vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-11 01:24:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge branch 'master' into concedo_experimental

Browse source 
# Conflicts:
#	.devops/nix/package.nix
#	README.md
#	ggml-metal.m
#	llama.cpp
#	scripts/sync-ggml.last
#	tests/test-backend-ops.cpp
This commit is contained in:
Concedo 2024-03-05 15:54:05 +08:00
commit ac43e0115c
28 changed files with 1225 additions and 309 deletions
Download patch file Download diff file Expand all files Collapse all files

55
examples/server/server.cpp
View file

@ -414,7 +414,7 @@ struct llama_server_context
int res = llama_chat_apply_template(model, nullptr, chat, 1, true, buf.data(), buf.size());
if (res < 0) {
LOG_ERROR("The chat template comes with this model is not yet supported, falling back to chatml. This may cause the model to output suboptimal responses", {});
sparams.chat_template = "<|im_start|>"; // llama_chat_apply_template only checks if <|im_start|> exist in the template
sparams.chat_template = "chatml";
}
}
@ -1211,7 +1211,7 @@ struct llama_server_context
queue_results.send(res);
}
void send_embedding(server_slot &slot)
void send_embedding(server_slot & slot, const llama_batch & batch)
{
task_result res;
res.id = slot.task_id;
@ -1220,6 +1220,7 @@ struct llama_server_context
res.stop = true;
const int n_embd = llama_n_embd(model);
if (!params.embedding)
{
LOG_WARNING("embedding disabled", {{"params.embedding", params.embedding}});
@ -1230,12 +1231,29 @@ struct llama_server_context
}
else
{
const float *data = llama_get_embeddings(ctx);
std::vector<float> embedding(data, data + n_embd);
res.result_json = json
{
{"embedding", embedding},
};
for (int i = 0; i < batch.n_tokens; ++i) {
if (!batch.logits[i] || batch.seq_id[i][0] != slot.id) {
continue;
}
const float * embd = llama_get_embeddings_seq(ctx, batch.seq_id[i][0]);
if (embd == NULL) {
embd = llama_get_embeddings_ith(ctx, i);
if (embd == NULL) {
LOG_ERROR("failed to get embeddings for token", {{"token", batch.token[i]}, {"seq_id", batch.seq_id[i][0]}});
res.result_json = json
{
{"embedding", std::vector<float>(n_embd, 0.0f)},
};
continue;
}
}
res.result_json = json
{
{"embedding", std::vector<float>(embd, embd + n_embd)},
};
}
}
queue_results.send(res);
}
@ -1846,7 +1864,7 @@ struct llama_server_context
ga_i += ga_w/ga_n;
}
}
llama_batch_add(batch, prefix_tokens[slot.n_past], system_tokens.size() + slot_npast, {slot.id }, false);
llama_batch_add(batch, prefix_tokens[slot.n_past], system_tokens.size() + slot_npast, { slot.id }, false);
slot_npast++;
}
@ -1882,7 +1900,7 @@ struct llama_server_context
for (int32_t i = 0; i < (int32_t) batch.n_tokens; i += n_batch)
{
const int32_t n_tokens = std::min(n_batch, (int32_t) (batch.n_tokens - i));
const int32_t n_tokens = std::min(n_batch, batch.n_tokens - i);
for (auto & slot : slots)
{
@ -1955,7 +1973,7 @@ struct llama_server_context
// prompt evaluated for embedding
if (slot.embedding)
{
send_embedding(slot);
send_embedding(slot, batch_view);
slot.release();
slot.i_batch = -1;
continue;
@ -2037,6 +2055,8 @@ static void server_print_usage(const char *argv0, const gpt_params &params,
printf(" --yarn-attn-factor N YaRN: scale sqrt(t) or attention magnitude (default: 1.0)\n");
printf(" --yarn-beta-slow N YaRN: high correction dim or alpha (default: %.1f)\n", params.yarn_beta_slow);
printf(" --yarn-beta-fast N YaRN: low correction dim or beta (default: %.1f)\n", params.yarn_beta_fast);
printf(" --pooling {none,mean,cls}\n");
printf(" pooling type for embeddings, use model default if unspecified\n");
printf(" -b N, --batch-size N batch size for prompt processing (default: %d)\n", params.n_batch);
printf(" --memory-f32 use f32 instead of f16 for memory key+value (default: disabled)\n");
printf(" not recommended: doubles context memory required and no measurable increase in quality\n");
@ -2277,6 +2297,18 @@ static void server_params_parse(int argc, char **argv, server_params &sparams,
}
params.yarn_beta_slow = std::stof(argv[i]);
}
else if (arg == "--pooling")
{
if (++i >= argc) {
invalid_param = true;
break;
}
std::string value(argv[i]);
/**/ if (value == "none") { params.pooling_type = LLAMA_POOLING_TYPE_NONE; }
else if (value == "mean") { params.pooling_type = LLAMA_POOLING_TYPE_MEAN; }
else if (value == "cls") { params.pooling_type = LLAMA_POOLING_TYPE_CLS; }
else { invalid_param = true; break; }
}
else if (arg == "--threads" || arg == "-t")
{
if (++i >= argc)
@ -2331,7 +2363,6 @@ static void server_params_parse(int argc, char **argv, server_params &sparams,
break;
}
params.n_batch = std::stoi(argv[i]);
params.n_batch = std::min(512, params.n_batch);
}
else if (arg == "--gpu-layers" || arg == "-ngl" || arg == "--n-gpu-layers")
{