ht37
ht37
I'm busy in CUDA (keras/tensorflow) setup for aged *NVIDIA Tesla K80 Azure Datacenter GPU* (2014) * [NC-series](https://learn.microsoft.com/en-us/azure/virtual-machines/nc-series) These GPU are rather old and near to be [dismissed by Microsoft](https://learn.microsoft.com/en-us/azure/virtual-machines/nc-series-retirement), but...
Maybe the question is easy to formulate, but the answer is not ... **Q:** _"Which is the latest tensorflow version that can be used (by CRAN-keras) with obsolete (470-driver line)...
It worked! But in a _"manual (hammered)"_ non-containerized setup ... --- One point is compatibility between (host) kernel driver from 470 line and user-mode `cuda-driver` (from `cuda-compat-470`, prepended in LD_LIBRARY_PATH...
_**It worked!**_ ### Sources #### Image Patch * [install_ubs-cuda-11-470.sh](https://gitlab.com/ub-dems-public/ds-labs/dve-sample-r/-/blob/main/docker/r-images/scripts/cuda/install_ubs-cuda-11-470.sh) #### Test Scripts * [dummy-keras.py](https://gitlab.com/ub-dems-public/ds-labs/dve-sample-r/-/blob/main/exec/dummy-keras.py) * [dummy-keras.R](https://gitlab.com/ub-dems-public/ds-labs/dve-sample-r/-/blob/main/exec/dummy-keras.R) Test based on Tensorflow guide: ["Use a GPU"](https://www.tensorflow.org/guide/gpu) ### Configuration #### Rocker-Project | Base...
@cboettig cuDNN compute capabilities requirements match with your GPU is critical. NVIDIA declares support for 3.5+ level (and 470 driver) for all CUDA-11 releases, but later components seem to break...
I found an issue in BLAS configuration. In base `nvidia-cuda` image, *OpenBLAS* libraries, while installed, are *disabled* in `/etc/alternatives` configuration. In the image, with *nvblas/cublas* enabled, system BLAS config is...
This test runs without any errors under this configuration: * [_"Segfault using scipy filter"_](https://github.com/rstudio/reticulate/issues/1190) ``` > sessionInfo() R version 4.3.0 (2023-04-21) Platform: x86_64-pc-linux-gnu (64-bit) Running under: Ubuntu 22.04.1 LTS Matrix...
With "rootless" Podman, it was possible to to run everything as root. This was required to share data between container and host, because, in rootless mode, host uid:gid are mapped...
In the _"workaround 2"_ above, there is a problem related to the user login (as root) made by the rstudio-server. This change of user context *"discards"* additional group membership of...
same hrere for a python 3.9, poetry project (tf 2.5, airsim 1.8.1) It seems to work with this workaround ```bash poetry lock poetry install # airsim fail with missing numpy...