Is there any reason for the weights_x for exogenous variables to be 0? I am posting this { "model": "timegpt-1", "freq": "D", "fh": 10, "y": { "2022-04-01": 0, "2022-04-02": 0, "2022-04-03": 0, "2022-04-04": 0, "2022-04-05": 0, "2022-04-06": 0, "2022-04-07": 2733.05, "2022-04-08": 3453.36, "2022-04-09": 4152.52, "2022-04-10": 5618.6 }, "x":{ "2022-04-01": [0, 0, 564], "2022-04-02": [24.57, 2, 506], "2022-04-03": [33.85, 2, 793], "2022-04-04": [0, 0, 520], "2022-04-05": [2.98, 1, 595], "2022-04-06": [50.8, 3, 542], "2022-04-07": [2773.61, 147, 15720], "2022-04-08": [3274.04, 170, 22292], "2022-04-09": [3820.76, 235, 24825], "2022-04-10": [3418.79, 187, 18084], "2022-04-11": [2489.05, 154, 20488], "2022-04-12": [3500.67, 180, 28601], "2022-04-13": [3280.38, 171, 36100], "2022-04-14": [3618.21, 193, 39644], "2022-04-15": [1765.26, 59, 5440], "2022-04-16": [643.62, 15, 4456], "2022-04-17": [2256.66, 80, 18934], "2022-04-18": [2947.8, 142, 30546], "2022-04-19": [3244.73, 165, 52656], "2022-04-20": [3915.74, 142, 45231] }, "clean_ex_first": true, "finetune_steps": 0, "finetune_loss": "default" } and getting this as a response { "timestamp": [ "2022-04-11 000000", "2022-04-12 000000", "2022-04-13 000000", "2022-04-14 000000", "2022-04-15 000000", "2022-04-16 000000", "2022-04-17 000000", "2022-04-18 000000", "2022-04-19 000000", "2022-04-20 000000" ], "value": [ 3883.482040874953, 3386.3937840390154, 3287.4763035702654, 3365.4116063046404, 3111.0009617733904, 3081.6581883358904, 3041.3454441952654, 2999.3578953671404, 2965.7890477108904, 3079.4033055233904 ], "input_tokens": 70, "output_tokens": 10, "finetune_tokens": 0, "weights_x": [ 0, 0, 0 ] }

Hello! Your input is too short to compute feature importances. We recommend having an input length of 25 timesteps + horizon. Given your horizon is 10, you should feed at least 35 timesteps for the best results and to get feature importances.

Oh that makes sense. Thanks a lot!