Add pydocstyle to the ruff rules (#52136)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com> Signed-off-by: Harry Mellor <hej.mellor@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
co-authored by
Claude Opus 5
parent
804e5377aa
commit
8c1557a79c
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
Demonstrates async reinforcement learning using vLLM and Ray,
|
||||
"""Demonstrates async reinforcement learning using vLLM and Ray,
|
||||
with native weight syncing APIs and batch-invariant generation.
|
||||
|
||||
The script separates training and inference workloads onto distinct GPUs
|
||||
|
||||
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
RLHF weight syncing against a `vllm serve` HTTP server, using CUDA IPC for the
|
||||
"""RLHF weight syncing against a `vllm serve` HTTP server, using CUDA IPC for the
|
||||
data plane.
|
||||
|
||||
* OpenAI-compatible API for inference requests
|
||||
|
||||
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
RLHF weight syncing against a `vllm serve` HTTP server, using NCCL for the
|
||||
"""RLHF weight syncing against a `vllm serve` HTTP server, using NCCL for the
|
||||
data plane.
|
||||
|
||||
* OpenAI-compatible API for inference requests
|
||||
|
||||
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
RLHF with FSDP2 training and vLLM expert-parallel inference using **CUDA IPC**
|
||||
"""RLHF with FSDP2 training and vLLM expert-parallel inference using **CUDA IPC**
|
||||
weight transfer and **packed** tensors.
|
||||
|
||||
Multi-rank version of `rlhf_http_ipc.py`: the trainer is 4 FSDP2 Ray actors
|
||||
|
||||
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
RLHF with FSDP2 training (4 GPUs) and vLLM expert-parallel inference (4 GPUs).
|
||||
"""RLHF with FSDP2 training (4 GPUs) and vLLM expert-parallel inference (4 GPUs).
|
||||
|
||||
8-GPU layout:
|
||||
Training — 4 GPUs, PyTorch FSDP2 (fully_shard), as Ray actors
|
||||
@@ -74,8 +73,7 @@ BASE_URL = f"http://localhost:{SERVER_PORT}"
|
||||
|
||||
@ray.remote(num_gpus=1)
|
||||
class FSDPTrainWorker:
|
||||
"""
|
||||
One FSDP2 training worker per GPU. Four of these form the FSDP group.
|
||||
"""One FSDP2 training worker per GPU. Four of these form the FSDP group.
|
||||
Rank 0 additionally drives weight transfer to the vLLM server.
|
||||
"""
|
||||
|
||||
|
||||
@@ -101,7 +101,8 @@ PROMPTS = ["The capital of France is", "The future of AI is"]
|
||||
class FSDPTrainWorker:
|
||||
"""One FSDP2 rank per GPU. Identical to the 8-GPU example's worker: the
|
||||
sharded-RDT engine owns the NIXL serve surface, so this stays a plain Ray
|
||||
actor with no producer mixin."""
|
||||
actor with no producer mixin.
|
||||
"""
|
||||
|
||||
def __init__(self, model_name, rank, world_size, master_addr, master_port):
|
||||
self.rank = rank
|
||||
|
||||
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
End-to-end example for routed experts capture with hybrid models.
|
||||
"""End-to-end example for routed experts capture with hybrid models.
|
||||
|
||||
Validates that:
|
||||
1. routed_experts is returned in CompletionOutput for MoE models.
|
||||
|
||||
Reference in New Issue
Block a user