy>=1.23.0, but you have numpy 1.22.0 which is incompatible pywavelets 1.5.0 requires numpy=1.22.4, but you have numpy 1.22.0 which is incompatible tensorflow 2.15.0 requires numpy=1.23.5, but you have numpy 1.22.0 which is incompatible
using version easy 10/25, i get this error: Authtoken saved to configuration file: /root/.config/ngrok/ngrok.yml Tensorboard NGROK URL:NgrokTunnel: ... it happen after running cell 4. and it stops there, nothing else happen. if i run the code cell again, it says there's already a process running.. any idea sir ?
That's the only drawback of using Google Colab's free account. Either you need to purchase Colab GPU or you can train a model on your system using your own GPU. On the other side, Google Colab won't get disconnected that easily. It can run for 12 hours without getting disconnected, but every 2 to 3 hours, it will ask for human validation. If you don't validate, the connection will get disconnected in a few minutes.
@@learnwithdeva "or you can train a model on your system using your own GPU" How is this done or attempted. Since we seem to have to connect Colab right now. And thank you for taking the time and effort to do these tutorials. Much appreciated.
I am getting this error , NameError Traceback (most recent call last) in () 2 f0method = "rmvpe_gpu" # @param ["pm", "harvest", "rmvpe", "rmvpe_gpu"] 3 get_ipython().run_line_magic('cd', '/content/drive/MyDrive/project-main') ----> 4 with open(f'./logs/{model_name}/extract_f0_feature.log','w') as f: 5 print("Starting...") 6 if f0method != "rmvpe_gpu": NameError: name 'model_name' is not defined
Sorry, I made those videos in Tamil because all those concepts are already available in English. So, I made them for the Tamil audience (which is my mother tongue). Regarding the RVC model, I couldn't find many videos on TH-cam, so I made it in English. Since many are interested in Deep-voice, Deep-Image, and Deep-video tutorials, I will post all upcoming videos in English. Stay tuned.
Was I wrong? I trained two models consecutively, each for 500 epochs, but the second one was halted. I remember reading somewhere that free Colab has a usage limit, so I let my account cool down. What I did was download the entire logs of the model and .pth file from the weights folder, and then upload it to the logs and weights of another Google Drive account that also has a copy of your notebook. I re-uploaded the dataset to Colab and ran the training. Although I didn’t do the coding part, it continued from the last epoch (250) and finished the training.
The behavior you're describing is unexpected, as the training code, when run in the same environment, should typically start from epoch 1 unless the pre-trained models are correctly loaded. Anyway, try using that model and index file and check the result as well. I am not exactly sure how it works; if you find out, let me know as well.
I was training a 500 epoch model that time sout at 492. I tired retraining as described but keep getting an error. Wondering if this is because free Colab has times out and I attempted to run it without GPU, figuring it's on;y 8 epochs, so I assume it would pick up from 450? Anyway, is there still no work around with freee Colab locking us out? I thought the purpose of storing a copy of Colab was to bpass them and run it off our pc with gpu. I'm familiar with some of the basic programming but really out in the cold on the rest. So any standalone pc only options for rvc training? Thanks
@@powerdistribution9590 Nope, this tutorial focuses on running RVC on colab without the webui, since colab has banned webui versions of AI tools like RVC and Stable Diffusion. They're no longer supporting those, claiming that free colab is meant for learning only. But yes, there is RVC for local machines. I initially tried training it on my RTX 3050, but it was taking about 8 minutes per epoch, which was too slow for me. Colab, on the other hand, can do an epoch in around 24 to 40 seconds, depending on dataset size. This saves me a ton on electricity bills. I train in colab, download the model and convert voice with my pc. As you can see, most tutorials are about training models, not voice conversion, since even a basic PC without dedicated gpu can handle that. So, I'm curious where you're doing your RVC voice conversion and how you know you have a good model if you don't have RVC on your local machine.
Just add more audios in a folder with the same name you used before. Run all the code from the beginning, and while running the "Train Model" code, just change those two files to D_XXXXXX.pth and G_XXXXXX.pth, as in this video.
Yes, you can increase the size of your dataset and continue training by running the preprocessing, feature extraction, training index, and training model codes.
I just did what you mention in the video and the process starts from where I left off but the training of the model stops and an error appears "TypeError: Unable to convert function return value to a Python type! The signature was () -> handle" And then something in "chinese"train.log 😢 And I can't continue the train model from 135 epochs to 250 epochs. It's so weird
I even tried a different colab account, and same result. ngrok seems to be the problem, it was reported before that colab blocked ngrok by error, maybe the same is happening ? @@learnwithdeva
I'm having some trouble for me it says that I have exceeded the time I thought this method was free also I have a question, I want to use the model I created of 250 epochs for an audio book but I believe the time limit for an audio to be converted with the model is 5 minutes do you know how can I fix this?
Yes, you can add. But after adding you need to run all the codes from beginning and when running Train model just alter those 2 paths as shown in this video.
It means you have used the free GPU for more than the time limit. Either you can wait for some time, or you can use a different Colab account using another Google account to train a model.
The "heartbeat timed out" error in Colab typically occurs when the notebook is idle for too long, and the connection between the Colab virtual machine and the browser is lost. Colab has a maximum active runtime of 12 hours, and if your notebook is left idle, it might disconnect.
Yes, if the session gets disconnected, unfortunately, we need to run all the codes again. That is the drawback of using online training. Hopefully, by tomorrow, I will post a video on offline model training using the local RVC GUI.
Reminder: If you are training a model for a long time, it will ask you to verify "You are human" approximately every 2 hours. So, you have to verify it; otherwise, in a few minutes, your session will time out.
Yes, you can run Google Colab using any web browser from your phone and train a model. Just follow this video: th-cam.com/video/79K9k8OSpIA/w-d-xo.html
@@learnwithdeva I downloaded goweb into download folder, and I double click that file. It runs into cmd screen. There is sign: click any key to continou.... When I click, the screen disappeared
No, bro, this is not the way of doing it. This video is just the continuation of the RVC series. Please check this video: th-cam.com/video/vHUrPgRwW5Q/w-d-xo.html
after changing value of those 2 values in 4.13 of this video.now i want to train another model but values are as last model and in the codes says No pretrained Generator No pretrained Discriminato still my last model information is in how to change it to normal value
After resuming the training of a model and later when you want to train a new model, you have to set the path to default as it was before and run all the code from Step 1 (Preprocessing Data).
I will train voice model and run it successfully thank you so much, but the problem is It's show me processing very log time but don’t show any result Plesse help me.
I got an error while training which shows "FileNotFoundError: [Errno 2] No such file or directory: '/content/drive/MyDrive/project-main/logs/SGmodel1/3_feature768/54_9.npy'" . Can you explain who to solve this? btw your videos are so helpful machaa
Seems like some files of feature extraction are missing. Just re-run the feature extraction and train index code, and then try training the model; it should work fine now.
Hii bro, i phases different problems in gradio downloading and you link of gradio is not giving like you gradio so i truly request to for solve this problem please😢🙏🙏🙏
Gradio ka Jo video aapne video banaya hai na bhaiya usme RVC beta.7z me sirf 1 zip file me deta hai or unzip file nahi deta jis tarah se RVC beta.7z voice modal ko paste Kiya hai apne us tarah se paste bhi nahi ho rha Mera voice modal (bhaiya mai Sara screen shot bhej deta hu kya) (apke pas telegram channel hai aap pin kar dijiye na bhaiya please)
Can you explain properly within downloading proof and clearly extraction,I thought that if you do it,probably anybody will converse his voice modal and i will give you a genuine subs
It will generate with respect to the save frequency you have set. Watch my previous video regarding training custom models using RVC for better understanding. This video is just the continuation of the previous video.
To resolve this warning, you can either choose to move the Pyngrok configuration file to the recommended XDG location or choose to ignore it if it's not causing any issues, as it's just a warning message.
using version easy 10/25, i get this error: Authtoken saved to configuration file: /root/.config/ngrok/ngrok.yml Tensorboard NGROK URL:NgrokTunnel: ... it happen after running cell 4. and it stops there, nothing else happen. if i run the code cell again, it says there's already a process running.. any idea bro ?
![RVC Local || Train Custom Voice Model in Local System [GPU Required]](http://i.ytimg.com/vi/bpcr0v1Nie8/mqdefault.jpg)
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behavior
chex 0.1.86 requires numpy>=1.24.1, but you have numpy 1.22.0 which is incompatible
plotly 5.15.0 requires tenacity>=6.2.0, but you have tenacity 5.1.5 which is incompatible
plotnine 0.12.4 requires numpy>=1.23.0, but you have numpy 1.22.0 which is incompatible
pywavelets 1.5.0 requires numpy=1.22.4, but you have numpy 1.22.0 which is incompatible
tensorflow 2.15.0 requires numpy=1.23.5, but you have numpy 1.22.0 which is incompatible
using version easy 10/25, i get this error: Authtoken saved to configuration file: /root/.config/ngrok/ngrok.yml
Tensorboard NGROK URL:NgrokTunnel: ...
it happen after running cell 4. and it stops there, nothing else happen. if i run the code cell again, it says there's already a process running.. any idea sir ?
What I need to get from the project main folder copied when I want to resume the training on a different Google Account?
It's showing
SystemError: initialization of _pywrap_checkpoint_reader raised unreported exception
Woahhh, it works!!! you did it again 🔥 now i know what to do when my colab disconnect 🙌
That's the only drawback of using Google Colab's free account. Either you need to purchase Colab GPU or you can train a model on your system using your own GPU. On the other side, Google Colab won't get disconnected that easily. It can run for 12 hours without getting disconnected, but every 2 to 3 hours, it will ask for human validation. If you don't validate, the connection will get disconnected in a few minutes.
@@learnwithdeva "or you can train a model on your system using your own GPU" How is this done or attempted. Since we seem to have to connect Colab right now. And thank you for taking the time and effort to do these tutorials. Much appreciated.
Just uploaded a video on how to train model on your system. Please check it out:
th-cam.com/video/bpcr0v1Nie8/w-d-xo.html
Nice videos 👍 its very helpful with detailed explanation 👏 keep doing more videos 🤘
Vazhthukal 👍
Sure, bro. நன்றி ❤️
comments are disabled ?
No, bro. See you can comment here 😋
If I paste my error my comments are not going thru unfortunately @@learnwithdeva
Maybe it contains some words which are blocked by TH-cam terms. DM me in Instagram
I noticed that my RAM is red and not green. why is that?
What if my folder vanish on the left side with the source mp3?
You did a cut in your video and the folder came back the neil
I am getting this error , NameError Traceback (most recent call last)
in ()
2 f0method = "rmvpe_gpu" # @param ["pm", "harvest", "rmvpe", "rmvpe_gpu"]
3 get_ipython().run_line_magic('cd', '/content/drive/MyDrive/project-main')
----> 4 with open(f'./logs/{model_name}/extract_f0_feature.log','w') as f:
5 print("Starting...")
6 if f0method != "rmvpe_gpu":
NameError: name 'model_name' is not defined
Does the quality of training improve as the number of epochs increases?
I mean, is it certain?
Yes
great explanation
you are making great videos. I was cheaking your other videos but all that was in Tamil language which I don't understand as a Bengali.
Sorry, I made those videos in Tamil because all those concepts are already available in English. So, I made them for the Tamil audience (which is my mother tongue). Regarding the RVC model, I couldn't find many videos on TH-cam, so I made it in English. Since many are interested in Deep-voice, Deep-Image, and Deep-video tutorials, I will post all upcoming videos in English. Stay tuned.
Was I wrong? I trained two models consecutively, each for 500 epochs, but the second one was halted. I remember reading somewhere that free Colab has a usage limit, so I let my account cool down. What I did was download the entire logs of the model and .pth file from the weights folder, and then upload it to the logs and weights of another Google Drive account that also has a copy of your notebook. I re-uploaded the dataset to Colab and ran the training. Although I didn’t do the coding part, it continued from the last epoch (250) and finished the training.
The behavior you're describing is unexpected, as the training code, when run in the same environment, should typically start from epoch 1 unless the pre-trained models are correctly loaded. Anyway, try using that model and index file and check the result as well. I am not exactly sure how it works; if you find out, let me know as well.
I was training a 500 epoch model that time sout at 492. I tired retraining as described but keep getting an error. Wondering if this is because free Colab has times out and I attempted to run it without GPU, figuring it's on;y 8 epochs, so I assume it would pick up from 450? Anyway, is there still no work around with freee Colab locking us out? I thought the purpose of storing a copy of Colab was to bpass them and run it off our pc with gpu. I'm familiar with some of the basic programming but really out in the cold on the rest. So any standalone pc only options for rvc training? Thanks
@@powerdistribution9590 Nope, this tutorial focuses on running RVC on colab without the webui, since colab has banned webui versions of AI tools like RVC and Stable Diffusion. They're no longer supporting those, claiming that free colab is meant for learning only.
But yes, there is RVC for local machines. I initially tried training it on my RTX 3050, but it was taking about 8 minutes per epoch, which was too slow for me. Colab, on the other hand, can do an epoch in around 24 to 40 seconds, depending on dataset size. This saves me a ton on electricity bills. I train in colab, download the model and convert voice with my pc.
As you can see, most tutorials are about training models, not voice conversion, since even a basic PC without dedicated gpu can handle that. So, I'm curious where you're doing your RVC voice conversion and how you know you have a good model if you don't have RVC on your local machine.
Just uploaded a video on training a model on the local system. Check it out:
th-cam.com/video/bpcr0v1Nie8/w-d-xo.html
its not working showing error did we use same token we use last time bro pls reply
Not necessarily. What error you are facing over here?
youre a legend, over time i can make my voice 1000 epoch
Yes, you can even train more then 1000 epochs. Watch my next video on how to increase epochs more than 1000.
@@learnwithdeva does it sound better after 1000
If you have around 20mins audio of using different words then you can just train 300 epochs then it will be perfect.
@@learnwithdeva what if i have a very small dataset, only few minute, then i should do as much epoch as possible?
Even if you train more number of times with small dataset after certain number of epochs there will be no improvement in the result.
please make video how to add more database and epochs to the trained model if can
Just add more audios in a folder with the same name you used before. Run all the code from the beginning, and while running the "Train Model" code, just change those two files to D_XXXXXX.pth and G_XXXXXX.pth, as in this video.
.index file is not downloading #
can i add more audio files to the drive before these steps and repeat preproccess and all other steps?
Yes, you can increase the size of your dataset and continue training by running the preprocessing, feature extraction, training index, and training model codes.
thank you brother@@learnwithdeva
After I created my model what platform do you use to use your voice model
I have posted another video on it. Check it out:
th-cam.com/video/vHUrPgRwW5Q/w-d-xo.html
Do i have to purchase more compute units? For using it
No, you can use free GPU of Google Colab. Watch this playlist:
th-cam.com/play/PLUp0XhK0KED2CD9dx36LMfno2KdGa3cW8.html
I just did what you mention in the video and the process starts from where I left off but the training of the model stops and an error appears "TypeError: Unable to convert function return value to a Python type! The signature was () -> handle" And then something in "chinese"train.log 😢 And I can't continue the train model from 135 epochs to 250 epochs. It's so weird
Please copy and paste the entire error message.
I'm getting the same.
Please copy and paste the entire error message so that I can identify the issue.
wondering if it'S the Easy 10/25 version that is broken, or if ngrok that is banned by colab.. ?
This is the updated version for resuming training epochs.
Easy 10/25 is not broken, watch this video:
th-cam.com/video/79K9k8OSpIA/w-d-xo.html
I even tried a different colab account, and same result. ngrok seems to be the problem, it was reported before that colab blocked ngrok by error, maybe the same is happening ?
@@learnwithdeva
It's working fine when running the file that I have given in my previous video. Please use that file and check whether you are facing the same issue.
I'm having some trouble for me it says that I have exceeded the time I thought this method was free also I have a question, I want to use the model I created of 250 epochs for an audio book but I believe the time limit for an audio to be converted with the model is 5 minutes do you know how can I fix this?
No, brother. It does not have any time limit for voice conversion. Do you mean the training time or the conversion audio time?
can i add more database to last model trained
Yes, you can add. But after adding you need to run all the codes from beginning and when running Train model just alter those 2 paths as shown in this video.
I have a question.
Why is it showing Cannot connect to GPU backend. You cannot currently connect to a GPU due to usage limits in colab?
It means you have used the free GPU for more than the time limit. Either you can wait for some time, or you can use a different Colab account using another Google account to train a model.
@@learnwithdeva if you have to use another account do you have to start over
This is the paid version or free
It's completely free. Watch this video to train your own model for free.
th-cam.com/video/79K9k8OSpIA/w-d-xo.htmlfeature=shared
Hey. I followed the steps shown in your video but training stopped due to error: heartbeat timed out. Any ideas? Thanks.
The "heartbeat timed out" error in Colab typically occurs when the notebook is idle for too long, and the connection between the Colab virtual machine and the browser is lost. Colab has a maximum active runtime of 12 hours, and if your notebook is left idle, it might disconnect.
How many hours are you running this model?
I cant remember but I think I left the session open for quite a few hours. Thanks for the info, its much appreciated! @@learnwithdeva
Yes, if the session gets disconnected, unfortunately, we need to run all the codes again. That is the drawback of using online training. Hopefully, by tomorrow, I will post a video on offline model training using the local RVC GUI.
Reminder: If you are training a model for a long time, it will ask you to verify "You are human" approximately every 2 hours. So, you have to verify it; otherwise, in a few minutes, your session will time out.
Can I create RVC models in mobile phone
Yes, you can run Google Colab using any web browser from your phone and train a model. Just follow this video: th-cam.com/video/79K9k8OSpIA/w-d-xo.html
@@learnwithdeva thanks buddy
Hi Bro, I cannot run go-web bat from my drive. Please help. Thank you
Please let me know what error it shows.
@@learnwithdeva I downloaded goweb into download folder, and I double click that file. It runs into cmd screen. There is sign: click any key to continou.... When I click, the screen disappeared
No, bro, this is not the way of doing it. This video is just the continuation of the RVC series. Please check this video:
th-cam.com/video/vHUrPgRwW5Q/w-d-xo.html
@@learnwithdeva OK. Is this only the way how to get pth file and index file?
No, you can also train your own model and index file for your custom voice. Follow this video:
th-cam.com/video/79K9k8OSpIA/w-d-xo.html
after changing value of those 2 values in 4.13 of this video.now i want to train another model but values are as last model and in the codes says No pretrained Generator
No pretrained Discriminato still my last model information is in how to change it to normal value
After resuming the training of a model and later when you want to train a new model, you have to set the path to default as it was before and run all the code from Step 1 (Preprocessing Data).
tnx bro u are really good man@@learnwithdeva
❤️
I will train voice model and run it successfully thank you so much, but the problem is It's show me processing very log time but don’t show any result Plesse help me.
You mean the audio conversion is taking time? Usually, it does take time according to the length of the target audio.
I got an error while training which shows "FileNotFoundError: [Errno 2] No such file or directory: '/content/drive/MyDrive/project-main/logs/SGmodel1/3_feature768/54_9.npy'" . Can you explain who to solve this?
btw your videos are so helpful machaa
Seems like some files of feature extraction are missing. Just re-run the feature extraction and train index code, and then try training the model; it should work fine now.
ok.. thanks bro@@learnwithdeva
If you solve this perfecyt voice modaling then you will become mass level youtuber
Please mention your query clearly; I will try my best to help you with it.
Hii bro, i phases different problems in gradio downloading and you link of gradio is not giving like you gradio so i truly request to for solve this problem please😢🙏🙏🙏
Can you give me another idea for doing audio transformation of modal. I want this. I truly waited your this video and again I will be must be waiting
Could you please explain what's the problem you are facing?
Please elaborate on the video you want on audio transformation.
Gradio ka Jo video aapne video banaya hai na bhaiya usme RVC beta.7z me sirf 1 zip file me deta hai or unzip file nahi deta jis tarah se RVC beta.7z voice modal ko paste Kiya hai apne us tarah se paste bhi nahi ho rha Mera voice modal (bhaiya mai Sara screen shot bhej deta hu kya) (apke pas telegram channel hai aap pin kar dijiye na bhaiya please)
I gave you a new subscriber Bhaiya.
Can you explain properly within downloading proof and clearly extraction,I thought that if you do it,probably anybody will converse his voice modal and i will give you a genuine subs
Please mention your query clearly; I will try my best to help you with it.
Sir .Pth file not generating
It will generate with respect to the save frequency you have set. Watch my previous video regarding training custom models using RVC for better understanding. This video is just the continuation of the previous video.
WARNING:pyngrok config file found at legacy location, move to XDG location"
please help
To resolve this warning, you can either choose to move the Pyngrok configuration file to the recommended XDG location or choose to ignore it if it's not causing any issues, as it's just a warning message.
@@learnwithdeva I actually just ran all the cells again with the correct paths and it worked
@@OpenizeMusicthat solution also worked for me too.
using version easy 10/25, i get this error: Authtoken saved to configuration file: /root/.config/ngrok/ngrok.yml
Tensorboard NGROK URL:NgrokTunnel: ...
it happen after running cell 4. and it stops there, nothing else happen. if i run the code cell again, it says there's already a process running.. any idea bro ?
Follow this video:
th-cam.com/video/79K9k8OSpIA/w-d-xo.html
@@learnwithdeva i did already these steps i watched ur video then i made but got that roblem