Natural Language Processing with Transformers by Lewis Tunstall, Leandro von Werra, Thomas Wolf.
Sunday, January 15, 2023
docker run --gpus all -it -v $(realpath ~/):/tf/All -v /home/rob/Data2:/home/rob/Data2 --env HF_DATASETS_CACHE=/home/rob/Data2/huggingface/datasets --env TRANSFORMERS_CACHE=/home/rob/Data2/huggingface/transformers -p 8888:8888 -p 6006:6006 hfptl:20221222
01_Introduction.ipynb Run Date: Sunday, January 15, 2023 Run Time: 00:00:29
02_classification.ipynb Run Date: Sunday, January 15, 2023 Run Time: 00:04:38
03_transformer-anatomy.ipynb Run Date: Sunday, January 15, 2023 Run Time: 00:00:10
04_multilingual-ner.ipynb Having out of memory issues with this notebook. Numerous installs to this image ... So saved the file, exited docker, want to restart the container ... so ...
(base) rob@KAUWITB:~$ docker container ps -a CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES c135df6d6af8 hfptl:20221222 "/opt/nvidia/nvidia_…" About an hour ago Exited (0) 2 minutes ago determined_wing
docker container start determined_wing docker container logs determined_wing
Sigh ... nope ... continues to run out of memory, even with a reduced batch size.
05_text-generation.ipynb Nope! Runs out of GPU ram ...
06_summarization.ipynb Nope! Runs out of GPU ram ...
Trying 05_textgeneration.ipynb again in the newest huggingface transformers docker image which is running inside the container hfpt_Oct28.
Hmm interesting ... it now runs! With the 2070 Super! Nice!
Gonna also try the 06_summarization.ipynb again ...
I had to run 'pip install py7zr'
1:49PM Wow ... so I renamed Data2/HuggingFace to Data2/HuggingFace_ then reran this notebook ... AND IT STILL RUNS! So where I thought the models were being saved to is wrong! ... Are they being saved into the container??
Installed the 4090, and now notebooks 05 and 06 run just fine. Nice!