[HN Gopher] Ask HN: Self-hosted/open-source ChatGPT alternative?...
___________________________________________________________________
Ask HN: Self-hosted/open-source ChatGPT alternative? Like Stable
Diffusion
Are there any self-hosted and open-source ChatGPT alternatives?
Author : jakearmitage
Score : 33 points
Date : 2022-12-12 14:45 UTC (8 hours ago)
| fswd wrote:
| What is your use case? I could build this for you, email is in
| profile.
| jollyllama wrote:
| The project has only one dependency!
| https://github.com/labteral/chatgpt-python
| boredemployee wrote:
| I tried character.ai and while it has some good answers, also it
| can enter in a weird and nonsense loop of answers.
| kanwisher wrote:
| None yet, at least that are competitive. The expensive of
| creating the model is over 10million+.
| burn_cycle wrote:
| Is this compute alone? Also, how likely do you think it is that
| we'll see anything in the next 12(ish) months?
| lioeters wrote:
| I know nothing, but have heard Hugging Face is in that direction.
|
| https://github.com/huggingface/transformers
|
| >Transformers provides thousands of pretrained models to perform
| tasks on different modalities such as text, vision, and audio.
|
| > These models can be applied on:
|
| > - Text, for tasks like text classification, information
| extraction, question answering, summarization, translation, text
| generation, in over 100 languages.
|
| > - Images, for tasks like image classification, object
| detection, and segmentation.
|
| > - Audio, for tasks like speech recognition and audio
| classification.
|
| ---
|
| Also I read about GPT-J, whose capability is comparable with
| GTP-3.
|
| https://github.com/kingoflolz/mesh-transformer-jax
|
| But I believe it requires buying or renting GPUs.
| theCrowing wrote:
| The best options are probably GPT-J and BLOOM but they are both
| so big they need at least a 24GB of VRAM and they are not as
| suffisticated.
| musicale wrote:
| > not as suffisticated
|
| ChatGPT in a nutshell.
|
| (I hadn't heard the slang before, but it seems descriptive for
| a false air of sophistication, or [possibly over-]sophisticated
| complexity that nevertheless leads to bad results.)
| canadianfella wrote:
| tough wrote:
| you can fine-tune in 8bits weights for use in collab or
| similarly more restrained environments
| https://github.com/huggingface/transformers/issues/14839
| https://huggingface.co/hivemind/gpt-j-6B-8bit
|
| by fine-tuning you should be able to attain niche-specific
| better results than default chatGPT theoretically, if you're
| lucky and thorough I guess
___________________________________________________________________
(page generated 2022-12-12 23:02 UTC)