AI model to predict the output, anything available?
- Thread starter Purush
- Start date
Purush
Elite Member
- Jul 12, 2016
- 6,750
- 2,427
3.5 mean chatgpt u r referring here.
Here is the chatgpt response.
Unfortunately, as of my last training data in September 2021, you can't directly train or fine-tune GPT models, like ChatGPT, via the OpenAI API. OpenAI's API only allows you to use pretrained models for inference (i.e., generating responses to text prompts). Fine-tuning and training have been reserved for specific partnerships and applications, usually carried out internally by OpenAI.
August 22, 2023 they released fine tuning for 3.5. Don't ask ChatGPT about anything after 2021, it doesn't even know it exists.
CB20C3
Banned - Selling via PM
- Nov 20, 2022
- 329
- 167
nice , do you recommended one which is free and consume less computer resources.
Purush
Elite Member
- Jul 12, 2016
- 6,750
- 2,427
Again it's prompt like option but my case will be two page document
- May 27, 2016
- 11,184
- 2,543
That sure is valuable information. Time to fine tune it!
MAF Publishing
Regular Member
- Jun 26, 2023
- 480
- 354
Fine tuning jobs on 3.5 are surprisingly cheap. I put together a small tuning for $6
- Mar 2, 2021
- 4,080
- 1,206
More reason why ChatGPT is getting boring and less of advantage to users. Only good for imaginations and past histories.
The new 3.5 Completions is incredibly powerful. I don't know why they're deciding to focus on "Chat GPT" when GPT-3 Davinci was a million times better. Instruct is way better at not sucking than Chat GPT.