vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-10 00:04:34 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix n_embd cannot be divided by quantized block size

Browse source
This commit is contained in:
Li, Zonghang 2025-06-03 14:06:31 +04:00
parent 421b3deca5
commit b30f749e5e
2 changed files with 32 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

3
common/common.cpp
View file

@ -1588,6 +1588,7 @@ struct llama_init_result llama_init_from_gpt_params(gpt_params & params) {
if (n_world == 1) {
uint32_t n_layers = llama_model_n_layers(model);
// assign all layers to this device
params.n_layer_window[0] = n_layers;
cparams.n_layer_window[0] = n_layers;
@ -1596,6 +1597,8 @@ struct llama_init_result llama_init_from_gpt_params(gpt_params & params) {
#if defined(GGML_USE_METAL) || defined(GGML_USE_CUDA)
params.n_gpu_layers = std::min((int32_t)n_layers, params.n_gpu_layers);
cparams.n_gpu_layers = params.n_gpu_layers;
mparams.n_gpu_layers = params.n_gpu_layers;
#endif
} else {