A 20M RWKV v6 can do nonogram
Pure language model can do nonogram
Nonograms / 数织 are picture logic puzzles in which cells in a grid must be colored or left blank according to numbers at the edges of the grid to reveal a hidden picture. In this puzzle, the numbers are a form of discrete tomography that measures how many unbroken lines of filled-in squares there are in any given row or column.
数织是一种逻辑游戏，以猜谜的方式绘画黑白位图。在一个网格中，每一行和列都有一组数，玩家需根据它们来填满或留空格子，最后就可以由此得出一幅图画。例如，“4 8 3”的意思就是指该行或列上有三条独立的线，分别占了4、8和3格，而每条线最少要由一个空格分开。传统上，玩家是以黑色填满格子，和以“×”号标记一定不需要填充的格子。数织是一个NP完全的问题。
You can play nonogram at here.
Try RWKV-nonogram at Colab
Our 20M RWKV v6 can do 5x5 nonograms (for v1 version) like:
Put the above process into a more intuitive way:
this test sample are not in the train dataset
rwkv-11.pth is the weighs file.((N_LAYER=12,N_EMBD=384))
rwkv_vocab_nonogram.txt is the vocab file.
pip install rwkv
run the code in rwkv-nonogram.ipynb
generate_train_data.py is the code to generate the train data.
And I use https://github.com/BlinkDL/RWKV-LM/blob/main/RWKV-v5/demo-training-run.sh to train.