Bronze V
1,372 Elo
- #8,574 on the ladder
- +2 today · −16 this week
- Peak 1,425
An implementation of model parallel GPT-2 and GPT-3-style models using the mesh-tensorflow library.
EleutherAI/gpt-neo is #8,574 on the bestofgithub Elo ladder, in Bronze V, with a rating of 1,372 (−16 this week). It has 8,248 stars.
1,372 Elo
8,248
git clone https://github.com/EleutherAI/gpt-neo.gitRWKV (pronounced RwaKuv) is an RNN with great LLM performance, which can also be directly trained like a GPT transformer (parallelizable). We are at RWKV-7 "Goose". So it's combining the best of RNN and transformer - great performance, linear time, constant space (no kv-cache), fast training, infinite ctx_len, and free sentence embedding.
git clone https://github.com/BlinkDL/RWKV-LM.gitAn implementation of model parallel autoregressive transformers on GPUs, based on the Megatron and DeepSpeed libraries
git clone https://github.com/EleutherAI/gpt-neox.gitCode for loralib, an implementation of "LoRA: Low-Rank Adaptation of Large Language Models"
git clone https://github.com/microsoft/LoRA.git💥 Fast State-of-the-Art Tokenizers optimized for Research and Production
git clone https://github.com/huggingface/tokenizers.git🤖 Awesome list for ChatGPT — an artificial intelligence chatbot developed by OpenAI [SUBMISSIONS ARE TEMPORARILY PAUSED BECAUSE OF TOO MANY LOW QUALITY SUBMISSIONS]
git clone https://github.com/sindresorhus/awesome-chatgpt.gitThis repository contains demos I made with the Transformers library by HuggingFace.
git clone https://github.com/NielsRogge/Transformers-Tutorials.gitLoading README…