diff options
Diffstat (limited to 'llama.cpp')
| -rw-r--r-- | llama.cpp | 8 |
1 files changed, 6 insertions, 2 deletions
@@ -2495,7 +2495,7 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s if (quantized_type == GGML_TYPE_Q2_K || quantized_type == GGML_TYPE_Q3_K || quantized_type == GGML_TYPE_Q4_K || quantized_type == GGML_TYPE_Q5_K || quantized_type == GGML_TYPE_Q6_K) { int nx = tensor.ne.at(0); - int ny = tensor.ne.at(0); + int ny = tensor.ne.at(1); if (nx % QK_K != 0 || ny % QK_K != 0) { fprintf(stderr, "\n\n========================= Tensor sizes %d x %d are not divisible by %d\n",nx,ny,QK_K); fprintf(stderr, "This is required to be able to use k-quants for now!\n"); @@ -2504,7 +2504,11 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s } } if (tensor.name == "output.weight") { - new_type = GGML_TYPE_Q6_K; + int nx = tensor.ne.at(0); + int ny = tensor.ne.at(1); + if (nx % QK_K == 0 && ny % QK_K == 0) { + new_type = GGML_TYPE_Q6_K; + } } else if (tensor.name.find("attention.wv.weight") != std::string::npos) { if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q2_K) new_type = GGML_TYPE_Q4_K; else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L) new_type = GGML_TYPE_Q5_K; |