LLM RAG Evaluation with MLflow Example Notebook

In this notebook, we will demonstrate how to evaluate various a RAG system with MLflow.

Download this Notebook

We need to set our OpenAI API key.

In order to set your private key safely, please be sure to either export your key through a command-line terminal for your current instance, or, for a permanent addition to all user-based sessions, configure your favored environment management configuration file (i.e., .bashrc, .zshrc) to have the following entry:

OPENAI_API_KEY=<your openai API key>

If using Azure OpenAI, you will instead need to set

OPENAI_API_TYPE="azure"

OPENAI_API_VERSION=<YYYY-MM-DD>

OPENAI_API_KEY=<https://<>.<>.<>.com>

OPENAI_API_DEPLOYMENT_NAME=<deployment name>

Notebook compatibility

With rapidly changing libraries such as langchain, examples can become outdated rather quickly and will no longer work. For the purposes of demonstration, here are the critical dependencies that are recommended to use to effectively run this notebook:

Package	Version
langchain	0.1.16
lanchain-community	0.0.33
langchain-openai	0.0.8
openai	1.12.0
mlflow	2.12.1
chromadb	0.4.24

If you attempt to execute this notebook with different versions, it may function correctly, but it is recommended to use the precise versions above to ensure that your code executes properly.

Create a RAG system

Use Langchain and Chroma to create a RAG system that answers questions based on the MLflow documentation.

import pandas as pd
from langchain.chains import RetrievalQA
from langchain.document_loaders import WebBaseLoader
from langchain.text_splitter import CharacterTextSplitter
from langchain.vectorstores import Chroma
from langchain_openai import OpenAI, OpenAIEmbeddings

import mlflow

loader = WebBaseLoader("https://mlflow.org/docs/latest/index.html")

documents = loader.load()
text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0)
texts = text_splitter.split_documents(documents)

embeddings = OpenAIEmbeddings()
docsearch = Chroma.from_documents(texts, embeddings)

qa = RetrievalQA.from_chain_type(
    llm=OpenAI(temperature=0),
    chain_type="stuff",
    retriever=docsearch.as_retriever(),
    return_source_documents=True,
)

Evaluate the RAG system using mlflow.evaluate()

Create a simple function that runs each input through the RAG chain

def model(input_df):
    answer = []
    for index, row in input_df.iterrows():
        answer.append(qa(row["questions"]))

    return answer

Create an eval dataset

eval_df = pd.DataFrame(
    {
        "questions": [
            "What is MLflow?",
            "How to run mlflow.evaluate()?",
            "How to log_table()?",
            "How to load_table()?",
        ],
    }
)

Create a faithfulness metric

from mlflow.metrics.genai import EvaluationExample, faithfulness

# Create a good and bad example for faithfulness in the context of this problem
faithfulness_examples = [
    EvaluationExample(
        input="How do I disable MLflow autologging?",
        output="mlflow.autolog(disable=True) will disable autologging for all functions. In Databricks, autologging is enabled by default. ",
        score=2,
        justification="The output provides a working solution, using the mlflow.autolog() function that is provided in the context.",
        grading_context={
            "context": "mlflow.autolog(log_input_examples: bool = False, log_model_signatures: bool = True, log_models: bool = True, log_datasets: bool = True, disable: bool = False, exclusive: bool = False, disable_for_unsupported_versions: bool = False, silent: bool = False, extra_tags: Optional[Dict[str, str]] = None) → None[source] Enables (or disables) and configures autologging for all supported integrations. The parameters are passed to any autologging integrations that support them. See the tracking docs for a list of supported autologging integrations. Note that framework-specific configurations set at any point will take precedence over any configurations set by this function."
        },
    ),
    EvaluationExample(
        input="How do I disable MLflow autologging?",
        output="mlflow.autolog(disable=True) will disable autologging for all functions.",
        score=5,
        justification="The output provides a solution that is using the mlflow.autolog() function that is provided in the context.",
        grading_context={
            "context": "mlflow.autolog(log_input_examples: bool = False, log_model_signatures: bool = True, log_models: bool = True, log_datasets: bool = True, disable: bool = False, exclusive: bool = False, disable_for_unsupported_versions: bool = False, silent: bool = False, extra_tags: Optional[Dict[str, str]] = None) → None[source] Enables (or disables) and configures autologging for all supported integrations. The parameters are passed to any autologging integrations that support them. See the tracking docs for a list of supported autologging integrations. Note that framework-specific configurations set at any point will take precedence over any configurations set by this function."
        },
    ),
]

faithfulness_metric = faithfulness(model="openai:/gpt-4", examples=faithfulness_examples)
print(faithfulness_metric)

EvaluationMetric(name=faithfulness, greater_is_better=True, long_name=faithfulness, version=v1, metric_details=
Task:
You must return the following fields in your response one below the other:
score: Your numerical score for the model's faithfulness based on the rubric
justification: Your step-by-step reasoning about the model's faithfulness score

You are an impartial judge. You will be given an input that was sent to a machine
learning model, and you will be given an output that the model produced. You
may also be given additional information that was used by the model to generate the output.

Your task is to determine a numerical score called faithfulness based on the input and output.
A definition of faithfulness and a grading rubric are provided below.
You must use the grading rubric to determine your score. You must also justify your score.

Examples could be included below for reference. Make sure to use them as references and to
understand them before completing the task.

Input:
{input}

Output:
{output}

{grading_context_columns}

Metric definition:
Faithfulness is only evaluated with the provided output and provided context, please ignore the provided input entirely when scoring faithfulness. Faithfulness assesses how much of the provided output is factually consistent with the provided context. A higher score indicates that a higher proportion of claims present in the output can be derived from the provided context. Faithfulness does not consider how much extra information from the context is not present in the output.

Grading rubric:
Faithfulness: Below are the details for different scores:
- Score 1: None of the claims in the output can be inferred from the provided context.
- Score 2: Some of the claims in the output can be inferred from the provided context, but the majority of the output is missing from, inconsistent with, or contradictory to the provided context.
- Score 3: Half or more of the claims in the output can be inferred from the provided context.
- Score 4: Most of the claims in the output can be inferred from the provided context, with very little information that is not directly supported by the provided context.
- Score 5: All of the claims in the output are directly supported by the provided context, demonstrating high faithfulness to the provided context.

Examples:

Example Input:
How do I disable MLflow autologging?

Example Output:
mlflow.autolog(disable=True) will disable autologging for all functions. In Databricks, autologging is enabled by default.

Additional information used by the model:
key: context
value:
mlflow.autolog(log_input_examples: bool = False, log_model_signatures: bool = True, log_models: bool = True, log_datasets: bool = True, disable: bool = False, exclusive: bool = False, disable_for_unsupported_versions: bool = False, silent: bool = False, extra_tags: Optional[Dict[str, str]] = None) → None[source] Enables (or disables) and configures autologging for all supported integrations. The parameters are passed to any autologging integrations that support them. See the tracking docs for a list of supported autologging integrations. Note that framework-specific configurations set at any point will take precedence over any configurations set by this function.

Example score: 2
Example justification: The output provides a working solution, using the mlflow.autolog() function that is provided in the context.


Example Input:
How do I disable MLflow autologging?

Example Output:
mlflow.autolog(disable=True) will disable autologging for all functions.

Additional information used by the model:
key: context
value:
mlflow.autolog(log_input_examples: bool = False, log_model_signatures: bool = True, log_models: bool = True, log_datasets: bool = True, disable: bool = False, exclusive: bool = False, disable_for_unsupported_versions: bool = False, silent: bool = False, extra_tags: Optional[Dict[str, str]] = None) → None[source] Enables (or disables) and configures autologging for all supported integrations. The parameters are passed to any autologging integrations that support them. See the tracking docs for a list of supported autologging integrations. Note that framework-specific configurations set at any point will take precedence over any configurations set by this function.

Example score: 5
Example justification: The output provides a solution that is using the mlflow.autolog() function that is provided in the context.


You must return the following fields in your response one below the other:
score: Your numerical score for the model's faithfulness based on the rubric
justification: Your step-by-step reasoning about the model's faithfulness score
    )

Create a relevance metric. You can see the full grading prompt by printing the metric or by accessing the metric_details attribute of the metric.

from mlflow.metrics.genai import EvaluationExample, relevance

relevance_metric = relevance(model="openai:/gpt-4")
print(relevance_metric)

EvaluationMetric(name=relevance, greater_is_better=True, long_name=relevance, version=v1, metric_details=
Task:
You must return the following fields in your response one below the other:
score: Your numerical score for the model's relevance based on the rubric
justification: Your step-by-step reasoning about the model's relevance score

You are an impartial judge. You will be given an input that was sent to a machine
learning model, and you will be given an output that the model produced. You
may also be given additional information that was used by the model to generate the output.

Your task is to determine a numerical score called relevance based on the input and output.
A definition of relevance and a grading rubric are provided below.
You must use the grading rubric to determine your score. You must also justify your score.

Examples could be included below for reference. Make sure to use them as references and to
understand them before completing the task.

Input:
{input}

Output:
{output}

{grading_context_columns}

Metric definition:
Relevance encompasses the appropriateness, significance, and applicability of the output with respect to both the input and context. Scores should reflect the extent to which the output directly addresses the question provided in the input, given the provided context.

Grading rubric:
Relevance: Below are the details for different scores:- Score 1: The output doesn't mention anything about the question or is completely irrelevant to the provided context.
- Score 2: The output provides some relevance to the question and is somehow related to the provided context.
- Score 3: The output mostly answers the question and is largely consistent with the provided context.
- Score 4: The output answers the question and is consistent with the provided context.
- Score 5: The output answers the question comprehensively using the provided context.

Examples:

Example Input:
How is MLflow related to Databricks?

Example Output:
Databricks is a data engineering and analytics platform designed to help organizations process and analyze large amounts of data. Databricks is a company specializing in big data and machine learning solutions.

Additional information used by the model:
key: context
value:
MLflow is an open-source platform for managing the end-to-end machine learning (ML) lifecycle. It was developed by Databricks, a company that specializes in big data and machine learning solutions. MLflow is designed to address the challenges that data scientists and machine learning engineers face when developing, training, and deploying machine learning models.

Example score: 2
Example justification: The output provides relevant information about Databricks, mentioning it as a company specializing in big data and machine learning solutions. However, it doesn't directly address how MLflow is related to Databricks, which is the specific question asked in the input. Therefore, the output is only somewhat related to the provided context.


Example Input:
How is MLflow related to Databricks?

Example Output:
MLflow is a product created by Databricks to enhance the efficiency of machine learning processes.

Additional information used by the model:
key: context
value:
MLflow is an open-source platform for managing the end-to-end machine learning (ML) lifecycle. It was developed by Databricks, a company that specializes in big data and machine learning solutions. MLflow is designed to address the challenges that data scientists and machine learning engineers face when developing, training, and deploying machine learning models.

Example score: 4
Example justification: The output provides a relevant and accurate statement about the relationship between MLflow and Databricks. While it doesn't provide extensive detail, it still offers a substantial and meaningful response. To achieve a score of 5, the response could be further improved by providing additional context or details about how MLflow specifically functions within the Databricks ecosystem.


You must return the following fields in your response one below the other:
score: Your numerical score for the model's relevance based on the rubric
justification: Your step-by-step reasoning about the model's relevance score
    )

results = mlflow.evaluate(
    model,
    eval_df,
    model_type="question-answering",
    evaluators="default",
    predictions="result",
    extra_metrics=[faithfulness_metric, relevance_metric, mlflow.metrics.latency()],
    evaluator_config={
        "col_mapping": {
            "inputs": "questions",
            "context": "source_documents",
        }
    },
)
print(results.metrics)

2023/11/16 09:05:21 INFO mlflow.models.evaluation.base: Evaluating the model with the default evaluator.
2023/11/16 09:05:21 INFO mlflow.models.evaluation.default_evaluator: Computing model predictions.
2023/11/16 09:05:28 INFO mlflow.models.evaluation.default_evaluator: Testing metrics on first row...
Using default facebook/roberta-hate-speech-dynabench-r4-target checkpoint

2023/11/16 09:05:58 INFO mlflow.models.evaluation.default_evaluator: Evaluating builtin metrics: token_count
2023/11/16 09:05:58 INFO mlflow.models.evaluation.default_evaluator: Evaluating builtin metrics: toxicity
2023/11/16 09:05:58 INFO mlflow.models.evaluation.default_evaluator: Evaluating builtin metrics: flesch_kincaid_grade_level
2023/11/16 09:05:58 INFO mlflow.models.evaluation.default_evaluator: Evaluating builtin metrics: ari_grade_level
2023/11/16 09:05:58 INFO mlflow.models.evaluation.default_evaluator: Evaluating builtin metrics: exact_match
2023/11/16 09:05:58 INFO mlflow.models.evaluation.default_evaluator: Evaluating metrics: faithfulness

2023/11/16 09:06:12 INFO mlflow.models.evaluation.default_evaluator: Evaluating metrics: relevance

{'toxicity/v1/mean': 0.00022622970209340565, 'toxicity/v1/variance': 3.84291113351624e-09, 'toxicity/v1/p90': 0.0002859298692783341, 'toxicity/v1/ratio': 0.0, 'flesch_kincaid_grade_level/v1/mean': 8.1, 'flesch_kincaid_grade_level/v1/variance': 8.815, 'flesch_kincaid_grade_level/v1/p90': 11.48, 'ari_grade_level/v1/mean': 11.649999999999999, 'ari_grade_level/v1/variance': 19.527499999999993, 'ari_grade_level/v1/p90': 16.66, 'faithfulness/v1/mean': 4.0, 'faithfulness/v1/variance': 3.0, 'faithfulness/v1/p90': 5.0, 'relevance/v1/mean': 4.5, 'relevance/v1/variance': 0.25, 'relevance/v1/p90': 5.0}

results.tables["eval_results_table"]

	questions	outputs	source_documents	latency	token_count	toxicity/v1/score	flesch_kincaid_grade_level/v1/score	ari_grade_level/v1/score	faithfulness/v1/score	faithfulness/v1/justification	relevance/v1/score	relevance/v1/justification
0	What is MLflow?	MLflow is an open-source platform, purpose-bu...	[{'lc_attributes': {}, 'lc_namespace': ['langc...	1.989822	53	0.000137	12.5	18.4	5	The output provided by the model is a direct e...	5	The output provides a comprehensive answer to ...
1	How to run mlflow.evaluate()?	The mlflow.evaluate() API allows you to valid...	[{'lc_attributes': {}, 'lc_namespace': ['langc...	1.945368	55	0.000200	9.1	12.6	5	The output provided by the model is completely...	4	The output provides a relevant and accurate ex...
2	How to log_table()?	You can log a table with MLflow using the log...	[{'lc_attributes': {}, 'lc_namespace': ['langc...	1.521511	32	0.000289	5.0	6.8	1	The output claims that you can log a table wit...	5	The output provides a comprehensive answer to ...
3	How to load_table()?	You can't load_table() with MLflow. MLflow is...	[{'lc_attributes': {}, 'lc_namespace': ['langc...	1.105279	27	0.000279	5.8	8.8	5	The output claim that "You can't load_table() ...	4	The output provides a relevant and accurate re...