Bug: Illegal instruction (core dumped) ,

[Siyamfahad]

I am trying to install stt for training my own model but I am facing this issue called Illegal instruction (core dumped),

I tried using docker and manually both ways but got the same result, I also tried with TensorFlow GPU and CPU version same result :( can anyone help me with this?

terminal log

**(coqui-stt-train-venv) siyamfahad@siyamfahad-VirtualBox:~/STT$ ./bin/run-ldc93s1.sh

[ ! -f train.py ] [ ! -f data/smoke_test/ldc93s1.csv ] checkpoint_dir=/home/siyamfahad/.local/share/stt/ldc93s1 export CUDA_VISIBLE_DEVICES=0 python -m coqui_stt_training.train --alphabet_config_path data/alphabet.txt --show_progressbar false --train_files data/smoke_test/ldc93s1.csv --test_files data/smoke_test/ldc93s1.csv --train_batch_size 1 --test_batch_size 1 --n_hidden 100 --epochs 200 --checkpoint_dir /home/siyamfahad/.local/share/stt/ldc93s1 Illegal instruction (core dumped) (coqui-stt-train-venv) siyamfahad@siyamfahad-VirtualBox:/STT$ python -m coqui_stt_training.train --help Illegal instruction (core dumped) (coqui-stt-train-venv) siyamfahad@siyamfahad-VirtualBox:/STT$** My CPU is i7-4770 my GPU is gtx-1050ti

[Siyamfahad]

I found the issue was for my GPU, after removing it its working well, i don't know why my gpu makes this issue ,please reopen the issue if anyone can help

[fquirin]

Can we reopen this please.

Some of my SEPIA STT-Server users experience the same issue when running Coqui on a VM or as Docker container in a Synology DS920+ (CPU Intel Celeron J4125).

For the Proxmox VM the solution was "setting the hypervisor CPU type to host instead of default". I think this Stackoverflow issue might be related as well ("Intel Celeron J4125 [...] does not support AVX and AVX2 instructions").

It seems that some dependency of Coqui STT has requirements for a certain CPU instructions set. Although I don't understand how this would relate to the OP's GPU issue 🤔