python 3
.pipenv
installedhow to install
.- to run with real MlaaS providers, you should provide valid credentials to each used provider:
- create a
credentials.py
file following the structure ofcredentials.example.py
. - fill the credentials to Amazon, Microsoft and Google providers.
- create a
- activate the environmentment:
pipenv shell
. - install dependencies:
pipenv sync
. The process may take several minutes. - with the environmentment activated, run the notebook:
- jupyter-lab with command
jupyter-lab
. - open the file
notebook_experiment.ipynb
. - follow notebook's instructions.
- jupyter-lab with command
- Open: https://colab.research.google.com/github/evaluating-effectiveness-cloud-nlp/replication_package/blob/master/notebook_rq1.ipynb
- Follow notebook's instructions.
- you don't need to download the
glove.twitter
word embedding model if you remove the noise algorithmWordEmbeddings
fromnoise_list
variable in both experiments. - by default, the MLaaS providers are mocked with random predictions, so you can easily try the algorithm, but in a real use you should register an account in each one and fill the
creditials.py
file. - in both experiments, in case of error (eg. network error) is possible to continue from a previously running experiment by filling the
Continue from
text area in the notebook with a/outputs
folder. - is possible to manipulate the
progress.json
file in order to re-run some task of the experiment.