Adapt ggml-org/llama.cpp#3560

Sources/llmfarm_core_cpp/ggml/ggml.c

@@ -13537,7 +13537,7 @@ static void ggml_compute_forward_rope_f16(
                         dst_data[n_dims]     = GGML_FP32_TO_FP16(x2*cos_block_theta - x3*sin_block_theta);
                         dst_data[n_dims/2*3] = GGML_FP32_TO_FP16(x2*sin_block_theta + x3*cos_block_theta);
                     }
-                } if (!is_neox) {
+                } else if (!is_neox) {
                     for (int64_t i0 = 0; i0 < ne0; i0 += 2) {
                         const float cos_theta = cosf(theta);
                         const float sin_theta = sinf(theta);
@@ -19170,6 +19170,7 @@ void ggml_graph_export(const struct ggml_cgraph * cgraph, const char * fname) {
 
                             if (idx == -1) {
                                 fprintf(stderr, "%s: failed to find tensor, arg = %d, node = %d\n", __func__, j, i);
+                                fclose(fout);
                                 return;
                             }
 

Sources/llmfarm_core_cpp/ggml/train.cpp

@@ -1376,11 +1376,11 @@ void train_opt_callback(void * vdata, int accum_step, float * sched, bool * canc
     struct ggml_opt_context        * opt    = train->opt;
     int n_batch = params->n_batch;
     int n_ctx = params->n_ctx;
-            
+
     bool swift_callback_res = false;
-    
+
 //    swift_callback_res = data->swiftcallback("Begin");
-    
+
     if (accum_step == 0) {
         // time measurement
         int64_t now = ggml_time_ms();
@@ -1392,9 +1392,9 @@ void train_opt_callback(void * vdata, int accum_step, float * sched, bool * canc
                 const double gain = 0.7;
                 data->millis_per_iter = data->millis_per_iter*(1.0-gain) + dt*gain;
             }
-        }        
+        }
         double remaining_millis = 0.0;
-        
+
         if (data->millis_per_iter > 0.0) {
             const int n_iter = params->adam_n_iter;
             const int done_iter = opt->iter - data->first_iter;
@@ -1431,11 +1431,11 @@ void train_opt_callback(void * vdata, int accum_step, float * sched, bool * canc
 
         int impr_plot = -(int)(1 + (opt->loss_before - opt->loss_after) * 10.0f + 0.5f);
         if (impr_plot > 0) impr_plot = 0;
-        if (std::isnan(opt->loss_before) || std::isnan(opt->loss_before)) impr_plot = 0;
+        if (std::isnan(opt->loss_before) || std::isnan(opt->loss_after)) impr_plot = 0;
         printf("%s: iter=%6d sample=%zu/%zu sched=%f loss=%f",
             __func__, opt->iter, std::min(1+train->shuffle_next_sample, train->shuffle_sample_count), train->shuffle_sample_count,
             *sched, opt->loss_after);
-        
+
         char descr[500]; \
         sprintf(descr, "%s: iter=%6d sample=%zu/%zu sched=%f loss=%f",
                 __func__, opt->iter, std::min(1+train->shuffle_next_sample, train->shuffle_sample_count), train->shuffle_sample_count,
@@ -1444,8 +1444,8 @@ void train_opt_callback(void * vdata, int accum_step, float * sched, bool * canc
         if (swift_callback_res == true){
             *cancel = true;
         }
-        
-        
+
+
 
         if (data->millis_per_iter > 0) {
             printf(" dt=");

Sources/llmfarm_core_cpp/grammar-parser.cpp

@@ -399,7 +399,7 @@ namespace grammar_parser {
     void print_grammar(FILE * file, const parse_state & state) {
         try {
             std::map<uint32_t, std::string> symbol_id_names;
-            for (auto kv : state.symbol_ids) {
+            for (const auto & kv : state.symbol_ids) {
                 symbol_id_names[kv.second] = kv.first;
             }
             for (size_t i = 0, end = state.rules.size(); i < end; i++) {

Sources/llmfarm_core_cpp/llama/llama.cpp

@@ -6587,7 +6587,6 @@ struct llm_tokenizer_bpe {
                 llm_symbol sym;
                 size_t char_len = std::min(word.size() - offset, (size_t) ::utf8_len(word[offset]));
                 sym.text = word.c_str() + offset;
-                sym.n = 1;
                 sym.n = char_len;
                 offset += sym.n;
                 sym.prev = index - 1;
@@ -7317,7 +7316,7 @@ static std::vector<llama_grammar_candidate> llama_grammar_reject_candidates_for_
     std::vector<llama_grammar_candidate> rejects;
 
     if (stack.empty()) {
-        for (auto tok : candidates) {
+        for (auto const & tok : candidates) {
             if (*tok.code_points != 0 || tok.partial_utf8.n_remain != 0) {
                 rejects.push_back(tok);
             }
@@ -7328,7 +7327,7 @@ static std::vector<llama_grammar_candidate> llama_grammar_reject_candidates_for_
     const llama_grammar_element * stack_pos = stack.back();
 
     std::vector<llama_grammar_candidate> next_candidates;
-    for (auto tok : candidates) {
+    for (auto const & tok : candidates) {
         if (*tok.code_points == 0) {
             // reached end of full codepoints in token, reject iff it ended in a partial sequence
             // that cannot satisfy this position in grammar
@@ -7354,7 +7353,7 @@ static std::vector<llama_grammar_candidate> llama_grammar_reject_candidates_for_
     llama_grammar_advance_stack(rules, stack_after, next_stacks);
 
     auto next_rejects = llama_grammar_reject_candidates(rules, next_stacks, next_candidates);
-    for (auto tok : next_rejects) {
+    for (auto const & tok : next_rejects) {
         rejects.push_back({ tok.index, tok.code_points - 1, tok.partial_utf8 });
     }