matt/ollama37
1
0
Fork 0
mirror of https://github.com/dogkeeper886/ollama37.git synced 2025-12-10 15:57:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'upstream/main' into pr3702

Browse Source
This commit is contained in:
Daniel Hiltgen
2024-05-08 16:44:35 -07:00
parent c496967e56 ee49844d09
commit 920a4b0794
132 changed files with 7701 additions and 4766 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
llm/ext_server/server.cpp vendored
View File

@@ -1032,7 +1032,7 @@ struct llama_server_context
slot.has_next_token = false;
}
if (!slot.cache_tokens.empty() && result.tok == llama_token_eos(model))
if (!slot.cache_tokens.empty() && llama_token_is_eog(model, result.tok))
{
slot.stopped_eos = true;
slot.has_next_token = false;
@@ -1144,12 +1144,15 @@ struct llama_server_context
res.result_json = json
{
{"content", tkn.text_to_send},
{"stop", false},
{"slot_id", slot.id},
{"multimodal", multimodal}
};
if (!llama_token_is_eog(model, tkn.tok)) {
res.result_json["content"] = tkn.text_to_send;
}
if (slot.sparams.n_probs > 0)
{
std::vector<completion_token_output> probs_output = {};
@@ -1183,8 +1186,6 @@ struct llama_server_context
{"model", params.model_alias},
{"tokens_predicted", slot.n_decoded},
{"tokens_evaluated", slot.n_prompt_tokens},
{"generation_settings", get_formated_generation(slot)},
{"prompt", slot.prompt},
{"truncated", slot.truncated},
{"stopped_eos", slot.stopped_eos},
{"stopped_word", slot.stopped_word},
@@ -2644,18 +2645,18 @@ static void server_params_parse(int argc, char **argv, server_params &sparams,
if (strncmp(sep, "int:", 4) == 0) {
sep += 4;
kvo.tag = LLAMA_KV_OVERRIDE_TYPE_INT;
kvo.int_value = std::atol(sep);
kvo.val_i64 = std::atol(sep);
} else if (strncmp(sep, "float:", 6) == 0) {
sep += 6;
kvo.tag = LLAMA_KV_OVERRIDE_TYPE_FLOAT;
kvo.float_value = std::atof(sep);
kvo.val_f64 = std::atof(sep);
} else if (strncmp(sep, "bool:", 5) == 0) {
sep += 5;
kvo.tag = LLAMA_KV_OVERRIDE_TYPE_BOOL;
if (std::strcmp(sep, "true") == 0) {
kvo.bool_value = true;
kvo.val_bool = true;
} else if (std::strcmp(sep, "false") == 0) {
kvo.bool_value = false;
kvo.val_bool = false;
} else {
fprintf(stderr, "error: Invalid boolean value for KV override: %s\n", argv[i]);
invalid_param = true;