neuralforecast

neuralforecast copied to clipboard

### What happened + What you expected to happen **What happened:** In the documentation of the `fit` method of NHITS, there are arguments `val_size` and `test_size`. However, when I try...

stephanielees

Adding MLflow loading and forecasting capability

6

### Description It will be good to add MLflow loading and forecasting capability. @jmoralez ### Use case To use MLflow to deploy models, this could also be leveraged with ONNX.

cargecla1

### What happened + What you expected to happen Report autoformer when I am training fedformer, not a big deal though. ### Versions / Dependencies  ### Reproduction script use...

Raphealshuai

ValueError: There are missing combinations of ids and times in `futr_df`. You can run the `make_future_dataframe()` method to get the expected combinations or the `get_missing_future(futr_df)` method to get the missing combinations.

3

### What happened + What you expected to happen 1.the bug: We are using Informer model with neural forecast. I get the mentioned error when I predict. forecasts = nf.predict(futr_df=test_data)...

Mohan16071996

add option to remove windows with poor data quality

12

This review adds the parameters data_availability_threshold (defaults to 0.0 to maintain currently functionality) to all models that inherit the BaseWindows class. This parameters allows us to discard windows where the...

jasminerienecker

[FEAT] enable setting gpu optimization variables

4

This is to allow adjusting the torch pin_memory and prefetch_factor variables to optimize gpu usage. Note: by adjusting these variables I am now able to increase GPU usage to 95%...

jasminerienecker

### Description Hi could you add a number of layers parameters in the TFT's LSTM so we can do stacking of LSTM? Like num_layers=1 by defaults? ### Use case I...

Otsutsukii

Neural Forecast: Tweedie Distribution Loss Broken (NBEATSx)

1

### What happened + What you expected to happen Attempting to train NBEATsx with NeuralForecast and Tweedie loss and encounter error claiming that "an output must be a tensor, not...

jrodenberg0

### Description model2=[TiDE(h=horizon, input_size=12, loss=MAE(), scaler_type='minmax', learning_rate=1e-3, max_steps=100, val_check_steps=100, early_stop_patience_steps=4, futr_exog_list=['vacation_num'], batch_size=32)] model3=[TiDE(h=horizon, input_size=12, loss=MAE(), scaler_type='minmax', learning_rate=1e-3, max_steps=100, val_check_steps=100, early_stop_patience_steps=4, futr_exog_list=['timeinformation'], batch_size=32)] I added two covariates separately, but why are...

HAUNGTOFU

MLflow loading and serving to address issue #1027

1

Started PR to add MLflow flavor based on mlforecast flavor. To address issue https://github.com/Nixtla/neuralforecast/issues/1027 @jmoralez Please confirm if the flavor location works for you, current path: .\GitHub\neuralforecast\neuralforecast **Note**: this work...

cargecla1