Our great sponsors
-
WorkOS
The modern identity platform for B2B SaaS. The APIs are flexible and easy-to-use, supporting authentication, user identity, and complex enterprise features like SSO and SCIM provisioning.
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 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.
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 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.
The project has only one dependency! https://github.com/labteral/chatgpt-python
https://github.com/KoboldAI/KoboldAI-Client
It's not a direct alternative, but it's the closest that can be run on consumer hardware. I run it locally but I have heard a lot of success about it being run on Collab as well and there are notebooks maintained in the repo. It can use OPT, Fairseq Dense and the older models like GPT-J. There are fine tuned models for NSFW content as well, my understanding is that those models were motivated to move away from AI Dungeon that censored and read the text of user's stories. The models on hugging face come in all sizes, 8bg VRAM is enough for a 2.7B model, but there's options ranging from 350M [1] to a OPT 66B in the FB/Meta AI repo released on May 3rd [2], the 66B parameter one is openly available and the full 175B parameter model is available via request. I found 2.7B and 6.7B impressive personally. 66B would take over hundreds of gigabytes of vram to run at a similar speed to how ChatGPT works. I think that's part of the reason why it's not as popular, very few people can run it at all.
[1] https://huggingface.co/KoboldAI
[2] https://huggingface.co/facebook/opt-66b
https://ai.facebook.com/blog/democratizing-access-to-large-s...
Related posts
- [P] Discord Chatbot for LLaMA 4-bit quantized that runs 13b in <9 GiB VRAM
- Lossless Acceleration of LLM via Adaptive N-Gram Parallel Decoding
- AI enthusiasm #6 - Finetune any LLM you want💡
- Zephyr 141B, a Mixtral 8x22B fine-tune, is now available in Hugging Chat
- Schedule-Free Learning – A New Way to Train