Datasets:

LLM360
/

TxT360

License:

victormiller commited on 9 days ago

Commit

83de6dd

•

1 Parent(s): a1061a9

Update README.md

Files changed (1) hide show

README.md CHANGED Viewed

@@ -28,7 +28,7 @@ Complete details on the dataset can be found in our blog post [here](https://hug
 ## Upsampling Experiment with Comparison to FineWeb
 To evaluate the training efficiency of our dataset, we sampled 1.5T tokens from both FineWeb and TxT360 (using the aforementioned weighting) and conducted a training ablation on an 8x8B Mixture-of-Experts architecture, similar to Mixtral. We compared the learning curves by tracking training loss, validation scores, and performance across a wide array of diverse evaluation benchmarks. The validation set was sampled independently from SlimPajama. Note that this experiment is done on a slightly earlier version of the dataset.
-<center><img src="txt360tofinewebtweet.png" alt="comparison" /></center>
 ## Initial Data Representation

 ## Upsampling Experiment with Comparison to FineWeb
 To evaluate the training efficiency of our dataset, we sampled 1.5T tokens from both FineWeb and TxT360 (using the aforementioned weighting) and conducted a training ablation on an 8x8B Mixture-of-Experts architecture, similar to Mixtral. We compared the learning curves by tracking training loss, validation scores, and performance across a wide array of diverse evaluation benchmarks. The validation set was sampled independently from SlimPajama. Note that this experiment is done on a slightly earlier version of the dataset.
+<center><img src=" txttofineweb.png" alt="comparison" /></center>
 ## Initial Data Representation