aparler/c-llm
1
0
Fork 0
mirror of https://github.com/karpathy/llm.c.git synced 2026-07-28 20:35:09 -04:00
Code Issues Projects Releases Packages Wiki Activity Actions

remove duplicate workspace allocation

Browse source
This commit is contained in:
Erik Schultheis 2024-05-17 01:15:47 +03:00
parent 88c3bea890
commit b24279c4db
1 changed files with 0 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

2
train_gpt2.cu
View file

@ -2409,8 +2409,6 @@ void common_start(bool override_enable_tf32 = true, bool print_device_info = tru
bool enable_tf32 = PRECISION_MODE == PRECISION_FP32 && deviceProp.major >= 8 && override_enable_tf32;
cublasCheck(cublasSetMathMode(cublas_handle, enable_tf32 ? CUBLAS_TF32_TENSOR_OP_MATH : CUBLAS_DEFAULT_MATH));
cublas_compute = enable_tf32 ? CUBLAS_COMPUTE_32F_FAST_TF32 : CUBLAS_COMPUTE_32F;
// setup the (global) cuBLASLt workspace
cudaCheck(cudaMalloc(&cublaslt_workspace, cublaslt_workspace_size));
create_cudnn();
}