[RFW0092] Tracking Mt model and creating documentation

[TenzinGayche]

RFW0092: Tracking Mt model and creating documentation of the MT training pipeline

Named Concepts

hyperparameter: A hyperparameter is a machine learning parameter that is chosen before a learning algorithm is trained. Hyperparameters are used to improve the learning of a model. They control how the model is training and determine the values of model parameters that a learning algorithm ends up learning

Summary

Tracking the model version and its hyperparameter and documenting of hyperparameter tuning and experiments Documentation is for both Technical/Non Technical

We need a document to keep track of all the experiments and settings we've adjusted. This way, anyone can easily join in and help with the project.

Input

The training data and benchmark data set with its distribution report. The model that is being trained

Expected Output

documentation of hyperparameter tuning and experiments Documentation :

• Standardised version naming (version of dataset)
• model link
• Data content distribution (eg Buddhist literature )
• Hyperparameter tuned
• Evaluation Metric and History
• Graphs

Expected Timeline

References

[kaldan007]

@TenzinGayche can we know the need of rfw briefly in summary

[kaldan007]

does this rfw demands documentation file only or we need a to make a script which will update the documentation once every experiment.