TBNilles/LEANN
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix mem compare

Browse Source
This commit is contained in:
yichuan520030910320
2025-07-14 22:55:10 -07:00
parent f2feccdbd0
commit e5a9ca8787
6 changed files with 147 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
README.md
View File

@@ -196,6 +196,8 @@ LEANN can create a searchable index of your Chrome browser history, allowing you
<details>
<summary><strong>📋 Click to expand: Command Examples</strong></summary>
Note you need to quit google right now to successfully run this.
```bash
# Use default Chrome profile (auto-finds all profiles) and recommand method to run this because usually default file is enough
python examples/google_history_reader_leann.py

79
examples/compare_faiss_vs_leann.py
View File

@@ -11,6 +11,7 @@ import psutil
import gc
import subprocess
from pathlib import Path
from llama_index.core.node_parser import SentenceSplitter
# Setup logging
logging.basicConfig(stream=sys.stdout, level=logging.INFO)
@@ -110,6 +111,72 @@ def test_leann_hnsw():
tracker.checkpoint("After imports")
from llama_index.core import SimpleDirectoryReader
from leann.api import LeannBuilder, LeannSearcher
# Load and parse documents
documents = SimpleDirectoryReader(
"examples/data",
recursive=True,
encoding="utf-8",
required_exts=[".pdf", ".txt", ".md"],
).load_data()
tracker.checkpoint("After document loading")
# Parse into chunks
node_parser = SentenceSplitter(
chunk_size=256, chunk_overlap=20, separator=" ", paragraph_separator="\n\n"
)
all_texts = []
for doc in documents:
nodes = node_parser.get_nodes_from_documents([doc])
for node in nodes:
all_texts.append(node.get_content())
tracker.checkpoint("After text chunking")
# Build LEANN index
INDEX_DIR = Path("./test_leann_comparison")
INDEX_PATH = str(INDEX_DIR / "comparison.leann")
# Check if index already exists
if os.path.exists(INDEX_PATH + ".meta.json"):
print("Loading existing LEANN HNSW index...")
tracker.checkpoint("After loading existing index")
else:
print("Building new LEANN HNSW index...")
# Clean up previous index
import shutil
if INDEX_DIR.exists():
shutil.rmtree(INDEX_DIR)
builder = LeannBuilder(
backend_name="hnsw",
embedding_model="facebook/contriever",
graph_degree=32,
complexity=64,
is_compact=True,
is_recompute=True,
num_threads=1,
)
tracker.checkpoint("After builder setup")
print("Building LEANN HNSW index...")
for chunk_text in all_texts:
builder.add_text(chunk_text)
builder.build_index(INDEX_PATH)
del builder
gc.collect()
tracker.checkpoint("After index building")
# Find existing LEANN index
index_paths = [
"./test_leann_comparison/comparison.leann",
@@ -124,10 +191,18 @@ def test_leann_hnsw():
print("❌ LEANN index not found. Please build it first")
return {"peak_memory": float("inf"), "error": "Index not found"}
# Measure runtime memory overhead
print("\nMeasuring runtime memory overhead...")
runtime_start_mem = get_memory_usage()
print(f"Before load memory: {runtime_start_mem:.1f} MB")
tracker.checkpoint("Before load memory")
# Load searcher
searcher = LeannSearcher(index_path)
tracker.checkpoint("After searcher loading")
print("Running search queries...")
queries = [
"什么是盘古大模型以及盘古开发过程中遇到了什么阴暗面，任务令一般在什么城市颁发",
@@ -143,7 +218,11 @@ def test_leann_hnsw():
print(f"Query {i + 1} time: {query_time:.3f}s")
tracker.checkpoint(f"After query {i + 1}")
runtime_end_mem = get_memory_usage()
runtime_overhead = runtime_end_mem - runtime_start_mem
peak_memory = tracker.summary()
print(f"Runtime Memory Overhead: {runtime_overhead:.1f} MB")
# Get storage size before cleanup
storage_size = 0

70
examples/faiss_only.py
View File

@@ -5,6 +5,7 @@ import sys
import time
import psutil
import gc
import os
def get_memory_usage():
@@ -44,7 +45,10 @@ def main():
VectorStoreIndex,
StorageContext,
Settings,
node_parser,
Document,
)
from llama_index.core.node_parser import SentenceSplitter
from llama_index.vector_stores.faiss import FaissVectorStore
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
@@ -68,15 +72,63 @@ def main():
).load_data()
tracker.checkpoint("After document loading")
print("Building Faiss HNSW index...")
vector_store = FaissVectorStore(faiss_index=faiss_index)
storage_context = StorageContext.from_defaults(vector_store=vector_store)
index = VectorStoreIndex.from_documents(documents, storage_context=storage_context)
tracker.checkpoint("After index building")
# Parse into chunks using the same splitter as LEANN
node_parser = SentenceSplitter(
chunk_size=256, chunk_overlap=20, separator=" ", paragraph_separator="\n\n"
)
index.storage_context.persist("./storage_faiss")
tracker.checkpoint("After index saving")
all_texts = []
for doc in documents:
nodes = node_parser.get_nodes_from_documents([doc])
for node in nodes:
all_texts.append(node.get_content())
tracker.checkpoint("After text chunking")
# Check if index already exists and try to load it
index_loaded = False
if os.path.exists("./storage_faiss"):
print("Loading existing Faiss HNSW index...")
try:
# Use the correct Faiss loading pattern from the example
vector_store = FaissVectorStore.from_persist_dir("./storage_faiss")
storage_context = StorageContext.from_defaults(
vector_store=vector_store, persist_dir="./storage_faiss"
)
from llama_index.core import load_index_from_storage
index = load_index_from_storage(storage_context=storage_context)
print(f"Index loaded from ./storage_faiss")
tracker.checkpoint("After loading existing index")
index_loaded = True
except Exception as e:
print(f"Failed to load existing index: {e}")
print("Cleaning up corrupted index and building new one...")
# Clean up corrupted index
import shutil
if os.path.exists("./storage_faiss"):
shutil.rmtree("./storage_faiss")
if not index_loaded:
print("Building new Faiss HNSW index...")
# Use the correct Faiss building pattern from the example
vector_store = FaissVectorStore(faiss_index=faiss_index)
storage_context = StorageContext.from_defaults(vector_store=vector_store)
index = VectorStoreIndex.from_documents(
documents, storage_context=storage_context
)
tracker.checkpoint("After index building")
# Save index to disk using the correct pattern
index.storage_context.persist(persist_dir="./storage_faiss")
tracker.checkpoint("After index saving")
# Measure runtime memory overhead
print("\nMeasuring runtime memory overhead...")
runtime_start_mem = get_memory_usage()
print(f"Before load memory: {runtime_start_mem:.1f} MB")
tracker.checkpoint("Before load memory")
query_engine = index.as_query_engine(similarity_top_k=20)
queries = [
"什么是盘古大模型以及盘古开发过程中遇到了什么阴暗面，任务令一般在什么城市颁发",
@@ -91,8 +143,12 @@ def main():
print(f"Query {i + 1} time: {query_time:.3f}s")
tracker.checkpoint(f"After query {i + 1}")
runtime_end_mem = get_memory_usage()
runtime_overhead = runtime_end_mem - runtime_start_mem
peak_memory = tracker.summary()
print(f"Peak Memory: {peak_memory:.1f} MB")
print(f"Runtime Memory Overhead: {runtime_overhead:.1f} MB")
if __name__ == "__main__":

4
examples/google_history_reader_leann.py
View File

@@ -199,7 +199,7 @@ async def query_leann_index(index_path: str, query: str):
query: The query string
"""
print(f"\n[PHASE 2] Starting Leann chat session...")
chat = LeannChat(index_path=index_path, llm_config={"type": "hf", "model": "Qwen/Qwen3-0.6B"})
chat = LeannChat(index_path=index_path)
print(f"You: {query}")
chat_response = chat.ask(
@@ -270,8 +270,6 @@ async def main():
# Example queries
queries = [
"What websites did I visit about machine learning?",
"Show me my recent shopping history",
"What news sites did I visit this week?",
"Find my search history about programming"
]

2
packages/leann-backend-hnsw/leann_backend_hnsw/hnsw_embedding_server.py
View File

@@ -680,7 +680,7 @@ def create_hnsw_embedding_server(
f"ERROR: Passage ID {nid} not found in passages dict"
)
print(
f"ERROR: Available passage IDs: {list(passages.keys())[:10]}..."
f"ERROR: Available passage IDs: {list(passages.keys())}..."
)
raise RuntimeError(
f"FATAL: Passage with ID {nid} not found"

2
packages/leann-core/src/leann/api.py
View File

@@ -290,7 +290,7 @@ class LeannSearcher:
)
)
print(
f" {i + 1}. passage_id='{string_id}' -> SUCCESS: {passage_data['text'][:60]}..."
f" {i + 1}. passage_id='{string_id}' -> SUCCESS: {passage_data['text']}..."
)
except KeyError:
print(