Skip to content

Instantly share code, notes, and snippets.

@acanakoglu
Created May 22, 2020 17:48
Show Gist options
  • Save acanakoglu/22e84d9d15aa97e0cfbeeeba79d85851 to your computer and use it in GitHub Desktop.
Save acanakoglu/22e84d9d15aa97e0cfbeeeba79d85851 to your computer and use it in GitHub Desktop.
run rasa with GPU with correct cudatoolkit version
# create env
conda create -n agent python=3.7
# goto directory
cd ~/Prj/GecoBot
# installation of all the requirements, I have rasa==1.9.6,rasa-sdk==1.9.0 and their dependencies.
pip install -r requirements.txt
# installed the correct cudo tool kit, which I found from the cuda conf.
# I ran the commend `nvidia-smi` and it is written upper right, in my case it is 10.1
conda install -c anaconda cudatoolkit=10.1
conda install -c anaconda cudnn
# I downloaded all english versions:
python -m spacy download en_core_web_sm
python -m spacy download en_core_web_md
python -m spacy download en_core_web_lg
# I set to small:
python -m spacy link en_core_web_sm en
# to set another one:
# python -m spacy link en_core_web_lg en --force
# and run my code to train:
python trainer.py
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment