SensiML Python Client¶
Current Release¶
2020.2.1 (07/07/2020)¶
What’s New
- Adds API to delete knowledge Packs
dsk.delete_knowledgepack(uuid) kp = dsk.get_knowledgepack(uuid) kp.delete()
- Adds API to view featurefiles stored on the server
dsk.list_featurefiles() dsk.get_featurefile(uuid)
- Adds API to view datafiles stored on the server
dsk.list_datafiles() dsk.get_datafiles(uuid)
Past Releases¶
2020.2.0 (07/07/2020)¶
What’s New
Adds tensorflow helper functions
Adds support for local queries against Data Capture Lab Project
Adds support for bulk updates/creates for Labels, Label_Values, Capture Label and Metadata Relationships
2020.1.2 (03/03/2020)¶
What’s New
Ability to query by segment_uuid
Adds more detailed statistics for queries
2020.1.0 (01/20/2020)¶
What’s New
Adds the ability to visually plot features in the model explore widget
Improvements to debug log outputs and addition of profiling option
Adds option for automating the creation of a hierarchical model which can provide increased accuracy but may increase the size of the model
2019.3.6 (11/13/2019)¶
Bug Fixes
Specifies a version for pywin32 as the latest version is breaking installs
2019.3.5 (10/21/2019)¶
What’s New
Minor Features
Better rendering of error messages in the Dashboard
Dashboard now resizes the notebook width on refresh
API updates
2019.3.4 (09/19/2019)¶
What’s New
Major Features
Autosense pipeline now runs asynchronously and has a terminate button
Project and Pipeline are now locked while autosense is running
2019.3.3 (09/19/2019)¶
What’s New
Major Features
Additional Pipeline settings
Ability to specify custom features for the feature family in the model building step
Added ability to use data files in the model explore widget
2019.3.1 (08/22/2019)¶
What’s New
Major Features
Additional plotting that shows the number of samples for each capture events along with the number of segments
Improvements to the Model Explore widget to allow selecting multiple capture files for model evaluation
Addition of new model selection algorithms Metadata K-fold and Stratified Metadata K-fold to Model Creation Widget
Minor visual improvements
2019.3.0 (07/30/2019)¶
What’s New
Major Features
Adds support for SensorTile 1.0 Knowledge Pack creation
Adds support for SensorTile Firmware 1.0 flashing
Overhaul of the Model Creation Widget, which now supports selecting an optimizing for a specific metric along with the Classifier Size in bytes
Addition of the Project Explorer to the Data Exploration Widget
2019.2.0 (06/11/2019)¶
What’s New
Major Features
Adds support for choosing the validation method used in the auto sense pipeline
Adds support for balancing data as part of the AutoSense pipeline
Adds support for query by capture_uuid
Adds post processing to model explore widget for test data that uses a majority voting algorithm over the last N examples
Minor Features
Adds support for displaying the created date to projects, pipelines, Knowledge Packs and captures
Bug Fixes
Fixes issue where downloading a Knowledge Pack after running the autosense pipeline was resetting the settings
2019.1.3 (06/04/2019)¶
What’s New
Minor Features
Adds status messages information to widget output as well as log output
Previous results for Auto Sense pipeline will now be displayed when logging into the dashboard if available
Sets the width of jupyter notebooks to 95% when import sensiml dashboard
Switch from project level to pipeline level for displaying Knowledge Packs in explore models and create knowledgepack tabs
All list functions now also display the uuid of the object (ie: list_captures, list_capture_configurations…)
Adds a model summary view to the model exploration tab
2019.1.2 (05/16/2019)¶
What’s New
Model Exploration Widget
Enables viewing confusion matrix in UI
Enables viewing feature summary in UI
Enables testing a model with Test Data captured through the DCL
2019.1.0 (05/05/2019)¶
What’s New
Minor Features
UI improvements to DashBoard widget
Additional documentation
Capture Configuration is selectable in the download widget
Bug Fixes
Removes Metadata from the query widget label field
2.5.6 (03/18/2019)¶
Bug Fixes
Locking down jupyter notebook to 5.7.5 until his issue is resolved with 5.7.6 https://github.com/jupyter/notebook/issues/4467
2.5.5 (03/14/2019)¶
What’s New
Minor Features
UI improvements to DashBoard widget
Adds additional documentation
Bug Fixes
Disable jedi autocomplete in ipython as it is causing errors with autocompletion
2.5.2 (02/04/2019)¶
Bug Fixes
Improvements to DashBoard widget stability
SMLRunner bug fix for get_model
Fixed issue in DashBoard widget where user could accidentally queue multiple pipelines by clicking ‘Optimize Knowledge Pack’ multiple times
2.5.1 (01/15/2019)¶
What’s New
Updates to Dashboard widget layout
SMLRunner improved state checking
Provides version compatibility checking against the SensiML Cloud server
Generating a Knowledge Pack now displays the local directory path
2.4.0 (10/24/2018)¶
What’s New
Improvements to Download Widget
Bug Fixes
Adding some sanity checks to SMLRunner that prevent getting Knowledge Pack in a bad state
2.3.15 (10/09/2018)¶
Bug Fixes
Fixes bug where sml_runner would enter debug mode
Flash widget now finds zip files from downloads and unzips for Nordic thingy
Improved error reporting from server
Minor stability improvements
Fixes some python 2/3 compatibility issues
2.3.13 (09/10/2018)¶
Bug Fixes
Fixed bug in feature visualization that was causing incorrect number of plots to show in Python 3
Added DashBoard to top level sensiml import
2.3.0¶
What’s New
Major Features
SensiML python client now supports python 3.4+ (pip install sensiml)
Adding SMLRunner for calling a Knowledge Pack c library from sensiml client
Minor Features
Pipeline status is returned during pipeline execution
Deprecation Warning
pipeline.set_input_data no longer allows passing of a dataframe, use this practice instead:
dsk.upload_dataframe('file_name', df) dsk.set_input_data('file_name', data_columns=...)NOTE: There are minor differences in syntax between python 2 and 3. The most obvious being print statements now require parentheses. ie print(“message”) instead of print “message” cheat sheet for 2-3: http://python-future.org/compatible_idioms.html
2.2.2¶
What’s New
Major Features
Adding ability to use the emulator for hierarchical models via recognize_signal
SensiML python client can now be installed/updated through pip (pip install sensiml)
Minor Features
Adding ability to rehydrate a Knowledge Pack summary (this is only the part of the Knowledge Pack that is put on the device)
General stability improvements and bug fixes
2.2.1¶
What’s New
Minor Features
ModelVisualizationWidget improvements with support for Windows Test App and Feature Vectors DataFrame
2.2.0¶
What’s New
Major Features
SensiML Dashboard Widget
Minor Features
Visual improvements/bug fixes to all widgets
KP download now includes option to explicitly define source (Audio, Motion, Custom)
SensiML Labs (Experimental Features)
Model Visualization Widget for viewing feature values in real time (early concept)
Bug Fixes
Generating OTA file from Flash widget works without installing nRFgo Studio
2.1.3¶
What’s New
Major Features
Improvements to all widgets (Query, Download, Flash, AutoSense)
Creation of Dashboard widget
Bug Fixes
Flash widget now working on windows
2.1.2¶
What’s New
Major Features
Adds support for hierarchical models in Download Widget
Adds ability to upload project data from Data Capture Lab Project
2.1.1¶
What’s New
Major Features
Added visualizations for comparing features as well as neuron/feature placement
Minor Features
Added ability to specify a capture file as the test data in recognize signal
Added ability to upload a dcl project from the client
Widget optimizations
2.1.0¶
What’s New
Major Features
Kbclient has been replaced by SensiML as the python client for connecting to SensiML’s Rest APIs
Minor Features
Label column is now part of the query
Widget improvements
Query statistics are now displayed when adding a query through the widget
statistics_segments() now detailed information for all segment in that query
statistics() now returns summary information for query
2.0.0¶
What’s New
Major Features
Widgets - KB Client now has some built in widgets for bring a GUI to some basic functions. Widgets have been created for Querying, Pipeline Automation, and Knowledge Pack Generation