npm - cui-llama.rn - Versions diffs - 1.1.4 → 1.1.6 - Mend

cui-llama.rn 1.1.4 → 1.1.6

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Files changed (20) hide show

package/android/src/main/CMakeLists.txt +1 -0
package/android/src/main/jni.cpp +3 -4
package/cpp/common.cpp +183 -1990
package/cpp/common.h +101 -130
package/cpp/ggml-impl.h +32 -0
package/cpp/ggml-metal.m +38 -28
package/cpp/ggml-quants.c +275 -84
package/cpp/ggml.c +89 -35
package/cpp/ggml.h +30 -67
package/cpp/llama-impl.h +1 -0
package/cpp/llama-sampling.cpp +218 -102
package/cpp/llama.cpp +599 -120
package/cpp/llama.h +33 -25
package/cpp/log.cpp +401 -0
package/cpp/log.h +85 -703
package/cpp/rn-llama.hpp +9 -11
package/cpp/sampling.cpp +12 -9
package/cpp/sampling.h +4 -56
package/cpp/sgemm.cpp +38 -0
package/package.json +1 -1

package/android/src/main/CMakeLists.txt CHANGED Viewed

@@ -12,6 +12,7 @@ set(
     ${RNLLAMA_LIB_DIR}/llama-grammar.cpp
     ${RNLLAMA_LIB_DIR}/llama-sampling.cpp
     ${RNLLAMA_LIB_DIR}/llama-vocab.cpp
+    ${RNLLAMA_LIB_DIR}/log.cpp
     ${RNLLAMA_LIB_DIR}/ggml-aarch64.c
     ${RNLLAMA_LIB_DIR}/ggml-alloc.c

package/android/src/main/jni.cpp CHANGED Viewed

@@ -523,7 +523,7 @@ Java_com_rnllama_LlamaContext_doCompletion(
         }
     }
-    llama_perf_print(llama->ctx, LLAMA_PERF_TYPE_CONTEXT);
+    llama_perf_context_print(llama->ctx);
     llama->is_predicting = false;
     auto result = createWriteableMap(env);
@@ -538,7 +538,7 @@ Java_com_rnllama_LlamaContext_doCompletion(
     putString(env, result, "stopping_word", llama->stopping_word.c_str());
     putInt(env, result, "tokens_cached", llama->n_past);
-    const auto timings_token = llama_get_token_timings(llama->ctx);
+    const auto timings_token = llama_perf_context(llama -> ctx);
     auto timingsResult = createWriteableMap(env);
     putInt(env, timingsResult, "prompt_n", timings_token.n_p_eval);
@@ -635,8 +635,7 @@ Java_com_rnllama_LlamaContext_embedding(
     llama->rewind();
-    // llama_reset_timings(llama->ctx);
-    llama_perf_reset(llama->ctx, LLAMA_PERF_TYPE_CONTEXT);
+    llama_perf_context_reset(llama->ctx);
     gpt_sampler_reset(llama->ctx_sampling);