examples : dedup simple

ggerganov · Aug 16, 2023 · 795ec70 · 795ec70
1 parent c290f3e
commit 795ec70
Show file tree

Hide file tree

Showing 3 changed files with 12 additions and 140 deletions.
diff --git a/Makefile b/Makefile
@@ -1,5 +1,5 @@
 # Define the default target now so that it is always the first target
-BUILD_TARGETS = main quantize quantize-stats perplexity embedding vdot train-text-from-scratch convert-llama2c-to-ggml simple server embd-input-test gguf gguf-llama-simple gptneox-main
+BUILD_TARGETS = main quantize quantize-stats perplexity embedding vdot train-text-from-scratch convert-llama2c-to-ggml simple server embd-input-test gguf gptneox-main
 
 # Binaries only useful for tests
 TEST_TARGETS = tests/test-grammar-parser tests/test-double-float tests/test-grad0 tests/test-opt tests/test-quantize-fns tests/test-quantize-perf tests/test-sampling tests/test-tokenizer-0
@@ -388,9 +388,6 @@ embd-input-test: $(LIB_PRE)embdinput$(DSO_EXT) examples/embd-input/embd-input-te
 gguf: examples/gguf/gguf.cpp                                  build-info.h ggml.o llama.o $(OBJS)
 	$(CXX) $(CXXFLAGS) $(filter-out %.h,$^) -o $@ $(LDFLAGS)
 
-gguf-llama-simple: examples/gguf/gguf-llama-simple.cpp        build-info.h ggml.o llama.o common.o $(OBJS)
-	$(CXX) $(CXXFLAGS) $(filter-out %.h,$^) -o $@ $(LDFLAGS)
-
 gptneox-main: gptneox-main.cpp ggml.o $(OBJS)
 	$(CXX) $(CXXFLAGS) $(filter-out %.h,$^) -o $@ $(LDFLAGS)
 

diff --git a/examples/gguf/gguf-llama-simple.cpp b/examples/gguf/gguf-llama-simple.cpp
diff --git a/examples/simple/simple.cpp b/examples/simple/simple.cpp
@@ -36,16 +36,17 @@ int main(int argc, char ** argv) {
 
     llama_backend_init(params.numa);
 
-    llama_model * model;
-    llama_context * ctx;
+    llama_context_params ctx_params = llama_context_default_params();
 
-    std::tie(model, ctx) = llama_init_from_gpt_params(params);
+    llama_model * model = llama_load_model_from_file(params.model.c_str(), ctx_params);
 
     if (model == NULL) {
-        fprintf(stderr, "%s: error: unable to load model\n", __func__);
+        fprintf(stderr , "%s: error: unable to load model\n" , __func__);
         return 1;
     }
 
+    llama_context * ctx = llama_new_context_with_model(model, ctx_params);
+
     // tokenize the prompt
 
     std::vector<llama_token> tokens_list;
@@ -54,7 +55,7 @@ int main(int argc, char ** argv) {
     const int max_context_size     = llama_n_ctx(ctx);
     const int max_tokens_list_size = max_context_size - 4;
 
-    if ((int)tokens_list.size() > max_tokens_list_size) {
+    if ((int) tokens_list.size() > max_tokens_list_size) {
         fprintf(stderr, "%s: error: prompt too long (%d tokens, max %d)\n", __func__, (int) tokens_list.size(), max_tokens_list_size);
         return 1;
     }
@@ -74,7 +75,9 @@ int main(int argc, char ** argv) {
     // tokens (see "infinite text generation via context swapping" in the main example), but in this minimalist
     // example, we will just stop the loop once this cache is full or once an end of stream is detected.
 
-    while (llama_get_kv_cache_token_count( ctx ) < max_context_size) {
+    const int n_gen = std::min(32, max_context_size);
+
+    while (llama_get_kv_cache_token_count(ctx) < n_gen) {
         // evaluate the transformer
 
         if (llama_eval(ctx, tokens_list.data(), int(tokens_list.size()), llama_get_kv_cache_token_count(ctx), params.n_threads)) {
@@ -114,13 +117,14 @@ int main(int argc, char ** argv) {
 
         // push this new token for next evaluation
         tokens_list.push_back(new_token_id);
-
     }
 
     llama_free(ctx);
     llama_free_model(model);
 
     llama_backend_free();
 
+    fprintf(stderr, "\n\n");
+
     return 0;
 }