[Bugfix][AMD] Update torch_bindings so that scaled_fp4_quant isn't build on ROCm (#13235)

This commit is contained in:
Sage Moore
2025-02-14 20:30:42 -08:00
committed by GitHub
parent 0c73026844
commit c9f9d5b397
3 changed files with 12 additions and 10 deletions
+4 -4
View File
@@ -177,6 +177,10 @@ void cutlass_scaled_sparse_mm(torch::Tensor& out, torch::Tensor const& a,
std::optional<torch::Tensor> const& bias);
std::vector<torch::Tensor> cutlass_sparse_compress(torch::Tensor const& a);
void scaled_fp4_quant(torch::Tensor& output, torch::Tensor const& input,
torch::Tensor& output_scale,
torch::Tensor const& input_scale);
#endif
void static_scaled_int8_quant(torch::Tensor& out, torch::Tensor const& input,
@@ -194,10 +198,6 @@ torch::Tensor gptq_gemm(torch::Tensor a, torch::Tensor b_q_weight,
void gptq_shuffle(torch::Tensor q_weight, torch::Tensor q_perm, int64_t bit);
void scaled_fp4_quant(torch::Tensor& output, torch::Tensor const& input,
torch::Tensor& output_scale,
torch::Tensor const& input_scale);
void static_scaled_fp8_quant(torch::Tensor& out, torch::Tensor const& input,
torch::Tensor const& scale);