Skip to content

Issues: NVIDIA/NeMo-RL

Beta
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

103 Open 84 Closed
103 Open 84 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Issues list

OOM when trying to reproduce the grpo-deepscaler run bug Something isn't working
#456 opened May 29, 2025 by LeonMalteW
@abukharin3
1
gemma-3-4b-it failed on 'Gemma3TextModel' object has no attribute 'model' bug Something isn't working
#453 opened May 29, 2025 by yuki-666
@yuki-666
Random crashes of HFVerifyWorker bug Something isn't working
#451 opened May 28, 2025 by terrykong
@terrykong
4
Create a separate eval function to replace train + eval_mode=True
#448 opened May 27, 2025 by ashors1
@ashors1
Use async vllm for multi-turn and reward pipelining
#447 opened May 27, 2025 by parthchadha
@parthchadha
Support exp_manager.max_time_per_run to guarantee save before deadline
#444 opened May 27, 2025 by terrykong
init_process_group timeout on 32B model 16 nodes bug Something isn't working
#443 opened May 27, 2025 by yuki-666
Allow saving checkpoints in sft without running validation enhancement New feature or request
#441 opened May 23, 2025 by yfw
Feature to mark invalid outputs from the reward system
#431 opened May 21, 2025 by suhara
Sglang support
#429 opened May 21, 2025 by terrykong
Notebook series
#428 opened May 21, 2025 by terrykong
@terrykong
DAPO features
#425 opened May 20, 2025 by terrykong
Qwen3 Moe with Megatron backend help wanted Extra attention is needed
#424 opened May 20, 2025 by terrykong
[Feature] Add Reward Model Hosting as Environment
#421 opened May 20, 2025 by abukharin-nv
gemma-3-4b-it got nan probs_ratio in both FSDP1/FSDP2 bug Something isn't working
#419 opened May 20, 2025 by yuki-666
@yfw
3
[Feature] Explicit failure for unmatched model and checkpoints bug Something isn't working
#415 opened May 19, 2025 by KiddoZhu
@parthchadha
Support MoE models in FSDP2 enhancement New feature or request new model
#413 opened May 19, 2025 by yuki-666
Isolated ray worker creation doesn't allow us to patch functions easily anymore. bug Something isn't working
#408 opened May 16, 2025 by SahilJain314
@parthchadha
Create RL playbook as a follow up to DAPT enhancement New feature or request good first issue Good for newcomers
#404 opened May 16, 2025 by snowmanwwg
1
Allow proper vocab padding to permit training on 32+ nodes bug Something isn't working
#403 opened May 15, 2025 by alexandery-nvidia
DCP to HF script should also propagate the tokenizer bug Something isn't working
#395 opened May 15, 2025 by terrykong
@terrykong
Add non-colocated refit enhancement New feature or request
#394 opened May 15, 2025 by parthchadha
@parthchadha
[Feature] Add LLM as Judge Environment enhancement New feature or request
#392 opened May 15, 2025 by yashaswikarnati
@yashaswikarnati
Remove val_batch_size bug Something isn't working
#384 opened May 14, 2025 by parthchadha
@parthchadha
Gemma 27B OOM with dynamic batching (in get_logprobs) bug Something isn't working
#383 opened May 14, 2025 by terrykong
@jiemingz
ProTip! Updated in the last three days: updated:>2025-05-28.