File size: 684 Bytes
060b78b 0be695e 060b78b 0be695e 8ce9232 0be695e 060b78b 0be695e a07420d 0be695e 1e8417a 605a997 29fbd06 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 |
---
language:
- en
tags:
- pytorch
- text-generation
- causal-lm
- rwkv
license: apache-2.0
datasets:
- the_pile
---
# RWKV-4 7B
## Model Description
RWKV-4 7B is a L32-D4096 causal language model trained on the Pile. See https://github.com/BlinkDL/RWKV-LM for details.
Use https://github.com/BlinkDL/ChatRWKV to run it.
ctx_len = 1024
n_layer = 32
n_embd = 4096
(there are ctx_len 2048 and 4096 models though they might be slightly weaker at generating short contents)
Final checkpoint: RWKV-4-Pile-7B-20221115-8047.pth : Trained on the Pile for 332B tokens.
* Pile loss 1.8415
* LAMBADA ppl 4.38, acc 67.18%
* PIQA acc 76.06%
* SC2016 acc 73.44%
* Hellaswag acc_norm 65.51%
|