Update README.md

README.md

@@ -113,4 +113,30 @@ configs:
   data_files:
   - split: train
     path: wiki/train-*
+task_categories:
+- text-generation
+language:
+- en
+size_categories:
+- 10M<n<100M
 ---
+
+A small, aggressively cleaned and de-duped pre-training corpus for academic settings. It aims to recreate something akin to [The Pile](https://huggingface.co/datasets/EleutherAI/pile) but prioritizes quality for the constrained token budget academic researchers live with.
+
+It has seven config subsets and an eighth `all` subset that combines them for a total of ~91B tokens (GPT2 Tokenizer estimate). These splits are as follows:
+
+1. `c4_realnews`: The RealNews domain subset of the C4 dataset containing news articles.
+2. `openwebtext`: The OpenWebText dataset containing the contents of the links mentioned in Reddit posts with at least 3 upvotes.
+3. `peS2o`: The PeS2o dataset containing academic articles from Semantic Scholar.
+4. `redpajama_books`: The books subset of RedPajama V1.
+5. `stackexchange`: The EN StackExchange non-code subset of the BigScience ROOTs dataset.
+6. `uspto`: The EN USPTO patent applications contents' subset of the BigScience ROOTs dataset.
+7. `wiki`: The EN Wiki subset of the BigScience ROOTs dataset.
+
+The following processing and filtering steps have been applied:
+
+1. Removed citation text and bibliography information for academic texts.
+2. Ran a perplexity filter using a KenLM model trained on the EN OSCAR corpus and removed documents with a perplexity of more than 325 and less than 7.
+3. Removed samples which have a repeating <=4-gram proportion of 15%.
+4. Removed samples which have lower than 99% confidence of being EN using the lingua language detector.
+5. Performed an aggressive MinHash de-dupe using a shingle size of 8 and a low threshold of 0.5.