Back: Catalog

rllm

rLLM is an open-source framework that supports reinforced learning processes for language models. It allows users to train language agents on different test environments and sandbox solutions. With the help of a single flag, the distributed multi-GPU allows switching between different training infrastructures, such as a single machine.

RL

rllm

New80

rllm-org · AI Agent Frameworks

STARS5.9K
FORKS635
COMMITS0
CONTRIBUTORS0
VERSIONv0.3.0-pre
UPDATED14 days ago
PythonReinforcement LearningLLMAgent TrainingApache-2.0

Common commercial products in this category: OpenAI Assistants API, Microsoft Copilot Studio, Amazon Bedrock Agents

What can it do?
  • You can evaluate agents on different CLI test environments and integrated benchmarks.
  • Using sandbox options such as Docker, Daytona or Modal, you can reduce costs on the educational scale.
  • You can run multiple training methods such as GRPO, REINFORCE and SFT over a single API.
User reviews
  • • NeutralGitHub Issue

    🎉 Reproduce the results successfully, avg score 56.4

    “I have reproduced the results in the repo on an 8x80G A800, achieving an average score of **56.4**. Here is the huggingface hub's model: Thanks for your generous contribution of open-source code to the repository. Your efforts have made it possible for others like me to learn,…”

  • ↑ PositiveGitHub Issue

    eval for LCB

    “Great Work:) I use the eval.sh for eval on LCB(preview-1.5b|preview-14b), but the score is lower than the reported. Could you please share the command for reproduce the score for LCB for (preview-1.5b|preview-14b). ![Image]( ![Image](”

Website GitHub

My note

Alternative pages

OpenAI Assistants API alternativesMicrosoft Copilot Studio alternativesAmazon Bedrock Agents alternatives

More in: AI Agent Frameworks