This model is finetuned based on RWKV4 world 7B english model, using wizard datset to fit in 32k ctx.
should do some complex instruction and can feed in 32k token, but as we know wizard dataset only have max 4K length per sample ,
i'll do more test on long prompt and generation
use https://github.com/josStorer/RWKV-Runner to run this model.