Update

Makefile

@@ -27,3 +27,8 @@ pypi:
 
 hf:
 	python $(DIR)/make/hf.py
+
+build:
+	format
+	pypi
+

examples/example9.py

@@ -0,0 +1,14 @@
+from src.unitxt.text_utils import print_dict
+from datasets import load_dataset
+
+dataset = load_dataset('unitxt/data', 'card=wnli_card,template_item=0,num_demos=5,demos_pool_size=100')
+
+print_dict(dataset['train'][0])
+
+import evaluate
+
+metric = evaluate.load('unitxt/metric')
+
+results = metric.compute(predictions=['entailment' for t in dataset['test']], references=dataset['test'])
+
+print_dict(results[0])

prepare/cards/wnli_card.py

@@ -15,7 +15,7 @@
 card = TaskCard(
         loader=LoadHF(path='glue', name='wnli'),
         preprocess_steps=[
-            SplitRandomMix({'train': 'train[:95%]', 'validation': 'train[95%:]', 'test': 'test'}),
+            SplitRandomMix({'train': 'train[95%]', 'validation': 'train[5%]', 'test': 'validation'}),
             MapInstanceValues(mappers={'label': {"0": 'entailment', "1": 'not_entailment'}}),
             AddFields(
             fields={

setup.py

@@ -14,10 +14,10 @@
     description="Load any mixture of text to text data in one line of code",
     long_description=long_description,
     long_description_content_type="text/markdown",
-    url="https://github.ibm.com/IBM-Research-AI/unitext",
+    url="https://github.com/ibm/unitxt",
     packages=setuptools.find_packages('src'),
     package_dir={'': 'src'},
-    package_data={'unitext': ['catalog/*.json']},
+    package_data={'unitxt': ['catalog/*.json']},
     classifiers=[
         "Programming Language :: Python :: 3",
         "Operating System :: OS Independent",

src/unitxt/__init__.py

@@ -2,10 +2,5 @@
 
 register_blocks()
 
-from . import dataset
-
-dataset_url = dataset.__file__
-
-from . import metric
-
-metric_url = metric.__file__
+dataset_url = 'unitxt/data'
+metric_url = 'unitxt/metric'

src/unitxt/catalog.py

@@ -7,7 +7,11 @@ class Catalog(Artifactory):
     name: str = None
     location: str = None
 
-catalog_path = os.path.dirname(__file__) + "/catalog"
+try:
+    import unitxt
+    catalog_path = os.path.dirname(unitxt.__file__) + "/catalog"
+except ImportError:
+    catalog_path = os.path.dirname(__file__) + "/catalog"
 
 class LocalCatalog(Catalog):
     name: str = "local"

src/unitxt/catalog/cards/wnli_card.json

@@ -25,9 +25,9 @@
         {
             "type": "split_random_mix",
             "mix": {
-                "train": "train[:95%]",
-                "validation": "train[95%:]",
-                "test": "test"
+                "train": "train[95%]",
+                "validation": "train[5%]",
+                "test": "validation"
             }
         },
         {

src/unitxt/common.py

@@ -11,6 +11,7 @@
 
 
 class CommonRecipe(Recipe, SourceOperator):
+
     card: TaskCard
     demos_pool_name: str = "demos_pool"
     demos_pool_size: int = None
@@ -37,8 +38,8 @@ def prepare(self):
             steps.append(
                 SliceSplit(
                     slices={
-                        self.demos_pool_name: f"train[:{self.demos_pool_size}]",
-                        "train": f"train[{self.demos_pool_size}:]",
+                        self.demos_pool_name: f"train[:{int(self.demos_pool_size)}]",
+                        "train": f"train[{int(self.demos_pool_size)}:]",
                         "validation": "validation",
                         "test": "test",
                     }
@@ -47,7 +48,7 @@ def prepare(self):
 
         if self.num_demos is not None:
             if self.sampler_type == "random":
-                sampler = RandomSampler(sample_size=self.num_demos)
+                sampler = RandomSampler(sample_size=int(self.num_demos))
 
             steps.append(
                 SpreadSplit(

src/unitxt/dataset.py

@@ -65,10 +65,16 @@
 #############
 
 from .register import register_blocks
-from .artifact import Artifact
+from .artifact import Artifact, fetch_artifact, UnitxtArtifactNotFoundError
 
 import datasets
 
+def fetch(artifact_name):
+    try:
+        artifact, _ = fetch_artifact(artifact_name)
+        return artifact
+    except UnitxtArtifactNotFoundError:
+        return None
 
 def parse(query: str):
     """
@@ -87,7 +93,7 @@ def parse(query: str):
     return result
 
 
-class Unitext(datasets.GeneratorBasedBuilder):
+class Dataset(datasets.GeneratorBasedBuilder):
     """TODO: Short description of my dataset."""
 
     VERSION = datasets.Version("1.1.1")
@@ -97,10 +103,12 @@ class Unitext(datasets.GeneratorBasedBuilder):
     def generators(self):
         register_blocks()
         if not hasattr(self, "_generators") or self._generators is None:
-            args = parse(self.config.name)
-            if "type" not in args:
-                args["type"] = "common_recipe"
-            recipe = Artifact.from_dict(args)
+            recipe = fetch(self.config.name)
+            if recipe is None:
+                args = parse(self.config.name)
+                if "type" not in args:
+                    args["type"] = "common_recipe"
+                recipe = Artifact.from_dict(args)
             self._generators = recipe()
         return self._generators
 

src/unitxt/metric.py

@@ -119,7 +119,7 @@ def prepare(self):
 
 
 # @evaluate.utils.file_utils.add_start_docstrings(_DESCRIPTION, _KWARGS_DESCRIPTION)
-class UnitextMetric(evaluate.Metric):
+class Metric(evaluate.Metric):
     def _info(self):
         return evaluate.MetricInfo(
             description="_DESCRIPTION",