autogluon.timeseries 1.2.1b20250204__py3-none-any.whl → 1.2.1b20250206__py3-none-any.whl

autogluon/timeseries/models/chronos/model.py

@@ -9,6 +9,7 @@ import numpy as np
 import pandas as pd
 
 from autogluon.common.loaders import load_pkl
+from autogluon.common.space import Space
 from autogluon.timeseries.dataset.ts_dataframe import TimeSeriesDataFrame
 from autogluon.timeseries.models.abstract import AbstractTimeSeriesModel
 from autogluon.timeseries.utils.warning_filters import disable_duplicate_logs, warning_filter
@@ -192,39 +193,11 @@ class ChronosModel(AbstractTimeSeriesModel):
         model_path_input = hyperparameters.get("model_path", self.default_model_path)
         self.model_path = MODEL_ALIASES.get(model_path_input, model_path_input)
 
-        # TODO: automatically determine batch size based on GPU / memory availability
-        self.batch_size = hyperparameters.get("batch_size", self.default_batch_size)
-        self.num_samples = hyperparameters.get("num_samples", self.default_num_samples)
-        self.device = hyperparameters.get("device")
-
-        # if the model requires a GPU, set the torch dtype to bfloat16
-        self.torch_dtype = hyperparameters.get("torch_dtype", self.default_torch_dtype)
-
-        self.data_loader_num_workers = hyperparameters.get("data_loader_num_workers", 0)
-        self.optimization_strategy: Optional[Literal["onnx", "openvino"]] = hyperparameters.get(
-            "optimization_strategy", None
-        )
-        if self.optimization_strategy is not None:
-            warnings.warn(
-                (
-                    "optimization_strategy is deprecated and will be removed in a future release. "
-                    "We recommend using Chronos-Bolt models for fast inference on the CPU."
-                ),
-                category=FutureWarning,
-                stacklevel=3,
-            )
-        self.context_length = hyperparameters.get("context_length")
-
-        if self.context_length is not None and self.context_length > self.maximum_context_length:
-            logger.info(
-                f"\tContext length {self.context_length} exceeds maximum context length {self.maximum_context_length}."
-                f"Context length will be set to {self.maximum_context_length}."
-            )
-            self.context_length = self.maximum_context_length
-
-        # we truncate the name to avoid long path errors on Windows
-        model_path_safe = str(model_path_input).replace("/", "__").replace(os.path.sep, "__")[-50:]
-        name = (name if name is not None else "Chronos") + f"[{model_path_safe}]"
+        name = name if name is not None else "Chronos"
+        if not isinstance(model_path_input, Space):
+            # we truncate the name to avoid long path errors on Windows
+            model_path_safe = str(model_path_input).replace("/", "__").replace(os.path.sep, "__")[-50:]
+            name += f"[{model_path_safe}]"
 
         super().__init__(
             path=path,
@@ -332,6 +305,7 @@ class ChronosModel(AbstractTimeSeriesModel):
         self._model_pipeline = pipeline
 
     def persist(self) -> "ChronosModel":
+        # TODO: Check the model has been fit before persist
         self.load_model_pipeline()
         return self
 
@@ -344,6 +318,14 @@ class ChronosModel(AbstractTimeSeriesModel):
         """Gets params that are passed to the inner model."""
         init_args = super()._get_model_params().copy()
 
+        init_args.setdefault("batch_size", self.default_batch_size)
+        init_args.setdefault("num_samples", self.default_num_samples)
+        init_args.setdefault("device", None)
+        # if the model requires a GPU, set the torch dtype to bfloat16
+        init_args.setdefault("torch_dtype", self.default_torch_dtype)
+        init_args.setdefault("data_loader_num_workers", 0)
+        init_args.setdefault("context_length", None)
+        init_args.setdefault("optimization_strategy", None)
         init_args.setdefault("fine_tune", False)
         init_args.setdefault("keep_transformers_logs", False)
         init_args.setdefault("fine_tune_lr", 1e-5)
@@ -370,7 +352,7 @@ class ChronosModel(AbstractTimeSeriesModel):
             report_to="none",
             max_steps=init_args["fine_tune_steps"],
             gradient_accumulation_steps=1,
-            dataloader_num_workers=self.data_loader_num_workers,
+            dataloader_num_workers=init_args["data_loader_num_workers"],
             tf32=self._has_tf32(),
             save_only_model=True,
             prediction_loss_only=True,
@@ -389,6 +371,36 @@ class ChronosModel(AbstractTimeSeriesModel):
 
         return init_args
 
+    def _validate_and_assign_attributes(self, model_params: dict):
+        # we validate the params here because their values are concrete,
+        # unlike in the constructor where they may be a search space
+
+        # TODO: automatically determine batch size based on GPU / memory availability
+        self.batch_size = model_params["batch_size"]
+        self.num_samples = model_params["num_samples"]
+        self.device = model_params["device"]
+        self.torch_dtype = model_params["torch_dtype"]
+        self.data_loader_num_workers = model_params["data_loader_num_workers"]
+        self.optimization_strategy: Optional[Literal["onnx", "openvino"]] = model_params["optimization_strategy"]
+
+        if self.optimization_strategy is not None:
+            warnings.warn(
+                (
+                    "optimization_strategy is deprecated and will be removed in a future release. "
+                    "We recommend using Chronos-Bolt models for fast inference on the CPU."
+                ),
+                category=FutureWarning,
+                stacklevel=3,
+            )
+        self.context_length = model_params["context_length"]
+
+        if self.context_length is not None and self.context_length > self.maximum_context_length:
+            logger.info(
+                f"\tContext length {self.context_length} exceeds maximum context length {self.maximum_context_length}."
+                f"Context length will be set to {self.maximum_context_length}."
+            )
+            self.context_length = self.maximum_context_length
+
     def _fit(
         self,
         train_data: TimeSeriesDataFrame,
@@ -419,13 +431,14 @@ class ChronosModel(AbstractTimeSeriesModel):
 
         self._check_fit_params()
 
-        fine_tune_args = self._get_model_params()
-        do_fine_tune = fine_tune_args["fine_tune"]
+        model_params = self._get_model_params()
+        self._validate_and_assign_attributes(model_params)
+        do_fine_tune = model_params["fine_tune"]
 
         if do_fine_tune:
             assert train_data is not None, "train_data cannot be None when fine_tune=True"
 
-        eval_during_fine_tune = val_data is not None and fine_tune_args["eval_during_fine_tune"]
+        eval_during_fine_tune = val_data is not None and model_params["eval_during_fine_tune"]
 
         if do_fine_tune:
             context_length = self._get_context_length(train_data)
@@ -461,7 +474,7 @@ class ChronosModel(AbstractTimeSeriesModel):
             else:
                 raise ValueError(f"Unsupported model pipeline: {type(self.model_pipeline)}")
 
-            fine_tune_trainer_kwargs = fine_tune_args["fine_tune_trainer_kwargs"]
+            fine_tune_trainer_kwargs = model_params["fine_tune_trainer_kwargs"]
             fine_tune_trainer_kwargs["use_cpu"] = str(self.model_pipeline.inner_model.device) == "cpu"
 
             if fine_tune_trainer_kwargs["use_cpu"]:
@@ -493,7 +506,7 @@ class ChronosModel(AbstractTimeSeriesModel):
                 # the original Chronos models otherwise the data is returned in ChronosBolt's format
                 tokenizer=getattr(self.model_pipeline, "tokenizer", None),
                 mode="training",
-            ).shuffle(fine_tune_args["fine_tune_shuffle_buffer_size"])
+            ).shuffle(model_params["fine_tune_shuffle_buffer_size"])
 
             callbacks = []
             if time_limit is not None:
@@ -503,8 +516,8 @@ class ChronosModel(AbstractTimeSeriesModel):
                 callbacks.append(EvaluateAndSaveFinalStepCallback())
                 # evaluate on a randomly-sampled subset
                 fine_tune_eval_max_items = (
-                    min(val_data.num_items, fine_tune_args["fine_tune_eval_max_items"])
-                    if fine_tune_args["fine_tune_eval_max_items"] is not None
+                    min(val_data.num_items, model_params["fine_tune_eval_max_items"])
+                    if model_params["fine_tune_eval_max_items"] is not None
                     else val_data.num_items
                 )
 
@@ -548,7 +561,7 @@ class ChronosModel(AbstractTimeSeriesModel):
             logger.info(f"\tSaving fine-tuned model to {fine_tuned_ckpt_path}")
             self.model_pipeline.inner_model.save_pretrained(Path(self.path) / self.fine_tuned_ckpt_name)
 
-            if not fine_tune_args["keep_transformers_logs"]:
+            if not model_params["keep_transformers_logs"]:
                 logger.debug(f"Removing transformers_logs directory {output_dir}")
                 shutil.rmtree(output_dir)
 

autogluon/timeseries/version.py

@@ -1,4 +1,4 @@
 """This is the autogluon version file."""
 
-__version__ = "1.2.1b20250204"
+__version__ = "1.2.1b20250206"
 __lite__ = False

{autogluon.timeseries-1.2.1b20250204.dist-info → autogluon.timeseries-1.2.1b20250206.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.timeseries
-Version: 1.2.1b20250204
+Version: 1.2.1b20250206
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community
@@ -55,9 +55,9 @@ Requires-Dist: fugue>=0.9.0
 Requires-Dist: tqdm<5,>=4.38
 Requires-Dist: orjson~=3.9
 Requires-Dist: tensorboard<3,>=2.9
-Requires-Dist: autogluon.core[raytune]==1.2.1b20250204
-Requires-Dist: autogluon.common==1.2.1b20250204
-Requires-Dist: autogluon.tabular[catboost,lightgbm,xgboost]==1.2.1b20250204
+Requires-Dist: autogluon.core[raytune]==1.2.1b20250206
+Requires-Dist: autogluon.common==1.2.1b20250206
+Requires-Dist: autogluon.tabular[catboost,lightgbm,xgboost]==1.2.1b20250206
 Provides-Extra: all
 Provides-Extra: chronos-onnx
 Requires-Dist: optimum[onnxruntime]<1.20,>=1.17; extra == "chronos-onnx"

{autogluon.timeseries-1.2.1b20250204.dist-info → autogluon.timeseries-1.2.1b20250206.dist-info}/RECORD

@@ -1,4 +1,4 @@
-autogluon.timeseries-1.2.1b20250204-py3.9-nspkg.pth,sha256=cQGwpuGPqg1GXscIwt-7PmME1OnSpD-7ixkikJ31WAY,554
+autogluon.timeseries-1.2.1b20250206-py3.9-nspkg.pth,sha256=cQGwpuGPqg1GXscIwt-7PmME1OnSpD-7ixkikJ31WAY,554
 autogluon/timeseries/__init__.py,sha256=_CrLLc1fkjen7UzWoO0Os8WZoHOgvZbHKy46I8v_4k4,304
 autogluon/timeseries/evaluator.py,sha256=l642tYfTHsl8WVIq_vV6qhgAFVFr9UuZD7gLra3A_Kc,250
 autogluon/timeseries/learner.py,sha256=PDAHFlos6q5JukwRE86tKoH0zxYf3nLzy7qfD_a5NYY,13849
@@ -6,7 +6,7 @@ autogluon/timeseries/predictor.py,sha256=HTE8a_R_9U0z-KlxyoELm-64BXNRzFu3mIEbTab
 autogluon/timeseries/regressor.py,sha256=dIXttb0SOGS8IAwZOMANNDc796spN0LMysGUvuKgskU,9623
 autogluon/timeseries/splitter.py,sha256=yzPca9p2bWV-_VJAptUyyzQsxu-uixAdpMoGQtDzMD4,3205
 autogluon/timeseries/trainer.py,sha256=IAX6zJBpljdWvgm0gjlI9EiOzU3BMSIttNydzy66zHg,57333
-autogluon/timeseries/version.py,sha256=etlL75BgGblMklwNZJWSVpR1mEowTtAQIUD5d_plfag,91
+autogluon/timeseries/version.py,sha256=HW8IYQYcNP5gBk-bKSY_rkK5vHCBYKLhFTjTRVYE_j4,91
 autogluon/timeseries/configs/__init__.py,sha256=BTtHIPCYeGjqgOcvqb8qPD4VNX-ICKOg6wnkew1cPOE,98
 autogluon/timeseries/configs/presets_configs.py,sha256=cLat8ecLlWrI-SC5KLBDCX2SbVXaucemy2pjxJAtSY0,2543
 autogluon/timeseries/dataset/__init__.py,sha256=UvnhAN5tjgxXTHoZMQDy64YMDj4Xxa68yY7NP4vAw0o,81
@@ -26,7 +26,7 @@ autogluon/timeseries/models/autogluon_tabular/mlforecast.py,sha256=H2UlpnJcIIEi_
 autogluon/timeseries/models/autogluon_tabular/transforms.py,sha256=XVoy8KpvoeX38lHHAXq4Be9LCxKjxZ36SOFeSAICRFM,2524
 autogluon/timeseries/models/autogluon_tabular/utils.py,sha256=Fn3Vu_Q0PCtEUbtNgLp1xIblg7dOdpFlF3W5kLHgruI,63
 autogluon/timeseries/models/chronos/__init__.py,sha256=wT77HzTtmQxW3sw2k0mA5Ot6PSHivX-Uvn5fjM05EU4,60
-autogluon/timeseries/models/chronos/model.py,sha256=Is5WF0LWebH0dlNO1d8I27ogM-fAcoy8__FMirlrxBQ,30506
+autogluon/timeseries/models/chronos/model.py,sha256=ijupqg4S6kRxdJWanNt6bnyPoGAaJluUHHmZpyQrfDE,31211
 autogluon/timeseries/models/chronos/pipeline/__init__.py,sha256=N-YZH9BGBoi99r5cznJe1zEEjwjIg7cOYIHZkKuJq44,247
 autogluon/timeseries/models/chronos/pipeline/base.py,sha256=14OAKHmio6LmO4mVom2mPGB0CvIrOjMGJzb-MVSAq-s,5596
 autogluon/timeseries/models/chronos/pipeline/chronos.py,sha256=uFJLsSb2WQiSrmDZ0g2mO-lhTFUlq7vplGRBXZ9_VBk,22591
@@ -58,11 +58,11 @@ autogluon/timeseries/utils/datetime/base.py,sha256=3NdsH3NDq4cVAOSoy3XpaNixyNlbj
 autogluon/timeseries/utils/datetime/lags.py,sha256=gQDk5_zmsY5DUWDUpSaCKYkQ9nHKKY-LsywJQRAoYSk,5988
 autogluon/timeseries/utils/datetime/seasonality.py,sha256=YK_2k8hvYIMW-sJPnjGWRtCnvIOthwA2hATB3nwVoD4,834
 autogluon/timeseries/utils/datetime/time_features.py,sha256=MjLi3zQ00uWWJtXH9oGX2GJkTbvjdSiuabSa4kcVuxE,2672
-autogluon.timeseries-1.2.1b20250204.dist-info/LICENSE,sha256=CeipvOyAZxBGUsFoaFqwkx54aPnIKEtm9a5u2uXxEws,10142
-autogluon.timeseries-1.2.1b20250204.dist-info/METADATA,sha256=efEPQU0dus9o80QwXGGqlvXiQWYaWDhY8TKCH81AsP4,12662
-autogluon.timeseries-1.2.1b20250204.dist-info/NOTICE,sha256=7nPQuj8Kp-uXsU0S5so3-2dNU5EctS5hDXvvzzehd7E,114
-autogluon.timeseries-1.2.1b20250204.dist-info/WHEEL,sha256=tZoeGjtWxWRfdplE7E3d45VPlLNQnvbKiYnx7gwAy8A,92
-autogluon.timeseries-1.2.1b20250204.dist-info/namespace_packages.txt,sha256=giERA4R78OkJf2ijn5slgjURlhRPzfLr7waIcGkzYAo,10
-autogluon.timeseries-1.2.1b20250204.dist-info/top_level.txt,sha256=giERA4R78OkJf2ijn5slgjURlhRPzfLr7waIcGkzYAo,10
-autogluon.timeseries-1.2.1b20250204.dist-info/zip-safe,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
-autogluon.timeseries-1.2.1b20250204.dist-info/RECORD,,
+autogluon.timeseries-1.2.1b20250206.dist-info/LICENSE,sha256=CeipvOyAZxBGUsFoaFqwkx54aPnIKEtm9a5u2uXxEws,10142
+autogluon.timeseries-1.2.1b20250206.dist-info/METADATA,sha256=H-VEWEC46px1FIkn59ZDKYa_ZayhvGIEmLlnR0s9nv0,12662
+autogluon.timeseries-1.2.1b20250206.dist-info/NOTICE,sha256=7nPQuj8Kp-uXsU0S5so3-2dNU5EctS5hDXvvzzehd7E,114
+autogluon.timeseries-1.2.1b20250206.dist-info/WHEEL,sha256=tZoeGjtWxWRfdplE7E3d45VPlLNQnvbKiYnx7gwAy8A,92
+autogluon.timeseries-1.2.1b20250206.dist-info/namespace_packages.txt,sha256=giERA4R78OkJf2ijn5slgjURlhRPzfLr7waIcGkzYAo,10
+autogluon.timeseries-1.2.1b20250206.dist-info/top_level.txt,sha256=giERA4R78OkJf2ijn5slgjURlhRPzfLr7waIcGkzYAo,10
+autogluon.timeseries-1.2.1b20250206.dist-info/zip-safe,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
+autogluon.timeseries-1.2.1b20250206.dist-info/RECORD,,