https://github.com/warner-benjamin/nlpbookstudygroup
NLP with Transformers Study Group Materials & Resources
https://github.com/warner-benjamin/nlpbookstudygroup
Last synced: over 1 year ago
JSON representation
NLP with Transformers Study Group Materials & Resources
- Host: GitHub
- URL: https://github.com/warner-benjamin/nlpbookstudygroup
- Owner: warner-benjamin
- License: apache-2.0
- Created: 2023-04-25T16:09:56.000Z (about 3 years ago)
- Default Branch: main
- Last Pushed: 2023-06-26T19:14:48.000Z (almost 3 years ago)
- Last Synced: 2025-03-17T16:49:27.485Z (over 1 year ago)
- Language: Jupyter Notebook
- Homepage:
- Size: 2.86 MB
- Stars: 10
- Watchers: 1
- Forks: 2
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# *NLP With Transformers* Study Group Resources
[NLP with Transformers](https://transformersbook.com) study group materials and resources
## Setup with PyTorch 2.0
The following Conda + pip commands will install all the required packages for the
book's notebooks. Or run `setup-lite.sh` from bash. You may need to change
`source activate nlpbook` if conda isn't setup for bash.
```bash
conda create -n nlpbook python=3.10 "pytorch>=2.0.1" torchvision torchaudio torchtext \
pytorch-cuda=11.8 "transformers>=4.28.1" "datasets>=2.11.0" "tokenizers>=0.13.3" \
sentencepiece optuna scikit-learn onnxruntime matplotlib ipywidgets jupyterlab timm \
umap-learn seqeval nltk sacrebleu py7zr nlpaug scikit-multilearn psutil accelerate \
-c pytorch -c nvidia/label/cuda-11.8.0 -c huggingface -c conda-forge
conda activate nlpbook
pip install rouge-score bertviz
```
These commands will additionally install Cuda, fastai, blurr, fastxtend, Composer,
& Weights and Biases. Or run `setup-full.sh` from bash. You may need to change
`source activate nlpbook` if conda isn't setup for bash.
```bash
conda create -n nlpbook python=3.10 "pytorch>=2.0.1" torchvision torchaudio torchtext \
pytorch-cuda=11.8 cuda "transformers>=4.28.1" "datasets>=2.11.0" "tokenizers>=0.13.3" \
fastai sentencepiece optuna scikit-learn onnxruntime matplotlib ipywidgets jupyterlab \
umap-learn seqeval nltk sacrebleu py7zr nlpaug scikit-multilearn psutil accelerate \
wandb openpyxl xlrd torchmetrics requests coolname tabulate py-cpuinfo timm \
importlib-metadata \
-c pytorch -c nvidia/label/cuda-11.8.0 -c huggingface -c fastai -c conda-forge
conda activate nlpbook
pip install rouge-score bertviz fastxtend torch_optimizer composer
pip install ohmeow-blurr --no-deps
```
`--no-deps` is so blurr and Composer won't uninstall PyTorch 2.0 for 1.13.