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

rerank : use [SEP] token instead of [BOS] (#9737)

Browse source 
* rerank : use [SEP] token instead of [BOS]

ggml-ci

* common : sanity check for non-NULL tokens

ggml-ci

* ci : adjust rank score interval

ggml-ci

* ci : add shebang to run.sh

ggml-ci
This commit is contained in:
Georgi Gerganov 2024-10-05 15:55:04 +03:00 • committed by GitHub
parent 58b16695e1
commit 8c475b97b8
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
5 changed files with 45 additions and 16 deletions
Download patch file Download diff file Expand all files Collapse all files

30
common/common.cpp
View file

@ -838,6 +838,31 @@ struct llama_init_result llama_init_from_gpt_params(gpt_params & params) {
return iparams;
}
if (params.reranking) {
bool ok = true;
if (llama_token_bos(model) == LLAMA_TOKEN_NULL) {
LOG_WRN("%s: warning: model does not have a BOS token, reranking will not work\n", __func__);
ok = false;
}
if (llama_token_eos(model) == LLAMA_TOKEN_NULL) {
LOG_WRN("%s: warning: model does not have an EOS token, reranking will not work\n", __func__);
ok = false;
}
if (llama_token_sep(model) == LLAMA_TOKEN_NULL) {
LOG_WRN("%s: warning: model does not have a SEP token, reranking will not work\n", __func__);
ok = false;
}
if (!ok) {
llama_free_model(model);
return iparams;
}
}
auto cparams = llama_context_params_from_gpt_params(params);
llama_context * lctx = llama_new_context_with_model(model, cparams);
@ -855,6 +880,7 @@ struct llama_init_result llama_init_from_gpt_params(gpt_params & params) {
if (cvec.n_embd == -1) {
llama_free(lctx);
llama_free_model(model);
return iparams;
}
@ -867,6 +893,7 @@ struct llama_init_result llama_init_from_gpt_params(gpt_params & params) {
if (err) {
llama_free(lctx);
llama_free_model(model);
return iparams;
}
}
@ -889,7 +916,7 @@ struct llama_init_result llama_init_from_gpt_params(gpt_params & params) {
llama_lora_adapters_apply(lctx, iparams.lora_adapters);
}
if (params.sparams.ignore_eos && llama_token_eos(model) == -1) {
if (params.sparams.ignore_eos && llama_token_eos(model) == LLAMA_TOKEN_NULL) {
LOG_WRN("%s: warning: model does not have an EOS token, ignoring --ignore-eos\n", __func__);
params.sparams.ignore_eos = false;
}
@ -930,6 +957,7 @@ struct llama_init_result llama_init_from_gpt_params(gpt_params & params) {
iparams.model = model;
iparams.context = lctx;
return iparams;
}