6c00ee5a0d
* add pre-commit config Signed-off-by: Sun, Xuehao <xuehao.sun@intel.com> * add prettier Signed-off-by: Sun, Xuehao <xuehao.sun@intel.com> * update prettier Signed-off-by: Sun, Xuehao <xuehao.sun@intel.com> * fix codespell Signed-off-by: Sun, Xuehao <xuehao.sun@intel.com> --------- Signed-off-by: Sun, Xuehao <xuehao.sun@intel.com>
1.2 KiB
1.2 KiB
Performance measurements of chain with langsmith
Pre-requisite: Signup in langsmith [https://www.langchain.com/langsmith] and get the api token
Steps to run perf measurements
- Build langchain-rag container with most updated Dockerfile
- Start tgi server on system with Gaudi
- Statr redis container with docker-compose-redis.yml
- Add your hugging face access token in docker-compose-langchain.yml and start langchain-rag-server container
- enter into langchain-rag-server container and start jupyter notebook server (can specify needed IP address and jupyter will run on port 8888)
docker exec -it langchain-rag-server bash
cd /test
jupyter notebook --allow-root --ip=X.X.X.X
- Launch jupyter notebook in your browser and open the tgi_gaudi.ipynb notebook
- Add langsmith api key in first cell of the notebook [os.environ["LANGCHAIN_API_KEY"] = "add-your-langsmith-key" # Your API key]
- Clear all the cells and run all the cells
- The output of the last cell which calls client.run_on_dataset() will run the langchain Q&A test and captures measurements in the langsmith server. The URL to access the test result can be obtained from the output of the command