Upload folder using huggingface_hub

.gitattributes

@@ -76,3 +76,12 @@ data/datasets/lilac/imdb/text/near_dup/data-00000-of-00001.parquet filter=lfs di
 data/datasets/lilac/imdb/text/pii/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
 data/datasets/lilac/imdb/text/spacy_ner/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
 data/datasets/lilac/imdb/text/text_statistics/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/wikitext-2-raw-v1/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/wikitext-2-raw-v1/text/gte-small/hnsw.hnswlib.bin filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/wikitext-2-raw-v1/text/gte-small/hnsw.lookup.pkl filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/wikitext-2-raw-v1/text/gte-small/spans.pkl filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/wikitext-2-raw-v1/text/lang_detection/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/wikitext-2-raw-v1/text/near_dup/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/wikitext-2-raw-v1/text/pii/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/wikitext-2-raw-v1/text/spacy_ner/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/wikitext-2-raw-v1/text/text_statistics/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text

data/datasets/lilac/wikitext-2-raw-v1/config.yml

@@ -0,0 +1,32 @@
+embeddings:
+- embedding: gte-small
+  path: text
+name: wikitext-2-raw-v1
+namespace: local
+settings:
+  preferred_embedding: gte-small
+  ui:
+    media_paths:
+    - text
+signals:
+- path: text
+  signal:
+    signal_name: pii
+- path: text
+  signal:
+    signal_name: near_dup
+- path: text
+  signal:
+    signal_name: lang_detection
+- path: text
+  signal:
+    signal_name: text_statistics
+- path: text
+  signal:
+    signal_name: spacy_ner
+source:
+  config_name: wikitext-2-raw-v1
+  dataset_name: wikitext
+  source_name: huggingface
+tags:
+- machine-learning

data/datasets/lilac/wikitext-2-raw-v1/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d2722d7f4c6b75fdd39f4477d67ed046ab1da08f1fa6d757109d8df77a1a4c2
+size 9171984

data/datasets/lilac/wikitext-2-raw-v1/manifest.json

@@ -0,0 +1,18 @@
+{
+  "files": [
+    "data-00000-of-00001.parquet"
+  ],
+  "data_schema": {
+    "fields": {
+      "text": {
+        "dtype": "string"
+      },
+      "__hfsplit__": {
+        "dtype": "string"
+      },
+      "__rowid__": {
+        "dtype": "string"
+      }
+    }
+  }
+}

data/datasets/lilac/wikitext-2-raw-v1/text/gte-small/hnsw.hnswlib.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9c5174ef8444ec37b0cc39acb01225d77ef8e131d6af4501ac85e904d80a9b2
+size 91233592

data/datasets/lilac/wikitext-2-raw-v1/text/gte-small/hnsw.lookup.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6de9264ae1f50b8dd78b4fec2601cd88625cfe253eaab24542890a870ecd900
+size 1848900

data/datasets/lilac/wikitext-2-raw-v1/text/gte-small/signal_manifest.json

@@ -0,0 +1,35 @@
+{
+  "files": [],
+  "parquet_id": "gte-small(text)",
+  "data_schema": {
+    "fields": {
+      "__rowid__": {
+        "dtype": "string"
+      },
+      "text": {
+        "fields": {
+          "gte-small": {
+            "repeated_field": {
+              "fields": {
+                "embedding": {
+                  "dtype": "embedding"
+                }
+              },
+              "dtype": "string_span"
+            },
+            "signal": {
+              "signal_name": "gte-small"
+            }
+          }
+        }
+      }
+    }
+  },
+  "signal": {
+    "signal_name": "gte-small"
+  },
+  "enriched_path": [
+    "text"
+  ],
+  "vector_store": "hnsw"
+}

data/datasets/lilac/wikitext-2-raw-v1/text/gte-small/spans.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:064d18ca7f0f3b021ebfa93323bcbe9f50271da2ea2234393581a8113322683a
+size 1629720

data/datasets/lilac/wikitext-2-raw-v1/text/lang_detection/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68e4d6290a46a7dac484299be15b8536324cb677ced9721934bd150a277ddaf9
+size 1532348

data/datasets/lilac/wikitext-2-raw-v1/text/lang_detection/signal_manifest.json

@@ -0,0 +1,31 @@
+{
+  "files": [
+    "data-00000-of-00001.parquet"
+  ],
+  "parquet_id": "lang_detection(text)",
+  "data_schema": {
+    "fields": {
+      "__rowid__": {
+        "dtype": "string"
+      },
+      "text": {
+        "fields": {
+          "lang_detection": {
+            "dtype": "string",
+            "signal": {
+              "split_by_paragraph": false,
+              "signal_name": "lang_detection"
+            }
+          }
+        }
+      }
+    }
+  },
+  "signal": {
+    "split_by_paragraph": false,
+    "signal_name": "lang_detection"
+  },
+  "enriched_path": [
+    "text"
+  ]
+}

data/datasets/lilac/wikitext-2-raw-v1/text/near_dup/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f316a8a3f031470cf9fa99475b81a04e962f2da7704d61de8a44089bbf7e59b
+size 1703637

data/datasets/lilac/wikitext-2-raw-v1/text/near_dup/signal_manifest.json

@@ -0,0 +1,36 @@
+{
+  "files": [
+    "data-00000-of-00001.parquet"
+  ],
+  "parquet_id": "near_dup(text)",
+  "data_schema": {
+    "fields": {
+      "__rowid__": {
+        "dtype": "string"
+      },
+      "text": {
+        "fields": {
+          "near_dup": {
+            "fields": {
+              "cluster_id": {
+                "dtype": "uint32",
+                "categorical": true
+              }
+            },
+            "signal": {
+              "threshold": 0.85,
+              "signal_name": "near_dup"
+            }
+          }
+        }
+      }
+    }
+  },
+  "signal": {
+    "threshold": 0.85,
+    "signal_name": "near_dup"
+  },
+  "enriched_path": [
+    "text"
+  ]
+}

data/datasets/lilac/wikitext-2-raw-v1/text/pii/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb257c0e65788f446b23d8f389db299c51a896baf0dcbd4d37114b625868f0ad
+size 1517126

data/datasets/lilac/wikitext-2-raw-v1/text/pii/signal_manifest.json

@@ -0,0 +1,45 @@
+{
+  "files": [
+    "data-00000-of-00001.parquet"
+  ],
+  "parquet_id": "pii(text)",
+  "data_schema": {
+    "fields": {
+      "__rowid__": {
+        "dtype": "string"
+      },
+      "text": {
+        "fields": {
+          "pii": {
+            "fields": {
+              "emails": {
+                "repeated_field": {
+                  "dtype": "string_span"
+                }
+              },
+              "ip_addresses": {
+                "repeated_field": {
+                  "dtype": "string_span"
+                }
+              },
+              "secrets": {
+                "repeated_field": {
+                  "dtype": "string_span"
+                }
+              }
+            },
+            "signal": {
+              "signal_name": "pii"
+            }
+          }
+        }
+      }
+    }
+  },
+  "signal": {
+    "signal_name": "pii"
+  },
+  "enriched_path": [
+    "text"
+  ]
+}

data/datasets/lilac/wikitext-2-raw-v1/text/spacy_ner/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84a98ab86da1750c73ce707415d10487541b2b3265844bc7d2d0a5a2eefe8d15
+size 2718467

data/datasets/lilac/wikitext-2-raw-v1/text/spacy_ner/signal_manifest.json

@@ -0,0 +1,38 @@
+{
+  "files": [
+    "data-00000-of-00001.parquet"
+  ],
+  "parquet_id": "spacy_ner(text)",
+  "data_schema": {
+    "fields": {
+      "__rowid__": {
+        "dtype": "string"
+      },
+      "text": {
+        "fields": {
+          "spacy_ner": {
+            "repeated_field": {
+              "fields": {
+                "label": {
+                  "dtype": "string"
+                }
+              },
+              "dtype": "string_span"
+            },
+            "signal": {
+              "model": "en_core_web_sm",
+              "signal_name": "spacy_ner"
+            }
+          }
+        }
+      }
+    }
+  },
+  "signal": {
+    "model": "en_core_web_sm",
+    "signal_name": "spacy_ner"
+  },
+  "enriched_path": [
+    "text"
+  ]
+}

data/datasets/lilac/wikitext-2-raw-v1/text/text_statistics/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96b6d56208f0c74e7c9bc1d858ebabaae941eaf2c95cdf9207a7bdd1418a59ce
+size 1827010

data/datasets/lilac/wikitext-2-raw-v1/text/text_statistics/signal_manifest.json

@@ -0,0 +1,59 @@
+{
+  "files": [
+    "data-00000-of-00001.parquet"
+  ],
+  "parquet_id": "text_statistics(text)",
+  "data_schema": {
+    "fields": {
+      "__rowid__": {
+        "dtype": "string"
+      },
+      "text": {
+        "fields": {
+          "text_statistics": {
+            "fields": {
+              "num_characters": {
+                "dtype": "int32"
+              },
+              "readability": {
+                "dtype": "float32"
+              },
+              "log(type_token_ratio)": {
+                "dtype": "float32"
+              },
+              "frac_non_ascii": {
+                "dtype": "float32",
+                "bins": [
+                  [
+                    "Low",
+                    null,
+                    0.15
+                  ],
+                  [
+                    "Medium",
+                    0.15,
+                    0.3
+                  ],
+                  [
+                    "High",
+                    0.3,
+                    null
+                  ]
+                ]
+              }
+            },
+            "signal": {
+              "signal_name": "text_statistics"
+            }
+          }
+        }
+      }
+    }
+  },
+  "signal": {
+    "signal_name": "text_statistics"
+  },
+  "enriched_path": [
+    "text"
+  ]
+}