Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-19 01:05:09 +02:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'master' into concedo_experimental

Browse Source 
# Conflicts:
#	.flake8
#	.github/workflows/python-lint.yml
#	flake.lock
#	ggml-cuda.cu
#	ggml-quants.c
#	llama.cpp
#	pocs/vdot/q8dot.cpp
#	pocs/vdot/vdot.cpp
#	tests/test-quantize-fns.cpp
#	tests/test-quantize-perf.cpp
This commit is contained in:
Concedo
2024-02-13 00:14:22 +08:00
parent fdc54efcf0 43fe07c1a4
commit 3cec37c2e0
35 changed files with 2814 additions and 1749 deletions
Download Patch File Download Diff File Expand all files Collapse all files
examples/server/README.md
+1 -1
View File
@@ -185,7 +185,7 @@ node index.js
`ignore_eos`: Ignore end of stream token and continue generating (default: false).
`logit_bias`: Modify the likelihood of a token appearing in the generated text completion. For example, use `"logit_bias": [[15043,1.0]]` to increase the likelihood of the token 'Hello', or `"logit_bias": [[15043,-1.0]]` to decrease its likelihood. Setting the value to false, `"logit_bias": [[15043,false]]` ensures that the token `Hello` is never produced (default: []).
`logit_bias`: Modify the likelihood of a token appearing in the generated text completion. For example, use `"logit_bias": [[15043,1.0]]` to increase the likelihood of the token 'Hello', or `"logit_bias": [[15043,-1.0]]` to decrease its likelihood. Setting the value to false, `"logit_bias": [[15043,false]]` ensures that the token `Hello` is never produced. The tokens can also be represented as strings, e.g. `[["Hello, World!",-0.5]]` will reduce the likelihood of all the individual tokens that represent the string `Hello, World!`, just like the `presence_penalty` does. (default: []).
`n_probs`: If greater than 0, the response also contains the probabilities of top N tokens for each generated token (default: 0)
examples/server/oai.hpp
+6 -2
View File
@@ -15,9 +15,13 @@
using json = nlohmann::json;
inline static json oaicompat_completion_params_parse(
const json &body /* openai api json semantics */)
const json &body, /* openai api json semantics */
const std::string &chat_template)
{
json llama_params;
std::string formatted_prompt = chat_template == "chatml"
? format_chatml(body["messages"]) // OpenAI 'messages' to chatml (with <|im_start|>,...)
: format_llama2(body["messages"]); // OpenAI 'messages' to llama2 (with [INST],...)
llama_params["__oaicompat"] = true;
@@ -30,7 +34,7 @@ inline static json oaicompat_completion_params_parse(
// https://platform.openai.com/docs/api-reference/chat/create
llama_sampling_params default_sparams;
llama_params["model"] = json_value(body, "model", std::string("unknown"));
llama_params["prompt"] = format_chatml(body["messages"]); // OpenAI 'messages' to llama.cpp 'prompt'
llama_params["prompt"] = formatted_prompt;
llama_params["cache_prompt"] = json_value(body, "cache_prompt", false);
llama_params["temperature"] = json_value(body, "temperature", 0.0);
llama_params["top_k"] = json_value(body, "top_k", default_sparams.top_k);
examples/server/server.cpp
+45 -9
View File
@@ -37,6 +37,7 @@ struct server_params
std::string hostname = "127.0.0.1";
std::vector<std::string> api_keys;
std::string public_path = "examples/server/public";
std::string chat_template = "chatml";
int32_t port = 8080;
int32_t read_timeout = 600;
int32_t write_timeout = 600;
@@ -626,18 +627,36 @@ struct llama_server_context
const int n_vocab = llama_n_vocab(model);
for (const auto &el : *logit_bias)
{
if (el.is_array() && el.size() == 2 && el[0].is_number_integer())
if (el.is_array() && el.size() == 2)
{
llama_token tok = el[0].get<llama_token>();
if (tok >= 0 && tok < n_vocab)
float bias;
if (el[1].is_number())
{
if (el[1].is_number())
bias = el[1].get<float>();
}
else if (el[1].is_boolean() && !el[1].get<bool>())
{
bias = -INFINITY;
}
else
{
continue;
}
if (el[0].is_number_integer())
{
llama_token tok = el[0].get<llama_token>();
if (tok >= 0 && tok < n_vocab)
{
slot->sparams.logit_bias[tok] = el[1].get<float>();
slot->sparams.logit_bias[tok] = bias;
}
else if (el[1].is_boolean() && !el[1].get<bool>())
}
else if (el[0].is_string())
{
auto toks = llama_tokenize(model, el[0].get<std::string>(), false);
for (auto tok : toks)
{
slot->sparams.logit_bias[tok] = -INFINITY;
slot->sparams.logit_bias[tok] = bias;
}
}
}
@@ -1860,6 +1879,8 @@ static void server_print_usage(const char *argv0, const gpt_params &params,
printf(" types: int, float, bool. example: --override-kv tokenizer.ggml.add_bos_token=bool:false\n");
printf(" -gan N, --grp-attn-n N set the group attention factor to extend context size through self-extend(default: 1=disabled), used together with group attention width `--grp-attn-w`");
printf(" -gaw N, --grp-attn-w N set the group attention width to extend context size through self-extend(default: 512), used together with group attention factor `--grp-attn-n`");
printf(" --chat-template FORMAT_NAME");
printf(" set chat template, possible valus is: llama2, chatml (default %s)", sparams.chat_template.c_str());
printf("\n");
}
@@ -2291,6 +2312,21 @@ static void server_params_parse(int argc, char **argv, server_params &sparams,
log_set_target(stdout);
LOG_INFO("logging to file is disabled.", {});
}
else if (arg == "--chat-template")
{
if (++i >= argc)
{
invalid_param = true;
break;
}
std::string value(argv[i]);
if (value != "chatml" && value != "llama2") {
fprintf(stderr, "error: chat template can be \"llama2\" or \"chatml\", but got: %s\n", value.c_str());
invalid_param = true;
break;
}
sparams.chat_template = value;
}
else if (arg == "--override-kv")
{
if (++i >= argc) {
@@ -2744,13 +2780,13 @@ int main(int argc, char **argv)
// TODO: add mount point without "/v1" prefix -- how?
svr.Post("/v1/chat/completions", [&llama, &validate_api_key](const httplib::Request &req, httplib::Response &res)
svr.Post("/v1/chat/completions", [&llama, &validate_api_key, &sparams](const httplib::Request &req, httplib::Response &res)
{
res.set_header("Access-Control-Allow-Origin", req.get_header_value("Origin"));
if (!validate_api_key(req, res)) {
return;
}
json data = oaicompat_completion_params_parse(json::parse(req.body));
json data = oaicompat_completion_params_parse(json::parse(req.body), sparams.chat_template);
const int task_id = llama.queue_tasks.get_new_id();
llama.queue_results.add_waiting_task_id(task_id);
examples/server/utils.hpp
+30
View File
@@ -167,6 +167,34 @@ static T json_value(const json &body, const std::string &key, const T &default_v
: default_value;
}
inline std::string format_llama2(std::vector<json> messages)
{
std::ostringstream output;
bool is_inside_turn = false;
for (auto it = messages.begin(); it != messages.end(); ++it) {
if (!is_inside_turn) {
output << "[INST] ";
}
std::string role = json_value(*it, "role", std::string("user"));
std::string content = json_value(*it, "content", std::string(""));
if (role == "system") {
output << "<<SYS>>\n" << content << "\n<<SYS>>\n\n";
is_inside_turn = true;
} else if (role == "user") {
output << content << " [/INST]";
is_inside_turn = true;
} else {
output << " " << content << " </s>";
is_inside_turn = false;
}
}
LOG_VERBOSE("format_llama2", {{"text", output.str()}});
return output.str();
}
inline std::string format_chatml(std::vector<json> messages)
{
std::ostringstream chatml_msgs;
@@ -180,6 +208,8 @@ inline std::string format_chatml(std::vector<json> messages)
chatml_msgs << "<|im_start|>assistant" << '\n';
LOG_VERBOSE("format_chatml", {{"text", chatml_msgs.str()}});
return chatml_msgs.str();
}