This project has been archived by its maintainers, and is no longer receiving any updates.
MlFoundry
This guide is to give an idea of how you can log metrics, paramaters, predictions, models, dataset using MlFoundry. MlFoundry supports logging above mentioned logs to S3 asynchronously.
API Workflow:
Using mlfoundry you ca create multiple projects and each projects can have multiple run.
Example
- Project 1
- run 1
- run 2
- Project 2
- run 1
- run 2
- run 3
- run 4
Each run under each project will have a unique run_id.
Quickstart
1. Install mlfoundry
# Install via pip
pip install mlfoundry --extra-index-url https://api.packagr.app/public
2. Setup AWS credentials
To setup AWS credentials MlLogs_foundry's CLI can be used. AWS credentials are required to store the artifacts like model, dataset, whylogs_metrics etc in S3 Bucket.
The CLI requires AWS credentials(SECRET_ACCESS_KEY, ACCESS_KEY_ID)
All the credentials can be set by running:
mlfoundry init
3. Initialise mlfoundry
Create an project and use the project_name to create a run:
import mlfoundry as mlf
# create a run
mlf_run = mlf.create_run(project_name=<project-name>)
If you want to use a previously created run:
# printing all the runs available and get the run_id
show_all_runs()
run = get_run(run_id=<run_id>)
4. Start logging
To log a model:
run.log_model(sklearn_model, framework=mlf.ModelFramework.SKLEARN)
To log parameters:
run.log_params({'learning_rate':0.01,
'n_epochs:10'
})
To log metrics:
run.log_metrics({'accuracy':87,
'f1_score':0.84,
})
Log predictions synchronously:
feature_df: a pd.DataFrame, the input given to the model to make predictions
predictions: must be a list or pd.Series
To log predictions synchronously:
run.log_predictions(self, feature_df, predictions)
To log predictions asynchronously:
responses = run.log_predictions_async(self, feature_df, predictions)
#### To confirm that the log request completed successfully, await for futures to resolve: This is a blocking call
import concurrent.futures as cf
for response in cf.as_completed(responses):
res = response.result()
Users can additionaly pass in feature_names argument which is a list of feture names for feature_df.
To release as Python package
Use Github Releases to create a tag on main and release it. This will trigger the workflow and publish the pip package
Tags must be of format vx.x.x, example v0.1.0
Development instructions
git clone https://github.com/truefoundry/mlfoundry.git
cd mlfoundry
virtualenv venv
source venv/bin/activate
pip install "poetry>=1.8.0,<2.0.0"
poetry install
pre-commit install
Run Manual QA
git clone https://github.com/truefoundry/mlf-test
cd mlf-test
virtualenv venv
source venv/bin/activate
pip install -r requirements.txt
cd mlf_examples
python main.py In-this-step-check-for-error-logs
mlfoundry ui In-this-step-go-to-the-listed-url-and-play-around-with-the-ui
2 maintainers
Release files for mlfoundry 0.11.5
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| mlfoundry-0.11.5.tar.gz | 94.5 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| mlfoundry-0.11.5-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 217.8 kB
Release files / mlfoundry-0.11.5.tar.gz
| Download URL | mlfoundry-0.11.5.tar.gz |
|---|---|
| Size | 94.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
0da5061b79d4707107ec1fe56996dbc424b329b5340408bb53db755eace7752f
|
|
BLAKE2b-256 checksum How to use checksums |
5dc22fd11b5533068347e1eb5e12945c3f618987963a42a917e092e19d0cd590
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/5.1.1 CPython/3.12.5
|
Release files / mlfoundry-0.11.5-py3-none-any.whl
| Download URL | mlfoundry-0.11.5-py3-none-any.whl |
|---|---|
| Size | 123.4 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
91d7b666e16cf640bc8d76cedca3ebdad2a62b46f095f918ff4b797e177f3b6a
|
|
BLAKE2b-256 checksum How to use checksums |
325aca8c05e8ada49425ae31cdd3d5f6c592a93d5bf3069c450fdad7074a5d3e
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/5.1.1 CPython/3.12.5
|
