xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-07-25 00:57:24 +08:00
Code Issues Packages Projects Releases Wiki Activity

random uuid

Browse Source 
Signed-off-by: Woosuk Kwon <woosuk@thinkingmachines.ai>
This commit is contained in:
Woosuk Kwon 2025-09-20 11:17:45 +00:00
parent 98ef239486
commit 158a46888e
1 changed files with 2 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
vllm/v1/worker/gpu/input_batch.py
View File

@ -11,6 +11,7 @@ import torch
import triton import triton
import triton.language as tl import triton.language as tl
from vllm.utils import random_uuid
from vllm.v1.utils import CpuGpuBuffer from vllm.v1.utils import CpuGpuBuffer
@ -80,7 +81,7 @@ class InputBatch:
device: torch.device, device: torch.device,
) -> "InputBatch": ) -> "InputBatch":
assert 0 < num_reqs <= num_tokens assert 0 < num_reqs <= num_tokens
req_ids = [f"req_{i}" for i in range(num_reqs)] req_ids = [f"req_{i}_{random_uuid()}" for i in range(num_reqs)]
idx_mapping_np = np.arange(num_reqs, dtype=np.int32) idx_mapping_np = np.arange(num_reqs, dtype=np.int32)
idx_mapping = torch.tensor(idx_mapping_np, device=device) idx_mapping = torch.tensor(idx_mapping_np, device=device)
num_scheduled_tokens = np.full(num_reqs, num_scheduled_tokens = np.full(num_reqs,