Datasets:

Languages:
English
ArXiv:
License:
bhatta1 commited on
Commit
19113fc
·
verified ·
1 Parent(s): 6583d0f

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +2 -0
README.md CHANGED
@@ -103,6 +103,8 @@ For those in the greater than 5 Trillion token set size, in Fig 8 we show the pe
103
 
104
  In Fig 9, we show the progression of accuracy with training for High Signal Tasks for 1.4 billion parameter model for 350 billion tokens. We see that for all three datasets compared, the accuracy increases over time and the accuracy of GneissWeb is consistently higher than FineWeb and FineWeb-Edu-score-2.
105
 
 
 
106
  **Figure 9:** Average evaluation score on High-Signal tasks versus the number of tokens for 1.4 Billion parameter models. The model trained on GneissWeb consistently outperforms the ones trained on FineWeb.V1.1 and FineWeb-Edu-score-2.
107
 
108
  **At 3 and 7 Billion Model Size with 100 Billion Tokens**
 
103
 
104
  In Fig 9, we show the progression of accuracy with training for High Signal Tasks for 1.4 billion parameter model for 350 billion tokens. We see that for all three datasets compared, the accuracy increases over time and the accuracy of GneissWeb is consistently higher than FineWeb and FineWeb-Edu-score-2.
105
 
106
+ <img src="1.4Bmodels_100B.png" alt="1.4Bmodels_100B.png" style="width:1000px;"/>
107
+
108
  **Figure 9:** Average evaluation score on High-Signal tasks versus the number of tokens for 1.4 Billion parameter models. The model trained on GneissWeb consistently outperforms the ones trained on FineWeb.V1.1 and FineWeb-Edu-score-2.
109
 
110
  **At 3 and 7 Billion Model Size with 100 Billion Tokens**