common : refactor cli arg parsing (#7675)

* common : gpt_params_parse do not print usage

* common : rework usage print (wip)

* common : valign

* common : rework print_usage

* infill : remove cfg support

* common : reorder args

* server : deduplicate parameters

ggml-ci

* common : add missing header

ggml-ci

* common : remote --random-prompt usages

ggml-ci

* examples : migrate to gpt_params

ggml-ci

* batched-bench : migrate to gpt_params

* retrieval : migrate to gpt_params

* common : change defaults for escape and n_ctx

* common : remove chatml and instruct params

ggml-ci

* common : passkey use gpt_params

1 files changed, 8 insertions, 61 deletions

diff --git a/examples/main/main.cpp b/examples/main/main.cpp
index 44949ba8..b97b7b79 100644
--- a/examples/main/main.cpp
+++ b/examples/main/main.cpp
@@ -122,8 +122,10 @@ int main(int argc, char ** argv) {
     g_params = &params;
 
     if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
+
     llama_sampling_params & sparams = params.sparams;
 
 #ifndef LOG_DISABLE_LOGS
@@ -180,9 +182,6 @@ int main(int argc, char ** argv) {
     LOG_TEE("%s: seed  = %u\n", __func__, params.seed);
 
     std::mt19937 rng(params.seed);
-    if (params.random_prompt) {
-        params.prompt = string_random_prompt(rng);
-    }
 
     LOG("%s: llama backend init\n", __func__);
     llama_backend_init();
@@ -250,11 +249,8 @@ int main(int argc, char ** argv) {
 
     std::vector<llama_token> embd_inp;
 
-    if (params.interactive_first || params.instruct || params.chatml || !params.prompt.empty() || session_tokens.empty()) {
+    if (params.interactive_first || !params.prompt.empty() || session_tokens.empty()) {
         LOG("tokenize the prompt\n");
-        if (params.chatml) {
-            params.prompt = "<|im_start|>system\n" + params.prompt + "<|im_end|>";
-        }
         embd_inp = ::llama_tokenize(ctx, params.prompt, true, true);
     } else {
         LOG("use session tokens\n");
@@ -332,37 +328,13 @@ int main(int argc, char ** argv) {
     }
 
     // number of tokens to keep when resetting context
-    if (params.n_keep < 0 || params.n_keep > (int) embd_inp.size() || params.instruct || params.chatml) {
+    if (params.n_keep < 0 || params.n_keep > (int) embd_inp.size()) {
         params.n_keep = (int)embd_inp.size();
     } else {
         params.n_keep += add_bos; // always keep the BOS token
     }
 
-    // prefix & suffix for instruct mode
-    const auto inp_pfx = ::llama_tokenize(ctx, "\n\n### Instruction:\n\n", true,  true);
-    const auto inp_sfx = ::llama_tokenize(ctx, "\n\n### Response:\n\n",    false, true);
-
-    LOG("inp_pfx: %s\n", LOG_TOKENS_TOSTR_PRETTY(ctx, inp_pfx).c_str());
-    LOG("inp_sfx: %s\n", LOG_TOKENS_TOSTR_PRETTY(ctx, inp_sfx).c_str());
-
-    // chatml prefix & suffix
-    const auto cml_pfx = ::llama_tokenize(ctx, "\n<|im_start|>user\n", true, true);
-    const auto cml_sfx = ::llama_tokenize(ctx, "<|im_end|>\n<|im_start|>assistant\n", false, true);
-
-    LOG("cml_pfx: %s\n", LOG_TOKENS_TOSTR_PRETTY(ctx, cml_pfx).c_str());
-    LOG("cml_sfx: %s\n", LOG_TOKENS_TOSTR_PRETTY(ctx, cml_sfx).c_str());
-
-    // in instruct mode, we inject a prefix and a suffix to each input by the user
-    if (params.instruct) {
-        params.interactive_first = true;
-        params.antiprompt.emplace_back("### Instruction:\n\n");
-    }
-    // similar for chatml mode
-    else if (params.chatml) {
-        params.interactive_first = true;
-        params.antiprompt.emplace_back("<|im_start|>user\n");
-    }
-    else if (params.conversation) {
+    if (params.conversation) {
         params.interactive_first = true;
     }
 
@@ -823,15 +795,13 @@ int main(int argc, char ** argv) {
 
                     is_interacting = true;
                     printf("\n");
-                } else if (params.instruct || params.chatml) {
-                    is_interacting = true;
                 }
             }
 
             if (n_past > 0 && is_interacting) {
                 LOG("waiting for user input\n");
 
-                if (params.conversation || params.instruct || params.chatml) {
+                if (params.conversation) {
                     printf("\n> ");
                 }
 
@@ -874,24 +844,12 @@ int main(int argc, char ** argv) {
 
                     const size_t original_size = embd_inp.size();
 
-                    // instruct mode: insert instruction prefix
-                    if (params.instruct && !is_antiprompt) {
-                        LOG("inserting instruction prefix\n");
-                        n_consumed = embd_inp.size();
-                        embd_inp.insert(embd_inp.end(), inp_pfx.begin(), inp_pfx.end());
-                    }
-                    // chatml mode: insert user chat prefix
-                    if (params.chatml && !is_antiprompt) {
-                        LOG("inserting chatml prefix\n");
-                        n_consumed = embd_inp.size();
-                        embd_inp.insert(embd_inp.end(), cml_pfx.begin(), cml_pfx.end());
-                    }
                     if (params.escape) {
                         string_process_escapes(buffer);
                     }
 
                     const auto line_pfx = ::llama_tokenize(ctx, params.input_prefix, false, true);
-                    const auto line_inp = ::llama_tokenize(ctx, buffer,              false, params.interactive_specials);
+                    const auto line_inp = ::llama_tokenize(ctx, buffer,              false, false);
                     const auto line_sfx = ::llama_tokenize(ctx, params.input_suffix, false, true);
 
                     LOG("input tokens: %s\n", LOG_TOKENS_TOSTR_PRETTY(ctx, line_inp).c_str());
@@ -900,17 +858,6 @@ int main(int argc, char ** argv) {
                     embd_inp.insert(embd_inp.end(), line_inp.begin(), line_inp.end());
                     embd_inp.insert(embd_inp.end(), line_sfx.begin(), line_sfx.end());
 
-                    // instruct mode: insert response suffix
-                    if (params.instruct) {
-                        LOG("inserting instruction suffix\n");
-                        embd_inp.insert(embd_inp.end(), inp_sfx.begin(), inp_sfx.end());
-                    }
-                    // chatml mode: insert assistant chat suffix
-                    if (params.chatml) {
-                        LOG("inserting chatml suffix\n");
-                        embd_inp.insert(embd_inp.end(), cml_sfx.begin(), cml_sfx.end());
-                    }
-
                     for (size_t i = original_size; i < embd_inp.size(); ++i) {
                         const llama_token token = embd_inp[i];
                         output_tokens.push_back(token);
@@ -935,7 +882,7 @@ int main(int argc, char ** argv) {
         }
 
         // end of generation
-        if (!embd.empty() && llama_token_is_eog(model, embd.back()) && !(params.instruct || params.interactive || params.chatml)) {
+        if (!embd.empty() && llama_token_is_eog(model, embd.back()) && !(params.interactive)) {
             LOG_TEE(" [end of text]\n");
             break;
         }