OpenNMT Forum

Speech to text training model on gpu error

I am facing this error (The NVIDIA driver on your system is too old (found version 8000).
Please update your GPU driver by downloading and installing a new
version from the URL: http://www.nvidia.com/Download/index.aspx
Alternatively, go to: https://pytorch.org to install
a PyTorch version that has been compiled with your version
of the CUDA driver.). When I tried to run the command on gpu machine.
On CPU the following command doing good --python train.py -model_type audio -enc_rnn_size 1024 -dec_rnn_size 1024 -audio_enc_pooling 1,1,1,2,2,2 -dropout 0.2 -enc_layers 6 -dec_layers 2 -rnn_type LSTM -data data/speech/demofiles_vctk/demo -save_model models/exp_mix_vctk/demo-model_vctkmix -global_attention general -batch_size 5 -save_checkpoint 5000 -optim adam -max_grad_norm 100 -learning_rate 0.00025 -learning_rate_decay 0.5 -decay_method rsqrt -train_steps 80000 -encoder_type brnn -decoder_type rnn -normalization tokens -bridge -window_size 0.015 -image_channel_size 1 -gpu_ranks 0 -world_size 1.
But on GPU machine it gave error.

The GPU model is NVIDIA Tesla M60.

Hi,
This is a driver / pytorch issue, not an OpenNMT one. You need to follow what it says: have compatible driver / cuda / pytorch versions.
You can check which cuda version your pytorch was compiled with by doing this in a python shell:

import torch
print(torch.version.cuda)

Then, you need to get the proper drivers / cuda version from nvidia.

Hi,
Can you suggest me the compatible versions of cuda,driver and pytorch for speech to text model training?

This is not related to speech per se. It’s related to torch.

I will try this.