I was trying to modify the demo notebook below for 'Dynamic Section Retrieval with LlamaParse' (originally based on OpenAI's gpt-4o) so that it could work with an LLM on Groq and an open embedding model on Hugging Face:
https://github.com/run-llama/llama_parse/blob/main/examples/advanced_rag/dynamic_section_retrieval.ipynbMy modified notebook and data files are in this repo:
https://github.com/tlyim/DynSecRetrievalOn Google Colab (Python 3, T4 GPU), I could execute the modified notebook with expected results up to and before the chunk with this code line:
sections_dict = asyncio_run(acreate_sections(text_nodes_dict))When I ran this line repeatedly without changing anything, each time it could give me a different error message. In a random fashion, the error message could be one of the three types below.
Type 1:
https://github.com/tlyim/DynSecRetrieval/blob/main/errorTypes/error1.JPGType 2:
https://github.com/tlyim/DynSecRetrieval/blob/main/errorTypes/error2.JPGType 3:
https://github.com/tlyim/DynSecRetrieval/blob/main/errorTypes/error3.JPGOddly, when I executed the notebook on Github Codespaces (with 2-core, 8GB RAM) using LlamaParse without Premium Mode (i.e., with the option
premium_mode=True commented out in a chunk), I would get only an error message of Type 1 above. However, when this was done with the Premium Mode, I would again get random error messages but only of Type 1 and 2 above.
I'd appreciate very much any help/suggestion on understanding why the error message was not consistently the same and how it could be fixed. Thanks.
