{"payload":{"pageCount":5,"repositories":[{"type":"Public","name":"lm-evaluation-harness","owner":"EleutherAI","isFork":false,"description":"A framework for few-shot evaluation of language models.","topicNames":["transformer","language-model","evaluation-framework"],"topicsNotShown":0,"allTopics":["transformer","language-model","evaluation-framework"],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":58,"issueCount":193,"starsCount":5338,"forksCount":1388,"license":"MIT License","participation":[14,56,104,106,56,97,100,70,30,73,96,145,65,60,54,66,67,13,37,21,46,7,48,23,33,40,102,50,50,30,8,7,11,17,15,14,9,8,15,16,17,9,13,6,5,2,4,5,8,14,3,12],"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-26T10:54:55.500Z"}},{"type":"Public","name":"rnngineering","owner":"EleutherAI","isFork":false,"description":"Engineering the state of RNN language models (Mamba, RWKV, etc.)","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Jupyter Notebook","color":"#DA5B0B"},"pullRequestCount":0,"issueCount":0,"starsCount":31,"forksCount":2,"license":"MIT License","participation":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,2,3,1,13,7,0,0,0,0,0,0,0,0,0,2],"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-25T16:13:46.749Z"}},{"type":"Public","name":"website","owner":"EleutherAI","isFork":false,"description":"New website for EleutherAI based on Hugo static site generator","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"HTML","color":"#e34c26"},"pullRequestCount":2,"issueCount":0,"starsCount":4,"forksCount":6,"license":null,"participation":[0,0,0,2,0,3,0,0,8,0,0,0,2,0,0,0,12,0,0,9,8,9,0,2,11,0,14,0,20,21,0,0,0,0,0,2,0,0,0,8,0,0,0,25,0,0,21,0,0,0,0,15],"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-24T23:49:16.489Z"}},{"type":"Public","name":"sae","owner":"EleutherAI","isFork":false,"description":"Sparse autoencoders","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":0,"issueCount":0,"starsCount":2,"forksCount":0,"license":"MIT License","participation":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,4],"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-24T23:39:43.275Z"}},{"type":"Public","name":"aria-amt","owner":"EleutherAI","isFork":false,"description":"Efficient and robust implementation of seq-to-seq automatic piano transcription.","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":2,"issueCount":0,"starsCount":6,"forksCount":4,"license":"Apache License 2.0","participation":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,6,3,3,6,3,1,0,0,2,3,4,1,1,1,2],"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-24T17:07:53.548Z"}},{"type":"Public","name":"cookbook","owner":"EleutherAI","isFork":false,"description":"Deep learning for dummies. All the practical details and useful utilities that go into working with real models.","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":3,"issueCount":6,"starsCount":186,"forksCount":13,"license":"Apache License 2.0","participation":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,13,0,4,5,0,3,0,0,4,3,0,3,0,0,0,0,0,1,0,0,0,0,0],"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-24T06:59:44.529Z"}},{"type":"Public","name":"bayesian-adam","owner":"EleutherAI","isFork":false,"description":"Exactly what it says on the tin","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":0,"issueCount":0,"starsCount":1,"forksCount":0,"license":"MIT License","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-24T05:26:48.375Z"}},{"type":"Public","name":"tokengrams","owner":"EleutherAI","isFork":false,"description":"Efficiently computing & storing token n-grams from large corpora","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Rust","color":"#dea584"},"pullRequestCount":0,"issueCount":1,"starsCount":7,"forksCount":1,"license":"MIT License","participation":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,5,4,0,0,0,1,0,0,0,0,0,6,6,24,0,0,0,0,0,0,0],"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-24T01:30:57.180Z"}},{"type":"Public","name":"w2s","owner":"EleutherAI","isFork":false,"description":"","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":0,"issueCount":1,"starsCount":0,"forksCount":0,"license":"MIT License","participation":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,17,8,0],"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-24T01:22:10.576Z"}},{"type":"Public","name":"features-across-time","owner":"EleutherAI","isFork":false,"description":"Understanding how features learned by neural networks evolve throughout training","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":1,"issueCount":0,"starsCount":25,"forksCount":1,"license":"MIT License","participation":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0,1,32,69,69,3,0,0,0,0,0,4,0,0,0,0,0,0,0,0,0],"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-24T01:08:19.712Z"}},{"type":"Public","name":"elk-generalization","owner":"EleutherAI","isFork":false,"description":"Investigating the generalization behavior of LM probes trained to predict truth labels: (1) from one annotator to another, and (2) from easy questions to hard","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":0,"issueCount":1,"starsCount":21,"forksCount":3,"license":"MIT License","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-23T23:25:42.706Z"}},{"type":"Public","name":"tuned-lens","owner":"EleutherAI","isFork":true,"description":"Tools for understanding how transformer predictions are built layer-by-layer","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":1,"issueCount":0,"starsCount":2,"forksCount":36,"license":"MIT License","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-23T20:10:19.155Z"}},{"type":"Public","name":"cupbearer","owner":"EleutherAI","isFork":true,"description":"A library for mechanistic anomaly detection","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":0,"issueCount":0,"starsCount":0,"forksCount":3,"license":"MIT License","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-23T06:56:23.122Z"}},{"type":"Public","name":"concept-erasure","owner":"EleutherAI","isFork":false,"description":"Erasing concepts from neural representations with provable guarantees","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":1,"issueCount":2,"starsCount":194,"forksCount":15,"license":"MIT License","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-23T05:15:19.958Z"}},{"type":"Public","name":"SAELens","owner":"EleutherAI","isFork":true,"description":"Training Sparse Autoencoders on Language Models","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"HTML","color":"#e34c26"},"pullRequestCount":0,"issueCount":0,"starsCount":0,"forksCount":49,"license":"MIT License","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-22T19:41:40.681Z"}},{"type":"Public","name":"gpt-neox","owner":"EleutherAI","isFork":false,"description":"An implementation of model parallel autoregressive transformers on GPUs, based on the Megatron and DeepSpeed libraries","topicNames":["transformers","language-model","gpt-3","deepspeed-library"],"topicsNotShown":0,"allTopics":["transformers","language-model","gpt-3","deepspeed-library"],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":27,"issueCount":54,"starsCount":6631,"forksCount":961,"license":"Apache License 2.0","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-22T12:34:59.137Z"}},{"type":"Public","name":"pythia","owner":"EleutherAI","isFork":false,"description":"The hub for EleutherAI's work on interpretability and learning dynamics","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Jupyter Notebook","color":"#DA5B0B"},"pullRequestCount":2,"issueCount":14,"starsCount":2084,"forksCount":150,"license":"Apache License 2.0","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-22T01:45:07.678Z"}},{"type":"Public","name":"aria","owner":"EleutherAI","isFork":false,"description":"","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":2,"issueCount":0,"starsCount":36,"forksCount":6,"license":"Apache License 2.0","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-14T14:28:07.342Z"}},{"type":"Public","name":"elk","owner":"EleutherAI","isFork":false,"description":"Keeping language models honest by directly eliciting knowledge encoded in their activations.","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":10,"issueCount":13,"starsCount":171,"forksCount":32,"license":"MIT License","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-13T17:32:50.997Z"}},{"type":"Public","name":"weak-to-strong","owner":"EleutherAI","isFork":true,"description":"","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":2,"issueCount":2,"starsCount":6,"forksCount":291,"license":"MIT License","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-05-10T17:59:25.843Z"}},{"type":"Public","name":"semantic-memorization","owner":"EleutherAI","isFork":false,"description":"","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Jupyter Notebook","color":"#DA5B0B"},"pullRequestCount":5,"issueCount":0,"starsCount":34,"forksCount":3,"license":null,"participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-04-30T05:06:50.749Z"}},{"type":"Public","name":"variance-across-time","owner":"EleutherAI","isFork":false,"description":"Studying the variance in neural net predictions across training time","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":1,"issueCount":0,"starsCount":3,"forksCount":0,"license":"MIT License","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-04-24T07:12:33.410Z"}},{"type":"Public","name":"DeeperSpeed","owner":"EleutherAI","isFork":true,"description":"DeepSpeed is a deep learning optimization library that makes distributed training easy, efficient, and effective.","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":1,"issueCount":0,"starsCount":159,"forksCount":3899,"license":"Apache License 2.0","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-04-22T07:31:01.226Z"}},{"type":"Public","name":"improved-t5","owner":"EleutherAI","isFork":false,"description":"Experiments for efforts to train a new and improved t5","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":0,"issueCount":1,"starsCount":75,"forksCount":5,"license":null,"participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-04-15T05:51:48.509Z"}},{"type":"Public","name":"CAA","owner":"EleutherAI","isFork":true,"description":"Steering Llama 2 with Contrastive Activation Addition","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Jupyter Notebook","color":"#DA5B0B"},"pullRequestCount":0,"issueCount":0,"starsCount":0,"forksCount":19,"license":"MIT License","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-03-28T21:51:28.715Z"}},{"type":"Public","name":"math-lm","owner":"EleutherAI","isFork":false,"description":"","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":1,"issueCount":9,"starsCount":991,"forksCount":74,"license":"MIT License","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-03-12T05:10:40.751Z"}},{"type":"Public","name":"pd-books","owner":"EleutherAI","isFork":false,"description":"","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Jupyter Notebook","color":"#DA5B0B"},"pullRequestCount":1,"issueCount":0,"starsCount":1,"forksCount":1,"license":"Apache License 2.0","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-03-07T20:44:37.188Z"}},{"type":"Public","name":"RWKV-LM","owner":"EleutherAI","isFork":true,"description":"RWKV is an RNN with transformer-level LLM performance. It can be directly trained like a GPT (parallelizable). So it's combining the best of RNN and transformer - great performance, fast inference, saves VRAM, fast training, \"infinite\" ctx_len, and free sentence embedding.","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Python","color":"#3572A5"},"pullRequestCount":0,"issueCount":0,"starsCount":1,"forksCount":809,"license":"Apache License 2.0","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-02-21T10:01:35.441Z"}},{"type":"Public","name":"minetest","owner":"EleutherAI","isFork":true,"description":"Minetest is an open source voxel game engine with easy modding and game creation","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"C++","color":"#f34b7d"},"pullRequestCount":5,"issueCount":16,"starsCount":57,"forksCount":1960,"license":"Other","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-02-19T22:53:24.144Z"}},{"type":"Public","name":"conceptual-constraints","owner":"EleutherAI","isFork":false,"description":"Applying LEACE to models during training","topicNames":[],"topicsNotShown":0,"allTopics":[],"primaryLanguage":{"name":"Jupyter Notebook","color":"#DA5B0B"},"pullRequestCount":0,"issueCount":0,"starsCount":1,"forksCount":0,"license":"MIT License","participation":null,"lastUpdated":{"hasBeenPushedTo":true,"timestamp":"2024-02-17T01:46:51.746Z"}}],"repositoryCount":138,"userInfo":null,"searchable":true,"definitions":[],"typeFilters":[{"id":"all","text":"All"},{"id":"public","text":"Public"},{"id":"source","text":"Sources"},{"id":"fork","text":"Forks"},{"id":"archived","text":"Archived"},{"id":"template","text":"Templates"}],"compactMode":false},"title":"Repositories"}