LLMCompiler
Basic Information
LLMCompiler is a developer-focused framework for orchestrating parallel function calling with large language models. It automatically decomposes user-specified problems into tasks, identifies which tasks can run in parallel and which are interdependent, and computes an optimized orchestration for invoking user-provided functions (tools) during LLM reasoning. The project supports both open-source and closed-source models, provides example configurations and benchmarks, and includes scripts to run experiments and store results. It is intended to reproduce and extend the experiments from the associated paper, enabling lower latency, reduced API cost, and improved accuracy when multi-function calling is required. The repository includes installation instructions, runnable scripts, and example configs for HotpotQA, Movie Recommendation, and ParallelQA benchmarks.
Links
Categorization
App Details
-
$$$$ -
$$$$