zhangyiss/mlx
1
0
Fork 0
mirror of https://github.com/ml-explore/mlx.git synced 2025-12-16 01:49:05 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

WIP (io)

Browse Source
This commit is contained in:
Ronan Collobert
2025-10-31 13:20:47 -07:00
parent b48d298205
commit 4a1b1796b7
2 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
mlx/io/gguf.cpp
View File

@@ -238,7 +238,7 @@ std::unordered_map<std::string, array> load_arrays(gguf_ctx* ctx) {
return array_map;
}
GGUFLoad load_gguf(const std::string& file, StreamOrDevice s) {
GGUFLoad load_gguf(const std::string& file, StreamOrDevice /* s */) {
bool exists;
{
std::ifstream f(file.c_str());
@@ -440,7 +440,7 @@ void save_gguf(
}
const char* tensorname = key.c_str();
const uint64_t namelen = key.length();
const uint32_t num_dim = arr.ndim();
const int num_dim = arr.ndim();
std::vector<uint64_t> dim(num_dim);
for (int i = 0; i < num_dim; i++) {
dim[i] = arr.shape()[num_dim - 1 - i];

4
mlx/io/gguf_quants.cpp
View File

@@ -77,8 +77,8 @@ void extract_q8_0_data(
array& weights_arr,
array& scales_arr,
array& biases_arr) {
const uint64_t weights_per_block = 32;
const uint64_t bytes_per_block = 34; // 2 bytes scale, 32x1 byte weights
const int64_t weights_per_block = 32;
const int64_t bytes_per_block = 34; // 2 bytes scale, 32x1 byte weights
auto data = static_cast<uint8_t*>(tensor.weights_data);
auto weights = weights_arr.data<int8_t>();
auto scales = scales_arr.data<float16_t>();