Missing of the scripts of Training and Reproducibility

[Wizard759]

Hi authors,

Thanks for releasing DAPO — fantastic work and a very clear paper!

While trying to run the code, I noticed the “Training” link out-of-box in the README appears to be unavailable (404 / not accessible). Screenshots are attached below.

Could you please:

• share a working link to the training implementation, or
• point me to the corresponding entry points in VERL (configs + trainer) needed to run DAPO?

Many thanks for your time and for open-sourcing this work!

[dmndxld]

check verl repo, verl/recipe/dapo

[ksrinivs64]

Hi thanks for some very nice work. I am trying to just test the code on 2 or 1 GPUs and and I've modified your script to run on Qwen2.5-0.5B, but keep running into a problem in initialization.
` File "/tmp/ray/session_2025-12-15_12-10-35_410172_1449906/runtime_resources/working_dir_files/_ray_pkg_9eeef69c01829797/recipe/dapo/main_dapo.py", line 33, in main
run_ppo(config)
File "/tmp/ray/session_2025-12-15_12-10-35_410172_1449906/runtime_resources/working_dir_files/_ray_pkg_9eeef69c01829797/recipe/dapo/main_dapo.py", line 62, in run_ppo
ray.get(runner.run.remote(config))
File "/dccstor/m3/miniforge3/envs/verl/lib/python3.12/site-packages/ray/_private/auto_init_hook.py", line 22, in auto_init_wrapper
return fn(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^
File "/dccstor/m3/miniforge3/envs/verl/lib/python3.12/site-packages/ray/_private/client_mode_hook.py", line 104, in wrapper
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/dccstor/m3/miniforge3/envs/verl/lib/python3.12/site-packages/ray/_private/worker.py", line 2972, in get
values, debugger_breakpoint = worker.get_objects(
^^^^^^^^^^^^^^^^^^^
File "/dccstor/m3/miniforge3/envs/verl/lib/python3.12/site-packages/ray/_private/worker.py", line 1031, in get_objects
raise value.as_instanceof_cause()
ray.exceptions.RayTaskError(ActorDiedError): ray::TaskRunner.run() (pid=1481520, ip=9.47.193.138, actor_id=34c2443581502c900c3fd11a03000000, repr=<main_dapo.TaskRunner object at 0x14b25d7f8260>)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/tmp/ray/session_2025-12-15_12-10-35_410172_1449906/runtime_resources/working_dir_files/_ray_pkg_9eeef69c01829797/recipe/dapo/main_dapo.py", line 178, in run
trainer.init_workers()
File "/tmp/ray/session_2025-12-15_12-10-35_410172_1449906/runtime_resources/working_dir_files/_ray_pkg_9eeef69c01829797/verl/trainer/ppo/ray_trainer.py", line 773, in init_workers
self.actor_rollout_wg.init_model()
File "/tmp/ray/session_2025-12-15_12-10-35_410172_1449906/runtime_resources/working_dir_files/_ray_pkg_9eeef69c01829797/verl/single_controller/ray/base.py", line 53, in call
output = ray.get(output)
^^^^^^^^^^^^^^^
^^^^^^^^^^^^^^^^^^^
^^^^^^^^^^^^^^^^^^^^^
^^^^^^^^^^^^^^^^^^^
ray.exceptions.ActorDiedError: The actor died because of an error raised in its creation task, ray::rVcVMqWorkerDict_0:0:WorkerDict.init() (pid=1485957, ip=9.47.193.138, actor_id=d6e61de1862cb17b4c7c857b03000000, repr=<verl.single_controller.ray.base.WorkerDict object at 0x143ce1669250>)
File "/dccstor/m3/miniforge3/envs/verl/lib/python3.12/concurrent/futures/_base.py", line 449, in result
return self.__get_result()
^^^^^^^^^^^^^^^^^^^
File "/dccstor/m3/miniforge3/envs/verl/lib/python3.12/concurrent/futures/_base.py", line 401, in __get_result
raise self._exception
^^^^^^^^^^^^^^^^^^^^^
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/tmp/ray/session_2025-12-15_12-10-35_410172_1449906/runtime_resources/working_dir_files/_ray_pkg_9eeef69c01829797/verl/single_controller/ray/base.py", line 918, in init
self.worker_dict[key] = user_defined_cls(
^^^^^^^^^^^^^^^^^
File "/tmp/ray/session_2025-12-15_12-10-35_410172_1449906/runtime_resources/working_dir_files/_ray_pkg_9eeef69c01829797/verl/workers/fsdp_workers.py", line 167, in init
self.ulysses_device_mesh = init_device_mesh(
^^^^^^^^^^^^^^^^^
File "/dccstor/m3/miniforge3/envs/verl/lib/python3.12/site-packages/torch/distributed/device_mesh.py", line 1046, in init_device_mesh
device_mesh = DeviceMesh(
^^^^^^^^^^^
File "/dccstor/m3/miniforge3/envs/verl/lib/python3.12/site-packages/torch/distributed/device_mesh.py", line 463, in init
self._init_process_groups()
File "/dccstor/m3/miniforge3/envs/verl/lib/python3.12/site-packages/torch/distributed/device_mesh.py", line 552, in _init_process_groups
pg_ranks_by_dim = self.mesh.swapdims(-1, dim).reshape(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: cannot reshape tensor of 0 elements into shape [-1, 0] because the unspecified dimension size -1 can be any value and is ambiguous``

Any idea what might be wrong?