Update README.md
Browse files
README.md
CHANGED
|
@@ -43,6 +43,27 @@ The released weights were trained on ~70 billion tokens.
|
|
| 43 |
We plan to continue training up to 300 billion tokens and update the weights at every 20b tokens.
|
| 44 |
This training run is monolingual and uses c4en and english wikipedia datasets.
|
| 45 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
## Installation
|
| 47 |
|
| 48 |
```shell
|
| 43 |
We plan to continue training up to 300 billion tokens and update the weights at every 20b tokens.
|
| 44 |
This training run is monolingual and uses c4en and english wikipedia datasets.
|
| 45 |
|
| 46 |
+
## Test results
|
| 47 |
+
|
| 48 |
+
These are the results from [EleutherAI/lm-evaluation-harness](https://github.com/EleutherAI/lm-evaluation-harness) at 80B (tokens trained) checkpoint.
|
| 49 |
+
|
| 50 |
+
| Task |Version| Metric |Value | |Stderr|
|
| 51 |
+
|--------------|------:|--------|-----:|---|-----:|
|
| 52 |
+
|anli_r1 | 0|acc |0.3150|± |0.0147|
|
| 53 |
+
|anli_r2 | 0|acc |0.3380|± |0.0150|
|
| 54 |
+
|anli_r3 | 0|acc |0.3367|± |0.0136|
|
| 55 |
+
|hellaswag | 0|acc |0.4761|± |0.0050|
|
| 56 |
+
| | |acc_norm|0.6308|± |0.0048|
|
| 57 |
+
|lambada_openai| 0|ppl |8.9700|± |0.2606|
|
| 58 |
+
| | |acc |0.5628|± |0.0069|
|
| 59 |
+
|mathqa | 0|acc |0.2318|± |0.0077|
|
| 60 |
+
| | |acc_norm|0.2372|± |0.0078|
|
| 61 |
+
|piqa | 0|acc |0.7448|± |0.0102|
|
| 62 |
+
| | |acc_norm|0.7639|± |0.0099|
|
| 63 |
+
|winogrande | 0|acc |0.5935|± |0.0138|
|
| 64 |
+
|wsc | 0|acc |0.4038|± |0.0483|
|
| 65 |
+
|
| 66 |
+
|
| 67 |
## Installation
|
| 68 |
|
| 69 |
```shell
|