Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,18 @@ data/datasets/lilac/enron-emails/text/gte-small/spans.pkl filter=lfs diff=lfs me
 data/datasets/lilac/enron-emails/text/lang_detection/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
 data/datasets/lilac/enron-emails/text/near_dup/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
 data/datasets/lilac/enron-emails/text/pii/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/mmlu_professional_law/choices/gte-small/hnsw.hnswlib.bin filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/mmlu_professional_law/choices/gte-small/hnsw.lookup.pkl filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/mmlu_professional_law/choices/gte-small/spans.pkl filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/mmlu_professional_law/choices/lang_detection/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/mmlu_professional_law/choices/near_dup/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/mmlu_professional_law/choices/pii/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/mmlu_professional_law/choices/text_statistics/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/mmlu_professional_law/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/mmlu_professional_law/question/gte-small/hnsw.hnswlib.bin filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/mmlu_professional_law/question/gte-small/hnsw.lookup.pkl filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/mmlu_professional_law/question/gte-small/spans.pkl filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/mmlu_professional_law/question/lang_detection/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/mmlu_professional_law/question/near_dup/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/mmlu_professional_law/question/pii/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/mmlu_professional_law/question/text_statistics/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text

data/datasets/lilac/mmlu_professional_law/choices/gte-small/hnsw.hnswlib.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df9d6e2f5df4b8693544f31ca78a9d1936a4caf47acc2babeb1cb766131b7636
+size 684360968

data/datasets/lilac/mmlu_professional_law/choices/gte-small/hnsw.lookup.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2081ce5d760026fe341e0553cd9e40747ca902e4e7edb851cb747f350f19bb0d
+size 11174465

data/datasets/lilac/mmlu_professional_law/choices/gte-small/signal_manifest.json

@@ -0,0 +1,38 @@
+{
+  "files": [],
+  "parquet_id": "gte-small(choices)",
+  "data_schema": {
+    "fields": {
+      "__rowid__": {
+        "dtype": "string"
+      },
+      "choices": {
+        "repeated_field": {
+          "fields": {
+            "gte-small": {
+              "repeated_field": {
+                "fields": {
+                  "embedding": {
+                    "dtype": "embedding"
+                  }
+                },
+                "dtype": "string_span"
+              },
+              "signal": {
+                "signal_name": "gte-small"
+              }
+            }
+          }
+        }
+      }
+    }
+  },
+  "signal": {
+    "signal_name": "gte-small"
+  },
+  "enriched_path": [
+    "choices",
+    "*"
+  ],
+  "vector_store": "hnsw"
+}

data/datasets/lilac/mmlu_professional_law/choices/gte-small/spans.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02fb1662da21f33ea1429a0f9adf1301185da46f642a722717fe7c523314fa57
+size 11173475

data/datasets/lilac/mmlu_professional_law/choices/lang_detection/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:968d4f87c7b51b995d9e3a96423a06b91984e5ee4a47062cd53fe87cca5cafbe
+size 3469413

data/datasets/lilac/mmlu_professional_law/choices/lang_detection/signal_manifest.json

@@ -0,0 +1,34 @@
+{
+  "files": [
+    "data-00000-of-00001.parquet"
+  ],
+  "parquet_id": "lang_detection(choices)",
+  "data_schema": {
+    "fields": {
+      "__rowid__": {
+        "dtype": "string"
+      },
+      "choices": {
+        "repeated_field": {
+          "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": [
+    "choices",
+    "*"
+  ]
+}

data/datasets/lilac/mmlu_professional_law/choices/near_dup/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:316f3be499fcbb960bc1e83a201838ca0b3047a71d8e1c302fe4e0d833a3bf90
+size 5544176

data/datasets/lilac/mmlu_professional_law/choices/near_dup/signal_manifest.json

@@ -0,0 +1,39 @@
+{
+  "files": [
+    "data-00000-of-00001.parquet"
+  ],
+  "parquet_id": "near_dup(choices)",
+  "data_schema": {
+    "fields": {
+      "__rowid__": {
+        "dtype": "string"
+      },
+      "choices": {
+        "repeated_field": {
+          "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": [
+    "choices",
+    "*"
+  ]
+}

data/datasets/lilac/mmlu_professional_law/choices/pii/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cb41d4e9d0d82bd824abfa733d5be3a599e011098c5d41ebadeb1166a15f722
+size 3393096

data/datasets/lilac/mmlu_professional_law/choices/pii/signal_manifest.json

@@ -0,0 +1,48 @@
+{
+  "files": [
+    "data-00000-of-00001.parquet"
+  ],
+  "parquet_id": "pii(choices)",
+  "data_schema": {
+    "fields": {
+      "__rowid__": {
+        "dtype": "string"
+      },
+      "choices": {
+        "repeated_field": {
+          "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": [
+    "choices",
+    "*"
+  ]
+}

data/datasets/lilac/mmlu_professional_law/choices/text_statistics/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc00a68e0f835e25b214d90e7e48251b39d748f1e836af713440cd0ea2517ead
+size 4634821

data/datasets/lilac/mmlu_professional_law/choices/text_statistics/signal_manifest.json

@@ -0,0 +1,62 @@
+{
+  "files": [
+    "data-00000-of-00001.parquet"
+  ],
+  "parquet_id": "text_statistics(choices)",
+  "data_schema": {
+    "fields": {
+      "__rowid__": {
+        "dtype": "string"
+      },
+      "choices": {
+        "repeated_field": {
+          "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": [
+    "choices",
+    "*"
+  ]
+}

data/datasets/lilac/mmlu_professional_law/config.yml

@@ -0,0 +1,35 @@
+embeddings:
+- embedding: gte-small
+  path: [choices, '*']
+- {embedding: gte-small, path: question}
+name: mmlu_professional_law
+namespace: local
+settings:
+  preferred_embedding: gte-small
+  ui:
+    media_paths:
+    - question
+    - [choices, '*']
+signals:
+- path: question
+  signal: {signal_name: text_statistics}
+- path: question
+  signal: {signal_name: pii}
+- path: question
+  signal: {signal_name: near_dup}
+- path: [choices, '*']
+  signal: {signal_name: text_statistics}
+- path: [choices, '*']
+  signal: {signal_name: spacy_ner}
+- path: question
+  signal: {signal_name: lang_detection}
+- path: [choices, '*']
+  signal: {signal_name: near_dup}
+- path: [choices, '*']
+  signal: {signal_name: pii}
+- path: [choices, '*']
+  signal: {signal_name: lang_detection}
+- path: question
+  signal: {signal_name: spacy_ner}
+source: {config_name: professional_law, dataset_name: cais/mmlu, source_name: huggingface}
+tags: [legal]

data/datasets/lilac/mmlu_professional_law/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65cd2771cf0bb88dbed9ad66ceaff472115f07c9dfea866c7e3f65b68392e745
+size 50699938

data/datasets/lilac/mmlu_professional_law/manifest.json

@@ -0,0 +1,26 @@
+{
+  "files": [
+    "data-00000-of-00001.parquet"
+  ],
+  "data_schema": {
+    "fields": {
+      "question": {
+        "dtype": "string"
+      },
+      "choices": {
+        "repeated_field": {
+          "dtype": "string"
+        }
+      },
+      "answer": {
+        "dtype": "string"
+      },
+      "__hfsplit__": {
+        "dtype": "string"
+      },
+      "__rowid__": {
+        "dtype": "string"
+      }
+    }
+  }
+}

data/datasets/lilac/mmlu_professional_law/question/gte-small/hnsw.hnswlib.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b02300405fccc3011294e15ee869933dd81578173435defbcb19e3b40a65e93
+size 771802212

data/datasets/lilac/mmlu_professional_law/question/gte-small/hnsw.lookup.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f72169740d80ee2b2ea66589d7ebcc58c83381978a4640a27510c416a02bf6c7
+size 11296648

data/datasets/lilac/mmlu_professional_law/question/gte-small/signal_manifest.json

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

data/datasets/lilac/mmlu_professional_law/question/gte-small/spans.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b51cad455e94b167bc9cf130c262ed1b143a8f386c7074a61983e01cd93d277
+size 7911602

data/datasets/lilac/mmlu_professional_law/question/lang_detection/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf6cf8cdc246ce4406599aec8782d3be02f2585f1fbad74173faf0ffcb453a49
+size 3361922

data/datasets/lilac/mmlu_professional_law/question/lang_detection/signal_manifest.json

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

data/datasets/lilac/mmlu_professional_law/question/near_dup/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c4139f699d1a248cf5378c442ef6f17970913394d5d0c79bd7c6e6801ab548a
+size 3697516

data/datasets/lilac/mmlu_professional_law/question/near_dup/signal_manifest.json

@@ -0,0 +1,36 @@
+{
+  "files": [
+    "data-00000-of-00001.parquet"
+  ],
+  "parquet_id": "near_dup(question)",
+  "data_schema": {
+    "fields": {
+      "__rowid__": {
+        "dtype": "string"
+      },
+      "question": {
+        "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": [
+    "question"
+  ]
+}

data/datasets/lilac/mmlu_professional_law/question/pii/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2735c4a2c5d40973652d369140533af74425db6dd753f8a25850d4efeee4928e
+size 3369080

data/datasets/lilac/mmlu_professional_law/question/pii/signal_manifest.json

@@ -0,0 +1,45 @@
+{
+  "files": [
+    "data-00000-of-00001.parquet"
+  ],
+  "parquet_id": "pii(question)",
+  "data_schema": {
+    "fields": {
+      "__rowid__": {
+        "dtype": "string"
+      },
+      "question": {
+        "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": [
+    "question"
+  ]
+}

data/datasets/lilac/mmlu_professional_law/question/text_statistics/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:995b3ac42907ea244d9cb04c68a4715af8ddb7d72dcced056bc58dc9a9f05e7e
+size 4389031

data/datasets/lilac/mmlu_professional_law/question/text_statistics/signal_manifest.json

@@ -0,0 +1,59 @@
+{
+  "files": [
+    "data-00000-of-00001.parquet"
+  ],
+  "parquet_id": "text_statistics(question)",
+  "data_schema": {
+    "fields": {
+      "__rowid__": {
+        "dtype": "string"
+      },
+      "question": {
+        "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": [
+    "question"
+  ]
+}