huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-09-11 12:24:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix

Browse Source
This commit is contained in:
Felix Marty 2024-01-25 12:28:56 +00:00 committed by OlivierDehaene
parent d8f33e3c2b
commit 3c93b31959
2 changed files with 7 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
server/exllamav2_kernels/exllamav2_kernels/cuda/q_gemm.cu
View File

@ -43,11 +43,11 @@ void gemm_half_q_half_cuda_part
bool mul_r_weights bool mul_r_weights
) )
{ {
ofstream myfile; ofstream myfile("/tgi/server/exllamav2_kernels/log.txt");
myfile.open ("/tgi/server/exllamav2_kernels/log.txt");
if (!b->is_gptq) if (!b->is_gptq)
{ {
myfile << "go in is_gptq path" << "\n"; myfile << "go in is_gptq path" << "\n";
myfile.flush();
dim3 blockDim, gridDim; dim3 blockDim, gridDim;
blockDim.x = EXL2_BLOCK_KN_SIZE; blockDim.x = EXL2_BLOCK_KN_SIZE;
blockDim.y = 1; blockDim.y = 1;
@ -59,6 +59,7 @@ void gemm_half_q_half_cuda_part
fp_gemm_half_q_half_kernel kernel = pick_gemm_half_q_half_kernel(m_count, r_weights != NULL, mul_r_weights); fp_gemm_half_q_half_kernel kernel = pick_gemm_half_q_half_kernel(m_count, r_weights != NULL, mul_r_weights);
myfile << "launch kernel" << "\n"; myfile << "launch kernel" << "\n";
myfile.flush();
kernel<<<gridDim, blockDim>>> kernel<<<gridDim, blockDim>>>
( (
a, a,
@ -119,6 +120,7 @@ void gemm_half_q_half_cuda_part
r_weights_stride r_weights_stride
); );
} }
myfile.flush();
myfile.close(); myfile.close();
} }

5
server/exllamav2_kernels/exllamav2_kernels/ext.cpp
View File

@ -109,9 +109,9 @@ void gemm_half_q_half
bool force_cuda bool force_cuda
) )
{ {
ofstream myfile; ofstream myfile("/tgi/server/exllamav2_kernels/log.txt");
myfile.open ("/tgi/server/exllamav2_kernels/log.txt");
myfile << "start gemm_half_q_half" << "\n"; myfile << "start gemm_half_q_half" << "\n";
myfile.flush();
QMatrix* qm = reinterpret_cast<QMatrix*> (b); QMatrix* qm = reinterpret_cast<QMatrix*> (b);
@ -124,6 +124,7 @@ void gemm_half_q_half
const at::cuda::OptionalCUDAGuard device_guard(device_of(a)); const at::cuda::OptionalCUDAGuard device_guard(device_of(a));
myfile << "call gemm_half_q_half_cuda" << "\n"; myfile << "call gemm_half_q_half_cuda" << "\n";
myfile.flush();
gemm_half_q_half_cuda gemm_half_q_half_cuda
( (
at::cuda::getCurrentCUDABlasHandle(), at::cuda::getCurrentCUDABlasHandle(),