biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Bugfix] Allow 64-bit integer values for LoRA IDs to avoid overflow/truncation (#27876)

Browse Source 
Signed-off-by: Madeesh Kannan <shadeMe@users.noreply.github.com>
This commit is contained in:
Madeesh Kannan
2025-10-31 17:58:42 +01:00
committed by GitHub
parent 0384aa7150
commit 675704ac01
2 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/v1/worker/gpu_input_batch.py
View File

@@ -204,7 +204,7 @@ class InputBatch:
self.num_accepted_tokens_cpu = self.num_accepted_tokens_cpu_tensor.numpy()
# lora related
self.request_lora_mapping = np.zeros((self.max_num_reqs,), dtype=np.int32)
self.request_lora_mapping = np.zeros((self.max_num_reqs,), dtype=np.int64)
self.lora_id_to_request_ids: dict[int, set[str]] = {}
self.lora_id_to_lora_request: dict[int, LoRARequest] = {}

2
vllm/v1/worker/tpu_input_batch.py
View File

@@ -139,7 +139,7 @@ class InputBatch:
self.min_tokens: dict[int, tuple[int, set[int]]] = {}
# lora related
self.request_lora_mapping = np.zeros((self.max_num_reqs,), dtype=np.int32)
self.request_lora_mapping = np.zeros((self.max_num_reqs,), dtype=np.int64)
self.lora_id_to_request_ids: dict[int, set[str]] = {}
self.lora_id_to_lora_request: dict[int, LoRARequest] = {}