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 deepep dispatch (#5036)

Browse Source 
* fix dispatch

* fix dispatch

---------

Co-authored-by: yuanxiaolan <yuanxiaolan01@baidu.com>
This commit is contained in:
yangjianfengo1
2025-11-17 10:34:01 +08:00
committed by GitHub
parent 3b80a799ab
commit 3afb717995
3 changed files with 21 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
custom_ops/gpu_ops/moe/moe_fast_hardamard_kernel.cu
View File

@@ -228,3 +228,22 @@ template void MoeFastHardamardWrapper<phi::dtype::bfloat16, int8_t>(
const int hadamard_block_size,
int8_t *out,
cudaStream_t &stream);
template void
MoeFastHardamardWrapper<phi::dtype::bfloat16, phi::dtype::float8_e4m3fn>(
const phi::dtype::bfloat16 *x_data,
const int64_t *expert_idx_per_token,
const int64_t *recv_expert_count,
const phi::dtype::bfloat16 *shift,
const phi::dtype::bfloat16 *smooth,
const float *quant_scales,
const int quant_round_type,
const float quant_max_bound,
const float quant_min_bound,
const int64_t token_num,
const int64_t dim,
const int num_max_tokens_per_expert,
bool used_in_ep_low_latency,
const int hadamard_block_size,
phi::dtype::float8_e4m3fn *out,
cudaStream_t &stream);