-
Notifications
You must be signed in to change notification settings - Fork 3.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Feature 'cvt with .bf16' requires .target sm_80 or higher Error #3947 #18070
Comments
@lk1983823 What was the solution? I'm getting the same error, when using
I'm running this on a T4. According to https://arnon.dk/matching-sm-architectures-arch-and-gencode-for-various-nvidia-cards/ SM80 is only supported on A100. So maybe bfloat16 doesn't work on T4s? (I'm confused because the same model works in eager mode on the T4, but adding |
I also met the same error when I tried to run my model using Triton package. I am running it on V100. Any solutions so far? Thank you for the help. |
Bug description
I am running a code using lightning and deepspeed, the optimizer is set as:
optimizer = deepspeed.ops.adam.DeepSpeedCPUAdam(model.parameters(), lr=1e-3)
When the code runs the first time, it do as follows:
where in the step 1/3, it sets -gencode=arch=compute_75,code=sm_75, which is not satisfied in the following steps and the error shows:
However, in another computer with the same hardware. There is no such problem and the step 1/3 didn't appear.
[1/3] /usr/local/cuda-11.7/bin/nvcc -DTORCH_EXTENSION_NAME=cpu_adam -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE=\"_gcc\" -DPYBIND11_STDLIB=\"_libstdcpp\" -DPYBIND11_BUILD_ABI=\"_cxxabi1011\" -I/home/lk/anaconda3/envs/weather/lib/python3.10/site-packages/deepspeed/ops/csrc/includes -I/usr/local/cuda-11.7/include -isystem /home/lk/anaconda3/envs/weather/lib/python3.10/site-packages/torch/include -isystem /home/lk/anaconda3/envs/weather/lib/python3.10/site-packages/torch/include/torch/csrc/api/include -isystem /home/lk/anaconda3/envs/weather/lib/python3.10/site-packages/torch/include/TH -isystem /home/lk/anaconda3/envs/weather/lib/python3.10/site-packages/torch/include/THC -isystem /usr/local/cuda-11.7/include -isystem /home/lk/anaconda3/envs/weather/include/python3.10 -D_GLIBCXX_USE_CXX11_ABI=0 -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr -gencode=arch=compute_75,code=compute_75 -gencode=arch=compute_75,code=sm_75 --compiler-options '-fPIC' -O3 --use_fast_math -std=c++14 -U__CUDA_NO_HALF_OPERATORS__ -U__CUDA_NO_HALF_CONVERSIONS__ -U__CUDA_NO_HALF2_OPERATORS__ -gencode=arch=compute_75,code=sm_75 -gencode=arch=compute_75,code=compute_75 -c /home/lk/anaconda3/envs/weather/lib/python3.10/site-packages/deepspeed/ops/csrc/common/custom_cuda_kernel.cu -o custom_cuda_kernel.cuda.o
doesn't appear.
I have reinstalled the cuda-toolkit, but it didn't work.
Anyone know how to solve it? Thanks.
Ubuntu 18.04
Python 3.10
GPU RTX2080ti
CUDA 11.7
NVCC 2.14.3
lightning 2.0.2
deepspeed 0.9.2
What version are you seeing the problem on?
v2.0
How to reproduce the bug
No response
Error messages and logs
Environment
Current environment
More info
No response
The text was updated successfully, but these errors were encountered: