adding code

chat-t5.py

@@ -0,0 +1,62 @@
+import os
+import sys
+import pandas as pd
+import numpy as np
+import datasets
+
+_DESCRIPTION = """\
+This script is used to generate a dataset for the chatbot task. The dataset is combination of datasets used to train BlenderBot."""
+
+_VERSION = "1.0.0"
+
+
+class ChatT5Config(datasets.BuilderConfig):
+    """BuilderConfig for TapacoConfig."""
+
+    def __init__(self, **kwargs):
+        super(ChatT5Config, self).__init__(version=datasets.Version(_VERSION, ""), **kwargs),
+
+
+class ChatT5(datasets.GeneratorBasedBuilder):
+
+    BUILDER_CONFIGS = [
+        ChatT5Config(
+            name="chat_t5",
+            description="The dataset is combination of datasets used to train BlenderBot.",
+        )
+    ]
+    BUILDER_CONFIG_CLASS = ChatT5Config
+    DEFAULT_CONFIG_NAME = "chat_t5"
+
+    def _info(self):
+        features = datasets.Features({"texts": datasets.Value("string"), "labels": datasets.Value("string")})
+
+        return datasets.DatasetInfo(description=_DESCRIPTION, features=features, supervised_keys=None)
+
+    def _split_generators(self, dl_manager):
+        """Returns SplitGenerators."""
+        data_dir = os.path.join(os.curdir, "data")
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={"data_dir": os.path.join(data_dir, "train")},
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.TEST,
+                gen_kwargs={"data_dir": os.path.join(data_dir, "test")},
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.VALIDATION,
+                gen_kwargs={"data_dir": os.path.join(data_dir, "validation")},
+            ),
+        ]
+
+    def _generate_examples(self, data_dir):
+        """Yields examples."""
+        data = pd.DataFrame()
+        for file_path in os.listdir(data_dir):
+            data = pd.concat(data, pd.read_parquet(os.path.join(data_dir, file_path)))
+        id_ = -1
+        for _, row in data.iterrows():
+            id_ += 1
+            yield id_, {"texts": row["texts"], "labels": row["labels"]}

data/test/wizard_of_wikipedia_test.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d81abbcbea8aa23558992f9a3d9b3eb70259cf2483e148e942b9e1208652ef0
+size 1035006