replica
/
llama.cpp
mirror of https://github.com/ggml-org/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

server : fix handling of the ignore_eos flag (#14710) 

ggml-ci
This commit is contained in:
Georgi Gerganov 2025-07-16 12:13:57 +03:00 committed by GitHub
parent 5cae766541
commit 538cc77f7f
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 10 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
tools/server/server.cpp
View File

@ -127,7 +127,6 @@ struct slot_params {
std::vector<std::string> response_fields; std::vector<std::string> response_fields;
bool timings_per_token = false; bool timings_per_token = false;
bool post_sampling_probs = false; bool post_sampling_probs = false;
bool ignore_eos = false;
struct common_params_sampling sampling; struct common_params_sampling sampling;
struct common_params_speculative speculative; struct common_params_speculative speculative;
@ -441,7 +440,6 @@ struct server_task {
{ {
params.sampling.logit_bias.clear(); params.sampling.logit_bias.clear();
params.ignore_eos = json_value(data, "ignore_eos", false);
const auto & logit_bias = data.find("logit_bias"); const auto & logit_bias = data.find("logit_bias");
if (logit_bias != data.end() && logit_bias->is_array()) { if (logit_bias != data.end() && logit_bias->is_array()) {
@ -472,6 +470,16 @@ struct server_task {
} }
} }
} }
params.sampling.ignore_eos = json_value(data, "ignore_eos", params_base.sampling.ignore_eos);
if (params.sampling.ignore_eos) {
for (llama_token i = 0; i < llama_vocab_n_tokens(vocab); i++) {
if (llama_vocab_is_eog(vocab, i)) {
//SRV_DBG("%s: added %s logit bias = %f\n", __func__, common_token_to_piece(ctx, i).c_str(), -INFINITY);
params.sampling.logit_bias.push_back({i, -INFINITY});
}
}
}
} }
{ {
@ -2217,10 +2225,6 @@ struct server_context {
slot.params.n_predict = slot.n_predict; slot.params.n_predict = slot.n_predict;
} }
if (slot.params.ignore_eos && has_eos_token) {
slot.params.sampling.logit_bias.push_back({llama_vocab_eos(vocab), -INFINITY});
}
{ {
if (slot.smpl != nullptr) { if (slot.smpl != nullptr) {
common_sampler_free(slot.smpl); common_sampler_free(slot.smpl);