Update _predictor.py

Dockerfile

@@ -1,4 +1,4 @@
-FROM python:3
+FROM python:3.7
 
 ADD . /gluonts
 

README.md

@@ -1,6 +1,9 @@
 # GluonTS - Probabilistic Time Series Modeling in Python
 
-[![PyPI](https://img.shields.io/pypi/v/gluonts.svg?style=flat-square)](https://pypi.org/project/gluonts/) ![GitHub](https://img.shields.io/github/license/awslabs/gluon-ts.svg?style=flat-square)
+[![PyPI](https://img.shields.io/pypi/v/gluonts.svg?style=flat-square)](https://pypi.org/project/gluonts/)
+[![GitHub](https://img.shields.io/github/license/awslabs/gluon-ts.svg?style=flat-square)](./LICENSE)
+[![Static](https://img.shields.io/static/v1?label=docs&message=stable&color=blue&style=flat-square)][stable docs url]
+[![Static](https://img.shields.io/static/v1?label=docs&message=latest&color=blue&style=flat-square)][latest docs url]
 
 GluonTS is a Python toolkit for probabilistic time series modeling,
 built around [Apache MXNet (incubating)](https://mxnet.incubator.apache.org/).
@@ -9,9 +12,13 @@ GluonTS provides utilities for loading and iterating over time series datasets,
 state of the art models ready to be trained, and building blocks to define
 your own models and quickly experiment with different solutions.
 
-* [Documentation](https://gluon-ts.mxnet.io/)
+* [Documentation (stable version)][stable docs url]
+* [Documentation (latest)][latest docs url]
 * [Paper](https://arxiv.org/abs/1906.05264)
 
+[stable docs url]: https://gluon-ts.mxnet.io/
+[latest docs url]: https://gluon-ts.s3-accelerate.dualstack.amazonaws.com/master/index.html
+
 ## Installation
 
 GluonTS requires Python 3.6, and the easiest

docs/examples/extended_forecasting_tutorial/extended_tutorial.md

@@ -337,8 +337,8 @@ We can easily create the train and test datasets by simply filling in the correc
 ```python
 train_ds = ListDataset([{FieldName.TARGET: target, 
                          FieldName.START: start,
-                         FieldName.FEAT_DYNAMIC_REAL: fdr,
-                         FieldName.FEAT_STATIC_CAT: fsc} 
+                         FieldName.FEAT_DYNAMIC_REAL: [fdr],
+                         FieldName.FEAT_STATIC_CAT: [fsc]} 
                         for (target, start, fdr, fsc) in zip(target[:, :-custom_ds_metadata['prediction_length']], 
                                                              custom_ds_metadata['start'], 
                                                              feat_dynamic_real[:, :-custom_ds_metadata['prediction_length']], 
@@ -350,8 +350,8 @@ train_ds = ListDataset([{FieldName.TARGET: target,
 ```python
 test_ds = ListDataset([{FieldName.TARGET: target, 
                         FieldName.START: start,
-                        FieldName.FEAT_DYNAMIC_REAL: fdr,
-                        FieldName.FEAT_STATIC_CAT: fsc} 
+                        FieldName.FEAT_DYNAMIC_REAL: [fdr],
+                        FieldName.FEAT_STATIC_CAT: [fsc]} 
                        for (target, start, fdr, fsc) in zip(target, 
                                                             custom_ds_metadata['start'], 
                                                             feat_dynamic_real, 
@@ -693,7 +693,7 @@ print(f"Start date of the forecast window: {forecast_entry.start_date}")
 print(f"Frequency of the time series: {forecast_entry.freq}")
 ```
 
-We can also do calculations to summarize the sample paths, such computing the mean or a quantile for each of the 48 time steps in the forecast window.
+We can also do calculations to summarize the sample paths, such as computing the mean or a quantile for each of the 24 time steps in the forecast window.
 
 
 ```python

docs/examples/synthetic_data_generation_tutorial/tutorial.md

@@ -232,7 +232,7 @@ As recipes are just lists of expressions that evaluated sequentially, recipes ca
 
 ```python
 scaling = [
-    ("scale", rcp.RandomUniform(0, 1000)),
+    ("scale", rcp.RandomUniform(low=0, high=1000, shape=1)),
     ("z", "scale" * rcp.Ref("unscaled"))
 ]
 

examples/gp_synthetic_example.py

@@ -3,7 +3,7 @@
 
 # First-party imports
 from gluonts.kernels import RBFKernel
-from gluonts.gp import GaussianProcess
+from gluonts.model.gp_forecaster.gaussian_process import GaussianProcess
 
 # Third-party imports
 import mxnet.ndarray as nd

pytest.ini

@@ -5,3 +5,5 @@ markers =
     gpu: mark a test that requires GPU.
     integration: mark an integration test
     skip_master: mark a test that is temporarily skipped for mxnet master validation.
+
+timeout = 30

requirements/requirements-setup.txt

@@ -1,4 +1,4 @@
-black==19.3b0
-mypy==0.630
+black==19.10b0
+mypy==0.761
 setuptools>=40.1.0
 setuptools_scm>=3.3.3

requirements/requirements.txt

@@ -6,7 +6,7 @@ boto3~=1.0
 holidays>=0.9,<0.10
 matplotlib~=3.0
 numpy~=1.14
-pandas>=0.25,<0.26
+pandas~=1.0
 pydantic~=1.1
 tqdm~=4.23
 ujson~=1.35

src/gluonts/block/cnn.py

@@ -58,7 +58,7 @@ def __init__(
         channels: int,
         kernel_size: int,
         dilation: int = 1,
-        activation: Optional[str] = "relu",
+        activation: Optional[str] = None,
         **kwargs,
     ):
         super(CausalConv1D, self).__init__(**kwargs)

src/gluonts/core/component.py

@@ -35,10 +35,7 @@
 # Relative imports
 from . import fqname_for
 
-DEBUG = os.environ.get("DEBUG", "false").lower() == "true"
-
-logger = logging.getLogger()
-logger.setLevel(logging.DEBUG if DEBUG else logging.INFO)
+logger = logging.getLogger(__name__)
 
 A = TypeVar("A")
 
@@ -496,17 +493,17 @@ def num_gpus(refresh=False):
     return NUM_GPUS
 
 
+@functools.lru_cache()
 def get_mxnet_context(gpu_number=0) -> mx.Context:
     """
     Returns either CPU or GPU context
     """
-    n = num_gpus()
-    if n == 0:
-        logging.info("Using CPU")
-        return mx.context.cpu()
-    else:
-        logging.info("Using GPU")
+    if num_gpus():
+        logger.info("Using GPU")
         return mx.context.gpu(gpu_number)
+    else:
+        logger.info("Using CPU")
+        return mx.context.cpu()
 
 
 def check_gpu_support() -> bool:
@@ -516,7 +513,7 @@ def check_gpu_support() -> bool:
     """
     n = num_gpus()
     logger.info(f'MXNet GPU support is {"ON" if n > 0 else "OFF"}')
-    return False if n == 0 else True
+    return n != 0
 
 
 class DType:

src/gluonts/core/serde.py

@@ -560,7 +560,7 @@ def decode(r: Any) -> Any:
     # r = { 'class': ..., 'args': ... }
     # r = { 'class': ..., 'kwargs': ... }
     if type(r) == dict and r.get("__kind__") == kind_inst:
-        cls = locate(r["class"])
+        cls = cast(Any, locate(r["class"]))
         args = decode(r["args"]) if "args" in r else []
         kwargs = decode(r["kwargs"]) if "kwargs" in r else {}
         return cls(*args, **kwargs)

src/gluonts/dataset/artificial/recipe.py

@@ -165,16 +165,19 @@ def __rsub__(self, other):
         return LiftedSub(other, self)
 
     def __mul__(self, other):
-        return LiftedMul(self, other, operator.mul)
+        return LiftedMul(self, other)
 
     def __rmul__(self, other):
-        return LiftedMul(other, self, operator.mul)
+        return LiftedMul(other, self)
 
     def __truediv__(self, other):
-        return LiftedTruediv(self, other, operator.truediv)
+        return LiftedTruediv(self, other)
 
     def __rtruediv__(self, other):
-        return LiftedTruediv(other, self, operator.truediv)
+        return LiftedTruediv(other, self)
+
+    def __pow__(self, other):
+        return LiftedBinaryOp(self, other, operator.pow)
 
     def __call__(
         self,

src/gluonts/dataset/common.py

@@ -17,6 +17,7 @@
 from functools import lru_cache
 from pathlib import Path
 from typing import (
+    cast,
     Any,
     Callable,
     Dict,
@@ -25,8 +26,6 @@
     List,
     NamedTuple,
     Optional,
-    Sized,
-    cast,
     Union,
 )
 
@@ -35,21 +34,18 @@
 import pandas as pd
 import pydantic
 import ujson as json
-from pandas.tseries.frequencies import to_offset
 from pandas.tseries.offsets import Tick
 
 # First-party imports
 from gluonts.core.exception import GluonTSDataError
 from gluonts.dataset import jsonl, util
-from gluonts.dataset.stat import (
-    DatasetStatistics,
-    calculate_dataset_statistics,
-)
 
 # Dictionary used for data flowing through the transformations.
-# A Dataset is an iterable over such dictionaries.
 DataEntry = Dict[str, Any]
 
+# A Dataset is an iterable of DataEntry.
+Dataset = Iterable[DataEntry]
+
 
 class Timestamp(pd.Timestamp):
     # we need to sublcass, since pydantic otherwise converts the value into
@@ -152,21 +148,6 @@ class SourceContext(NamedTuple):
     row: int
 
 
-class Dataset(Sized, Iterable[DataEntry]):
-    """
-    An abstract class for datasets, i.e., iterable collection of DataEntry.
-    """
-
-    def __iter__(self) -> Iterator[DataEntry]:
-        raise NotImplementedError
-
-    def __len__(self):
-        raise NotImplementedError
-
-    def calc_stats(self) -> DatasetStatistics:
-        return calculate_dataset_statistics(self)
-
-
 class Channel(pydantic.BaseModel):
     metadata: Path
     train: Path
@@ -393,14 +374,11 @@ def __call__(self, data: DataEntry) -> DataEntry:
         value = data.get(self.name, None)
         if value is not None:
             value = np.asarray(value, dtype=self.dtype)
-            ddiff = self.req_ndim - value.ndim
 
-            if ddiff == 1:
-                value = np.expand_dims(a=value, axis=0)
-            elif ddiff != 0:
+            if self.req_ndim != value.ndim:
                 raise GluonTSDataError(
-                    f"JSON array has bad shape - expected {self.req_ndim} "
-                    f"dimensions, got {ddiff}"
+                    f"Array '{self.name}' has bad shape - expected "
+                    f"{self.req_ndim} dimensions, got {value.ndim}."
                 )
 
             data[self.name] = value
@@ -410,7 +388,7 @@ def __call__(self, data: DataEntry) -> DataEntry:
             return data
         else:
             raise GluonTSDataError(
-                f"JSON object is missing a required field `{self.name}`"
+                f"Object is missing a required field `{self.name}`"
             )