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

llama : remove token functions with context args in favor of model (#3720)

Browse source 
* added `llama_model_token_*` variants to all the `llama_token_*` functions.

* added `LLAMA_API`

* formatting

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

* removed old `llama_token` functions

* changed 3 more functions to take in model

- `llama_token_get_text`
- `llama_token_get_score`
- `llama_token_get_type`

* added back docs

* fixed main.cpp

* changed token functions to use new model variants

* changed token functions to use new model variants

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Marcus Dunn 2023-10-23 12:40:03 -07:00 • committed by GitHub
parent 6336701c93
commit 5be6c803fa
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
16 changed files with 81 additions and 79 deletions
Download patch file Download diff file Expand all files Collapse all files

8
common/common.cpp
View file

@ -880,13 +880,13 @@ std::tuple<struct llama_model *, struct llama_context *> llama_init_from_gpt_par
}
if (params.ignore_eos) {
params.sparams.logit_bias[llama_token_eos(lctx)] = -INFINITY;
params.sparams.logit_bias[llama_token_eos(model)] = -INFINITY;
}
{
LOG("warming up the model with an empty run\n");
std::vector<llama_token> tmp = { llama_token_bos(lctx), llama_token_eos(lctx), };
std::vector<llama_token> tmp = { llama_token_bos(model), llama_token_eos(model), };
llama_decode(lctx, llama_batch_get_one(tmp.data(), std::min(tmp.size(), (size_t) params.n_batch), 0, 0));
llama_kv_cache_tokens_rm(lctx, -1, -1);
llama_reset_timings(lctx);
@ -941,7 +941,7 @@ std::string llama_token_to_piece(const struct llama_context * ctx, llama_token t
}
std::string llama_detokenize_spm(llama_context * ctx, const std::vector<llama_token> & tokens) {
const llama_token bos_id = llama_token_bos(ctx);
const llama_token bos_id = llama_token_bos(llama_get_model(ctx));
std::string piece;
std::string result;
@ -1186,7 +1186,7 @@ void dump_non_result_info_yaml(FILE * stream, const gpt_params & params, const l
fprintf(stream, "hellaswag: %s # default: false\n", params.hellaswag ? "true" : "false");
fprintf(stream, "hellaswag_tasks: %zu # default: 400\n", params.hellaswag_tasks);
const auto logit_bias_eos = sparams.logit_bias.find(llama_token_eos(lctx));
const auto logit_bias_eos = sparams.logit_bias.find(llama_token_eos(llama_get_model(lctx)));
const bool ignore_eos = logit_bias_eos != sparams.logit_bias.end() && logit_bias_eos->second == -INFINITY;
fprintf(stream, "ignore_eos: %s # default: false\n", ignore_eos ? "true" : "false");