BlinkDL
|
71538e44a9
|
refactoring
|
4 years ago |
BlinkDL
|
5817d265c3
|
no message
|
4 years ago |
BlinkDL
|
0b6aec3da6
|
Default parameters for 8G VRAM
|
4 years ago |
PENG Bo
|
6aefe59c3d
|
Update README.md
|
4 years ago |
PENG Bo
|
da6f35f276
|
Update README.md
|
4 years ago |
PENG Bo
|
72a6f28add
|
Update README.md
|
4 years ago |
PENG Bo
|
d8234047e6
|
Update README.md
|
4 years ago |
PENG Bo
|
5ac82f37be
|
Update README.md
|
4 years ago |
PENG Bo
|
a86ac324de
|
Update README.md
|
4 years ago |
PENG Bo
|
9ca62d3a1e
|
Update README.md
|
4 years ago |
PENG Bo
|
4b1df60e94
|
Update README.md
|
4 years ago |
PENG Bo
|
780bed4e19
|
Update README.md
|
4 years ago |
BlinkDL
|
5f21ddf20d
|
RWKV v2 RNN is here. Probably the strongest LM as of now.
|
4 years ago |
PENG Bo
|
1f189a4034
|
Update README.md
|
4 years ago |
PENG Bo
|
5ac265691f
|
RWKV-v2-RNN introduction
|
4 years ago |
PENG Bo
|
b4fd1a7209
|
Update model.py
|
4 years ago |
PENG Bo
|
c8a751ed8b
|
Update README.md
|
4 years ago |
BlinkDL
|
0a0eae447d
|
+headQK (compatible with 2022-02-15 AI-Writer)
|
4 years ago |
BlinkDL
|
b48aa1d430
|
no message
|
4 years ago |
BlinkDL
|
a19be54bf5
|
no message
|
4 years ago |
BlinkDL
|
fcd01f8851
|
no message
|
4 years ago |
BlinkDL
|
76e241b71e
|
saves vocab.json, and the model every X epoch
|
4 years ago |
PENG Bo
|
689a6a924d
|
Update train.py
|
4 years ago |
PENG Bo
|
34fa2ec81b
|
Update README.md
|
4 years ago |
PENG Bo
|
58bdb908f9
|
Update README.md
|
4 years ago |
PENG Bo
|
3d8d0373b4
|
Update README.md
|
4 years ago |
BlinkDL
|
710d3e34b7
|
better init for RWKV
|
4 years ago |
BlinkDL
|
619ed00e4b
|
misc improvement
|
4 years ago |
PENG Bo
|
a36fc09fea
|
Update README.md
|
4 years ago |
PENG Bo
|
a91084efa9
|
Update README.md
|
4 years ago |
BlinkDL
|
3329161ed7
|
rapid convergence using ZERO initialization
|
4 years ago |
BlinkDL
|
7f391c5758
|
+ RWKV tiny-attn and now it's great for ctx 1024 or 2048
|
4 years ago |
PENG Bo
|
a9f39c112c
|
Update README.md
|
4 years ago |
PENG Bo
|
8fd4601dea
|
Update README.md
|
4 years ago |
BlinkDL
|
9b903db103
|
Merge branch 'main' of https://github.com/BlinkDL/RWKV-LM into main
|
4 years ago |
BlinkDL
|
8aec414db2
|
no message
|
4 years ago |
PENG Bo
|
9e959d0b8a
|
Update README.md
|
4 years ago |
BlinkDL
|
4ffd8f1b76
|
+ new comparison
|
4 years ago |
PENG Bo
|
04852faf04
|
Update README.md
|
4 years ago |
BlinkDL
|
ad627311f4
|
clean init code
|
4 years ago |
BlinkDL
|
c675b47705
|
misc improvements
|
4 years ago |
BlinkDL
|
ef29f4b9e8
|
fixed nan loss
|
4 years ago |
BlinkDL
|
4fd8716976
|
improve RWKV time_w initialization
|
4 years ago |
PENG Bo
|
1ea53a2f03
|
Update README.md
|
4 years ago |
BlinkDL
|
a31a3b2e92
|
+ MHA_shift
|
4 years ago |
PENG Bo
|
4096fff9ee
|
Update README.md
|
4 years ago |
PENG Bo
|
12ba06216d
|
Update README.md
|
4 years ago |
PENG Bo
|
639de69256
|
Create CITATION.cff
|
4 years ago |
PENG Bo
|
994170685b
|
Update README.md
|
4 years ago |
BlinkDL
|
3b9005ea11
|
RWKV: now faster and less params
|
4 years ago |