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

trying to fix ggjt v3

Browse source
This commit is contained in:
Concedo 2023-05-21 00:29:50 +08:00
parent c048bcfec4
commit 5032e0fd64
2 changed files with 10 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

6
gpttype_adapter.cpp
View file

@ -318,7 +318,11 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
//determine mem per token
const std::vector<int> tmp = {1, 2, 3, 4};
llama_eval(llama_ctx_v3, tmp.data(), tmp.size(), 0, params.n_threads);
auto er = llama_eval(llama_ctx_v3, tmp.data(), tmp.size(), 0, params.n_threads);
if(er!=0)
{
printf("\nLLAMA EVAL returned nonzero!\n");
}
return ModelLoadResult::SUCCESS;
}
else if (file_format == FileFormat::RWKV_1)

10
llama.cpp
View file

@ -1137,11 +1137,11 @@ static bool llama_eval_internal(
const int n_past,
const int n_threads) {
// enforce that the first token is BOS
if (n_past == 0 && tokens[0] != llama_token_bos()) {
fprintf(stderr, "%s: first token must be BOS\n", __func__);
return false;
}
// // enforce that the first token is BOS
// if (n_past == 0 && tokens[0] != llama_token_bos()) {
// fprintf(stderr, "%s: first token must be BOS\n", __func__);
// return false;
// }
const int64_t t_start_us = ggml_time_us();