https://gricad-gitlab.univ-grenoble-alpes.fr/coavouxm/flaubertagger.git
Revision c939ca9fac094ac3c379256ef3d3d4d14a5a4bf1 authored by m on 23 February 2024, 16:44:50 UTC, committed by m on 23 February 2024, 16:44:50 UTC
1 parent 2a04175
Tip revision: c939ca9fac094ac3c379256ef3d3d4d14a5a4bf1 authored by m on 23 February 2024, 16:44:50 UTC
up
up
Tip revision: c939ca9
README.md
# flaubertagger
FlauberTagger is a morphological tagger for French based on Flaubert language model.
## Install the tagger:
* Install anaconda or miniconda (follow instructions at [https://docs.anaconda.com/free/miniconda/index.html](https://docs.anaconda.com/free/miniconda/index.html).
* Run the following commands:
```
conda create -n tagger python=3.8
conda activate tagger
pip install torch transformers==4.6.0 six
```
## Run the tagger
```
cd flaubertagger/src
python tagger.py eval gsd_base_0.00001_0.5_8_aug <input> <output> [--cuda 0]
```
where:
- `gsd_base_0.00001_0.5_8_aug` is the path to the unzipped pretrained model
- `input` = path text file with 1 sentence per line. Sentences should be tokenized (e.g. with Stanza)
- `output` = path to output file that will be created
- `[--cuda]` use this option if you have a GPU available

Computing file changes ...