PyPI - autogluon.timeseries - Versions diffs - 1.2.1b20250407__tar.gz → 1.2.1b20250409__tar.gz - Mend

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.timeseries
-Version: 1.2.1b20250407
+Version: 1.2.1b20250409
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community

@@ -408,6 +408,8 @@ class ChronosModel(AbstractTimeSeriesModel):
         time_limit: Optional[int] = None,
         **kwargs,
     ) -> None:
+        import transformers
+        from packaging import version
         from transformers.trainer import PrinterCallback, Trainer, TrainingArguments
         from .pipeline import ChronosBoltPipeline, ChronosPipeline
@@ -496,6 +498,10 @@ class ChronosModel(AbstractTimeSeriesModel):
                 fine_tune_trainer_kwargs["load_best_model_at_end"] = False
                 fine_tune_trainer_kwargs["metric_for_best_model"] = None
+            if version.parse(transformers.__version__) >= version.parse("4.46"):
+                # transformers changed the argument name from `evaluation_strategy` to `eval_strategy`
+                fine_tune_trainer_kwargs["eval_strategy"] = fine_tune_trainer_kwargs.pop("evaluation_strategy")
             training_args = TrainingArguments(**fine_tune_trainer_kwargs, **pipeline_specific_trainer_kwargs)
             tokenizer_train_dataset = ChronosFineTuningDataset(
                 target_df=train_data,

@@ -82,7 +82,7 @@ class InstanceNorm(nn.Module):
         if loc_scale is None:
             loc = torch.nan_to_num(torch.nanmean(x, dim=-1, keepdim=True), nan=0.0)
             scale = torch.nan_to_num((x - loc).square().nanmean(dim=-1, keepdim=True).sqrt(), nan=1.0)
-            scale = torch.where(scale == 0, torch.abs(loc) + self.eps, scale)
+            scale = torch.where(scale == 0, self.eps, scale)
         else:
             loc, scale = loc_scale

@@ -218,9 +218,11 @@ class MultiWindowBacktestingModel(AbstractTimeSeriesModel):
     def _get_search_space(self):
         return self.model_base._get_search_space()
-    def _initialize_covariate_regressor_scaler(self, **kwargs) -> None:
+    def _initialize_transforms_and_regressor(self, **kwargs) -> None:
         # Do not initialize the target_scaler and covariate_regressor in the multi window model!
-        pass
+        self.target_scaler = None
+        self.covariate_scaler = None
+        self.covariate_regressor = None
     def _get_hpo_train_fn_kwargs(self, **train_fn_kwargs) -> dict:
         train_fn_kwargs["is_bagged_model"] = True

@@ -1,4 +1,4 @@
 """This is the autogluon version file."""
-__version__ = "1.2.1b20250407"
+__version__ = "1.2.1b20250409"
 __lite__ = False

@@ -2,11 +2,11 @@ joblib<2,>=1.1
 numpy<2.1.4,>=1.25.0
 scipy<1.16,>=1.5.4
 pandas<2.3.0,>=2.0.0
-torch<2.6,>=2.2
-lightning<2.6,>=2.2
+torch<2.7,>=2.2
+lightning<2.7,>=2.2
 pytorch_lightning
 transformers[sentencepiece]<4.50,>=4.38.0
-accelerate<1.0,>=0.34.0
+accelerate<2.0,>=0.34.0
 gluonts<0.17,>=0.15.0
 networkx<4,>=3.0
 statsforecast<2.0.1,>=1.7.0
@@ -17,9 +17,9 @@ fugue>=0.9.0
 tqdm<5,>=4.38
 orjson~=3.9
 tensorboard<3,>=2.9
-autogluon.core[raytune]==1.2.1b20250407
-autogluon.common==1.2.1b20250407
-autogluon.tabular[catboost,lightgbm,xgboost]==1.2.1b20250407
+autogluon.core[raytune]==1.2.1b20250409
+autogluon.common==1.2.1b20250409
+autogluon.tabular[catboost,lightgbm,xgboost]==1.2.1b20250409
 [all]

autogluon.timeseries 1.2.1b20250407__tar.gz → 1.2.1b20250409__tar.gz