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

Signed-off-by: ElizaWszola <eliza@neuralmagic.com>
2024-11-19 22:31:12 +01:00
parent efa9084628
commit b00b33d77e
11 changed files with 632 additions and 89 deletions
--- a/benchmarks/kernels/benchmark_machete.py
+++ b/benchmarks/kernels/benchmark_machete.py
@@ -210,7 +210,8 @@ def marlin_create_bench_fn(bt: BenchmarkTensors) -> Callable:
                                          size_m=bt.a.shape[0],
                                          size_n=bt.w_ref.shape[1],
                                          size_k=bt.w_ref.shape[0],
-                                          is_k_full=True)
+                                          is_k_full=True,
+                                          is_zp_float=False)
    else:
        assert bt.a.dtype == torch.int8
        assert bt.wtype == scalar_types.uint4b8