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

missed another thing

Browse source
This commit is contained in:
Concedo 2023-04-25 23:16:04 +08:00
parent 8c9c218609
commit a696b0a16c
1 changed files with 3 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

5
gpttype_adapter.cpp
View file

@ -357,7 +357,8 @@ generation_outputs gpttype_generate(const generation_inputs inputs, generation_o
file_format == FileFormat::GGJT ||
file_format == FileFormat::GPT2_2 ||
file_format == FileFormat::GPTJ_3 ||
file_format == FileFormat::NEOX_1);
file_format == FileFormat::NEOX_1 ||
file_format == FileFormat::NEOX_2);
bool blasmode = (approved_format && embd_inp.size() >= 32 && ggml_cpu_has_blas());
// bool blasmode = false;
int original_batch = params.n_batch;
@ -477,7 +478,7 @@ generation_outputs gpttype_generate(const generation_inputs inputs, generation_o
{
evalres = gpt2_eval(gpt2_ctx_v2, params.n_threads, n_past, embd, logits, mem_per_token, file_format);
}
else if(file_format==FileFormat::NEOX_1)
else if(file_format==FileFormat::NEOX_1 || file_format == FileFormat::NEOX_2)
{
evalres = stablelm_eval(neox_ctx, params.n_threads, n_past, embd, logits, mem_per_token);
}