RAGAS TimeoutError Exception while Evaluation

[kaan9700]

I have a self-created data set with questions, GT, contexts and answers and have started the evaluation with the RAGAS evaluate() method. The percentage increases a few points and then stops. Then I get several identical errors. Then the evaluation continues and freezes again after a few minutes.

Ragas version: 0.1.13 Python version: 3.10.13

Code to Reproduce

import pandas as pd
from datasets import Dataset
from ragas import evaluate
from ragas.metrics import (
    answer_relevancy,
    faithfulness,
    context_recall,
    context_precision,
    context_utilization,
    context_entity_recall,
    answer_correctness,
    answer_similarity,
)
from ragas.metrics.critique import harmfulness
from dotenv import load_dotenv

load_dotenv()

df = pd.read_excel('testset_full-200-ES.xlsx')

# Create the 'data_samples' dictionary structure
data_samples = {
    'question': df['question'].tolist(),
    'answer': df['answer'].tolist(),
    'contexts': df['contexts'].apply(lambda x: [x] if pd.notna(x) else []).tolist(),
    'ground_truth': df['ground_truth'].tolist()
}
dataset = Dataset.from_dict(data_samples)

result = evaluate(
    dataset,
    metrics=[
        context_precision,
        faithfulness,
        answer_relevancy,
        context_recall,
        context_utilization,
        context_entity_recall,
        answer_correctness,
        answer_similarity,
        harmfulness,
    ],
)

df = result.to_pandas()
print(result)
# save evaluation results to csv
df.to_csv('results_es.csv', index=False)

Error trace

Evaluating:  11%|█         | 1709/15453 [14:40<330:40:17, 86.61s/it]Exception raised in Job[13246]: TimeoutError()
Exception raised in Job[13243]: TimeoutError()
Exception raised in Job[5118]: TimeoutError()
Exception raised in Job[13245]: TimeoutError()
Exception raised in Job[5125]: TimeoutError()
Exception raised in Job[13241]: TimeoutError()
Exception raised in Job[5126]: TimeoutError()
Exception raised in Job[13227]: TimeoutError()
Exception raised in Job[5122]: TimeoutError()
Exception raised in Job[13244]: TimeoutError()
Exception raised in Job[13240]: TimeoutError()
Exception raised in Job[13236]: TimeoutError()
Exception raised in Job[5113]: TimeoutError()
Exception raised in Job[5124]: TimeoutError()
Exception raised in Job[14126]: TimeoutError()
Evaluating:  13%|█▎        | 2063/15453 [33:06<750:37:32, 201.81s/it]Exception raised in Job[8734]: TimeoutError()
Exception raised in Job[8731]: TimeoutError()
Exception raised in Job[4239]: TimeoutError()
Exception raised in Job[8727]: TimeoutError()
Exception raised in Job[1626]: TimeoutError()
Exception raised in Job[12353]: TimeoutError()
Exception raised in Job[4231]: TimeoutError()
Exception raised in Job[12354]: TimeoutError()
Exception raised in Job[4235]: TimeoutError()
Exception raised in Job[12358]: TimeoutError()
Exception raised in Job[4236]: TimeoutError()
Exception raised in Job[8732]: TimeoutError()
Exception raised in Job[1625]: TimeoutError()
Exception raised in Job[1630]: TimeoutError()
Exception raised in Job[1617]: TimeoutError()

[dosubot[bot]]

Hey @kaan9700! How can I assist you today?

I found some similar issues that might help you resolve the TimeoutError when using the evaluate() function in the ragas library.

1. Timeout Configuration: You need to set the timeout parameter in the RunConfig and ensure it is passed to the evaluation function. Here is an example of how to do it:

   from ragas import evaluate, RunConfig
   from ragas.metrics import (
      answer_relevancy,
      faithfulness,
      context_recall,
      context_precision,
      context_utilization,
      context_entity_recall,
      answer_correctness,
      answer_similarity,
   )
   from ragas.metrics.critique import harmfulness
   from datasets import Dataset
   import pandas as pd
   
   # Load your dataset
   df = pd.read_excel('testset_full-200-ES.xlsx')
   data_samples = {
      'question': df['question'].tolist(),
      'answer': df['answer'].tolist(),
      'contexts': df['contexts'].apply(lambda x: [x] if pd.notna(x) else []).tolist(),
      'ground_truth': df['ground_truth'].tolist()
   }
   dataset = Dataset.from_dict(data_samples)
   
   # Create a RunConfig with a specified timeout
   run_config = RunConfig(timeout=120)  # Set timeout to 120 seconds
   
   # Use the evaluate function with the custom run_config
   result = evaluate(
      dataset,
      metrics=[
          context_precision,
          faithfulness,
          answer_relevancy,
          context_recall,
          context_utilization,
          context_entity_recall,
          answer_correctness,
          answer_similarity,
          harmfulness,
      ],
      run_config=run_config
   )
   
   df = result.to_pandas()
   print(result)
   df.to_csv('results_es.csv', index=False)

2. Known Issues and Limitations: There are known issues with using certain configurations in the ragas library, such as the lack of support for asynchronous callbacks and specific configurations required for different types of models ^[1].

3. Debugging and Logging: Enable debugging and logging options to diagnose the TimeoutError. Set the log_tenacity parameter in RunConfig to True and enable the RAGAS_DEBUG environment variable:

   import os
   from ragas import RunConfig
   
   os.environ["RAGAS_DEBUG"] = "true"
   
   run_config = RunConfig(timeout=120, log_tenacity=True)

   This will provide comprehensive logging and debugging information ^[1].

By following these steps, you should be able to resolve the TimeoutError and successfully use the evaluate() function with your dataset.

To continue talking to Dosu, mention @dosu.

[mmiumiu]

The same error. Have you fixed it?

[CraigMChen]

+1

[jjmachan]

@CraigMChen does it happen with timeout too?

[CraigMChen]

@jjmachan

Here's my code:

from ragas.llms import LangchainLLMWrapper
from ragas import evaluate
from langchain.chat_models import ChatOllama
from ragas.metrics import (
    faithfulness,
    answer_relevancy,
    context_recall,
    context_precision,
)

llm = ChatOllama(model="llama3.2", base_url="http://127.0.0.1:11434", request_timeout=3000)
print(llm.invoke("hello")) # ok

result = evaluate(
    llm=LangchainLLMWrapper(llm),
    dataset=dataset,
    metrics=[
        context_precision,
        context_recall,
        faithfulness,
        answer_relevancy,
    ],
)

df = result.to_pandas()

Error:

Exception raised in Job[0]: TimeoutError()
Exception raised in Job[3]: TimeoutError()
Exception raised in Job[2]: TimeoutError()
Exception raised in Job[1]: TimeoutError()