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

[Model][Quantization] HQQ support through Marlin kernel expansion (#9766)

Browse Source 
Signed-off-by: ElizaWszola <eliza@neuralmagic.com>
This commit is contained in:
ElizaWszola
2024-11-19 22:31:12 +01:00
committed by GitHub
parent efa9084628
commit b00b33d77e
11 changed files with 632 additions and 89 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
csrc/torch_bindings.cpp
View File

@@ -244,7 +244,7 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
"Tensor b_zeros, Tensor g_idx, Tensor perm, Tensor workspace, "
"int b_q_type, "
"SymInt size_m, SymInt size_n, SymInt size_k, bool is_k_full, "
"bool has_zp, bool use_fp32_reduce) -> Tensor");
"bool has_zp, bool use_fp32_reduce, bool is_zp_float) -> Tensor");
// conditionally compiled so impl registration is in source file
// gptq_marlin repack from GPTQ.