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

removed server changes

Browse source
This commit is contained in:
VJHack 2024-09-15 21:26:46 -05:00
parent f5a23928c7
commit 2736688af4
3 changed files with 3 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

2
common/arg.cpp
View file

@ -691,7 +691,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
[](gpt_params & params) { [](gpt_params & params) {
params.ctx_shift = false; params.ctx_shift = false;
} }
).set_examples({LLAMA_EXAMPLE_MAIN, LLAMA_EXAMPLE_SERVER})); ).set_examples({LLAMA_EXAMPLE_MAIN}));
add_opt(llama_arg( add_opt(llama_arg(
{"--chunks"}, "N", {"--chunks"}, "N",
format("max number of chunks to process (default: %d, -1 = all)", params.n_chunks), format("max number of chunks to process (default: %d, -1 = all)", params.n_chunks),

2
examples/main/README.md
View file

@ -161,7 +161,7 @@ A value of -1 will enable infinite text generation, even though we have a finite
If the pause is undesirable, a value of -2 will stop generation immediately when the context is filled. If the pause is undesirable, a value of -2 will stop generation immediately when the context is filled.
The `--no-context-shift` options allows you to stop the inifinite text generation once the finite context window is full. The `--no-context-shift` option allows you to stop the infinite text generation once the finite context window is full.
It is important to note that the generated text may be shorter than the specified number of tokens if an End-of-Sequence (EOS) token or a reverse prompt is encountered. In interactive mode, text generation will pause and control will be returned to the user. In non-interactive mode, the program will end. In both cases, the text generation may stop before reaching the specified `--predict` value. If you want the model to keep going without ever producing End-of-Sequence on its own, you can use the `--ignore-eos` parameter. It is important to note that the generated text may be shorter than the specified number of tokens if an End-of-Sequence (EOS) token or a reverse prompt is encountered. In interactive mode, text generation will pause and control will be returned to the user. In non-interactive mode, the program will end. In both cases, the text generation may stop before reaching the specified `--predict` value. If you want the model to keep going without ever producing End-of-Sequence on its own, you can use the `--ignore-eos` parameter.

9
examples/server/server.cpp
View file

@ -1815,13 +1815,6 @@ struct server_context {
for (server_slot & slot : slots) { for (server_slot & slot : slots) {
if (slot.ga_n == 1) { if (slot.ga_n == 1) {
if (slot.is_processing() && (int) system_tokens.size() + slot.n_past >= slot.n_ctx - 1) { if (slot.is_processing() && (int) system_tokens.size() + slot.n_past >= slot.n_ctx - 1) {
if (!params.ctx_shift){
slot.release();
slot.print_timings();
send_final_response(slot);
metrics.on_prediction(slot);
continue;
}
// Shift context // Shift context
const int n_keep = slot.params.n_keep + add_bos_token; const int n_keep = slot.params.n_keep + add_bos_token;
const int n_left = (int) system_tokens.size() + slot.n_past - n_keep; const int n_left = (int) system_tokens.size() + slot.n_past - n_keep;
@ -3175,4 +3168,4 @@ int main(int argc, char ** argv) {
t.join(); t.join();
return 0; return 0;
} }