root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-26 03:14:35 +00:00
Code Issues Actions 7 Packages Projects Releases Wiki Activity

server: Fix has_next_line in JSON response (#10818)
Some checks are pending
flake8 Lint / Lint (push) Waiting to run
Details
Python Type-Check / pyright type-check (push) Waiting to run
Details

Browse Source 
* Update server JSON response.

* Add unit test to check `has_new_line` JSON response

* Remove `has_new_line` unit test changes.

* Address code review comment: type check for `has_new_line` in unit test
This commit is contained in:
Michelle Tan 2024-12-14 22:29:45 +00:00 committed by GitHub
parent e52aba537a
commit 89d604f2c8
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 4 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
examples/server/server.cpp
View File

@ -459,7 +459,7 @@ struct server_task_result_cmpl_final : server_task_result {
int32_t n_decoded; int32_t n_decoded;
int32_t n_prompt_tokens; int32_t n_prompt_tokens;
int32_t n_tokens_cached; int32_t n_tokens_cached;
int32_t has_new_line; bool has_new_line;
std::string stopping_word; std::string stopping_word;
stop_type stop = STOP_TYPE_NONE; stop_type stop = STOP_TYPE_NONE;

2
examples/server/tests/unit/test_completion.py
View File

@ -25,6 +25,7 @@ def test_completion(prompt: str, n_predict: int, re_content: str, n_prompt: int,
assert res.body["timings"]["prompt_n"] == n_prompt assert res.body["timings"]["prompt_n"] == n_prompt
assert res.body["timings"]["predicted_n"] == n_predicted assert res.body["timings"]["predicted_n"] == n_predicted
assert res.body["truncated"] == truncated assert res.body["truncated"] == truncated
assert type(res.body["has_new_line"]) == bool
assert match_regex(re_content, res.body["content"]) assert match_regex(re_content, res.body["content"])
@ -48,6 +49,7 @@ def test_completion_stream(prompt: str, n_predict: int, re_content: str, n_promp
assert data["timings"]["predicted_n"] == n_predicted assert data["timings"]["predicted_n"] == n_predicted
assert data["truncated"] == truncated assert data["truncated"] == truncated
assert data["stop_type"] == "limit" assert data["stop_type"] == "limit"
assert type(data["has_new_line"]) == bool
assert "generation_settings" in data assert "generation_settings" in data
assert server.n_predict is not None assert server.n_predict is not None
assert data["generation_settings"]["n_predict"] == min(n_predict, server.n_predict) assert data["generation_settings"]["n_predict"] == min(n_predict, server.n_predict)

2
examples/server/utils.hpp
View File

@ -22,7 +22,7 @@
#include <vector> #include <vector>
#include <memory> #include <memory>
#define DEFAULT_OAICOMPAT_MODEL "gpt-3.5-turbo-0613" #define DEFAULT_OAICOMPAT_MODEL "gpt-3.5-turbo"
using json = nlohmann::ordered_json; using json = nlohmann::ordered_json;