zhangirazerbayev commited on
Commit
b18f429
1 Parent(s): 664e7c5

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -1
README.md CHANGED
@@ -23,7 +23,7 @@ task_ids:
23
 
24
  # Dataset Description
25
  The `proof-pile` is a 13GB pre-training dataset of mathematical text that comprises 8.3 billion tokens (using the `gpt-neox` tokenizer). Models trained on this dataset are coming soon :) The dataset is composed of diverse sources of both informal and formal mathematics, namely
26
- - ArXiv.math (35GB)
27
  - Open-source math textbooks (50MB)
28
  - Formal mathematics libraries (500MB)
29
  - Lean mathlib and other Lean repositories
 
23
 
24
  # Dataset Description
25
  The `proof-pile` is a 13GB pre-training dataset of mathematical text that comprises 8.3 billion tokens (using the `gpt-neox` tokenizer). Models trained on this dataset are coming soon :) The dataset is composed of diverse sources of both informal and formal mathematics, namely
26
+ - ArXiv.math (10GB)
27
  - Open-source math textbooks (50MB)
28
  - Formal mathematics libraries (500MB)
29
  - Lean mathlib and other Lean repositories