Update README.md
Browse files
README.md
CHANGED
|
@@ -24,4 +24,11 @@ At this moment you have to use my Github code (https://github.com/BlinkDL/RWKV-L
|
|
| 24 |
|
| 25 |
ctx_len = 1024
|
| 26 |
n_layer = 32
|
| 27 |
-
n_embd = 2560
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
|
| 25 |
ctx_len = 1024
|
| 26 |
n_layer = 32
|
| 27 |
+
n_embd = 2560
|
| 28 |
+
|
| 29 |
+
Preview checkpoint: RWKV-4-Pile-3B-20220915-1207.pth : Trained on the Pile for 50B tokens.
|
| 30 |
+
* Pile loss 2.0902
|
| 31 |
+
* LAMBADA ppl 7.01, acc 57.11%
|
| 32 |
+
* PIQA acc 72.52%
|
| 33 |
+
* SC2016 acc 68.36%
|
| 34 |
+
* Hellaswag acc_norm 52.17%
|