Skip to content

server: disable context shift #9544

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 5 commits into from
Closed

server: disable context shift #9544

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
5688864
allow disable context shift for sever
VJHack Sep 19, 2024
2f2e4b3
Fixed error message to say 'enable context shift'
VJHack Sep 20, 2024
0cabcbe
fixed server 200 null response when context is exceeded
VJHack Sep 20, 2024
9880e3a
changed error message wording
VJHack Sep 20, 2024
4af076b
updated context shift error to ERROR_TYPE_INVALID_REQUEST
VJHack Sep 21, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion common/arg.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -691,7 +691,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
[](gpt_params & params) {
params.ctx_shift = false;
}
).set_examples({LLAMA_EXAMPLE_MAIN}));
).set_examples({LLAMA_EXAMPLE_MAIN, LLAMA_EXAMPLE_SERVER}));
add_opt(llama_arg(
{"--chunks"}, "N",
format("max number of chunks to process (default: %d, -1 = all)", params.n_chunks),
Expand Down
1 change: 1 addition & 0 deletions examples/server/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,7 @@ The project is under active development, and we are [looking for feedback and co
| `--poll-batch <0\|1>` | use polling to wait for work (default: same as --poll) |
| `-c, --ctx-size N` | size of the prompt context (default: 0, 0 = loaded from model)<br/>(env: LLAMA_ARG_CTX_SIZE) |
| `-n, --predict, --n-predict N` | number of tokens to predict (default: -1, -1 = infinity, -2 = until context filled)<br/>(env: LLAMA_ARG_N_PREDICT) |
| `--no-context-shift` | stop generation when context window is full (default: context shift is enabled)<br/> |
| `-b, --batch-size N` | logical maximum batch size (default: 2048)<br/>(env: LLAMA_ARG_BATCH) |
| `-ub, --ubatch-size N` | physical maximum batch size (default: 512)<br/>(env: LLAMA_ARG_UBATCH) |
| `--keep N` | number of tokens to keep from the initial prompt (default: 0, -1 = all) |
Expand Down
13 changes: 13 additions & 0 deletions examples/server/server.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1815,6 +1815,13 @@ struct server_context {
for (server_slot & slot : slots) {
if (slot.ga_n == 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
const int n_keep = slot.params.n_keep + add_bos_token;
const int n_left = (int) system_tokens.size() + slot.n_past - n_keep;
Expand Down Expand Up @@ -1940,6 +1947,12 @@ struct server_context {
send_final_response(slot);
continue;
}
// context shift is disabled and prompt is too large - discard it
if (!params.ctx_shift && slot.n_prompt_tokens > slot.n_ctx ){
slot.release();
send_error(slot, "Input is too large to process. Either disable context shift or increase context length. ", ERROR_TYPE_SERVER);
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Shouldn't the error say enable context shift, since it's already disabled?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, I made the change. Thanks for the correction.

continue;
}

if (slot.cmpl_type == SERVER_TASK_CMPL_TYPE_EMBEDDING) {
// this prompt is too large to process - discard it
Expand Down
Loading