biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Kernel] Fix conflicting macro names for gguf kernels (#15456)

Browse Source 
Signed-off-by: SzymonOzog <szymon.ozog@gmail.com>
This commit is contained in:
Szymon Ożóg
2025-03-25 14:50:49 +01:00
committed by GitHub
parent 3f04a7fbf2
commit a608160027
2 changed files with 90 additions and 90 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
csrc/quantization/gguf/gguf_kernel.cu
View File

@@ -375,25 +375,25 @@ torch::Tensor ggml_moe_a8(torch::Tensor X, // input
int64_t ggml_moe_get_block_size(int64_t type) {
switch (type) {
case 2:
return MMQ_X_Q4_0;
return MOE_X_Q4_0;
case 3:
return MMQ_X_Q4_1;
return MOE_X_Q4_1;
case 6:
return MMQ_X_Q5_0;
return MOE_X_Q5_0;
case 7:
return MMQ_X_Q5_1;
return MOE_X_Q5_1;
case 8:
return MMQ_X_Q8_0;
return MOE_X_Q8_0;
case 10:
return MMQ_X_Q2_K;
return MOE_X_Q2_K;
case 11:
return MMQ_X_Q3_K;
return MOE_X_Q3_K;
case 12:
return MMQ_X_Q4_K;
return MOE_X_Q4_K;
case 13:
return MMQ_X_Q5_K;
return MOE_X_Q5_K;
case 14:
return MMQ_X_Q6_K;
return MOE_X_Q6_K;
}
return 0;
}