Skip to content

GitLab

Commit 14ad512a authored by gaoqiong's avatar gaoqiong
Browse files

增加awq 多卡支持

parent 6ba90df9
Hide whitespace changes
Inline Side-by-side
Showing with 4 additions and 4 deletions
src/turbomind/utils/cublasMMWrapper.cc
View file @ 14ad512a
...@@ -66,7 +66,7 @@ cublasMMWrapper::cublasMMWrapper(cublasHandle_t cublas_handle, ...@@ -66,7 +66,7 @@ cublasMMWrapper::cublasMMWrapper(cublasHandle_t cublas_handle,
else if(m_weightlayout_switch ==2) else if(m_weightlayout_switch ==2)
{ {
printf("alloc space for ck workspace\n"); printf("alloc space for ck workspace\n");
printf("weight layout is tn pading ck\n"); printf("weight layout is tn ck\n");
ck_workspace_ = allocator_->reMalloc(ck_workspace_, CK_WORKSPACE_SIZE, false); ck_workspace_ = allocator_->reMalloc(ck_workspace_, CK_WORKSPACE_SIZE, false);
} }
} }
...@@ -125,7 +125,7 @@ cublasMMWrapper::cublasMMWrapper(cublasHandle_t cublas_handle, ...@@ -125,7 +125,7 @@ cublasMMWrapper::cublasMMWrapper(cublasHandle_t cublas_handle,
else if(m_weightlayout_switch ==2) else if(m_weightlayout_switch ==2)
{ {
printf("alloc space for ck workspace\n"); printf("alloc space for ck workspace\n");
printf("weight layout is tn pading ck\n"); printf("weight layout is tn ck\n");
ck_workspace_ = allocator_->reMalloc(ck_workspace_, CK_WORKSPACE_SIZE, false); ck_workspace_ = allocator_->reMalloc(ck_workspace_, CK_WORKSPACE_SIZE, false);
} }
} }
...@@ -182,7 +182,7 @@ cublasMMWrapper::cublasMMWrapper(const cublasMMWrapper& wrapper): ...@@ -182,7 +182,7 @@ cublasMMWrapper::cublasMMWrapper(const cublasMMWrapper& wrapper):
if(m_weightlayout_switch ==1||m_weightlayout_switch==0) if(m_weightlayout_switch ==1||m_weightlayout_switch==0)
{ {
//需要反量化后weight临时存储的空间 //需要反量化后weight临时存储的空间
printf("alloc space for deqeight\n"); //("alloc space for deqeight\n");
deweight_workspace_=allocator_->reMalloc(deweight_workspace_, DEQ_WORKSPACE_SIZE, false); deweight_workspace_=allocator_->reMalloc(deweight_workspace_, DEQ_WORKSPACE_SIZE, false);
if(m_weightlayout_switch ==1) if(m_weightlayout_switch ==1)
{ {
...@@ -194,7 +194,7 @@ cublasMMWrapper::cublasMMWrapper(const cublasMMWrapper& wrapper): ...@@ -194,7 +194,7 @@ cublasMMWrapper::cublasMMWrapper(const cublasMMWrapper& wrapper):
else if(m_weightlayout_switch ==2) else if(m_weightlayout_switch ==2)
{ {
printf("alloc space for ck workspace\n"); printf("alloc space for ck workspace\n");
printf("weight layout is tn pading ck\n"); printf("weight layout is tn ck\n");
ck_workspace_ = allocator_->reMalloc(ck_workspace_, CK_WORKSPACE_SIZE, false); ck_workspace_ = allocator_->reMalloc(ck_workspace_, CK_WORKSPACE_SIZE, false);
} }
} }
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment