Chris Lo

Hello there, a small code example of what I mean by the "curried function that dynamically creates a new subflow". **What I like about this design pattern:** - Assume I...

Also getting this same error (Prefect 2.2.0, KubernetesJob, tested on both default and Dask task runners), not sure if this helps but noticed that I only get this error for...

@madkinsz A few more observations that might help: This is what the flow looks like (I'm working on a reproducible example atm): ```python @flow def etl_flow(): series: List[pd.Series] = extract_raw_data()...

> Has the worker started at 5 minutes and 30 seconds? Or is it still creating? Maybe its related to the dask worker TTL? https://docs.dask.org/en/stable/configuration.html#distributed.scheduler.worker-ttl @caleb-recursion Unfortunately it doesn't seem...

@fkiraly > another suggestion: we could assume independenc residuals, and add variances of the components together, if they implement `predict_var`? This makes the most sense. But what would the resulting...

@fkiraly thanks for the clarification! I realised that I made a logical error with my "if" statement there about variance additivity and normality. Classic stats 101 mistake. Happy to work...

> Thanks! > > Will dig into this later. > > Quick question, is the idea that this is the same as `NaiveVariance(NaiveForecaster())`, but more efficiently computed due to explicit...

> `_predict_var` looks good! > > Change request: `_predict_quantiles` should call `_predict` and `_predict_var`, i.e., the private functions, not the public ones. @fkiraly I agree with your assessment to only...

@fkiraly Added an explanation of the predict variance / quantiles algorithm. > The three different cases need to be explained clearly. Instead of writing out the equations (which can be...

@TNTran92 I have this simple utility function for joining panel, time-series, and cross-sectional data that might help as well: ```python def reshape_X_into_panel( panel_idx: pd.MultiIndex, X_static: Optional[pd.DataFrame] = None, X_ts: Optional[pd.DataFrame]...