Issues

• [Important] NeMo-Aligner deprecation notice (see NeMo RL)

  NVIDIA/NeMo-Aligner#542 · Filter by author terrykong opened on May 16, 2025

• PPOTrainer can't be imported when the slurm job has more than one node

  bug
  NVIDIA/NeMo-Aligner#541 · Filter by author mrm-196 opened on May 1, 2025

• RuntimeError: Error(s) in loading state_dict for GPTModel

  bug
  NVIDIA/NeMo-Aligner#539 · Filter by author mrm-196 opened on Apr 17, 2025
  

• Fomula confusion in distill loss funtion

  NVIDIA/NeMo-Aligner#538 · Filter by author yspMing opened on Apr 14, 2025
  1

• ValueError: Expected a parent

  bug
  NVIDIA/NeMo-Aligner#535 · Filter by author starxa2 opened on Mar 18, 2025

• llama-70b SFT OSError: [Errno 5] Input/output error

  bug
  NVIDIA/NeMo-Aligner#502 · Filter by author songwangnlp opened on Feb 7, 2025

• Memory inefficiency when loading attention_mask, causing dataloader OOM with long context

  NVIDIA/NeMo-Aligner#488 · Filter by author shensimeteor opened on Jan 22, 2025
  1

• ImportError: cannot import name 'MoESubmodules' from 'megatron.core.transformer.moe.moe_layer'

  bug
  NVIDIA/NeMo-Aligner#483 · Filter by author Cppowboy opened on Jan 16, 2025

• Support other RL algorithms(REINFORCE, ReMax, RLOO, REINFORCE++)

  NVIDIA/NeMo-Aligner#481 · Filter by author Cppowboy opened on Jan 11, 2025
  3

• The version number is wrong

  bug
  NVIDIA/NeMo-Aligner#476 · Filter by author null-test-7 opened on Jan 6, 2025

• Out of Memory (OOM) During Training a LLaMA 7B Reward Model (8 A800 40GB GPUs)

  bug
  NVIDIA/NeMo-Aligner#444 · Filter by author qingyiaaaaa opened on Dec 11, 2024

• use lightning or pytorch-lightning

  bug
  NVIDIA/NeMo-Aligner#438 · Filter by author better629 opened on Dec 10, 2024
  6

• OOM when saving torch_dist checkpoint

  bug
  NVIDIA/NeMo-Aligner#436 · Filter by author Cppowboy opened on Dec 7, 2024
  5

• NeMo Trainer / PTL Based Trainer

  NVIDIA/NeMo-Aligner#426 · Filter by author ashvinnihalani opened on Nov 30, 2024

• Fix dev branch's build after PTL upgrade

  bug
  NVIDIA/NeMo-Aligner#418 · Filter by author terrykong opened on Nov 22, 2024
  

• Will you support online DPO?

  NVIDIA/NeMo-Aligner#414 · Filter by author Shiguang-Guo opened on Nov 21, 2024

• How can I use nvidia/Llama-3.1-Nemotron-70B-Reward-HF directly for inference?

  NVIDIA/NeMo-Aligner#360 · Filter by author arunasank opened on Oct 25, 2024
  6

• serve_reward_model goes down

  bug
  NVIDIA/NeMo-Aligner#351 · Filter by author AtsunoriFujita opened on Oct 18, 2024
  4

• attribute_annotate.py is not worked by KeyError: 'exceeded'

  bug
  NVIDIA/NeMo-Aligner#349 · Filter by author AtsunoriFujita opened on Oct 18, 2024

• Unable to pip install nemo-aligner

  NVIDIA/NeMo-Aligner#342 · Filter by author SCccc21 opened on Oct 11, 2024
  1

  

• [Question] Converting a Megatron-LM ckpt to nemo so we can use NeMo-Aligner for post-training

  NVIDIA/NeMo-Aligner#340 · Filter by author abgoswam opened on Oct 10, 2024

• Error during saving checkpoint with TensorRT-enabled PPO actor training

  bug
  NVIDIA/NeMo-Aligner#281 · Filter by author haizadinia opened on Sep 5, 2024
  3

• [Question] TransfomerEngine and Apex dependencies

  bug
  NVIDIA/NeMo-Aligner#278 · Filter by author peri044 opened on Sep 2, 2024

• make build_dataloader not take in cfg

  NVIDIA/NeMo-Aligner#273 · Filter by author gshennvm opened on Aug 28, 2024

• common class for aligner models

  NVIDIA/NeMo-Aligner#272 · Filter by author gshennvm opened on Aug 27, 2024