Update README.md
Browse files
README.md
CHANGED
@@ -144,7 +144,7 @@ We make several modifications to the exact substring deduplication implementatio
|
|
144 |
|
145 |
In Fig 14, we show the progression of accuracy with training for High Signal Tasks at 1.4 billion parameter model for 350 billion tokens. We see that for both datasets compared, the accuracy increases over time and the accuracy of the dataset with exact substring deduplication is consistently higher ending at 57.39 than the baseline which ends at 55.99
|
146 |
|
147 |
-
|
148 |
|
149 |
**Figure 14:** Ablation experiment comparing Exact Substring Deduplication against the FineWeb.V1.1 baseline at 1.4 Billion model size for 350 Billion tokens
|
150 |
|
|
|
144 |
|
145 |
In Fig 14, we show the progression of accuracy with training for High Signal Tasks at 1.4 billion parameter model for 350 billion tokens. We see that for both datasets compared, the accuracy increases over time and the accuracy of the dataset with exact substring deduplication is consistently higher ending at 57.39 than the baseline which ends at 55.99
|
146 |
|
147 |
+
<img src="RepRemoval.png" alt="RepRemoval.png" style="width:1000px;"/>
|
148 |
|
149 |
**Figure 14:** Ablation experiment comparing Exact Substring Deduplication against the FineWeb.V1.1 baseline at 1.4 Billion model size for 350 Billion tokens
|
150 |
|