From 49a18bdd147d8f470c83ea7bf0eb8faa069dc7a7 Mon Sep 17 00:00:00 2001 From: anon Date: Thu, 1 Jun 2023 09:41:35 -0300 Subject: [PATCH 1/3] remove unused parameter warning --- examples/server/server.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/examples/server/server.cpp b/examples/server/server.cpp index d0d0c0b79..8f0b702d5 100644 --- a/examples/server/server.cpp +++ b/examples/server/server.cpp @@ -888,7 +888,7 @@ int main(int argc, char **argv) } }); - svr.Options(R"(/.*)", [&llama](const Request &req, Response &res) + svr.Options(R"(/.*)", [&llama](const Request &, Response &res) { return res.set_content("", "application/json"); }); From 6025476e3910079c31e4a4f0c7194fdf2913c8a5 Mon Sep 17 00:00:00 2001 From: anon Date: Thu, 1 Jun 2023 09:49:16 -0300 Subject: [PATCH 2/3] default penalize_nl back to true --- examples/server/server.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/examples/server/server.cpp b/examples/server/server.cpp index 8f0b702d5..b16a49b10 100644 --- a/examples/server/server.cpp +++ b/examples/server/server.cpp @@ -638,7 +638,7 @@ bool parse_options_completion(json body, llama_server_context& llama, Response & if (!body["penalize_nl"].is_null()) { llama.params.penalize_nl = body["penalize_nl"].get(); } else { - llama.params.penalize_nl = false; + llama.params.penalize_nl = default_params.penalize_nl; } if (!body["n_keep"].is_null()) { llama.params.n_keep = body["n_keep"].get(); From 8cbc4be6c24fbcab20bb3b0b050b1c5f6bfdb97f Mon Sep 17 00:00:00 2001 From: anon Date: Thu, 1 Jun 2023 09:49:50 -0300 Subject: [PATCH 3/3] clear logit_bias between requests + print --- examples/server/server.cpp | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/examples/server/server.cpp b/examples/server/server.cpp index b16a49b10..b345b53c2 100644 --- a/examples/server/server.cpp +++ b/examples/server/server.cpp @@ -559,6 +559,7 @@ json format_generation_settings(llama_server_context &llama) { { "n_keep", llama.params.n_keep }, { "ignore_eos", ignore_eos }, { "stream", llama.stream }, + { "logit_bias", llama.params.logit_bias }, }; } @@ -650,10 +651,10 @@ bool parse_options_completion(json body, llama_server_context& llama, Response & } else { llama.params.seed = time(NULL); } + + llama.params.logit_bias.clear(); if (!body["ignore_eos"].is_null() && body["ignore_eos"].get()) { llama.params.logit_bias[llama_token_eos()] = -INFINITY; - } else { - llama.params.logit_bias.erase(llama_token_eos()); } if (body["logit_bias"].is_array()) { int n_vocab = llama_n_vocab(llama.ctx); @@ -665,6 +666,7 @@ bool parse_options_completion(json body, llama_server_context& llama, Response & } } } + if (!body["prompt"].is_null()) { llama.params.prompt = body["prompt"].get(); } else { @@ -673,6 +675,7 @@ bool parse_options_completion(json body, llama_server_context& llama, Response & res.status = 400; return false; } + llama.params.antiprompt.clear(); if (!body["stop"].is_null()) { const auto stop = body["stop"].get>();