Issues

Video MME Reproduce (Question about think mode usage)
NVlabs/Long-RL#39 · opened on Dec 29, 2025
How should I run SFT using VILA codebase
NVlabs/Long-RL#38 · opened on Dec 25, 2025
LongVideoReason Eval Dataset
NVlabs/Long-RL#37 · opened on Dec 16, 2025
Where is the implementation of Stage1 parallel encoding (video frame uniform splitting)?
NVlabs/Long-RL#36 · opened on Dec 4, 2025
How to perform SFT on custom data
NVlabs/Long-RL#35 · opened on Nov 20, 2025
Any plan to support Qwen3-VL dense model?
NVlabs/Long-RL#34 · opened on Oct 24, 2025
2
How to train with multi images?
NVlabs/Long-RL#33 · opened on Sep 30, 2025
If this structure fit for NVILA-LITE-8B
NVlabs/Long-RL#32 · opened on Sep 29, 2025
problem when run example/new_support/qwen2.5_omni_3b_grpo.sh
NVlabs/Long-RL#31 · opened on Sep 23, 2025
1
Why log_probs.unsqueeze in compute_policy_loss?
NVlabs/Long-RL#30 · opened on Sep 18, 2025
The weights synchronizing of rollout
NVlabs/Long-RL#29 · opened on Sep 18, 2025
Can not run the Qwen-2.5-VL-7B model
NVlabs/Long-RL#28 · opened on Sep 10, 2025
Wan text encoder is loaded on CPU?
NVlabs/Long-RL#27 · opened on Sep 8, 2025
Why is ground_truth not included in non_tensor_batch_keys in ray_trainer.py?
NVlabs/Long-RL#26 · opened on Sep 2, 2025
Request for problem_type of training data
NVlabs/Long-RL#23 · opened on Aug 20, 2025
The config of micro_batch_size_per_device_for_update and micro_batch_size_per_device_for_experience
NVlabs/Long-RL#21 · opened on Aug 18, 2025
3
Using Slurm For Training, World Size Settings
NVlabs/Long-RL#20 · opened on Aug 13, 2025
3
Runing qwen2_5_vl_3b_video_grpo.sh issue
NVlabs/Long-RL#19 · opened on Aug 5, 2025
12
Environment requirements
NVlabs/Long-RL#16 · opened on Aug 3, 2025
13
LongVideo-Reason Dataset
NVlabs/Long-RL#1 · opened on Jul 11, 2025
16