Datasets:
yury-zyphra
commited on
Commit
•
a0f17a0
1
Parent(s):
c408126
Update README.md (#114)
Browse files- Update README.md (b56faac15baad275c8e3dbae1bd91d760b9b2e73)
README.md
CHANGED
@@ -44,11 +44,10 @@ According to our evaluations, Zyda-2 is the most performant per-token open datas
|
|
44 |
|
45 |
|
46 |
<center>
|
47 |
-
<img src="https://cdn-uploads.huggingface.co/production/uploads/
|
48 |
</center>
|
49 |
|
50 |
|
51 |
-
|
52 |
For more information, please see our technical blog (-/TODO LINK)
|
53 |
|
54 |
## How to download
|
@@ -95,15 +94,15 @@ Each component has they're own individual schema. Please, consult with their res
|
|
95 |
|
96 |
However, in all components the document text is in `text` column, and unique document document id is in `nemo_id` column.
|
97 |
|
98 |
-
Our
|
99 |
|
100 |
### Source Data
|
101 |
|
102 |
Zyda-2 is comprised of four high quality open-source datasets:
|
103 |
|
104 |
-
|
105 |
|
106 |
-
Dolma-
|
107 |
|
108 |
DCLM-baseline: https://huggingface.co/datasets/mlfoundations/dclm-baseline-1.0
|
109 |
|
|
|
44 |
|
45 |
|
46 |
<center>
|
47 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/65455aca468722e935103b17/-nxHBcU38QJ-MNdKXPiYS.png" width="600" alt="Zyda-2 evaluation scores">
|
48 |
</center>
|
49 |
|
50 |
|
|
|
51 |
For more information, please see our technical blog (-/TODO LINK)
|
52 |
|
53 |
## How to download
|
|
|
94 |
|
95 |
However, in all components the document text is in `text` column, and unique document document id is in `nemo_id` column.
|
96 |
|
97 |
+
Our Zyda-1 and Dolma-CC versions also have two additional columns corresponding to prediction of Nvidia's quality model (https://huggingface.co/nvidia/quality-classifier-deberta): `quality_prob` and `quality_pred`.
|
98 |
|
99 |
### Source Data
|
100 |
|
101 |
Zyda-2 is comprised of four high quality open-source datasets:
|
102 |
|
103 |
+
Zyda-1: https://huggingface.co/datasets/Zyphra/Zyda
|
104 |
|
105 |
+
Dolma-CC v1.7: https://huggingface.co/datasets/allenai/dolma
|
106 |
|
107 |
DCLM-baseline: https://huggingface.co/datasets/mlfoundations/dclm-baseline-1.0
|
108 |
|