1
0
Fork 0
mirror of https://gitlab.com/niansa/discord_llama.git synced 2025-03-06 20:48:25 +01:00

Added context size config key

This commit is contained in:
niansa 2023-04-23 22:40:55 +02:00
parent 06b815db08
commit f0e55ee3c6
2 changed files with 6 additions and 2 deletions

View file

@ -8,3 +8,4 @@ mlock false
pool_size 2 pool_size 2
threads 4 threads 4
persistance true persistance true
ctx_size 1012

View file

@ -168,7 +168,7 @@ class Bot {
LM::Inference::Params llm_get_params() const { LM::Inference::Params llm_get_params() const {
return { return {
.n_threads = int(config.threads), .n_threads = int(config.threads),
.n_ctx = 1012, .n_ctx = int(config.ctx_size),
.n_repeat_last = 256, .n_repeat_last = 256,
.temp = 0.3f, .temp = 0.3f,
.repeat_penalty = 1.372222224f, .repeat_penalty = 1.372222224f,
@ -344,7 +344,8 @@ public:
language = "EN", language = "EN",
inference_model = "13B-ggml-model-quant.bin", inference_model = "13B-ggml-model-quant.bin",
translation_model = "13B-ggml-model-quant.bin"; translation_model = "13B-ggml-model-quant.bin";
unsigned pool_size = 2, unsigned ctx_size = 1012,
pool_size = 2,
threads = 4, threads = 4,
persistance = true; persistance = true;
bool mlock = false; bool mlock = false;
@ -466,6 +467,8 @@ int main(int argc, char **argv) {
cfg.pool_size = std::stoi(value); cfg.pool_size = std::stoi(value);
} else if (key == "threads") { } else if (key == "threads") {
cfg.threads = std::stoi(value); cfg.threads = std::stoi(value);
} else if (key == "ctx_size") {
cfg.ctx_size = std::stoi(value);
} else if (key == "mlock") { } else if (key == "mlock") {
cfg.mlock = (value=="true")?true:false; cfg.mlock = (value=="true")?true:false;
} else if (key == "persistance") { } else if (key == "persistance") {