[Docs] Improve docs for RLHF co-location example (#20599)

Signed-off-by: Ricardo Decal <rdecal@anyscale.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

examples/offline_inference/rlhf_colocate.py

@@ -1,14 +1,31 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """
-a simple demonstration to show how to co-locate
+Demonstrates how to co-locate a vLLM inference worker and training
-vLLM worker with training actors on the same GPUs,
+actors on the same set of GPUs for reinforcement learning from human feedback
-for RLHF-like applications.
+(RLHF) workloads.
-The key points:
+
-- Control the placement of the vLLM workers with Ray, by setting
+Ray serves as the distributed execution framework in this example. Ray
-    VLLM_RAY_PER_WORKER_GPUS and VLLM_RAY_BUNDLE_INDICES properly.
+placement groups allocate both training actors and vLLM workers to the
-- Use cuda-ipc to pass tensors, since NCCL does not work when we have
+same GPU bundles, enabling fast, in-GPU communication between the two
-    multiple processes on the same GPU.
+components.
+
+The script shows how to do the following:
+
+* Configure environment variables (`VLLM_RAY_PER_WORKER_GPUS` and
+  `VLLM_RAY_BUNDLE_INDICES`) so that vLLM workers land on the desired
+  devices.
+* Exchange tensors between processes by means of CUDA inter-process
+  communication (IPC). CUDA IPC sidesteps NCCL limitations that occur
+  when multiple processes share a single GPU.
+
+Note that this example assumes a single-node cluster with four GPUs, but Ray
+supports multi-node clusters. vLLM expects exclusive use of the GPUs during
+its initialization for memory profiling. Residual GPU activity interferes
+with vLLM memory profiling and causes unexpected behavior.
+
+Learn more about Ray placement groups:
+https://docs.ray.io/en/latest/placement-groups.html
 """
 
 import os
@@ -22,13 +39,24 @@ from vllm import LLM
 
 
 class MyLLM(LLM):
-    def __init__(self, *args, bundle_indices: list, **kwargs):
+    """Configure the vLLM worker for Ray placement group execution.
-        # a hack to make the script work.
+
-        # stop ray from manipulating CUDA_VISIBLE_DEVICES
+    The constructor sets environment variables that allow multiple vLLM
-        # at the top-level
+    workers to share a single physical GPU and that encode the bundle
+    indices assigned by the placement group.
+
+    Args:
+        *args: Positional arguments forwarded to `vllm.LLM`.
+        bundle_indices (list[int]): Placement-group bundle indices
+            assigned to this worker.
+        **kwargs: Keyword arguments forwarded to `vllm.LLM`.
+    """
+
+    def __init__(self, *args, bundle_indices: list[int], **kwargs):
+        # Prevent Ray from manipulating the top-level CUDA_VISIBLE_DEVICES variable
+        # so that vLLM can its own device placement inside the worker.
         os.environ.pop("CUDA_VISIBLE_DEVICES", None)
-        # every worker will use 0.4 GPU, so that we can schedule
+        # Each worker uses 0.4 GPU so that two instances fit on the same GPUs.
-        # 2 instances on the same GPUs.
         os.environ["VLLM_RAY_PER_WORKER_GPUS"] = "0.4"
         os.environ["VLLM_RAY_BUNDLE_INDICES"] = ",".join(map(str, bundle_indices))
         print(f"creating LLM with bundle_indices={bundle_indices}")
@@ -36,17 +64,25 @@ class MyLLM(LLM):
 
 
 class RayTrainingActor:
+    """Training actor that hosts a Facebook OPT-125M model from Hugging Face.
+
+    The model is loaded onto the first GPU assigned to this actor, and expose
+    the CUDA IPC handles so that colocated vLLM workers can map tensors
+    directly.
+    """
+
     def __init__(self):
-        # ray will set CUDA_VISIBLE_DEVICES to the assigned GPUs
+        # Ray sets CUDA_VISIBLE_DEVICES to the GPUs assigned to this actor.
         from transformers import AutoModelForCausalLM
 
         self.model = AutoModelForCausalLM.from_pretrained("facebook/opt-125m")
         self.model.to("cuda:0")
+        # Zero out all the parameters.
         for name, p in self.model.named_parameters():
             p.data.zero_()
         torch.cuda.synchronize()
-        # the argument for get_device_uuid is the index
+        # The argument for `get_device_uuid` is the index of the GPU in the
-        # of the GPU in the visible devices.
+        # list of visible devices.
         from vllm.platforms import current_platform
 
         self.device_uuid = current_platform.get_device_uuid(0)
@@ -59,23 +95,23 @@ class RayTrainingActor:
 
         data = {}
         for name, p in self.model.named_parameters():
-            # the training actor might only have a subset of the weights
+            # A training actor might hold only a subset of the weights and may
-            # and need to all-gather the weights from all the actors.
+            # need to gather weights from other actors. For demonstration
-            # for demonstration, here we assume all training actors have
+            # purposes, each training actor owns the full weight set.
-            # the full weights.
             data[name] = reduce_tensor(p.detach())
         return {self.device_uuid: data}
 
 
-# ray manages 4 GPUs
+# Ray manages four GPUs.
+
 os.environ["CUDA_VISIBLE_DEVICES"] = "0,1,2,3"
 ray.init()
 
-# we want to co-locate vLLM instance and the training actor
+# Co-locate vLLM instances and training actors on the same set of GPUs:
-# on the same set of GPUs.
+#   * GPU 0 and 1: training actor 0, training actor 1, and vLLM instance 0
-# the placement plan is as follows:
+#     (tensor parallelism = 2).
-# GPU 0 and 1: training actor 0, 1, and vLLM instance 0 (with TP=2)
+#   * GPU 2 and 3: training actor 2, training actor 3, and vLLM instance 1
-# GPU 2 and 3: training actor 2, 3, and vLLM instance 1 (with TP=2)
+#     (tensor parallelism = 2).
 
 pg = placement_group([{"GPU": 1, "CPU": 0}] * 4)
 ray.get(pg.ready())
@@ -104,10 +140,8 @@ for bundle_index, training_actor in enumerate(training_actors):
     training_actor_device_ids.append(device_id)
 
 for i, bundle_indices in enumerate([[0, 1], [2, 3]]):
-    # IMPORTANT: when creating vLLM instances, we need to
+    # Use the following syntax instead of the @ray.remote decorator so that
-    # make sure there are no GPU activities on the target GPUs,
+    # the placement group is customized for each bundle.
-    # otherwise, they will interfere with the vLLM memory profiling,
-    # and cause unexpected behaviors.
     llm = ray.remote(
         num_cpus=0,
         num_gpus=0,
@@ -125,8 +159,8 @@ for i, bundle_indices in enumerate([[0, 1], [2, 3]]):
         bundle_indices=bundle_indices,
     )
     inference_engines.append(llm)
-    # don't call any method on the inference engine here,
+    # Do not call any method on the inference engine at this point; the call
-    # otherwise it will block until the vLLM instance is created.
+    # blocks until the vLLM instance finishes initialization.
 
 for i, llm in enumerate(inference_engines):
     inference_engine_device_ids.append(
@@ -134,26 +168,25 @@ for i, llm in enumerate(inference_engines):
     )
     print(f"inference engine {i} is on {inference_engine_device_ids[-1]}")
 
-# check the placement
+# Verify placement: the first two training actors share the same GPUs as
-# the first two training actors should be
+# the first inference engine.
-# on the same GPUs as the first inference engine
 assert training_actor_device_ids[:2] == inference_engine_device_ids[0]
-# the last two training actors should be
+# Verify placement: the last two training actors share the same GPUs as
-# on the same GPUs as the second inference engine
+# the second inference engine.
 assert training_actor_device_ids[2:] == inference_engine_device_ids[1]
 
-print("gather all the IPC handles from the training actors")
+print("Gather all the IPC handles from the training actors.")
 ipc_handles = {}
 for actor in training_actors:
     ipc_handles.update(ray.get(actor.get_weight_ipc_handles.remote()))
 
-print("update the weights of the inference engines")
+print("Update the weights of the inference engines.")
 for llm in inference_engines:
     ray.get(
         llm.collective_rpc.remote(
             "update_weights_from_ipc_handles", args=(ipc_handles,)
         )
     )
-print("check if the weights are updated")
+print("Check if the weights are updated.")
 for llm in inference_engines:
     assert ray.get(llm.collective_rpc.remote("check_weights_changed", args=tuple()))