mirror of https://github.com/hwchase17/langchain
Add Example Notebook for LCP Client (#4207)
Add a notebook in the `experimental/` directory detailing: - How to capture traces with the v2 endpoint - How to create datasets - How to run traces over the datasetpull/4300/head^2
parent
f1401a6dff
commit
c73cec5ac1
@ -0,0 +1,983 @@
|
||||
{
|
||||
"cells": [
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "1a4596ea-a631-416d-a2a4-3577c140493d",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"# Running Chains on Traced Datasets\n",
|
||||
"\n",
|
||||
"Developing applications with language models can be uniquely challenging. To manage this complexity and ensure reliable performance, LangChain provides tracing and evaluation functionality. This notebook demonstrates how to run Chains, which are language model functions, as well as Chat models, and LLMs on previously captured datasets or traces. Some common use cases for this approach include:\n",
|
||||
"\n",
|
||||
"- Running an evaluation chain to grade previous runs.\n",
|
||||
"- Comparing different chains, LLMs, and agents on traced datasets.\n",
|
||||
"- Executing a stochastic chain multiple times over a dataset to generate metrics before deployment.\n",
|
||||
"\n",
|
||||
"Please note that this notebook assumes you have LangChain+ tracing running in the background. It is also configured to work only with the V2 endpoints. To set it up, follow the [tracing directions here](..\\/..\\/tracing\\/local_installation.md).\n",
|
||||
" \n",
|
||||
"We'll start by creating a client to connect to LangChain+."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 1,
|
||||
"id": "904db9a5-f387-4a57-914c-c8af8d39e249",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [
|
||||
{
|
||||
"name": "stdout",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"You can click the link below to view the UI\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"data": {
|
||||
"text/html": [
|
||||
"<a href=\"http://localhost\", target=\"_blank\" rel=\"noopener\">LangChain+ Client</a>"
|
||||
],
|
||||
"text/plain": [
|
||||
"LangChainPlusClient (API URL: http://localhost:8000)"
|
||||
]
|
||||
},
|
||||
"execution_count": 1,
|
||||
"metadata": {},
|
||||
"output_type": "execute_result"
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"from langchain.client import LangChainPlusClient\n",
|
||||
"\n",
|
||||
"client = LangChainPlusClient(\n",
|
||||
" api_url=\"http://localhost:8000\",\n",
|
||||
" api_key=None,\n",
|
||||
" # tenant_id=\"your_tenant_uuid\", # This is required when connecting to a hosted LangChain instance\n",
|
||||
")\n",
|
||||
"print(\"You can click the link below to view the UI\")\n",
|
||||
"client"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "2d77d064-41b4-41fb-82e6-2d16461269ec",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"source": [
|
||||
"## Capture traces\n",
|
||||
"\n",
|
||||
"If you have been using LangChainPlus already, you may have datasets available. To view all saved datasets, run:\n",
|
||||
"\n",
|
||||
"```\n",
|
||||
"datasets = client.list_datasets()\n",
|
||||
"print(datasets)\n",
|
||||
"```\n",
|
||||
"\n",
|
||||
"Datasets can be created in a number of ways, most often by collecting `Run`'s captured through the LangChain tracing API and converting a set of runs to a dataset.\n",
|
||||
"\n",
|
||||
"The V2 tracing API is currently accessible using the `tracing_v2_enabled` context manager. Assuming the server was succesfully started above, running LangChain Agents, Chains, LLMs, and other primitives will then automatically capture traces. We'll start with a simple math example."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 2,
|
||||
"id": "4417e0b8-a26f-4a11-b7eb-ba7a18e73885",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"from langchain.callbacks.manager import tracing_v2_enabled"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 3,
|
||||
"id": "7c801853-8e96-404d-984c-51ace59cbbef",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"from langchain.chat_models import ChatOpenAI\n",
|
||||
"from langchain.agents import initialize_agent, load_tools\n",
|
||||
"from langchain.agents import AgentType\n",
|
||||
"\n",
|
||||
"llm = ChatOpenAI(temperature=0)\n",
|
||||
"tools = load_tools(['serpapi', 'llm-math'], llm=llm)\n",
|
||||
"agent = initialize_agent(tools, llm, agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION, verbose=False)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 4,
|
||||
"id": "19537902-b95c-4390-80a4-f6c9a937081e",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [
|
||||
{
|
||||
"name": "stderr",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"/Users/wfh/code/lc/lckg/langchain/callbacks/manager.py:65: UserWarning: The experimental tracing v2 is in development. This is not yet stable and may change in the future.\n",
|
||||
" warnings.warn(\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"name": "stdout",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"The current population of Canada as of 2023 is 38,681,797.\n",
|
||||
"Anwar Hadid is Dua Lipa's boyfriend and his age raised to the 0.43 power is approximately 3.87.\n",
|
||||
"'age'. Please try again with a valid numerical expression\n",
|
||||
"The distance between Paris and Boston is approximately 3448 miles.\n",
|
||||
"unknown format from LLM: Assuming we don't have any information about the actual number of points scored in the 2023 super bowl, we cannot provide a mathematical expression to solve this problem.\n",
|
||||
"invalid syntax. Perhaps you forgot a comma? (<expr>, line 1). Please try again with a valid numerical expression\n",
|
||||
"0\n",
|
||||
"1.9347796717823205\n",
|
||||
"1.275494929129063\n",
|
||||
"1213 divided by 4345 is approximately 0.2791.\n"
|
||||
]
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"inputs = [\n",
|
||||
"'How many people live in canada as of 2023?',\n",
|
||||
" \"who is dua lipa's boyfriend? what is his age raised to the .43 power?\",\n",
|
||||
" \"what is dua lipa's boyfriend age raised to the .43 power?\",\n",
|
||||
" 'how far is it from paris to boston in miles',\n",
|
||||
" 'what was the total number of points scored in the 2023 super bowl? what is that number raised to the .23 power?',\n",
|
||||
" 'what was the total number of points scored in the 2023 super bowl raised to the .23 power?',\n",
|
||||
" 'how many more points were scored in the 2023 super bowl than in the 2022 super bowl?',\n",
|
||||
" 'what is 153 raised to .1312 power?',\n",
|
||||
" \"who is kendall jenner's boyfriend? what is his height (in inches) raised to .13 power?\",\n",
|
||||
" 'what is 1213 divided by 4345?'\n",
|
||||
"]\n",
|
||||
"with tracing_v2_enabled(session_name=\"search_and_math_chain\"):\n",
|
||||
" for input_example in inputs:\n",
|
||||
" try:\n",
|
||||
" print(agent.run(input_example))\n",
|
||||
" except Exception as e:\n",
|
||||
" # The agent sometimes makes mistakes! These will be captured by the tracing.\n",
|
||||
" print(e)\n",
|
||||
" "
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "6c43c311-4e09-4d57-9ef3-13afb96ff430",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Creating the Dataset\n",
|
||||
"\n",
|
||||
"Now that you've captured a session entitled 'search_and_math_chain', it's time to create a dataset:\n",
|
||||
"\n",
|
||||
" 1. Navigate to the UI by clicking on the link below.\n",
|
||||
" 2. Select the 'search_and_math_chain' session from the list.\n",
|
||||
" 3. Next to the fist example, click \"+ to Dataset\".\n",
|
||||
" 4. Click \"Create Dataset\" and create a title **\"calculator-example-dataset\"**.\n",
|
||||
" 5. Add the other examples to the dataset as well"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 9,
|
||||
"id": "d14a9881-2a01-404c-8c56-0b78565c3ff4",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"dataset_name = \"calculator-example-dataset\""
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 5,
|
||||
"id": "7bfb4699-62c3-400a-b3e7-7fb8ad3a68ad",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [
|
||||
{
|
||||
"data": {
|
||||
"text/html": [
|
||||
"<a href=\"http://localhost\", target=\"_blank\" rel=\"noopener\">LangChain+ Client</a>"
|
||||
],
|
||||
"text/plain": [
|
||||
"LangChainPlusClient (API URL: http://localhost:8000)"
|
||||
]
|
||||
},
|
||||
"execution_count": 5,
|
||||
"metadata": {},
|
||||
"output_type": "execute_result"
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"client"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "db79dea2-fbaa-4c12-9083-f6154b51e2d3",
|
||||
"metadata": {
|
||||
"jp-MarkdownHeadingCollapsed": true,
|
||||
"tags": []
|
||||
},
|
||||
"source": [
|
||||
"**Optional:** If you didn't run the trace above, you can also create datasets by uploading dataframes or CSV files."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "1baa677c-5642-4378-8e01-3aa1647f19d6",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# !pip install datasets > /dev/null\n",
|
||||
"# !pip install pandas > /dev/null"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "60d14593-c61f-449f-a38f-772ca43707c2",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# import pandas as pd\n",
|
||||
"# from langchain.evaluation.loading import load_dataset\n",
|
||||
"\n",
|
||||
"# dataset = load_dataset(\"agent-search-calculator\")\n",
|
||||
"# df = pd.DataFrame(dataset, columns=[\"question\", \"answer\"])\n",
|
||||
"# df.columns = [\"input\", \"output\"] # The chain we want to evaluate below expects inputs with the \"input\" key \n",
|
||||
"# df.head()"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "52a7ea76-79ca-4765-abf7-231e884040d6",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# dataset_name = \"calculator-example-dataset\"\n",
|
||||
"\n",
|
||||
"# if dataset_name not in set([dataset.name for dataset in client.list_datasets()]):\n",
|
||||
"# dataset = client.upload_dataframe(df, \n",
|
||||
"# name=dataset_name,\n",
|
||||
"# description=\"A calculator example dataset\",\n",
|
||||
"# input_keys=[\"input\"],\n",
|
||||
"# output_keys=[\"output\"],\n",
|
||||
"# )"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "07885b10",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"source": [
|
||||
"## Running a Chain on a Traced Dataset\n",
|
||||
"\n",
|
||||
"Once you have a dataset, you can run a compatible chain or other object over it to see its results. The run traces will automatically be associated with the dataset for easy attribution and analysis.\n",
|
||||
"\n",
|
||||
"**First, we'll define the chain we wish to run over the dataset.**\n",
|
||||
"\n",
|
||||
"In this case, we're using an agent, but it can be any simple chain."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 6,
|
||||
"id": "c2b59104-b90e-466a-b7ea-c5bd0194263b",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"from langchain.chat_models import ChatOpenAI\n",
|
||||
"from langchain.agents import initialize_agent, load_tools\n",
|
||||
"from langchain.agents import AgentType\n",
|
||||
"\n",
|
||||
"llm = ChatOpenAI(temperature=0)\n",
|
||||
"tools = load_tools(['serpapi', 'llm-math'], llm=llm)\n",
|
||||
"agent = initialize_agent(tools, llm, agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION, verbose=False)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "84094a4a-1d76-461c-bc37-8c537939b466",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"**Now we're ready to run the chain!**\n",
|
||||
"\n",
|
||||
"The docstring below hints ways you can configure the method to run."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 7,
|
||||
"id": "112d7bdf-7e50-4c1a-9285-5bac8473f2ee",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [
|
||||
{
|
||||
"data": {
|
||||
"text/plain": [
|
||||
"\u001b[0;31mSignature:\u001b[0m\n",
|
||||
"\u001b[0mclient\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0marun_on_dataset\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\u001b[0m\n",
|
||||
"\u001b[0;34m\u001b[0m \u001b[0mdataset_name\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;34m'str'\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\n",
|
||||
"\u001b[0;34m\u001b[0m \u001b[0mllm_or_chain\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;34m'Union[Chain, BaseLanguageModel]'\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\n",
|
||||
"\u001b[0;34m\u001b[0m \u001b[0mconcurrency_level\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;34m'int'\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m5\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\n",
|
||||
"\u001b[0;34m\u001b[0m \u001b[0mnum_repetitions\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;34m'int'\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\n",
|
||||
"\u001b[0;34m\u001b[0m \u001b[0msession_name\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;34m'Optional[str]'\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\n",
|
||||
"\u001b[0;34m\u001b[0m \u001b[0mverbose\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;34m'bool'\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mFalse\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\n",
|
||||
"\u001b[0;34m\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m->\u001b[0m \u001b[0;34m'Dict[str, Any]'\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
|
||||
"\u001b[0;31mDocstring:\u001b[0m\n",
|
||||
"Run the chain on a dataset and store traces to the specified session name.\n",
|
||||
"\n",
|
||||
"Args:\n",
|
||||
" dataset_name: Name of the dataset to run the chain on.\n",
|
||||
" llm_or_chain: Chain or language model to run over the dataset.\n",
|
||||
" concurrency_level: The number of async tasks to run concurrently.\n",
|
||||
" num_repetitions: Number of times to run the model on each example.\n",
|
||||
" This is useful when testing success rates or generating confidence\n",
|
||||
" intervals.\n",
|
||||
" session_name: Name of the session to store the traces in.\n",
|
||||
" Defaults to {dataset_name}-{chain class name}-{datetime}.\n",
|
||||
" verbose: Whether to print progress.\n",
|
||||
"\n",
|
||||
"Returns:\n",
|
||||
" A dictionary mapping example ids to the model outputs.\n",
|
||||
"\u001b[0;31mFile:\u001b[0m ~/code/lc/lckg/langchain/client/langchain.py\n",
|
||||
"\u001b[0;31mType:\u001b[0m method"
|
||||
]
|
||||
},
|
||||
"metadata": {},
|
||||
"output_type": "display_data"
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"?client.arun_on_dataset"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 10,
|
||||
"id": "a8088b7d-3ab6-4279-94c8-5116fe7cee33",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [
|
||||
{
|
||||
"name": "stderr",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"Chain failed for example bb17c39c-c90c-4706-8410-e501e89a81fb. Error: 'age'. Please try again with a valid numerical expression\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"name": "stdout",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"Processed examples: 1\r"
|
||||
]
|
||||
},
|
||||
{
|
||||
"name": "stderr",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"Chain failed for example e66a366d-f5d3-4dcc-9ec4-aae88369f45e. Error: invalid syntax. Perhaps you forgot a comma? (<expr>, line 1). Please try again with a valid numerical expression\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"name": "stdout",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"Processed examples: 6\r"
|
||||
]
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"chain_results = await client.arun_on_dataset(\n",
|
||||
" dataset_name=dataset_name,\n",
|
||||
" llm_or_chain=agent,\n",
|
||||
" verbose=True\n",
|
||||
")\n",
|
||||
"\n",
|
||||
"# Sometimes, the agent will error due to parsing issues, incompatible tool inputs, etc.\n",
|
||||
"# These are logged as warnings here and captured as errors in the tracing UI."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "d2737458-b20c-4288-8790-1f4a8d237b2a",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Reviewing the Chain Results\n",
|
||||
"\n",
|
||||
"You can review the results of the run in the tracing UI below and navigating to the session \n",
|
||||
"with the title 'calculator-example-dataset-AgentExecutor-YYYY-MM-DD-HH-MM-SS'"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 11,
|
||||
"id": "136db492-d6ca-4215-96f9-439c23538241",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [
|
||||
{
|
||||
"data": {
|
||||
"text/html": [
|
||||
"<a href=\"http://localhost\", target=\"_blank\" rel=\"noopener\">LangChain+ Client</a>"
|
||||
],
|
||||
"text/plain": [
|
||||
"LangChainPlusClient (API URL: http://localhost:8000)"
|
||||
]
|
||||
},
|
||||
"execution_count": 11,
|
||||
"metadata": {},
|
||||
"output_type": "execute_result"
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"# You can navigate to the UI by clicking on the link below\n",
|
||||
"client"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "c70cceb5-aa53-4851-bb12-386f092191f9",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"### Running a Chat Model over a Traced Dataset\n",
|
||||
"\n",
|
||||
"We've shown how to run a _chain_ over a dataset, but you can also run an LLM or Chat model over a datasets formed from runs. \n",
|
||||
"\n",
|
||||
"First, we'll show an example using a ChatModel. This is useful for things like:\n",
|
||||
"- Comparing results under different decoding parameters\n",
|
||||
"- Comparing model providers\n",
|
||||
"- Testing for regressions in model behavior\n",
|
||||
"- Running multiple times with a temperature to gauge stability \n",
|
||||
"\n",
|
||||
"To speed things up, we'll upload a dataset we've previously captured directly to the tracing service."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 12,
|
||||
"id": "64490d7c-9a18-49ed-a3ac-36049c522cb4",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [
|
||||
{
|
||||
"name": "stderr",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"Found cached dataset parquet (/Users/wfh/.cache/huggingface/datasets/LangChainDatasets___parquet/LangChainDatasets--two-player-dnd-2e84407830cdedfc/0.0.0/2a3b91fbd88a2c90d1dbbb32b460cf621d31bd5b05b934492fdef7d8d6f236ec)\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"data": {
|
||||
"application/vnd.jupyter.widget-view+json": {
|
||||
"model_id": "57ffb582606f47e9a3e96629a6d417c2",
|
||||
"version_major": 2,
|
||||
"version_minor": 0
|
||||
},
|
||||
"text/plain": [
|
||||
" 0%| | 0/1 [00:00<?, ?it/s]"
|
||||
]
|
||||
},
|
||||
"metadata": {},
|
||||
"output_type": "display_data"
|
||||
},
|
||||
{
|
||||
"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>prompts</th>\n",
|
||||
" <th>generations</th>\n",
|
||||
" </tr>\n",
|
||||
" </thead>\n",
|
||||
" <tbody>\n",
|
||||
" <tr>\n",
|
||||
" <th>0</th>\n",
|
||||
" <td>[System: Here is the topic for a Dungeons & Dr...</td>\n",
|
||||
" <td>[[{'generation_info': None, 'message': {'conte...</td>\n",
|
||||
" </tr>\n",
|
||||
" <tr>\n",
|
||||
" <th>1</th>\n",
|
||||
" <td>[System: Here is the topic for a Dungeons & Dr...</td>\n",
|
||||
" <td>[[{'generation_info': None, 'message': {'conte...</td>\n",
|
||||
" </tr>\n",
|
||||
" <tr>\n",
|
||||
" <th>2</th>\n",
|
||||
" <td>[System: Here is the topic for a Dungeons & Dr...</td>\n",
|
||||
" <td>[[{'generation_info': None, 'message': {'conte...</td>\n",
|
||||
" </tr>\n",
|
||||
" <tr>\n",
|
||||
" <th>3</th>\n",
|
||||
" <td>[System: Here is the topic for a Dungeons & Dr...</td>\n",
|
||||
" <td>[[{'generation_info': None, 'message': {'conte...</td>\n",
|
||||
" </tr>\n",
|
||||
" <tr>\n",
|
||||
" <th>4</th>\n",
|
||||
" <td>[System: Here is the topic for a Dungeons & Dr...</td>\n",
|
||||
" <td>[[{'generation_info': None, 'message': {'conte...</td>\n",
|
||||
" </tr>\n",
|
||||
" </tbody>\n",
|
||||
"</table>\n",
|
||||
"</div>"
|
||||
],
|
||||
"text/plain": [
|
||||
" prompts \\\n",
|
||||
"0 [System: Here is the topic for a Dungeons & Dr... \n",
|
||||
"1 [System: Here is the topic for a Dungeons & Dr... \n",
|
||||
"2 [System: Here is the topic for a Dungeons & Dr... \n",
|
||||
"3 [System: Here is the topic for a Dungeons & Dr... \n",
|
||||
"4 [System: Here is the topic for a Dungeons & Dr... \n",
|
||||
"\n",
|
||||
" generations \n",
|
||||
"0 [[{'generation_info': None, 'message': {'conte... \n",
|
||||
"1 [[{'generation_info': None, 'message': {'conte... \n",
|
||||
"2 [[{'generation_info': None, 'message': {'conte... \n",
|
||||
"3 [[{'generation_info': None, 'message': {'conte... \n",
|
||||
"4 [[{'generation_info': None, 'message': {'conte... "
|
||||
]
|
||||
},
|
||||
"execution_count": 12,
|
||||
"metadata": {},
|
||||
"output_type": "execute_result"
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"import pandas as pd\n",
|
||||
"from langchain.evaluation.loading import load_dataset\n",
|
||||
"\n",
|
||||
"chat_dataset = load_dataset(\"two-player-dnd\")\n",
|
||||
"chat_df = pd.DataFrame(chat_dataset)\n",
|
||||
"chat_df.head()"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 13,
|
||||
"id": "348acd86-a927-4d60-8d52-02e64585e4fc",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"chat_dataset_name = \"two-player-dnd\"\n",
|
||||
"\n",
|
||||
"if chat_dataset_name not in set([dataset.name for dataset in client.list_datasets()]):\n",
|
||||
" client.upload_dataframe(chat_df, \n",
|
||||
" name=chat_dataset_name,\n",
|
||||
" description=\"An example dataset traced from chat models in a multiagent bidding dialogue\",\n",
|
||||
" input_keys=[\"prompts\"],\n",
|
||||
" output_keys=[\"generations\"],\n",
|
||||
" )"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "927a43b8-e4f9-4220-b75d-33e310bc318b",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"#### Reviewing behavior with temperature\n",
|
||||
"\n",
|
||||
"Here, we will set `num_repetitions > 1` and set the temperature to 0.3 to see the variety of response types for a each example.\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 14,
|
||||
"id": "a69dd183-ad5e-473d-b631-db90706e837f",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"from langchain.chat_models import ChatAnthropic\n",
|
||||
"\n",
|
||||
"chat_model = ChatAnthropic(temperature=.3)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 15,
|
||||
"id": "063da2a9-3692-4b7b-8edb-e474824fe416",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [
|
||||
{
|
||||
"name": "stderr",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"/Users/wfh/code/lc/lckg/langchain/callbacks/manager.py:65: UserWarning: The experimental tracing v2 is in development. This is not yet stable and may change in the future.\n",
|
||||
" warnings.warn(\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"name": "stdout",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"Processed examples: 36\r"
|
||||
]
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"chat_model_results = await client.arun_on_dataset(\n",
|
||||
" dataset_name=chat_dataset_name,\n",
|
||||
" llm_or_chain=chat_model,\n",
|
||||
" concurrency_level=5, # Optional, sets the number of examples to run at a time\n",
|
||||
" num_repetitions=3,\n",
|
||||
" verbose=True\n",
|
||||
")\n",
|
||||
"\n",
|
||||
"# The 'experimental tracing v2' warning is expected, as we are still actively developing the v2 tracing API \n",
|
||||
"# Since we are running examples concurrently, you may run into some RateLimit warnings from your model\n",
|
||||
"# provider. In most cases, the tests will still run to completion (the wrappers have backoff)."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "de7bfe08-215c-4328-b9b0-631d9a41f0e8",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"source": [
|
||||
"## Reviewing the Chat Model Results\n",
|
||||
"\n",
|
||||
"You can review the latest runs by clicking on the link below and navigating to the \"two-player-dnd\" session."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 16,
|
||||
"id": "5b7a81f2-d19d-438b-a4bb-5678f746b965",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [
|
||||
{
|
||||
"data": {
|
||||
"text/html": [
|
||||
"<a href=\"http://localhost\", target=\"_blank\" rel=\"noopener\">LangChain+ Client</a>"
|
||||
],
|
||||
"text/plain": [
|
||||
"LangChainPlusClient (API URL: http://localhost:8000)"
|
||||
]
|
||||
},
|
||||
"execution_count": 16,
|
||||
"metadata": {},
|
||||
"output_type": "execute_result"
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"client"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "7896cbeb-345f-430b-ab5e-e108973174f8",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Running an LLM over a Traced Dataset\n",
|
||||
"\n",
|
||||
"You can run an LLM over a dataset in much the same way as the chain and chat models, provided the dataset you've captured is in the appropriate format. We've cached one for you here, but using application-specific traces will be much more useful for your use cases."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 17,
|
||||
"id": "d6805d0b-4612-4671-bffb-e6978992bd40",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"from langchain.llms import OpenAI\n",
|
||||
"\n",
|
||||
"llm = OpenAI(model_name='text-curie-001', temperature=0)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 18,
|
||||
"id": "5d7cb243-40c3-44dd-8158-a7b910441e9f",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [
|
||||
{
|
||||
"name": "stderr",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"Found cached dataset parquet (/Users/wfh/.cache/huggingface/datasets/LangChainDatasets___parquet/LangChainDatasets--state-of-the-union-completions-ae7542e7bbd0ae0a/0.0.0/2a3b91fbd88a2c90d1dbbb32b460cf621d31bd5b05b934492fdef7d8d6f236ec)\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"data": {
|
||||
"application/vnd.jupyter.widget-view+json": {
|
||||
"model_id": "17b798a06afb49e3af24ca62ee867e82",
|
||||
"version_major": 2,
|
||||
"version_minor": 0
|
||||
},
|
||||
"text/plain": [
|
||||
" 0%| | 0/1 [00:00<?, ?it/s]"
|
||||
]
|
||||
},
|
||||
"metadata": {},
|
||||
"output_type": "display_data"
|
||||
},
|
||||
{
|
||||
"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>prompts</th>\n",
|
||||
" <th>generations</th>\n",
|
||||
" <th>ground_truth</th>\n",
|
||||
" </tr>\n",
|
||||
" </thead>\n",
|
||||
" <tbody>\n",
|
||||
" <tr>\n",
|
||||
" <th>0</th>\n",
|
||||
" <td>[Putin may circle Kyiv with tanks, but he will...</td>\n",
|
||||
" <td>[[{'generation_info': {'finish_reason': 'stop'...</td>\n",
|
||||
" <td>The pandemic has been punishing. \\n\\nAnd so ma...</td>\n",
|
||||
" </tr>\n",
|
||||
" <tr>\n",
|
||||
" <th>1</th>\n",
|
||||
" <td>[Madam Speaker, Madam Vice President, our Firs...</td>\n",
|
||||
" <td>[[]]</td>\n",
|
||||
" <td>With a duty to one another to the American peo...</td>\n",
|
||||
" </tr>\n",
|
||||
" <tr>\n",
|
||||
" <th>2</th>\n",
|
||||
" <td>[With a duty to one another to the American pe...</td>\n",
|
||||
" <td>[[{'generation_info': {'finish_reason': 'stop'...</td>\n",
|
||||
" <td>He thought he could roll into Ukraine and the ...</td>\n",
|
||||
" </tr>\n",
|
||||
" <tr>\n",
|
||||
" <th>3</th>\n",
|
||||
" <td>[Madam Speaker, Madam Vice President, our Firs...</td>\n",
|
||||
" <td>[[{'generation_info': {'finish_reason': 'lengt...</td>\n",
|
||||
" <td>With a duty to one another to the American peo...</td>\n",
|
||||
" </tr>\n",
|
||||
" <tr>\n",
|
||||
" <th>4</th>\n",
|
||||
" <td>[Please rise if you are able and show that, Ye...</td>\n",
|
||||
" <td>[[]]</td>\n",
|
||||
" <td>And the costs and the threats to America and t...</td>\n",
|
||||
" </tr>\n",
|
||||
" </tbody>\n",
|
||||
"</table>\n",
|
||||
"</div>"
|
||||
],
|
||||
"text/plain": [
|
||||
" prompts \\\n",
|
||||
"0 [Putin may circle Kyiv with tanks, but he will... \n",
|
||||
"1 [Madam Speaker, Madam Vice President, our Firs... \n",
|
||||
"2 [With a duty to one another to the American pe... \n",
|
||||
"3 [Madam Speaker, Madam Vice President, our Firs... \n",
|
||||
"4 [Please rise if you are able and show that, Ye... \n",
|
||||
"\n",
|
||||
" generations \\\n",
|
||||
"0 [[{'generation_info': {'finish_reason': 'stop'... \n",
|
||||
"1 [[]] \n",
|
||||
"2 [[{'generation_info': {'finish_reason': 'stop'... \n",
|
||||
"3 [[{'generation_info': {'finish_reason': 'lengt... \n",
|
||||
"4 [[]] \n",
|
||||
"\n",
|
||||
" ground_truth \n",
|
||||
"0 The pandemic has been punishing. \\n\\nAnd so ma... \n",
|
||||
"1 With a duty to one another to the American peo... \n",
|
||||
"2 He thought he could roll into Ukraine and the ... \n",
|
||||
"3 With a duty to one another to the American peo... \n",
|
||||
"4 And the costs and the threats to America and t... "
|
||||
]
|
||||
},
|
||||
"execution_count": 18,
|
||||
"metadata": {},
|
||||
"output_type": "execute_result"
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"completions_dataset = load_dataset(\"state-of-the-union-completions\")\n",
|
||||
"completions_df = pd.DataFrame(completions_dataset)\n",
|
||||
"completions_df.head()"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 19,
|
||||
"id": "c7dcc1b2-7aef-44c0-ba0f-c812279099a5",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"completions_dataset_name = \"state-of-the-union-completions\"\n",
|
||||
"\n",
|
||||
"if completions_dataset_name not in set([dataset.name for dataset in client.list_datasets()]):\n",
|
||||
" client.upload_dataframe(completions_df, \n",
|
||||
" name=completions_dataset_name,\n",
|
||||
" description=\"An example dataset traced from completion endpoints over the state of the union address\",\n",
|
||||
" input_keys=[\"prompts\"],\n",
|
||||
" output_keys=[\"generations\"],\n",
|
||||
" )"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 20,
|
||||
"id": "e946138e-bf7c-43d7-861d-9c5740c933fa",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [
|
||||
{
|
||||
"name": "stderr",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"/Users/wfh/code/lc/lckg/langchain/callbacks/manager.py:65: UserWarning: The experimental tracing v2 is in development. This is not yet stable and may change in the future.\n",
|
||||
" warnings.warn(\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"name": "stdout",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"55 processed\r"
|
||||
]
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"# We also offer a synchronous method for running examples if a cahin's async methods aren't yet implemented\n",
|
||||
"completions_model_results = client.run_on_dataset(\n",
|
||||
" dataset_name=completions_dataset_name,\n",
|
||||
" llm_or_chain=llm,\n",
|
||||
" num_repetitions=1,\n",
|
||||
" verbose=True\n",
|
||||
")"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "cc86e8e6-cee2-429e-942b-289284d14816",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Reviewing the LLM Results\n",
|
||||
"\n",
|
||||
"You can once again inspect the latest runs by clicking on the link below and navigating to the \"two-player-dnd\" session."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 21,
|
||||
"id": "2bf96f17-74c1-4f7d-8458-ae5ab5c6bd36",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"outputs": [
|
||||
{
|
||||
"data": {
|
||||
"text/html": [
|
||||
"<a href=\"http://localhost\", target=\"_blank\" rel=\"noopener\">LangChain+ Client</a>"
|
||||
],
|
||||
"text/plain": [
|
||||
"LangChainPlusClient (API URL: http://localhost:8000)"
|
||||
]
|
||||
},
|
||||
"execution_count": 21,
|
||||
"metadata": {},
|
||||
"output_type": "execute_result"
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"client"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "df80cd88-cd6f-4fdc-965f-f74600e1f286",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": []
|
||||
}
|
||||
],
|
||||
"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": "3.11.2"
|
||||
}
|
||||
},
|
||||
"nbformat": 4,
|
||||
"nbformat_minor": 5
|
||||
}
|
Loading…
Reference in New Issue