vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-11 17:44:38 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 3

fixed clip quantize

Browse source
This commit is contained in:
Concedo 2025-04-30 20:45:40 +08:00
parent fda682fa12
commit dbb6bbf8ea
2 changed files with 17 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

20
examples/llava/clip.cpp
View file

@ -3489,12 +3489,24 @@ bool clip_model_quantize(const char * fname_inp, const char * fname_out, const i
assert(itype < GGML_TYPE_COUNT); assert(itype < GGML_TYPE_COUNT);
ggml_type type = static_cast<ggml_type>(itype); ggml_type type = static_cast<ggml_type>(itype);
auto * ctx_clip = clip_init(fname_inp, clip_context_params{ auto ccparams = clip_context_params{
/* use_gpu */ false, /* use_gpu */ false,
/* verbosity */ GGML_LOG_LEVEL_ERROR, /* verbosity */ GGML_LOG_LEVEL_DEBUG,
}); };
g_logger_state.verbosity_thold = ccparams.verbosity;
clip_ctx * ctx_clip = new clip_ctx(ccparams);
clip_model_loader loader(fname_inp, *ctx_clip);
try {
loader.load_hparams();
loader.load_tensors();
loader.alloc_compute_meta();
} catch (const std::exception & e) {
printf("%s: failed to load model '%s': %s\n", __func__, fname_inp, e.what());
delete ctx_clip;
return false;
}
const auto & ctx_src = ctx_clip->ctx_gguf.get(); const auto & ctx_src = loader.ctx_gguf.get();
const auto & ctx_data = ctx_clip->ctx_data.get(); const auto & ctx_data = ctx_clip->ctx_data.get();
auto * ctx_out = gguf_init_empty(); auto * ctx_out = gguf_init_empty();

2
examples/llava/quantclip.cpp
View file

@ -33,7 +33,7 @@ int main(int argc, char ** argv) {
printf("quantizing mmproj clip model to type=%d... ",type); printf("quantizing mmproj clip model to type=%d... ",type);
clip_model_quantize(fname_inp.c_str(), fname_out.c_str(), type); clip_model_quantize(fname_inp.c_str(), fname_out.c_str(), type);
printf("done\n"); printf("\ndone\n");
return 0; return 0;
} }