Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-17 16:25:01 +02:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'master' into concedo_experimental

Browse Source 
# Conflicts:
#	README.md
#	ci/run.sh
#	llama.cpp
#	scripts/sync-ggml.last
This commit is contained in:
Concedo
2024-03-01 17:41:37 +08:00
parent e5861e993d f105471ef6
commit 55af5446ad
23 changed files with 1069 additions and 743 deletions
Download Patch File Download Diff File Expand all files Collapse all files
examples/server/server.cpp
+234 -284
View File
@@ -34,8 +34,7 @@
using json = nlohmann::json;
struct server_params
{
struct server_params {
std::string hostname = "127.0.0.1";
std::vector<std::string> api_keys;
std::string public_path = "examples/server/public";
@@ -50,103 +49,50 @@ struct server_params
bool server_verbose = false;
bool server_log_json = true;
static size_t common_part(const std::vector<llama_token> &a, const std::vector<llama_token> &b)
{
size_t i;
for (i = 0; i < a.size() && i < b.size() && a[i] == b[i]; i++)
{
}
return i;
}
enum stop_type
{
enum stop_type {
STOP_FULL,
STOP_PARTIAL,
};
static bool ends_with(const std::string &str, const std::string &suffix)
{
return str.size() >= suffix.size() &&
0 == str.compare(str.size() - suffix.size(), suffix.size(), suffix);
}
// TODO: can become bool if we can't find use of more states
enum slot_state {
IDLE,
PROCESSING,
};
static size_t find_partial_stop_string(const std::string &stop,
const std::string &text)
{
if (!text.empty() && !stop.empty())
{
const char text_last_char = text.back();
for (int64_t char_index = stop.size() - 1; char_index >= 0; char_index--)
{
if (stop[char_index] == text_last_char)
{
const std::string current_partial = stop.substr(0, char_index + 1);
if (ends_with(text, current_partial))
{
return text.size() - char_index - 1;
}
}
}
}
return std::string::npos;
}
enum slot_command {
NONE,
LOAD_PROMPT,
RELEASE,
};
// TODO: reuse llama_detokenize
template <class Iter>
static std::string tokens_to_str(llama_context *ctx, Iter begin, Iter end)
{
std::string ret;
for (; begin != end; ++begin)
{
ret += llama_token_to_piece(ctx, *begin);
}
return ret;
}
struct slot_params {
bool stream = true;
bool cache_prompt = false; // remember the prompt to avoid reprocessing all prompt
// format incomplete utf-8 multibyte character for output
static std::string tokens_to_output_formatted_string(const llama_context *ctx, const llama_token token)
{
std::string out = token == -1 ? "" : llama_token_to_piece(ctx, token);
// if the size is 1 and first bit is 1, meaning it's a partial character
// (size > 1 meaning it's already a known token)
if (out.size() == 1 && (out[0] & 0x80) == 0x80)
{
std::stringstream ss;
ss << std::hex << (out[0] & 0xff);
std::string res(ss.str());
out = "byte: \\x" + res;
}
return out;
}
uint32_t seed = -1; // RNG seed
int32_t n_keep = 0; // number of tokens to keep from initial prompt
int32_t n_predict = -1; // new tokens to predict
// convert a vector of completion_token_output to json
static json probs_vector_to_json(const llama_context *ctx, const std::vector<completion_token_output> &probs)
{
json out = json::array();
for (const auto &prob : probs)
{
json probs_for_token = json::array();
for (const auto &p : prob.probs)
{
std::string tok_str = tokens_to_output_formatted_string(ctx, p.tok);
probs_for_token.push_back(json
{
{"tok_str", tok_str},
{"prob", p.prob},
});
}
std::string tok_str = tokens_to_output_formatted_string(ctx, prob.tok);
out.push_back(json{
{"content", tok_str},
{"probs", probs_for_token},
});
}
return out;
}
std::vector<std::string> antiprompt;
struct llama_client_slot
{
json input_prefix;
json input_suffix;
};
struct slot_image {
int32_t id;
bool request_encode_image = false;
float * image_embedding = nullptr;
int32_t image_tokens = 0;
clip_image_u8 * img_data;
std::string prefix_prompt; // before of this image
};
struct server_slot {
int id;
int task_id = -1;
@@ -166,8 +112,8 @@ struct llama_client_slot
int32_t i_batch = -1;
int32_t n_predict = -1;
int32_t num_prompt_tokens = 0;
int32_t num_prompt_tokens_processed = 0;
int32_t n_prompt_tokens = 0;
int32_t n_prompt_tokens_processed = 0;
json prompt;
std::string generated_text;
@@ -202,8 +148,8 @@ struct llama_client_slot
std::vector<slot_image> images;
// stats
size_t sent_count = 0;
size_t sent_token_probs_index = 0;
size_t n_sent_text = 0; // number of sent text character
size_t n_sent_token_probs = 0;
int64_t t_start_process_prompt;
int64_t t_start_genereration;
@@ -215,7 +161,7 @@ struct llama_client_slot
int multitask_id = -1;
void reset() {
num_prompt_tokens = 0;
n_prompt_tokens = 0;
generated_text = "";
truncated = false;
stopped_eos = false;
@@ -223,16 +169,15 @@ struct llama_client_slot
stopped_limit = false;
stopping_word = "";
n_past = 0;
sent_count = 0;
sent_token_probs_index = 0;
n_sent_text = 0;
n_sent_token_probs = 0;
infill = false;
ga_i = 0;
n_past_se = 0;
generated_token_probs.clear();
for (slot_image & img : images)
{
for (slot_image & img : images) {
free(img.image_embedding);
if (img.img_data) {
clip_image_u8_free(img.img_data);
@@ -244,19 +189,15 @@ struct llama_client_slot
}
bool has_budget(gpt_params &global_params) {
if (params.n_predict == -1 && global_params.n_predict == -1)
{
if (params.n_predict == -1 && global_params.n_predict == -1) {
return true; // limitless
}
n_remaining = -1;
if (params.n_predict != -1)
{
if (params.n_predict != -1) {
n_remaining = params.n_predict - n_decoded;
}
else if (global_params.n_predict != -1)
{
} else if (global_params.n_predict != -1) {
n_remaining = global_params.n_predict - n_decoded;
}
@@ -272,8 +213,7 @@ struct llama_client_slot
}
void add_token_string(const completion_token_output &token) {
if (command == RELEASE)
{
if (command == RELEASE) {
return;
}
cache_tokens.push_back(token.tok);
@@ -291,10 +231,10 @@ struct llama_client_slot
json get_formated_timings() {
return json
{
{"prompt_n", num_prompt_tokens_processed},
{"prompt_n", n_prompt_tokens_processed},
{"prompt_ms", t_prompt_processing},
{"prompt_per_token_ms", t_prompt_processing / num_prompt_tokens_processed},
{"prompt_per_second", 1e3 / t_prompt_processing * num_prompt_tokens_processed},
{"prompt_per_token_ms", t_prompt_processing / n_prompt_tokens_processed},
{"prompt_per_second", 1e3 / t_prompt_processing * n_prompt_tokens_processed},
{"predicted_n", n_decoded},
{"predicted_ms", t_token_generation},
@@ -305,18 +245,18 @@ struct llama_client_slot
void print_timings() const {
char buffer[512];
double t_token = t_prompt_processing / num_prompt_tokens_processed;
double n_tokens_second = 1e3 / t_prompt_processing * num_prompt_tokens_processed;
double t_token = t_prompt_processing / n_prompt_tokens_processed;
double n_tokens_second = 1e3 / t_prompt_processing * n_prompt_tokens_processed;
sprintf(buffer, "prompt eval time = %10.2f ms / %5d tokens (%8.2f ms per token, %8.2f tokens per second)",
t_prompt_processing, num_prompt_tokens_processed,
t_prompt_processing, n_prompt_tokens_processed,
t_token, n_tokens_second);
LOG_INFO(buffer, {
{"slot_id", id},
{"task_id", task_id},
{"t_prompt_processing", t_prompt_processing},
{"num_prompt_tokens_processed", num_prompt_tokens_processed},
{"t_token", t_token},
{"n_tokens_second", n_tokens_second},
{"slot_id", id},
{"task_id", task_id},
{"t_prompt_processing", t_prompt_processing},
{"n_prompt_tokens_processed", n_prompt_tokens_processed},
{"t_token", t_token},
{"n_tokens_second", n_tokens_second},
});
t_token = t_token_generation / n_decoded;
@@ -344,7 +284,7 @@ struct llama_client_slot
}
};
struct llama_metrics {
struct server_metrics {
uint64_t n_prompt_tokens_processed_total = 0;
uint64_t n_tokens_predicted_total = 0;
@@ -355,18 +295,16 @@ struct llama_metrics {
uint64_t t_tokens_generation = 0;
void on_prompt_eval(const llama_client_slot &slot) {
n_prompt_tokens_processed_total += slot.num_prompt_tokens_processed;
n_prompt_tokens_processed += slot.num_prompt_tokens_processed;
t_prompt_processing += slot.t_prompt_processing;
void on_prompt_eval(const server_slot &slot) {
n_prompt_tokens_processed_total += slot.n_prompt_tokens_processed;
n_prompt_tokens_processed += slot.n_prompt_tokens_processed;
t_prompt_processing += slot.t_prompt_processing;
}
void on_prediction(const llama_client_slot &slot) {
void on_prediction(const server_slot &slot) {
n_tokens_predicted_total += slot.n_decoded;
n_tokens_predicted += slot.n_decoded;
t_tokens_generation += slot.t_token_generation;
n_tokens_predicted += slot.n_decoded;
t_tokens_generation += slot.t_token_generation;
}
void reset_bucket() {
@@ -405,13 +343,13 @@ struct llama_server_context
std::string name_assistant;
// slots / clients
std::vector<llama_client_slot> slots;
std::vector<server_slot> slots;
json default_generation_settings_for_props;
llama_server_queue queue_tasks;
llama_server_queue queue_tasks;
llama_server_response queue_results;
llama_metrics metrics;
server_metrics metrics;
~llama_server_context()
{
@@ -488,7 +426,7 @@ struct llama_server_context
LOG_INFO("initializing slots", {{"n_slots", params.n_parallel}});
for (int i = 0; i < params.n_parallel; i++)
{
llama_client_slot slot;
server_slot slot;
slot.id = i;
slot.n_ctx = n_ctx_slot;
@@ -580,11 +518,11 @@ struct llama_server_context
return prompt_tokens;
}
llama_client_slot* get_slot(int id) {
server_slot* get_slot(int id) {
int64_t t_last = ggml_time_us();
llama_client_slot *last_used = nullptr;
server_slot *last_used = nullptr;
for (llama_client_slot & slot : slots)
for (server_slot & slot : slots)
{
if (slot.id == id && slot.available())
{
@@ -601,7 +539,7 @@ struct llama_server_context
return last_used;
}
bool launch_slot_with_data(llama_client_slot* &slot, json data) {
bool launch_slot_with_data(server_slot* &slot, json data) {
slot_params default_params;
llama_sampling_params default_sparams;
@@ -889,7 +827,7 @@ struct llama_server_context
clean_kv_cache = false;
}
void update_system_prompt() {
void system_prompt_update() {
kv_cache_clear();
system_tokens.clear();
@@ -934,9 +872,9 @@ struct llama_server_context
system_need_update = false;
}
void notify_system_prompt_changed() {
void system_prompt_notify() {
// release all slots
for (llama_client_slot &slot : slots)
for (server_slot &slot : slots)
{
slot.release();
}
@@ -944,17 +882,17 @@ struct llama_server_context
system_need_update = true;
}
void process_system_prompt_data(const json &sys_props) {
void system_prompt_process(const json &sys_props) {
system_prompt = sys_props.value("prompt", "");
name_user = sys_props.value("anti_prompt", "");
name_assistant = sys_props.value("assistant_name", "");
notify_system_prompt_changed();
system_prompt_notify();
}
static size_t find_stopping_strings(const std::string &text, const size_t last_token_size,
const stop_type type, llama_client_slot &slot)
const stop_type type, server_slot &slot)
{
size_t stop_pos = std::string::npos;
@@ -976,8 +914,8 @@ struct llama_server_context
{
if (type == STOP_FULL)
{
slot.stopped_word = true;
slot.stopping_word = word;
slot.stopped_word = true;
slot.stopping_word = word;
slot.has_next_token = false;
}
stop_pos = pos;
@@ -987,7 +925,7 @@ struct llama_server_context
return stop_pos;
}
bool process_token(completion_token_output &result, llama_client_slot &slot) {
bool process_token(completion_token_output &result, server_slot &slot) {
// remember which tokens were sampled - used for repetition penalties during sampling
const std::string token_str = llama_token_to_piece(ctx, result.tok);
slot.sampled = result.tok;
@@ -1033,7 +971,7 @@ struct llama_server_context
if (!incomplete)
{
size_t pos = std::min(slot.sent_count, slot.generated_text.size());
size_t pos = std::min(slot.n_sent_text, slot.generated_text.size());
const std::string str_test = slot.generated_text.substr(pos);
bool is_stop_full = false;
size_t stop_pos = find_stopping_strings(str_test, token_str.size(), STOP_FULL, slot);
@@ -1043,7 +981,7 @@ struct llama_server_context
slot.generated_text.erase(
slot.generated_text.begin() + pos + stop_pos,
slot.generated_text.end());
pos = std::min(slot.sent_count, slot.generated_text.size());
pos = std::min(slot.n_sent_text, slot.generated_text.size());
}
else
{
@@ -1056,7 +994,7 @@ struct llama_server_context
{
// no send the stop word in the response
result.text_to_send = slot.generated_text.substr(pos, std::string::npos);
slot.sent_count += result.text_to_send.size();
slot.n_sent_text += result.text_to_send.size();
// add the token to slot queue and cache
}
slot.add_token_string(result);
@@ -1100,7 +1038,7 @@ struct llama_server_context
return slot.has_next_token; // continue
}
bool process_images(llama_client_slot &slot) const
bool process_images(server_slot &slot) const
{
for (slot_image &img : slot.images)
{
@@ -1133,7 +1071,7 @@ struct llama_server_context
queue_results.send(res);
}
json get_formated_generation(llama_client_slot &slot)
json get_formated_generation(server_slot &slot)
{
const auto eos_bias = slot.sparams.logit_bias.find(llama_token_eos(model));
const bool ignore_eos = eos_bias != slot.sparams.logit_bias.end() &&
@@ -1180,7 +1118,7 @@ struct llama_server_context
};
}
void send_partial_response(llama_client_slot &slot, completion_token_output tkn)
void send_partial_response(server_slot &slot, completion_token_output tkn)
{
task_result res;
res.id = slot.task_id;
@@ -1200,13 +1138,13 @@ struct llama_server_context
{
std::vector<completion_token_output> probs_output = {};
const std::vector<llama_token> to_send_toks = llama_tokenize(ctx, tkn.text_to_send, false);
size_t probs_pos = std::min(slot.sent_token_probs_index, slot.generated_token_probs.size());
size_t probs_stop_pos = std::min(slot.sent_token_probs_index + to_send_toks.size(), slot.generated_token_probs.size());
size_t probs_pos = std::min(slot.n_sent_token_probs, slot.generated_token_probs.size());
size_t probs_stop_pos = std::min(slot.n_sent_token_probs + to_send_toks.size(), slot.generated_token_probs.size());
if (probs_pos < probs_stop_pos)
{
probs_output = std::vector<completion_token_output>(slot.generated_token_probs.begin() + probs_pos, slot.generated_token_probs.begin() + probs_stop_pos);
}
slot.sent_token_probs_index = probs_stop_pos;
slot.n_sent_token_probs = probs_stop_pos;
res.result_json["completion_probabilities"] = probs_vector_to_json(ctx, probs_output);
}
@@ -1219,7 +1157,7 @@ struct llama_server_context
queue_results.send(res);
}
void send_final_response(llama_client_slot &slot)
void send_final_response(server_slot &slot)
{
task_result res;
res.id = slot.task_id;
@@ -1234,7 +1172,7 @@ struct llama_server_context
{"stop", true},
{"model", params.model_alias},
{"tokens_predicted", slot.n_decoded},
{"tokens_evaluated", slot.num_prompt_tokens},
{"tokens_evaluated", slot.n_prompt_tokens},
{"generation_settings", get_formated_generation(slot)},
{"prompt", slot.prompt},
{"truncated", slot.truncated},
@@ -1272,7 +1210,7 @@ struct llama_server_context
queue_results.send(res);
}
void send_embedding(llama_client_slot &slot)
void send_embedding(server_slot &slot)
{
task_result res;
res.id = slot.task_id;
@@ -1283,9 +1221,7 @@ struct llama_server_context
const int n_embd = llama_n_embd(model);
if (!params.embedding)
{
LOG_WARNING("embedding disabled", {
{"params.embedding", params.embedding},
});
LOG_WARNING("embedding disabled", {{"params.embedding", params.embedding}});
res.result_json = json
{
{"embedding", std::vector<float>(n_embd, 0.0f)},
@@ -1297,7 +1233,7 @@ struct llama_server_context
std::vector<float> embedding(data, data + n_embd);
res.result_json = json
{
{"embedding", embedding },
{"embedding", embedding},
};
}
queue_results.send(res);
@@ -1346,7 +1282,7 @@ struct llama_server_context
}
// for multiple images processing
bool ingest_images(llama_client_slot &slot, int n_batch)
bool ingest_images(server_slot &slot, int n_batch)
{
int image_idx = 0;
@@ -1385,7 +1321,17 @@ struct llama_server_context
}
const int n_embd = llama_n_embd(model);
llama_batch batch_img = { n_eval, nullptr, (img.image_embedding + i * n_embd), nullptr, nullptr, nullptr, nullptr, slot.n_past, 1, 0, };
llama_batch batch_img = {
n_eval,
nullptr,
(img.image_embedding + i * n_embd),
nullptr,
nullptr,
nullptr,
nullptr,
slot.n_past,
1, 0
};
if (llama_decode(ctx, batch_img))
{
LOG_TEE("%s : failed to eval image\n", __func__);
@@ -1455,7 +1401,7 @@ struct llama_server_context
switch (task.type)
{
case TASK_TYPE_COMPLETION: {
llama_client_slot *slot = get_slot(json_value(task.data, "slot_id", -1));
server_slot *slot = get_slot(json_value(task.data, "slot_id", -1));
if (slot == nullptr)
{
// if no slot is available, we defer this task for processing later
@@ -1470,10 +1416,10 @@ struct llama_server_context
send_error(task, "system prompt can only be updated when all slots are idle");
break;
}
process_system_prompt_data(task.data["system_prompt"]);
system_prompt_process(task.data["system_prompt"]);
// reset cache_tokens for all slots
for (llama_client_slot &slot : slots)
for (server_slot &slot : slots)
{
slot.cache_tokens.clear();
slot.n_past = 0;
@@ -1513,20 +1459,20 @@ struct llama_server_context
int n_idle_slots = 0;
int n_processing_slots = 0;
for (llama_client_slot &slot: slots) {
for (server_slot &slot: slots) {
json slot_data = get_formated_generation(slot);
slot_data["id"] = slot.id;
slot_data["task_id"] = slot.task_id;
slot_data["state"] = slot.state;
slot_data["prompt"] = slot.prompt;
slot_data["next_token"] = {
{"has_next_token", slot.has_next_token},
{"n_remain", slot.n_remaining},
{"has_next_token", slot.has_next_token},
{"n_remain", slot.n_remaining},
{"num_tokens_predicted", slot.n_decoded},
{"stopped_eos", slot.stopped_eos},
{"stopped_word", slot.stopped_word},
{"stopped_limit", slot.stopped_limit},
{"stopping_word", slot.stopping_word},
{"stopped_eos", slot.stopped_eos},
{"stopped_word", slot.stopped_word},
{"stopped_limit", slot.stopped_limit},
{"stopping_word", slot.stopping_word},
};
if (slot_data["state"] == IDLE) {
n_idle_slots++;
@@ -1564,10 +1510,10 @@ struct llama_server_context
{ "n_tokens_predicted", metrics.n_tokens_predicted},
{ "t_tokens_generation", metrics.t_tokens_generation},
{ "kv_cache_tokens_count", llama_get_kv_cache_token_count(ctx)},
{ "kv_cache_used_cells", llama_get_kv_cache_used_cells(ctx)},
{ "kv_cache_tokens_count", llama_get_kv_cache_token_count(ctx)},
{ "kv_cache_used_cells", llama_get_kv_cache_used_cells(ctx)},
{ "slots", slots_data },
{ "slots", slots_data },
};
metrics.reset_bucket();
queue_results.send(res);
@@ -1598,7 +1544,7 @@ struct llama_server_context
if (system_need_update)
{
LOG_INFO("updating system prompt", {});
update_system_prompt();
system_prompt_update();
}
llama_batch_clear(batch);
@@ -1619,7 +1565,7 @@ struct llama_server_context
task.target_id = -1;
queue_tasks.post(task);
for (llama_client_slot &slot : slots)
for (server_slot &slot : slots)
{
if (slot.ga_n == 1)
{
@@ -1755,45 +1701,50 @@ struct llama_server_context
prompt_tokens = tokenize(slot.prompt, system_prompt.empty() && add_bos_token); // add BOS if there isn't system prompt
}
slot.num_prompt_tokens = prompt_tokens.size();
slot.n_prompt_tokens = prompt_tokens.size();
if (slot.params.n_keep < 0)
{
slot.params.n_keep = slot.num_prompt_tokens;
slot.params.n_keep = slot.n_prompt_tokens;
}
slot.params.n_keep = std::min(slot.n_ctx - 4, slot.params.n_keep);
// if input prompt is too big, truncate it
if (slot.num_prompt_tokens >= slot.n_ctx)
if (slot.n_prompt_tokens >= slot.n_ctx)
{
const int n_left = slot.n_ctx - slot.params.n_keep;
const int n_block_size = n_left / 2;
const int erased_blocks = (slot.num_prompt_tokens - slot.params.n_keep - n_block_size) / n_block_size;
const int erased_blocks = (slot.n_prompt_tokens - slot.params.n_keep - n_block_size) / n_block_size;
std::vector<llama_token> new_tokens(prompt_tokens.begin(), prompt_tokens.begin() + slot.params.n_keep);
new_tokens.insert(new_tokens.end(), prompt_tokens.begin() + slot.params.n_keep + erased_blocks * n_block_size, prompt_tokens.end());
std::vector<llama_token> new_tokens(
prompt_tokens.begin(),
prompt_tokens.begin() + slot.params.n_keep);
new_tokens.insert(
new_tokens.end(),
prompt_tokens.begin() + slot.params.n_keep + erased_blocks * n_block_size,
prompt_tokens.end());
LOG_VERBOSE("input truncated", {
{"n_ctx", slot.n_ctx},
{"n_keep", slot.params.n_keep},
{"n_left", n_left},
{"n_ctx", slot.n_ctx},
{"n_keep", slot.params.n_keep},
{"n_left", n_left},
{"new_tokens", tokens_to_str(ctx, new_tokens.cbegin(), new_tokens.cend())},
});
slot.truncated = true;
prompt_tokens = new_tokens;
slot.num_prompt_tokens = prompt_tokens.size();
GGML_ASSERT(slot.num_prompt_tokens < slot.n_ctx);
slot.n_prompt_tokens = prompt_tokens.size();
GGML_ASSERT(slot.n_prompt_tokens < slot.n_ctx);
}
if (!slot.params.cache_prompt)
{
llama_sampling_reset(slot.ctx_sampling);
slot.n_past = 0;
slot.n_past = 0;
slot.n_past_se = 0;
slot.ga_i = 0;
slot.num_prompt_tokens_processed = slot.num_prompt_tokens;
slot.ga_i = 0;
slot.n_prompt_tokens_processed = slot.n_prompt_tokens;
}
else
{
@@ -1812,7 +1763,7 @@ struct llama_server_context
slot.n_past -= 1;
}
slot.num_prompt_tokens_processed = slot.num_prompt_tokens - slot.n_past;
slot.n_prompt_tokens_processed = slot.n_prompt_tokens - slot.n_past;
if (slot.ga_n != 1)
{
@@ -1837,13 +1788,13 @@ struct llama_server_context
{ "slot_id", slot.id },
{ "task_id", slot.task_id },
{ "n_past", slot.n_past },
{ "num_prompt_tokens_processed", slot.num_prompt_tokens_processed }
{ "n_prompt_tokens_processed", slot.n_prompt_tokens_processed }
});
}
slot.cache_tokens = prompt_tokens;
if (slot.n_past == slot.num_prompt_tokens && slot.n_past > 0)
if (slot.n_past == slot.n_prompt_tokens && slot.n_past > 0)
{
// we have to evaluate at least 1 token to generate logits.
LOG_INFO("we have to evaluate at least 1 token to generate logits", {
@@ -1899,8 +1850,8 @@ struct llama_server_context
if (has_images && !ingest_images(slot, n_batch))
{
LOG_ERROR("failed processing images", {
"slot_id", slot.id,
"task_id", slot.task_id,
{"slot_id", slot.id},
{"task_id", slot.task_id},
});
// FIXME @phymbert: to be properly tested
// early returning without changing the slot state will block the slot for ever
@@ -2050,10 +2001,6 @@ struct llama_server_context
LOG_VERBOSE("slots updated", {});
return true;
}
void run_on_all_tasks_finished() {
update_slots();
}
};
static void server_print_usage(const char *argv0, const gpt_params &params,
@@ -2134,8 +2081,8 @@ static void server_print_usage(const char *argv0, const gpt_params &params,
printf(" --override-kv KEY=TYPE:VALUE\n");
printf(" advanced option to override model metadata by key. may be specified multiple times.\n");
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(" -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`\n");
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`\n");
printf(" --chat-template JINJA_TEMPLATE\n");
printf(" set custom jinja chat template (default: template taken from model's metadata)\n");
printf(" Note: only commonly used templates are accepted, since we don't have jinja parser\n");
@@ -2562,7 +2509,7 @@ static void server_params_parse(int argc, char **argv, server_params &sparams,
std::istreambuf_iterator<char>(),
std::back_inserter(systm_content)
);
llama.process_system_prompt_data(json::parse(systm_content));
llama.system_prompt_process(json::parse(systm_content));
}
else if (arg == "-ctk" || arg == "--cache-type-k") {
params.cache_type_k = argv[++i];
@@ -2693,7 +2640,7 @@ static void server_params_parse(int argc, char **argv, server_params &sparams,
/* llama.cpp completion api semantics */
static json format_partial_response(
llama_server_context &llama, llama_client_slot *slot, const std::string &content, const std::vector<completion_token_output> &probs
llama_server_context &llama, server_slot *slot, const std::string &content, const std::vector<completion_token_output> &probs
) {
json res = json
{
@@ -2749,14 +2696,7 @@ static void log_server_request(const httplib::Request &req, const httplib::Respo
});
}
struct token_translator
{
llama_context * ctx;
std::string operator()(llama_token tok) const { return llama_token_to_piece(ctx, tok); }
std::string operator()(const completion_token_output &cto) const { return (*this)(cto.tok); }
};
static void append_to_generated_text_from_generated_token_probs(llama_server_context &llama, llama_client_slot *slot)
static void append_to_generated_text_from_generated_token_probs(llama_server_context &llama, server_slot *slot)
{
auto & gtps = slot->generated_token_probs;
auto translator = token_translator{llama.ctx};
@@ -2773,7 +2713,16 @@ static void append_to_generated_text_from_generated_token_probs(llama_server_con
}
std::function<void(int)> shutdown_handler;
inline void signal_handler(int signal) { shutdown_handler(signal); }
std::atomic_flag is_terminating = ATOMIC_FLAG_INIT;
inline void signal_handler(int signal) {
if (is_terminating.test_and_set()) {
// in case it hangs, we can force terminate the server by hitting Ctrl+C twice
// this is for better developer experience, we can remove when the server is stable enough
fprintf(stderr, "Received second interrupt, terminating immediately.\n");
exit(1);
}
shutdown_handler(signal);
}
int main(int argc, char **argv)
{
@@ -3212,87 +3161,88 @@ int main(int argc, char **argv)
res.set_content(models.dump(), "application/json; charset=utf-8");
});
const auto 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(llama.model, json::parse(req.body), sparams.chat_template);
// TODO: add mount point without "/v1" prefix -- how?
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(llama.model, 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);
llama.request_completion(task_id, data, false, false, -1);
const int task_id = llama.queue_tasks.get_new_id();
llama.queue_results.add_waiting_task_id(task_id);
llama.request_completion(task_id, data, false, false, -1);
if (!json_value(data, "stream", false)) {
std::string completion_text;
task_result result = llama.queue_results.recv(task_id);
if (!json_value(data, "stream", false)) {
std::string completion_text;
task_result result = llama.queue_results.recv(task_id);
if (!result.error && result.stop) {
json oaicompat_result = format_final_response_oaicompat(data, result);
if (!result.error && result.stop) {
json oaicompat_result = format_final_response_oaicompat(data, result);
res.set_content(oaicompat_result.dump(-1, ' ', false,
json::error_handler_t::replace),
"application/json; charset=utf-8");
} else {
res.status = 500;
res.set_content(result.result_json["content"], "text/plain; charset=utf-8");
}
llama.queue_results.remove_waiting_task_id(task_id);
} else {
const auto chunked_content_provider = [task_id, &llama](size_t, httplib::DataSink &sink) {
while (true) {
task_result llama_result = llama.queue_results.recv(task_id);
if (!llama_result.error) {
std::vector<json> result_array = format_partial_response_oaicompat( llama_result);
res.set_content(oaicompat_result.dump(-1, ' ', false,
json::error_handler_t::replace),
"application/json; charset=utf-8");
} else {
res.status = 500;
res.set_content(result.result_json["content"], "text/plain; charset=utf-8");
}
llama.queue_results.remove_waiting_task_id(task_id);
} else {
const auto chunked_content_provider = [task_id, &llama](size_t, httplib::DataSink &sink) {
while (true) {
task_result llama_result = llama.queue_results.recv(task_id);
if (!llama_result.error) {
std::vector<json> result_array = format_partial_response_oaicompat( llama_result);
for (auto it = result_array.begin(); it != result_array.end(); ++it)
{
if (!it->empty()) {
const std::string str =
"data: " +
it->dump(-1, ' ', false, json::error_handler_t::replace) +
"\n\n";
LOG_VERBOSE("data stream", {{"to_send", str}});
if (!sink.write(str.c_str(), str.size())) {
llama.queue_results.remove_waiting_task_id(task_id);
return false;
}
}
}
if (llama_result.stop) {
break;
}
} else {
for (auto it = result_array.begin(); it != result_array.end(); ++it)
{
if (!it->empty()) {
const std::string str =
"error: " +
llama_result.result_json.dump(-1, ' ', false,
json::error_handler_t::replace) +
"data: " +
it->dump(-1, ' ', false, json::error_handler_t::replace) +
"\n\n";
LOG_VERBOSE("data stream", {{"to_send", str}});
if (!sink.write(str.c_str(), str.size())) {
llama.queue_results.remove_waiting_task_id(task_id);
return false;
}
break;
}
}
sink.done();
llama.queue_results.remove_waiting_task_id(task_id);
return true;
};
auto on_complete = [task_id, &llama](bool) {
// cancel request
llama.request_cancel(task_id);
llama.queue_results.remove_waiting_task_id(task_id);
};
res.set_chunked_content_provider("text/event-stream", chunked_content_provider, on_complete);
if (llama_result.stop) {
break;
}
} else {
const std::string str =
"error: " +
llama_result.result_json.dump(-1, ' ', false,
json::error_handler_t::replace) +
"\n\n";
LOG_VERBOSE("data stream", {{"to_send", str}});
if (!sink.write(str.c_str(), str.size())) {
llama.queue_results.remove_waiting_task_id(task_id);
return false;
}
break;
}
}
});
sink.done();
llama.queue_results.remove_waiting_task_id(task_id);
return true;
};
auto on_complete = [task_id, &llama](bool) {
// cancel request
llama.request_cancel(task_id);
llama.queue_results.remove_waiting_task_id(task_id);
};
res.set_chunked_content_provider("text/event-stream", chunked_content_provider, on_complete);
}
};
svr.Post("/chat/completions", chat_completions);
svr.Post("/v1/chat/completions", chat_completions);
svr.Post("/infill", [&llama, &validate_api_key](const httplib::Request &req, httplib::Response &res)
{
@@ -3517,8 +3467,8 @@ int main(int argc, char **argv)
&llama_server_context::process_single_task, &llama, std::placeholders::_1));
llama.queue_tasks.on_finish_multitask(std::bind(
&llama_server_context::on_finish_multitask, &llama, std::placeholders::_1));
llama.queue_tasks.on_all_tasks_finished(std::bind(
&llama_server_context::run_on_all_tasks_finished, &llama));
llama.queue_tasks.on_run_slots(std::bind(
&llama_server_context::update_slots, &llama));
llama.queue_results.on_multitask_update(std::bind(
&llama_server_queue::update_multitask,
&llama.queue_tasks,
examples/server/tests/features/parallel.feature
+22
View File
@@ -54,6 +54,28 @@ Feature: Parallel
| disabled | 128 |
| enabled | 64 |
Scenario Outline: Multi users OAI completions compatibility no v1
Given a system prompt You are a writer.
And a model tinyllama-2
Given a prompt:
"""
Write a very long book.
"""
And a prompt:
"""
Write another a poem.
"""
And <n_predict> max tokens to predict
And streaming is <streaming>
Given concurrent OAI completions requests no v1
Then the server is busy
Then the server is idle
Then all prompts are predicted with <n_predict> tokens
Examples:
| streaming | n_predict |
| disabled | 128 |
| enabled | 64 |
Scenario: Multi users with total number of tokens to predict exceeds the KV Cache size #3969
Given a prompt:
"""
examples/server/tests/features/steps/steps.py
+26 -2
View File
@@ -231,6 +231,7 @@ async def step_oai_chat_completions(context, api_error):
completion = await oai_chat_completions(context.prompts.pop(),
context.system_prompt,
context.base_url,
'/v1/chat',
False,
model=context.model if hasattr(context, 'model') else None,
@@ -288,6 +289,28 @@ async def step_oai_chat_completions(context):
# user_prompt is inserted automatically
context.system_prompt,
context.base_url,
'/v1/chat/completions',
True, # async_client
model=context.model
if hasattr(context, 'model') else None,
n_predict=context.n_predict
if hasattr(context, 'n_predict') else None,
enable_streaming=context.enable_streaming
if hasattr(context, 'enable_streaming') else None,
server_seed=context.server_seed
if hasattr(context, 'server_seed') else None,
user_api_key=context.user_api_key
if hasattr(context, 'user_api_key') else None)
@step(u'concurrent OAI completions requests no v1')
@async_run_until_complete
async def step_oai_chat_completions(context):
await concurrent_requests(context, oai_chat_completions,
# user_prompt is inserted automatically
context.system_prompt,
context.base_url,
'/chat/completions',
True, # async_client
model=context.model
if hasattr(context, 'model') else None,
@@ -497,6 +520,7 @@ async def request_completion(prompt,
async def oai_chat_completions(user_prompt,
system_prompt,
base_url,
base_path,
async_client,
debug=False,
model=None,
@@ -537,7 +561,7 @@ async def oai_chat_completions(user_prompt,
origin = 'llama.cpp'
headers = {'Authorization': f'Bearer {user_api_key}', 'Origin': origin}
async with aiohttp.ClientSession() as session:
async with session.post(f'{base_url}/v1/chat/completions',
async with session.post(f'{base_url}{base_path}',
json=payload,
headers=headers) as response:
if enable_streaming:
@@ -579,7 +603,7 @@ async def oai_chat_completions(user_prompt,
else:
try:
openai.api_key = user_api_key
openai.api_base = f'{base_url}/v1/chat'
openai.api_base = f'{base_url}{base_path}'
chat_completion = openai.Completion.create(
messages=payload['messages'],
model=model,
examples/server/utils.hpp
+122 -64
View File
@@ -37,10 +37,6 @@ extern bool server_log_json;
#define LOG_WARNING(MSG, ...) server_log("WARN", __func__, __LINE__, MSG, __VA_ARGS__)
#define LOG_INFO( MSG, ...) server_log("INFO", __func__, __LINE__, MSG, __VA_ARGS__)
//
// parallel
//
enum server_state {
SERVER_STATE_LOADING_MODEL, // Server is starting up, model not fully loaded yet
SERVER_STATE_READY, // Server is ready and model is loaded
@@ -78,51 +74,8 @@ struct task_multi {
std::vector<task_result> results{};
};
// TODO: can become bool if we can't find use of more states
enum slot_state
{
IDLE,
PROCESSING,
};
enum slot_command
{
NONE,
LOAD_PROMPT,
RELEASE,
};
struct slot_params
{
bool stream = true;
bool cache_prompt = false; // remember the prompt to avoid reprocessing all prompt
uint32_t seed = -1; // RNG seed
int32_t n_keep = 0; // number of tokens to keep from initial prompt
int32_t n_predict = -1; // new tokens to predict
std::vector<std::string> antiprompt;
json input_prefix;
json input_suffix;
};
struct slot_image
{
int32_t id;
bool request_encode_image = false;
float * image_embedding = nullptr;
int32_t image_tokens = 0;
clip_image_u8 * img_data;
std::string prefix_prompt; // before of this image
};
// completion token output with probabilities
struct completion_token_output
{
struct completion_token_output {
struct token_prob
{
llama_token tok;
@@ -134,8 +87,13 @@ struct completion_token_output
std::string text_to_send;
};
static inline void server_log(const char *level, const char *function, int line, const char *message, const nlohmann::ordered_json &extra)
{
struct token_translator {
llama_context * ctx;
std::string operator()(llama_token tok) const { return llama_token_to_piece(ctx, tok); }
std::string operator()(const completion_token_output &cto) const { return (*this)(cto.tok); }
};
static inline void server_log(const char *level, const char *function, int line, const char *message, const nlohmann::ordered_json &extra) {
std::stringstream ss_tid;
ss_tid << std::this_thread::get_id();
json log = nlohmann::ordered_json{
@@ -183,8 +141,7 @@ static inline void server_log(const char *level, const char *function, int line,
//
template <typename T>
static T json_value(const json &body, const std::string &key, const T &default_value)
{
static T json_value(const json &body, const std::string &key, const T &default_value) {
// Fallback null to default value
return body.contains(key) && !body.at(key).is_null()
? body.value(key, default_value)
@@ -200,8 +157,7 @@ inline bool verify_custom_template(const std::string & tmpl) {
}
// Format given chat. If tmpl is empty, we take the template from model metadata
inline std::string format_chat(const struct llama_model * model, const std::string & tmpl, const std::vector<json> & messages)
{
inline std::string format_chat(const struct llama_model * model, const std::string & tmpl, const std::vector<json> & messages) {
size_t alloc_size = 0;
// vector holding all allocated string to be passed to llama_chat_apply_template
std::vector<std::string> str(messages.size() * 2);
@@ -250,7 +206,7 @@ struct llama_server_queue {
// callback functions
std::function<void(task_server&)> callback_new_task;
std::function<void(task_multi&)> callback_finish_multitask;
std::function<void(void)> callback_all_task_finished;
std::function<void(void)> callback_run_slots;
// Add a new task to the end of the queue
int post(task_server task) {
@@ -283,14 +239,14 @@ struct llama_server_queue {
callback_new_task = callback;
}
// Register function to process a multitask
// Register function to process a multitask when it is finished
void on_finish_multitask(std::function<void(task_multi&)> callback) {
callback_finish_multitask = callback;
}
// Register the function to be called when the batch of tasks is finished
void on_all_tasks_finished(std::function<void(void)> callback) {
callback_all_task_finished = callback;
// Register the function to be called when all slots data is ready to be processed
void on_run_slots(std::function<void(void)> callback) {
callback_run_slots = callback;
}
// Call when the state of one slot is changed
@@ -312,7 +268,13 @@ struct llama_server_queue {
condition_tasks.notify_all();
}
// Start the main loop.
/**
* Main loop consists of these steps:
* - Wait until a new task arrives
* - Process the task (i.e. maybe copy data into slot)
* - Check if multitask is finished
* - Run all slots
*/
void start_loop() {
running = true;
while (true) {
@@ -331,8 +293,8 @@ struct llama_server_queue {
LOG_VERBOSE("callback_new_task", {{"task_id", task.id}});
callback_new_task(task);
}
LOG_VERBOSE("callback_all_task_finished", {});
// process and update all the multitasks
LOG_VERBOSE("update_multitasks", {});
// check if we have any finished multitasks
auto queue_iterator = queue_multitasks.begin();
while (queue_iterator != queue_multitasks.end())
{
@@ -349,8 +311,9 @@ struct llama_server_queue {
++queue_iterator;
}
}
// all tasks in the current loop is finished
callback_all_task_finished();
// all tasks in the current loop is processed, slots data is now ready
LOG_VERBOSE("callback_run_slots", {});
callback_run_slots();
}
LOG_VERBOSE("wait for new task", {});
// wait for new task
@@ -408,12 +371,14 @@ struct llama_server_response {
std::mutex mutex_results;
std::condition_variable condition_results;
// add the task_id to the list of tasks waiting for response
void add_waiting_task_id(int task_id) {
LOG_VERBOSE("waiting for task id", {{"task_id", task_id}});
std::unique_lock<std::mutex> lock(mutex_results);
waiting_task_ids.insert(task_id);
}
// when the request is finished, we can remove task associated with it
void remove_waiting_task_id(int task_id) {
LOG_VERBOSE("remove waiting for task id", {{"task_id", task_id}});
std::unique_lock<std::mutex> lock(mutex_results);
@@ -574,3 +539,96 @@ static std::string gen_chatcmplid()
chatcmplid << "chatcmpl-" << random_string();
return chatcmplid.str();
}
//
// other common utils
//
static size_t common_part(const std::vector<llama_token> &a, const std::vector<llama_token> &b)
{
size_t i;
for (i = 0; i < a.size() && i < b.size() && a[i] == b[i]; i++)
{
}
return i;
}
static bool ends_with(const std::string &str, const std::string &suffix)
{
return str.size() >= suffix.size() &&
0 == str.compare(str.size() - suffix.size(), suffix.size(), suffix);
}
static size_t find_partial_stop_string(const std::string &stop,
const std::string &text)
{
if (!text.empty() && !stop.empty())
{
const char text_last_char = text.back();
for (int64_t char_index = stop.size() - 1; char_index >= 0; char_index--)
{
if (stop[char_index] == text_last_char)
{
const std::string current_partial = stop.substr(0, char_index + 1);
if (ends_with(text, current_partial))
{
return text.size() - char_index - 1;
}
}
}
}
return std::string::npos;
}
// TODO: reuse llama_detokenize
template <class Iter>
static std::string tokens_to_str(llama_context *ctx, Iter begin, Iter end)
{
std::string ret;
for (; begin != end; ++begin)
{
ret += llama_token_to_piece(ctx, *begin);
}
return ret;
}
// format incomplete utf-8 multibyte character for output
static std::string tokens_to_output_formatted_string(const llama_context *ctx, const llama_token token)
{
std::string out = token == -1 ? "" : llama_token_to_piece(ctx, token);
// if the size is 1 and first bit is 1, meaning it's a partial character
// (size > 1 meaning it's already a known token)
if (out.size() == 1 && (out[0] & 0x80) == 0x80)
{
std::stringstream ss;
ss << std::hex << (out[0] & 0xff);
std::string res(ss.str());
out = "byte: \\x" + res;
}
return out;
}
// convert a vector of completion_token_output to json
static json probs_vector_to_json(const llama_context *ctx, const std::vector<completion_token_output> &probs)
{
json out = json::array();
for (const auto &prob : probs)
{
json probs_for_token = json::array();
for (const auto &p : prob.probs)
{
std::string tok_str = tokens_to_output_formatted_string(ctx, p.tok);
probs_for_token.push_back(json
{
{"tok_str", tok_str},
{"prob", p.prob},
});
}
std::string tok_str = tokens_to_output_formatted_string(ctx, prob.tok);
out.push_back(json{
{"content", tok_str},
{"probs", probs_for_token},
});
}
return out;
}