harrisonvanderbyl/rwkv-cpp-accelerated
A torchless, c++ rwkv implementation using 8bit quantization, written in cuda/hip/vulkan for maximum compatibility and minimum dependencies
C++MIT
Issues
- 5
NumCpp NdArray not initialized
#34 opened by montecarlo26 - 2
Questions about int8 quantification.
#39 opened by zzczzc20 - 2
converter failure
#38 opened by malv-c - 0
ACO ERROR: Unsupported opcode
#36 opened by Jipok - 1
- 3
something went wrong while convert model
#33 opened by innnk - 1
Radeon Open Compute support
#31 opened by erkinalp - 0
Training, in -cpp-cuda, on one machine?
#30 opened by SCRIER-org - 1
compiling for windows
#29 opened by RichardErkhov - 1
- 2
fail to load on windows
#27 opened by flizzywine - 2
输出都是attout浮点数是什么情况?
#26 opened by wangshankun - 2
Error converting to bin
#21 opened by sammyf - 1
Faster model loading
#13 opened by nenkoru - 6
Add Dockerfile
#2 opened by nenkoru - 3
Memory leak in rwkv.h
#17 opened by maksmaisak - 1
Cannot build
#11 opened by dillfrescott - 6
Endless <|endoftext|> bug
#4 opened by Murugurugan - 1
Not working on 7B & 14B models | Torch Binding
#10 opened by nenkoru - 1
Add basic example of a chat using Python
#6 opened by nenkoru - 0
Multi-GPU support
#3 opened by nenkoru