metadata/schema.json

{
  "excluded_columns": [
    "title",
    "abstract",
    "full_text"
  ],
  "format": "parquet",
  "license": "cc-by-4.0",
  "public_columns": [
    {
      "description": "PubMed identifier.",
      "name": "pmid",
      "type": "int64"
    },
    {
      "description": "Embedding vector for the selected dataset config/model.",
      "name": "embedding",
      "type": "fixed_size_list<float>"
    },
    {
      "description": "SHA-256 of the local title+abstract text used for embedding, when available.",
      "name": "text_sha256",
      "type": "string"
    },
    {
      "description": "Publication year, when available.",
      "name": "pub_year",
      "type": "int16"
    },
    {
      "description": "Publication month, when available.",
      "name": "pub_month",
      "type": "int8"
    },
    {
      "description": "Tokenizer count before runtime/model cap, when available.",
      "name": "raw_token_count",
      "type": "int32"
    },
    {
      "description": "Tokenizer count after runtime/model cap, when available.",
      "name": "used_token_count",
      "type": "int32"
    },
    {
      "description": "Whether the embedding input was truncated, when available.",
      "name": "was_truncated",
      "type": "bool"
    }
  ],
  "schema_version": "1.0.0"
}