73 lines
3.2 KiB
Markdown
73 lines
3.2 KiB
Markdown
# Train and deploy a SOTA model
|
|
|
|
Let's see GreatAI in action by going over the life-cycle of a simple service.
|
|
|
|
## Objectives
|
|
|
|
1. You will see how the [great_ai.utilities][] can integrate into your Data Science workflow.
|
|
2. You will use [great_ai.large_file][] to version and store your trained model.
|
|
3. You will use [GreatAI][great_ai.GreatAI] to prepare your model for a robust and responsible deployment.
|
|
|
|
## Overview
|
|
|
|
You are going to train a field of study (domain) classifier for scientific sentences. The exact task was proposed by the [SciBERT paper](https://arxiv.org/abs/1903.10676) in which SciBERT [achieved an F1-score of 0.6571](https://paperswithcode.com/sota/sentence-classification-on-paper-field). We are going to outperform it using a trivial text classification model: a [Linear SVM](https://scikit-learn.org/stable/modules/generated/sklearn.svm.LinearSVC.html).
|
|
|
|
We use the same synthetic dataset derived from the [Microsoft Academic Graph](https://www.microsoft.com/en-us/research/project/microsoft-academic-graph/). The dataset is [available here](https://github.com/allenai/scibert/tree/master/data/text_classification/mag).
|
|
|
|
|
|
!!! success
|
|
You are ready to start the tutorial. Feel free to come back to the [summary](#summary) section once you're finished.
|
|
|
|
<div style="display: flex; justify-content: space-evenly;" markdown>
|
|
[:fontawesome-solid-chart-simple: Train it](train.ipynb){ .md-button .md-button--primary }
|
|
|
|
[:material-cloud-tags: Deploy it](deploy.ipynb){ .md-button .md-button--primary }
|
|
</div>
|
|
|
|
|
|
## Summary
|
|
|
|
### The [training notebook](train.ipynb)
|
|
|
|
We load and preprocess the dataset while relying on [great_ai.utilities.clean][] for the heavy-lifting. Additionally, the preprocessing is parallelised using [great_ai.utilities.simple_parallel_map][]
|
|
|
|
After training and evaluating a model, it is exported using [great_ai.save_model][].
|
|
|
|
??? tip "Remote storage"
|
|
To store your model remotely, you need to set your credentials before calling `save_model`.
|
|
|
|
For example, to use [AWS S3](https://aws.amazon.com/s3/):
|
|
```python
|
|
from great_ai.large_file import LargeFileS3
|
|
|
|
LargeFileS3.configure(
|
|
aws_region_name='eu-west-2',
|
|
aws_access_key_id='MY_AWS_ACCESS_KEY',
|
|
aws_secret_access_key='MY_AWS_SECRET_KEY',
|
|
large_files_bucket_name='my_bucket_for_models'
|
|
)
|
|
|
|
from great_ai import save_model
|
|
|
|
save_model(model, key='my-domain-predictor')
|
|
```
|
|
|
|
### The [deployment notebook](deploy.ipynb)
|
|
|
|
We create an inference function that can be hardened by wrapping it in a [GreatAI][great_ai.GreatAI] instance.
|
|
|
|
```python
|
|
from great_ai import GreatAI, use_model
|
|
from great_ai.utilities import clean
|
|
|
|
@GreatAI.create
|
|
@use_model('my-domain-predictor') #(1)
|
|
def predict_domain(sentence, model):
|
|
inputs = [clean(sentence)]
|
|
return str(model.predict(inputs)[0])
|
|
```
|
|
|
|
1. [@use_model][great_ai.use_model] loads and injects your model into the `predict_domain` function's `model` argument.
|
|
You can freely reference it knowing that it is always given to the function.
|
|
|
|
Finally, we deploy the model, inference function, and the GreatAI wrapping all of these. For that we either use: `great-ai deploy.ipynb` or build a Docker image.
|