Shortening formatting
This commit is contained in:
parent
7212098755
commit
8f1b99fee8
1 changed files with 6 additions and 18 deletions
|
@ -15334,14 +15334,10 @@ static ggml_type llama_tensor_get_type(quantize_state_internal & qs, ggml_type n
|
|||
ftype == LLAMA_FTYPE_MOSTLY_IQ1_S || ftype == LLAMA_FTYPE_MOSTLY_IQ1_M) {
|
||||
new_type = GGML_TYPE_Q2_K;
|
||||
}
|
||||
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ2_S || ftype == LLAMA_FTYPE_MOSTLY_IQ2_M) {
|
||||
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ2_S || ftype == LLAMA_FTYPE_MOSTLY_IQ2_M || ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS) {
|
||||
new_type = GGML_TYPE_IQ3_S;
|
||||
}
|
||||
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS) {
|
||||
new_type = GGML_TYPE_IQ3_S;
|
||||
}
|
||||
else if (new_type == GGML_TYPE_Q4_0_4_4 || new_type == GGML_TYPE_Q4_0_4_8 ||
|
||||
new_type == GGML_TYPE_Q4_0_8_8) {
|
||||
else if (new_type == GGML_TYPE_Q4_0_4_4 || new_type == GGML_TYPE_Q4_0_4_8 || new_type == GGML_TYPE_Q4_0_8_8) {
|
||||
new_type = GGML_TYPE_Q4_0;
|
||||
}
|
||||
}
|
||||
|
@ -15381,12 +15377,8 @@ static ggml_type llama_tensor_get_type(quantize_state_internal & qs, ggml_type n
|
|||
else if ((ftype == LLAMA_FTYPE_MOSTLY_IQ3_XS) && (qs.model.hparams.n_gqa() >= 2 || qs.model.hparams.n_expert >= 2)) {
|
||||
new_type = GGML_TYPE_Q4_K;
|
||||
}
|
||||
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_S) {
|
||||
new_type = GGML_TYPE_Q4_K;
|
||||
}
|
||||
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_M) {
|
||||
new_type = GGML_TYPE_Q5_K;
|
||||
}
|
||||
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_S) new_type = GGML_TYPE_Q4_K;
|
||||
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_M) new_type = GGML_TYPE_Q5_K;
|
||||
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M) {
|
||||
new_type = qs.i_attention_wv < 2 ? GGML_TYPE_Q5_K : GGML_TYPE_Q4_K;
|
||||
}
|
||||
|
@ -15418,16 +15410,12 @@ static ggml_type llama_tensor_get_type(quantize_state_internal & qs, ggml_type n
|
|||
new_type = GGML_TYPE_IQ2_S;
|
||||
}
|
||||
} else if (name.find("attn_q.weight") != std::string::npos) {
|
||||
if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XS) {
|
||||
new_type = GGML_TYPE_IQ3_XXS;
|
||||
}
|
||||
if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XS) new_type = GGML_TYPE_IQ3_XXS;
|
||||
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS) new_type = GGML_TYPE_IQ2_S;
|
||||
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ2_XXS || ftype == LLAMA_FTYPE_MOSTLY_IQ2_XS || ftype == LLAMA_FTYPE_MOSTLY_IQ1_S ||
|
||||
ftype == LLAMA_FTYPE_MOSTLY_IQ2_S || ftype == LLAMA_FTYPE_MOSTLY_IQ2_M || ftype == LLAMA_FTYPE_MOSTLY_IQ1_M) {
|
||||
if (qs.model.hparams.n_expert >= 8) new_type = GGML_TYPE_Q3_K;
|
||||
}
|
||||
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS) {
|
||||
new_type = GGML_TYPE_IQ2_S;
|
||||
}
|
||||
} else if (name.find("ffn_down") != std::string::npos) {
|
||||
auto info = layer_info(qs.i_ffn_down, qs.n_ffn_down, name.c_str());
|
||||
int i_layer = info.first, n_layer = info.second;
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue