[BREAKING][core] move single_controller into verl directory

.github/workflows/yapf_format.yml

@@ -42,4 +42,4 @@ jobs:
           pip install toml==0.10.2
       - name: Running yapf
         run: |
-          yapf -r -vv -d --style=./.style.yapf verl tests single_controller examples
+          yapf -r -vv -d --style=./.style.yapf verl tests examples

README.md

@@ -180,7 +180,7 @@ pip3 install yapf
 ```
 Then, make sure you are at top level of verl repo and run
 ```bash
-yapf -ir -vv --style ./.style.yapf verl single_controller examples
+yapf -ir -vv --style ./.style.yapf verl examples
 ```
 
 

docs/advance/dpo_extension.rst

@@ -47,8 +47,8 @@ Implementation details:
 
 .. code:: python
 
-   from single_controller.base import Worker
-   from single_controller.ray import RayWorkerGroup, RayClassWithInitArgs, RayResourcePool
+   from verl.single_controller.base import Worker
+   from verl.single_controller.ray import RayWorkerGroup, RayClassWithInitArgs, RayResourcePool
    import ray
 
    @ray.remote
@@ -75,7 +75,7 @@ API: compute reference log probability
 
 .. code:: python
 
-   from single_controller.base import Worker
+   from verl.single_controller.base import Worker
    import ray
 
    @ray.remote
@@ -93,7 +93,7 @@ API: Update actor model parameters
 
 .. code:: python
 
-   from single_controller.base import Worker
+   from verl.single_controller.base import Worker
    import ray
 
    @ray.remote
@@ -184,7 +184,7 @@ registered into the worker_group**
 
 .. code:: python
 
-   from single_controller.base.decorator import register
+   from verl.single_controller.base.decorator import register
 
    def dispatch_data(worker_group, data):
        return data.chunk(worker_group.world_size)
@@ -214,11 +214,11 @@ computation, and data collection.
 
 Furthermore, the model parallelism size of each model is usually fixed,
 including dp, tp, pp. So for these common distributed scenarios, we have
-pre-implemented specific dispatch and collect methods,in `decorator.py <https://github.com/volcengine/verl/blob/main/single_controller/base/decorator.py>`_, which can be directly used to wrap the computations.
+pre-implemented specific dispatch and collect methods,in `decorator.py <https://github.com/volcengine/verl/blob/main/verl/single_controller/base/decorator.py>`_, which can be directly used to wrap the computations.
 
 .. code:: python
 
-   from single_controller.base.decorator import register, Dispatch
+   from verl.single_controller.base.decorator import register, Dispatch
 
    @register(dispatch_mode=Dispatch.DP_COMPUTE_PROTO)
    def generate_sequences(self, data: DataProto) -> DataProto:

docs/examples/ppo_code_architecture.rst

@@ -49,13 +49,13 @@ Define worker classes
    if config.actor_rollout_ref.actor.strategy == 'fsdp': # for FSDP backend
        assert config.actor_rollout_ref.actor.strategy == config.critic.strategy
        from verl.trainer.ppo.workers.fsdp_workers import ActorRolloutRefWorker, CriticWorker
-       from single_controller.ray import RayWorkerGroup
+       from verl.single_controller.ray import RayWorkerGroup
        ray_worker_group_cls = RayWorkerGroup
 
    elif config.actor_rollout_ref.actor.strategy == 'megatron': # for Megatron backend
        assert config.actor_rollout_ref.actor.strategy == config.critic.strategy
        from verl.trainer.ppo.workers.megatron_workers import ActorRolloutRefWorker, CriticWorker
-       from single_controller.ray.megatron import NVMegatronRayWorkerGroup
+       from verl.single_controller.ray.megatron import NVMegatronRayWorkerGroup
        ray_worker_group_cls = NVMegatronRayWorkerGroup # Ray worker class for Megatron-LM
 
    else:

docs/workers/megatron_workers.rst

@@ -40,7 +40,7 @@ We implement various of APIs for each ``Worker`` class decorated by the
 ``@register(dispatch_mode=)`` . These APIs can be called by the ray
 driver process. The data can be correctly collect and dispatch following
 the ``dispatch_mode`` on each function. The supported dispatch_model
-(i.e., transfer protocols) can be found in `decorator.py <https://github.com/volcengine/verl/blob/main/single_controller/base/decorator.py>`_.
+(i.e., transfer protocols) can be found in `decorator.py <https://github.com/volcengine/verl/blob/main/verl/single_controller/base/decorator.py>`_.
 
 ActorRolloutRefWorker
 ^^^^^^^^^^^^^^^^^^^^^

examples/ray/tutorial.ipynb

@@ -232,8 +232,8 @@
    },
    "outputs": [],
    "source": [
-    "from single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup, merge_resource_pool\n",
-    "from single_controller.base import Worker"
+    "from verl.single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup, merge_resource_pool\n",
+    "from verl.single_controller.base import Worker"
    ]
   },
   {
@@ -437,7 +437,7 @@
    },
    "outputs": [],
    "source": [
-    "from single_controller.ray.decorator import register, Dispatch, Execute"
+    "from verl.single_controller.ray.decorator import register, Dispatch, Execute"
    ]
   },
   {
@@ -518,7 +518,7 @@
    },
    "outputs": [],
    "source": [
-    "from single_controller.ray.decorator import register, Dispatch, collect_all_to_all, Execute"
+    "from verl.single_controller.ray.decorator import register, Dispatch, collect_all_to_all, Execute"
    ]
   },
   {
@@ -723,10 +723,10 @@
    },
    "outputs": [],
    "source": [
-    "from single_controller.ray.decorator import register, Dispatch, Execute\n",
-    "from single_controller.ray.megatron import NVMegatronRayWorkerGroup\n",
-    "from single_controller.base.megatron.worker import MegatronWorker\n",
-    "from single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup\n",
+    "from verl.single_controller.ray.decorator import register, Dispatch, Execute\n",
+    "from verl.single_controller.ray.megatron import NVMegatronRayWorkerGroup\n",
+    "from verl.single_controller.base.megatron.worker import MegatronWorker\n",
+    "from verl.single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup\n",
     "from omegaconf import OmegaConf\n",
     "from megatron.core import parallel_state as mpu"
    ]

examples/split_placement/main_ppo_split.py

@@ -121,13 +121,13 @@ def main_task(config):
     if config.actor_rollout_ref.actor.strategy == 'fsdp':
         assert config.actor_rollout_ref.actor.strategy == config.critic.strategy
         from verl.trainer.ppo.workers.fsdp_workers import ActorRolloutRefWorker, CriticWorker
-        from single_controller.ray import RayWorkerGroup
+        from verl.single_controller.ray import RayWorkerGroup
         ray_worker_group_cls = RayWorkerGroup
 
     elif config.actor_rollout_ref.actor.strategy == 'megatron':
         assert config.actor_rollout_ref.actor.strategy == config.critic.strategy
         from verl.trainer.ppo.workers.megatron_workers import ActorRolloutRefWorker, CriticWorker
-        from single_controller.ray.megatron import NVMegatronRayWorkerGroup
+        from verl.single_controller.ray.megatron import NVMegatronRayWorkerGroup
         ray_worker_group_cls = NVMegatronRayWorkerGroup
 
     else:

examples/split_placement/split_monkey_patch.py

@@ -16,7 +16,7 @@
 """
 import os
 from pprint import pprint
-from single_controller.ray import RayResourcePool, RayWorkerGroup, RayClassWithInitArgs
+from verl.single_controller.ray import RayResourcePool, RayWorkerGroup, RayClassWithInitArgs
 from verl import DataProto
 from verl.trainer.ppo.ray_trainer import compute_advantage, apply_kl_penalty, reduce_metrics, compute_data_metrics, Role, create_colocated_worker_cls
 from codetiming import Timer

tests/ray/check_worker_alive/main.py

@@ -18,9 +18,9 @@
 
 import ray
 
-from single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup
-from single_controller.base.worker import Worker
-from single_controller.base.decorator import register, Dispatch
+from verl.single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup
+from verl.single_controller.base.worker import Worker
+from verl.single_controller.base.decorator import register, Dispatch
 
 
 @ray.remote

tests/ray/detached_worker/client.py

@@ -19,8 +19,8 @@
 import torch
 
 from verl import DataProto
-from single_controller.ray import RayClassWithInitArgs
-from single_controller.ray.megatron import NVMegatronRayWorkerGroup
+from verl.single_controller.ray import RayClassWithInitArgs
+from verl.single_controller.ray.megatron import NVMegatronRayWorkerGroup
 
 from tensordict import TensorDict
 

tests/ray/detached_worker/server.py

@@ -25,10 +25,10 @@
 from torch import nn
 
 import ray
-from single_controller.ray import RayClassWithInitArgs, RayResourcePool
-from single_controller.ray.megatron import NVMegatronRayWorkerGroup
-from single_controller.base.megatron.worker import MegatronWorker
-from single_controller.ray.decorator import register, Dispatch
+from verl.single_controller.ray import RayClassWithInitArgs, RayResourcePool
+from verl.single_controller.ray.megatron import NVMegatronRayWorkerGroup
+from verl.single_controller.base.megatron.worker import MegatronWorker
+from verl.single_controller.ray.decorator import register, Dispatch
 from verl import DataProto
 from verl.models.llama.megatron import ParallelLlamaForCausalLMRmPadPP
 

tests/ray/test_colocated_workers.py

@@ -14,9 +14,9 @@
 
 import ray
 
-from single_controller.base import Worker
-from single_controller.base.decorator import register, Dispatch
-from single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup, create_colocated_worker_cls
+from verl.single_controller.base import Worker
+from verl.single_controller.base.decorator import register, Dispatch
+from verl.single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup, create_colocated_worker_cls
 
 from verl import DataProto
 

tests/ray/test_data_transfer.py

@@ -15,10 +15,10 @@
 In this test, we instantiate a data parallel worker with 8 GPUs
 """
 
-from single_controller.base import Worker
-from single_controller.ray import RayWorkerGroup, RayClassWithInitArgs, RayResourcePool
+from verl.single_controller.base import Worker
+from verl.single_controller.ray import RayWorkerGroup, RayClassWithInitArgs, RayResourcePool
 
-from single_controller.base.decorator import Dispatch, register
+from verl.single_controller.base.decorator import Dispatch, register
 
 import ray
 import torch

tests/ray/test_driverfunc_to_worker.py

@@ -18,9 +18,9 @@
 from verl import DataProto
 from tensordict import TensorDict
 
-from single_controller.base.worker import Worker
-from single_controller.ray.base import RayResourcePool, RayClassWithInitArgs
-from single_controller.ray import RayWorkerGroup
+from verl.single_controller.base.worker import Worker
+from verl.single_controller.ray.base import RayResourcePool, RayClassWithInitArgs
+from verl.single_controller.ray import RayWorkerGroup
 
 os.environ['RAY_DEDUP_LOGS'] = '0'
 os.environ['NCCL_DEBUG'] = 'WARN'

tests/ray/test_high_level_scheduling_api.py

@@ -16,8 +16,8 @@
 
 import ray
 
-from single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup, merge_resource_pool
-from single_controller.base.worker import Worker
+from verl.single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup, merge_resource_pool
+from verl.single_controller.base.worker import Worker
 
 
 @ray.remote

tests/ray/test_ray_local_envs.py

@@ -12,14 +12,14 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 """
-e2e test single_controller.ray
+e2e test verl.single_controller.ray
 """
 import os
 import ray
 
-from single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup
-from single_controller.base.worker import Worker
-from single_controller.ray.decorator import register, Dispatch, collect_all_to_all, Execute
+from verl.single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup
+from verl.single_controller.base.worker import Worker
+from verl.single_controller.ray.decorator import register, Dispatch, collect_all_to_all, Execute
 
 
 @ray.remote

tests/ray/test_remote_api.py

@@ -12,9 +12,9 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from single_controller.remote import remote, RemoteBackend, SharedResourcePool
-from single_controller.base.decorator import register, Dispatch
-from single_controller.base.worker import Worker
+from verl.single_controller.remote import remote, RemoteBackend, SharedResourcePool
+from verl.single_controller.base.decorator import register, Dispatch
+from verl.single_controller.base.worker import Worker
 
 
 @remote(process_on_nodes=[3], use_gpu=True, name_prefix="actor", sharing=SharedResourcePool)

tests/ray/test_worker_group_basics.py

@@ -12,15 +12,15 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 """
-e2e test single_controller.ray
+e2e test verl.single_controller.ray
 """
 
 import torch
 import ray
 
-from single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup
-from single_controller.base.worker import Worker
-from single_controller.ray.decorator import register, Dispatch, collect_all_to_all, Execute
+from verl.single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup
+from verl.single_controller.base.worker import Worker
+from verl.single_controller.ray.decorator import register, Dispatch, collect_all_to_all, Execute
 
 
 def two_to_all_dispatch_fn(worker_group, *args, **kwargs):

tests/ray/test_worker_group_torch.py

@@ -21,8 +21,8 @@
 import torch.distributed
 import ray
 
-from single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup
-from single_controller.base.worker import Worker
+from verl.single_controller.ray.base import RayResourcePool, RayClassWithInitArgs, RayWorkerGroup
+from verl.single_controller.base.worker import Worker
 
 
 @ray.remote