Closed kadir-gunel closed 3 years ago
Some info about CUDA :
julia> CUDA.versioninfo()
CUDA toolkit 11.1.1, artifact installation
CUDA driver 11.1.0
NVIDIA driver 455.45.1
Libraries:
- CUBLAS: 11.3.0
- CURAND: 10.2.2
- CUFFT: 10.3.0
- CUSOLVER: 11.0.1
- CUSPARSE: 11.3.0
- CUPTI: 14.0.0
- NVML: 11.0.0+455.45.1
- CUDNN: 8.0.4 (for CUDA 11.1.0)
- CUTENSOR: 1.2.1 (for CUDA 11.1.0)
Toolchain:
- Julia: 1.5.2
- LLVM: 9.0.1
- PTX ISA support: 3.2, 4.0, 4.1, 4.2, 4.3, 5.0, 6.0, 6.1, 6.3, 6.4
- Device support: sm_35, sm_37, sm_50, sm_52, sm_53, sm_60, sm_61, sm_62, sm_70, sm_72, sm_75
2 devices
That's a Zygote error, you're probably differentiating code that isn't supported (like the ccall
to the CUDA driver here). I recommend you open a Discourse post, and if it turns out to be a bug, an issue on e.g. Flux or Zygote.jl.
Hello,
I have a flux model and when I take the gradients of the model by using Zygote I get
ERROR: this intrinsic must be compiled to be called
Below is tested only on GPU not on CPU.For the mwe please use the following code :
When taking the gradients I get :
And with the suggestion of @simeonschaub
Zygote.@adjoint CUDA.zeros(x...) = CUDA.zeros(x...), _ -> map(_ -> nothing, x)
I managed to get the gradients. And since @simeoneschaub told me to open an issue for CUDA.jl, here it is.