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

[Kernel/Quant] Remove AQLM (#22943)

Browse Source 
Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
This commit is contained in:
Michael Goin
2025-08-16 15:38:21 -04:00
committed by GitHub
parent 3253ae765e
commit 4fc722eca4
16 changed files with 0 additions and 1534 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
csrc/torch_bindings.cpp
View File

@@ -207,21 +207,6 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
// Quantization ops
#ifndef USE_ROCM
// Quantized GEMM for AQLM.
ops.def(
"aqlm_gemm(Tensor input, Tensor codes, Tensor codebooks, "
"Tensor scales, int[] codebook_partition_sizes, Tensor? bias) "
"-> Tensor",
{stride_tag});
ops.impl("aqlm_gemm", torch::kCUDA, &aqlm_gemm);
// Decompression method for AQLM.
ops.def(
"aqlm_dequant(Tensor codes, Tensor codebooks, "
"int[] codebook_partition_sizes) -> Tensor",
{stride_tag});
ops.impl("aqlm_dequant", torch::kCUDA, &aqlm_dequant);
// Quantized GEMM for AWQ.
ops.def(
"awq_gemm(Tensor _in_feats, Tensor _kernel, Tensor _scaling_factors, "