Skip to content

Commit

Permalink
Add hyperlinks and paths validation. (#132)
Browse files Browse the repository at this point in the history
Signed-off-by: ZePan110 <[email protected]>
Signed-off-by: aasavari <[email protected]>
  • Loading branch information
ZePan110 authored and adkakne committed Sep 20, 2024
1 parent 44a12f0 commit 6502c68
Show file tree
Hide file tree
Showing 6 changed files with 176 additions and 5 deletions.
123 changes: 123 additions & 0 deletions .github/workflows/pr-path-detection.yml
Original file line number Diff line number Diff line change
@@ -0,0 +1,123 @@
# Copyright (C) 2024 Intel Corporation
# SPDX-License-Identifier: Apache-2.0

name: Check Paths and Hyperlinks

on:
pull_request:
branches: [main]
types: [opened, reopened, ready_for_review, synchronize]

jobs:
check-the-validity-of-hyperlinks-in-README:
runs-on: ubuntu-latest
steps:
- name: Clean Up Working Directory
run: sudo rm -rf ${{github.workspace}}/*

- name: Checkout Repo GenAIEval
uses: actions/checkout@v4

- name: Check the Validity of Hyperlinks
run: |
cd ${{github.workspace}}
fail="FALSE"
url_lines=$(grep -Eo '\]\(http[s]?://[^)]+\)' --include='*.md' -r .|grep -Ev 'GenAIEval/blob/main')
if [ -n "$url_lines" ]; then
for url_line in $url_lines; do
url=$(echo "$url_line"|cut -d '(' -f2 | cut -d ')' -f1|sed 's/\.git$//')
path=$(echo "$url_line"|cut -d':' -f1 | cut -d'/' -f2-)
response=$(curl -L -s -o /dev/null -w "%{http_code}" "$url")
if [ "$response" -ne 200 ]; then
echo "**********Validation failed, try again**********"
response_retry=$(curl -s -o /dev/null -w "%{http_code}" "$url")
if [ "$response_retry" -eq 200 ]; then
echo "*****Retry successfully*****"
else
echo "Invalid link from ${{github.workspace}}/$path: $url"
fail="TRUE"
fi
fi
done
fi
if [[ "$fail" == "TRUE" ]]; then
exit 1
else
echo "All hyperlinks are valid."
fi
shell: bash

check-the-validity-of-relative-path:
runs-on: ubuntu-latest
steps:
- name: Clean up Working Directory
run: sudo rm -rf ${{github.workspace}}/*

- name: Checkout Repo GenAIEval
uses: actions/checkout@v4

- name: Checking Relative Path Validity
run: |
cd ${{github.workspace}}
fail="FALSE"
repo_name=${{ github.event.pull_request.head.repo.full_name }}
if [ "$(echo "$repo_name"|cut -d'/' -f1)" != "opea-project" ]; then
owner=$(echo "${{ github.event.pull_request.head.repo.full_name }}" |cut -d'/' -f1)
branch="https://github.com/$owner/GenAIEval/tree/${{ github.event.pull_request.head.ref }}"
else
branch="https://github.com/opea-project/GenAIEval/blob/${{ github.event.pull_request.head.ref }}"
fi
link_head="https://github.com/opea-project/GenAIEval/blob/main"
png_lines=$(grep -Eo '\]\([^)]+\)' --include='*.md' -r .|grep -Ev 'http')
if [ -n "$png_lines" ]; then
for png_line in $png_lines; do
refer_path=$(echo "$png_line"|cut -d':' -f1 | cut -d'/' -f2-)
png_path=$(echo "$png_line"|cut -d '(' -f2 | cut -d ')' -f1)
if [[ "${png_path:0:1}" == "/" ]]; then
check_path=${{github.workspace}}$png_path
elif [[ "${png_path:0:1}" == "#" ]]; then
check_path=${{github.workspace}}/$refer_path$png_path
else
check_path=${{github.workspace}}/$(dirname "$refer_path")/$png_path
fi
real_path=$(realpath $check_path)
if [ $? -ne 0 ]; then
echo "Path $png_path in file ${{github.workspace}}/$refer_path does not exist"
fail="TRUE"
else
url=$link_head$(echo "$real_path" | sed 's|.*/GenAIEval||')
response=$(curl -I -L -s -o /dev/null -w "%{http_code}" "$url")
if [ "$response" -ne 200 ]; then
echo "**********Validation failed, try again**********"
response_retry=$(curl -s -o /dev/null -w "%{http_code}" "$url")
if [ "$response_retry" -eq 200 ]; then
echo "*****Retry successfully*****"
else
echo "Retry failed. Check branch ${{ github.event.pull_request.head.ref }}"
url_dev=$branch$(echo "$real_path" | sed 's|.*/GenAIEval||')
response=$(curl -I -L -s -o /dev/null -w "%{http_code}" "$url_dev")
if [ "$response" -ne 200 ]; then
echo "**********Validation failed, try again**********"
response_retry=$(curl -s -o /dev/null -w "%{http_code}" "$url_dev")
if [ "$response_retry" -eq 200 ]; then
echo "*****Retry successfully*****"
else
echo "Invalid path from ${{github.workspace}}/$refer_path: $png_path"
fail="TRUE"
fi
else
echo "Check branch ${{ github.event.pull_request.head.ref }} successfully."
fi
fi
fi
fi
done
fi
if [[ "$fail" == "TRUE" ]]; then
exit 1
else
echo "All hyperlinks are valid."
fi
shell: bash
2 changes: 1 addition & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -67,7 +67,7 @@ results = evaluate(args)

#### remote service usage

1. setup a separate server with [GenAIComps](https://github.com/opea-project/GenAIComps/tree/main/comps/llms/lm-eval)
1. setup a separate server with [GenAIComps](https://github.com/opea-project/GenAIComps/tree/main/comps/llms/utils/lm-eval)

```
# build cpu docker
Expand Down
2 changes: 1 addition & 1 deletion doc/platform-optimization/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -98,7 +98,7 @@ Let us consider isolating AI inference and reranking containers in
application's Gaudi accelerated pipeline.

In the
[manifest](https://github.com/opea-project/GenAIExamples/blob/main/ChatQnA/kubernetes/manifests/gaudi/chatqna.yaml)
[manifest](https://github.com/opea-project/GenAIExamples/blob/main/ChatQnA/kubernetes/intel/hpu/gaudi/manifest/chatqna.yaml)
there are "tgi", "tei" and "teirerank" containers in "chatqna-tgi" and
"chatqna-tei" and "chatqna-teirerank" deployments that will need a lot
of CPUs. They implement text-generation-interface and
Expand Down
4 changes: 2 additions & 2 deletions evals/evaluation/autorag/evaluation/README.md
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
# AutoRAG to evaluate the RAG system performance

AutoRAG is help to end-to-end evaluate the performance of the whole system. Currently, we support to evaluate the performance from 4 perspectives, answer_relevancy, faithfulness, context_recall, context_precision. Before using this service, the use should firstly prepare the groundtruth dataset in the [standard format](https://github.com/opea-project/GenAIEval/blob/main/evals/benchmark/ground_truth.jsonl). We also provide a [script](https://github.com/opea-project/GenAIEval/blob/main/evals/evaluation/autorag/data_generation/gen_eval_dataset.py) to automatically generate the groundtruth query and answer.
AutoRAG is help to end-to-end evaluate the performance of the whole system. Currently, we support to evaluate the performance from 4 perspectives, answer_relevancy, faithfulness, context_recall, context_precision. Before using this service, the use should firstly prepare the groundtruth dataset in the [standard format](https://github.com/opea-project/GenAIEval/blob/main/evals/benchmark/ragas/ground_truth.jsonl). We also provide a [script](https://github.com/opea-project/GenAIEval/blob/main/evals/evaluation/autorag/data_generation/gen_eval_dataset.py) to automatically generate the groundtruth query and answer.

## Service preparation
The evaluation for the RAG system is based on the set up of the RAG services. Please follow [the steps](https://github.com/opea-project/GenAIExamples/blob/main/ChatQnA/README.md) to set up your RAG services.
Expand All @@ -12,7 +12,7 @@ At this moment, we provide a solution that test the single group of parameters a
python -u ragas_evaluation_benchmark.py --ground_truth_file ground_truth.jsonl --search_type mmr --k 1 --fetch_k 5 --score_threshold 0.3 --top_n 1 --temperature 0.01 --top_k 5 --top_p 0.95 --repetition_penalty 1.1 --use_openai_key True
```

For evaluating multiple groups of parameters, please use [this script](https://github.com/opea-project/GenAIEval/blob/main/evals/benchmark/run_rag_benchmark.py).
For evaluating multiple groups of parameters, please use [this script](https://github.com/opea-project/GenAIEval/blob/main/evals/benchmark/ragas/run_rag_benchmark.py).
```bash
python -u run_rag_benchmark.py --config config.yaml
```
Expand Down
2 changes: 1 addition & 1 deletion evals/evaluation/rag_eval/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@
- [Prerequisites](#prerequisites)
- [MultiHop (English dataset)](#multihop)
- [Launch Service of RAG System](#launch-service-of-rag-system)
- [Launch Service of LLM-as-a-Judge](launch-service-of-llm)
- [Launch Service of LLM-as-a-Judge](#launch-service-of-llm-as-a-judge)
- [Prepare Dataset](#prepare-dataset)
- [Evaluation](#evaluation)
- [CRUD (Chinese dataset)](#crud)
Expand Down
48 changes: 48 additions & 0 deletions examples/AudioQnA/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,48 @@
# AudioQnA accuracy Evaluation

## Dataset


We evaluate the ASR accuracy on the test set of librispeech [dataset](https://huggingface.co/datasets/andreagasparini/librispeech_test_only), which contains 2620 records of audio and texts.

## Metrics

We evaluate the WER (Word Error Rate) metric of the ASR microservice.

## Evaluation

### Launch ASR microservice

Launch the ASR microserice with the following commands. For more details please refer to [doc](https://github.com/opea-project/GenAIComps/tree/main/comps/asr).

```bash
git clone https://github.com/opea-project/GenAIComps
cd GenAIComps
docker build -t opea/whisper:latest --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy -f comps/asr/whisper/Dockerfile .
# change the name of model by editing model_name_or_path you want to evaluate
docker run -p 7066:7066 --ipc=host -e http_proxy=$http_proxy -e https_proxy=$https_proxy opea/whisper:latest --model_name_or_path "openai/whisper-tiny"
```

### Evaluate

Install dependencies:

```
pip install -r requirements.txt
```

Evaluate the performance with the LLM:
```py
# validate the offline model
# python offline_evaluate.py
# validate the online asr microservice accuracy
python online_evaluate.py
```

### Performance Result
Here is the tested result for your reference
|| WER |
| --- | ---- |
|whisper-large-v2| 2.87|
|whisper-large| 2.7 |
|whisper-medium| 3.45 |

0 comments on commit 6502c68

Please sign in to comment.