ludwig
ludwig copied to clipboard
ludwig-ai
Low-code framework for building custom LLMs, neural networks, and other AI models
Currently, I am using KubeRay to start a distributed Ray cluster on kubernetes and get the endpoint which can be assessed from external either using LoadBalancer or NodePort. Then I...
The current "expanded" Ludwig config (saved as `model_hyperparameters.json`) contains user config-specified parameters merged with preprocesing and metadata parameters. However, there aren't any module-specific parameters (unless they were explicitly specified by...
Model-agnostic, predict-only parameters like `top_k` and `threshold` are "free" parameters since they have no bearing on how the model is architected or how the model is trained. However, they are...
While a precise time to completion would be difficult to estimate as convergence depends on a mix of non-deterministic factors like early stopping criteria, there’s likely a reasonably good "max...
In several locations in Ludwig code, there are manual checks for the Ray version that workaround backwards-incompatible changes between different versions of Ray. #2173 compiles a partial list. As we...
A dataset with a split column with missing 0 (train) values, yields the following error during Ludwig preprocessing: ```python-traceback File "/home/ray/anaconda3/lib/python3.8/site-packages/ludwig/api.py", line 434, in train preprocessed_data = self.preprocess( File "/home/ray/anaconda3/lib/python3.8/site-packages/ludwig/api.py",...
Add integrations for experiment management for hyperopt. Notes & Requirements: - Should work with a ray backend - The current tensorboard `tfevents` file only stores the tune logs, and not...
Factor out lightweight functions that enable convenient fetching of default parameters from schemas, [motivating example](https://github.com/predibase/predibase/blob/0b1de5eee84e6a7792900389eca385e38ace78f7/engine/predibase_engine/automl/config.py#L30-L40).
In the push to create a fully fledged schema using Marshmallow, we've run into a bit of an issue. When specifying things like optimizers, the structure of the config works...