Update README.md
Browse files
README.md
CHANGED
@@ -26,9 +26,9 @@ ctx_len = 1024
|
|
26 |
n_layer = 32
|
27 |
n_embd = 2560
|
28 |
|
29 |
-
|
30 |
-
* Pile loss 1.
|
31 |
-
* LAMBADA ppl 5.
|
32 |
-
* PIQA acc
|
33 |
-
* SC2016 acc 70.
|
34 |
-
* Hellaswag acc_norm
|
|
|
26 |
n_layer = 32
|
27 |
n_embd = 2560
|
28 |
|
29 |
+
Final checkpoint: RWKV-4-Pile-3B-20221008-8023.pth : Trained on the Pile for 331B tokens.
|
30 |
+
* Pile loss 1.9469
|
31 |
+
* LAMBADA ppl 5.24, acc 63.94%
|
32 |
+
* PIQA acc 73.72%
|
33 |
+
* SC2016 acc 70.28%
|
34 |
+
* Hellaswag acc_norm 59.63%
|