diff options
| -rw-r--r-- | src/llama.cpp | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/src/llama.cpp b/src/llama.cpp index 72662fc9..0eea948a 100644 --- a/src/llama.cpp +++ b/src/llama.cpp @@ -15690,7 +15690,7 @@ static ggml_type llama_tensor_get_type(quantize_state_internal & qs, ggml_type n new_type = GGML_TYPE_Q4_K; } else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS) { - new_type = qs.model.hparams.n_gqa() >= 4 ? GGML_TYPE_Q4_K : new_type = qs.model.hparams.n_gqa() >= 2 ? GGML_TYPE_IQ3_K + new_type = qs.model.hparams.n_gqa() >= 4 ? GGML_TYPE_Q4_K : qs.model.hparams.n_gqa() >= 2 ? GGML_TYPE_IQ3_K : !qs.has_imatrix ? GGML_TYPE_IQ3_S : GGML_TYPE_IQ3_XXS; } else if ((ftype == LLAMA_FTYPE_MOSTLY_IQ3_XS || ftype == LLAMA_FTYPE_MOSTLY_IQ3_S || ftype == LLAMA_FTYPE_MOSTLY_IQ3_K) && qs.model.hparams.n_gqa() >= 2) { |