You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm a beginner so sorry if this is the wrong place.
Branch/Tag/Commit
main?
Docker Image Version
nvcr.io/nvidia/pytorch:22.03-py3
GPU name
NVIDIA GeForce GTX TITAN X
CUDA Driver
Cuda compilation tools, release 11.6, V11.6.112 ?
Reproduced Steps
1. docker run -ti --gpus all nvcr.io/nvidia/pytorch:22.03-py3 bash
2. git clone https://github.com/NVIDIA/FasterTransformer.git
3. cd FasterTransformer mkdir build &&cd build
4. cmake -DSM=80 -DCMAKE_BUILD_TYPE=Release .. && make -j12
5. ./bin/bert_example 32 12 32 12 64 0 0
root@1423b5c7567a:/workspace/FasterTransformer/build# ./bin/bert_example 32 12 32 12 64 0 0
[INFO] Device: NVIDIA GeForce GTX TITAN X
Before loading model: free: 11.54 GB, total: 11.93 GB, used: 0.39 GB
[WARNING] gemm_config.in is not found; using default GEMM algo
terminate called after throwing an instance of 'std::runtime_error'what(): [FT][ERROR] CUDA runtime error: operation not supported /workspace/FasterTransformer/src/fastertransformer/utils/allocator.h:160
Aborted (core dumped)
root@1423b5c7567a:/workspace/FasterTransformer/build# /usr/local/cuda/bin/nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2022 NVIDIA Corporation
Built on Thu_Feb_10_18:23:41_PST_2022
Cuda compilation tools, release 11.6, V11.6.112
Build cuda_11.6.r11.6/compiler.30978841_0
zpc is my PC, usually running Arch with multi-seat, but I didn't manage to install the nvidia driver on Arch (likely due to latest linux kernel, also tried linux-lts) so I booted into Ubuntu ... =)
I'm having same/similar issue, tried lots of things. According to NVIDIA CUDA/Triton docs I should be able to run the versions I've tried but I always encounter the same error above
I don't have access to update the NVIDIA drivers, so I'm using FasterTransformer v5.0
Using v100d-32gb's
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 510.85.02 Driver Version: 510.85.02 CUDA Version: 11.6 |
|-------------------------------+----------------------+----------------------+
what(): [FT][ERROR] CUDA runtime error: operation not supported /workspace/build/fastertransformer_backend/build/_deps/repo-ft-src/src/fastertransformer/utils/allocator.h:181
The same issue.
root@3bf52cde87a7:/workspace/FasterTransformer/build# ./bin/swin_example 2 0 0 8 256 2
[FT][INFO] Device GRID T4-8C
terminate called after throwing an instance of 'std::runtime_error'
what(): [FT][ERROR] CUDA runtime error: operation not supported /workspace/FasterTransformer/src/fastertransformer/utils/allocator.h:160
I'm a beginner so sorry if this is the wrong place.
Branch/Tag/Commit
main?
Docker Image Version
nvcr.io/nvidia/pytorch:22.03-py3
GPU name
NVIDIA GeForce GTX TITAN X
CUDA Driver
Cuda compilation tools, release 11.6, V11.6.112 ?
Reproduced Steps
zpc is my PC, usually running Arch with multi-seat, but I didn't manage to install the nvidia driver on Arch (likely due to latest linux kernel, also tried linux-lts) so I booted into Ubuntu ... =)
I suspect it has to do with my old CUDA Version: 11.4 ... I will try upgrading ...
After upgrading to version 530 of nvidia drivers:
Still the same error after upgrading nvidia driver. Or did I miss something ?
The text was updated successfully, but these errors were encountered: