ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

llama : rework embeddings logic (#14208)

Browse source 
* llama : rework embeddings logic

ggml-ci

* cont : fix rerank

ggml-ci

* cont : engrish [no ci]

* cont : fix rerank

ggml-ci

* server : support both embeddings and completions with single model

ggml-ci

* cont : avoid embeddings_org

ggml-ci
This commit is contained in:
Georgi Gerganov 2025-06-16 14:14:00 +03:00 • committed by GitHub
parent 3ba0d843c6
commit d3e64b9f49
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
16 changed files with 159 additions and 114 deletions
Download patch file Download diff file Expand all files Collapse all files

9
common/arg.cpp
View file

@ -988,10 +988,6 @@ static bool common_params_parse_ex(int argc, char ** argv, common_params_context
params.tensor_buft_overrides.push_back({nullptr, nullptr});
}
if (params.reranking && params.embedding) {
throw std::invalid_argument("error: either --embedding or --reranking can be specified, but not both");
}
if (!params.chat_template.empty() && !common_chat_verify_template(params.chat_template, params.use_jinja)) {
throw std::runtime_error(string_format(
"error: the supplied chat template is not supported: %s%s\n",
@ -2747,9 +2743,10 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_EMBEDDINGS"));
add_opt(common_arg(
{"--reranking", "--rerank"},
string_format("enable reranking endpoint on server (default: %s)", params.reranking ? "enabled" : "disabled"),
string_format("enable reranking endpoint on server (default: %s)", "disabled"),
[](common_params & params) {
params.reranking = true;
params.embedding = true;
params.pooling_type = LLAMA_POOLING_TYPE_RANK;
}
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_RERANKING"));
add_opt(common_arg(