Error after checking for tensorflow 2.10 installation

[desmodus1984]

Hi,
I tried running Tiberius and it failed several times. I tried installing tensorflow 2.10 as stated,
https://github.com/Gaius-Augustus/Tiberius/blob/main/docs/install_tensorflow.md

and after doing the installation check, I got the following error message in conda:

python3 -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
2025-07-02 11:35:49.013384: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 AVX512F AVX512_VNNI FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2025-07-02 11:35:49.162298: I tensorflow/core/util/util.cc:169] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0.
2025-07-02 11:35:49.192731: E tensorflow/stream_executor/cuda/cuda_blas.cc:2981] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2025-07-02 11:35:50.762821: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer.so.7'; dlerror: libnvinfer.so.7: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: :/data/common/juanpablo.aguilar/apps/miniconda3/lib/:/data/common/juanpablo.aguilar/apps/miniconda3/envs/tf2_10/lib/
2025-07-02 11:35:50.763065: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer_plugin.so.7'; dlerror: libnvinfer_plugin.so.7: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: :/data/common/juanpablo.aguilar/apps/miniconda3/lib/:/data/common/juanpablo.aguilar/apps/miniconda3/envs/tf2_10/lib/
2025-07-02 11:35:50.763100: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Cannot dlopen some TensorRT libraries. If you would like to use Nvidia GPU with TensorRT, please make sure the missing libraries mentioned above are installed properly.
2025-07-02 11:35:51.768417: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcuda.so.1'; dlerror: libcuda.so.1: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: :/data/common/juanpablo.aguilar/apps/miniconda3/lib/:/data/common/juanpablo.aguilar/apps/miniconda3/envs/tf2_10/lib/
2025-07-02 11:35:51.768453: W tensorflow/stream_executor/cuda/cuda_driver.cc:263] failed call to cuInit: UNKNOWN ERROR (303)
2025-07-02 11:35:51.768489: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:156] kernel driver does not appear to be running on this host (pirate.cas.hamptonu.edu): /proc/driver/nvidia/version does not exist
[]

I did the installation for a 2nd time since I thought I did it wrong or there was something different from the other instruction - which are identical, (both links seem to be for conda and under a different python setup), and I got the error.

Did I miss something?

Hope this can be fixed.

[LarsGab]

Hi,

if I understand correctly, you haven't installed Tiberius itself yet and only installed Tensorflow 2.10 using the instructions?

Unfortunately, . I tested it on our machine and couldn’t reproduce the error, can you share your environment details (e.g., GPU model)?

If you have Singularity available, running the Singularity image is the easiest fix. Otherwise, upgrading to TensorFlow 2.17 and using a shorter sequence length (259,992) with tiberius.py should also work without any notable loss in accuracy. A newer Tensorflow version should be installed automatically, if you install Tiberius with pip install . .

Best,
Lars

[desmodus1984]

Hi,

I have Singularity but when I tried getting the container it downloaded a lot of GBS of data until the point that my account was saturated of files, and hence it didn't finish. Do you know much data will Singularity download to make sure I am downloading the right thing? One file/hash was like 4.5 GB and several 1.5 GB.

In regards of the error, could you tell if it is something relevant or not?
Furthermore, I don't have a GPU, only CPU, which migtht be the issue the error is stating.

My next question is, if I want to have Tensorflow 2.10 because I have a contiguous assembly and there is no logical and coherent way to break my contigs safely without inadvertently damaging a gene/exon, how do I install Tiberius if not with pip install . ?

[LarsGab]

Hi,

the Singularity image has 11Gb.

If you do not have a GPU, the installation of tf 2.10 should have worked and you can ignore the warnings. It just did not find a GPU.

You can install Tiberius now with pip install . .

Best,
Lars

[desmodus1984]

Hi,
Are you sure it will work?
You mentioned earlier that after installing tensorflow 2.10 when I run pip

A newer Tensorflow version should be installed automatically, if you install Tiberius with pip install . .
And I want to keep Tensorflow 2.10 because I don't want to be forced to break my contigs anywhere senselessly.

I tried running Tiberius in the new environment as stated in the instructions of tensorflow 2.10, with pip install . , the code I used was:

#activate conda environment
conda activate tf2_10

# Change directory
cd /data/common/j-lar/apps/Tiberius

# Save starting time
start_time=$(date +%s)
echo "Tiberius-tf2.10: Job started: $(date)"

#Test Tiberius annotation

python tiberius.py --genome /data/common/j-lar/data/fx/HydLep.fx.fasta --out HydLep.tiberius.gtf \
         --batch_size 16  --codingseq CDS_HydLep.fasta --protseq PROT-HydLep.fasta

and I got this error:

Traceback (most recent call last):
File "/data/common/juanpablo.aguilar/apps/Tiberius/tiberius.py", line 2, in
import yaml
ModuleNotFoundError: No module named 'yaml'

So, I even installed yaml on the conda environment and it didn't work.

Any reason why now yaml is not detected?

I then, tried download the singularity using the code on the main page, and I got this error:

INFO: Starting build...
FATAL: While performing build: conveyor failed to get: pinging container registry registry-1.docker.io: Get "https://registry-1.docker.io/v2/": x509: certificate signed by unknown authority

[LarsGab]

A large update has been released since this comment was made. Please try installing the latest version of Tiberius or running it with Singularity.

You can also run Tiberius with a newer version of TensorFlow, but you will need to reduce the sequence length to '--seq_len 259.992'.