readme
Browse files
README.md
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
This file contains the embeddings for the full fineweb-edu dataset.
|
| 2 |
+
|
| 3 |
+
The dataset has been deduplicated (using only exact deduplication).
|
| 4 |
+
|
| 5 |
+
The emb folder contains for each parquet file a new_{parquet_name}.npy and old_{parquet}.npy file. The old refers to text that has been seen in the smaller 10B/100B/350B data samples.
|
| 6 |
+
|
| 7 |
+
Cohere embed-multilingual-v3.0 model has been used.
|
| 8 |
+
|
| 9 |
+
The index of the dataset can be found here:
|
| 10 |
+
https://huggingface.co/datasets/Cohere/fineweb-edu-index
|
| 11 |
+
|
| 12 |
+
The corpus can be found here:
|
| 13 |
+
https://huggingface.co/datasets/Cohere/fineweb-edu-corpus
|