apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-12-24 13:28:13 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

FIX] Fix CUDA error(700): 'cudaErrorIllegalAddress' in CascadeAppendWriteCacheKVQKV cache_kernel(). Continue when batch_id_per_token[token_idx] is default value -1. (#4218)

Browse Source
This commit is contained in:
Yohanna
2025-09-24 14:08:49 +08:00
committed by GitHub
parent f1b5392e20
commit 44010cee13
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
custom_ops/gpu_ops/append_attn/encoder_write_cache_with_rope_impl.cuh
View File

@@ -1004,7 +1004,8 @@ __global__ void cache_kernel(
const uint32_t qkv_bias = bias % hidden_size;
const uint32_t hi = qkv_bias / head_size;
const uint32_t h_bias = qkv_bias % head_size;
const uint32_t ori_bi = batch_id_per_token[token_idx];
const int32_t ori_bi = batch_id_per_token[token_idx];
if (ori_bi == -1) continue; // skip batch_id_per_token[token_idx]=-1
if (seq_lens[ori_bi] == 0) continue;
const uint32_t ori_seq_id = (token_idx - cu_seqlens_q[ori_bi]) + seq_lens_decoder[ori_bi];