mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-06-10 16:28:38 +00:00
0df3523ef1
[CK_TILE] Add save_matrix_txt() and extract HostTensor I/O to free functions (#6807) ## Summary - Extract `loadtxt`, `savetxt`, and `save_matrix_txt` from `HostTensor` member functions into standalone free functions in `host_tensor_io.hpp` (Single Responsibility Principle) - Add `save_matrix_txt()` for writing 2D tensors to space-separated text files with configurable output limit (default 256x256, pass 0 to dump all) - Supports float, int, and int8_t output formats via a `dtype` parameter - Validate dtype early and throw on unsupported values in all three functions - Update callers in `15_fused_moe/main.cpp` to use free function syntax
57 lines
2.9 KiB
C++
57 lines
2.9 KiB
C++
// Copyright (c) Advanced Micro Devices, Inc., or its affiliates.
|
|
// SPDX-License-Identifier: MIT
|
|
#pragma once
|
|
|
|
#include "ck_tile/host/arg_parser.hpp"
|
|
#include "ck_tile/host/check_err.hpp"
|
|
#include "ck_tile/host/concat.hpp"
|
|
#include "ck_tile/host/convolution_host_tensor_descriptor_helper.hpp"
|
|
#include "ck_tile/host/convolution_parameter.hpp"
|
|
#include "ck_tile/host/device_memory.hpp"
|
|
#include "ck_tile/host/device_prop.hpp"
|
|
#include "ck_tile/host/fill.hpp"
|
|
#include "ck_tile/host/flush_icache.hpp"
|
|
#include "ck_tile/host/high_res_cpu_clock.hpp"
|
|
#include "ck_tile/host/hip_check_error.hpp"
|
|
#include "ck_tile/host/host_tensor.hpp"
|
|
#include "ck_tile/host/host_tensor_io.hpp"
|
|
#include "ck_tile/host/joinable_thread.hpp"
|
|
#include "ck_tile/host/kernel_launch.hpp"
|
|
#include "ck_tile/host/permute_pk_int4.hpp"
|
|
#include "ck_tile/host/pinned_host_releaser.hpp"
|
|
#include "ck_tile/host/ranges.hpp"
|
|
#include "ck_tile/host/reference/reference_batched_contraction.hpp"
|
|
#include "ck_tile/host/reference/reference_batched_dropout.hpp"
|
|
#include "ck_tile/host/reference/reference_batched_dropout_randval.hpp"
|
|
#include "ck_tile/host/reference/reference_batched_elementwise.hpp"
|
|
#include "ck_tile/host/reference/reference_batched_gemm.hpp"
|
|
#include "ck_tile/host/reference/reference_batched_masking.hpp"
|
|
#include "ck_tile/host/reference/reference_batched_mx_descale.hpp"
|
|
#include "ck_tile/host/reference/reference_batched_rotary_position_embedding.hpp"
|
|
#include "ck_tile/host/reference/reference_batched_softmax.hpp"
|
|
#include "ck_tile/host/reference/reference_batched_transpose.hpp"
|
|
#include "ck_tile/host/reference/reference_blocked_attention.hpp"
|
|
#include "ck_tile/host/reference/reference_elementwise.hpp"
|
|
#include "ck_tile/host/reference/reference_fused_moe.hpp"
|
|
#include "ck_tile/host/reference/reference_gemm.hpp"
|
|
#include "ck_tile/host/reference/reference_grouped_conv_bwd_data.hpp"
|
|
#include "ck_tile/host/reference/reference_grouped_conv_bwd_weight.hpp"
|
|
#include "ck_tile/host/reference/reference_grouped_conv_fwd.hpp"
|
|
#include "ck_tile/host/reference/reference_im2col.hpp"
|
|
#include "ck_tile/host/reference/reference_layernorm2d_fwd.hpp"
|
|
#include "ck_tile/host/reference/reference_moe_gemm.hpp"
|
|
#include "ck_tile/host/reference/reference_moe_sorting.hpp"
|
|
#include "ck_tile/host/reference/reference_permute.hpp"
|
|
#include "ck_tile/host/reference/reference_pool.hpp"
|
|
#include "ck_tile/host/reference/reference_reduce.hpp"
|
|
#include "ck_tile/host/reference/reference_rmsnorm2d_fwd.hpp"
|
|
#include "ck_tile/host/reference/reference_rowwise_quantization2d.hpp"
|
|
#include "ck_tile/host/reference/reference_softmax.hpp"
|
|
#include "ck_tile/host/reference/reference_topk.hpp"
|
|
#include "ck_tile/host/reference/reference_transpose.hpp"
|
|
#include "ck_tile/host/rotating_buffers.hpp"
|
|
#include "ck_tile/host/stream_config.hpp"
|
|
#include "ck_tile/host/stream_utils.hpp"
|
|
#include "ck_tile/host/tensor_shuffle_utils.hpp"
|
|
#include "ck_tile/host/timer.hpp"
|