vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

iq4_xs: Added forgotten check for 256 divisibility

Browse source
This commit is contained in:
Iwan Kawrakow 2024-02-27 12:16:17 +02:00
parent 801f998bd9
commit d7bb4b6d4d
1 changed files with 5 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

9
llama.cpp
View file

@ -10996,10 +10996,11 @@ static ggml_type get_k_quant_type(quantize_state_internal & qs, ggml_type new_ty
case GGML_TYPE_IQ3_S:
case GGML_TYPE_IQ1_S:
case GGML_TYPE_Q2_K:
case GGML_TYPE_Q3_K: new_type = GGML_TYPE_IQ4_NL; break;
case GGML_TYPE_Q4_K: new_type = GGML_TYPE_Q5_0; break;
case GGML_TYPE_Q5_K: new_type = GGML_TYPE_Q5_1; break;
case GGML_TYPE_Q6_K: new_type = GGML_TYPE_Q8_0; break;
case GGML_TYPE_Q3_K:
case GGML_TYPE_IQ4_XS: new_type = GGML_TYPE_IQ4_NL; break;
case GGML_TYPE_Q4_K: new_type = GGML_TYPE_Q5_0; break;
case GGML_TYPE_Q5_K: new_type = GGML_TYPE_Q5_1; break;
case GGML_TYPE_Q6_K: new_type = GGML_TYPE_Q8_0; break;
default: throw std::runtime_error("\nUnsupported tensor size encountered\n");
}
LLAMA_LOG_WARN(" - using fallback quantization %s\n", ggml_type_name(new_type));