npm - @fugood/llama.node - Versions diffs - 1.4.11 → 1.4.13 - Mend

@fugood/llama.node 1.4.11 → 1.4.13

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 (69) hide show

package/package.json +15 -15
package/scripts/llama.cpp.patch +31 -31
package/src/llama.cpp/common/arg.cpp +128 -59
package/src/llama.cpp/common/arg.h +1 -0
package/src/llama.cpp/common/chat-parser.cpp +11 -0
package/src/llama.cpp/common/chat.cpp +36 -7
package/src/llama.cpp/common/chat.h +1 -0
package/src/llama.cpp/common/common.cpp +42 -23
package/src/llama.cpp/common/common.h +11 -1
package/src/llama.cpp/common/llguidance.cpp +10 -6
package/src/llama.cpp/common/regex-partial.cpp +13 -13
package/src/llama.cpp/common/sampling.cpp +58 -14
package/src/llama.cpp/common/sampling.h +3 -1
package/src/llama.cpp/ggml/CMakeLists.txt +13 -1
package/src/llama.cpp/ggml/include/ggml-backend.h +1 -1
package/src/llama.cpp/ggml/src/CMakeLists.txt +23 -9
package/src/llama.cpp/ggml/src/ggml-cpu/CMakeLists.txt +12 -2
package/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-impl.h +1 -1
package/src/llama.cpp/ggml/src/ggml-cpu/kleidiai/kernels.cpp +86 -25
package/src/llama.cpp/ggml/src/ggml-cpu/kleidiai/kleidiai.cpp +15 -8
package/src/llama.cpp/ggml/src/ggml-cpu/llamafile/sgemm.cpp +768 -0
package/src/llama.cpp/ggml/src/ggml-cpu/simd-mappings.h +0 -4
package/src/llama.cpp/include/llama.h +100 -12
package/src/llama.cpp/src/CMakeLists.txt +4 -0
package/src/llama.cpp/src/llama-adapter.cpp +12 -3
package/src/llama.cpp/src/llama-adapter.h +7 -1
package/src/llama.cpp/src/llama-arch.cpp +78 -0
package/src/llama.cpp/src/llama-arch.h +8 -0
package/src/llama.cpp/src/llama-chat.cpp +11 -0
package/src/llama.cpp/src/llama-chat.h +1 -0
package/src/llama.cpp/src/llama-context.cpp +637 -49
package/src/llama.cpp/src/llama-context.h +43 -1
package/src/llama.cpp/src/llama-grammar.cpp +40 -13
package/src/llama.cpp/src/llama-grammar.h +2 -0
package/src/llama.cpp/src/llama-graph.cpp +173 -5
package/src/llama.cpp/src/llama-graph.h +71 -6
package/src/llama.cpp/src/llama-hparams.cpp +4 -0
package/src/llama.cpp/src/llama-hparams.h +12 -5
package/src/llama.cpp/src/llama-kv-cache.h +1 -1
package/src/llama.cpp/src/llama-mmap.cpp +11 -4
package/src/llama.cpp/src/llama-model-loader.cpp +23 -0
package/src/llama.cpp/src/llama-model-loader.h +2 -0
package/src/llama.cpp/src/llama-model-saver.cpp +3 -0
package/src/llama.cpp/src/llama-model.cpp +337 -26
package/src/llama.cpp/src/llama-model.h +13 -2
package/src/llama.cpp/src/llama-sampling.cpp +1259 -186
package/src/llama.cpp/src/llama-sampling.h +19 -7
package/src/llama.cpp/src/llama-vocab.cpp +101 -33
package/src/llama.cpp/src/llama-vocab.h +2 -0
package/src/llama.cpp/src/llama.cpp +87 -64
package/src/llama.cpp/src/models/afmoe.cpp +9 -5
package/src/llama.cpp/src/models/bert.cpp +4 -2
package/src/llama.cpp/src/models/cogvlm.cpp +5 -3
package/src/llama.cpp/src/models/cohere2-iswa.cpp +3 -0
package/src/llama.cpp/src/models/deepseek2.cpp +1 -1
package/src/llama.cpp/src/models/gemma-embedding.cpp +2 -6
package/src/llama.cpp/src/models/gemma2-iswa.cpp +5 -2
package/src/llama.cpp/src/models/gemma3.cpp +3 -4
package/src/llama.cpp/src/models/gemma3n-iswa.cpp +4 -7
package/src/llama.cpp/src/models/llama-iswa.cpp +6 -2
package/src/llama.cpp/src/models/llama.cpp +19 -6
package/src/llama.cpp/src/models/maincoder.cpp +117 -0
package/src/llama.cpp/src/models/mimo2-iswa.cpp +123 -0
package/src/llama.cpp/src/models/models.h +18 -0
package/src/llama.cpp/src/models/modern-bert.cpp +116 -0
package/src/llama.cpp/src/models/openai-moe-iswa.cpp +5 -2
package/src/llama.cpp/src/models/plamo3.cpp +128 -0
package/src/llama.cpp/src/models/smallthinker.cpp +11 -5
package/src/llama.cpp/src/unicode.cpp +23 -14

package/src/llama.cpp/src/llama-model-loader.h CHANGED Viewed

@@ -131,6 +131,8 @@ struct llama_model_loader {
     template<typename T>
     bool get_key_or_arr(enum llm_kv kid, T & result, uint32_t n, bool required = true);
+    bool get_key_or_arr(enum llm_kv kid, uint32_t & result, bool required = true);
     std::string get_arch_name() const;
     enum llm_arch get_arch() const;

package/src/llama.cpp/src/llama-model-saver.cpp CHANGED Viewed

@@ -146,6 +146,9 @@ void llama_model_saver::add_kv_from_model() {
     add_kv(LLM_KV_VOCAB_SIZE,                        vocab.n_tokens());
     add_kv(LLM_KV_CONTEXT_LENGTH,                    hparams.n_ctx_train);
     add_kv(LLM_KV_EMBEDDING_LENGTH,                  hparams.n_embd);
+    if (hparams.n_embd_out > 0) {
+        add_kv(LLM_KV_EMBEDDING_LENGTH_OUT,          hparams.n_embd_out);
+    }
     add_kv(LLM_KV_BLOCK_COUNT,                       hparams.n_layer);
     add_kv(LLM_KV_LEADING_DENSE_BLOCK_COUNT,         hparams.n_layer_dense_lead);
     add_kv(LLM_KV_FEED_FORWARD_LENGTH,               hparams.n_ff_arr, true);