BlinkDL
/

rwkv-4-pile-14b

Text Generation

Model card Files Files and versions Community

BlinkDL commited on Nov 18, 2022

Commit

74a3f96

•

1 Parent(s): bb4450e

Update README.md

Files changed (1) hide show

README.md +6 -6

README.md CHANGED Viewed

@@ -26,9 +26,9 @@ ctx_len = 1024
 n_layer = 40
 n_embd = 5120
-Preview checkpoint: RWKV-4-Pile-14B-20221103-1214.pth : Trained on the Pile for 50B tokens.
-* Pile loss 1.9533
-* LAMBADA ppl 5.07, acc 64.08%
-* PIQA acc 73.99%
-* SC2016 acc 72.05%
-* Hellaswag acc_norm 61.07%

 n_layer = 40
 n_embd = 5120
+Preview checkpoint: RWKV-4-Pile-14B-20221116-2258.pth : Trained on the Pile for 93B tokens.
+* Pile loss 1.8758
+* LAMBADA ppl 4.35, acc 67.98%
+* PIQA acc 75.84%
+* SC2016 acc 72.96%
+* Hellaswag acc_norm 65.02%