-
Notifications
You must be signed in to change notification settings - Fork 574
All issues
Issue creation is restricted in this repository
Issues
is:issue state:open
is:issue state:open
Search results
exact_answer_alphanumeric_reward returns 1.0 when both the answer and the ground truth normalize to an empty string
bugSomething isn't workingSomething isn't workingStatus: Open.#4275 In NVIDIA-NeMo/RL;Leader broadcast corrupts noncontiguous tensor values
bugSomething isn't workingSomething isn't workingStatus: Open.#4262 In NVIDIA-NeMo/RL;Support DeepSeek V4.1 Flash
enhancementNew feature or requestNew feature or requestStatus: Open.#4246 In NVIDIA-NeMo/RL;Nemotron-3-Nano-30B-A3B: train/generation logprob divergence persists on vLLM 0.20
bugSomething isn't workingSomething isn't workingDocumentationImprovements or additions to documentationImprovements or additions to documentationStatus: Open.#4243 In NVIDIA-NeMo/RL;Normalize the in-loss survivor denominator once instead of rescaling gradients by G/K
enhancementNew feature or requestNew feature or requestStatus: Open.#4242 In NVIDIA-NeMo/RL;Track OPD loss refactoring and full-vocabulary configuration documentation
DocumentationImprovements or additions to documentationImprovements or additions to documentationStatus: Open.#4241 In NVIDIA-NeMo/RL;DTensor v2 fails to load Nemotron-H checkpoints (backbone.embedding vs backbone.embeddings)
bugSomething isn't workingSomething isn't workingStatus: Open.#4211 In NVIDIA-NeMo/RL;Evaluation-path processors render a duplicated default system block (two chat-template calls)
bugSomething isn't workingSomething isn't workingStatus: Open.#4185 In NVIDIA-NeMo/RL;- Status: Open.#4178 In NVIDIA-NeMo/RL;
[Feature] Provide a validated Megatron GRPO recipe for the original Nemotron 3.5 Super VL checkpoint
Status: Open.#4175 In NVIDIA-NeMo/RL;[Bug] select_teacher can pick different teachers after changing only the microbatch size
bugSomething isn't workingSomething isn't workingwaiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respondStatus: Open.#4173 In NVIDIA-NeMo/RL;[Bug] Dynamic CE–KD scaling changes when the same training batch is split into microbatches
bugSomething isn't workingSomething isn't workingwaiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respondStatus: Open.#4172 In NVIDIA-NeMo/RL;