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

support qk norm (#3145)

Browse Source
This commit is contained in:
Yuan Xiaolan
2025-08-05 16:46:14 +08:00
committed by GitHub
parent 4a10e29804
commit 7ce00e597c
17 changed files with 791 additions and 201 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
custom_ops/gpu_ops/append_attn/template_instantiation/encoder_write_cache_with_rope_bfloat16_bfloat16_kernel.cu
View File

@@ -43,4 +43,7 @@ EncoderWriteCacheWithRopeKernel<paddle::bfloat16, paddle::bfloat16>(
cudaStream_t& stream,
paddle::Tensor* qkv_out,
paddle::Tensor* key_cache_out,
paddle::Tensor* value_cache_out);
paddle::Tensor* value_cache_out,
const paddle::optional<paddle::Tensor>& q_norm_weight,
const paddle::optional<paddle::Tensor>& k_norm_weight,
const float rms_norm_eps);

5
custom_ops/gpu_ops/append_attn/template_instantiation/encoder_write_cache_with_rope_bfloat16_int_kernel.cu
View File

@@ -42,4 +42,7 @@ template void EncoderWriteCacheWithRopeKernel<paddle::bfloat16, int>(
cudaStream_t& stream,
paddle::Tensor* qkv_out,
paddle::Tensor* key_cache_out,
paddle::Tensor* value_cache_out);
paddle::Tensor* value_cache_out,
const paddle::optional<paddle::Tensor>& q_norm_weight,
const paddle::optional<paddle::Tensor>& k_norm_weight,
const float rms_norm_eps);

5
custom_ops/gpu_ops/append_attn/template_instantiation/encoder_write_cache_with_rope_float16_float16_kernel.cu
View File

@@ -42,4 +42,7 @@ template void EncoderWriteCacheWithRopeKernel<paddle::float16, paddle::float16>(
cudaStream_t& stream,
paddle::Tensor* qkv_out,
paddle::Tensor* key_cache_out,
paddle::Tensor* value_cache_out);
paddle::Tensor* value_cache_out,
const paddle::optional<paddle::Tensor>& q_norm_weight,
const paddle::optional<paddle::Tensor>& k_norm_weight,
const float rms_norm_eps);

5
custom_ops/gpu_ops/append_attn/template_instantiation/encoder_write_cache_with_rope_float16_int_kernel.cu
View File

@@ -42,4 +42,7 @@ template void EncoderWriteCacheWithRopeKernel<paddle::float16, int>(
cudaStream_t& stream,
paddle::Tensor* qkv_out,
paddle::Tensor* key_cache_out,
paddle::Tensor* value_cache_out);
paddle::Tensor* value_cache_out,
const paddle::optional<paddle::Tensor>& q_norm_weight,
const paddle::optional<paddle::Tensor>& k_norm_weight,
const float rms_norm_eps);