apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 08:37:06 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[FIX]fix rejection sampling when topp=0 using _SAMPLING_EPS (#2967)

Browse Source 
* fix rejection sampling when topp=0

* fix
This commit is contained in:
Sunny-bot1
2025-07-22 20:53:37 +08:00
committed by GitHub
parent dbe6225b33
commit 7c5e34e72d
3 changed files with 9 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
custom_ops/gpu_ops/sample_kernels/sampling.cuh
View File

@@ -292,7 +292,7 @@ __global__ void TopKTopPSamplingFromProbKernel(DType* probs, IdType* output,
curand_init(philox_seed, bx, philox_offset, &state);
const uint32_t row_idx = bx;
const uint32_t k = top_k_arr[row_idx] == 0 ? d : top_k_arr[row_idx];
const float p = top_p_arr[row_idx] == 0 ? 1e-6 : top_p_arr[row_idx];
const float p = top_p_arr[row_idx];
extern __shared__ __align__(
alignof(SamplingTempStorage<BLOCK_THREADS, SCAN_ALGORITHM, REDUCE_ALGORITHM>))