apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-06 09:07:10 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[XPU] Add docs for gm_default_size and fix type (#1962)

Browse Source 
Add description of gm_default_size and change kunlunxin_gm_default_size‘s type to int64_t
This commit is contained in:
linkk08
2023-05-18 20:30:06 +08:00
committed by GitHub
parent 04ef882ee1
commit c8ff8b63e8
6 changed files with 7 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
c_api/fastdeploy_capi/runtime/runtime_option.cc
View File

@@ -108,7 +108,7 @@ void FD_C_RuntimeOptionWrapperUseKunlunXin(
std::string(precision),
bool(adaptive_seqlen),
bool(enable_multi_stream),
gm_default_size);
int64_t(gm_default_size));
}
void FD_C_RuntimeOptionWrapperUseSophgo(

1
c_api/fastdeploy_capi/runtime/runtime_option.h
View File

@@ -126,6 +126,7 @@ FASTDEPLOY_CAPI_EXPORT extern void FD_C_RuntimeOptionWrapperUseAscend(
/// \param[in] adaptive_seqlen Is the input of multi_encoder variable length
/// \param[in] enable_multi_stream Whether to enable the multi stream of
/// KunlunXin XPU.
/// \param[in] gm_default_size The default size of context global memory of KunlunXin XPU.
///
FASTDEPLOY_CAPI_EXPORT extern void FD_C_RuntimeOptionWrapperUseKunlunXin(
__fd_keep FD_C_RuntimeOptionWrapper* fd_c_runtime_option_wrapper,

1
csharp/fastdeploy/runtime_option.cs
View File

@@ -103,6 +103,7 @@ public class RuntimeOption {
/// \param adaptive_seqlen Is the input of multi_encoder variable length
/// \param enable_multi_stream Whether to enable the multi stream of
/// KunlunXin XPU.
/// \param gm_default_size The default size of context global memory of KunlunXin XPU.
///
public void
UseKunlunXin(int kunlunxin_id = 0, int l3_workspace_size = 0xfffc00,

4
fastdeploy/runtime/backends/lite/option.h
View File

@@ -75,12 +75,12 @@ struct LiteBackendOption {
std::string kunlunxin_autotune_file = "";
/// kunlunxin_precision
std::string kunlunxin_precision = "int16";
/// kunlunxin_gm_default_size
int kunlunxin_gm_default_size = 0;
/// kunlunxin_adaptive_seqlen
bool kunlunxin_adaptive_seqlen = false;
/// kunlunxin_enable_multi_stream
bool kunlunxin_enable_multi_stream = false;
/// kunlunxin_gm_default_size
int64_t kunlunxin_gm_default_size = 0;
/// Optimized model dir for CxxConfig
std::string optimized_model_dir = "";

1
fastdeploy/runtime/runtime_option.h
View File

@@ -106,6 +106,7 @@ struct FASTDEPLOY_DECL RuntimeOption {
/// \param adaptive_seqlen Is the input of multi_encoder variable length
/// \param enable_multi_stream Whether to enable the multi stream of
/// KunlunXin XPU.
/// \param gm_default_size The default size of global memory of KunlunXin XPU.
///
void UseKunlunXin(int kunlunxin_id = 0, int l3_workspace_size = 0xfffc00,
bool locked = false, bool autotune = true,

1
python/fastdeploy/runtime.py
View File

@@ -242,6 +242,7 @@ class RuntimeOption:
:param precision: (str)Calculation accuracy of multi_encoder
:param adaptive_seqlen: (bool)adaptive_seqlen Is the input of multi_encoder variable length
:param enable_multi_stream: (bool)Whether to enable the multi stream of KunlunXin XPU.
:param gm_default_size The default size of context global memory of KunlunXin XPU.
"""
return self._option.use_kunlunxin(device_id, l3_workspace_size, locked,
autotune, autotune_file, precision,