Product Release: January 31, 2024
We are excited to announce our latest release, which includes several updates and new features to improve your experience.
SDK Enhancements
Update SDK version to v2.0.2
Please update SDK to latest version. Refer Upgrade MarkovML SDK .
Note: SDK versions v1.x.x and older are no longer supported. Please make sure you are on the latest version.
Upload & Download Model Files using MarkovML SDK
We have enhanced our SDK to allow upload and download trained model files. This feature helps you track files associated with your models and handle versioning, environment deployment, and related tasks.
You can store your model files against the model_id
and retrieve them using either the model_id
or model_name
, as shown in the code snippet below.
# Train and save model locally
X_train, y_train = get_dataset() # Assume get_dataset is defined
model = get_model() # Assume get_model is defined
model.fit(X_train, y_train)
local_path = "/path/to/save/model"
model.save(local_path)
# Track / upload model files using Markov SDK
import markov
mkv_model = markov.Model(name="My model")
mkv_model.register()
mkv_model.upload_model_files(local_file_paths=local_path)
# In a different a file
# Download the model files using Markov SDk
import markov
mkv_model = markov.Model.get_by_id("my-model-id") # Paste the id (can be found on Markov UI)
local_download_path = "/path/to/download/model"
mkv_model.download_model_files(local_destination_path=local_download_path)
Auto-Labeling
This feature enables you to label an unlabeled dataset using semi-supervised learning.
Manually label a small subset of your unlabeled dataset, comprising at least 200 data points, using MarkovML's interactive relabeling interface. Once labeled, you label the rest of the dataset in a single click.
Note: To effectively use this feature, ensure that the subset selected for labeling includes at least 200 data points and encompasses at least two distinct classes.
Redesigned Dashboard
We have redesigned the dashboard and sidebar navigation to make MarkovML features more accessible. The sidebar now highlights the essential features, while the rest can be accessed by clicking the 'More' button.
Bug Fixes and Performance Improvements
Additionally, we have made several bug fixes and performance improvements to make your experience with Markov better than ever before.