ggml : sync latest repo (mostly refactoring changes)

Author: ggerganov

examples/common.cpp

@@ -39,6 +39,10 @@ bool gpt_params_parse(int argc, char ** argv, gpt_params & params) {
             params.top_p = std::stof(argv[++i]);
         } else if (arg == "--temp") {
             params.temp = std::stof(argv[++i]);
+        } else if (arg == "--repeat-last-n") {
+            params.repeat_last_n = std::stof(argv[++i]);
+        } else if (arg == "--repeat-penalty") {
+            params.repeat_penalty = std::stof(argv[++i]);
         } else if (arg == "-b" || arg == "--batch_size") {
             params.n_batch = std::stoi(argv[++i]);
         } else if (arg == "-m" || arg == "--model") {
@@ -90,6 +94,8 @@ void gpt_print_usage(int /*argc*/, char ** argv, const gpt_params & params) {
     fprintf(stderr, "  --top_k N             top-k sampling (default: %d)\n", params.top_k);
     fprintf(stderr, "  --top_p N             top-p sampling (default: %.1f)\n", params.top_p);
     fprintf(stderr, "  --temp N              temperature (default: %.1f)\n", params.temp);
+    fprintf(stderr, "  --repeat-last-n N     last n tokens to consider for penalize (default: %d, 0 = disabled)\n", params.repeat_last_n);
+    fprintf(stderr, "  --repeat-penalty N    penalize repeat sequence of tokens (default: %.2f, 1.0 = disabled)\n", (double)params.repeat_penalty);
     fprintf(stderr, "  -b N, --batch_size N  batch size for prompt processing (default: %d)\n", params.n_batch);
     fprintf(stderr, "  -m FNAME, --model FNAME\n");
     fprintf(stderr, "                        model path (default: %s)\n", params.model.c_str());

examples/common.h

@@ -23,6 +23,8 @@ struct gpt_params {
     int32_t top_k = 40;
     float   top_p = 0.9f;
     float   temp  = 0.9f;
+    int32_t repeat_last_n  = 64;
+    float   repeat_penalty = 1.00f;
 
     int32_t n_batch = 8; // batch size for prompt processing
 

examples/quantize/quantize.cpp

@@ -57,7 +57,7 @@ bool whisper_model_quantize(const std::string & fname_inp, const std::string & f
     {
         uint32_t magic;
         finp.read((char *) &magic, sizeof(magic));
-        if (magic != 0x67676d6c) {
+        if (magic != GGML_FILE_MAGIC) {
             fprintf(stderr, "%s: invalid model file '%s' (bad magic)\n", __func__, fname_inp.c_str());
             return false;
         }