GFM-RAG Configuration¶

An example configuration file for GFM-RAG is shown below:

Example

gfmrag/workflow/config/stage3_qa_ircot_inference.yaml

hydra:
  run:
    dir: outputs/qa_agent_inference/${dataset.data_name}/${now:%Y-%m-%d}/${now:%H-%M-%S} # Output directory

defaults:
  - _self_
  - doc_ranker: idf_topk_ranker # The document ranker to use
  - agent_prompt: hotpotqa_ircot # The agent prompt to use
  - qa_prompt: hotpotqa # The QA prompt to use
  - ner_model: llm_ner_model # The NER model to use
  - el_model: colbert_el_model # The EL model to use
  - qa_evaluator: hotpotqa # The QA evaluator to use

seed: 1024

dataset:
  root: ./data # data root directory
  data_name: hotpotqa_test # data name

llm:
  _target_: gfmrag.llms.ChatGPT # The language model to use
  model_name_or_path: gpt-3.5-turbo # The model name or path
  retry: 5 # Number of retries

graph_retriever:
  model_path: rmanluo/GFM-RAG-8M # Checkpoint path of the pre-trained GFM-RAG model
  doc_ranker: ${doc_ranker} # The document ranker to use
  ner_model: ${ner_model} # The NER model to usek
  el_model: ${el_model} # The EL model to use
  qa_evaluator: ${qa_evaluator} # The QA evaluator to use
  init_entities_weight: True # Whether to initialize the entities weight

test:
  top_k: 10 # Number of documents to retrieve
  max_steps: 2 # Maximum number of steps
  max_test_samples: -1 # -1 for all samples
  resume: null # Resume from previous prediction

General Configuration¶

Parameter	Options	Note
`run.dir`	None	The output directory of the log

Defaults¶

Parameter	Options	Note
`ner_model`	None	The config of the ner_model
`el_model`	None	The config of the el_model
`doc_ranker`	None	The config of the doc_ranker
`qa_evaluator`	None	The config of the qa_evaluator
`agent_prompt`	None	The config of the PromptBuilder used for IRCOT
`qa_prompt`	None	The config of the PromptBuilder

Dataset¶

Parameter	Options	Note
`root`	None	The data root directory
`data_name`	None	The data name

LLM¶

Parameter	Options	Note
`_target_`	None	The language model to use
`model_name_or_path`	None	The model name or path
Additional parameters	None	Parameters to initialize a language model

Please refer to the LLMs page for more details.

Graph Retriever¶

Parameter	Options	Note
`_target_`	None	The graph retriever to use
`model_path`	None	Checkpoint path of the pre-trained GFM-RAG model
`doc_ranker`	None	The document ranker to use
`ner_model`	None	The NER model to use
`el_model`	None	The EL model to use
`qa_evaluator`	None	The QA evaluator to use
`init_entities_weight`	`True`,`False`	Whether to initialize the entities weight

Test¶

Parameter	Options	Note
`top_k`	None	Number of documents to retrieve
`max_steps`	None	Maximum number of steps, `1` for single step
`max_test_samples`	None	Maximum number of samples to test (`-1` for all samples)
`resume`	None	Resume from previous prediction