For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Mojo function

q6_k_dequantize_impl

def q6_k_dequantize_impl[output_rank: Int](input_tt: TileTensor[DType.uint8, Storage=input_tt.Storage, linear_idx_type=input_tt.linear_idx_type], output_tt: TileTensor[DType.float32, Storage=output_tt.Storage, linear_idx_type=output_tt.linear_idx_type], output_shape: IndexList[output_rank])

Dequantizes a Q6_K encoded tensor into a float32 output tensor.

Parameters:

output_rank (Int): The rank of the output tensor.

Args:

input_tt (TileTensor[DType.uint8, Storage=input_tt.Storage, linear_idx_type=input_tt.linear_idx_type]): The input tensor containing Q6_K encoded data.
output_tt (TileTensor[DType.float32, Storage=output_tt.Storage, linear_idx_type=output_tt.linear_idx_type]): The output tensor to write dequantized float32 values to.
output_shape (IndexList[output_rank]): The shape of the output tensor.