Spaces:
Sleeping
Sleeping
File size: 9,115 Bytes
3b69792 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 | """
graph/write_graph.py
Phase 3: Write entities and relationships to Neo4j.
Reads from data/graph/embeddings_<year>.npy
and data/graph/embedding_index_<year>.jsonl
No GPU needed β pure network writes to Neo4j.
IDEMPOTENT: Safe to re-run β relationships only update when ark_id not already present.
Run:
python -m graph.write_graph --year 1900
python -m graph.write_graph --all
"""
from __future__ import annotations
import argparse
import functools
import json
import os
import time
from itertools import combinations
from pathlib import Path
import numpy as np
# Flush all print output immediately so logs update in real time
print = functools.partial(print, flush=True)
# Prevent numpy from spawning extra threads (avoids memory bloat on cluster)
os.environ["OMP_NUM_THREADS"] = "1"
os.environ["MKL_NUM_THREADS"] = "1"
from graph.neo4j_client import get_session, create_schema
OUTPUT_DIR = Path("data/graph")
BATCH_SIZE = 100 # documents per Neo4j transaction
def write_all(year: int = None, suffix: str = None):
print("write_all() started")
if suffix:
file_suffix = suffix
else:
file_suffix = str(year) if year else "all"
emb_file = OUTPUT_DIR / f"embeddings_{file_suffix}.npy"
index_file = OUTPUT_DIR / f"embedding_index_{file_suffix}.jsonl"
print(f"\n{'='*60}")
print(f"BPL Graph β Phase 3: Write to Neo4j")
print(f" Embeddings : {emb_file}")
print(f" Index : {index_file}")
print(f" Batch size : {BATCH_SIZE}")
print(f"{'='*60}\n")
if not emb_file.exists() or not index_file.exists():
raise FileNotFoundError(
f"Missing files. Run Phase 1 and Phase 2 first."
)
create_schema()
# Load embeddings
print("Loading embeddings...")
embeddings = np.load(emb_file)
print(f" Shape: {embeddings.shape}")
# Load index
print("Loading index...")
records = []
with open(index_file, "r", encoding="utf-8") as f:
for line in f:
line = line.strip()
if line:
records.append(json.loads(line))
print(f" Found {len(records)} documents\n")
start_time = time.monotonic()
total_written = 0
for batch_start in range(0, len(records), BATCH_SIZE):
batch = records[batch_start:batch_start + BATCH_SIZE]
batch_end = min(batch_start + BATCH_SIZE, len(records))
# Build all data for the batch upfront
docs_data = []
for record in batch:
entities = record["entities"]
emb_indices = record["emb_indices"]
if not entities:
continue
doc_embs = embeddings[emb_indices]
top_entities = entities[:10]
docs_data.append({
"ark_id": record["ark_id"],
"title": record["title"],
"year": record["year"][0] if record["year"] else None,
"institution": record["institution"],
"source_url": record["source_url"],
"issue_date": record["issue_date"],
"entities": [
{
"name": ent["text"],
"type": ent["type"],
"count": ent["count"],
"embedding": doc_embs[i].tolist(),
}
for i, ent in enumerate(entities)
],
"pairs": [
{
"name1": e1["text"], "type1": e1["type"],
"name2": e2["text"], "type2": e2["type"],
}
for e1, e2 in combinations(top_entities, 2)
],
})
with get_session() as session:
# Single round trip for all document + entity upserts in the batch
session.run(
"""
UNWIND $docs AS doc
MERGE (d:Document {ark_id: doc.ark_id})
SET d.title = doc.title,
d.year = doc.year,
d.institution = doc.institution,
d.source_url = doc.source_url,
d.issue_date = doc.issue_date
WITH d, doc
UNWIND doc.entities AS ent
MERGE (e:Entity {name: ent.name, type: ent.type})
ON CREATE SET e.embedding = ent.embedding
MERGE (d)-[r:MENTIONS]->(e)
ON CREATE SET r.count = ent.count,
r.documents = [doc.ark_id]
ON MATCH SET r.count = CASE
WHEN NOT doc.ark_id IN coalesce(r.documents, [])
THEN coalesce(r.count, 0) + ent.count
ELSE r.count
END,
r.documents = CASE
WHEN NOT doc.ark_id IN coalesce(r.documents, [])
THEN coalesce(r.documents, []) + [doc.ark_id]
ELSE r.documents
END
""",
docs=docs_data,
)
# Co-occurrence β canonicalize pair ordering and make idempotent
all_pairs_with_ark = []
for doc in docs_data:
for p in doc["pairs"]:
all_pairs_with_ark.append({
"ark_id": doc["ark_id"],
"name1": p["name1"],
"type1": p["type1"],
"name2": p["name2"],
"type2": p["type2"],
})
if all_pairs_with_ark:
session.run(
"""
UNWIND $pairs AS pair
MATCH (e1:Entity {name: pair.name1, type: pair.type1})
MATCH (e2:Entity {name: pair.name2, type: pair.type2})
// Canonicalize: ensure consistent ordering (smaller name/type first)
WITH e1, e2, pair,
CASE
WHEN pair.name1 < pair.name2 THEN e1
WHEN pair.name1 > pair.name2 THEN e2
WHEN pair.type1 <= pair.type2 THEN e1
ELSE e2
END AS a,
CASE
WHEN pair.name1 < pair.name2 THEN e2
WHEN pair.name1 > pair.name2 THEN e1
WHEN pair.type1 <= pair.type2 THEN e2
ELSE e1
END AS b
MERGE (a)-[r:CO_OCCURS_WITH]-(b)
ON CREATE SET r.weight = 1,
r.documents = [pair.ark_id]
ON MATCH SET r.weight = CASE
WHEN NOT pair.ark_id IN coalesce(r.documents, [])
THEN coalesce(r.weight, 0) + 1
ELSE r.weight
END,
r.documents = CASE
WHEN NOT pair.ark_id IN coalesce(r.documents, [])
THEN coalesce(r.documents, []) + [pair.ark_id]
ELSE r.documents
END
""",
pairs=all_pairs_with_ark,
)
total_written += len(docs_data)
elapsed = time.monotonic() - start_time
remaining = (elapsed / total_written) * (len(records) - total_written) if total_written else 0
print(
f" [{batch_end}/{len(records)}] "
f"Written {total_written} docs | "
f"ETA: {remaining/60:.1f}min"
)
print(f"\nβ Graph write complete.")
print(f" Documents written : {total_written}")
print(f" Total time : {(time.monotonic()-start_time)/60:.1f} min")
# ββ CLI βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
if __name__ == "__main__":
parser = argparse.ArgumentParser(description="Phase 3: Write graph to Neo4j")
parser.add_argument("--year", type=int, default=None)
parser.add_argument("--all", action="store_true")
parser.add_argument("--suffix", type=str, default=None, help="Explicit file suffix e.g. 'all_gpt' or 'metadata'")
args = parser.parse_args()
write_all(
year = None if (args.all or args.suffix) else (args.year or 1900),
suffix = args.suffix,
) |