Fine tuning a model like llama is super easy. You can fine tune models up to 33B parameters on a single consumer GPU at around 3 epochs per minute (less than 6 hours for 1000 epochs).
You can even fine tune on a free Google Colab GPU.
Running it fully locally isn't easy because you need a complex setup with GPUs. Can you give some more color on exactly what you are trying to do (number of PDFs, types of queries, etc.)? With some more info we can give you some guidance.
oobabooga's textgen webui has a tab for fine tuning now. You only need a single consumer GPU to fine tune up to 33B parameter models at a rate of about 200 epochs per hour, per GPU.
You are right. We are automating the creation of offer letters, employment contracts and confirmation letters by telling the developer exactly what to do and then testing what he did. And then still we are already working 18 months on this! 3 different developers already on the project. So if you try to create a solution without the directions from someone who actually does the work, good luck to ya!
I want to search youtube transcripts. Is that possible? Like google search but instead of articles it searching in the transcripts of all youtube videos
We have a whole Concur team of like 5 people at least. All fte's. Employees still have to declare their own expenses. Almost 30k employees. So your suggestion of having just 1 person handling all declarations won't work unless you are suggesting 5 fte's for just Concur isn't enough.
For 30k not enough. Usually an office admin person can process that kind of stuff as part time responsibility. And you have one for every 50-100 employees. So that would be 300-600 PTE users for 30k
Yeah that will never happen. Overhead is already a few percentage points above the industry average. Self service is the industry trend as I see it in this business. Whatever vendor facilitates that will win our business.
