Bugfix setup.py and down/upgrade packages (#16)

* Add "evalem.pipelines" to setup.py

- Also, add **kwargs to QA wrapper post processor

* Update torch to 2.0

* Update transformers lib to 4.28.1

evalem/misc/datasets.py

@@ -36,7 +36,7 @@ def get_squad_v2(
     data = data.shuffle(seed=42) if shuffle else data
     data = data.select(range(nsamples)) if nsamples > 0 else data
 
-    inputs = [dict(question=d["question"], context=d["context"]) for d in data]
+    inputs = [dict(question=d["question"].lstrip(), context=d["context"]) for d in data]
     references = [d["answers"]["text"] for d in data]
 
     inputs, references = zip(*filter(lambda x: len(x[1]) > 0, zip(inputs, references)))

evalem/models/defaults.py

@@ -48,6 +48,7 @@ def __init__(
     def _postprocess_predictions(
         self,
         predictions: Union[dict, List[dict]],
+        **kwargs,
     ) -> Iterable[QAPredictionDTO]:
         """
         This method converts the pipeline's default output format

requirements.txt

@@ -1,13 +1,16 @@
+arrow==1.2.3
 bert-score==0.3.13
-datasets==2.9.0
+datasets==2.3.2
+evaluate==0.2.2
 jury==2.2.3
 loguru==0.6.0
 numpy==1.24.2
 pandas==1.5.3
+pyarrow==9.0.0
 pytest==7.2.1
 sacrebleu==2.3.1
 scikit-learn==1.2.1
 sentencepiece==0.1.97
 seqeval==1.2.2
-torch==1.13.1
-transformers==4.26.1
+torch==2.0.0
+transformers==4.28.1

setup.py

@@ -21,10 +21,11 @@
     python_requires=">=3.8",
     packages=[
         "evalem",
-        "evalem.misc",
-        "evalem.models",
         "evalem.evaluators",
         "evalem.metrics",
+        "evalem.misc",
+        "evalem.models",
+        "evalem.pipelines",
     ],
     install_requires=required,
     classifiers=[