| 934 | // |
| 935 | |
| 936 | llama_model_quantize_params llama_model_quantize_default_params() { |
| 937 | llama_model_quantize_params result = { |
| 938 | /*.nthread =*/ 0, |
| 939 | /*.ftype =*/ LLAMA_FTYPE_MOSTLY_Q5_1, |
| 940 | /*.output_tensor_type =*/ GGML_TYPE_COUNT, |
| 941 | /*.token_embedding_type =*/ GGML_TYPE_COUNT, |
| 942 | /*.allow_requantize =*/ false, |
| 943 | /*.quantize_output_tensor =*/ true, |
| 944 | /*.only_copy =*/ false, |
| 945 | /*.pure =*/ false, |
| 946 | /*.keep_split =*/ false, |
| 947 | /*.imatrix =*/ nullptr, |
| 948 | /*.kv_overrides =*/ nullptr, |
| 949 | /*.tensor_type =*/ nullptr, |
| 950 | }; |
| 951 | |
| 952 | return result; |
| 953 | } |
| 954 | |
| 955 | uint32_t llama_model_quantize( |
| 956 | const char * fname_inp, |