[17:37:20] loaded 2001 [17:37:23] loaded 2002 [17:37:26] loaded 2003 [17:37:30] loaded 2004 [17:37:33] loaded 2005 [17:37:36] loaded 2006 [17:37:39] loaded 2007 [17:37:42] loaded 2008 [17:37:45] loaded 2009 [17:37:48] loaded 2010 [17:37:52] loaded 2011 [17:37:55] loaded 2012 [17:37:58] loaded 2013 [17:38:01] loaded 2014 [17:38:04] loaded 2015 [17:38:07] loaded 2016 [17:38:11] loaded 2017 [17:38:14] loaded 2018 [17:38:17] loaded 2019 [17:38:20] loaded 2020 [17:38:23] loaded 2021 [17:38:26] loaded 2022 [17:38:30] loaded 2023 [17:38:33] loaded 2024 [17:38:36] loaded 2025 /root/trendex-ca/scripts/30_train_ca.py:67: RuntimeWarning: Mean of empty slice anom = lambda x: x - np.nanmean(x, axis=1, keepdims=True) # noqa: E731 /root/trendex-ca/scripts/30_train_ca.py:76: RuntimeWarning: Mean of empty slice {f"{k}_mean": np.nanmean(v[:, :, j], axis=1).astype("float32") for k, v in raw_m.items()} | /root/trendex-ca/scripts/30_train_ca.py:77: RuntimeWarning: Mean of empty slice {f"{k}_mean": np.nanmean(v, axis=1).astype("float32") for k, v in raw_y.items()}) [17:39:37] eligible cells 3432538, sample 150000 /opt/conda/lib/python3.11/site-packages/xgboost/core.py:751: UserWarning: [17:41:03] WARNING: /__w/xgboost/xgboost/src/common/error_msg.cc:62: Falling back to prediction using DMatrix due to mismatched devices. This might lead to higher memory usage and slower performance. XGBoost is running on: cuda:0, while the input data is on: cpu. Potential solutions: - Use a data structure that matches the device ordinal in the booster. - Set the device for booster before call to inplace_predict. This warning will only be shown once. return func(**kwargs) [17:45:07] tuning 20 trials, xgb 0.0519 vs ridge 0.0524, n_est 189 [17:46:23] temporal skill vs ridge {'point': 0.05526524782180786, 'ci_lo': 0.03597217798233032, 'ci_hi': 0.07688862085342407, 'units': 6} [17:49:02] spatial skill vs ridge {'point': 0.11963582038879395, 'ci_lo': 0.11683060228824615, 'ci_hi': 0.12261748313903809, 'units': 4115} [17:49:02] GATE passed=True -> xgb /root/trendex-ca/scripts/30_train_ca.py:276: RuntimeWarning: Mean of empty slice clim_all = np.nanmean(f["ndvi"], axis=1, keepdims=True) # (n, 1, 5) /root/trendex-ca/scripts/30_train_ca.py:297: RuntimeWarning: Mean of empty slice summer_res[chunk[:, None], yi[None, :]] = np.nanmean(res[:, :, summer_idx], axis=2) [17:50:44] oof block 2001-2005 done [17:52:11] oof block 2006-2010 done [17:53:37] oof block 2011-2015 done [17:55:04] oof block 2016-2020 done [17:56:29] oof block 2021-2025 done [17:56:35] TRAIN_CA_DONE