BlinkDL
/

rwkv-4-pile-7b

Text Generation

Model card Files Files and versions Community

BlinkDL commited on Oct 20, 2022

Commit

428b12d

•

1 Parent(s): 8ce9232

Update README.md

Files changed (1) hide show

README.md +5 -5

README.md CHANGED Viewed

@@ -26,9 +26,9 @@ ctx_len = 1024
 n_layer = 32
 n_embd = 4096
-Preview checkpoint: RWKV-4-Pile-7B-20221010-3787.pth : Trained on the Pile for 156B tokens.
-* Pile loss 1.8935
-* LAMBADA ppl 4.60, acc 65.52%
 * PIQA acc 75.63%
-* SC2016 acc 72.21%
-* Hellaswag acc_norm 63.22%

 n_layer = 32
 n_embd = 4096
+Preview checkpoint: RWKV-4-Pile-7B-20221019-5163.pth : Trained on the Pile for 213B tokens.
+* Pile loss 1.8684
+* LAMBADA ppl 4.46, acc 66.45%
 * PIQA acc 75.63%
+* SC2016 acc 72.63%
+* Hellaswag acc_norm 64.58%