Spaces:
Sleeping
Sleeping
File size: 54,413 Bytes
ed4d993 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 |
{
"cells": [
{
"cell_type": "markdown",
"id": "f1571abe-8e84-44d1-b222-e4121fdbb4be",
"metadata": {},
"source": [
"# Advanced RAG Eval\n",
"\n",
"The cookbook walks through the process of running eval(s) on advanced RAG. \n",
"\n",
"This can be very useful to determine the best RAG approach for your application."
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "0d8415ee-709c-407f-9ac2-f03a9d697aaf",
"metadata": {},
"outputs": [],
"source": [
"! pip install -U langchain openai chromadb langchain-experimental # (newest versions required for multi-modal)"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "191f8465-fd6b-4017-8f0e-d284971b45ae",
"metadata": {},
"outputs": [],
"source": [
"# lock to 0.10.19 due to a persistent bug in more recent versions\n",
"! pip install \"unstructured[all-docs]==0.10.19\" pillow pydantic lxml pillow matplotlib tiktoken open_clip_torch torch"
]
},
{
"cell_type": "markdown",
"id": "45949db5-d9b6-44a9-85f8-96d83a288616",
"metadata": {},
"source": [
"## Data Loading\n",
"\n",
"Let's look at an [example whitepaper](https://sgp.fas.org/crs/misc/IF10244.pdf) that provides a mixture of tables, text, and images about Wildfires in the US."
]
},
{
"cell_type": "markdown",
"id": "961a42b9-c16b-472e-b994-3c3f73afbbcb",
"metadata": {},
"source": [
"### Option 1: Load text"
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "12f24fc0-c176-4201-982b-8a84b278ff1b",
"metadata": {},
"outputs": [],
"source": [
"# Path\n",
"path = \"/Users/rlm/Desktop/cpi/\"\n",
"\n",
"# Load\n",
"from langchain_community.document_loaders import PyPDFLoader\n",
"\n",
"loader = PyPDFLoader(path + \"cpi.pdf\")\n",
"pdf_pages = loader.load()\n",
"\n",
"# Split\n",
"from langchain_text_splitters import RecursiveCharacterTextSplitter\n",
"\n",
"text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=0)\n",
"all_splits_pypdf = text_splitter.split_documents(pdf_pages)\n",
"all_splits_pypdf_texts = [d.page_content for d in all_splits_pypdf]"
]
},
{
"cell_type": "markdown",
"id": "92fc1870-1836-4bc3-945a-78e2c16ad823",
"metadata": {},
"source": [
"### Option 2: Load text, tables, images \n",
"\n"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "7d863632-f894-4471-b4cc-a1d9aa834d29",
"metadata": {},
"outputs": [],
"source": [
"from unstructured.partition.pdf import partition_pdf\n",
"\n",
"# Extract images, tables, and chunk text\n",
"raw_pdf_elements = partition_pdf(\n",
" filename=path + \"cpi.pdf\",\n",
" extract_images_in_pdf=True,\n",
" infer_table_structure=True,\n",
" chunking_strategy=\"by_title\",\n",
" max_characters=4000,\n",
" new_after_n_chars=3800,\n",
" combine_text_under_n_chars=2000,\n",
" image_output_dir_path=path,\n",
")\n",
"\n",
"# Categorize by type\n",
"tables = []\n",
"texts = []\n",
"for element in raw_pdf_elements:\n",
" if \"unstructured.documents.elements.Table\" in str(type(element)):\n",
" tables.append(str(element))\n",
" elif \"unstructured.documents.elements.CompositeElement\" in str(type(element)):\n",
" texts.append(str(element))"
]
},
{
"cell_type": "markdown",
"id": "65f399c5-bd91-4ed4-89c6-c89d2e17466e",
"metadata": {},
"source": [
"## Store\n",
"\n",
"### Option 1: Embed, store text chunks"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "7d7ecdb2-0bb5-46b8-bcff-af8fc272e88e",
"metadata": {},
"outputs": [],
"source": [
"from langchain_community.vectorstores import Chroma\n",
"from langchain_openai import OpenAIEmbeddings\n",
"\n",
"baseline = Chroma.from_texts(\n",
" texts=all_splits_pypdf_texts,\n",
" collection_name=\"baseline\",\n",
" embedding=OpenAIEmbeddings(),\n",
")\n",
"retriever_baseline = baseline.as_retriever()"
]
},
{
"cell_type": "markdown",
"id": "6a0eaefe-5e4b-4853-94c7-5abd6f7fbeac",
"metadata": {},
"source": [
"### Option 2: Multi-vector retriever\n",
"\n",
"#### Text Summary"
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "3d4b4b43-e96e-48ab-899d-c39d0430562e",
"metadata": {},
"outputs": [],
"source": [
"from langchain_core.output_parsers import StrOutputParser\n",
"from langchain_core.prompts import ChatPromptTemplate\n",
"from langchain_openai import ChatOpenAI\n",
"\n",
"# Prompt\n",
"prompt_text = \"\"\"You are an assistant tasked with summarizing tables and text for retrieval. \\\n",
"These summaries will be embedded and used to retrieve the raw text or table elements. \\\n",
"Give a concise summary of the table or text that is well optimized for retrieval. Table or text: {element} \"\"\"\n",
"prompt = ChatPromptTemplate.from_template(prompt_text)\n",
"\n",
"# Text summary chain\n",
"model = ChatOpenAI(temperature=0, model=\"gpt-4\")\n",
"summarize_chain = {\"element\": lambda x: x} | prompt | model | StrOutputParser()\n",
"\n",
"# Apply to text\n",
"text_summaries = summarize_chain.batch(texts, {\"max_concurrency\": 5})\n",
"\n",
"# Apply to tables\n",
"table_summaries = summarize_chain.batch(tables, {\"max_concurrency\": 5})"
]
},
{
"cell_type": "markdown",
"id": "bdb5c903-5b4c-4ddb-8f9a-e20f5155dfb9",
"metadata": {},
"source": [
"#### Image Summary"
]
},
{
"cell_type": "code",
"execution_count": 9,
"id": "4570578c-531b-422c-bedd-cc519d9b7887",
"metadata": {},
"outputs": [],
"source": [
"# Image summary chain\n",
"import base64\n",
"import io\n",
"import os\n",
"from io import BytesIO\n",
"\n",
"from langchain_core.messages import HumanMessage\n",
"from PIL import Image\n",
"\n",
"\n",
"def encode_image(image_path):\n",
" \"\"\"Getting the base64 string\"\"\"\n",
" with open(image_path, \"rb\") as image_file:\n",
" return base64.b64encode(image_file.read()).decode(\"utf-8\")\n",
"\n",
"\n",
"def image_summarize(img_base64, prompt):\n",
" \"\"\"Image summary\"\"\"\n",
" chat = ChatOpenAI(model=\"gpt-4-vision-preview\", max_tokens=1024)\n",
"\n",
" msg = chat.invoke(\n",
" [\n",
" HumanMessage(\n",
" content=[\n",
" {\"type\": \"text\", \"text\": prompt},\n",
" {\n",
" \"type\": \"image_url\",\n",
" \"image_url\": {\"url\": f\"data:image/jpeg;base64,{img_base64}\"},\n",
" },\n",
" ]\n",
" )\n",
" ]\n",
" )\n",
" return msg.content\n",
"\n",
"\n",
"# Store base64 encoded images\n",
"img_base64_list = []\n",
"\n",
"# Store image summaries\n",
"image_summaries = []\n",
"\n",
"# Prompt\n",
"prompt = \"\"\"You are an assistant tasked with summarizing images for retrieval. \\\n",
"These summaries will be embedded and used to retrieve the raw image. \\\n",
"Give a concise summary of the image that is well optimized for retrieval.\"\"\"\n",
"\n",
"# Apply to images\n",
"for img_file in sorted(os.listdir(path)):\n",
" if img_file.endswith(\".jpg\"):\n",
" img_path = os.path.join(path, img_file)\n",
" base64_image = encode_image(img_path)\n",
" img_base64_list.append(base64_image)\n",
" image_summaries.append(image_summarize(base64_image, prompt))"
]
},
{
"cell_type": "markdown",
"id": "87e03f07-4c82-4743-a3c6-d0597fb55107",
"metadata": {},
"source": [
"### Option 2a: Multi-vector retriever w/ raw images\n",
"\n",
"* Return images to LLM for answer synthesis"
]
},
{
"cell_type": "code",
"execution_count": 11,
"id": "6bf8a07d-203f-4397-8b0b-a84ec4d0adab",
"metadata": {},
"outputs": [],
"source": [
"import uuid\n",
"from base64 import b64decode\n",
"\n",
"from langchain.retrievers.multi_vector import MultiVectorRetriever\n",
"from langchain.storage import InMemoryStore\n",
"from langchain_core.documents import Document\n",
"\n",
"\n",
"def create_multi_vector_retriever(\n",
" vectorstore, text_summaries, texts, table_summaries, tables, image_summaries, images\n",
"):\n",
" # Initialize the storage layer\n",
" store = InMemoryStore()\n",
" id_key = \"doc_id\"\n",
"\n",
" # Create the multi-vector retriever\n",
" retriever = MultiVectorRetriever(\n",
" vectorstore=vectorstore,\n",
" docstore=store,\n",
" id_key=id_key,\n",
" )\n",
"\n",
" # Helper function to add documents to the vectorstore and docstore\n",
" def add_documents(retriever, doc_summaries, doc_contents):\n",
" doc_ids = [str(uuid.uuid4()) for _ in doc_contents]\n",
" summary_docs = [\n",
" Document(page_content=s, metadata={id_key: doc_ids[i]})\n",
" for i, s in enumerate(doc_summaries)\n",
" ]\n",
" retriever.vectorstore.add_documents(summary_docs)\n",
" retriever.docstore.mset(list(zip(doc_ids, doc_contents)))\n",
"\n",
" # Add texts, tables, and images\n",
" # Check that text_summaries is not empty before adding\n",
" if text_summaries:\n",
" add_documents(retriever, text_summaries, texts)\n",
" # Check that table_summaries is not empty before adding\n",
" if table_summaries:\n",
" add_documents(retriever, table_summaries, tables)\n",
" # Check that image_summaries is not empty before adding\n",
" if image_summaries:\n",
" add_documents(retriever, image_summaries, images)\n",
"\n",
" return retriever\n",
"\n",
"\n",
"# The vectorstore to use to index the summaries\n",
"multi_vector_img = Chroma(\n",
" collection_name=\"multi_vector_img\", embedding_function=OpenAIEmbeddings()\n",
")\n",
"\n",
"# Create retriever\n",
"retriever_multi_vector_img = create_multi_vector_retriever(\n",
" multi_vector_img,\n",
" text_summaries,\n",
" texts,\n",
" table_summaries,\n",
" tables,\n",
" image_summaries,\n",
" img_base64_list,\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 32,
"id": "84d5b4ea-51b8-49cf-8ad1-db8f7a50e3cf",
"metadata": {},
"outputs": [],
"source": [
"# Testing on retrieval\n",
"query = \"What percentage of CPI is dedicated to Housing, and how does it compare to the combined percentage of Medical Care, Apparel, and Other Goods and Services?\"\n",
"suffix_for_images = \" Include any pie charts, graphs, or tables.\"\n",
"docs = retriever_multi_vector_img.invoke(query + suffix_for_images)"
]
},
{
"cell_type": "code",
"execution_count": 19,
"id": "8db51ac6-ec0c-4c5d-a9a7-0316035e139d",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<img src=\"data:image/jpeg;base64,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\" />"
],
"text/plain": [
"<IPython.core.display.HTML object>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"from IPython.display import HTML, display\n",
"\n",
"\n",
"def plt_img_base64(img_base64):\n",
" # Create an HTML img tag with the base64 string as the source\n",
" image_html = f'<img src=\"data:image/jpeg;base64,{img_base64}\" />'\n",
"\n",
" # Display the image by rendering the HTML\n",
" display(HTML(image_html))\n",
"\n",
"\n",
"plt_img_base64(docs[1])"
]
},
{
"cell_type": "markdown",
"id": "48b268ec-db04-4107-9833-ea1615f6dbd1",
"metadata": {},
"source": [
"### Option 2b: Multi-vector retriever w/ image summaries\n",
"\n",
"* Return text summary of images to LLM for answer synthesis"
]
},
{
"cell_type": "code",
"execution_count": 20,
"id": "ae57c804-0dd1-4806-b761-a913efc4f173",
"metadata": {},
"outputs": [],
"source": [
"# The vectorstore to use to index the summaries\n",
"multi_vector_text = Chroma(\n",
" collection_name=\"multi_vector_text\", embedding_function=OpenAIEmbeddings()\n",
")\n",
"\n",
"# Create retriever\n",
"retriever_multi_vector_img_summary = create_multi_vector_retriever(\n",
" multi_vector_text,\n",
" text_summaries,\n",
" texts,\n",
" table_summaries,\n",
" tables,\n",
" image_summaries,\n",
" image_summaries,\n",
")"
]
},
{
"cell_type": "markdown",
"id": "580a3d55-5025-472d-9c14-cec7a384379f",
"metadata": {},
"source": [
"### Option 3: Multi-modal embeddings"
]
},
{
"cell_type": "code",
"execution_count": 22,
"id": "8dbed5dc-f7a3-4324-9436-1c3ebc24f9fd",
"metadata": {},
"outputs": [],
"source": [
"from langchain_experimental.open_clip import OpenCLIPEmbeddings\n",
"\n",
"# Create chroma w/ multi-modal embeddings\n",
"multimodal_embd = Chroma(\n",
" collection_name=\"multimodal_embd\", embedding_function=OpenCLIPEmbeddings()\n",
")\n",
"\n",
"# Get image URIs\n",
"image_uris = sorted(\n",
" [\n",
" os.path.join(path, image_name)\n",
" for image_name in os.listdir(path)\n",
" if image_name.endswith(\".jpg\")\n",
" ]\n",
")\n",
"\n",
"# Add images and documents\n",
"if image_uris:\n",
" multimodal_embd.add_images(uris=image_uris)\n",
"if texts:\n",
" multimodal_embd.add_texts(texts=texts)\n",
"if tables:\n",
" multimodal_embd.add_texts(texts=tables)\n",
"\n",
"# Make retriever\n",
"retriever_multimodal_embd = multimodal_embd.as_retriever()"
]
},
{
"cell_type": "markdown",
"id": "647abb6c-adf3-4d29-acd2-885c4925fa12",
"metadata": {},
"source": [
"## RAG\n",
"\n",
"### Text Pipeline"
]
},
{
"cell_type": "code",
"execution_count": 23,
"id": "73440ca0-4330-4c16-9d9d-6f27c249ae58",
"metadata": {},
"outputs": [],
"source": [
"from operator import itemgetter\n",
"\n",
"from langchain_core.runnables import RunnablePassthrough\n",
"\n",
"# Prompt\n",
"template = \"\"\"Answer the question based only on the following context, which can include text and tables:\n",
"{context}\n",
"Question: {question}\n",
"\"\"\"\n",
"rag_prompt_text = ChatPromptTemplate.from_template(template)\n",
"\n",
"\n",
"# Build\n",
"def text_rag_chain(retriever):\n",
" \"\"\"RAG chain\"\"\"\n",
"\n",
" # LLM\n",
" model = ChatOpenAI(temperature=0, model=\"gpt-4\")\n",
"\n",
" # RAG pipeline\n",
" chain = (\n",
" {\"context\": retriever, \"question\": RunnablePassthrough()}\n",
" | rag_prompt_text\n",
" | model\n",
" | StrOutputParser()\n",
" )\n",
"\n",
" return chain"
]
},
{
"cell_type": "markdown",
"id": "14b358ad-42fd-4c6d-b2c0-215dba135707",
"metadata": {},
"source": [
"### Multi-modal Pipeline"
]
},
{
"cell_type": "code",
"execution_count": 24,
"id": "ae89ce84-283e-4634-8169-9ff16f152807",
"metadata": {},
"outputs": [],
"source": [
"import re\n",
"\n",
"from langchain_core.documents import Document\n",
"from langchain_core.runnables import RunnableLambda\n",
"\n",
"\n",
"def looks_like_base64(sb):\n",
" \"\"\"Check if the string looks like base64.\"\"\"\n",
" return re.match(\"^[A-Za-z0-9+/]+[=]{0,2}$\", sb) is not None\n",
"\n",
"\n",
"def is_image_data(b64data):\n",
" \"\"\"Check if the base64 data is an image by looking at the start of the data.\"\"\"\n",
" image_signatures = {\n",
" b\"\\xff\\xd8\\xff\": \"jpg\",\n",
" b\"\\x89\\x50\\x4e\\x47\\x0d\\x0a\\x1a\\x0a\": \"png\",\n",
" b\"\\x47\\x49\\x46\\x38\": \"gif\",\n",
" b\"\\x52\\x49\\x46\\x46\": \"webp\",\n",
" }\n",
" try:\n",
" header = base64.b64decode(b64data)[:8] # Decode and get the first 8 bytes\n",
" for sig, format in image_signatures.items():\n",
" if header.startswith(sig):\n",
" return True\n",
" return False\n",
" except Exception:\n",
" return False\n",
"\n",
"\n",
"def split_image_text_types(docs):\n",
" \"\"\"Split base64-encoded images and texts.\"\"\"\n",
" b64_images = []\n",
" texts = []\n",
" for doc in docs:\n",
" # Check if the document is of type Document and extract page_content if so\n",
" if isinstance(doc, Document):\n",
" doc = doc.page_content\n",
" if looks_like_base64(doc) and is_image_data(doc):\n",
" b64_images.append(doc)\n",
" else:\n",
" texts.append(doc)\n",
" return {\"images\": b64_images, \"texts\": texts}\n",
"\n",
"\n",
"def img_prompt_func(data_dict):\n",
" # Joining the context texts into a single string\n",
" formatted_texts = \"\\n\".join(data_dict[\"context\"][\"texts\"])\n",
" messages = []\n",
"\n",
" # Adding image(s) to the messages if present\n",
" if data_dict[\"context\"][\"images\"]:\n",
" image_message = {\n",
" \"type\": \"image_url\",\n",
" \"image_url\": {\n",
" \"url\": f\"data:image/jpeg;base64,{data_dict['context']['images'][0]}\"\n",
" },\n",
" }\n",
" messages.append(image_message)\n",
"\n",
" # Adding the text message for analysis\n",
" text_message = {\n",
" \"type\": \"text\",\n",
" \"text\": (\n",
" \"Answer the question based only on the provided context, which can include text, tables, and image(s). \"\n",
" \"If an image is provided, analyze it carefully to help answer the question.\\n\"\n",
" f\"User-provided question / keywords: {data_dict['question']}\\n\\n\"\n",
" \"Text and / or tables:\\n\"\n",
" f\"{formatted_texts}\"\n",
" ),\n",
" }\n",
" messages.append(text_message)\n",
" return [HumanMessage(content=messages)]\n",
"\n",
"\n",
"def multi_modal_rag_chain(retriever):\n",
" \"\"\"Multi-modal RAG chain\"\"\"\n",
"\n",
" # Multi-modal LLM\n",
" model = ChatOpenAI(temperature=0, model=\"gpt-4-vision-preview\", max_tokens=1024)\n",
"\n",
" # RAG pipeline\n",
" chain = (\n",
" {\n",
" \"context\": retriever | RunnableLambda(split_image_text_types),\n",
" \"question\": RunnablePassthrough(),\n",
" }\n",
" | RunnableLambda(img_prompt_func)\n",
" | model\n",
" | StrOutputParser()\n",
" )\n",
"\n",
" return chain"
]
},
{
"cell_type": "markdown",
"id": "5e8b0e26-bb7e-420a-a7bd-8512b7eef92f",
"metadata": {},
"source": [
"### Build RAG Pipelines"
]
},
{
"cell_type": "code",
"execution_count": 25,
"id": "4f1ec8a9-f0fe-4f08-928f-23504803897c",
"metadata": {},
"outputs": [],
"source": [
"# RAG chains\n",
"chain_baseline = text_rag_chain(retriever_baseline)\n",
"chain_mv_text = text_rag_chain(retriever_multi_vector_img_summary)\n",
"\n",
"# Multi-modal RAG chains\n",
"chain_multimodal_mv_img = multi_modal_rag_chain(retriever_multi_vector_img)\n",
"chain_multimodal_embd = multi_modal_rag_chain(retriever_multimodal_embd)"
]
},
{
"cell_type": "markdown",
"id": "448d943c-a1b1-4300-9197-891a03232ee4",
"metadata": {},
"source": [
"## Eval set"
]
},
{
"cell_type": "code",
"execution_count": 34,
"id": "9aabf72f-26be-437f-9372-b06dc2509235",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>Question</th>\n",
" <th>Answer</th>\n",
" <th>Source</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>What percentage of CPI is dedicated to Housing?</td>\n",
" <td>Housing occupies 42% of CPI.</td>\n",
" <td>Figure 1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>Medical Care and Transportation account for wh...</td>\n",
" <td>Transportation accounts for 18% of CPI. Medica...</td>\n",
" <td>Figure 1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>Based on the CPI Owners' Equivalent Rent and t...</td>\n",
" <td>The FHFA Purchase Only Price Index appears to ...</td>\n",
" <td>Figure 2</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" Question \\\n",
"0 What percentage of CPI is dedicated to Housing? \n",
"1 Medical Care and Transportation account for wh... \n",
"2 Based on the CPI Owners' Equivalent Rent and t... \n",
"\n",
" Answer Source \n",
"0 Housing occupies 42% of CPI. Figure 1 \n",
"1 Transportation accounts for 18% of CPI. Medica... Figure 1 \n",
"2 The FHFA Purchase Only Price Index appears to ... Figure 2 "
]
},
"execution_count": 34,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# Read\n",
"import pandas as pd\n",
"\n",
"eval_set = pd.read_csv(path + \"cpi_eval.csv\")\n",
"eval_set.head(3)"
]
},
{
"cell_type": "code",
"execution_count": 35,
"id": "7fdeb77a-e185-47d2-a93f-822f1fc810a2",
"metadata": {},
"outputs": [],
"source": [
"from langsmith import Client\n",
"\n",
"# Dataset\n",
"client = Client()\n",
"dataset_name = f\"CPI Eval {str(uuid.uuid4())}\"\n",
"dataset = client.create_dataset(dataset_name=dataset_name)\n",
"\n",
"# Populate dataset\n",
"for _, row in eval_set.iterrows():\n",
" # Get Q, A\n",
" q = row[\"Question\"]\n",
" a = row[\"Answer\"]\n",
" # Use the values in your function\n",
" client.create_example(\n",
" inputs={\"question\": q}, outputs={\"answer\": a}, dataset_id=dataset.id\n",
" )"
]
},
{
"cell_type": "code",
"execution_count": 36,
"id": "3c4faf4b-f29f-4a42-9cf2-bfbb5158ab59",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"View the evaluation results for project 'CPI Eval 9648e7fe-5ae2-469f-8701-33c63212d126-baseline' at:\n",
"https://smith.langchain.com/o/1fa8b1f4-fcb9-4072-9aa9-983e35ad61b8/projects/p/533846be-d907-4d9c-82db-ce2f1a18fdbf?eval=true\n",
"\n",
"View all tests for Dataset CPI Eval 9648e7fe-5ae2-469f-8701-33c63212d126 at:\n",
"https://smith.langchain.com/datasets/d1762232-5e01-40e7-9978-63002a4c95a3\n",
"[------------------------------------------------->] 4/4View the evaluation results for project 'CPI Eval 9648e7fe-5ae2-469f-8701-33c63212d126-mv_text' at:\n",
"https://smith.langchain.com/o/1fa8b1f4-fcb9-4072-9aa9-983e35ad61b8/projects/p/f5caeede-6f8e-46f7-b4f2-9f23daa31eda?eval=true\n",
"\n",
"View all tests for Dataset CPI Eval 9648e7fe-5ae2-469f-8701-33c63212d126 at:\n",
"https://smith.langchain.com/datasets/d1762232-5e01-40e7-9978-63002a4c95a3\n",
"[------------------------------------------------->] 4/4View the evaluation results for project 'CPI Eval 9648e7fe-5ae2-469f-8701-33c63212d126-mv_img' at:\n",
"https://smith.langchain.com/o/1fa8b1f4-fcb9-4072-9aa9-983e35ad61b8/projects/p/48cf1002-7ae2-451d-a9b1-5bd8088f6a69?eval=true\n",
"\n",
"View all tests for Dataset CPI Eval 9648e7fe-5ae2-469f-8701-33c63212d126 at:\n",
"https://smith.langchain.com/datasets/d1762232-5e01-40e7-9978-63002a4c95a3\n",
"[------------------------------------------------->] 4/4View the evaluation results for project 'CPI Eval 9648e7fe-5ae2-469f-8701-33c63212d126-mm_embd' at:\n",
"https://smith.langchain.com/o/1fa8b1f4-fcb9-4072-9aa9-983e35ad61b8/projects/p/aaa1c2e3-79b0-43e0-b5d5-8e3d00a51d50?eval=true\n",
"\n",
"View all tests for Dataset CPI Eval 9648e7fe-5ae2-469f-8701-33c63212d126 at:\n",
"https://smith.langchain.com/datasets/d1762232-5e01-40e7-9978-63002a4c95a3\n",
"[------------------------------------------------->] 4/4"
]
}
],
"source": [
"from langchain.smith import RunEvalConfig\n",
"\n",
"eval_config = RunEvalConfig(\n",
" evaluators=[\"qa\"],\n",
")\n",
"\n",
"\n",
"def run_eval(chain, run_name, dataset_name):\n",
" _ = client.run_on_dataset(\n",
" dataset_name=dataset_name,\n",
" llm_or_chain_factory=lambda: (lambda x: x[\"question\"] + suffix_for_images)\n",
" | chain,\n",
" evaluation=eval_config,\n",
" project_name=run_name,\n",
" )\n",
"\n",
"\n",
"for chain, run in zip(\n",
" [chain_baseline, chain_mv_text, chain_multimodal_mv_img, chain_multimodal_embd],\n",
" [\"baseline\", \"mv_text\", \"mv_img\", \"mm_embd\"],\n",
"):\n",
" run_eval(chain, dataset_name + \"-\" + run, dataset_name)"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3 (ipykernel)",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.9.16"
}
},
"nbformat": 4,
"nbformat_minor": 5
}
|