วีดีโอ

ебать как у тебя всё просто, хуй у меня что работает так...

heyy..i am working on something similar using whisperx but i am alsomaking a sample html file for start and stop recording and to print the transcription into the text area..but i have run into a problem..so basically this html sends around 10-15 seconds chunk to my python code as a .webm file which is internally converted to wave file which is then used by whisperx to transcribe but whisperx isnt transcribing..like in a 16 second audio it just print "like it". Is there anyway i could make it do it accurately and is there anyway i could use the loop snippet in my script tag from your github, thank youu

I'm a teacher of philosophy, psychology, and political science at a highschool. I want to introduce this perspective to my students for discussion, but I'm so illiterate I don't know where to even begin. What do I do from the second I sit down at my computer?

❤

thankyou mate 👍

i have been mucking about for days trying to do one simple thing with so many different approaches and it just worked out with one simple command!! whisper_mic --model small --loop --dictate thanks , good work

i can't download the Models

DOkerAnd

I can't get this to run with your exact commands and weird file formats. It keeps throwing an error, "response template not set" Which is odd because there is no variable for response template. I tuned off complition_complete it ran the fine tune. Edit: I'm an idiot on the saving checkpoints but the completion complete part still wasn't working for me.

Going to play with this today! Thanks so much for posting this and your great technical work \m/

does not work probably with python 3.12 the code some time works fine and in other time it hang with no result even after the timeout is passed for listening

"I will assume you already have a docker installed" a what?

Hi, this video have been make since years ago so I didnt hope a awnser ... but I have this error when I make the 5:58 . exec /opt/nvidia/nvidia_entrypoint.sh: exec format error can you help me ^^

I recommend watching the whole video first(stupid me didn't). I don't know if this order makes any logical sense but. - run the command to check version first to see if WSL 2 is already installed - try the bash nvidia smi command and see if you get the result like in the video - if not then you might need to install wsl

but the link isn't up in the corner right now :'(

This simply doesn't work anymore. At least not on WSL. A repo link in the DockerFile is broken. When you go to run the image, there are so many failures.

Hi, thanks for the video. Are there any differences in requirements if I only need to interrogate the model? For example through a chat? I do not need to fine tune it

Do you have a tutorial for fine tuning GPT Neo on custom data set ?

Hi Blake, Thank you very much for the video. Could you please upload a tutorial on text-generation-inference? and in your previous LLM finetuning you were using Deepspeed and finetuning the whole model could you please advise if the same can be done on Mixtral 8x7B?

2-3090's and nvlink seem like the lowest entry to llama 2 70b. 2 used 3090's are about the price of a single 4090. Still too expensive for my wallet but at least something I can dream about.

Спасибо но это всё сложно когда уже появится установка в один клик?

can you also please add the commands in the description of your video so its easier to copy paste?

Hi, I was trying to replicate the LLaMA 2 70B fine-tuning with 2 4090s, even with the --split_model flag, the model is loaded to one GPU only before OOM. I tried with the 7B model, which is loaded to one GPU then replicate to the other. It seems it's running in data parallel not model parallel. Is nvlink required for it to work correctly?

Great work! Some questions from my side: How do I turn of the loop, when I am calling whisper_mic from another program? And how can I speed up the mic selection?

This is great, but when I run the listen_loop, it's nowhere near as fast as it is in the video. What needs to be changed to get that fresh quick real-time transcription?

Could someone also use this code and not only transcribe the audio but also diarize it? For example with 2 users speaking the one after the other.

Mate, this is absolutely fantastic. Thank you so much for your efforts on this repo. I was struggling to find a better way to do this coming off the normal speech_recognition library technique everyone uses but didn't want to send all my audio to google. Which would sometimes just not respond at all. I'm so glad I found your project, great stuff man love your work!

Hello, the content you make is impressive, I always wanted to try these text generation models and I have a question, I am about to buy a PC and I was thinking of buying an 8GB RX 580 (unfortunately I only have the budget for that in my country) and I was thinking if I could buy two of those I could do fine-tuning to the 1.3b version plus 32 GB of RAM, anyway thank you very much for your content it is very informative

Hello! Thanks for the video. I hear you tell on the video but I need a confirmation, because I have to buy and here it ts vert expensive: Onde single card RTX 3090 can handle the model GPT-J-6B?

Promo_SM

cool! how about 8 v100 16g. 128g vram in total.

What is the training time of this model in 3090?

very well explained!

faster-whisper implement?

Can somebody please help? When i try to generate a image it says at the end: ImportError: DLL load failed: The specified module could not be found.

I see that after finetuning the model i get a .json and .bin adapter file, how would I run my model using these? Can I use them with llama cpp or how do I go about using the finetuning. I guess my goal is a chat that uses my finetuned model.

It'd be great to allow for a udp stream as an alternative to a device. Wondering if it'd be feasible to replace the timeout with a sort of voice activity detection timeout. These two additions would make this a fantastic component for a home automation system.

Thanks for an interesting concept! Did you try to improve math reasoning for 70B llama2? I have a small cluster with 8 GPUs NVIDIA A100 40G and try to find a dataset to improve the base model

Discord: discord.gg/F7pjXfVJwZ

amazing video!

great video!

Will this work on 2 tesla v100 64GB?

Nice i ll try that in Touchdesigner i have a chat gpt reader and i was looking for something to send my voice to chat chpt. This will work amazing

Can i run it by 4090 ??

you need 2 x 3090 or 48GB VRAM to finetune 70B model ? So for 13B model, I should be able to do the same with 1 3090 card? I hope to have more hardware requirement details so that I can determine with this procedure is useful to me. Thanks!

Hello, I just found this repo today and I want to use it translator for korean... how to set up this language to others?

what's that terminal? just i cant use theese commands

I'd like to do some testing and research with this, but my machine doesn't have a GPU, I have approx 30 RAM and a multi-core CPU. I don't want to pay a virtual machine in the cloud for research, it will be wasted money, I prefer to understand how it works on my local machine and later when I understand how to do it, pay to train it. What are you suggesting? I think I've read that you can train without a GPU, but I don't know where to start.

nice. how would i stop the loop in my assistant?

This video is just installing WSL and Docker. To use NVIDIA gpus you have to take additional steps including the NVIDIA Container Toolkit