Behavior-Sequence-Transformer-Pytorch
Basic-UI-for-GPT-J-6B-with-low-vram
Our great sponsors
Behavior-Sequence-Transformer-Pytorch | Basic-UI-for-GPT-J-6B-with-low-vram | |
---|---|---|
1 | 4 | |
129 | 113 | |
- | - | |
0.0 | 0.0 | |
almost 2 years ago | over 2 years ago | |
Jupyter Notebook | Jupyter Notebook | |
MIT License | Apache License 2.0 |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
Behavior-Sequence-Transformer-Pytorch
Basic-UI-for-GPT-J-6B-with-low-vram
-
How to run this service with a local GPU?
You need a lot of VRAM to run the AI models, scaling somewhat with the amount of parameters a model uses. The most advanced model Pygmalion has is 6 billion parameters, which requires a minimum of 16GB of VRAM to run locally at decent speeds. There are methods of running 6b locally on low VRAM machines as listed here: https://github.com/arrmansa/Basic-UI-for-GPT-J-6B-with-low-vram but even then, the generations would be excruciatingly slow, and the lowest VRAM card used with this method has 6GB of VRAM.
-
Tesla M40 and GPT-J-6B
While waiting however I came across https://github.com/arrmansa/Basic-UI-for-GPT-J-6B-with-low-vram which allows you to use some of system memory to run the model. I was able to get a version working with 2.7B on my 2060 6GB with KoboldAI. The github above has an error that prevents it from working (https://github.com/arrmansa/Basic-UI-for-GPT-J-6B-with-low-vram/issues/1), but other than that it works.
-
How is any of this even possible?
Just to add to this, there is a low VRAM version of GPT-J here (suggest 16GB RAM + 8GB GPU).
-
GPT-J 6B locally on my computer
I found this yesterday, is it somehow possible to use this with KoboldAI to run GPT-J on weaker graphics cards?
What are some alternatives?
Stock-Prediction-Models - Gathers machine learning and deep learning models for Stock forecasting including trading bots and simulations
gpt-neo_dungeon - Colab notebooks to run a basic AI Dungeon clone using gpt-neo-2.7B
pytorch-seq2seq - Tutorials on implementing a few sequence-to-sequence (seq2seq) models with PyTorch and TorchText.
adaptnlp - An easy to use Natural Language Processing library and framework for predicting, training, fine-tuning, and serving up state-of-the-art NLP models.
nn - 🧑🏫 60 Implementations/tutorials of deep learning papers with side-by-side notes 📝; including transformers (original, xl, switch, feedback, vit, ...), optimizers (adam, adabelief, sophia, ...), gans(cyclegan, stylegan2, ...), 🎮 reinforcement learning (ppo, dqn), capsnet, distillation, ... 🧠
clip-italian - CLIP (Contrastive Language–Image Pre-training) for Italian
pytorch-sentiment-analysis - Tutorials on getting started with PyTorch and TorchText for sentiment analysis.
pytorch-generative - Easy generative modeling in PyTorch.
Eleya - Artificial Intelligence That Generate Novel Biomedical Text