Datasets:

Modalities:
Text
Languages:
English
Libraries:
Datasets
License:
zhangirazerbayev commited on
Commit
3246dea
1 Parent(s): 187fe1c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +8 -3
README.md CHANGED
@@ -25,7 +25,7 @@ Note: this repo is a WIP and does not yet implement all features described below
25
  # Dataset Card for Proof-pile
26
 
27
  # Dataset Description
28
- The `proof-pile` is a 45GB pre-training dataset of mathematical text. The dataset is composed of diverse sources of both informal and formal mathematics, namely
29
  - ArXiv.math (40GB)
30
  - Open-source math textbooks (50MB)
31
  - Formal mathematics libraries (500MB)
@@ -47,10 +47,15 @@ This dataset is intended to be used for pre-training language models. We envisio
47
  # Languages
48
  All informal mathematics in the `proof-pile` is written in English and LaTeX (arXiv articles in other languages are filtered out using [languagedetect](https://github.com/shuyo/language-detection/blob/wiki/ProjectHome.md)). Formal theorem proving languages represented in this dataset are Lean 3, Isabelle, Coq, HOL Light, Metamath, and Mizar.
49
 
50
- # Splits
51
  The data is sorted into `"arxiv", "books", "formal", "stack-exchange", "wiki",` and `"math-dataset"` configurations. This is so that it is easy to upsample particular configurations during pre-training with the `datasets.interleave_datasets()` function.
52
 
53
- Note that in the `"stack-exchange"`, `"wiki"`, and `"stack-exchange"` configurations, multiple documents are included in them same instance separated by the string `"<|endoftext|>"`.
 
 
 
 
 
54
 
55
  ## Contributions
56
  Authors: Zhangir Azerbayev, Edward Ayers, Bartosz Piotrowski.
 
25
  # Dataset Card for Proof-pile
26
 
27
  # Dataset Description
28
+ The `proof-pile` is a 40GB pre-training dataset of mathematical text. The dataset is composed of diverse sources of both informal and formal mathematics, namely
29
  - ArXiv.math (40GB)
30
  - Open-source math textbooks (50MB)
31
  - Formal mathematics libraries (500MB)
 
47
  # Languages
48
  All informal mathematics in the `proof-pile` is written in English and LaTeX (arXiv articles in other languages are filtered out using [languagedetect](https://github.com/shuyo/language-detection/blob/wiki/ProjectHome.md)). Formal theorem proving languages represented in this dataset are Lean 3, Isabelle, Coq, HOL Light, Metamath, and Mizar.
49
 
50
+ # Configurations
51
  The data is sorted into `"arxiv", "books", "formal", "stack-exchange", "wiki",` and `"math-dataset"` configurations. This is so that it is easy to upsample particular configurations during pre-training with the `datasets.interleave_datasets()` function.
52
 
53
+ # Evaluation
54
+ The version of `set.mm` in this dataset has 10% of proofs replaced with `?` in order to preserve a validation and test set for Metamath provers pre-trained on the `proof-pile`. The precise split can be found here: [validation](https://github.com/zhangir-azerbayev/mm-extract/blob/main/valid_decls.json) and [test](https://github.com/zhangir-azerbayev/mm-extract/blob/main/test_decls.json).
55
+
56
+ The Lean mathlib commit used in this dataset is `6313863`. Theorems created in subsequent commits can be used for evaluating Lean theorem provers.
57
+
58
+ This dataset contains only the training set of the [MATH dataset](https://github.com/hendrycks/math). However, because this dataset contains ProofWiki, the Stacks Project, Trench's Analysis, and Stein's Number Theory, the `proof-pile` contains the validation and test sets of the [NaturalProofs dataset](https://github.com/wellecks/naturalproofs).
59
 
60
  ## Contributions
61
  Authors: Zhangir Azerbayev, Edward Ayers, Bartosz Piotrowski.