Skip to content

[fix]: fix moe.hpp load from file bug. #1780

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ouqingliang merged 2 commits into from
Jan 4, 2026
Merged

[fix]: fix moe.hpp load from file bug. #1780

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f2baa0c
[fix]: fix moe.hpp load from file bug.
ouqingliang Dec 29, 2025
f8bd1fa
[fix]: fix all moe hpp init bug.
ouqingliang Jan 4, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 6 additions & 5 deletions kt-kernel/operators/amx/moe.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,8 +32,6 @@ class AMX_MOE_TP : public AMX_MOE_BASE<T, AMX_MOE_TP<T>> {
using Base::down_bc_;
using Base::m_local_num_;

std::filesystem::path prefix;

void* gate_proj_; // [expert_num * intermediate_size * hidden_size ( /32 if
// quantized)]
void* up_proj_; // [expert_num * intermediate_size * hidden_size ( /32 if
Expand Down Expand Up @@ -144,7 +142,7 @@ class AMX_MOE_TP : public AMX_MOE_BASE<T, AMX_MOE_TP<T>> {
auto& load = config_.load;
auto& save = config_.save;

prefix = config_.path;
std::filesystem::path prefix = config_.path;
prefix = prefix / ("_layer_" + std::to_string(config_.layer_idx)) / ("_numa_" + std::to_string(tp_part_idx));
if (save) {
std::cout << "Creating " << prefix << std::endl;
Expand Down Expand Up @@ -260,6 +258,9 @@ class AMX_MOE_TP : public AMX_MOE_BASE<T, AMX_MOE_TP<T>> {
} else {
int nth = T::recommended_nth(config_.intermediate_size);
static uint8_t mat_type_all = 3, mat_split = 1;
std::filesystem::path prefix = config_.path;
prefix = prefix / ("_layer_" + std::to_string(config_.layer_idx)) / ("_numa_" + std::to_string(tp_part_idx));
Comment on lines +259 to +260
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot Jan 4, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

This logic to construct the prefix path is duplicated from lines 145-146 within the same function. To improve maintainability and avoid redundancy, consider defining and initializing the prefix variable once at the beginning of the load_weights function, before the if (config_.quant_config.bits == 0) block.


if (config_.load) {
std::cout << "Loading from " << prefix << std::endl;
for (int task_id = 0; task_id < config_.expert_num * mat_type_all * mat_split; task_id++) {
Expand Down Expand Up @@ -335,7 +336,7 @@ class AMX_MOE_TP : public AMX_MOE_BASE<T, AMX_MOE_TP<T>> {
if (config_.save) {
pool->do_work_stealing_job(
config_.expert_num * mat_type_all, nullptr,
[this, physical_to_logical_map](int task_id) {
[this, physical_to_logical_map, prefix](int task_id) {
int64_t expert_idx = task_id / mat_type_all;
expert_idx = expert_map(physical_to_logical_map, expert_idx);
uint8_t mat_class = task_id % mat_type_all;
Expand Down Expand Up @@ -426,7 +427,7 @@ class TP_MOE<AMX_MOE_TP<K>> : public TP_MOE<AMX_MOE_BASE<K, AMX_MOE_TP<K>>> {

this->weights_loaded = true;
} else if (config.path != "") {
printf("TP Load from file\n");
printf("TP Load from file %s\n", config.path.c_str());
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot Jan 4, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

For consistency with the logging style in this file (e.g., lines 148, 265), it's better to use std::cout instead of printf. This also avoids mixing C-style I/O with C++ streams and leverages C++ features more effectively.

      std::cout << "TP Load from file " << config.path << std::endl;

DO_TPS_LOAD_WEIGHTS(pool);
this->weights_loaded = true;
} else {
Expand Down