token usage parser for llama_index

[tarunn2799]

[ *] I checked the documentation and related resources and couldn't find an answer to my question.

Your Question
what is unclear to you? What would you like to know?

How do I use the token usage parser when I'm using llama_index evaluate?

Code Examples
This community speaks code. Share your code snippets to help us understand your question better.

Additional context
Anything else you want to share with us?

[dosubot]

Hey @tarunn2799! 👋 I'm here to help you with any bugs, questions, or contributions while you wait for a human maintainer. Let's tackle this together!

To use the token usage parser for ragas.integrations.llama_index.evaluate with get_token_usage_for_openai, follow these steps:

1. Import the necessary modules and functions:

   from datasets import load_dataset
   from ragas.metrics import (
       answer_relevancy,
       faithfulness,
       context_recall,
       context_precision,
   )
   from ragas import evaluate
   from ragas.cost import get_token_usage_for_openai
   from langchain_openai.chat_models import ChatOpenAI
   from langchain_core.prompt_values import StringPromptValue

2. Load the dataset:

   amnesty_qa = load_dataset("explodinggradients/amnesty_qa", "english_v2")

3. Initialize the LLM model:

   gpt4o = ChatOpenAI(model="gpt-4o")

4. Perform the evaluation with the token usage parser:

   result = evaluate(
       amnesty_qa["eval"],
       metrics=[
           context_precision,
           faithfulness,
           answer_relevancy,
           context_recall,
       ],
       llm=gpt4o,
       token_usage_parser=get_token_usage_for_openai,
   )

5. Get the total tokens used:

   total_tokens = result.total_tokens()

6. Compute the total cost:

   total_cost = result.total_cost(cost_per_input_token=5 / 1e6, cost_per_output_token=15 / 1e6)

Here is the complete code snippet:

from datasets import load_dataset
from ragas.metrics import (
    answer_relevancy,
    faithfulness,
    context_recall,
    context_precision,
)
from ragas import evaluate
from ragas.cost import get_token_usage_for_openai
from langchain_openai.chat_models import ChatOpenAI
from langchain_core.prompt_values import StringPromptValue

# Load the dataset
amnesty_qa = load_dataset("explodinggradients/amnesty_qa", "english_v2")

# Initialize the LLM model
gpt4o = ChatOpenAI(model="gpt-4o")

# Perform the evaluation with the token usage parser
result = evaluate(
    amnesty_qa["eval"],
    metrics=[
        context_precision,
        faithfulness,
        answer_relevancy,
        context_recall,
    ],
    llm=gpt4o,
    token_usage_parser=get_token_usage_for_openai,
)

# Get the total tokens used
total_tokens = result.total_tokens()

# Compute the total cost
total_cost = result.total_cost(cost_per_input_token=5 / 1e6, cost_per_output_token=15 / 1e6)

This code will help you evaluate the dataset using the specified metrics and compute the token usage and cost using the get_token_usage_for_openai parser ^[1].

To continue talking to Dosu, mention @dosu.

[tarunn2799]

That's just the general documentation - ragas.integrations.llama_index.evaluate does not support token_usage_parser as an argument. That's my problem.

[jjmachan]

@tarunn2799 currently llama_index doesnot support token_parser sadly. I will take a look at it but right now we are working on #1009 so this might take some time sadly 🙁

if you want to take a stab at it I can help you out, would that be something you would be interested?

[tarunn2799]

Yes it would be! Let me know how I can contribute. Thanks :)

[jjmachan]

awesome! this is my cal, can you find a time there?
https://cal.com/jjmachan/huddle

we can go over how you might do this 🙂

[anistark]

Closing as resolved.