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

use const auto &kv instead of auto &kv, statements into braces

Browse source
This commit is contained in:
mendax0110 2023-06-26 21:59:08 +02:00
parent d38e451578
commit 565afba411
1 changed files with 16 additions and 13 deletions
Download patch file Download diff file Expand all files Collapse all files

29
llama.cpp
View file

@ -2533,23 +2533,26 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
new_type = GGML_TYPE_Q6_K; new_type = GGML_TYPE_Q6_K;
} }
} else if (tensor.name.find("attention.wv.weight") != std::string::npos) { } 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; 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; } else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L) { new_type = GGML_TYPE_Q5_K;
else if ((ftype == LLAMA_FTYPE_MOSTLY_Q4_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q5_K_M) && } else if ((ftype == LLAMA_FTYPE_MOSTLY_Q4_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q5_K_M) &&
use_more_bits(i_attention_wv, n_attention_wv)) new_type = GGML_TYPE_Q6_K; use_more_bits(i_attention_wv, n_attention_wv)) { new_type = GGML_TYPE_Q6_K;
else if (QK_K == 64 && (ftype == LLAMA_FTYPE_MOSTLY_Q4_K_S || ftype == LLAMA_FTYPE_MOSTLY_Q3_K_S) && } else if (QK_K == 64 && (ftype == LLAMA_FTYPE_MOSTLY_Q4_K_S || ftype == LLAMA_FTYPE_MOSTLY_Q3_K_S) &&
(i_attention_wv < n_attention_wv/8 || i_attention_wv >= 7*n_attention_wv/8)) new_type = GGML_TYPE_Q6_K; (i_attention_wv < n_attention_wv/8 || i_attention_wv >= 7*n_attention_wv/8)) { new_type = GGML_TYPE_Q6_K;
}
++i_attention_wv; ++i_attention_wv;
} else if (tensor.name.find("feed_forward.w2.weight") != std::string::npos) { } else if (tensor.name.find("feed_forward.w2.weight") != std::string::npos) {
if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q2_K) new_type = GGML_TYPE_Q4_K; 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; } } else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L) { new_type = GGML_TYPE_Q5_K;
else if ((ftype == LLAMA_FTYPE_MOSTLY_Q4_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q5_K_M) && } else if ((ftype == LLAMA_FTYPE_MOSTLY_Q4_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q5_K_M) &&
use_more_bits(i_feed_forward_w2, n_feed_forward_w2)) new_type = GGML_TYPE_Q6_K; use_more_bits(i_feed_forward_w2, n_feed_forward_w2)) { new_type = GGML_TYPE_Q6_K;
}
//else if (ftype == LLAMA_FTYPE_MOSTLY_Q4_K_S && i_feed_forward_w2 < n_feed_forward_w2/8) new_type = GGML_TYPE_Q6_K; //else if (ftype == LLAMA_FTYPE_MOSTLY_Q4_K_S && i_feed_forward_w2 < n_feed_forward_w2/8) new_type = GGML_TYPE_Q6_K;
++i_feed_forward_w2; ++i_feed_forward_w2;
} else if (tensor.name.find("attention.wo.weight") != std::string::npos) { } else if (tensor.name.find("attention.wo.weight") != std::string::npos) {
if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q2_K) new_type = GGML_TYPE_Q4_K; 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; } else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L) { new_type = GGML_TYPE_Q5_K;
}
} }
#endif #endif
@ -2874,7 +2877,7 @@ int llama_apply_lora_from_file_internal(const struct llama_model & model, const
// create a name -> tensor map of the model to accelerate lookups // create a name -> tensor map of the model to accelerate lookups
std::unordered_map<std::string, struct ggml_tensor*> model_tensors; std::unordered_map<std::string, struct ggml_tensor*> model_tensors;
for (auto & kv: model.tensors_by_name) { for (const auto & kv: model.tensors_by_name) {
model_tensors.insert(kv); model_tensors.insert(kv);
} }