move param parser to common

Author: ngxson

common/common.cpp

@@ -1576,6 +1576,7 @@ bool gpt_params_find_arg(int argc, char ** argv, const std::string & arg, gpt_pa
             return true;
         }
         params.out_file = argv[i];
+        params.cvector_outfile = argv[i];
         return true;
     }
     if (arg == "-ofreq" || arg == "--output-frequency") {
@@ -1610,6 +1611,55 @@ bool gpt_params_find_arg(int argc, char ** argv, const std::string & arg, gpt_pa
         params.i_chunk = std::stoi(argv[i]);
         return true;
     }
+    // control-vector-generator params
+    if (arg == "--completions-file") {
+        if (++i >= argc) {
+            invalid_param = true;
+            return true;
+        }
+        params.cvector_completions_file = argv[i];
+        return true;
+    }
+    if (arg == "--positive-file") {
+        if (++i >= argc) {
+            invalid_param = true;
+            return true;
+        }
+        params.cvector_positive_file = argv[i];
+        return true;
+    }
+    if (arg == "--negative-file") {
+        if (++i >= argc) {
+            invalid_param = true;
+            return true;
+        }
+        params.cvector_negative_file = argv[i];
+        return true;
+    }
+    if (arg == "--num-completions") {
+        if (++i >= argc) {
+            invalid_param = true;
+            return true;
+        }
+        params.n_completions = std::stoi(argv[i]);
+        return true;
+    }
+    if (arg == "--pca-batch") {
+        if (++i >= argc) {
+            invalid_param = true;
+            return true;
+        }
+        params.n_pca_batch = std::stoi(argv[i]);
+        return true;
+    }
+    if (arg == "--pca-iter") {
+        if (++i >= argc) {
+            invalid_param = true;
+            return true;
+        }
+        params.n_pca_iterations = std::stoi(argv[i]);
+        return true;
+    }
 #ifndef LOG_DISABLE_LOGS
     // Parse args for logging parameters
     if (log_param_single_parse(argv[i])) {
@@ -1931,6 +1981,15 @@ void gpt_params_print_usage(int /*argc*/, char ** argv, const gpt_params & param
     options.push_back({ "logging",     "       --log-append",           "Don't truncate the old log file." });
 #endif // LOG_DISABLE_LOGS
 
+    options.push_back({ "control-vector-generator" });
+    options.push_back({ "control-vector-generator", "-o,  --output FNAME",    "output file (default: '%s')", params.cvector_outfile.c_str() });
+    options.push_back({ "control-vector-generator", "--positive-file FNAME",  "positive prompts file, one prompt per line (default: '%s')", params.cvector_positive_file.c_str() });
+    options.push_back({ "control-vector-generator", "--negative-file FNAME",  "negative prompts file, one prompt per line (default: '%s')", params.cvector_negative_file.c_str() });
+    options.push_back({ "control-vector-generator", "--completions-file",     "completions file (default: '%s')", params.cvector_completions_file.c_str() });
+    options.push_back({ "control-vector-generator", "--num-completions N",    "number of lines of completions file to use (default: %d)", params.n_completions });
+    options.push_back({ "control-vector-generator", "--batch-pca N",          "batch size used for PCA. Larger batch runs faster, but uses more memory (default: %d)", params.n_pca_batch });
+    options.push_back({ "control-vector-generator", "--iter-pca N",           "number of iterations used for PCA (default: %d)", params.n_pca_iterations });
+
     printf("usage: %s [options]\n", argv[0]);
 
     for (const auto & o : options) {

common/common.h

@@ -232,6 +232,15 @@ struct gpt_params {
 
     bool process_output = false; // collect data for the output tensor
     bool compute_ppl    = true;  // whether to compute perplexity
+
+    // control-vector-generator params
+    int n_completions = 64;
+    int n_pca_batch = 20;
+    int n_pca_iterations = 1000;
+    std::string cvector_outfile          = "control_vector.gguf";
+    std::string cvector_completions_file = "examples/control-vector-generator/completions.txt";
+    std::string cvector_positive_file    = "examples/control-vector-generator/positive.txt";
+    std::string cvector_negative_file    = "examples/control-vector-generator/negative.txt";
 };
 
 void gpt_params_handle_model_default(gpt_params & params);

examples/control-vector-generator/README.md

@@ -5,7 +5,7 @@ This example demonstrates how to generate a control vector using gguf models.
 Related PRs:
 - [Add support for control vectors](https://github.com/ggerganov/llama.cpp/pull/5970)
 - (Issue) [Generate control vector using llama.cpp](https://github.com/ggerganov/llama.cpp/issues/6880)
-- [Add control-vector-generator](https://github.com/ggerganov/llama.cpp/pull/7514)
+- [Add control-vector-generator example](https://github.com/ggerganov/llama.cpp/pull/7514)
 
 Example:
 
@@ -14,13 +14,12 @@ Example:
 ./control-vector-generator -m ./dolphin-2.0-mistral-7b.Q4_K_M.gguf
 
 # With GPU
-./control-vector-generator --num-completions 2 --pca-iter 40 -m ./dolphin-2.0-mistral-7b.Q4_K_M.gguf -ngl 99
+./control-vector-generator -m ./dolphin-2.0-mistral-7b.Q4_K_M.gguf -ngl 99
 
 # With advanced options
-# Please note that the ORDER of arguments does matter
-# example-related options (i.e., --num-completions, --pca-iter) always come before model options (i.e., -m, -ngl)
-./control-vector-generator --num-completions 128 --pca-iter 2000 --batch-pca 100 -m ./dolphin-2.0-mistral-7b.Q4_K_M.gguf -ngl 99
+./control-vector-generator -m ./dolphin-2.0-mistral-7b.Q4_K_M.gguf -ngl 99 --num-completions 128 --pca-iter 2000 --batch-pca 100
 
 # To see help message
 ./control-vector-generator -h
+# Then, have a look at "control-vector-generator" section
 ```