Merge commit 'df270ef74596da8f1178f08991f4c51f18c9ee82' into concedo_experimental

# Conflicts:
#	Makefile
#	common/CMakeLists.txt
#	common/common.h
#	common/sampling.cpp
#	common/sampling.h
#	examples/infill/infill.cpp
#	examples/llama-bench/llama-bench.cpp
#	examples/quantize-stats/quantize-stats.cpp
#	examples/server/server.cpp
#	include/llama.h
#	src/llama-sampling.cpp
#	src/llama-sampling.h
#	src/llama.cpp
#	tests/test-grammar-integration.cpp
#	tests/test-grammar-parser.cpp
#	tests/test-json-schema-to-grammar.cpp
#	tests/test-llama-grammar.cpp
#	tests/test-sampling.cpp

examples/passkey/README.md

@@ -1,15 +0,0 @@
-# llama.cpp/example/passkey
-
-A passkey retrieval task is an evaluation method used to measure a language
-models ability to recall information from long contexts.
-
-See the following PRs for more info:
-
-- https://github.com/ggerganov/llama.cpp/pull/3856
-- https://github.com/ggerganov/llama.cpp/pull/4810
-
-### Usage
-
-```bash
-make -j && ./llama-passkey -m ./models/llama-7b-v2/ggml-model-f16.gguf --junk 250
-```

examples/passkey/passkey.cpp

@@ -26,8 +26,6 @@ int main(int argc, char ** argv) {
         return 1;
     }
 
-    srand(params.seed == LLAMA_DEFAULT_SEED ? time(NULL) : params.seed);
-
     int n_junk = params.n_junk;
     int n_keep = params.n_keep;
     int n_grp  = params.grp_attn_n;
@@ -80,12 +78,17 @@ int main(int argc, char ** argv) {
     GGML_ASSERT(ctx_params.n_batch % n_grp == 0 && "n_batch must be divisible by n_grp");
 
     llama_context * ctx = llama_new_context_with_model(model, ctx_params);
-
     if (ctx == NULL) {
         fprintf(stderr , "%s: error: failed to create the llama_context\n" , __func__);
         return 1;
     }
 
+    auto sparams = llama_sampler_chain_default_params();
+
+    llama_sampler * smpl = llama_sampler_chain_init(sparams);
+
+    llama_sampler_chain_add(smpl, llama_sampler_init_greedy());
+
     // tokenize the prompt
     std::vector<llama_token> tokens_list;
     tokens_list = ::llama_tokenize(ctx, params.prompt, true);
@@ -217,20 +220,9 @@ int main(int argc, char ** argv) {
     while (n_cur <= n_len) {
         // sample the next token
         {
-            auto   n_vocab = llama_n_vocab(model);
-            auto * logits  = llama_get_logits_ith(ctx, batch.n_tokens - 1);
+            const llama_token new_token_id = llama_sampler_sample(smpl, ctx, batch.n_tokens - 1);
 
-            std::vector<llama_token_data> candidates;
-            candidates.reserve(n_vocab);
-
-            for (llama_token token_id = 0; token_id < n_vocab; token_id++) {
-                candidates.emplace_back(llama_token_data{ token_id, logits[token_id], 0.0f });
-            }
-
-            llama_token_data_array candidates_p = { candidates.data(), candidates.size(), false };
-
-            // sample the most likely token
-            const llama_token new_token_id = llama_sample_token_greedy(ctx, &candidates_p);
+            llama_sampler_accept(smpl, new_token_id);
 
             // is it an end of generation?
             if (llama_token_is_eog(model, new_token_id) || n_cur == n_len) {
@@ -267,10 +259,13 @@ int main(int argc, char ** argv) {
     LOG_TEE("%s: decoded %d tokens in %.2f s, speed: %.2f t/s\n",
             __func__, n_decode, (t_main_end - t_main_start) / 1000000.0f, n_decode / ((t_main_end - t_main_start) / 1000000.0f));
 
-    llama_print_timings(ctx);
+    LOG_TEE("\n");
+    llama_perf_print(ctx, LLAMA_PERF_TYPE_CONTEXT);
 
     fprintf(stderr, "\n");
 
+    llama_sampler_free(smpl);
+
     llama_batch_free(batch);
 
     llama_free(ctx);