AIBionics
AIBionics
When training Qwen, I encountered a CUDA error after few steps. I have set ```VLLM_ATTENTION_BACKEND to XFORMERS``` and confirmed that the ```environment variable``` is in effect. Before running ray start,...
I encountered an error while running the 7B Qwen2.5 model on a 2-node 16-H800-80G GRPO training setup using the latest version of the framework (250209). The error involves an **illegal...
Hello, I would like to propose a feature enhancement aimed at **improving the efficiency of reward calculation**, particularly for **time-consuming methods**. Currently, the system waits for **all rollouts** to complete...
It cost too much time to get adv when the output lenghth over 10k. ``` timing_s/adv:10751.941 ``` what might happen?
I've tried the method of [vLLMRollout.generate_sequences to implement tool calling](https://github.com/volcengine/verl/issues/176) with verl 0.2 and vllm 0.6.3, However, it randomly hangs after running for 1 to 2 steps. Specifically, the GPU...
``` Traceback (most recent call last): File "", line 198, in _run_module_as_main File "", line 88, in _run_code File "/tmp/ray/session_2025-03-03_19-25-20_906534_72705/runtime_resources/working_dir_files/_ray_pkg_cf22861b38986192/openrlhf/cli/train_ppo_ray.py", line 437, in train(args) File "/tmp/ray/session_2025-03-03_19-25-20_906534_72705/runtime_resources/working_dir_files/_ray_pkg_cf22861b38986192/openrlhf/cli/train_ppo_ray.py", line 179, in train...
#### This occurs when the ```--vllm_sync_with_ray``` flag is set. --- ``` (raylet) It looks like you're creating a detached actor in an anonymous namespace. In order to access this actor...
#674 端口冲突，按照 ``` def _get_free_port(): ports=[] while len(ports)