vLLM upgrade

README.md

@@ -195,9 +195,9 @@ cd PipelineRL
 
 Create the environments with dependencies.
 ```bash
-conda create -n pipeline-rl -y python=3.11
-conda run --no-capture-output -n pipeline-rl pip install torch==2.6.0 
-conda run --no-capture-output -n pipeline-rl pip install -e . --no-build-isolation
+conda create -n pipeline-rl -y python=3.12
+conda run --no-capture-output -n pipeline-rl pip install -e .
+conda run --no-capture-output -n pipeline-rl pip install flash-attn==2.8.3 --no-build-isolation
 ```
 
 By default Pipeline-RL will use the file system as the medium for streaming the generated data to the trainer processes. This works on one node, but the files can get quite large. To use Redis instead you will need to install the Redis server in the same conda environment:

conf/base.yaml

@@ -57,14 +57,11 @@ test_llm:
     top_k: 50
 
 vllm_config:
-  use_v1: false
+  use_v1: true
   quantization: null  # or bf16_last_layer_fp32
   vllm_kwargs:
     dtype: bfloat16
     gpu-memory-utilization: 0.9
-    num-scheduler-steps: 1
-    disable-log-requests: ""
-    disable-frontend-multiprocessing: ""
     max-num-seqs: ${actor.llm_max_rollouts}
     max-num-batched-tokens: 1024
     enable-chunked-prefill: ""
@@ -73,6 +70,12 @@ vllm_config:
     pipeline-parallel-size: 1
     generation-config: vllm
     max_model_len: 10000
+    # V1 specific settings
+    # logprobs-mode: processed_logprobs
+    # V0 specific settings
+    # num-scheduler-steps: 1
+    # disable-log-requests: ""
+    # disable-frontend-multiprocessing: ""
 
 world:
   replicas: 1

pipelinerl/finetune/checkpoints.py

@@ -179,7 +179,7 @@ def load_model(args, model_class, current_dir):
         is_ds_zero_3 = get_accelerator().state.deepspeed_plugin.zero_stage == 3  # type: ignore
 
     if args.load_as_bf16:
-        loading_args["torch_dtype"] = torch.bfloat16
+        loading_args["dtype"] = torch.bfloat16
     if args.auto_device_map:
         loading_args["device_map"] = "auto"
     model_cls = get_auto_model_class(model_class)

pipelinerl/finetune/lora.py

@@ -147,7 +147,7 @@ def merge_lora(lora_model_path):
     assert os.path.exists(lora_model_config), f"{lora_model_config} does not exists"
 
     logger.info(f"Merge lora checkpoint {lora_model_path}")
-    model = lora_load_and_merge(lora_model_path, torch_dtype=torch.bfloat16, low_cpu_mem_usage=True)
+    model = lora_load_and_merge(lora_model_path, dtype=torch.bfloat16, low_cpu_mem_usage=True)
     tokenizer = AutoTokenizer.from_pretrained(lora_model_path)
 
     tmp_dir = f"{lora_model_path}_merged"

pipelinerl/finetune_loop.py

@@ -27,8 +27,9 @@
 from ring_flash_attn import substitute_hf_flash_attn, update_ring_flash_attn_params
 
 from pipelinerl.finetune.value_model import AutoModelForCausalLMWithValueHead
-import pipelinerl.torch_utils
+from pipelinerl import torch_utils
 from pipelinerl.finetune.types import PipelineBatchEncoding
+
 from pipelinerl.finetune.checkpoints import (
     load_model,
     load_tokenizer,
@@ -212,7 +213,8 @@ def send_weight_update(
             for name, parameter in named_parameters.items():
                 with deepspeed.zero.GatheredParameters([parameter]):
                     if get_accelerator().is_main_process:
-                        dist.broadcast(parameter.data, src=0, group=self.actor_update_group)
+                        # Use PyNcclCommunicator's broadcast method instead of torch.distributed
+                        self.actor_update_group.broadcast(parameter.data, src=0, stream=torch.cuda.current_stream())
             if get_accelerator().is_main_process:
                 logger.info("Wait for HTTP requests")
                 for future in futures:  # type: ignore
@@ -254,8 +256,8 @@ def send_weight_update(
                 futures = self.request_weight_updates(messages)
                 logger.info(f"Published weight update request for version {version}")
                 for _, parameter in named_parameters.items():
-                    dist.broadcast(parameter.data, src=0, group=self.actor_update_group)
-                dist.barrier(self.actor_update_group)
+                    # Use PyNcclCommunicator's broadcast method instead of torch.distributed
+                    self.actor_update_group.broadcast(parameter.data, src=0, stream=torch.cuda.current_stream())
                 for future in futures:
                     future.result()
                 logger.info("Finished broadcasting weights")
@@ -408,13 +410,15 @@ def run_finetuning_loop(
     get_accelerator().wait_for_everyone()
 
     if get_accelerator().is_main_process and args.send_weight_updates:
-        logger.info("Initializing actor process group")
-        actor_update_group = pipelinerl.torch_utils.init_extra_process_group(
-            group_name="actor",
-            backend="nccl",
+        current_device = get_accelerator().device
+        torch.cuda.set_device(current_device)
+        logger.info("Initializing actor process group using StatelessProcessGroup")
+        logger.info(f"Set CUDA device to {current_device} for actor process group (rank 0)")
+        actor_update_group = torch_utils.stateless_init_process_group(
             init_method=cfg.me.weight_update_group_init_method,
             rank=0,
             world_size=cfg.me.weight_update_group_world_size,
+            device=current_device,
         )
         logger.info("Actor process group initialized")
     else:
@@ -493,8 +497,7 @@ def run_finetuning_loop(
     finally:
         if weight_update_manager is not None:
             weight_update_manager.shutdown()
-        if actor_update_group:
-            dist.destroy_process_group(actor_update_group)
+        # PyNcclCommunicator doesn't need explicit destroy like torch.distributed process groups
 
 
 def rl_finetuning_worker(

pipelinerl/torch_utils.py

@@ -1,14 +1,51 @@
+import logging
 from datetime import timedelta
 from typing import Any, Optional, Union
+from urllib.parse import urlparse
+
+import torch
+import torch.distributed as dist
 from torch.distributed.distributed_c10d import (
     Backend,
     PrefixStore,
+    ProcessGroupNCCL,
     Store,
     _new_process_group_helper,
     _world,
     default_pg_timeout,
     rendezvous,
 )
+from vllm.distributed.device_communicators.pynccl import PyNcclCommunicator
+from vllm.distributed.utils import StatelessProcessGroup
+
+logger = logging.getLogger(__name__)
+
+
+def stateless_init_process_group(init_method, rank, world_size, device):
+    """
+    vLLM provides `StatelessProcessGroup` to create a process group
+    without considering the global process group in torch.distributed.
+    It is recommended to create `StatelessProcessGroup`, and then initialize
+    the data-plane communication (NCCL) between external (train processes)
+    and vLLM workers.
+
+    Args:
+        init_method: TCP init method string (e.g., "tcp://localhost:9000")
+        rank: The rank of this process in the group
+        world_size: Total number of processes in the group
+        device: The CUDA device to use for NCCL communication
+    """
+    # Parse master_address and master_port from init_method (e.g., "tcp://localhost:9000")
+    parsed = urlparse(init_method)
+    master_address = parsed.hostname or "localhost"
+    master_port = parsed.port or 9000
+    logger.debug(f"Parsed master_address: {master_address}, master_port: {master_port}")
+
+    pg = StatelessProcessGroup.create(
+        host=master_address, port=master_port, rank=rank, world_size=world_size
+    )
+    pynccl = PyNcclCommunicator(pg, device=device)
+    return pynccl
 
 
 # Copy from pytorch to allow creating multiple main groups.
@@ -49,6 +86,13 @@ def init_extra_process_group(
         # different systems (e.g. RPC) in case the store is multi-tenant.
         store = PrefixStore(group_name, store)
 
+    # Create NCCL-specific options if using NCCL backend
+    logger.info(f"[{group_name}] Backend: {backend}, str(backend): {str(backend)}")
+    if pg_options is None and str(backend) == "nccl":
+        pg_options = ProcessGroupNCCL.Options()
+        pg_options.is_high_priority_stream = False
+        logger.info(f"[{group_name}] Created NCCL options: {pg_options}")
+
     pg, _ = _new_process_group_helper(
         world_size,
         rank,
@@ -59,6 +103,7 @@ def init_extra_process_group(
         backend_options=pg_options,
         timeout=timeout,
     )
+    logger.info(f"[{group_name}] Process group created successfully")
 
     _world.pg_group_ranks[pg] = {i: i for i in range(world_size)}
 

pipelinerl/vllm0.py

@@ -1,3 +1,36 @@
+"""
+DEPRECATED - Kept only for backward compatibility with older vLLM versions.
+
+This module provides a custom vLLM inference server with dynamic weight updates using the legacy V0 engine architecture.
+
+Compatibility:
+    - vLLM versions <= 0.8.x only
+    - The V0 engine was removed in vLLM 0.11.0
+    - Use vllm1.py instead
+"""
+import warnings
+from packaging import version as version_parser
+import vllm
+
+# Check vLLM version compatibility
+vllm_version = version_parser.parse(vllm.__version__)
+
+if vllm_version >= version_parser.parse("0.9.0"):
+    raise ImportError(
+        f"pipelinerl.vllm0 is not compatible with vLLM {vllm.__version__}. "
+        "This module only works with vLLM <= 0.8.x. "
+        "Please use pipelinerl.vllm1 for vLLM >= 0.11.0 instead."
+    )
+
+# Only show deprecation warning for compatible versions
+warnings.warn(
+    "pipelinerl.vllm0 is DEPRECATED and will be removed in a future version. "
+    "This module only works with vLLM <= 0.8.x. "
+    "Please use pipelinerl.vllm1 as it is actively maintained.",
+    DeprecationWarning,
+    stacklevel=2,
+)
+
 import asyncio
 import json
 import logging
@@ -14,7 +47,6 @@
 )
 from vllm.entrypoints.launcher import serve_http
 from vllm.entrypoints.openai.api_server import (
-    run_server,
     create_server_socket,
     build_app,
     init_app_state,

pipelinerl/vllm1.py

@@ -1,15 +1,16 @@
+import asyncio
 import logging
 import signal
 import torch
 import uvloop
-from vllm.utils import FlexibleArgumentParser, set_ulimit
+from vllm.utils.argparse_utils import FlexibleArgumentParser
+from vllm.utils.system_utils import set_ulimit
 from vllm.entrypoints.openai.cli_args import (
     make_arg_parser,
     validate_parsed_serve_args,
 )
 from vllm.entrypoints.launcher import serve_http
 from vllm.entrypoints.openai.api_server import (
-    run_server,
     create_server_socket,
     build_app,
     init_app_state,
@@ -26,8 +27,8 @@
 
 from pipelinerl.finetune_loop import WeightUpdateRequest
 from pipelinerl.vllm_quantization import string_to_dtype  # reuse mapping
+from pipelinerl.torch_utils import stateless_init_process_group
 from typing import Any, Protocol, runtime_checkable
-import pipelinerl.torch_utils
 import pipelinerl.vllm_quantization  # Register bf16_last_layer_fp32 quantization config
 
 logger = logging.getLogger(__name__)
@@ -46,7 +47,7 @@ class LikeWorker(Protocol):
     rank: int
     local_rank: int
     device: torch.device
-    model_runner: GPUModelRunner 
+    model_runner: GPUModelRunner
     pg_rank: int
     process_group: Any
     model_config: ModelConfig
@@ -72,34 +73,47 @@ def init_actor_update_group(
             prefix
             + f"Weight update group init method: {weight_update_group_init_method}, world size: {weight_update_group_world_size}"
         )
-        self.process_group = pipelinerl.torch_utils.init_extra_process_group(
-            group_name="actor",
-            backend="nccl",
+
+        # Use vLLM's StatelessProcessGroup instead of torch.distributed
+        self.model_update_group = stateless_init_process_group(
             init_method=weight_update_group_init_method,
             rank=self.pg_rank,
             world_size=weight_update_group_world_size,
+            device=self.device,
         )
+        logger.info(prefix + "Actor update process group initialized")
 
-    def receive_weight_update(self: LikeWorker, request: WeightUpdateRequest):
+    def receive_weight_update(self: LikeWorker, request_json: str):
+        request = WeightUpdateRequest.model_validate_json(request_json)
         torch.cuda.synchronize(self.device)
         logger.info("Start receiving weight update")
         expected_dtypes = (torch.bfloat16, torch.float32, torch.float16)
+
         for info in request.parameters_info:
             target_dtype = string_to_dtype(info.dtype)
             if target_dtype not in expected_dtypes:
                 logger.warning(f"Unexpected dtype for {info.name}: {info.dtype}")
             buffer = torch.empty(tuple(info.shape), dtype=target_dtype, device=self.device)
-            torch.distributed.broadcast(buffer, src=0, group=self.process_group)
-            loaded_params = self.model_runner.model.load_weights(weights=[(info.name, buffer)]) # type: ignore
+            self.model_update_group.broadcast(buffer, src=0, stream=torch.cuda.current_stream())
+            loaded_params = self.model_runner.model.load_weights(weights=[(info.name, buffer)])  # type: ignore
             if len(loaded_params) != 1:
                 raise ValueError(f"model {info.name} not found in model state dict")
+
         pipelinerl.vllm_quantization.invalidate_fp32_cache()
         logger.info("Weight update received")
 
+    def close_communicator(self):
+        """Closes the communicator when weight synchronization is no longer needed."""
+        if hasattr(self, "model_update_group") and self.model_update_group is not None:
+            del self.model_update_group
+            self.model_update_group = None
+            logger.info("Weight update communicator closed")
+
 
 class WeightUpdateManager:
-    def __init__(self, args, engine_client: AsyncMPClient):
+    def __init__(self, args, engine: AsyncLLM, engine_client: AsyncMPClient):
         self.args = args
+        self.engine = engine
         self.engine_client = engine_client
 
     async def input_process_groups(self):
@@ -114,11 +128,16 @@ async def input_process_groups(self):
         )
 
     async def receive_weight_update(self, request: WeightUpdateRequest):
+        logger.info("Starting weight update...")
         await self.engine_client.collective_rpc_async(
-            "receive_weight_update", args=(request,)
+            "receive_weight_update", args=(request.model_dump_json(),)
         )
         logger.info("Weight update processed")
 
+    async def close_communicator(self):
+        """Closes the communicator when weight synchronization is no longer needed."""
+        await self.engine_client.collective_rpc_async("close_communicator")
+
 
 async def run_server(args, **uvicorn_kwargs) -> None:
     # COPIED FROM vllm/entrypoints/openai/api_server.py, vllm version 0.6.6.post1
@@ -157,11 +176,11 @@ def signal_handler(*_) -> None:
         vllm_config=engine_config,
         usage_context=UsageContext.OPENAI_API_SERVER,
         disable_log_stats=engine_args.disable_log_stats,
-        disable_log_requests=engine_args.disable_log_requests,
+        enable_log_requests=engine_args.enable_log_requests,
     )
     assert isinstance(engine.engine_core, AsyncMPClient)
 
-    weight_update_manager = WeightUpdateManager(args, engine.engine_core)
+    weight_update_manager = WeightUpdateManager(args, engine, engine.engine_core)
     if not args.disable_weight_updates:
         await weight_update_manager.input_process_groups()
 
@@ -172,10 +191,12 @@ def signal_handler(*_) -> None:
 
     @app.post("/receive_weight_update")
     async def _receive_weight_update(request: WeightUpdateRequest):
-        await weight_update_manager.receive_weight_update(request)
+        # Fire-and-forget: return immediately, weight update happens in background
+        logger.info("Received weight update request (fire-and-forget)")
+        asyncio.create_task(weight_update_manager.receive_weight_update(request))
         return {"status": "ok"}
 
-    await init_app_state(engine, engine_config, app.state, args)
+    await init_app_state(engine, app.state, args)
     shutdown_task = await serve_http(
         app,
         sock,
@@ -194,11 +215,11 @@ async def _receive_weight_update(request: WeightUpdateRequest):
     # NB: Await server shutdown only after the backend context is exited
     await shutdown_task
 
+    # Cleanup
+    if not args.disable_weight_updates:
+        await weight_update_manager.close_communicator()
     sock.close()
 
-    # TODO: proper cleanup
-    # dist.destroy_process_group(actor_update_group)
-
 
 def run_llm():
     parser = FlexibleArgumentParser(description="vLLM OpenAI-Compatible RESTful API server.")