Spaces:
Paused
Paused
File size: 19,015 Bytes
399f3c6 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 |
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# 🚀 GraphRAG GPU检测与测试 - Google Colab版本\n",
"\n",
"本Notebook用于在Google Colab上检测GPU可用性并测试GraphRAG系统的性能。\n",
"\n",
"## 📋 使用步骤\n",
"\n",
"1. **启用GPU**: 运行时 → 更改运行时类型 → 硬件加速器 → GPU (T4)\n",
"2. **运行所有单元格**: 依次执行下面的代码\n",
"3. **查看结果**: 检查GPU加速效果\n",
"\n",
"---"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 1️⃣ GPU环境检测"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# 检测GPU可用性\n",
"import torch\n",
"import subprocess\n",
"import sys\n",
"\n",
"print(\"=\"*60)\n",
"print(\"🔍 GPU环境检测\")\n",
"print(\"=\"*60)\n",
"\n",
"# PyTorch GPU检测\n",
"cuda_available = torch.cuda.is_available()\n",
"print(f\"\\n✅ CUDA可用: {cuda_available}\")\n",
"\n",
"if cuda_available:\n",
" print(f\" GPU数量: {torch.cuda.device_count()}\")\n",
" print(f\" 当前GPU: {torch.cuda.current_device()}\")\n",
" print(f\" GPU名称: {torch.cuda.get_device_name(0)}\")\n",
" print(f\" CUDA版本: {torch.version.cuda}\")\n",
" \n",
" # 显存信息\n",
" total_memory = torch.cuda.get_device_properties(0).total_memory / (1024**3)\n",
" print(f\" 总显存: {total_memory:.2f} GB\")\n",
" \n",
" # nvidia-smi信息\n",
" print(\"\\n📊 nvidia-smi 输出:\")\n",
" print(\"-\"*60)\n",
" !nvidia-smi\n",
"else:\n",
" print(\"\\n⚠️ 警告: 未检测到GPU\")\n",
" print(\" 请检查: 运行时 → 更改运行时类型 → 硬件加速器 → GPU\")\n",
"\n",
"print(\"\\n\" + \"=\"*60)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 2️⃣ GPU性能基准测试"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# GPU vs CPU 性能对比\n",
"import time\n",
"import numpy as np\n",
"\n",
"print(\"=\"*60)\n",
"print(\"⚡ GPU vs CPU 矩阵运算性能测试\")\n",
"print(\"=\"*60)\n",
"\n",
"# 测试参数\n",
"matrix_size = 5000\n",
"\n",
"# CPU测试\n",
"print(f\"\\n🔵 CPU测试 (矩阵大小: {matrix_size}x{matrix_size})\")\n",
"a_cpu = torch.randn(matrix_size, matrix_size)\n",
"b_cpu = torch.randn(matrix_size, matrix_size)\n",
"\n",
"start = time.time()\n",
"c_cpu = torch.mm(a_cpu, b_cpu)\n",
"cpu_time = time.time() - start\n",
"print(f\" CPU时间: {cpu_time:.2f} 秒\")\n",
"\n",
"# GPU测试\n",
"if cuda_available:\n",
" print(f\"\\n🟢 GPU测试 (矩阵大小: {matrix_size}x{matrix_size})\")\n",
" a_gpu = torch.randn(matrix_size, matrix_size).cuda()\n",
" b_gpu = torch.randn(matrix_size, matrix_size).cuda()\n",
" \n",
" # 预热GPU\n",
" _ = torch.mm(a_gpu, b_gpu)\n",
" torch.cuda.synchronize()\n",
" \n",
" start = time.time()\n",
" c_gpu = torch.mm(a_gpu, b_gpu)\n",
" torch.cuda.synchronize()\n",
" gpu_time = time.time() - start\n",
" print(f\" GPU时间: {gpu_time:.2f} 秒\")\n",
" \n",
" speedup = cpu_time / gpu_time\n",
" print(f\"\\n🚀 加速比: {speedup:.2f}x\")\n",
" print(f\" GPU比CPU快 {speedup:.1f} 倍!\")\n",
"else:\n",
" print(\"\\n⚠️ 跳过GPU测试（GPU不可用）\")\n",
"\n",
"print(\"\\n\" + \"=\"*60)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 3️⃣ 安装GraphRAG依赖"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# 克隆项目（如果需要）\n",
"import os\n",
"\n",
"print(\"📦 安装GraphRAG依赖...\\n\")\n",
"\n",
"# 安装核心依赖\n",
"!pip install -q langchain langchain-community langchain-core langgraph\n",
"!pip install -q chromadb sentence-transformers transformers\n",
"!pip install -q tiktoken beautifulsoup4 requests\n",
"!pip install -q tavily-python python-dotenv\n",
"!pip install -q networkx python-louvain\n",
"!pip install -q torch --index-url https://download.pytorch.org/whl/cu118\n",
"\n",
"print(\"\\n✅ 依赖安装完成!\")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 4️⃣ 上传项目文件\n",
"\n",
"**选项A**: 从GitHub克隆\n",
"```python\n",
"!git clone https://github.com/your-repo/adaptive_RAG.git\n",
"%cd adaptive_RAG\n",
"```\n",
"\n",
"**选项B**: 手动上传文件到Colab\n",
"- 使用左侧文件浏览器上传以下核心文件：\n",
" - `config.py`\n",
" - `entity_extractor.py`\n",
" - `knowledge_graph.py`\n",
" - `graph_indexer.py`\n",
" - `graph_retriever.py`\n",
" - `.env` (包含API密钥)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# 创建必要的目录\n",
"!mkdir -p data\n",
"\n",
"# 如果使用选项A，运行下面的命令\n",
"# !git clone YOUR_REPO_URL\n",
"# %cd adaptive_RAG\n",
"\n",
"print(\"✅ 目录准备完成\")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 5️⃣ 配置API密钥"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# 设置API密钥（替换为您的真实密钥）\n",
"import os\n",
"from getpass import getpass\n",
"\n",
"print(\"🔑 配置API密钥\\n\")\n",
"\n",
"# 方式1: 直接设置（不安全，仅用于测试）\n",
"# os.environ['TAVILY_API_KEY'] = 'your_tavily_api_key_here'\n",
"\n",
"# 方式2: 安全输入\n",
"if 'TAVILY_API_KEY' not in os.environ:\n",
" os.environ['TAVILY_API_KEY'] = getpass('输入 TAVILY_API_KEY: ')\n",
" print(\"✅ TAVILY_API_KEY 已设置\")\n",
"else:\n",
" print(\"✅ TAVILY_API_KEY 已存在\")\n",
"\n",
"print(\"\\n注意: GraphRAG在Colab上使用HuggingFace嵌入，不需要NOMIC_API_KEY\")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 6️⃣ 简化版GraphRAG测试代码"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# 简化版GraphRAG核心组件\n",
"# 适用于Colab快速测试，无需完整项目文件\n",
"\n",
"from typing import List, Dict\n",
"import networkx as nx\n",
"from sentence_transformers import SentenceTransformer\n",
"import torch\n",
"\n",
"class SimpleGraphRAG:\n",
" \"\"\"简化版GraphRAG用于GPU性能测试\"\"\"\n",
" \n",
" def __init__(self, use_gpu=True):\n",
" print(\"🚀 初始化SimpleGraphRAG...\")\n",
" \n",
" # 检测设备\n",
" self.device = 'cuda' if use_gpu and torch.cuda.is_available() else 'cpu'\n",
" print(f\" 设备: {self.device.upper()}\")\n",
" \n",
" # 加载嵌入模型\n",
" print(f\" 加载嵌入模型...\")\n",
" self.embedder = SentenceTransformer(\n",
" 'sentence-transformers/all-MiniLM-L6-v2',\n",
" device=self.device\n",
" )\n",
" \n",
" # 知识图谱\n",
" self.graph = nx.Graph()\n",
" self.entities = {}\n",
" \n",
" print(\"✅ 初始化完成!\")\n",
" \n",
" def add_sample_data(self):\n",
" \"\"\"添加示例数据\"\"\"\n",
" print(\"\\n📊 添加示例数据...\")\n",
" \n",
" # 示例实体\n",
" entities = [\n",
" {\"name\": \"LLM\", \"type\": \"CONCEPT\", \"desc\": \"大语言模型\"},\n",
" {\"name\": \"GPT\", \"type\": \"TECHNOLOGY\", \"desc\": \"生成式预训练转换器\"},\n",
" {\"name\": \"Transformer\", \"type\": \"CONCEPT\", \"desc\": \"注意力机制架构\"},\n",
" {\"name\": \"OpenAI\", \"type\": \"ORGANIZATION\", \"desc\": \"人工智能研究公司\"},\n",
" {\"name\": \"Attention\", \"type\": \"CONCEPT\", \"desc\": \"注意力机制\"},\n",
" ]\n",
" \n",
" for entity in entities:\n",
" self.graph.add_node(\n",
" entity[\"name\"],\n",
" type=entity[\"type\"],\n",
" description=entity[\"desc\"]\n",
" )\n",
" self.entities[entity[\"name\"]] = entity\n",
" \n",
" # 示例关系\n",
" relations = [\n",
" (\"GPT\", \"LLM\", \"IS_A\"),\n",
" (\"GPT\", \"Transformer\", \"USES\"),\n",
" (\"Transformer\", \"Attention\", \"CONTAINS\"),\n",
" (\"OpenAI\", \"GPT\", \"DEVELOPS\"),\n",
" ]\n",
" \n",
" for source, target, rel_type in relations:\n",
" self.graph.add_edge(source, target, relation=rel_type)\n",
" \n",
" print(f\" ✅ 添加了 {len(entities)} 个实体\")\n",
" print(f\" ✅ 添加了 {len(relations)} 个关系\")\n",
" \n",
" def test_gpu_embedding(self, texts: List[str]):\n",
" \"\"\"测试GPU嵌入性能\"\"\"\n",
" print(f\"\\n⚡ 测试嵌入性能 ({len(texts)} 个文本)...\")\n",
" \n",
" import time\n",
" \n",
" start = time.time()\n",
" embeddings = self.embedder.encode(\n",
" texts,\n",
" show_progress_bar=True,\n",
" batch_size=32\n",
" )\n",
" elapsed = time.time() - start\n",
" \n",
" print(f\" ✅ 完成! 耗时: {elapsed:.2f}秒\")\n",
" print(f\" 📊 嵌入维度: {embeddings.shape}\")\n",
" print(f\" 🚀 速度: {len(texts)/elapsed:.1f} 文本/秒\")\n",
" \n",
" return embeddings\n",
" \n",
" def query(self, question: str):\n",
" \"\"\"简单查询\"\"\"\n",
" print(f\"\\n🔍 查询: {question}\")\n",
" \n",
" # 简单的关键词匹配\n",
" results = []\n",
" for entity_name in self.entities:\n",
" if entity_name.lower() in question.lower():\n",
" neighbors = list(self.graph.neighbors(entity_name))\n",
" results.append({\n",
" \"entity\": entity_name,\n",
" \"info\": self.entities[entity_name],\n",
" \"neighbors\": neighbors\n",
" })\n",
" \n",
" print(f\"\\n📋 找到 {len(results)} 个相关实体:\")\n",
" for r in results:\n",
" print(f\" • {r['entity']} ({r['info']['type']})\")\n",
" print(f\" 描述: {r['info']['desc']}\")\n",
" print(f\" 关联: {', '.join(r['neighbors'])}\")\n",
" \n",
" return results\n",
"\n",
"print(\"✅ SimpleGraphRAG类定义完成\")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 7️⃣ 运行GPU性能测试"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# 初始化GraphRAG（GPU版本）\n",
"print(\"=\"*60)\n",
"print(\"🎯 GraphRAG GPU性能测试\")\n",
"print(\"=\"*60)\n",
"\n",
"graph_rag = SimpleGraphRAG(use_gpu=True)\n",
"\n",
"# 添加示例数据\n",
"graph_rag.add_sample_data()\n",
"\n",
"# 准备测试文本\n",
"test_texts = [\n",
" \"Large Language Models are transforming AI\",\n",
" \"GPT uses Transformer architecture\",\n",
" \"Attention mechanism is key to modern NLP\",\n",
" \"OpenAI develops cutting-edge AI models\",\n",
"] * 25 # 100个文本\n",
"\n",
"print(f\"\\n准备了 {len(test_texts)} 个测试文本\")\n",
"\n",
"# GPU嵌入测试\n",
"embeddings = graph_rag.test_gpu_embedding(test_texts)\n",
"\n",
"# 测试查询\n",
"graph_rag.query(\"What is GPT?\")\n",
"graph_rag.query(\"Tell me about Transformer\")\n",
"\n",
"print(\"\\n\" + \"=\"*60)\n",
"print(\"✅ GPU性能测试完成!\")\n",
"print(\"=\"*60)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 8️⃣ CPU vs GPU 性能对比"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# CPU vs GPU 嵌入性能对比\n",
"import time\n",
"\n",
"print(\"=\"*60)\n",
"print(\"📊 CPU vs GPU 嵌入性能对比\")\n",
"print(\"=\"*60)\n",
"\n",
"# 准备大量测试文本\n",
"large_test_texts = test_texts * 10 # 1000个文本\n",
"print(f\"\\n测试数据: {len(large_test_texts)} 个文本\\n\")\n",
"\n",
"# CPU测试\n",
"print(\"🔵 CPU测试...\")\n",
"graph_rag_cpu = SimpleGraphRAG(use_gpu=False)\n",
"start = time.time()\n",
"embeddings_cpu = graph_rag_cpu.embedder.encode(\n",
" large_test_texts,\n",
" show_progress_bar=False,\n",
" batch_size=32\n",
")\n",
"cpu_time = time.time() - start\n",
"print(f\" CPU时间: {cpu_time:.2f}秒\")\n",
"print(f\" 速度: {len(large_test_texts)/cpu_time:.1f} 文本/秒\")\n",
"\n",
"# GPU测试\n",
"if cuda_available:\n",
" print(\"\\n🟢 GPU测试...\")\n",
" graph_rag_gpu = SimpleGraphRAG(use_gpu=True)\n",
" start = time.time()\n",
" embeddings_gpu = graph_rag_gpu.embedder.encode(\n",
" large_test_texts,\n",
" show_progress_bar=False,\n",
" batch_size=32\n",
" )\n",
" gpu_time = time.time() - start\n",
" print(f\" GPU时间: {gpu_time:.2f}秒\")\n",
" print(f\" 速度: {len(large_test_texts)/gpu_time:.1f} 文本/秒\")\n",
" \n",
" speedup = cpu_time / gpu_time\n",
" print(f\"\\n🚀 加速比: {speedup:.2f}x\")\n",
" print(f\" GPU比CPU快 {speedup:.1f} 倍!\")\n",
" \n",
" # 节省的时间\n",
" time_saved = cpu_time - gpu_time\n",
" print(f\" ⏱️ 节省时间: {time_saved:.2f}秒\")\n",
"else:\n",
" print(\"\\n⚠️ GPU不可用，跳过GPU测试\")\n",
"\n",
"print(\"\\n\" + \"=\"*60)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 9️⃣ 显存使用监控"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# 监控GPU显存使用\n",
"if cuda_available:\n",
" print(\"=\"*60)\n",
" print(\"💾 GPU显存使用情况\")\n",
" print(\"=\"*60)\n",
" \n",
" allocated = torch.cuda.memory_allocated(0) / (1024**3)\n",
" reserved = torch.cuda.memory_reserved(0) / (1024**3)\n",
" total = torch.cuda.get_device_properties(0).total_memory / (1024**3)\n",
" \n",
" print(f\"\\n已分配: {allocated:.2f} GB\")\n",
" print(f\"已保留: {reserved:.2f} GB\")\n",
" print(f\"总显存: {total:.2f} GB\")\n",
" print(f\"使用率: {(allocated/total)*100:.1f}%\")\n",
" \n",
" print(\"\\n详细信息:\")\n",
" print(torch.cuda.memory_summary(0, abbreviated=True))\n",
" \n",
" print(\"\\n\" + \"=\"*60)\n",
"else:\n",
" print(\"⚠️ GPU不可用\")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 🔟 性能总结报告"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# 生成性能报告\n",
"print(\"=\"*60)\n",
"print(\"📈 GraphRAG GPU性能测试报告\")\n",
"print(\"=\"*60)\n",
"\n",
"print(\"\\n🖥️ 硬件信息:\")\n",
"if cuda_available:\n",
" print(f\" GPU型号: {torch.cuda.get_device_name(0)}\")\n",
" print(f\" 显存: {torch.cuda.get_device_properties(0).total_memory / (1024**3):.2f} GB\")\n",
" print(f\" CUDA版本: {torch.version.cuda}\")\n",
"else:\n",
" print(\" ⚠️ GPU不可用\")\n",
"\n",
"print(f\"\\n PyTorch版本: {torch.__version__}\")\n",
"print(f\" Python版本: {sys.version.split()[0]}\")\n",
"\n",
"print(\"\\n⚡ 性能测试结果:\")\n",
"print(f\" 矩阵运算加速: ~{speedup if cuda_available else 'N/A'}x\")\n",
"print(f\" 文本嵌入加速: ~{cpu_time/gpu_time if cuda_available else 'N/A'}x\")\n",
"\n",
"print(\"\\n💡 建议:\")\n",
"if cuda_available:\n",
" print(\" ✅ GPU运行良好！建议在Colab上运行完整的GraphRAG索引构建\")\n",
" print(\" ✅ 预计索引构建时间将大幅缩短\")\n",
" print(\" ✅ 可以处理更大规模的文档集\")\n",
"else:\n",
" print(\" ⚠️ 建议启用GPU以获得最佳性能\")\n",
" print(\" ⚠️ 路径: 运行时 → 更改运行时类型 → GPU\")\n",
"\n",
"print(\"\\n\" + \"=\"*60)\n",
"print(\"✅ 测试完成!\")\n",
"print(\"=\"*60)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"---\n",
"\n",
"## 📚 下一步\n",
"\n",
"如果GPU测试成功，您可以：\n",
"\n",
"1. **上传完整项目**: 将整个adaptive_RAG项目上传到Colab\n",
"2. **运行GraphRAG索引**: 使用GPU加速构建知识图谱\n",
"3. **保存结果**: 将构建好的图谱下载到本地\n",
"\n",
"### 运行完整GraphRAG的命令:\n",
"\n",
"```python\n",
"# 上传项目后运行\n",
"!python main_graphrag.py\n",
"```\n",
"\n",
"### 预期加速效果:\n",
"\n",
"- 实体提取: 使用GPU的LLM推理会更快\n",
"- 文本嵌入: **5-10倍加速**\n",
"- 向量相似度计算: **10-20倍加速**\n",
"- 总体索引构建时间: **3-5倍加速**\n",
"\n",
"---"
]
}
],
"metadata": {
"accelerator": "GPU",
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"name": "python",
"version": "3.10.0"
}
},
"nbformat": 4,
"nbformat_minor": 0
}
|