llama-cpp-capacitor 0.0.9 → 0.0.10

package/android/src/main/jni.cpp

@@ -154,13 +154,82 @@ Java_ai_annadata_plugin_capacitor_LlamaCpp_initContextNative(
         cparams.use_mmap = true;
         cparams.use_mlock = false;
         cparams.numa = LM_GGML_NUMA_STRATEGY_DISABLED;
+        cparams.ctx_shift = false;
+        cparams.chat_template = "";
+        cparams.embedding = false;
+        cparams.cont_batching = false;
+        cparams.parallel = false;
+        cparams.grammar = "";
+        cparams.grammar_penalty.clear();
+        cparams.antiprompt.clear();
+        cparams.lora_adapter.clear();
+        cparams.lora_base = "";
+        cparams.mul_mat_q = true;
+        cparams.f16_kv = true;
+        cparams.logits_all = false;
+        cparams.vocab_only = false;
+        cparams.rope_scaling_type = LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED;
+        cparams.rope_scaling_factor = 0.0f;
+        cparams.rope_scaling_orig_ctx_len = 0;
+        cparams.yarn_ext_factor = -1.0f;
+        cparams.yarn_attn_factor = 1.0f;
+        cparams.yarn_beta_fast = 32.0f;
+        cparams.yarn_beta_slow = 1.0f;
+        cparams.yarn_orig_ctx = 0;
+        cparams.offload_kqv = true;
+        cparams.flash_attn = false;
+        cparams.flash_attn_kernel = false;
+        cparams.flash_attn_causal = true;
+        cparams.mmproj = "";
+        cparams.image = "";
+        cparams.export = "";
+        cparams.export_path = "";
+        cparams.seed = -1;
+        cparams.n_keep = 0;
+        cparams.n_discard = -1;
+        cparams.n_draft = 0;
+        cparams.n_chunks = -1;
+        cparams.n_parallel = 1;
+        cparams.n_sequences = 1;
+        cparams.p_accept = 0.5f;
+        cparams.p_split = 0.1f;
+        cparams.n_gqa = 8;
+        cparams.rms_norm_eps = 5e-6f;
+        cparams.model_alias = "unknown";
+        cparams.ubatch_size = 512;
+        cparams.ubatch_seq_len_max = 1;
         
         // Load model
-        if (!context->loadModel(cparams)) {
-            throw_java_exception(env, "java/lang/RuntimeException", "Failed to load model");
-            return -1;
+        LOGI("Attempting to load model from: %s", model_path_str.c_str());
+        LOGI("Model parameters: n_ctx=%d, n_batch=%d, n_gpu_layers=%d", 
+             cparams.n_ctx, cparams.n_batch, cparams.n_gpu_layers);
+        
+        // Try to load the model
+        bool load_success = context->loadModel(cparams);
+        if (!load_success) {
+            LOGE("Model loading failed for: %s", model_path_str.c_str());
+            
+            // Try with minimal parameters as fallback
+            LOGI("Trying with minimal parameters...");
+            common_params minimal_params;
+            minimal_params.model.path = model_path_str;
+            minimal_params.n_ctx = 512;
+            minimal_params.n_batch = 256;
+            minimal_params.n_gpu_layers = 0;
+            minimal_params.use_mmap = true;
+            minimal_params.use_mlock = false;
+            minimal_params.numa = LM_GGML_NUMA_STRATEGY_DISABLED;
+            
+            load_success = context->loadModel(minimal_params);
+            if (!load_success) {
+                LOGE("Model loading failed even with minimal parameters");
+                throw_java_exception(env, "java/lang/RuntimeException", "Failed to load model");
+                return -1;
+            }
         }
         
+        LOGI("Model loaded successfully: %s", model_path_str.c_str());
+        
         // Store context
         jlong context_id = next_context_id++;
         contexts[context_id] = std::move(context);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "llama-cpp-capacitor",
-  "version": "0.0.9",
+  "version": "0.0.10",
   "description": "A native Capacitor plugin that embeds llama.cpp directly into mobile apps, enabling offline AI inference with comprehensive support for text generation, multimodal processing, TTS, LoRA adapters, and more.",
   "main": "dist/plugin.cjs.js",
   "module": "dist/esm/index.js",