autogluon.timeseries 1.1.2b20241121__tar.gz → 1.1.2b20241123__tar.gz

{autogluon.timeseries-1.1.2b20241121 → autogluon.timeseries-1.1.2b20241123}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.timeseries
-Version: 1.1.2b20241121
+Version: 1.1.2b20241123
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community

{autogluon.timeseries-1.1.2b20241121 → autogluon.timeseries-1.1.2b20241123}/setup.py

@@ -30,7 +30,7 @@ install_requires = [
     "pytorch_lightning",  # version range defined in `core/_setup_utils.py`
     "transformers[sentencepiece]",  # version range defined in `core/_setup_utils.py`
     "accelerate",  # version range defined in `core/_setup_utils.py`
-    "gluonts==0.16.0",
+    "gluonts>=0.15.0,<0.17",
     "networkx",  # version range defined in `core/_setup_utils.py`
     "statsforecast>=1.7.0,<1.8",
     "mlforecast==0.13.4",

{autogluon.timeseries-1.1.2b20241121 → autogluon.timeseries-1.1.2b20241123}/src/autogluon/timeseries/models/chronos/model.py

@@ -1,6 +1,7 @@
 import logging
 import os
 import shutil
+import warnings
 from pathlib import Path
 from typing import Any, Dict, Literal, Optional, Union
 
@@ -15,6 +16,11 @@ from autogluon.timeseries.utils.warning_filters import disable_duplicate_logs, w
 
 logger = logging.getLogger("autogluon.timeseries.models.chronos")
 
+# TODO: Replace `evaluation_strategy` with `eval_strategy` when upgrading to `transformers>=4.41` + remove warning filter
+warnings.filterwarnings("ignore", category=FutureWarning, message="`evaluation_strategy` is deprecated")
+# TODO: Remove warning filter when upgrading to `transformers>=4.40`
+warnings.filterwarnings("ignore", category=FutureWarning, message="Passing the following arguments to ")
+
 
 # allowed HuggingFace model paths with custom parameter definitions
 MODEL_CONFIGS = {
@@ -104,7 +110,7 @@ class ChronosModel(AbstractTimeSeriesModel):
 
     Other Parameters
     ----------------
-    model_path: str, default = "autogluon/chronos-t5-small"
+    model_path: str, default = "autogluon/chronos-bolt-small"
         Model path used for the model, i.e., a HuggingFace transformers ``name_or_path``. Can be a
         compatible model name on HuggingFace Hub or a local path to a model directory. Original
         Chronos models (i.e., ``autogluon/chronos-t5-{model_size}``) can be specified with aliases
@@ -140,8 +146,9 @@ class ChronosModel(AbstractTimeSeriesModel):
         for more information.
     fine_tune : bool, default = False
         If True, the pretrained model will be fine-tuned
-    fine_tune_lr: float, default = 0.0001
-        The learning rate used for fine-tuning
+    fine_tune_lr: float, default = 1e-5
+        The learning rate used for fine-tuning. This default is suitable for Chronos-Bolt models; for the original
+        Chronos models, we recommend using a higher learning rate such as ``1e-4``
     fine_tune_steps : int, default = 1000
         The number of gradient update steps to fine-tune for
     fine_tune_batch_size : int, default = 32
@@ -163,7 +170,7 @@ class ChronosModel(AbstractTimeSeriesModel):
 
     # default number of samples for prediction
     default_num_samples: int = 20
-    default_model_path = "autogluon/chronos-t5-small"
+    default_model_path = "autogluon/chronos-bolt-small"
     default_max_time_limit_ratio = 0.8
     maximum_context_length = 2048
     fine_tuned_ckpt_name: str = "fine-tuned-ckpt"
@@ -198,7 +205,7 @@ class ChronosModel(AbstractTimeSeriesModel):
         self.context_length = hyperparameters.get("context_length")
 
         if self.context_length is not None and self.context_length > self.maximum_context_length:
-            logger.warning(
+            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}."
             )
@@ -321,7 +328,7 @@ class ChronosModel(AbstractTimeSeriesModel):
 
         init_args.setdefault("fine_tune", False)
         init_args.setdefault("keep_transformers_logs", False)
-        init_args.setdefault("fine_tune_lr", 1e-4)
+        init_args.setdefault("fine_tune_lr", 1e-5)
         init_args.setdefault("fine_tune_steps", 1000)
         init_args.setdefault("fine_tune_batch_size", 32)
         init_args.setdefault("eval_during_fine_tune", False)
@@ -341,6 +348,7 @@ class ChronosModel(AbstractTimeSeriesModel):
             logging_dir=str(output_dir),
             logging_strategy="steps",
             logging_steps=100,
+            disable_tqdm=True,
             report_to="none",
             max_steps=init_args["fine_tune_steps"],
             gradient_accumulation_steps=1,
@@ -434,7 +442,6 @@ class ChronosModel(AbstractTimeSeriesModel):
                     )
 
             fine_tune_trainer_kwargs = fine_tune_args["fine_tune_trainer_kwargs"]
-            fine_tune_trainer_kwargs["disable_tqdm"] = fine_tune_trainer_kwargs.get("disable_tqdm", (verbosity < 3))
             fine_tune_trainer_kwargs["use_cpu"] = str(self.model_pipeline.inner_model.device) == "cpu"
 
             if fine_tune_trainer_kwargs["use_cpu"]:

{autogluon.timeseries-1.1.2b20241121 → autogluon.timeseries-1.1.2b20241123}/src/autogluon/timeseries/models/chronos/pipeline/utils.py

@@ -317,7 +317,7 @@ class TimeLimitCallback(TrainerCallback):
     def on_step_end(self, args, state, control, **kwargs):
         elapsed_time = time.monotonic() - self.start_time
         if elapsed_time > self.time_limit:
-            logger.info("\tStopping fine-tuning since time_limit is reached")
+            logger.log(15, "Stopping fine-tuning since time_limit is reached")
             control.should_training_stop = True
 
 

{autogluon.timeseries-1.1.2b20241121 → autogluon.timeseries-1.1.2b20241123}/src/autogluon/timeseries/models/presets.py

@@ -145,7 +145,6 @@ def get_default_hps(key):
         },
         "default": {
             "SeasonalNaive": {},
-            "Croston": {},
             "AutoETS": {},
             "NPTS": {},
             "DynamicOptimizedTheta": {},

{autogluon.timeseries-1.1.2b20241121 → autogluon.timeseries-1.1.2b20241123}/src/autogluon/timeseries/version.py

@@ -1,3 +1,3 @@
 """This is the autogluon version file."""
-__version__ = '1.1.2b20241121'
+__version__ = '1.1.2b20241123'
 __lite__ = False

{autogluon.timeseries-1.1.2b20241121 → autogluon.timeseries-1.1.2b20241123}/src/autogluon.timeseries.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.timeseries
-Version: 1.1.2b20241121
+Version: 1.1.2b20241123
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community

{autogluon.timeseries-1.1.2b20241121 → autogluon.timeseries-1.1.2b20241123}/src/autogluon.timeseries.egg-info/requires.txt

@@ -1,13 +1,13 @@
 joblib<2,>=1.1
-numpy<1.29,>=1.21
-scipy<1.13,>=1.5.4
+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
 pytorch_lightning
 transformers[sentencepiece]<5,>=4.38.0
 accelerate<1.0,>=0.32.0
-gluonts==0.16.0
+gluonts<0.17,>=0.15.0
 networkx<4,>=3.0
 statsforecast<1.8,>=1.7.0
 mlforecast==0.13.4
@@ -17,9 +17,9 @@ fugue>=0.9.0
 tqdm<5,>=4.38
 orjson~=3.9
 tensorboard<3,>=2.9
-autogluon.core[raytune]==1.1.2b20241121
-autogluon.common==1.1.2b20241121
-autogluon.tabular[catboost,lightgbm,xgboost]==1.1.2b20241121
+autogluon.core[raytune]==1.1.2b20241123
+autogluon.common==1.1.2b20241123
+autogluon.tabular[catboost,lightgbm,xgboost]==1.1.2b20241123
 
 [all]
 optimum[onnxruntime]<1.20,>=1.17