[NVIDIA] Support nvfp4 quantization (#12784)
This commit is contained in:
@@ -195,6 +195,10 @@ 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);
|
||||
|
||||
|
||||
Reference in New Issue
Block a user